Geode Sy

Martin Vermeer
GEODESY
The science underneath
Aalto University publication series
SCIENCE + TECHNOLOGY 6/2019
Geodesy
The science underneath
Martin Vermeer
Aalto University
School of Engineering
Department of Built Environment
Aalto University publication series
SCIENCE + TECHNOLOGY 6/2019
© 2019 Martin Vermeer
ISBN (pdf) 978-952-60-8872-3

ISSN 1799-490X (pdf)
http://urn.fi/URN:ISBN:978-952-60-8872-3
Graphic design: Cover: Tarja Paalanen
Helsinki 2019
Finland
Abstract
Aalto University, P.O. Box 11000, FI-00076 Aalto www.aalto.fi
Author
Martin Vermeer
Name of the publication
Geodesy: The Science Underneath
Publisher School of Engineering
Unit Department of Built Environment
Series Aalto University publication series SCIENCE + TECHNOLOGY 6/2019
Field of research Geodesy
Language English
Abstract
Geodesy is the science of precisely measuring and mapping the Earth’s surface and
locations of objects on it, the figure of the Earth and her gravity field, and changes in all
these over time. Geodesy is an old science, going back to the days when land was taken
into agricultural use and needed to be mapped. It is also a modern science, serving vital
infrastructure needs of our developing global technological society.
This text aims to describe the foundations of both traditional geodesy, mapping the
Earth within the constraints of the human living space, and modern geodesy, exploiting
space technology for mapping and monitoring our planet as a whole, in a unified three-
dimensional fashion. The approach is throughout at conveying an understanding of the
concepts, of both the science and mathematics of measuring and mapping the Earth
and the technologies used for doing so. The history of the science is not neglected, and
the perspective of the presentation is unapologetically Finnish.
Keywords geodesy, land surveying, mapping, geodetic measurements, networks, co-ordinates,

height, least-squares estimation, the built environment, figure of the Earth, GPS,
space geodesy, gravity, geodynamics, geophysics
ISBN (pdf) 978-952-60-8872-3
ISSN (PDF) 1799-490X
Location of publisher Helsinki Location of printing Helsinki Year 2019
Pages 608 urn http://urn.fi/URN:ISBN:978-952-60-8872-3
Preface
A LTHOUGH F INLAND IS, and has been since independence, a superpower

in the field of geodesy, there does not seem to be a modern geodesy
textbook in the Finnish-language area. Finnish-language textbooks, as
well as and popular books, do exist, but they are either badly outdated or
treat only a certain sub-area of geodesy. Of these, we may mention the
writings by Martti Tikka on measurement and instrument techniques
and geodetic computation (Tikka, 1991, 1985), now largely obsolete, and
Salmenperä (1998). The work on satellite positioning by Poutanen (1998)
has now been updated (2017) and is very useful. The book by Kallio
(1998) explains the least-squares statistical computation technique used
in geodesy. All these sources have been helpful in writing this book.
Internationally, clearly more geodesy textbooks are on offer, and
we have benefited from Torge (2001), Vaníček and Krakiwsky (1986),
Kahmen and Faig (1988) in measurement and instrumental techniques,
Heiskanen and Moritz (1967) in physical geodesy, and Hofmann-
Wellenhof et al. (2001) in satellite geodesy.
The material in this book divides naturally into two parts: classical
geodesy and modern geodesy. Each could be the textbook for its own
course, which would each be worth three ECTS points.
The subjects discussed in the classical geodesy part (chapters 1–10) are
the history of geodesy, the figure of the Earth and gravity, the reference el-
lipsoid, co-ordinates and heights, basics of geodetic measurements, units
of measurement, uncertainty of measurement; Helmert transformations,
the direct and inverse geodetic problems; levels and levelling, height
systems, the geoid; theodolites and total stations, angle measurements;
distance measurement using electromagnetic radiation and propagation
of the measurement ray in the atmosphere; geodetic networks, measure-
ment classes, network hierarchy; base and mapping measurements; area
–i–
ii P REFACE
and volume calculations.

In the part on modern geodesy (chapters 11–18), we concentrate instead
on the development, during the past century, from geodesy closely tied
to the Earth’s surface to genuinely three-dimensional geodesy, compris-
ing space and satellite geodesy and truly three-dimensional positioning
methods based on electromagnetism. We discuss three-dimensional refer-
ence systems, hyperbolic positioning systems and the global positioning
system GPS; GPS satellites, orbits, signals, receivers; measurements of
pseudo-range and carrier phase, measurement geometry, differencing of
observations, integer-valued ambiguities and their fixing; processing GPS
observations, relative and differential as well as real-time positioning.
We also dive deeper into the statistical foundations of geodesy, including
the least-squares method, residuals, statistical testing, outlier detection,
reliability, and planning of measurement networks. Finally, we look at
the borderlands between geodesy and geophysics, comprising the gravity
field of the Earth and the gravimetric geoid; space geodesy, the rotational
and orbital motions and deformations of the Earth; satellite orbits and
the role of geodesy in geophysical research.
We have chosen in this text to concentrate on conceptual and funda-
mental matters. That also means describing the internal workings of
instruments and processes which are in today’s systems handled auto-
matically by smart software — even if to some, this may feel like we are
teaching outdated skills.
Helsinki, 9th December, 2020,
Martin Vermeer
Second edition
A second, extensively corrected and improved edition was published on

25th September, 2022. No substantive content was added. The original
text was archived.
í ¤. û
P REFACE
iii
Acknowledgements
In drafting the various manuscript versions, lecturing transparencies and

other materials from professors Teuvo Parm and Martti Martikainen were
a great help. Assistant professor Jaakko Santala, teaching assistants
Mauri Väisänen, Panu Salo and Henri Turto, post-doctoral researcher Oc-
tavian Andrei, professor Markku Poutanen, National Land Survey chief
expert Marko Ollikainen, as well as students from many years of teaching
are gratefully acknowledged for many useful remarks, corrections, and
pieces of information.
Nicolàs de Hilster is gratefully acknowledged for contributing fig-
ure 9.8.
Several map images were drawn using the Generic Mapping Tools
(Wessel et al., 2013).
The English language was competently checked by the Finnish Trans-
lation Agency Aakkosto Oy. Tarja Paalanen designed the cover. Laura
Mure and Matti Yrjölä helped with the practicalities of publishing.
This content is licenced under the Creative Commons Attribution 4.0
International (CC BY 4.0) licence, except as noted in the text or otherwise
apparent.
í ¤. û
Contents
Chapters
Õ! 1. The history and societal status of geodesy . . . . . . . . . . . . . 1
Õ! 2. Geodetic measurements and co-ordinates . . . . . . . . . . . . . 23
Õ! 3. Map projections, datums and transformations . . . . . . . . . . 53
Õ! 4. Height measurement and the levelling instrument . . . . . . . 89
Õ! 5. The theodolite . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
Õ! 6. Angle measurement . . . . . . . . . . . . . . . . . . . . . . . . . . 157
Õ! 7. Distance measurement . . . . . . . . . . . . . . . . . . . . . . . . 185
Õ! 8. Base-network and detail-survey measurement . . . . . . . . . . 211
Õ! 9. Construction surveying . . . . . . . . . . . . . . . . . . . . . . . . 235
Õ! 10. Digital terrain models and volume calculation . . . . . . . . . . 251
Õ! 11. The third dimension . . . . . . . . . . . . . . . . . . . . . . . . . . 265
Õ! 12. The Global Positioning System (GPS) . . . . . . . . . . . . . . . 283
Õ! 13. Processing GPS observations . . . . . . . . . . . . . . . . . . . . 335
Õ! 14. Adjustment calculus in geodesy . . . . . . . . . . . . . . . . . . . 357
Õ! 15. Statistical methods in geodesy . . . . . . . . . . . . . . . . . . . . 395
Õ! 16. Gravity in geodesy . . . . . . . . . . . . . . . . . . . . . . . . . . . 425
Õ! 17. Space geodesy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 457
Õ! 18. Geodesy and geophysics . . . . . . . . . . . . . . . . . . . . . . . 479
Õ ! A. Properties of matrices . . . . . . . . . . . . . . . . . . . . . . . . . 513

Õ ! B. A short introduction to magnetohydrodynamics . . . . . . . . . 521
Õ ! C. The Kepler orbital elements for satellites . . . . . . . . . . . . . 525
Preface i
List of Tables xii
–v–
vi C ONTENTS
List of Figures xiii
Acronyms xxiii
1. The history and societal status of geodesy 1

1.1 The figure of the Earth, early conceptions . . . . . . . . . . 1
1.2 Newton’s laws and the figure of the Earth . . . . . . . . . . 5
1.3 The mathematical figure of the Earth or geoid . . . . . . . 9
1.4 The geodesic . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.5 The flattening of the Earth and gravity . . . . . . . . . . . 13
1.6 Reference surfaces and reference systems . . . . . . . . . . 14
1.7 The sub-fields of geodesy . . . . . . . . . . . . . . . . . . . . 14
1.8 Topographic surveying: from terrain to map . . . . . . . . 16
Self-test questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2. Geodetic measurements and co-ordinates 23

2.1 Units of measurement . . . . . . . . . . . . . . . . . . . . . . 23
2.2 Measurement error and uncertainty . . . . . . . . . . . . . 27
2.3 Stochastic quantities . . . . . . . . . . . . . . . . . . . . . . . 30
2.4 Statistical distributions . . . . . . . . . . . . . . . . . . . . . 32
2.5 Geodetic observables . . . . . . . . . . . . . . . . . . . . . . . 42
2.6 About co-ordinates . . . . . . . . . . . . . . . . . . . . . . . . 46
2.7 Why we use plane co-ordinates . . . . . . . . . . . . . . . . 46
2.8 Co-ordinates of location in three dimensions . . . . . . . . 47
Exercise 2–1: Co-ordinates and a street address . . . . . . . . . . 51
3. Map projections, datums and transformations 53

3.1 Map projections . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.2 The various co-ordinate solutions used in Finland . . . . . 55
3.3 Map projections used in Finland . . . . . . . . . . . . . . . . 58
3.4 More about plane co-ordinates . . . . . . . . . . . . . . . . . 65
3.5 The geodetic forward and inverse problems . . . . . . . . . 68
3.6 The similarity co-ordinate transformation . . . . . . . . . . 71
3.7 Determining the transformation parameters . . . . . . . . 73
3.8 Datums and datum transformations . . . . . . . . . . . . . 77
3.9 Map projections and height systems. . . . . . . . . . . . . . . 83
3.10 The time co-ordinate . . . . . . . . . . . . . . . . . . . . . . . 84
Exercise 3–1: Distances . . . . . . . . . . . . . . . . . . . . . . . . 86
í ¤. û
C ONTENTS
vii
4. Height measurement and the levelling instrument 89
4.1 Height, geopotential, and the geoid . . . . . . . . . . . . . . 89
4.2 Orthometric height . . . . . . . . . . . . . . . . . . . . . . . . 92
4.3 Height determination and levelling . . . . . . . . . . . . . . 93
4.4 The levelling instrument (“level”) . . . . . . . . . . . . . . . 97
4.5 The measuring telescope . . . . . . . . . . . . . . . . . . . . 98
4.6 The tubular level . . . . . . . . . . . . . . . . . . . . . . . . . 101
4.7 Checking and adjusting a levelling instrument . . . . . . . 102
4.8 Self-levelling instrument . . . . . . . . . . . . . . . . . . . . 105
4.9 Digital levelling instrument . . . . . . . . . . . . . . . . . . 107
4.10 The levelling staff . . . . . . . . . . . . . . . . . . . . . . . . 108
4.11 Levelling methods . . . . . . . . . . . . . . . . . . . . . . . . 111
Exercise 4–1: Heights . . . . . . . . . . . . . . . . . . . . . . . . . . 118
5. The theodolite 119

5.1 Horizontal angles and zenith angles . . . . . . . . . . . . . 119
5.2 The axes of a theodolite . . . . . . . . . . . . . . . . . . . . . 121
5.3 Construction of a theodolite . . . . . . . . . . . . . . . . . . 122
5.4 Theodolite handling in the field . . . . . . . . . . . . . . . . 124
5.5 Taking readings . . . . . . . . . . . . . . . . . . . . . . . . . 136
5.6 Instrumental errors of a theodolite . . . . . . . . . . . . . . 142
5.7 Electronic theodolites . . . . . . . . . . . . . . . . . . . . . . 148
5.8 Case: Leica robotic tacheometer TCA2003 . . . . . . . . . . 152
6. Angle measurement 157

6.1 Horizontal angle measurement . . . . . . . . . . . . . . . . 157
6.2 Intersection and resection . . . . . . . . . . . . . . . . . . . 158
6.3 Observation method of complete sets . . . . . . . . . . . . . 161
6.4 Station adjustment of horizontal angles . . . . . . . . . . . 162
6.5 Traverse measurement and computation . . . . . . . . . . 170
6.6 Open traverse . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
6.7 Closed traverse . . . . . . . . . . . . . . . . . . . . . . . . . . 174
6.8 Zenith angles and refraction . . . . . . . . . . . . . . . . . . 179
6.9 Heights of instrument and signal . . . . . . . . . . . . . . . 183
í ¤. û
viii C ONTENTS
7. Distance measurement 185

7.1 Mechanical distance measurement . . . . . . . . . . . . . . 185
7.2 Electromagnetic radiation . . . . . . . . . . . . . . . . . . . 188
7.3 Väisälä interferometry . . . . . . . . . . . . . . . . . . . . . 191
7.4 Electronic distance measurement . . . . . . . . . . . . . . . 194
7.5 Ray propagation in the atmosphere . . . . . . . . . . . . . . 201
7.6 “Curvature corrections” . . . . . . . . . . . . . . . . . . . . . 204
7.7 Geometric reductions . . . . . . . . . . . . . . . . . . . . . . 206
8. Base-network and detail-survey measurement 211

8.1 Objective and planning of base-network measurement . . 211
8.2 Guidance and standards . . . . . . . . . . . . . . . . . . . . 213
8.3 Network hierarchy and classification . . . . . . . . . . . . . 214
8.4 The terrain, the ellipsoid and the map plane . . . . . . . . 218
8.5 Detail survey . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
8.6 Carrying out a detail survey . . . . . . . . . . . . . . . . . . 228
9. Construction surveying 235

9.1 Zoning plans and setting out . . . . . . . . . . . . . . . . . . 235
9.2 Setting out and infrastructure . . . . . . . . . . . . . . . . . 236
9.3 Straight lines, circular arcs, rounding of corners . . . . . . 239
9.4 Transition curve . . . . . . . . . . . . . . . . . . . . . . . . . 243
9.5 Road and street surveying . . . . . . . . . . . . . . . . . . . 244
9.6 Construction surveying . . . . . . . . . . . . . . . . . . . . . 245
9.7 Other measurements . . . . . . . . . . . . . . . . . . . . . . 246
10. Digital terrain models and volume calculation 251

10.1 Terrain-model measurement, construction, presentation . 253
10.2 Use of terrain models . . . . . . . . . . . . . . . . . . . . . . 255
10.3 Calculating surface areas . . . . . . . . . . . . . . . . . . . . 256
10.4 Volume calculations . . . . . . . . . . . . . . . . . . . . . . . 260
11. The third dimension 265

11.1 Geocentric co-ordinate reference systems . . . . . . . . . . 265
11.2 Topocentric co-ordinates . . . . . . . . . . . . . . . . . . . . 268
11.3 Three-dimensional transformations . . . . . . . . . . . . . . 270
í ¤. û
C ONTENTS
ix
11.4 Transformation in the case of small rotation angles . . . . 271
11.5 The transformation between two reference ellipsoids . . . 272
11.6 Laplace azimuth measurements . . . . . . . . . . . . . . . . 275
11.7 Traditional “2D+1D” co-ordinates . . . . . . . . . . . . . . . 277
11.8 Case: the transformation between ED50 and EUREF89 . 280
11.9 Case: the transformation between ITRF and ETRF . . . . 280
Exercise 11–1: Greenwich: explain this . . . . . . . . . . . . . . . 282
12. The Global Positioning System (GPS) 283

12.1 Radio navigation and hyperbolic systems . . . . . . . . . . 284
12.2 The GPS satellite . . . . . . . . . . . . . . . . . . . . . . . . . 287
12.3 The GPS segments . . . . . . . . . . . . . . . . . . . . . . . . 289
12.4 The GPS constellation . . . . . . . . . . . . . . . . . . . . . . 291
12.5 Codes in the GPS signal . . . . . . . . . . . . . . . . . . . . . 291
12.6 GPS receivers . . . . . . . . . . . . . . . . . . . . . . . . . . . 297
12.7 Observables of GPS . . . . . . . . . . . . . . . . . . . . . . . 300
12.8 GPS measurement geometry . . . . . . . . . . . . . . . . . . 308
12.9 Measurement geometry and sensitivity of observations . . 309
12.10 Orbits of the GPS satellites . . . . . . . . . . . . . . . . . . . 324
12.11 The International GNSS Service IGS . . . . . . . . . . . . . 331
Exercise 12–1: Calculation of DOP quantities . . . . . . . . . . . 333
13. Processing GPS observations 335

13.1 Forming difference observations . . . . . . . . . . . . . . . . 335
13.2 Relative (static) GPS . . . . . . . . . . . . . . . . . . . . . . . 340
13.3 Fixing ambiguities . . . . . . . . . . . . . . . . . . . . . . . . 342
13.4 Real-time positioning . . . . . . . . . . . . . . . . . . . . . . 344
13.5 SBAS systems . . . . . . . . . . . . . . . . . . . . . . . . . . 351
13.6 Real-time support services in Finland . . . . . . . . . . . . 352
13.7 TheGDGPS system . . . . . . . . . . . . . . . . . . . . . . . . 353
Exercise 13–1: Geodetic GPS positioning . . . . . . . . . . . . . . 354
14. Adjustment calculus in geodesy 357

14.1 Why adjustment? . . . . . . . . . . . . . . . . . . . . . . . . . 357
14.2 The average . . . . . . . . . . . . . . . . . . . . . . . . . . . . 359
14.3 Linear regression . . . . . . . . . . . . . . . . . . . . . . . . . 361
14.4 Theory of least-squares adjustment . . . . . . . . . . . . . . 362
í ¤. û
x C ONTENTS
14.5 Examples of the least-squares method . . . . . . . . . . . . 366

14.6 Linearisation of geodetic models . . . . . . . . . . . . . . . . 370
14.7 Propagation of variances . . . . . . . . . . . . . . . . . . . . 376
14.8 Example: the forward geodetic problem . . . . . . . . . . . 378
14.9 Observables and observation equations in practice . . . . 382
14.10 Tacheometer measurement . . . . . . . . . . . . . . . . . . . 388
14.11 Helmert transformation in the plane . . . . . . . . . . . . . 389
Exercise 14–1: Helmert transformation parameter estimation . 392
15. Statistical methods in geodesy 395

15.1 The method of least-squares . . . . . . . . . . . . . . . . . . 395
15.2 The residuals from the adjustment . . . . . . . . . . . . . . 397
15.3 Testing, hypotheses for testing . . . . . . . . . . . . . . . . . 400
15.4 Overall validation . . . . . . . . . . . . . . . . . . . . . . . . 402
15.5 Locating gross errors . . . . . . . . . . . . . . . . . . . . . . 405
15.6 Calculation example: linear regression . . . . . . . . . . . . 406
15.7 Adding a gross error . . . . . . . . . . . . . . . . . . . . . . . 408
15.8 Significance level of the test . . . . . . . . . . . . . . . . . . 409
15.9 Reliability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 412
15.10 Deformation analysis . . . . . . . . . . . . . . . . . . . . . . 417
16. Gravity in geodesy 425

16.1 Measuring gravity . . . . . . . . . . . . . . . . . . . . . . . . 425
16.2 Gravity and geopotential . . . . . . . . . . . . . . . . . . . . 428
16.3 Gravity anomalies . . . . . . . . . . . . . . . . . . . . . . . . 433
16.4 The gravimetric geoid . . . . . . . . . . . . . . . . . . . . . . 435
16.5 The gravity field and heights . . . . . . . . . . . . . . . . . . 439
16.6 Bouguer anomalies . . . . . . . . . . . . . . . . . . . . . . . . 447
16.7 Astronomical position determination . . . . . . . . . . . . . 448
16.8 Measuring the gravity gradient . . . . . . . . . . . . . . . . 450
Exercise 16–1: Gravimetric geoid computation . . . . . . . . . . 454
17. Space geodesy 457

17.1 Earth rotation, orbital motion, sidereal time . . . . . . . . 457
17.2 Heavenly and Earthly co-ordinates . . . . . . . . . . . . . . 460
17.3 Väisälä’s stellar triangulation . . . . . . . . . . . . . . . . . 462
17.4 Variations in the Earth’s rotation . . . . . . . . . . . . . . . 465
í ¤. û
C ONTENTS
xi
17.5 Precession and nutation of the Earth . . . . . . . . . . . . . 469
17.6 Space weather . . . . . . . . . . . . . . . . . . . . . . . . . . 470
17.7 Satellite orbital motion . . . . . . . . . . . . . . . . . . . . . 473
17.8 Choosing a satellite orbit . . . . . . . . . . . . . . . . . . . . 475
17.9 Satellite orbital precession, Sun-synchronous orbit . . . . 476
18. Geodesy and geophysics 479

18.1 Geodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . 479
18.2 Plate tectonics . . . . . . . . . . . . . . . . . . . . . . . . . . 483
18.3 Glacial isostatic adjustment (GIA) . . . . . . . . . . . . . . 488
18.4 Local geodynamics . . . . . . . . . . . . . . . . . . . . . . . . 492
18.5 Deformation monitoring . . . . . . . . . . . . . . . . . . . . . 493
18.6 Studying the Earth’s gravity field from orbit . . . . . . . . 494
18.7 Atmospheric research and GNSS . . . . . . . . . . . . . . . 499
18.8 Long-term variations in Earth rotation axis and orbit . . . 503
18.9 Land-ice research and climate change . . . . . . . . . . . . 505
18.10 Geodetic oceanography . . . . . . . . . . . . . . . . . . . . . 506
A. Properties of matrices 513

A.1 Adding and subtracting matrices . . . . . . . . . . . . . . . 513
A.2 Matrices and vectors . . . . . . . . . . . . . . . . . . . . . . . 514
A.3 The unit matrix . . . . . . . . . . . . . . . . . . . . . . . . . . 514
A.4 Matrix multiplication . . . . . . . . . . . . . . . . . . . . . . 515
A.5 The transpose . . . . . . . . . . . . . . . . . . . . . . . . . . . 516
A.6 The inverse matrix . . . . . . . . . . . . . . . . . . . . . . . . 516
A.7 The determinant of a matrix . . . . . . . . . . . . . . . . . . 518
A.8 Vectorial products . . . . . . . . . . . . . . . . . . . . . . . . 519
B. A short introduction to magnetohydrodynamics 521

B.1 Plasma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 521
B.2 Maxwell’s equations . . . . . . . . . . . . . . . . . . . . . . . 521
B.3 “Frozen-in” magnetic field . . . . . . . . . . . . . . . . . . . 522
B.4 History of the field . . . . . . . . . . . . . . . . . . . . . . . . 523
C. The Kepler orbital elements for satellites 525

C.1 Angular elements describing the orientation in space of
the orbit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 525
í ¤. û
xii L IST OF T ABLES
C.2 Elements describing the orbit’s size and shape . . . . . . . 525

C.3 Elements describing the satellite’s place in its orbit, its
“time table” . . . . . . . . . . . . . . . . . . . . . . . . . . . . 526
Bibliography 527
Index 549
List of Tables
1.1 Topographic surveying . . . . . . . . . . . . . . . . . . . . . . . 21
2.1 Measured quantities, units and their symbols . . . . . . . . . 24

2.2 Prefixes indicating order of magnitude in the SI system . . . 25
2.3 Non-SI units accepted for use with the SI . . . . . . . . . . . 25
2.4 Die-cast statistics . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.5 On correlation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.1 Alternative vertical datums A and B . . . . . . . . . . . . . . 80
4.1 Classification of levelling instruments . . . . . . . . . . . . . 98

4.2 Classification of levelling staffs . . . . . . . . . . . . . . . . . . 110
6.1 Calculation template for station adjustment . . . . . . . . . . 169

6.2 Traverse calculation template . . . . . . . . . . . . . . . . . . 178
7.1 Calculating the constant and frequency error of a range-

finder by linear regression . . . . . . . . . . . . . . . . . . . . 200
7.2 Examples of distance reductions . . . . . . . . . . . . . . . . . 207
8.1 Methods for base-network measurement . . . . . . . . . . . . 214

8.2 Goodness of approximation by the reference ellipsoid . . . . 220
8.3 Classification of topographic data . . . . . . . . . . . . . . . . 233
11.1 Transformation parameters between EUREF89 and ED50 . 280

11.2 Transformation parameters between ITRF2005 and ETRF-
2005 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
12.1 Codes included in the GPS signal . . . . . . . . . . . . . . . . 292

12.2 How does dendrochronology work . . . . . . . . . . . . . . . . 294
12.3 Start of a RINEX file . . . . . . . . . . . . . . . . . . . . . . . . 300
í ¤. û
L IST OF F IGURES
xiii
12.4 Properties of carrier waves . . . . . . . . . . . . . . . . . . . . 304
12.5 Deriving the influence formula by linearisation . . . . . . . . 312
12.6 Variants of the DOP quantity . . . . . . . . . . . . . . . . . . . 314
12.7 Precise ephemeris in the original SP3 format . . . . . . . . . 329
12.8 DOP calculation script . . . . . . . . . . . . . . . . . . . . . . . 332
13.1 Effect of forming difference observations on the magnitude

of various errors . . . . . . . . . . . . . . . . . . . . . . . . . . . 338
13.2 Summary of GPS observables and difference quantities . . . 340
13.3 Relation between orbit error, length of vector, and position-
ing error . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341
14.1 Measurement results for linear regression . . . . . . . . . . . 369

14.2 Point set given in two different co-ordinate frames . . . . . . 392
14.3 Calculation script for Helmert transformation . . . . . . . . 394
15.1 The planning and measurement process . . . . . . . . . . . . 401

15.2 Rejection bounds for significance levels in a two-sided test
based on the standard normal distribution . . . . . . . . . . . 406
15.3 Example of linear regression . . . . . . . . . . . . . . . . . . . 406
15.4 Values of the chi-squared distribution . . . . . . . . . . . . . 407
15.5 Example of linear regression calculation . . . . . . . . . . . . 408
15.6 Example of linear regression with a simulated gross error . 409
15.7 Rejection bound and significance level, normal distribution . 411
15.8 Size of gross error and corresponding power of test . . . . . . 411
15.9 Horizontal deformation analysis, co-ordinates . . . . . . . . . 422
16.1 Normal potential and normal gravity according to GRS80 . 431

16.2 Properties of various height types . . . . . . . . . . . . . . . . 446
17.1 Kepler’s third law for Earth satellites . . . . . . . . . . . . . . 476
List of Figures
1.1 A lunar eclipse . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 The grade measurement of Eratosthenes . . . . . . . . . . . . 2
1.3 The Snellius grade measurement . . . . . . . . . . . . . . . . 4
1.4 Astronomically determining the difference of plumb-line di-
rections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.5 Different mass distribution models for the Earth . . . . . . . 6
1.6 Parameters of an ellipsoid of revolution . . . . . . . . . . . . . 7
í ¤. û
xiv L IST OF F IGURES
1.7 The Lapland grade measurement . . . . . . . . . . . . . . . . 9

1.8 The northernmost point of the Struve chain . . . . . . . . . . 10
1.9 Deviations of the plumb line and shape of the geoid . . . . . 11
1.10 The geodesic in the plane, on the sphere and on the ellipsoid 12
1.11 Spatial planning . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.1 A public standard metre in Paris . . . . . . . . . . . . . . . . . 24

2.2 Examples of different error types . . . . . . . . . . . . . . . . 30
2.3 A stochastic quantity on a continuous 2-D value set . . . . . 33
2.4 Probability density distribution as limit of histograms . . . . 33
2.5 Properties of the normal distribution . . . . . . . . . . . . . . 34
2.6 Probability values for the normal distribution . . . . . . . . . 35
2.7 Some examples of correlation . . . . . . . . . . . . . . . . . . . 37
2.8 A two-dimensional probability density distribution . . . . . . 39
2.9 Triangulation by means of a plane table and alidade . . . . . 43
2.10 Forming a stereo model in photogrammetry . . . . . . . . . . 44
2.11 The Greenwich meridian for tourists . . . . . . . . . . . . . . 48
2.12 Rectangular and geodetic co-ordinates . . . . . . . . . . . . . 49
2.13 Geodetic co-ordinates . . . . . . . . . . . . . . . . . . . . . . . . 50
3.1 Depicting the curved surface of the Earth to the map plane
using different projections . . . . . . . . . . . . . . . . . . . . . 55
3.2 Systematic shift between road network and aerial-photo-
graph base . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.3 Imaging the Earth’s surface as a narrow zone onto a plane . 58
3.4 Zone division of the KKJ Gauss–Krüger projection . . . . . . 59
3.5 Geometry of one zone of the Finnish KKJ system . . . . . . . 60
3.6 Scale distortion of Gauss–Krüger and UTM projections . . . 63
3.7 Triangulated affine transformation of the Finnish NLS . . . 64
3.8 Geodetic plane co-ordinates and the quadrants of the plane 65
3.9 A local co-ordinate frame . . . . . . . . . . . . . . . . . . . . . 66
3.10 Temporary co-ordinates . . . . . . . . . . . . . . . . . . . . . . 67
3.11 Forward geodetic problem in the plane . . . . . . . . . . . . . 68
3.12 Half-angle formula for the arc tangent . . . . . . . . . . . . . 70
3.13 Friedrich Robert Helmert . . . . . . . . . . . . . . . . . . . . . 72
3.14 Similarity or Helmert transformation in the plane . . . . . . 72
3.15 Stages of the Helmert transformation in the plane . . . . . . 74
3.16 Fundamental benchmark PP2000 of the N2000 height da-
tum at Metsähovi research station . . . . . . . . . . . . . . . . 79
í ¤. û
L IST OF F IGURES
xv
3.17 Alternative vertical datums . . . . . . . . . . . . . . . . . . . . 80
3.18 Two different datums of a horizontal network . . . . . . . . . 82
4.1 Different height types map geopotential numbers in different

ways to metric heights . . . . . . . . . . . . . . . . . . . . . . . 91
4.2 Orthometric heights are metric distances from the geoid . . 92
4.3 Important reference surfaces and height concepts . . . . . . 93
4.4 The Finnish geoid model FIN2000 . . . . . . . . . . . . . . . . 94
4.5 The geometry of levelling . . . . . . . . . . . . . . . . . . . . . 96
4.6 Levelling instrument . . . . . . . . . . . . . . . . . . . . . . . . 98
4.7 Telescope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
4.8 Measuring telescope . . . . . . . . . . . . . . . . . . . . . . . . 100
4.9 Parallax of a measuring telescope . . . . . . . . . . . . . . . . 101
4.10 Tubular spirit level . . . . . . . . . . . . . . . . . . . . . . . . . 102
4.11 The geometry of a field check . . . . . . . . . . . . . . . . . . . 103
4.12 Adjusting the horizon of a levelling instrument . . . . . . . . 104
4.13 An early self-levelling levelling instrument . . . . . . . . . . 106
4.14 A modern self-levelling levelling instrument . . . . . . . . . . 106
4.15 Principle of operation of the pendulum compensator . . . . . 107
4.16 Levelling staffs and foliage . . . . . . . . . . . . . . . . . . . . 108
4.17 Graduation alternatives for the staff scale . . . . . . . . . . . 109
4.18 Various temporary levelling-staff supports . . . . . . . . . . . 111
4.19 Traverse levelling . . . . . . . . . . . . . . . . . . . . . . . . . . 112
4.20 Area levelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
4.21 Self-calculating levelling staff . . . . . . . . . . . . . . . . . . 113
4.22 Principle of operation of a laser level . . . . . . . . . . . . . . 115
4.23 A profile and cross-sections . . . . . . . . . . . . . . . . . . . . 115
4.24 Metsähovi research station from the air . . . . . . . . . . . . 117
5.1 An old-fashioned theodolite . . . . . . . . . . . . . . . . . . . . 120

5.2 Horizontal angle and zenith angle . . . . . . . . . . . . . . . . 121
5.3 The axes and circles of a theodolite . . . . . . . . . . . . . . . 122
5.4 Theodolite construction . . . . . . . . . . . . . . . . . . . . . . 123
5.5 Forced-centring device or plate . . . . . . . . . . . . . . . . . . 125
5.6 Various monument types . . . . . . . . . . . . . . . . . . . . . 126
5.7 Theodolite axes . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
5.8 Precise levelling of a theodolite using the alidade level . . . 128
5.9 String plummet and rod plummet . . . . . . . . . . . . . . . . 129
5.10 Optical plummet . . . . . . . . . . . . . . . . . . . . . . . . . . 131
í ¤. û
xvi L IST OF F IGURES
5.11 A benchmark seen though an optical plummet . . . . . . . . 131

5.12 Optical plummet and bull’s-eye level used together . . . . . . 132
5.13 Problem situation . . . . . . . . . . . . . . . . . . . . . . . . . . 132
5.14 Principle of forced centring . . . . . . . . . . . . . . . . . . . . 133
5.15 Measuring a network using forced centring . . . . . . . . . . 134
5.16 Checking an optical plummet . . . . . . . . . . . . . . . . . . . 135
5.17 Good targets for horizontal angles . . . . . . . . . . . . . . . . 136
5.18 Targeting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
5.19 Various types of reading microscopes . . . . . . . . . . . . . . 138
5.20 Optical micrometer and its reading . . . . . . . . . . . . . . . 139
5.21 Reading the graduation circle. One circle location . . . . . . 140
5.22 Reading the graduation circle. Two opposite circle locations 141
5.23 Turning the sight axis by shifting the crosshairs . . . . . . . 142
5.24 Trunnion-axis tilt . . . . . . . . . . . . . . . . . . . . . . . . . . 145
5.25 Observing a zenith angle . . . . . . . . . . . . . . . . . . . . . 146
5.26 Index error . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
5.27 The Gray code . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
5.28 An absolute and an incremental encoding circle . . . . . . . . 150
5.29 Electronic readout of the horizontal circle . . . . . . . . . . . 151
5.30 A spinning circle converts measurement of angles into one
of time differences . . . . . . . . . . . . . . . . . . . . . . . . . 152
5.31 Leica TCA2003 robot tacheometer control panel . . . . . . . 153
5.32 Leica TCA2003 robot tacheometer . . . . . . . . . . . . . . . . 153
6.1 Use situations for horizontal angle measurement . . . . . . . 158

6.2 Intersection and resection . . . . . . . . . . . . . . . . . . . . . 159
6.3 Observation method of complete sets . . . . . . . . . . . . . . 162
6.4 An open and a closed traverse . . . . . . . . . . . . . . . . . . 171
6.5 An open traverse . . . . . . . . . . . . . . . . . . . . . . . . . . 172
6.6 The geometry of a closed traverse . . . . . . . . . . . . . . . . 175
6.7 Effects of refraction and Earth curvature on zenith-angle
measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
6.8 Trigonometric levelling traverse . . . . . . . . . . . . . . . . . 182
6.9 Heights of instrument and signal . . . . . . . . . . . . . . . . 183
7.1 Sag correction of measuring tape . . . . . . . . . . . . . . . . 186

7.2 Slope reduction of slant ranges . . . . . . . . . . . . . . . . . . 188
7.3 The phase of a wave motion . . . . . . . . . . . . . . . . . . . . 188
7.4 The electromagnetic radiation spectrum . . . . . . . . . . . . 190
í ¤. û
L IST OF F IGURES
xvii
7.5 Polarisation of electromagnetic radiation . . . . . . . . . . . . 192
7.6 Väisälä’s interference method . . . . . . . . . . . . . . . . . . 193
7.7 Dendrochronology . . . . . . . . . . . . . . . . . . . . . . . . . . 193
7.8 Fizeau’s method for measuring the speed of light . . . . . . . 195
7.9 One method of electronic phase measurement . . . . . . . . . 196
7.10 Ambiguities, or integer unknowns, are resolved by using
several different wavelengths . . . . . . . . . . . . . . . . . . . 197
7.11 Corner-cube prism . . . . . . . . . . . . . . . . . . . . . . . . . 198
7.12 A prism pack for measurement over long distances . . . . . . 199
7.13 Targeting a signal with a corner-cube prism . . . . . . . . . . 199
7.14 The second velocity correction . . . . . . . . . . . . . . . . . . 205
7.15 The terrain correction of distance measurement . . . . . . . 205
7.16 Reduction of distance measurement to a reference level . . . 206
8.1 Significance of network hierarchy, mistakes often made . . . 215

8.2 Finnish continuously operating GNSS network FinnRef . . 216
8.3 Finnish EUREF-FIN first-stage densification network . . . 217
8.4 A triangulation network and a traverse in space . . . . . . . 219
8.5 Use of reference ellipsoid and map projection plane when
mapping the Earth . . . . . . . . . . . . . . . . . . . . . . . . . 220
8.6 Transferring the geometry for adjustment of a small network
to the map projection plane . . . . . . . . . . . . . . . . . . . . 222
8.7 Tools of the right-angle survey method . . . . . . . . . . . . . 223
8.8 Right-angle survey . . . . . . . . . . . . . . . . . . . . . . . . . 224
8.9 Tie-in survey . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
8.10 Radial survey . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
8.11 Free-stationing survey . . . . . . . . . . . . . . . . . . . . . . . 228
8.12 Workflow diagram of detail survey . . . . . . . . . . . . . . . . 230
8.13 The encoding process for topographic data . . . . . . . . . . . 232
8.14 Attribute data of objects in multiple layers . . . . . . . . . . . 233
9.1 Setting out onto the terrain, process description. . . . . . . . 237

9.2 Zoning-plan calculation — an example . . . . . . . . . . . . . 238
9.3 Setting out using the radial survey method . . . . . . . . . . 238
9.4 Straight setting-out method . . . . . . . . . . . . . . . . . . . . 239
9.5 Rounding of corners with a circular arc . . . . . . . . . . . . . 240
9.6 Rounding of corners with a compound curve . . . . . . . . . . 241
9.7 Principle of the clothoid . . . . . . . . . . . . . . . . . . . . . . 243
9.8 Machine guidance, case Easter Scheldt . . . . . . . . . . . . . 248
í ¤. û
xviii L IST OF F IGURES
10.1 Terrain model ETOPO2 version 2 on the Finnish territory . 253

10.2 Presentation of terrain models: triangulated network or
point-grid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
10.3 Use of setting-out measures in calculating surface areas . . 257
10.4 Calculating surface area . . . . . . . . . . . . . . . . . . . . . . 258
10.5 A polar planimeter from 1908 . . . . . . . . . . . . . . . . . . 259
10.6 Graphical proof of the planimeter equation . . . . . . . . . . 259
10.7 Simpson’s integration rule in volume calculation . . . . . . . 261
10.8 Alternatives for quadrature . . . . . . . . . . . . . . . . . . . . 262
10.9 Volume calculation from digital terrain models . . . . . . . . 263
11.1 Inertial and terrestrial co-ordinate reference systems . . . . 266

11.2 A right-handed co-ordinate frame . . . . . . . . . . . . . . . . 267
11.3 Geocentric and geodetic latitude and transversal radius of
curvature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268
11.4 Topocentric and geocentric co-ordinate frames . . . . . . . . 269
11.5 Differential connection between rectangular and geodetic
co-ordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 274
11.6 Effect of a datum transformation on geodetic quantities . . . 275
11.7 Deviation of local plumb line from ellipsoidal normal . . . . 276
11.8 The Laplace phenomenon: the effect of the plumb-line devia-
tion on the azimuth . . . . . . . . . . . . . . . . . . . . . . . . . 278
11.9 Greenwich geometry: longitude is a direction, not a place . . 282
12.1 The Decca system . . . . . . . . . . . . . . . . . . . . . . . . . . 285

12.2 A Decca receiver . . . . . . . . . . . . . . . . . . . . . . . . . . 285
12.3 The NNSS Transit system . . . . . . . . . . . . . . . . . . . . 287
12.4 Positioning satellites . . . . . . . . . . . . . . . . . . . . . . . . 288
12.5 The three segments of the GPS . . . . . . . . . . . . . . . . . . 290
12.6 The original GPS constellation . . . . . . . . . . . . . . . . . . 292
12.7 Principle of phase modulation . . . . . . . . . . . . . . . . . . 293
12.8 Correlation method for determining the travel time of the
GPS signal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 294
12.9 The various frequencies and effective wavelengths of the
GPS signal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 296
12.10 Control panel of the Ashtech Z-12 . . . . . . . . . . . . . . . . 298
12.11 Choke-ring GNSS antenna for precise geodetic work . . . . . 299
12.12 The electric centre of an antenna is not a self-evident thing 300
12.13 Pseudo-range observation . . . . . . . . . . . . . . . . . . . . . 303
í ¤. û
L IST OF F IGURES
xix
12.14 Measurement of the phase of the GPS signal’s carrier wave 305
12.15 Propagation of a wave packet in a dispersive medium . . . . 306
12.16 Geometry of GPS positioning . . . . . . . . . . . . . . . . . . . 308
12.17 Geometry between a GPS satellite and an observation site . 311
12.18 DOP and error ellipsoids and mean error of unit weight . . . 317
12.19 DOP ellipsoid, principal axes along co-ordinate axes . . . . . 317
12.20 The circle singularity or “dangerous circle” for GPS . . . . . 322
12.21 Calculation example of DOP quantities . . . . . . . . . . . . . 323
12.22 The six orbital planes of GPS satellites in the Helsinki sky . 325
12.23 Satellite orbital motion described by position and velocity
vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 326
12.24 The tracking stations of the IGS . . . . . . . . . . . . . . . . . 330
13.1 The “common-mode” error assumption . . . . . . . . . . . . . 335

13.2 Forming difference observations of various types . . . . . . . 337
13.3 Double difference, short distance between GPS receivers . . 341
13.4 One-dimensional ambiguity resolution, ranging . . . . . . . . 342
13.5 Various ambiguity-resolution methods . . . . . . . . . . . . . 343
13.6 Principle of operation of the DGPS method . . . . . . . . . . 346
13.7 Principle of operation of the RTK method . . . . . . . . . . . 348
13.8 Satellite-based augmentation systems (SBAS) . . . . . . . . 352
14.1 Triangulation network . . . . . . . . . . . . . . . . . . . . . . . 358

14.2 Metaphor: a large weight means a small correction . . . . . 360
14.3 The idea of linear regression . . . . . . . . . . . . . . . . . . . 361
14.4 Linear regression, definitions of quantities . . . . . . . . . . . 362
14.5 Calculation example of linear regression . . . . . . . . . . . . 369
14.6 One-dimensional mapping and linearisation . . . . . . . . . . 372
14.7 A two-dimensional mapping . . . . . . . . . . . . . . . . . . . 374
14.8 Quantities related to the error ellipse . . . . . . . . . . . . . . 378
14.9 Geometry of azimuth measurement . . . . . . . . . . . . . . . 385
14.10 Geometry of zenith-angle measurement . . . . . . . . . . . . 387
15.1 Least-squares adjustment as an orthogonal projection . . . . 399

15.2 Chi-squared distribution with four degrees of freedom . . . . 403
15.3 Example of linear regression, error in observation 3 . . . . . 410
15.4 Statistical testing based on normal distribution . . . . . . . . 410
15.5 Harmonisation of the significance levels of the overall vali-
dation and per-observation tests . . . . . . . . . . . . . . . . . 413
15.6 An example of reliability . . . . . . . . . . . . . . . . . . . . . . 413
í ¤. û
xx L IST OF F IGURES
15.7 Another example of reliability . . . . . . . . . . . . . . . . . . 414

15.8 Height deformation monitoring network . . . . . . . . . . . . 419
15.9 Horizontal deformation monitoring network . . . . . . . . . . 421
16.1 Absolute or ballistic gravimeter . . . . . . . . . . . . . . . . . 427

16.2 Relative or spring gravimeter . . . . . . . . . . . . . . . . . . . 427
16.3 The terrain height depicted by height contours, and height
gradients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 428
16.4 Geopotential table . . . . . . . . . . . . . . . . . . . . . . . . . 429
16.5 Normal gravity field of the Earth . . . . . . . . . . . . . . . . 431
16.6 Equipotential surfaces and lines of force of geopotential and
normal potential . . . . . . . . . . . . . . . . . . . . . . . . . . 432
16.7 Equipotential surfaces of true and normal gravity field . . . 433
16.8 True and normal gravity vectors . . . . . . . . . . . . . . . . . 434
16.9 Relationship between gravity and geoid height variations . 436
16.10 Geometry of the Stokes integral equation . . . . . . . . . . . 437
16.11 Global geoid model EGM2008 . . . . . . . . . . . . . . . . . . 439
16.12 The gravity vector is the gradient of the geopotential . . . . 440
16.13 The path integral of work . . . . . . . . . . . . . . . . . . . . . 441
16.14 Heights and equipotential surfaces . . . . . . . . . . . . . . . 442
16.15 Free-air and Bouguer anomalies for Southern Finland . . . 448
16.16 Root of a mountain range and its effect on the plumb line . . 449
16.17 A levelling instrument converted to astrolabe . . . . . . . . . 450
16.18 Gravitational-gradient or tidal force field . . . . . . . . . . . . 451
17.1 Earth orbit around the Sun, and apparent path of the Sun
across the celestial sphere . . . . . . . . . . . . . . . . . . . . . 458
17.2 The vernal equinox and its movement, precession . . . . . . 459
17.3 Hour angle and declination on the celestial sphere . . . . . . 460
17.4 Yrjö Väisälä’s stellar triangulation (a) . . . . . . . . . . . . . 462
17.5 Yrjö Väisälä’s stellar triangulation (b) . . . . . . . . . . . . . 463
17.6 Satellite geodesy from NASA photographic archives . . . . . 465
17.7 Polar motion for the period 1970–2000 . . . . . . . . . . . . . 466
17.8 Polar motion causes variations in station latitudes . . . . . . 467
17.9 The Earth’s precession . . . . . . . . . . . . . . . . . . . . . . . 470
17.10 The corona of the Sun . . . . . . . . . . . . . . . . . . . . . . . 471
17.11 Sunspots and their magnetic field lines . . . . . . . . . . . . . 472
17.12 Space weather, the magnetosphere . . . . . . . . . . . . . . . 473
17.13 Ellipse, definition and how to draw one . . . . . . . . . . . . . 474
í ¤. û
L IST OF F IGURES
xxi
17.14 Kepler’s orbital elements . . . . . . . . . . . . . . . . . . . . . 475
17.15 Sun-synchronous orbit . . . . . . . . . . . . . . . . . . . . . . . 477
18.1 A LAGEOS satellite . . . . . . . . . . . . . . . . . . . . . . . . 481

18.2 Principle of operation of very long baseline interferometry . 482
18.3 Metsähovi radio telescope used for VLBI measurements . . 483
18.4 Wegener’s continental drift theory, Mid-Atlantic Ridge . . . 484
18.5 Internal structure of the Earth . . . . . . . . . . . . . . . . . . 485
18.6 Palaeomagnetism and sea-floor spreading . . . . . . . . . . . 486
18.7 Global plate tectonics . . . . . . . . . . . . . . . . . . . . . . . 487
18.8 Mechanisms of plate tectonics . . . . . . . . . . . . . . . . . . 488
18.9 Post-glacial land uplift in Fennoscandia . . . . . . . . . . . . 490
18.10 Horizontal and vertical motions in Fennoscandia as deter-
mined by the BIFROST project . . . . . . . . . . . . . . . . . 491
18.11 InSAR image . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 495
18.12 Determining the Earth’s gravitational field by tracking the
orbit of a low-flying satellite . . . . . . . . . . . . . . . . . . . 496
18.13 Basic idea of the GRACE satellite pair . . . . . . . . . . . . . 497
18.14 Determining the Earth’s gravitational field with the gravita-
tional gradiometer on-board the GOCE satellite . . . . . . . 498
18.15 Sea-surface topography map produced by GOCE . . . . . . . 499
18.16 Saturation partial pressure of water vapour . . . . . . . . . . 501
18.17 Use of GNSS for studying the troposphere . . . . . . . . . . . 501
18.18 GNSS radio occultation . . . . . . . . . . . . . . . . . . . . . . 503
18.19 Milanković cycles over the past 800 000 years . . . . . . . . . 504
18.20 Measurement geometry of satellite radar altimetry . . . . . 506
18.21 Link between sea-surface topography and ocean currents . . 507
18.22 Tide gauges of the Baltic Sea, Seasat ground tracks, and
mean sea level . . . . . . . . . . . . . . . . . . . . . . . . . . . . 511
18.23 Schematic of the Earth’s orbital changes (IPCC) . . . . . . . . 512
í ¤. û
Acronyms
ABCDEFGHIJKLMNOPRSTUVWY
A
ANNA 1B (“Army, Navy, NASA, Air Force”), 1962-60A, geodetic satellite 464
APPS Automatic Precise Positioning Service (JPL) 355
ARP antenna reference point 298
ATR automatic target recognition 150, 154, 155
AUSPOS Australian online GPS processing service 355, 356
B
BGI Bureau Gravimétrique International, International Gravimetric Bureau 448
BIFROST Baseline Inferences for Fennoscandian Rebound, Sea level, and Tec-
tonics, a Nordic geodynamics research project 491
BIPM Bureau International des Poids et Mesures, International Bureau of
Weights and Measures 23
C
C/A code Coarse / Acquisition, Civilian Access GPS code 291–293, 295–297, 300,
303, 347
CAD computer-aided design 21, 53, 255
Caltech California Institute of Technology 495
CCD charge-coupled device, image sensor type 107, 149, 150, 154
CDMA code division multiple access 293, 344
CERN Organisation européenne pour la recherche nucléaire, European Organiza-
tion for Nuclear Research 114
CHAMP 2000-39B, Challenging Minisatellite Payload, German satellite 494, 502
CIO Conventional International Origin, reference pole for polar motion 468
D
Decca marine navigation system 45, 284–287, 331, 348
DEM digital elevation model 251, 256
DGPS differential GPS 345–347, 349, 350, 353, 354
– xxiii –
xxiv ABCDEFGHIJKLMNOPRSTUVWY A CRONYMS
DHM digital height model 251

DNA deoxyribonucleic acid, helix-shaped macromolecule that carries and repli-
cates the genetic information of almost all Earth’s organisms 189
DOP dilution of precision, a measure for the geometric strength of satellite
positioning. xxv, xxvii, xxviii, 312–318, 323, 332, 333
DORIS Doppler Orbitography and Radiopositioning Integrated by Satellite, a
French satellite positioning system 214, 480, 483
DTM digital terrain model 251, 256, 447
DVD Digital Versatile Disc 230
DWT discrete wavelet transform 255
E
ECEF Earth-centred, Earth-fixed 266
ED50 European Datum 1950 11, 272, 277, 279–281
EEST Eastern European Summer Time 48
EET Eastern European Time 48
EGM2008 Earth Gravity Model 2008 118, 439, 448
EGM96 Earth Gravity Model 1996 118
EGNOS European Geostationary Navigation Overlay System, an SBAS for the
European area 351
eLoran navigation system, planned (“Enhanced Loran”) 284
EOP Earth orientation parameters 85, 331, 468
ERT electrical resistivity tomography 248
ET ephemeris time, efemeridiaika 468
ETRF European Terrestrial Reference Frame 77, 281
ETRS European Terrestrial Reference System. Coincides with ITRS for the epoch
1989.0. Also called ETRS89 xxiv, 57, 58, 77, 86, 279, 281
ETRS-GK Gauss–Krüger map projection system for Finland 62, 235, 392
ETRS-TM35FIN UTM map projection for Finland, zone 35 61, 62, 87
EUREF IAG Reference Frame Subcommission for Europe 57, 86, 281, 354
EUREF89 First European realisation of ETRS89 57, 280
EUREF-FIN Finnish national realisation of ETRS89 48, 57, 58, 61, 85, 213, 214,
217, 233, 235, 279–281
F
FAT file allocation table (file system) 155
FDMA frequency division multiple access 344
FGI Finnish Geodetic Institute, 1918–2015, Finnish Geospatial Research Insti-
tute, 2015– 79, 216, 217, 354, 483, 510
FIN2000 Finnish geoid model 94
FIN2005N00 Finnish geoid model 279
FRS Fellow of the Royal Society (of London) 188, 194, 395, 420, 486, 504, 522
í ¤. û
A CRONYMS ABCDEFGHIJKLMNOPRSTUVWY
xxv
FRSE Fellow of the Royal Society of Edinburgh 188, 194, 420, 486, 522
G
GAST Greenwich Apparent Sidereal Time 459, 461, 478
GCM general circulation model 256
GCP ground control point 254
GDOP geometric DOP 312, 314, 316
GIA glacial isostatic adjustment 85, 488, 490, 509
GIS geographic information system 21
GLONASS Global Navigation Satellite System (Russian) 283, 284, 318, 330, 341,
344, 350, 352, 479
GMT Greenwich Mean Time 48
GNSS Global Navigation Satellite Systems, generic name xxv, xxvii, 21, 45, 47,
85, 86, 118, 126, 133, 158, 170, 184, 198, 204, 212, 214, 216–218, 221,
228, 234, 236, 247, 280, 282, 284, 302, 328, 330, 331, 341, 342, 344,
353–355, 433, 465, 468, 479, 480, 482, 491–493, 497, 499–503, 506, 507,
509–511
GOCE Gravity Field and Steady-State Ocean Circulation Explorer 95, 497–499,
508, 511
GPR ground-penetrating radar 248
GPS Global Positioning System ii, xxiii, xxv, xxvii, xxviii, 14, 45, 46, 48, 93, 118,
163, 192, 193, 197, 203, 217, 234, 281, 283, 284, 286, 287, 289–300, 302,
303, 305–309, 311–314, 317–319, 321, 322, 324–328, 331–333, 335, 338,
340–345, 347, 349–351, 354, 382, 469, 473, 476, 479, 482, 488, 494, 496,
502, 503, 525
GPS/MET 1995-17C, GPS radio occultation satellite 502
2
GPU geopotential unit, 10 m /s2 441, 442
GRACE 2002-12A, 2002-12B, Gravity Recovery and Climate Experiment. This
was a pair of satellites 496, 497, 505, 511
GRS80 Geodetic Reference System 1980 49, 61, 84, 86, 94, 118, 219, 279, 431,
439, 476
GSI (Leica) Geo Serial Interface 154, 155
H
HDOP horizontal DOP 314–316, 318, 321, 324, 333
I
IAG International Association of Geodesy xxiv, 11, 57, 202, 331, 466, 480
IB inverted barometer 507
IERS International Earth Rotation and Reference Systems Service 466, 468
IGS International GNSS Service 328, 330, 331, 355, 503
i.i.d. independent and identically distributed 364
í ¤. û
xxvi ABCDEFGHIJKLMNOPRSTUVWY A CRONYMS
InSAR interferometric synthetic-aperture radar 494, 495

INSPIRE Infrastructure for Spatial Information in the European Community
252
IONEX Ionosphere Map Exchange Format 503
IP Internet Protocol 481
IPCC Intergovernmental Panel on Climate Change 512
ITRF International Terrestrial Reference Frame, a realisation of the ITRS 58, 86,
281
ITRS International Terrestrial Reference System xxiv, xxvi, 272
J
Jason 1–3, Joint Altimetry Satellite Oceanography Network, series of radar
altimetric satellites 509, 512
JHS Julkisen hallinnon suositukset, Recommendations for Public Administra-
tion 213
JPEG 2000 image format 255
JPL Jet Propulsion Laboratory, NASA xxiii, 331, 353, 355, 495
JUHTA Julkisen hallinnon tietohallinnon neuvottelukunta, Advisory Committee
on Information Management in Public Administration 213
K
KKJ Kartastokoordinaattijärjestelmä, Finnish National Map Grid Co-ordinate
System (obsolete) xxviii, 46, 56–62, 67, 85, 233, 277, 279, 280, 350, 392
KM10 Finnish national terrain model, resolution 10 m 252
KM2 Finnish national terrain model, resolution 2 m 252
L
LAGEOS 1–2, Laser Geodynamics Satellite 481
LAST Local Apparent Sidereal Time 459, 460
LCD liquid crystal display 154
LED light-emitting diode 44, 196
LHC Large Hadron Collider 114
LoD length of day 85, 331, 466
Loran-C marine navigation system 284
M
MHD magnetohydrodynamics 472, 523
MIF member of the Institut de France. The Institute comprises five learned
academies including the Academy of Sciences 194
MRI magnetic resonance imaging 469
MSAS Multi-functional Satellite Augmentation System (Japan) 351
í ¤. û
A CRONYMS ABCDEFGHIJKLMNOPRSTUVWY
xxvii
N
N2000 Finnish height system, epoch 2000.0 78, 79, 93, 96, 111, 213, 235, 279
N60 Finnish height system, epoch 1960.0 78, 93, 96, 111, 277, 279
NAP Normaal Amsterdams Peil, Amsterdam Ordnance Datum, a Western Euro-
pean height datum 78, 93
NASA National Aeronautics and Space Administration, US xxiii, xxvi, 353, 464,
465, 495
NGS National Geodetic Survey, US 86
NLS National Land Survey of Finland 86, 87
NNSS Navy Navigation Satellite System, ”Transit”, ”Doppler” 284, 286, 287
NOAA National Oceanic and Atmospheric Administration, US 252, 499
NUVEL global plate-motion model 487
NWP numerical weather prediction 256, 466
O
Omega marine navigation system 284
P
P code Precise / Protected GPS code 291–293, 295–297, 300, 303, 347
PAGEOS 1966-56A, Passive Geodetic Earth Orbiting Satellite 464
PC personal computer 155, 300
PCMCIA memory card bus standard 155
PDOP position DOP 314–316, 318, 333
PPP Precise Point Positioning, a precise geodetic positioning technique using a
single GNSS receiver 319, 353
R
RINEX Receiver-Independent Exchange Format 300, 302, 333, 353–355
RS232 Recommended Standard 232, serial interface 155
RTCM or RTCM-SC104, “Radio Technical Commission for Maritime Services
Special Committee 104”, a popular differential GNSS standard xxvii,
352
RTK real-time kinematic positioning xxviii, 21, 170, 214, 218, 222, 228, 234, 236,
319, 347–350, 352–354
S
SA selective availability (GPS) 345
SAR synthetic-aperture radar 252, 494, 505
SBAS satellite-based augmentation systems xxiv, xxviii, 351, 352, 354
Seasat 1978-64A, radar altimetric satellite 511
SI Système International d’Unités, International System of Units 23–26, 185,
426, 441, 442, 446
SMS Short Message Service (mobile telephony) 21
í ¤. û
xxviii ABCDEFGHIJKLMNOPRSTUVWY A CRONYMS
SoL Safety of Life 291, 351

SOPAC Scripps Orbit and Permanent Array Center, San Diego, USA 354
SP3 Standard Product 3, precise ephemeris data format 328–330
SRTM Shuttle Radar Topography Mission 252, 264
SST satellite-to-satellite tracking 497
T
TAI Temps Atomique International, International Atomic Time 468, 469
TDOP time DOP 314, 316
TEC total electron content 503
TIN triangulated irregular network 255
TOPEX/Poseidon 1992-52A, radar altimetric satellite 509, 510, 512
U
UDP User Datagram Protocol 481
UNAVCO A non-profit university-governed consortium that facilitates geoscience
research and education using geodesy. Formerly University NAVSTAR
Consortium 118, 488
USB Universal Serial Bus 155, 230
UTC Universal Time, Co-ordinated, global civil time scale 468, 469
UTM Universal Transverse Mercator (map projection) xxiv, 58, 61–63, 154, 208,
221, 279
V
VDOP vertical DOP 314, 315, 318, 321, 324, 333
VGOS VLBI Global Observing System 483
VHS Video Home System 481
VLBI very long baseline interferometry xxviii, 193, 214, 465, 468, 480, 482, 483
VRS-RTK virtual reference station RTK 352
VVJ Vanha valtion järjestelmä (obsolete) 57, 61, 279
W
WAAS Wide Area Augmentation System, an SBAS for the North American area
351
WADGPS wide-area differential GPS 351
WGS84 World Geodetic System 1984 58
Y
YKJ Yhtenäiskoordinaatisto, KKJ’s Uniform Co-ordinate System 61, 62
x
x
í ¤. û
^ The history and societal status of
geodesy
1
[. . . ] Nous avions été sur le fleuve, fort incommodés de
grosses Mouches à tête verte, qui tirent le sang par-tout où
elles picquent ; nous nous trouvâmes sur Niwa [Nivavaara],
persécutés de plusieurs autres espèces encore plus cruelles.
Maupertuis (1738), PDF page 44, page 16
^ 1.1 The figure of the Earth, early conceptions

In traditional societies, undoubtedly the most common conception of
the figure of the Earth was that the Earth is a flat disc extending to
the horizon, with the sky curving like a dome over her. On the inner
surface of the dome, the celestial bodies describe their complicated orbits.
Children also have generally the same conception. Only with formal
education does this “naïve world model” give way. Psychologically, from
the viewpoint of childhood development, this is by no means an easy
process, surely as difficult as it was for society as a whole, back in time
as a stage in the historical development of science.
However, the antique Hellenes were already aware of the spherical
shape of the Earth. Free of preconceptions, they had observed how, during
a lunar eclipse, the Earth cast her shadow on the surface of the Moon.
They also observed that a lunar eclipse that was high in the sky at one
end of the Mediterranean happened near the horizon at the other end.
Assuming that this was one and the same event, this could only mean
that the Earth’s surface must be curved at least in the east-west direction.
And the colder climes found further north were a sign to early observers
that the Earth is also curved in the south-north direction.
Eratosthenes, the “father of geography”, lived 276–195 BCE.1 He was 1
–1–
2 1 T HE HISTORY AND SOCIETAL STATUS OF GEODESY
F IGURE 1.1. A lunar eclipse. The shape of the shadow, always circular, shows
^ that the Earth must be a sphere.
the first to measure the size, or radius, of this spherical Earth. The
astemittaus measurement was the same in principle as the later grade measurements:
measure the length of an arc on the surface of the Earth by geodetic
luotiviiva means, and the difference in direction between the plumb lines at the
ends of the arc by astronomical means. By combining the length ℓ of the
arc and the difference between the plumb-line directions γ one obtains
γ Plumb line
Direction North Pole
of the Sun
ℓ
Alexandria
Earth
Equa X
tor centre
Syene
^ F IGURE 1.2. The grade measurement of Eratosthenes.
1 “Before the common (or Christian) era”.
í Õ ! ¤. û
The figure of the Earth, early conceptions 1.1
3
for the radius of the Earth
ℓ
R = γ. (1.1)
See figure 1.2.

The information on the directions of the plumb lines was obtained from
the midsummer Sun, which in Syene (today’s Aswan) did not throw any
shadows at all.2 In Alexandria, on the other hand, the Sun was not in 2
the zenith but, based on the lengths of shadows, some fiftieth part of a
circle further to the south. Eratosthenes obtained a value for the radius
of the Earth3 of 6317–7420 km — pretty close to the current best value of 3
6371 km.
More information can be found in Torge (2001), pages 5–6.
The principle of triangulation, so important in geodesy, that, in a kolmiomittaus
network consisting of triangles, the geometry may be uniquely determined
if, in addition to the angles of the triangle, only one distance is measured,
was presumably discovered by Gemma Frisius4 in 1533 (Crane, 2002, 4
pages 56–57).
The use of the method for grade measurement also happened for the
first time in the Netherlands, using the numerous church towers dotting
the prosperous but flat country. Snellius5 was among the first to use tri- 5
angulation to determine the length of an arc ℓ. By measuring one length
in the network, and otherwise only angles, he managed to determine the
distance between two cities, Bergen op Zoom and Alkmaar, although the
cities are separated by the broad river branches of the Rhine delta. See
figure 1.3.
The secret of triangulation is that with the aid of angle measurements
one can build, either computationally or graphically, a scale model of
the whole measurement network, where all proportions and shapes are
correct. To determine the true scale, it suffices to measure just one
distance in the model also in reality. In the case of Snellius, this was the
distance pq, in the meadow by Leiden, a baseline of only 326 “roeden”.6 perusviiva
6
2 The story that he used the circumstance that the Sun illuminated the bottom of a well
is apparently a misunderstanding (Dreyer, 1914).
3 In fact he obtained his results in a unit called the stadium, the length of which varies.
The length used by Eratosthenes is controversial.
4 Gemma Frisius (1508–1555) was a Dutch polymath.
5 Willebrord Snell van Rooyen (1580–1626) was a Dutch astronomer and mathematician.
6 The version of the unit used by Snellius was the “Rijnlandse roede”, 3.766 m.
í Õ ! ¤. û
Polaris
Alkmaar Plumb line
Alkmaar
Base
extension Amsterdam
network
Baseline
Leiden pq
Arc length
W
VZ Utrecht
The Hague
Gouda
Rotterdam
Dordrecht
Polaris
Breda
Bergen op Zoom Plumb line
Bergen op Zoom
F IGURE 1.3. The Snellius grade measurement. The length of baseline pq is

326.45 roeden (1229 m). This length was derived through a lo-
cal base extension network from the only measured length, the
original baseline of length 87.05 roeden (328 m) (personal commu-
nication L. Aardoom).
^ V = Voorschoten, W = Wassenaar, Z = Zoeterwoude.
Using astronomical position determination, one may determine the

difference between the directions of the plumb line in two locations,
see figure 1.4. When travelling along the meridian in the north-south
direction, the absolute direction of the local plumb line, the direction
with respect to the stars, changes. The local plane of the horizon, always
perpendicular to the plumb line, the local direction of gravity, also turns
by the same amount in the same direction.
The direction in space of the rotation axis of the Earth is very stable
due to the gyroscope phenomenon. It points to a place in the sky near
í Õ ! ¤. û
Newton’s laws and the figure of the Earth 1.2
5
α UMi (Polaris)
North Pole
Pole height
Plumb line
d Plane of horizon
∆Φ
F IGURE 1.4. Astronomically determining the difference in the north-south direc-

tion ∆Φ of plumb-line directions. From the direction difference and
the metric distance d, one can determine the radius of curvature
of the Earth R = d ∆Φ , assumed constant.
/︁
^
the star α Ursae Minoris, or Polaris. This star gives us the direction
of the north. The latitude Φ of a location is obtained by determining
astronomically the elevation angle of this celestial pole above the horizon.
This is easiest to do using Polaris, although a precise determination is a
little more involved.
By thus measuring astronomically the difference in plumb-line direc-
tions between Alkmaar and Bergen of Zoom, and combining this with
the metric distance obtained by triangulation, Snellius managed to deter-
mine the radius of curvature of the Earth. The method is referred to as
grade measurement.
^ 1.2 Newton’s laws and the figure of the Earth

The understanding of the figure of the Earth made a great leap for-
ward when Newton7 published in 1687 his main work, the Principia 7
(Philosophiæ Naturalis Principia Mathematica, “Mathematical principles
of natural philosophy [physics]”). In this opus he created the foundations
of the whole of classical mechanics, including celestial mechanics.
7 Sir Isaac Newton (1642–1727) was an English physicist and mathematician, the father
of classical mechanics.
í Õ ! ¤. û
Newton: 1 230 .
/︁
Huygens: 1 578 .
/︁
Mass evenly distributed
All mass in the centre
Current understanding: 1 298 .

/︁
Denser core inside mantle, thin

crust (green)
F IGURE 1.5. Different mass distribution models for the Earth, and their theo-
^ retical flattening values.
The universal law of gravitation Between two masses m 1 and m 2

acts an attraction F of strength
m1 m2
F =G ,
r2
in which r is the distance separating the masses. The constant G
is Newton’s universal gravitational constant, the value of which is
6.674 · 10−11 m3 kg s2 .
/︁
This attraction acts between all pairs of masses. So, not only does the
Earth’s attraction act on the Moon and the Sun’s attraction on the Earth,
but the Moon’s attraction also affects the Earth, etc. In geophysics again,
we know that the attraction works between all parts of the Earth: the
sea, atmosphere, mountains all affect the gravitational field surrounding
the Earth. And, because our Earth consists of materials that — however
more of less reluctantly — deform under the influence of external force,
gravitation also shapes the physical figure of the Earth.
In the Principia, Newton calculated, using his famous laws, that a
í Õ ! ¤. û
Newton’s laws and the figure of the Earth 1.2
7
Meridian Length of a degree in Lapland
Length of a
degree in Peru
rL b
rP
a
a
^ F IGURE 1.6. Parameters of an ellipsoid of revolution.
homogeneous, liquid Earth, in equilibrium and rotating once in 24 hours,

with gravitation acting between its elements of liquid, would be flat-
tened at the poles by centrifugal force (figure 1.6). The definition of the
flattening (oblateness) is
a−b
f= a , (1.2)
in which a and b are the semi-major and semi-minor axes of the Earth iso- ja
ellipsoid; in other words, the equatorial and polar radii. pikkuakselin
puolikkaat
The theoretical flattening calculated by Newton was f = 1 230 .
/︁
The assumption that the Earth is of homogeneous density is not cor-

rect. Christiaan Huygens calculated in 1690, by assuming that all the
Earth’s mass is concentrated in her centre — or at least, that the Earth’s
attraction emanates from her centre — that the flattening would only be
f = 1 578 . As we know today, the truth lies between these two extremes:
/︁
the density of the Earth’s crust is about 2.7 g cm3 , that of the underlying
/︁
mantle is 3.0–5.4 g cm3 , and the density of the iron core of the Earth is
/︁
10–13 g cm3 . The average density of the whole Earth is about 5.4 g cm3 . So,
/︁ /︁
while the density increases a great deal toward the centre of the Earth, a
large part of the Earth’s mass is nevertheless far from her centre.
In Newton’s days there were influential scientists, like the astronomer
Cassini,8 who believed that the Earth was elongated like a rugby ball, 8
b > a, and not flattened. An empirical answer to the question was needed!
The flattening issue remained unsolved until half a century later, when
8 Jean Dominique (Giovanni Domenico) Cassini (1625–1712) was an Italian-French

astronomer — explorer of the Saturn system — mathematician, map maker, and
engineer.
í Õ ! ¤. û
the French Academy of Sciences organised two expeditions, one to Finnish

Lapland — then part of the Swedish empire — (1736–1737), the other
to Peru, South America (1735–1744). The goal of the expeditions was to
measure, by geodetic and astronomical means, the length of a meridian
arc of one degree on two different latitudes, one close to the equator in
Peru, the other close to the North Pole in Lapland in the Torne river
valley. This was thus a similar grade measurement to the one Snellius
had carried out over a century earlier. . . . but this measurement took
place far away from the home country, in strange lands in different
climate zones, one of them even beyond the ocean.
The idea of the measurement is illustrated in figure 1.6. Using as-
tronomical measurements, a baseline is established in the north-south
direction, at the end points of which the directions of the plumb line differ
from each other by one degree. Over land, the distance between the points
9 is measured in metres.9 If Newton was right, the length of a degree close
to the North Pole would be greater than that of one close to the equator,
in other words, the radius of curvature of the Earth would, at the poles,
be longer than at the equator:
rL > rP .
The joint result of both expeditions was an empirical flattening of f =

1 210 . For comparison, the current best value for the flattening of the
/︁
Earth is f ≈ 1 298.257 .
/︁
Much has been written about the adventures of the expedition led
by Pierre L. M. de Maupertuis in the Torne river valley 1736–1737, for
example Rovaniemi, The Degree Measurement Expedition, and in the
French original (Maupertuis 1738).
10 Of the later grade measurements we may mention Struve’s10 Russian-
Nordic grade measurement (the “Struve chain”) 1816–1855, which
extended from Norway’s Atlantic coast all the way to the Black Sea
(Wikipedia, Struve Geodetic Arc). Some points of the chain have also
been preserved on Finnish territory.
9 In reality, the French Academy of Sciences measurements used the toise as the unit of
length, as the metre had not been invented yet.
10 Friedrich Georg Wilhelm von Struve (1793–1864) was a Russian astronomer and
geodesist.
í Õ ! ¤. û
The mathematical figure of the Earth or geoid 1.3
9
Meridian
Kittisvaara
Lapland
grade measurement
Pullinki
Niemivaara
Arctic circle
Poiki-Torni Horilankero
Rectory of Higher Tornio Aavasaksa

Baseline
Luppio Huitaperi
x
Kaakamavaara
Nivavaara
Peru
grade measurement
Tornio
F IGURE 1.7. The grade measurement project of the French Academy of Sciences:
^ the Lapland grade measurement network.
^ 1.3 The mathematical figure of the Earth or geoid

The changes from place to place in the direction of the plumb line along
an arc on the Earth’s surface can thus be used to find out about the
true figure of the Earth. In the previous section we described how the
grade-measurement project of the French Academy of Sciences exploited
this phenomenon for determining the figure of the Earth, assuming that
the Earth has the figure of an ellipsoid of revolution.
With the aid of more precise geodetic measurements it was noticed that
this assumption does not precisely apply. Already in the context of the
Peru grade measurement, Pierre Bouguer11 noticed that the direction 11
of the plumb line on both sides of the Andes had a tendency to deflect
11 Pierre Bouguer (1698–1758) was a French polymath, mostly a geophysicist and

shipbuilder.
í Õ ! ¤. û
F IGURE 1.8. The northernmost point of the Struve chain in Fuglenes, Norway,
^ Franz (2005).
towards the mountain range, and he interpreted this correctly as an

12 expression of Newtonian gravitation or attraction. George Everest12 in
India noticed the same phenomenon near the Himalayas. As geodetic

measurements, especially astronomical determinations of the direction of
the plumb line, progressed, the understanding spread that the figure of
the Earth is irregular.
People started to speak about the “mathematical figure of the Earth”
13 or geoid (J. B. Listing,13 1873), the continuation of mean sea level under
the continental masses, a surface that is everywhere perpendicular to

plumb lines, and along which a fluid at rest — like sea water — would
settle. See figure 1.9.
14 In 1862, under the leadership of the Prussian J. J. Baeyer,14 the “Mittel-
europäische Gradmessung” (“Central European Grade Measurement”)
12 Sir George Everest (1790–1866) was a geodesist and geographer born in Wales,
director-general of the Survey of India. In 1865 the highest peak in the Himalayas was
named Mount Everest in his honour but against his protestations.
13 Johann Benedict Listing (1808–1882) was a German mathematician, the inventor of
topology.
14 Johann Jacob Baeyer (1794–1885) was a Prussian military officer, geodesist, and
diplomat of geodesy.
í Õ ! ¤. û
The mathematical figure of the Earth or geoid 1.3
11
Geoid Topography (terrain)
Plumb line
Plumb-line direction
Con
tine Reference ellipsoid
nt
Mean sea surface
Se
a
^ F IGURE 1.9. Deviations of the plumb line and the shape of the geoid.
was established, which later developed into the global organisation IAG,
International Association of Geodesy. Its task was determining the figure
of the Earth or geoid, especially on the European territory, and uniting
the geodetic networks of Europe into a single network. This objective was
not properly achieved until 1950, when the first joint European network
adjustment ED50, “European Datum 1950”, was completed, even though verkkotasoitus
only on the Western European territory.
Elsewhere, for example in North America, continental-scale trian- kolmiomittaus-
gulation networks were being measured, to determine the figure and verkko
flattening of the Earth as well as the locations of points on the Earth’s
surface in support of map-making. Determining the general figure of the
Earth is however difficult from the Earth’s surface using classical geode-
tic techniques, because extended networks on the Earth’s surface are not
geometrically strong, and their unification across oceans is impossible.
Satellites have fundamentally changed this picture: satellite tech-
niques have provided precise data on, for example, the flattening of the
Earth by exploiting the rapid precession of the satellite orbital plane it ratatason
causes. Several weeks after the launch of Sputnik, much better values prekessio
were already becoming available for the flattening, and the American
Vanguard 1 satellite showed the Earth to be “pear shaped” — although
only very, very slightly.
í Õ ! ¤. û
A
B
B
B
A
F IGURE 1.10. The geodesic in the plane, on the sphere and on the ellipsoid of
revolution. The arrows depict the local normal to the surface.
The plane of the normal section is not uniquely defined for the
^ ellipsoid of revolution.
^ 1.4 The geodesic

In the plane, the shortest path between two points is the straight line.
pyrähdys- On the curved surface of a sphere, or of an ellipsoid of revolution, the
ellipsoidi shortest path is a curve. In the case of the sphere it is an arc of a great
isoympyrä
circle;15 in the case of the ellipsoid it is a surface curve that is ever so
15
slightly S-shaped.
Figure 1.10 shows this general concept of “the shortest path within a
geodeettinen surface” or geodesic.
viiva
The figure shows another curiosity: both in the plane and on the sphere,
the plumb lines, or normals, to the surface at the end points A and B lie
in the same plane, together with the connecting line or curve itself. This
normaali- plane is called the normal section. In the case of an ellipsoid of revolution,
leikkaus this is however not the case: in the general case there is no normal section
containing both surface normals. This effect is indeed extremely small

and it can be ignored in all but the most precise calculations.
Traditional geodetic measurement networks on the Earth’s surface,
especially triangulation networks, can be considered to consist of mea-
15 Between any two points there are two great-circle arcs, one the shortest and the other
the longest between the points. Only in the case of antipodes is there an infinity of
great-circle arcs, all 180◦ long.
í Õ ! ¤. û
The flattening of the Earth and gravity 1.5
13
surement lines that are geodesics on some reference surface, usually a
reference ellipsoid.
In practice, the geodetic instruments and signals are never precisely on tähys
the reference surface, but are some distance above or, more rarely, below
it. Then, a reduction of the raw observations to this reference surface
must be made. This applies to both angle and distance observations.
Generally, the corrections needed are small.
^ 1.5 The flattening of the Earth and gravity

As we explained above, the Hellenes were already well aware of the
sphericity of the Earth and even her approximate size. During the 17th
and 18th centuries, the idea developed of the flattening of the Earth, that
is, the ellipsoid of revolution as descriptive of the figure of the Earth.
Astronomers observed the flattening of Jupiter, describing it correctly as
a dynamic phenomenon caused by the planet’s fast rotation.
On a flattened Earth, of course also gravity changes with latitude. This
was in fact observed with a pendulum clock, the pendulum of which had
to be shortened when travelling to Cayenne in French Guyana (Jean
Richer,16 1672), so the clock would run on time again. Gravity is weaker 16
near the equator than in France. Upon returning to France, Richer had
again to lengthen his pendulum, so it would swing on time.
Newton and Huygens calculated theoretically a value for the Earth’s
flattening f , equation 1.2. A. C. Clairaut17 again derived his famous equa- 17
tion giving the relationship between flattening f and “gravity flattening”
γb − γa
β= γa .
Here, γa and γb are the accelerations of gravity on the equator and on the
poles, respectively. Clairaut’s theorem’s approximate but elegant form is
(Heiskanen and Moritz, 1967, equation 2-99)
5 ω2 a
f +β = ,
2 γa
in which ω is the rotation rate or angular velocity of the Earth. The

2 /︁
expression ω a γa represents the centrifugal force on the equator caused
by the Earth’s rotation as a fraction of gravity there.
16 Jean Richer (1630–1696) was a French astronomer.

17 Alexis Clairaut (1713–1765) was a French mathematician and astronomer.
í Õ ! ¤. û
^ 1.6 Reference surfaces and reference systems

As a reference surface for heights, the geoid is used, that level surface, or
equipotential surface, of the gravity field of the Earth, which is on average
on the same level as the mean sea level. This is thus the surface, the
minimum state of potential energy, which sea water would reach if there
suolaisuus were no currents in the ocean, no salinity or temperature differences,
no variations in air pressure above it, and so on. In reality, all these
disturbing factors exist and the global mean sea level deviates by over
one metre from this equipotential surface, both above and below it. Of
this deviation, part varies in time, like the tides, part is permanent, the
meritopografia sea-surface topography.
The levelled heights of countries may be understood as heights above

this geoid surface. In practice, the heights are tied to sea-level obser-
vations by tide gauges or mareographs operating on the coast, and the
reference level is transferred inland by levelling, thus creating a height
system.
More about the geoid and heights can be found in chapter 4.
A space geodetic measurement method, like GPS, provides a way to
determine the height of a point from the reference ellipsoid, because the
ellipsoid of revolution is a simple mathematical surface in space. The ref-
erence ellipsoid is also otherwise a good approximation of the true figure
of the Earth, which has traditionally been used as a reference surface for
tasoitus the adjustment of national or continental triangulation networks.
In very small areas and for many purposes, the “flat Earth approxima-
tion”, the assumption that the curvature of the Earth may be neglected,
continues to serve us. The locations of points may be described by two
plane co-ordinates and the height by one height co-ordinate, the vertical
distance in metres from the reference surface. Depending on the applica-
tion, a “small area” may be a building site, a city, or all of Finland — or,
in special situations, taking special care, even an area the size of Europe,
for example Strang van Hees (1990).
^ 1.7 The sub-fields of geodesy

Geodesy is defined as “the science of measuring and mapping the Earth’s
surface” (Helmert, 1880; see Torge, 2001 page 1). This definition contin-
merenpohja ues to hold today. Geodesy also includes the mapping of the sea floor, as
well as the determination of the gravity field of the Earth, or geopotential.
í Õ ! ¤. û
The sub-fields of geodesy 1.7
15
More and more, the study of changes in the shape of the Earth and of the
physical mechanisms causing these, geodynamics, has also become a part
of the geodetic research field.
Thus, geodesy, in particular physical geodesy, belongs to the Earth sci-
ences. However, geodesy also clearly belongs to the engineering sciences.
In Finland, geodesy is currently (2018) being taught both at the Univer-
sity of Helsinki (two docents, external lecturers) and at Aalto University
(one professor, docents, external lecturers).
According to Torge, geodesy may be divided into three sub-fields:
Global geodesy, also “measuring the Earth” (geomensuration, German
Erdmessung). More precisely (Torge, 2001, page 2):
“The problem of geodesy is to determine the figure and

external gravity field of the earth and of other celestial
bodies as a function of time, from observations on and
exterior to the surfaces of these bodies.”
Figure of the Earth:

◦ The physical figure of the Earth: the solid surface of the Earth,
the interface between solid and gaseous or liquid matter —
atmosphere, ocean — with all its mountains and depths.
◦ The mathematical figure of the Earth or geoid: the equipo-
tential surface of the Earth’s gravity field that on average
coincides with the mean sea surface, and that may be consid-
ered the continuation of the mean sea surface under the land
masses. See section 1.3.
Geodetic surveying, surveying science. To this belongs the measure-
ment of national — and today, international — geodetic and gravi-
metric base networks as a basis for mapping. In this work, the runkoverkko
Earth’s curvature and gravity field must be taken into account.
Ordinary surveying (“plane surveying”). To this belong, for example,
topographic surveying and engineering surveying measurements. maastomittaus
These measurements are of such areal extent and accuracy level,
that in all calculations the curvature of the Earth may be neglected
or taken into account by simple correction formulas.
These measurements are made, besides by geodetic means, often
also photogrammetrically by means of aerial photography. The
objective of the work is always the production of geometrically
precise and correct mapping material for use by society.
í Õ ! ¤. û
Computations may be done using plane co-ordinates x and y. Maps

for small areas may be drawn without the use of a map projec-
tion, and the separate height co-ordinate H may be assumed to
be purely metric (“metres above sea level”) without unpleasant
consequences.
^ 1.8 Topographic surveying: from terrain to map

The booklet “Maastomittaus ja kartoitus” (“Topographic Surveying and
Mapping”) (Heiskanen and Härmälä, 1963) states:
maastomittaus “Topographic surveying and the mapping that goes with it

most often serve to provide of a larger or smaller area a
depiction that is as correct and precise as possible, in the form
of a map.”
Society needs maps and location-based information for many purposes.

In today’s society, the ownership of real estate, its buying and selling, and
hypoteekki especially its use as collateral (mortgage) for loans aimed at maintaining
and developing the property’s value, are foundational for a modern society
with a high level of investment. To this end exists the cadastral system,
which registers as reliably as possible the state of properties and the
rights tied to them. There are millions of real-estate properties and
18 parcels and their collective monetary value is astronomical.18
paikkatieto Another use of maps and geospatial information that is important

19 to society, is the planning and construction of infrastructure:19 roads,
railways, bridges, tunnels, canals, airports, harbours, power plants, water

works, telephone and data networks and so on. This is a vital public
service from the viewpoint of economic productivity.
In every developed society, construction is limited in some sense. You
kaavoitus may not build what and how you wish, even on land that you own. Zoning
regulates in a co-ordinated way the purposes for which land may and
kaava may not be used. These regulations are contained in zoning plans, the
18 According to a report by RAKLI ry (2014) “All of Finland’s building stock together has
a value, including parcels, of some 480 billion euros”. This means almost a hundred
thousand euros for every Finnish man, woman and child.
19 There are many definitions of infrastructure. The most perceptive one is undoubtedly
that it is “those things the importance of which does not reach our awareness until they
stop working”.
í Õ ! ¤. û
Topographic surveying: from terrain to map 1.8
17
approval process of which is statutorily prescribed, public and consisting
of several stages. The reason for this is that zoning affects the value
of real-estate property, so the legal status of property owners requires
that the democratic approval process of zoning plans contains adequate
instruments of appeal. Maps and other surveying-based information
sources are essential for this process.
In Finland, the planning of land use and the associated local infrastruc-
ture construction happens for the most part in the public administration,
most often in municipalities. We speak of spatial planning. yhdyskunta-
suunnittelu
Topographic surveying is, in terms of volume, the overwhelmingly
largest field of application of land-surveying.
^ 1.8.1 Spatial planning

Spatial planning is a continuous activity, to which belongs zoning, plan-
ning of land use and the built environment, and managing building
activity. In spatial planning a major role falls to the municipality. City
planning and regional planning are forms of spatial planning.
The technical planning and building of a community require reliable
information on the environment which is being planned and built. Figure
1.11 gives an idea of all the places where geospatial information plays a
role in the continuous process of spatial planning and construction.
Topographic surveying is present in the whole process:
◦ Topographic information must be measured onto the map in a maastotieto
certain co-ordinate reference frame.
◦ When the zoning plan is ready, it has to be set out on the terrain.
◦ The properties have to be measured and mapped.
◦ The technical structures must be placed onto the terrain.
Topographic surveying has the following technical tasks:

◦ Creating a base network for mapping: getting the measurements
into a certain, known co-ordinate frame.
◦ Mapping terrain details while using base-network points: detail kartoitus-
survey. mittaus
◦ Setting out: the transfer of a plan onto the terrain, to be realised in maastoon
its correct location. Setting or staking out is, in a way, the inverse merkintä
problem of mapping.
í Õ ! ¤. û
^ 1.8.2 Carrying out the tasks of topographic surveying

Successfully carrying out the tasks of topographic surveying requires
that the surveyor understands the following things:
◦ geodetic instruments
◦ measurement technique, measurement planning, measurement
conditions
◦ geodetic computation: co-ordinates and transformations, derived
quantities, accuracy
◦ zoning-plan calculations, setting out
◦ geographic information systems
◦ maps, printing, reporting.
Professor Matti Martikainen drafted an attractive tableau, here pre-

sented as tableau 1.1, on the role of the measurement plan and the
place of topographic surveying in the whole of the measurement and
mapping process. Our version is slightly modernised. The various parts
of the tableau belong to the sub-fields of geodesy, photogrammetry, and
cartography.
^ 1.8.3 The end product of topographic surveying

The visible end product of topographic surveying is a map. A map must
be, before all else, correct, but also clearly drawn, give all relevant infor-
mation, and it is nice if it is attractive as well (Heiskanen and Härmälä,
1963).
A suitable scale is chosen for the map, which defines the level of accu-
racy of the information presented. The scale is chosen to suit the purpose
yleistys of use of the map. The objects to be presented need to be generalised in a
suitable way: too-small details should be taken out, however, essential
20 details should be made clearer.20
Many maps are digital. Then, the significance of the scale is not quite
as clear.
Furthermore there is much location-related information in numerical
sijainti- ja form (geospatial information, consisting of location and attribute data)
ominaisuustieto
20 For example, on a road map, the widths of the drawn roads have no relationship
whatever with the widths of the roads in the terrain! The drawn width expresses the
importance of the road to traffic. This is how generalisation works.
í Õ ! ¤. û
x
◦ Geodetic measurements
Topographic surveying: from terrain to map

◦ Topographic survey
◦ Aerial mapping
◦ Topographic mapping
Observation, measurement
◦ Processing of measurements
Maintenance of maps
◦ Entry into GIS system
and geographic data
í
Spatial planning:
x Management of geospatial info
Õ ! ¤. û
Land use
◦ Maps
◦ Real estate Land-use planning as a
◦ Geographic information
◦ Municipal continuous process
systems
technology
◦ Infrastructure
◦ Zoning
Management of ◦ Spatial planning
◦ Spatial planning and
land use and planning ◦ Technical planning (roads, streets,
building
street furniture, sewers, city
◦ Property formation and
Zoning plan and base map heating, telephony and data
registration
networks)
X
1.8
^ F IGURE 1.11. The roles of the map and topographic surveying in spatial planning and civil engineering.
19
as well as “metadata”: data describing other data, for example map

information. The legend on a paper map is an example of metadata.
The following items of information are or can be part of the end product:
◦ Plane co-ordinates. These state the location inside a municipality,
a country, the world. On the map, co-ordinate curves, a co-ordinate
grid.
korkeuskäyrä ◦ Height information, for example height contours, height values of
points, possibly profiles.
◦ The forms of the physical Earth’s surface, presented in various
ways.
ominaisuustieto ◦ Attribute data. The measured objects are shown on the map in ac-
cordance with an agreed presentational style. A suitable identifier
or symbol is given to every piece of information.
The information shown on the map may also be divided into natural and
cultural data. See table 8.3.
^ Self-test questions
1. You have a car with an FM radio, a clock, a passport, a few weeks
of free time, and money for food and lodgings and fuel. How do you
establish to your own satisfaction that the Earth is round, not flat?
Leah (2017).
2. How does Newton’s universal law of gravitation explain that the
larger celestial bodies are approximately spherical? Why are they
not precisely spherical?
3. Describe the hypotheses of Newton and Huygens on the interior
mass distribution of the Earth and its effect on the Earth’s flatten-
ing.
4. Describe the idea of grade measurement as well as the objective of
the grade-measurement project of the French Academy of Sciences.
5. What is, in equation 1.1, the unit in which the angle γ is expressed?
6. What is the “mathematical figure of the Earth”, and how is it related
to the direction of the plumb line?
7. What is the geoid, and what is the reference ellipsoid?
8. Clairaut’s theorem gives the relationship between the rotation of
the Earth, her flattening, and her “gravity flattening”. Intuitively,
í Õ ! ¤. û
Self-test questions
^ T ABLEAU 1.1. Topographic surveying as part of the whole measurement and mapping process.
Measurement plan
Base network measurement (GNSS), computation
í
Collection of Mapping surveying Existing maps Digital imagery Laser Photographs

topographic data (often GNSS-RTK) scanning
Õ ! ¤. û
Processing of Geodetic computation Digitisation Image processing Processing Stereo

topographic data photogrammetry
Data integration Geographic information system (GIS)
Products Graphical Numerical Textual Metadata
Data presentation to Topographic maps Point data bases Reports Map legend
the end user Thematic maps Elevation models SMS alerts
Special maps Digital CAD models Navigator voice
Customer printouts 3D instructions
visualisation
21
give two reasons why gravity on the equator should be weaker than
at the poles, and one reason why it should be stronger.
9. What is a geodesic?
10. What is the task of topographic surveying according to Heiskanen
and Härmälä?
11. Name the three main fields of application of topographic surveying
in society.
12. What are the three technical tasks of topographic surveying?
13. What is generalisation of map information? Give an example.
14. Think of reasons why society would want to disallow building a
bicycle factory in the middle of the residential area of Eira, Helsinki.
í Õ ! ¤. û
^ Geodetic measurements and
co-ordinates
2 METPΩ XPΩ
”Use measure”, Pittacus of Mytilene (640–568 BCE),

appears in the seal of the BIPM
^ 2.1 Units of measurement
^ 2.1.1 Definitions
When we talk in geodesy, like more generally in physics, about measure-
ment units to be used, we distinguish between units and quantities. For
example, length is a quantity, the unit of which may be, for example, the
metre [m].1 So, for example: 1
The length of the distance AB is 15 metres, or 15 m.
Quantity Value Unit Symbol
Length 15 Metre m
In the literature, the term dimension is also used, for example the
dimension of volume is length3 , the dimension of acceleration is length ×
time−2 . This way is used to express how the definition of a certain
quantity depends on the definitions of other quantities. For example,
if one wants to precisely measure accelerations, one has to precisely
measure both lengths (distances) and time intervals. This belongs to the
field of metrology, the science of measurement.
1 The official symbol of a unit according to the SI system is always written in upright
letters, not in italics! Italics is used for mathematical symbols. So E = mc2 , but
J = kg m2 s2 .
/︁
– 23 –
24 2 G EODETIC MEASUREMENTS AND CO - ORDINATES
F IGURE 2.1. A public standard metre in Paris. Wikimedia Commons, Standard

^ metre in Paris.
In Finland, as in most countries of the world, the SI system or Interna-

tional System of Units (SI = Système International d’Unités) is used.
The system consists of base units and derived units. There are seven
base units, BIPM, SI base units, table 2.1.
^ 2.1.2 Prefixes
2 One may add to the SI units (but not to the additional units,2 see subsec-
etuliite tion 2.1.3!) a prefix indicating the order of magnitude according to table
2.2. The table is not complete.
^ T ABLE 2.1. Measured quantities, units and their symbols. The list of derived
quantities is incomplete.
Quantity Unit Symbol How derived
Base units Length Metre m

Mass Kilogram kg
Time Second s
Electric current Ampere A
Temperature Kelvin K
Luminous intensity Candela cd
Amount of substance Mole mol
Derived units Plane angle Radian rad
Solid angle Steradian sr
Frequency Hertz Hz s−1
Force Newton N kg m s−2
Pressure Pascal Pa N m−2
Energy Joule J Nm
Power Watt W J s−1
Electric tension Volt V W A−1
Electrical resistance Ohm Ω V A−1
í Õ ! ¤. û
Units of measurement 2.1
25
^ T ABLE 2.2. Prefixes indicating decimal order of magnitude in the SI system.
Value Prefix Symbol Value Prefix Symbol
+1 deca da +6 mega M
+2 hecto h +9 giga G
+3 kilo k +12 tera T
−1 deci d −6 micro µ
−2 centi c −9 nano n
−3 milli m −12 pico p
Example: 1 MHz = 10+6 Hz.
^ 2.1.3 Non-SI units accepted for use with SI

In daily life, and in many scientific disciplines, a generous number of
non-SI units are in widespread use and are not going away soon. For
example, a calendar, or a clock, would not be very practical in kilo- and
megaseconds. The international metrological community, realising this,
has created a category for these units, “Non-SI units accepted for use with
the SI” (Wikipedia, Non-SI units mentioned in the SI).
Table 2.3 gives some often-used additional units “accepted for use with lisäyksikkö
the SI”.
^ T ABLE 2.3. Non-SI units accepted for use with the SI.
Quantity Base unit Accepted unit Symbol
Time s Hour, minute, second hms

Day d
Year a
◦ ′ ′′
Plane angle rad Degree, minute, second
Gon gon, g
◦
Temperature K Degree Celsius C
Volume m3 Litre l
Mass kg Tonne t
2 This is nevertheless widely done, for example kcal means kilocalorie, a traditional unit
of energy content of chemical substances and food. In computing again, the prefix k, or
sometimes K, expresses the non-standard binary order of magnitude 1024× (Wikipedia,
Kilobyte). And the monetary unit k$ is also used!
í Õ ! ¤. û
Celsius temperature is obtained from temperature in kelvin by sub-

tracting 273.15 K from it:
T1 = 0 ◦ C = 273.15 K, T2 = 0 K = −273.15 ◦ C.
Temperature differences are the same on the Celcius and Kelvin scales:
∆T = 1 ◦ C = 1 K.
^ 2.1.4 Length and time

Of the base units, both length and time are based on atomic phenomena.
The metre is the distance travelled by light in a vacuum in
1 299 792 458 seconds. In practice, the metre is realised by an iodine-
/︁
stabilised helium-neon laser, the wavelength of which is very precisely

(2.5 parts in 1011 ) known (Penzes, undated).
The second, again, is “the duration of 9 192 631 770 oscillation periods of
the radiation corresponding to the transition between the two hyperfine
levels of the ground state of the 133 Cs atom” [official SI definition]: it is
based on the use of a caesium clock (Hardis, 2018).
^ 2.1.5 Units of angle

The radian and steradian units are dimensionless numbers (“bare num-
bers”) because they are ratios. For example, an angle in radians is the
ratio between the length of a circular arc and its radius, and thus dimen-
sionless. There exist however other units of angle, like the degree and
the gon. Therefore it makes sense, in a way, to also treat all of these as
units.
The situation is somewhat similar for logarithmic scales: Richter (total
energy of earthquakes), the magnitude scale of stars, the decibel (dB)
scale, the ◦ DIN scale for the light sensitivity of photographic emulsions,
and the pH, the degree of acidity of a solution.
Degrees In ordinary life we use as unit of angle the degree, symbol ◦ .
Geographical latitude and longitude as read from a map are also
oikokulma commonly given in degrees. A right angle is 90◦ , and a straight
angle 180◦ . In addition to degrees, we have as traditional units the
minute (′ ) and the second (′′ ). These behave like their namesakes
in time measurement: one degree is 60 minutes and one minute
60 seconds.
í Õ ! ¤. û
Measurement error and uncertainty 2.2
27
Calculation example Convert degrees, minutes, seconds to de-
grees and decimals:
)︂◦ (︂ )︂◦
47 33
(︂
56◦ 47′ 33′′ = 56◦ + + =
60 60 × 60
= 56◦ + 0◦. 783333 . . . + 0◦. 0091666 . . . = 56◦. 7924999 . . .
Convert back in the other direction (note the rounding er-

ror!):
′
56◦. 7925 = 56◦ + (60 × 0.7925) =
= 56◦ + 47′.55 =
= 56◦ 47′ + 0′.55 =
′′
= 56◦ 47′ + (60 × 0.55) = 56◦ 47′ 33′′ .
Gon In geodesy and geodetic instrumentation, the gon is often used

as a measurement unit.3 Sometimes the name “new degree” is 3
used. A new minute, or centigon, is 0.01 gon, a new second, or
decimilligon, 0.0001 gon. One notation is 1.2345 gon = 1g 23c 45cc .
In units of gon, a right angle is 100g .
Radians A full circle contains 2π radians, 360 degrees (360◦ ) and 400
gon (400g ). A right angle thus has 2π 4 = π 2 radians.
/︁ /︁
Here are some useful equations to convert an angle given in radians to

gon and degrees. See Kahmen and Faig (1988).
A full circle is
2π rad = 400g = 360◦ ,

)︂g
360 ◦ 400
(︂ )︂ (︂
1 rad = ≈ 57◦. 2957795 = ≈ 63g.6619772,
2π 2π
)︂◦ (︂ )︂g
360 400
(︂
α rad = α = α .
2π 2π
And
)︂g )︂◦
400 360
(︂ (︂
◦ g g
α = α = 1.1111 · · · × α , α = α = 0.9 × α◦ .
360 400
^ 2.2 Measurement error and uncertainty

No measurement is perfectly accurate. Land surveying, being a human
activity, is also prone to error. Just like in computer programming, one
3 Another name used is grad. The unit was taken into use by the French in connection
with the Revolution and introduction of the metric system. Today, it is only used in the
land-surveying field.
í Õ ! ¤. û
should not even try to measure completely errorlessly. A more realistic

goal is, in addition to the usual carefulness, to develop methods by which
◦ Errors of a certain size can be noticed and removed from the obser-
vational material (statistical testing).
◦ The impact of not removed, or not even noticed, errors on the end
tasoituslasku result can be evaluated and minimised (adjustment calculus).
With these methods, measurement errors may be taken into account and
as correct as possible measurement results be produced, the quality of
which, their accuracy, is known, and the magnitude of any errors possibly
still hiding out in them is also known or at least guessable.
tasoitus The processing of the measurements, the adjustment, will yield the
following results:
◦ The “best” value for the unknown quantities, based on the measure-
ments:
– the most likely value
odotusarvo – the statistical expectancy
– a value with respect to which the remaining deviations are as
little damaging as possible.
◦ A judgement on the “goodness” of the original measurements, their
4 precision.4 We speak of standard deviation or mean error. Nowa-
days the official term is standard uncertainty.
◦ Similarly, a judgement on the “goodness” or precision of the end
results, the computed values (estimates) of the unknown quantities
of interest.
karkea virhe ◦ A judgement on the possible occurrence of gross errors and their
maximal magnitude after statistical testing has been done: relia-
bility.
◦ A judgement on the possible presence of systematic errors and their
magnitude.
5 Measurement error is the difference between the value to be measured5
and the measured value. The measured value is the result of an, often
4 In English the terms “precision” and “accuracy” are used. Precision refers only to
statistical spread, the variability of measurement values, whereas accuracy refers to
the deviance of the observation values from the “correct values”. The latter concept thus
also includes systematic errors.
5 The “true value”, although one may philosophically ask if it even exists.
í Õ ! ¤. û
Measurement error and uncertainty 2.2
29
complex,6 measurement process. 6
^ 2.2.1 Types of errors

We may divide the errors arising from this measurement and reduction
process into the following categories:
Random errors represent the inherent, natural imprecision of the mea-
surement process. Often it is assumed that these random errors
are normally distributed, meaning the distribution of errors is
a pretty Gaussian bell curve or normal distribution — on which
more later on.
If this is so — and one should not assume this without further
testing! — one has available the least-squares adjustment method, pienimmän
which minimises the joint impact of random errors in the end neliösumman
tasoitus
result.
Gross errors are caused by human mistakes or measurement device karkea virhe
malfunctions. They happen only now and then, and they cannot
be described by a statistical distribution. An example of a gross
error is writing a digit wrong into an observation notebook.
One strives to eliminate gross errors by statistical testing. Sta-
tistical testing is a broad scientific subject. As a rule of thumb,
one may say that if a function of measurements differs from its
known expectancy (for example zero) by more than three times its
own, known standard deviation (mean error) σ, there is reason to
suspect that one or more measurements are in error. This criterion
is called the three-sigma rule.
Systematic errors are a sign that there are deficiencies in the theories
used to describe the measurement process, observation geometry
and physics of the observation situation. The occurrence of signifi-
cant systematic errors should lead to a study and re-assessment
of the theoretical assumptions. As an example, we know that the
sum of the angles of a triangle must be 180◦ : the triangle condition.
If the sum of the measured angles of a plane triangle differs sig-
nificantly and consistently from this value, one must, for example,
suspect
◦ lateral refraction in the atmosphere
6 Sometimes the measurement process includes a pretty complex reduction chain or

modelling effort.
í Õ ! ¤. û
(a) (b) (c) (d) (e)

Small random Large random Correlated Systematic Gross
F IGURE 2.2. Examples of different error types. Random precise, random impre-
^ cise, correlated, systematic, gross.
◦ the triangle is so large, that on the curved surface of the

Earth we no longer have a plane triangle and the theorem no
longer applies.
Uncertainty, knowledge concerning the possible size of errors, can also be

divided into different categories (GUM, 2008):
Type A uncertainty is standard uncertainty or standard error calcu-
lated for example from repeated measurements. This type of
uncertainty can be handled by means of statistical theory: the
concept corresponds to that of random errors.
Type B uncertainty is uncertainty that can not be determined by sta-
tistical means from the measurements, but needs to be assessed
in some other way. For example, systematic errors can be deter-
mined by calibration, obtained from information by the device
manufacturer, etc.
There is no uncertainty concept corresponding to gross errors. They

should be avoided or eliminated using statistical testing.
^ 2.3 Stochastic quantities
^ 2.3.1 Discrete stochastic quantities

A measurement process is a random or stochastic quantity. A stochastic
quantity is obtained by doing something the outcome of which is random.
For example, casting a die creates a stochastic quantity n, the realisations
(“throws”) of which are n 1 , n 2 , n 3 , . . . .
{︁ }︁
The possible values of a die cast are the integer numbers 1, 2, 3, 4, 5, 6 .
It is the same with a coin flip, if we take as the values heads being a 0
í Õ ! ¤. û
Stochastic quantities 2.3
31
^ T ABLE 2.4. Die-cast statistics.
Number of Value Mean ± standard

throws 1 2 3 4 5 6 deviation
60 14 8 9 12 9 8 10 ± 2.4
% 13.33 23.33 13.33 15.00 20.00 15.00 16.67 ± 4.08
600 103 91 106 114 87 99 100 ± 9.9
% 16.50 17.17 15.17 17.67 19.00 14.50 16.67 ± 1.65
6000 973 1007 1003 962 1015 1040 1000 ± 28.5
% 17.33 16.22 16.78 16.72 16.03 16.92 16.67 ± 0.47
60 000 10 138 9936 10 057 10 029 9925 9915 10 000 ± 89.5
% 16.52 16.90 16.56 16.76 16.72 16.54 16.67 ± 0.15
{︁ }︁
and tails being a 1. We say that the value set or codomain is 0, 1 , a arvojoukko
discrete value set.
If we perform a die cast again and again — meaning that we collect
realisations n i , i = 1, 2, 3, . . . — we may always after a certain total
number of casts tabulate the results. We obtain the example table 2.4,
which expresses how many casts of the total number were ones, how
many twos, etc.
According to experience, the greater the number of die casts, the
smaller tends to be the deviation of the end result from the ideal outcome,
the outcome in which the frequency of occurrence of every value in per-
cents would be 16.666 . . . %, or 16 . This empirical result is called the law of
large numbers.
^ 2.3.2 Expectancy
Based on this, we may assign to the outcomes of die casts a theoretical
probability value, which expresses how often, “in the long run”, a certain
value will happen. For a balanced (“fair”) die, the probability values are
p(1) = p(2) = . . . = p(6) = 16 .
In the discrete case, the expectancy — the value around which the casts odotusarvo
group themselves, their “centre of mass” — is computed by the equation7 7
N
∑︂
def
E { n} = i · p( i ) (2.1)
i =1
í Õ ! ¤. û
21
which yields 6 = 3.5. Here, N is the number of alternatives. The value
3.5 is not even a possible cast value!
In the case of a fair coin
p(0) = p(1) = 0.5,
and the expectancy is, by the same equation, 0.5. This value is also not a
possible throw.
^ 2.4 Statistical distributions
^ 2.4.1 Real-valued quantities and density distributions

Geodetic measurements are generally stochastic quantities, the value set
or codomain of which is a sub-set of the real numbers R: a continuous set.
For example, the result of a distance measurement is a distance in metres,
8 and it is a real8 number. It is the same for angle measurements. The
[︁ )︁
value set may be bounded, α ∈ 0, 360◦ , but it is in any case continuous.
In the case of a continuous value set, we speak of a probability density
distribution, or a distribution in short. If we make a large number of
measurements of the same object, we may draw a histogram. Divide
up the value set into “baskets”, and count and show the numbers of
measurement results that fall into each basket.
As an example of a continuous statistical distribution may serve the
places of the goals of a football match scored in the goal gate, figure 2.3.
The more measurements, the more bars we may draw, and the narrower
we may make them. In the limit for an infinite number of measurements
we obtain a continuous curve, the probability density distribution, which
expresses the probability with which a measurement result will fall
[︁ ]︁
within some value interval x1 , x2 . See figure 2.4 for an example.
[︁ ]︁
If on a certain interval x1 , x2 the integral
ˆ x2
p( x) dx
x1
7 More generally,
N
∑︂
def
E { n} = v i · p ( i ),
i =1
in which v i is the value alternative with serial number i . For a die, v i = i .

8 Actually a rational number ∈ Q.
í Õ ! ¤. û
Statistical distributions 2.4
33
y
y-histogram
x-histogram
Two-dimensional histogram
F IGURE 2.3. An example of a stochastic quantity on a continuous (two-

dimensional) value set. Also histograms are drawn, both according
to arguments x and y and according to both together. The blue
^ curve is a possible probability density distribution function.
is zero, we say that it is impossible for a value x in this interval to occur

as a realisation of the stochastic quantity. If the integral is 1, we say that
[︁ ]︁
the occurrence of some x ∈ x1 , x2 is certain. In all other cases, the value
of the integral gives the probability that the realisation of the stochastic
quantity would hit inside this interval. The value of the probability is
{︁ }︁
E x
x (second series)
x (first series)
p ( x)
F IGURE 2.4. The probability density distribution p( x) of a real-valued stochas-

tic quantity as the limit of histograms. The expectancy E { x} and
the averages of the two measurement series, the second containing
^ twice the number of measurements of the first, are also drawn.
í Õ ! ¤. û
Expectancy
{︁ }︁
µ=E x
Probability
density p( x) −σ +σ Mean error
Value space x
σ σ
σ σ
σ σ
More precise Less precise
^ F IGURE 2.5. Important properties of the normal distribution.
9 thus always between zero and one.9
Always ˆ +∞
p( x) dx = 1,
−∞
because the joint probability that the measurement value will be any real
number, is 1, in other words, 100 %: it is certain.
For a continuous stochastic quantity one can also compute an ex-
pectancy. The expectancy integral looks similar to the discrete coun-
terpart 2.1: ˆ +∞
E { x} = x · p( x) dx.
−∞
The expectancy is the x co-ordinate of the centre of mass of the area under
the density distribution curve p( x).
^ 2.4.2 The Gaussian bell curve

In geodesy we use almost always the normal or Gaussian distribution
(“bell curve”).
In figure 2.5 we see the expectancy µ = E { x}. The expectancy may be
understood as the “true value” of the observed quantity x, which is not
itself measurable, but around which the measurement values group under
the influence of their random errors. In the figure we also see the mean
10 error10 σ, which describes the tendency of an individual measurement
9 Mathematically, probability is a measure, like, for example, surface area or volume.

10 . . . or standard deviation, or standard uncertainty.
í Õ ! ¤. û
35
Area total 32 % Total 4.5 % Total 0.27 %
σ σ 2σ 2σ 3σ 3σ
^ F IGURE 2.6. Probability values for the normal distribution.
to differ from the expectancy. In the case of the normal distribution,

deviations equal in magnitude to the mean error µ ± σ are located in the
curve’s inflection points. taivutuspiste
A concept by the name of variance is also used, the square of the mean
error:
Var{ x} = σ2 ,
the formal definition of which is11 11
{︂(︁ )︁2 }︂
def
Var{ x} = E x − E { x} . (2.2)
Here E {·} is the expectancy operator as defined above.

Like expectancy, variance is also the kind of “true value” that theoreti-
cally exists, but that we never actually really measure. When we have
available a finite number of observations or sample, we may calculate the otos
sample mean and sample variance, that, for the number of observations
increasing, move ever closer and closer to the expectancy and variance.
This phenomenon is also called the law of large numbers.
The mathematical expression p( x) for the normal distribution — which
we do not present here — gives through integration the following proba-
bility values:
◦ the probability that x deviates from E { x} more than an amount σ
(in either direction): 32 %
◦ the probability that x deviates from E { x} more than an amount 2σ:
4.5 %
◦ the probability that x deviates from E { x} more than an amount 3σ:
0.27 %.
11 So, the expectancy of the square of the deviation of a stochastic quantity from its
own expectancy. This is a kind of cost function: if the cost of the “error” x − E { x} is
proportional to its square, then Var{ x} is the expected cost of the error.
í Õ ! ¤. û
In practice, a normally distributed stochastic quantity really never

deviates from its expectancy by over three times its mean error.
This rule of thumb, the three-sigma rule, is made use of in statistical

testing.
There exist a large number of different statistical distributions in
addition to the normal distribution. These are contained in software
packages used for statistical computation, such as MATLAB and R, and in
12 the numerical libraries accompanying general programming languages.12
^ 2.4.3 Covariance and correlation

When there are two stochastic quantities x and y, one can, besides their
separate behaviours, also study how they behave together.
Let us give the name covariance to the expression
{︃ }︃
{︁ }︁ def (︁ )︁(︂ {︁ }︁)︂
Cov x, y = E x − E { x} y−E y . (2.3)
This definition is analogous to the one for variance, equation 2.2, but
describes the “behaving in the same way” property of the quantities x
and y; the similarity in their random behaviours.
Often it makes sense to scale this covariance relative to the variances
of the quantities x and y in the following way:
{︁ }︁
{︁ }︁ def Cov x, y
Corr x, y = √︂ {︁ }︁ .
Var{ x} Var y
This is how correlation is defined between the quantities x and y. Corre-

[︁ ]︁ [︁ ]︁
lation always lies in the interval −1, 1 , or −100 %, 100 % .
The correlation (and covariance) between statistically independent
quantities x and y equals 0. The non-vanishing of correlation may be a
sign of a cause-and-effect relationship — but one cannot say that x is the
cause and y the effect! y may be the cause of x, or x and y may have a
common cause z.
If the correlation is 1, or 100 %, we speak of perfect correlation. In this
case there exists an exact functional relationship between x and y. If the
realisation value of one quantity is given, the corresponding realisation
value of the other can be calculated precisely.
12 A good overview is Easton and McColl.
í Õ ! ¤. û
37
y
(a) (b) (c)

Non-existent Weak Strong
(d) (e) (f)

100 % Anticorrelation −100 %
^ F IGURE 2.7. Some examples of correlation.
If the correlation is negative, we speak of anticorrelation. If the cor-

relation is −1, or −100 %, there exists a perfect correlation between the
quantities x and − y, and we speak of perfect anticorrelation.
Figure 2.7 shows some examples of correlation.
^ 2.4.4 Propagation law of errors

An important property of mean errors is the propagation law of errors. virheiden
kasautumislaki
If a stochastic quantity z is a linear combination of two other stochastic
quantities x and y,
z = ax + b y,
we may also write

{︁ }︁
E { z} = aE { x} + bE y ,
and thus (︂
(︁ )︁ (︁ )︁ {︁ }︁)︂
z − E { z } = a x − E { x} + b y − E y .
This is probably intuitively clear. Now according to the definition of

variance, equation 2.2, it follows that
{︂(︁ )︁2 }︂
Var{ z} = E z − E { z} =
í Õ ! ¤. û
{︃(︂ }︃
2
{︂(︁ )︁2 }︂ 2
{︁ }︁)︂2
=a E x − E { x} +b E y−E y +
{︃ }︃
(︁ )︁(︂ {︁ }︁)︂
+ 2ab E x − E { x} y−E y .
The definitions 2.2 and 2.3 now yield
Var{ z} = a2 Var{ x} + b2 Var y + 2ab Cov x, y .

{︁ }︁ {︁ }︁
(2.4)
varianssien This equation is called the law of propagation of variances.

kasautumislaki
In the case where Cov{ x, y} = 0 — meaning that the quantities x and y
are uncorrelated — we obtain
Var{ z} = a2 Var{ x} + b2 Var y ,

{︁ }︁
or, in a more compact notation,
σz2 = a2 σx2 + b2 σy2 .
This can be generalised to more than two quantities. For example, if
w = ax + b y + cz,
we obtain
2
σw = a2 σx2 + b2 σy2 + c2 σ2z.
A special case is the situation in which z is the sum or difference of

x and y, so a = ±1, b = ±1, and x and y are statistically independent of
each other. Then we get a simple, Pythagoras-like, much-used equation:
σz2 = σx2 + σy2 .
Often, this special case is referred to as the law of error propagation.

Example We have measured the distance between points A and B in
three parts,
s AB = s1 + s2 + s3
(so a = b = c = 1), in which
s1 = 10 m ± 0.05 m, s2 = 20 m ± 0.2 m, s3 = 15 m ± 0.1 m.
We observe that
σ1 = 0.05 m, σ2 = 0.2 m, σ3 = 0.1 m,
í Õ ! ¤. û
39
1.0
Standard ellipse
0.8
0.6
0.4
0.2
0.0
10
5
−10 0
−5
0 −5
5 −10
10
Cross-section
^ F IGURE 2.8. A two-dimensional probability density distribution.
and
2 2 2 2
σAB = σ12 + σ22 + σ32 = (0.05 m) + (0.2 m) + (0.1 m) =
= 0.0525 m2 ,
√︂
2
so σAB = σAB ≈ 0.23 m, and the result is
s AB = 45 m ± 0.23 m.
(The three measurements are assumed to be statistically indepen-

dent of each other.)
^ 2.4.5 Multi-dimensional distributions

Often, like in the above football example (figure 2.3), we speak of stochas-
tic quantities consisting of several components. An example is given by
the co-ordinates ( x, y) of a point in a plane. In the same way as described
above, we may draw two-dimensional histograms and speak about the
probability density distribution p( x, y).
Figure 2.8 shows a two-dimensional normal density distribution, and
its standard ellipse or error ellipse. It is the two-dimensional counterpart
to the mean error. A vertical cross-section again always produces a one-
dimensional Gaussian bell curve, the mean-error points µ ± σ of which
are precisely the intersections of the cutting plane and the error ellipse.
í Õ ! ¤. û
^ T ABLEAU 2.5. On correlation.
Correlation is a curious thing. It is always good to remember that statistical

dependence does not necessarily mean that there is a direct cause-and-effect
relationship, only that the observations have something in common. The con-
nection may be indirect, like in the famous case in which the ice-cream sales
over summer correlated with drowning statistics. The connection may also be
uninteresting, like the correlation between two phenomena both trending in
time, for example, between the Ethiopian gross domestic product and Danish
wind-power output.
Whereas the mean-error points of the one-dimensional distribution are

at x = µ ± σ, the equation for the error ellipse is more complicated. The
(︁ )︁
expectancy (“true value”) of the stochastic co-ordinate pair x, y is itself
a co-ordinate pair, (µ x , µ y ), the centre of the error ellipse.
In the general case where the axes of the error ellipse are not parallel
to the co-ordinate axes (and of different lengths!), the pair of stochastic
quantities x, y can no longer be treated as independent variables: they
correlate with each other. This means that the most probable value of
x will depend on the actual value of y, and vice versa. In adjustment
calculus, this statistical dependence must be taken into account in order
to arrive at the best possible — optimal — estimates both of µ x and of µ y .
In three-dimensional space there exist probability density distributions
of three arguments p( x, y, z), the visual representations of which are
triaxial (and three-dimensional) error ellipsoids.
^ 2.4.6 The variance matrix

The variances of multi-dimensional stochastic quantities themselves also
13 become multi-dimensional: they become variance matrices. Let13
[︄ ]︄
def x
x= ,
y
then the variance (or variance-covariance) matrix is

[︄ {︁ }︁ ]︄ [︄ 2 ]︄
def Var{ x} Cov x, y σx σx y
Var{x} = = = Σxx ,
σy2
{︁ }︁ {︁ }︁
Cov x, y Var y σx y
[︄ ]︄
13 . . . µx
and of course also the expectancy is a vector: E {x} = !
µy
í Õ ! ¤. û
41
in which again apply the definitions
{︂(︁ )︁2 }︂
def
Var{ x} = E x − E { x} , {︃ }︃
{︁ }︁ def (︁ )︁(︂ {︁ }︁)︂
{︁ }︁ def
{︃(︂
{︁ }︁)︂2
}︃ Cov x, y = E x − E { x} y − E y ,
Var y = E y−E y ,
and in which we see the often-used notation14 14
Var{ x} = σx2 , Var y = σy2 ,

{︁ }︁ {︁ }︁
Cov x, y = σx y .
In appendix A a short explanation is given of the basics of matrix

calculus.
In fact, the error ellipse is a graphical representation of this 2 × 2 sized
variance matrix. If σx y = 0, we say that x and y do not correlate with
each other, or, carelessly,15 that they are statistically independent of each 15
other.
In addition, the propagation law of variances, equation 2.4,
Var{ z} = a2 Var{ x} + b2 Var y + 2ab Cov x, y ,

{︁ }︁ {︁ }︁
may be written into a more general form: if we form the row vector
[︂ ]︂
def
a= a b ,
and the corresponding column vector16 16

[︄ ]︄
[︂ ]︂T a
aT = a b = ,
b
we obtain
[︄ {︁ }︁ ]︄ [︄ ]︄
[︂ ]︂ Var{ x} Cov x, y a
Var{ z} = a b · {︁ }︁ {︁ }︁ · = a · Var{x} · aT .
Cov x, y Var y b
Understanding this requires knowing how to multiply matrices (row

× column), see appendix A. Writing it out gives us back the original
equation 2.4.
This is presented in the literature symbolically in somewhat varying
notations:
σz2 = a Var{x}aT = a Σxx aT .
We will discuss the propagation of variances in linear models of many varianssien

variables more generally in section 14.7. kasautuminen
14 Sometimes we see in older texts Var{ x} = mx2 , Var y = m2y , Cov x, y = mx y , as back
{︁ }︁ {︁ }︁
then, typewriting Greek letters was difficult.

15 . . . because there may be a more complicated statistical dependence that does not
í Õ ! ¤. û
^ 2.5 Geodetic observables
^ 2.5.1 Angles, directions

Traditionally, the large national or continental geodetic networks have
kolmiomittaus- always been triangulation networks. Directions are measured between
verkko the network stations, and the measurements in the network are adjusted:
from the measured values, small contradictions caused by the uncertainty

of measurement are computationally removed, leading to the solution of
the triangulation. Back in time, distance measurement was difficult over
larger distances, as the only working technique was measurement using
mechanical means, such as measuring tapes and rods.
In triangulation, the procedure is that in a network, all feasible di-
rections are measured, plus one single distance. From these measured
values, all the other distances and point locations are computed. This is
astemittaus how Snellius carried out his famous grade measurement between Bergen
op Zoom and Alkmaar, cities 130 km apart, by measuring directly only
perusviiva the baseline 87 roeden long, built by him in a meadow! The Lapland
grade measurement was similar: the baseline was built during winter
1736 on the ice of the Torne river, and all other measurements were
directions between the points of the triangulation network.
kolmiomittaus The basic idea of triangulation can be explained using the plane table.17
17
At point A on the terrain, a plane table is erected, a sheet of transparent
mittapöytä paper placed on it, and all directions to the terrain points B, C , D and
18 E 18 are drawn. Move to point B, and draw a similar set of directions to
targets A , C , D and E . In the office, the sheets of paper are placed on top
of each other. The result is a miniature image of the true geometry of the
terrain — at a scale of m = ab : AB. So, if the distance AB in the terrain
is 3 km, and on the superposed papers ab = 30 cm, then the scale of the
“map” thus obtained is 1 : 10 000.
Nowadays, directions are measured using a theodolite, a precise instru-
ment for measuring horizontal and vertical angles, and creating the map
takes place computationally.
show as correlation.
16 We often use the transpose notation to make it easier to write a column vector in
running text.
17 German Messtisch, French planchette, Dutch meettafel, Swedish lantmätartavla,
Danish, Norwegian målebord.
18 The drawing is made easier by using an alidade, German kippregel.
í Õ ! ¤. û
Geodetic observables 2.5
43
C
Superimposed: E
B
d c
a
e b
F IGURE 2.9. Triangulation by means of a plane table and alidade. Centre image
^ Wikimedia Commons, Plane table with alidade.
In photogrammetry there is a three-dimensional analogue to the plane

table method, stereo model restitution, which however nowadays is also
invariably realised fully digitally. Figure 2.10 shows how from two aerial
images a stereo model is formed using a viewing device. Similarly, in a
stereo restitution instrument, a model is formed, inside which a floating avaruus-
mark may be moved around three-dimensionally. The co-ordinates of the mittamerkki
mark are continuously output to a computer hooked up to the instrument,
and a map can be drafted immediately.
Aerotriangulation is also based on forming stereo models. ilmakolmiointi
A fairly modern geodetic technique based on the idea of the plane
table but in three dimensions, is Yrjö Väisälä’s19 stellar triangulation. 19
This technique, in which the stellar background is used for measuring tähti-
directions, is discussed in section 17.3. kolmiomittaus
19 Yrjö Väisälä (1891–1971), “the Wizard of Tuorla”, was a Finnish astronomer, physicist,
geodesist, metrologist, builder of telescopes, finder of comets and asteroids, recreational
sailor and Esperanto practitioner.
í Õ ! ¤. û
Image 1 Image 2
Aerial photography
Terrain
Stereo viewer
Stereo model
Stereo image pair
^ F IGURE 2.10. Forming a stereo model in photogrammetry.
^ 2.5.2 Distances, distance differences

In the past, distances could only be measured mechanically, using mea-
suring wires, tapes, or rods. The precision achieved in the measurements
was often impressive thanks to careful procedures.
In local surveying, steel measuring tapes continue to be used, as they
are inexpensive and easy to use and carry along. One should only remem-
ber to clean and grease them after use. The lengths are 20–60 m. The
precision measuring wires of old were made of invar, a steel alloy with a
very low thermal expansion coefficient.
Today, distance measurements are done electronically or electro-
optically. Only a straight line of sight between points is needed. The
devices may use microwaves (Tellurometer) or more often visible light
20 (Kern Mekometer), laser light (Geodimeter),20 or, in the case of most
etäisyysmittari modern range-finders (distance measurement devices) and total stations,

takymetri infrared light produced by a light-emitting diode (LED). The effect of
hohtodiodi atmospheric delays on signal propagation must always be carefully taken
20 Geodimeter™ is a Swedish invention and the name is a trademark of Trimble AB,

Danderyd, Sweden.
í Õ ! ¤. û
Geodetic observables 2.5
45
into account in processing the observations.
Terrestrial geodetic measurement instruments are able to measure
both horizontal and vertical angles as well as slant ranges. They are
called electronic tacheometers,21 and they thus combine the properties 21
of a theodolite and a range-finder into one integrated, fully electronic,
highly automated instrument.
Satellite geodesy also uses electronic distance measurement. The GNSS,
Global Navigation Satellite Systems — like the Global Positioning System
GPS — in broad use today are based on distance measurements made
by microwaves — more precisely, distance difference measurements.22 22
Satellite lasers on the other hand (like the instrument at the Metsähovi
research station) measure the travelling time from the observation station
to a satellite reflecting the light, and back.
The advantage of using electronics is, that the incredibly high precision
of frequency measurement is thus harnessed as distance measurement
precision. When the precision of frequency measurement can well be
1 : 1012 , it is understandable that with GPS, we measure intercontinental
distances with a relative precision of even 1 : 109 . Satellite techniques
are even more precise than terrestrial ones, because most of the signal
propagation takes place outside the denser parts of the atmosphere.
^ 2.5.3 Potential differences, levelling

Measurement of geopotential differences is traditionally done by levelling:
one measures the height difference between two points in metres, and
converts this difference into a potential difference using local gravity g.
If the height difference is ∆ H and the difference in geopotential is ∆C ,
we use the equation23 23
∆C = g ∆ H.
The measurements are repeated along a line, and the differences ∆C i

obtained are summed up. From the lines, a levelling network is built, in
which the potential differences calculated around closed loops must sum
to zero. The network is adjusted using this condition.
21 Tacheometer, Greek rapid measurer. Called more often total stations.

22 Hyperbolic positioning systems, like Decca, Transit/Doppler, and also GPS and the
other satellite positioning systems, are based on distance difference measurement.
23 This is an example of “work is force times path”. Work (per unit of mass) is ∆C , path
is ∆ H , and force (per unit of mass: according to Newton’s law F = ma, acceleration) is g.
í Õ ! ¤. û
In subsection 4.3.1 a few more exotic height determination alternatives

to traditional levelling are presented, many of which are based on the
direct physical comparison of geopotential values, for example using a
fluid surface.
^ 2.6 About co-ordinates

In geodesy we use co-ordinates in order to describe the figure and size of
the Earth and to determine the locations of points on or near the Earth’s
surface.
The co-ordinate reference systems used in geodesy are generally three-
dimensional, as the Earth is a three-dimensional object in a three-dimen-
sional space. As an example we may mention latitude, longitude, and
height (ϕ, λ, h), which let us present the location of a point in an intuitive
way.
“Two-dimensional” co-ordinate reference systems are really map pro-
jection co-ordinates, derived quantities not of direct interest to geodesy.
They belong more to the field of cartography, although they are used very
broadly in applied surveying. For example, on older Finnish topographic
maps one encounters KKJ co-ordinates, which are map-plane co-ordinates,
the kind of co-ordinates ( x, y) used on a map sheet, directly measurable
by ruler.
In addition to spatial co-ordinates, time, for describing processes of
change, and in physical geodesy, as a kind of co-ordinate, geopotential, the
potential of the gravity field of the Earth, are also used. See subsection
4.1.1.
^ 2.7 Why we use plane co-ordinates

Because the Earth is a three-dimensional object, geodesy is a three-
dimensional science. The Earth and points associated with her are located
in three-dimensional space, and the task of geodesy is to determine and
present these locations using three-dimensional co-ordinates, such as
( X , Y, Z ). Modern measurement systems, like the global positioning
system GPS, are able to directly measure three-dimensional co-ordinates.
Geodesy, however, is also an applied branch of science and engineer-
ing serving people. Humankind lives, forced by gravity, in a quasi-two-
dimensional “subspace”, in which there is freedom of movement almost
í Õ ! ¤. û
Co-ordinates of location in three dimensions 2.8
47
only in the horizontal direction, along the surface of the Earth. Further-
more, the important medium of communication called paper is uncondi-
tionally two-dimensional, and maps are commonly drawn on paper!
For this reason, plane co-ordinates are very generally used in geodesy
and surveying: rectangular, two-dimensional co-ordinates in the horizon-
tal plane.
There exist a multitude of practical plane co-ordinate reference systems
suitable for surveying work. The main differences between them are:
◦ The location of the origin and the orientation of the axes. The origin,
the starting point where x = y = 0, must be known. Generally the
axes are x to the north and y to the east, but not always, and not
necessarily accurately.
◦ The technology of determination, in other words, geocentricity:
modern plane co-ordinates are obtained from geocentric, three-
dimensional co-ordinates produced using GNSS.
Geodetic plane co-ordinates are, in fact, map projection co-ordinates: they
are calculated, using map projection formulas, from originally three-
dimensional co-ordinates through geodetic latitude and longitude (ϕ, λ).
In a very small area, like a building site, no proper map projection is
needed. In those kinds of areas, plane co-ordinates may be understood
as a special, rectangular case of topocentric (observation-site centric)
co-ordinates.
^ 2.8 Co-ordinates of location in three dimensions

Three-dimensional co-ordinates can be of different types, for example
rectangular or geodetic, also called geographical. The most common co-
ordinate reference system used is the three-dimensional rectangular or
Cartesian ( X , Y, Z ) system. Often it is also geocentric.
Geocentricity means literally, that the origin coincides (to a certain
accuracy) with the centre of mass of the Earth. In addition, the Z axis
of a geocentric system points along the Earth’s axis of rotation, to the
celestial North Pole.
In a geocentric co-ordinate reference system, the direction of the X
axis is in principle arbitrary and thus conventionally agreed. For a
geocentric system on the Earth’s surface, the meridian which contains
the direction of the plumb line at the Royal Observatory Greenwich serves luotiviiva
as an international standard24 for fixing the direction of the X axis. 24
í Õ ! ¤. û
F IGURE 2.11. The Greenwich meridian for tourists. Wikimedia Commons, The
^ prime meridian (Greenwich).
The X axis lies both in the plane of the Greenwich meridian and in the
equatorial plane, and is thus perpendicular to the Z axis. The Y axis in
turn is perpendicular to both the Z and X axes, so that all three axes are
mutually perpendicular. See figure 2.12.
Three-dimensional rectangular co-ordinates X , Y and Z may be gen-
erally applicable, but they are not particularly intuitive. For example,
the co-ordinates of the GPS antenna at the Metsähovi research station
are, expressed in the Finnish EUREF-FIN reference frame (see subsection
3.2.3):
X = 2 892 571.1204 m, Y = 1 311 843.2621 m, Z = 5 512 633.9521 m.
The numbers are interesting looking, but do not give a very enlightening
or easy to grasp answer to the question “where is Metsähovi?”. . .
As a first step to more practical co-ordinates, we construct geodetic
co-ordinates. First we construct mathematically a reference ellipsoid,
24 The treaty of Washington DC of 1884 made the Greenwich meridian the world’s zero
or reference meridian. At the same time, a “world time” or universal time was agreed:
Greenwich Mean Time, GMT. The basic idea was that civil times of countries would
differ from GMT by a whole number of hours, for Finland +2 h in winter (EET) and +3 h
in summer (EEST). Without this time-zone system, international traffic (by sea, air, or
telephone) would be cumbersome.
í Õ ! ¤. û
Co-ordinates of location in three dimensions 2.8
49
North Z
Pole Ellipsoidal
normal
h P
ϕ
Y
ϕ
λ
Greenwich
observatory
X
Greenwich
meridian
^ F IGURE 2.12. Rectangular and geodetic co-ordinates.
a suitably flattened ellipsoid of revolution the measures of which are pyörähdys-

reasonably close to those of the real Earth.25 ellipsoidi
25
A point is projected along the ellipsoidal normal onto the surface; the
distance of projection, h, is the ellipsoidal height, the direction angles
of the projection line or ellipsoidal normal26 are ϕ, the geodetic latitude 26
reckoned from the equator, and λ, the geodetic longitude, reckoned from
the Greenwich meridian. The triplet (ϕ, λ, h) — or often only the pair
(ϕ, λ) — is referred to as geodetic co-ordinates. The geodetic co-ordinates
of, for example, Metsähovi are
ϕ = 60◦ 13′ 2′′ .89046, λ = 24◦ 23′ 43′′ .13336, h = 94.568 m, (2.5)
. . . and this tells people already a whole lot more about where this point
is located!
There is a simple mathematical relationship between rectangular and
geodetic co-ordinates: each can be converted into the other without losing
precision. They are equivalent27 presentations of the location of a point: 27
25 For example, the GRS80 reference ellipsoid: equatorial radius 6 378 137.0 m, polar
radius 6 356 752.3141 m (some 21 km shorter), and flattening 1 : 298.257 222 101.
26 The ellipsoidal normal generally does not go through the centre of the ellipsoid! See
figure 2.12.
27 So, if ( X , Y , Z ) is given, then (ϕ, λ, h) can be calculated, and the reverse is also true.
í Õ ! ¤. û
^ F IGURE 2.13. Geodetic co-ordinates.
equation
( X , Y, Z ) ←−−−−→ (ϕ, λ, h)
equation−1
. . . and both are geocentric and three-dimensional. The differences

between them are:
◦ Rectangular co-ordinates are more convenient to handle in numeri-
cal work.
◦ Geodetic co-ordinates are more intuitive, closer to the human un-
derstanding.
In addition to geocentric co-ordinates, topocentric co-ordinates often are
encountered; co-ordinates reckoned relative to some local origin. Topocen-
tric co-ordinates are naturally produced by measurements in which the lo-
cation of the measuring instrument becomes the origin of the co-ordinate
system: one measures a direction — azimuth or horizontal direction
korkeuskulma angle, and elevation angle above the horizon — and a distance from the
instrument’s location to the points to be measured.
1. What is a measurement unit, a measured quantity, and its dimen-
sion? Describe and explain the differences.
2. “Three score years and ten”. How much is this in gigaseconds?
í Õ ! ¤. û
Exercise 2–1: Co-ordinates and a street address
51
3. Scour the Internet for examples of confusion around physical units
causing damage, loss or accidents.
4. The mean temperature of the Earth for 2016 was an estimated
14.8 ◦ C. How much is this in kelvin?
5. Explain random errors, gross errors, and systematic errors.
6. Consider a pack of 52 playing cards, with the numbers 2–10, a jack
counted as 11, a queen as 12, a king as 13, and an ace as 1. What is
the expectancy for a card drawn at random from the pack?
7. Look up the equation for the standard normal distribution’s prob-
ability density p( x). What is the function value p( x) for x = µ ± σ?
And for x = µ? And what is the ratio p(µ ± σ) p(µ) ?
/︁
The standard normal distribution is the normal distribution for the

parameter values µ = 0 and σ = 1.
8. Would you expect that the incidence of bush fires in Australia
and the thickness of snow cover in Lapland would be correlated?
Anticorrelated? Uncorrelated? Why?
9. Why do we use also two-dimensional (plane) co-ordinates in
geodesy?
10. How is geocentricity defined?
11. What is the advantage of using geodetic co-ordinates (ϕ, λ, h) over
rectangular co-ordinates ( X , Y, Z )?
^ Exercise 2–1: Co-ordinates and a street address

A riddle for seafarers:
1. To what street address do the geodetic co-ordinates in the accompa-
nying picture, figure 2.13, refer?
2. What is the general nature of the object at the address? Give a
picture!
3. Are the co-ordinates geocentric? On which reference ellipsoid?
4. Find the “Easter egg” in the index of this text.
í Õ ! ¤. û
^ Map projections, datums and
transformations
3
Le 24 février dernier, la Chambre des députés adoptait un
projet de loi, dû à l’initiative parlementaire et ayant pour
objet de fixer à nouveau l’heure légale de notre pays. Il était
ainsi formulé : l’heure légale, en France et en Algérie, est
l’heure, temps moyen, de Paris, retardée de 9 minutes 21
secondes.
How France legislated in 1898 civil time in France and

Algeria to be Greenwich Mean Time. Dastre (1898)
^ 3.1 Map projections

Although geodetic co-ordinates are more user friendly than rectangular
ones, even better co-ordinates, closer to the human user, can be produced
by using the method of map projection.
The practice of depicting the Earth’s surface on a two-dimensional
plane or map is an old one. Cartography is the science that has grown
around this art. If the area to be depicted is small, the method of depiction
is straightforward and error-free: local horizontal rectangular landscape
co-ordinates ( x, y) can be depicted, through a scale, onto the paper map
plane. We speak of a plan.1 In most countries, including Finland, the x 1
axis points north and the y axis east.
For a larger area, we use a map projection, a mathematical method
to map the location, or latitude and longitude (ϕ, λ), of a point to the
map plane. Thus we can draw a graphical representation of the Earth’s
surface onto paper, a map. Many digital applications, like CAD software,
1 Swedish plankarta (archaic; nowadays the term refers to zoning maps), Dutch platte-
grond.
– 53 –
54 3 M AP PROJECTIONS, DATUMS AND TRANSFORMATIONS
that do not actually even require the use of a paper map, are nevertheless
based on the intuitive use of the map plane.
In the simplest approach we use the co-ordinate pair (ϕ, λ) directly
as map co-ordinates x and y: x = S ϕ, y = S λ, S being the nominal scale.
This is a pathetic solution, because
◦ The co-ordinates ϕ and λ are in degrees, angular units, whereas
the map co-ordinates have to be in metric units.
◦ One degree of longitude λ expressed in kilometres diminishes to-
ward the poles. At the latitude of Helsinki, a degree of longitude
is only 55 km, when at the equator it is 111 km. A slightly better
approach is to use the co-ordinate pair (ϕ, λ cos ϕ).
karttaprojektio- Better solutions are offered by map projection science. Thus we may map
oppi the parameter pair (ϕ, λ) on the surface of the reference ellipsoid onto the
pair ( x, y) in the map plane in a sensible way. Unfortunately a method

that depicts everything exactly correctly does not exist. Something is
always distorted (Wikipedia, Theorema Egregium).
In map projections we always make approximations. We always lose
something. There are no projections that do not distort anything or
have the same scale throughout the map plane. A projection is chosen
according to the purpose of use, so that something that is important to
the user is preserved: the shapes of objects, their surface areas, some
distances, the compass direction.
kulmatarkka ◦ If angles and ratios of distances are preserved, we speak of a con-
formal or angle-preserving projection. In this case both the linear
scale — distances — and the scales of surface areas are distorted,
except at some special points of the map.
pinta-alatarkka ◦ If surface area is preserved, we speak of an equivalent or equal-area
projection. Here, angles and shapes are distorted, again with the
exception of special points or lines.
etäisyystarkka ◦ If distances are preserved, we speak of an equidistant projection.
A projection can be equidistant only along certain lines, never
everywhere.
◦ In navigation it is important that compass directions are preserved.
In that case, loxodromes or rhumb lines (Wikipedia, Rhumb line),
lines of constant compass direction, are straight. The normally
oriented Mercator projection has this property.
isoympyrä ◦ If great circles are depicted as straight lines, we have the gnomonic
í Õ ! ¤. û
The various co-ordinate solutions used in Finland 3.2
55
x
ϕ
F IGURE 3.1. Depicting the curved surface of the Earth to the map plane using
^ different projections. Something will always be distorted!
projection. It is used in aviation and in connection with meteor

observations — the track of an aircraft on the Earth’s surface or
the track of a meteor on the sky is mapped to a straight line, which
may be drawn with a ruler.
It is important to be aware that there is no “correct” projection! The
choice is dictated by the purpose of use and the distortions considered
acceptable. In a way they are all “wrong”. On the other hand, they are
also all useful.2 2
Philosophically one may be of the view, that map co-ordinates ( x, y) are

not actually geodetic quantities. Genuinely geodetic quantities are always
three-dimensional. Map projection co-ordinates are derived quantities.
A high-level introduction into the mathematics of map projections is
offered by Grafarend et al. (2014).
^ 3.2 The various co-ordinate solutions used in Finland

In Finland, like everywhere on Earth, several different co-ordinate refer-
ence frames3 are in use for historical reasons. This complicates the use of 3
geographic information and co-ordinate materials. It is a precondition for
the use of co-ordinate data, that we know in which co-ordinate reference
frame the material is. If necessary, one transforms from one frame to
2 “Essentially, all models are wrong, but some are useful” – George E. P. Box (1919–2013),
British statistician.
3 Here, we do not yet make any clear distinction between co-ordinate reference systems
and reference frames. Find more on this in section 3.8.
í Õ ! ¤. û
F IGURE 3.2. The systematic shift between the road network and the aerial pho-
tograph base could have something to do with the use of different
co-ordinate reference frames. Google Earth™. Google terms of use.
Map data © 2009 Google. Image © 2010 DigitalGlobe, © 2010 Tele
^ Atlas, © 2010 Europa Technologies.
another before use.

The most fundamental distinction between different co-ordinate ref-
erence frames is, whether we have a “traditional” one, created before
the satellite era using traditional geodetic measurement methods, or a
“modern”, geocentric, one created using satellite positioning technology.
In the following, we explain more about co-oordinate reference frame
alternatives and the transformations between them.
paikkatieto However, even if it is clear in principle what co-ordinate frame geospa-
tial data is in, small differences in how the data was processed and
connected to known locations on the ground may lead to visible discrep-
ancies in the resulting co-ordinates. See figure 3.2 for an example.
^ 3.2.1 National Map Grid Co-ordinate System (KKJ)

In Finland, the old, in its time official, and now largely obsolete, co-
ordinate reference frame is KKJ, the National Map Grid Co-ordinate
System (Parm, 1988). The system was created based on the results of
kolmiomittaus the national triangulation effort. It is also a good example of the use of a
map projection for mapping a national territory onto a two-dimensional
í Õ ! ¤. û
The various co-ordinate solutions used in Finland 3.2
57
map plane. More about this map projection in subsection 3.3.1.
^ 3.2.2 Helsinki system (VVJ)

The Helsinki system, or VVJ, was the predecessor of KKJ. The national
triangulation was a huge project (Puupponen and Järvinen, 2008), lasting
over half a century. The users of accurate co-ordinates, however, could not
wait. Thus, VVJ was created “on the fly” in support of practical mapping
work. The co-ordinates were calculated stage after stage, as the Finnish
triangulation progressed from Southern Finland towards the north. Not
until the 1970s, when the national triangulation was almost completed
and its complete adjustment carried out, was the KKJ system formed as a tasoitus
nationally unified system.
The co-ordinate differences between the Helsinki system and KKJ are
up to a few metres (Häkli et al., 2009, page 18).
Ancient as the Helsinki system may be, it was still in use in many
municipalities in the 2010s: for example Tuusula did not move away from
it — directly to EUREF-FIN — until 2013. . .
^ 3.2.3 The EUREF-FIN reference frame

With the spread of satellite positioning, as early as in the 1990s it was
possible to determine three-dimensional co-ordinates for everywhere on
Earth, in a system that was geocentric on the several-centimetre level:
the origin of the three-dimensional co-ordinates ( X , Y, Z ) coincides to
that level of precision with the centre of mass of the Earth. In addition,
the Z axis of a geocentric co-ordinate system points in the direction of the
Earth’s axis of rotation, whereas the X and Y axis directions lie within
the equatorial plane.
Geocentric systems are global. In Europe, the International Association
of Geodesy IAG4 created the geocentric reference system ETRS89, the 4
European Terrestrial Reference System 1989. The system has been
realised by measurements many times: the first Europe-wide realisation,
or co-ordinate reference frame, was EUREF89, which was based on a
space geodetic measurement campaign (Overgaauw et al., 1994). Later,
Finland created its own national realisation under the name EUREF-FIN,5 5
4 This was done by its European reference-frame subcommission EUREF. The name
EUREF was proposed by the Danish geodesist Knud Poder. Knud Poder (1925–2019)
was an eminent Danish land surveyor and mathematical geodesist and a pioneer in
computational geodesy.
í Õ ! ¤. û
Map projection
Greenwich or
prime meridian Equator
λ0 Central
meridian,
longitude λ0
F IGURE 3.3. Mapping the curved Earth’s surface as a narrow zone onto a plane.
This is the principle of both the Gauss–Krüger and UTM projections.
Both projections are used in Finland. The distortions remain
^ acceptably small only in a relatively narrow area.
which has in recent decades been taken into use by the national mapping
agencies and most other actors.
In subsection 3.3.3 we explain more about the map projections chosen
for use with the EUREF-FIN reference frame.
^ 3.3 Map projections used in Finland

In Finland we encounter currently two different plane or projected co-
ordinate reference frames: the old KKJ, and the new system based on
EUREF-FIN. In the old KKJ, the Gauss–Krüger projection was used, in the
new system is used, in addition to Gauss–Krüger, also the UTM (Universal
Transverse Mercator) map projection. See figure 3.3.
^ 3.3.1 The map projection system of the KKJ

In order to map the Earth’s surface to the map plane without large
projektiokaista distortions, Finland was divided into six KKJ projection zones, every one
of which has its own co-ordinate frame in the projection plane. The zones
5 The often-used name WGS84 refers to the system — with actually half a dozen reali-
sations — maintained by the US defence authorities. The system’s latest realisations
are consistent with ITRF frames on the centimetre level. Consistency with the various
ETRS89 realisations is poorer than that, although better than a metre. See Malys et al.
(2016). Often the name WGS84 is used, erroneously, as shorthand for all of these.
í Õ ! ¤. û
Map projections used in Finland 3.3
59
1 2 3 4 x
◦
70 N 7 769 187 m
65◦ N 7 211 519 m
60◦ N 6 654 228 m
y
21◦ E 24◦ E 27◦ E 30◦ E
F IGURE 3.4. The zone division of the Finnish KKJ system’s Gauss–Krüger pro-
jection. The zones 0 (central meridian 18◦ ) and 5 (33◦ ) have been
^ left out.
were numbered from zero to five: the central meridians of the zones are
at longitudes 18◦ , 21◦ , 24◦ , 27◦ , 30◦ and 33◦ east.
The co-ordinates within a projection zone are x (northing) and y (east-
ing). The projection used is conformal and goes by the name of Gauss–
Krüger, one type of transversal Mercator projection. The reference ellip- poikittainen
soid used in the projection calculations is the Hayford or International
Ellipsoid of 1924.
In the several zones, the same co-ordinates appear multiple times.
Therefore, to obtain unique values, the zone number is prepended to the
y co-ordinate as its first decimal (except when it was 0). See figures 3.4,
3.5. This multi-zone system is referred to as the basic co-ordinate system perus-
of the KKJ. koordinaatisto
The x co-ordinate origin is on the equator, and the values grow along
the central meridian. Because of this, the x co-ordinates in Finland are
in the range 6 600 000–7 800 000 m.
í Õ ! ¤. û
x 7 800 000
northing
420 000
580 000
meridian
Central
6 600 000
O1 y
easting
500 000
False easting
F IGURE 3.5. The geometry of one zone of the Finnish KKJ system (stretched in
^ the east-west direction).
The y co-ordinate describes distance from the central meridian. In order

to avoid negative y co-ordinates, 500 km is added to them, so that a point
on the central meridian has a y co-ordinate of 500 000 m (false easting).
The y co-ordinates could theoretically be in the interval 0–1 000 000 m; in
practice, however, the range of values is 420 000–580 000 m, due to the
narrowness of the zones at Finnish latitudes.
For example, according to figure 3.5, the bottom-left corner co-ordinates
of the area of the first zone would be
x = 6 600 000.000 m, y = 1 420 000.000 m.
Similarly the bottom-left corner of the third zone:
x = 6 600 000.000 m, y = 3 420 000.000 m.
On the central meridian of the projection, the distortion is zero: the

nominal scale of the map applies exactly.
In the case where KKJ co-ordinates were used within a small area,
the leftmost digits were often left off, as they are all the same. Thus
í Õ ! ¤. û
61
one obtained truncated co-ordinates, which may still be encountered in katkaistut
municipal calculation documents. koordinaatit
The map projection system employed by the Helsinki system (VVJ) was
similar to that of KKJ, except that the notation was y = 21◦ 420 000, so
the y co-ordinate was prefixed with the longitude of the central meridian
itself.
^ 3.3.2 The KKJ’s Uniform Co-ordinate System

In addition to KKJ, for small-scale maps6 only, the Uniform Co-ordinate 6
System (Finnish acronym: YKJ) of KKJ was taken into use, which mapped yhtenäis-
the whole territory of Finland using one and the same Gauss–Krüger koordinaatisto
projection with a central meridian of 27◦ . As there is only one zone,
nothing is prepended to the y co-ordinate.
^ 3.3.3 The map projections of EUREF-FIN

In connection with the new reference frame we use EUREF-FIN map
projection co-ordinates (JUHTA, 2016b), which thus are plane co-ordi-
nates ( x, y). The new map projections, which use the geocentric GRS80
reference ellipsoid, are, depending on the map scale, either the familiar
Gauss–Krüger, or alternatively UTM, Universal Transverse Mercator, a
projection type that is also conformal and is in broad international use.
◦ For small-scale maps which depict all of or a large part of Finland,
one chooses the new ETRS-TM35FIN7 projection co-ordinate frame, 7
which is based on the three-dimensional EUREF-FIN co-ordinate
frame and the UTM projection for central meridian 27◦ E. This
replaces the old KKJ Uniform Co-ordinate System. The projection
also forms the basis for the whole Finnish map-sheet division.
◦ ETRS-TM35FIN is also used for topographic maps.
◦ For large-scale maps, intended for local use, one uses the Gauss–
Krüger projection like before, but with a zone width of 1◦ . These
6A small-scale map is a map the scale number M of which is large, if the scale is 1 : M .
A scale of 1 : 1 000 000 is small: even large objects look small on the map, but the area
mapped is large. A scale of 1 : 2000 is large: even small details are well discernible, but
the area covered by the map is small.
7 The name is a tongue and memory breaker. Alternative names circulating in the user
community are “Finnish UTM” and “Finnish Map Grid”.
í Õ ! ¤. û
projections are named ETRS-GK n, in which n is the degree num-

ber of the central meridian, for example 23◦ E, in which case the
name of the projection is ETRS-GK23. For practical reasons, however,
this rule is bent; for example, in order to get a whole municipality
mapped onto a single zone. Due to the choice of projection type,
these maps are useful for applications where no larger distortions
kaavoitus of the scale can be accepted, like zoning and infrastructure construc-
tion, where map co-ordinates are moved straight to construction
projects as terrain co-ordinates.
The UTM system is also much used internationally. UTM differs from
Gauss–Krüger in two ways:
◦ The scale on the central meridian is 0.9996 rather than 1.0. This
means that on the central meridian the map depicts details 400 ppm
(parts per million) smaller than they ought to be based on the map’s
nominal scale. That amounts to 40 cm for every kilometre.
◦ The width of the projection zones is 6◦ , not 3◦ or 1◦ . This means
that the scale distortion, which on the central meridian amounts
to −400 ppm, turns positive going to the edges of the zone, to about
+1000 ppm, calculated on the equator. At Finnish latitudes the
distortion remains quite a bit smaller than that.
These two choices aim to keep the scale distortion over the whole area
of the zone within certain bounds, in spite of the zone’s large width.
Both Gauss–Krüger and UTM are conformal projections: angles and
length ratios are preserved locally, so circles map to circles and squares
8 to squares — but only if they are small.8
^ 3.3.4 The triangulated affine transformation of the Finnish National

Land Survey
The Finnish National Land Survey provides on its web-site a service
that allows one to transform point co-ordinate sets between the older
KKJ (actually, the Uniform Co-ordinate System YKJ) and the new ETRS-
TM35FIN. The service is based on a division of the Finnish territory
9 into triangles using a Delaunay9 triangulation, and an affine or bilinear
8 If the square is too large, the nonlinearity of the map projection will cause it to be
mapped to a curved quadrangle. A large circle maps to a circle only for the stereographic
projection, otherwise it too will be deformed.
í Õ ! ¤. û
63
1500
Gauss–Krüger projection
UTM projection
1000
Scale distortion (ppm)
500
-500
0 50 100 150 200 250 300 350
Distance from central meridian (km)
^ F IGURE 3.6. Scale distortion of Gauss–Krüger and UTM projections.
Delaunay’n
transformation within every triangle. The transformation is linear within kolmiointi
every triangle and continuous across triangle boundaries, JUHTA (2016b).
See figure 3.7.
The method is simple: for the triangle △ ABC we define first for the
vertices A , B, and C the co-ordinate differences ∆ x and ∆ y between the
co-ordinate frames:
def
∆ x i = x i,YKJ − x i, ETRS−TM35FIN
}︄
def
i = A, B, C.
∆ yi = yi,YKJ − yi, ETRS−TM35FIN
After this, we define barycentric co-ordinates for an arbitrary point P ,

co-ordinates ( x, y):
def ω
A def ω
B def ω
C
pA = ω , pB = ω , pC = ω ,
with (the vertices A , B, and C of the triangle are assumed to be in
9 Boris Nikolayevich Delaunay (1890–1980) was a Russian mathematician and moun-

taineer. His grandson Vadim Nikolayevich Delaunay (1947–1983) was a poet who
participated in the 1968 Red Square demonstration against the occupation of Czechoslo-
vakia. Apparently not related to the French celestial mechanicist Charles-Eugène
Delaunay (1816–1872).
í Õ ! ¤. û
16 ◦
20 ◦ 24◦ 28◦ 32◦
72 ◦ 72
◦
70 ◦ ◦
70
68 ◦ ◦
68
66 ◦ ◦
66
64 ◦ ◦
64
62 ◦ ◦
62
60 ◦ ◦
60
58 ◦ ◦
16 ◦ 58
20 ◦ 32◦
24◦ 28◦
F IGURE 3.7. The triangulated affine transformation of the Finnish National

^ Land Survey. Triangle network used.
clockwise order):
⎡ ⎤ ⎡ ⎤
1 1 1 1 1 1
def 1 def 1
ωA = 2 det ⎣ xB xC x ⎦, ωB = 2 det ⎣ xC xA x ⎦,
⎢ ⎥ ⎢ ⎥
yB yC y yC yA y
⎡ ⎤ ⎡ ⎤
1 1 1 1 1 1
def 1 def 1
ωC = 2 det ⎣ xA xB x ⎦, ω= 2 det ⎣ xA xB xC ⎦ ,
⎢ ⎥ ⎢ ⎥
yA yB y yA yB yC
all determinants. Note that ω A , ωB , and ωC , like p A , pB , and pC , are

linear functions of the co-ordinates ( x, y) of the arbitrary point P . If we
í Õ ! ¤. û
More about plane co-ordinates 3.4
65
x x
P
Quadrant IV Quadrant I
s y
Northing
y Quadrant III Quadrant II

O (Origin) Easting
^ F IGURE 3.8. Geodetic plane co-ordinates and the quadrants of the plane I–IV.
write, for point P , the co-ordinate corrections as

def
∆ x P = p A ∆ x A + p B ∆ xB + p C ∆ xC ,
def
(3.1)
∆ yP = p A ∆ yA + pB ∆ yB + pC ∆ yC ,
we are carrying out a bilinear interpolation between the points A , B, and

C.
It is easy to show that
◦ Equations 3.1 reproduce the corrections for the vertices: ∆ xP → ∆ x A
if P → A , etc.
◦ The corrections are continuous across triangle edges: for example,
on edge AB, a straight line, the functions (∆ xP , ∆ yP ) are linear
along the edge, and reproduce (∆ x A , ∆ yA ) , (∆ xB , ∆ yB ) at both ends.
They must be the same for both triangles sharing the edge AB.
◦ As a result of this, the interpolation is continuous for the whole
triangulation cover. kolmiointi
^ 3.4 More about plane co-ordinates

The plane co-ordinates used in geodesy differ a little from the familiar
mathematical ( x, y) system. See figure 3.8. While in mathematics, the x
axis points to the right and the y axis to the upper edge of the paper, in
geodesy it is the habit that the x axis points north (“northing”) and the y
axis east (“easting”).
In addition to rectangular co-ordinates ( x, y), polar co-ordinates (α, s)
are used. Sometimes the symbols used are ( A, s). In geodesy, the azimuth
or horizontal direction angle α (or A ) turns clockwise from the north,10 10
10 In astronomy, sometimes, especially in older texts, the azimuth turns from the south
í Õ ! ¤. û
(City area)
x = 15 000 m
O y = 20 000 m y
^ F IGURE 3.9. A local co-ordinate frame.
that is through the east, unlike in mathematics. s is the distance from

the co-ordinate origin O .
The following trigonometric relations exist between rectangular and
polar co-ordinates:
y
y = s sin α =⇒ sin α = s ,
x
x = s cos α =⇒ cos α = s ,
sin α y y
tan α = cos α = x =⇒ α = arctan x + k · 200g , x ̸= 0.
In this equation, the small integer k is chosen such, that the result α lies
in the correct quadrant: arctan( y x ) is always in the interval − π , + π ,
/︁ (︁ )︁
2 2
that is in the quadrants I or IV. This is easiest to verify by making a
sketch.
The Pythagoras theorem yields the distance s:
√︁
s= x2 + y2 .
^ 3.4.1 Local co-ordinates

Local, often old, co-ordinate frames were long used in many Finnish
municipalities, and one comes across them in older documents.
to the west, that is also clockwise. In geodesy, too, practices vary: one must always
check.
The word “azimuth” originates from the Arabic as-sumût, “directions”.
í Õ ! ¤. û
More about plane co-ordinates 3.4
67
x x′
O′
y′
O y
^ F IGURE 3.10. Temporary co-ordinates.
The origin is generally located so, that everywhere in the municipality

there are only positive x and y values. Often the origin is just a computa-
tional point without realisation on the terrain, for example “truncated”
KKJ co-ordinates.
The connection to the national frame may be that one also knows the
co-ordinates of a church spire or some other landmark in the KKJ system.
Then, one can transform local and national co-ordinates into each other
by adding constant shifts, translations, to both co-ordinates x and y. siirto
In more precise work, one point is not enough: a sufficient number of

common points is needed, and their co-ordinates must be known in both
the local and the national system.
^ 3.4.2 Temporary co-ordinates

Sometimes it makes sense to use in measurements a temporary co-ordi-
nate frame that deviates from the general system. Even the directions of
the axes may deviate from the customary north and east directions.
A temporary or project-specific co-ordinate frame is only used during
measurement or, for example, during a construction project. Later on, the
co-ordinates may be transformed to a more permanent, local or national,
correctly oriented system.
The origin and axes orientation can be chosen in accordance with the
measurement at hand, for example along the walls of a building.
í Õ ! ¤. û
x
P2
α s ∆ x = x2 − x1
P1 ∆ y = y2 − y1
^ F IGURE 3.11. The forward geodetic problem in the plane.
^ 3.5 The geodetic forward and inverse problems

geodeettinen The “direct or forward geodetic problem” means the determination of the
päätehtävä co-ordinates of an unknown point, when the co-ordinates of a starting
point and both the azimuth (horizontal direction) and the distance from
the starting point are given.
In the general case, on an arbitrary curved surface, the forward geodetic
problem has no easy solution. On a sphere, however, a closed — though
not particularly simple — solution already exists. On the surface of a
reference ellipsoid the solution is obtained numerically. An on-line service
is offered by, among others, the US National Geodetic Survey on their
web-site NGS, Computation utilities.
In a plane co-ordinate system, two-dimensionally, the forward geodetic
problem is simpler, as we shall see next.
^ 3.5.1 The forward geodetic problem in the plane

Let there be in the plane two points P1 and P2 , figure 3.11. The plane co-
ordinates ( x1 , y1 ) of starting point P1 , as well as the azimuth α and length
−−→
s of vector P1 P2 , are given. The co-ordinates ( x2 , y2 ) of the unknown
point P2 are to be calculated.
The solution is obtained as follows:
∆y
sin α = s =⇒ ∆ y = s sin α,
∆x
cos α = s =⇒ ∆ x = s cos α,
and
x2 = x1 + ∆ x = x1 + s cos α,
y2 = y1 + ∆ y = y1 + s sin α.
í Õ ! ¤. û
The geodetic forward and inverse problems 3.5
69
Example Given point A with co-ordinates
x A = 6 800 000 m, yA = 400 000 m.
If the distance from point B is s = 2828.427 m, and the azimuth

(direction angle) α = 50g , solve the forward geodetic problem for
the points A and B.
Solution
∆ x = s cos α = 2828.427 m · cos (50g ) = 2000 m,

∆ y = s sin α = 2828.427 m · sin (50g ) = 2000 m,
xB = x A + ∆ x = 6 800 000 m + 2000 m = 6 802 000 m,
yB = yA + ∆ y = 400 000 m + 2000 m = 402 000 m.
^ 3.5.2 The inverse geodetic problem in the plane

The inverse geodetic problem means the determination of the azimuth geodeettinen
(horizontal direction) and distance between two given points. käänteistehtävä
Let there be given again two points in the plane, P1 and P2 (figure 3.11).
Let their rectangular co-ordinates be ( x1 , y1 ) and ( x2 , y2 ). To be calculated
are α and s.
Solution
√︂
2 2
√︁
s= ∆ x2 + ∆ y2 = ( x2 − x1 ) + ( y2 − y1 ) ,
∆ y y2 − y1
tan α = = . (3.2)
∆ x x2 − x1
We rather do not use sin α = ∆ y s or cos α = ∆ x s for determining
/︁ /︁
α: the sine formula becomes imprecise when α ≈ 100g + k · 200g ,

and the cosine formula when α ≈ 0g + k · 200g , k ∈ Z.
In those points, the functions sin α and cos α are stationary: a
large change in α causes only a small change in the function
value sin α = ∆ y s , and therefore, from the given values ∆ y and s,
/︁
precise as they may be, one can calculate α only imprecisely.

It is better to calculate α from equation 3.2:
⎧
⎨0 if ( x − x ) > 0,
y −y
(︂ )︂ 2 1
α = arctan x2 − x1 + k · 200g , k=
2 1 ⎩1 if ( x2 − x1 ) < 0,
⎧
⎨100g if x = x and y > y ,
1 2 2 1
α=
⎩300g if x1 = x2 and y2 < y1 .
í Õ ! ¤. û
∆y
∆x
α s
s + ∆x
s
α 2
/︁
^ F IGURE 3.12. The half-angle formula for the arc tangent.
It must be remembered here that the values of the arctan function

are, according to the definition, always in the interval − π2 , + π2 .
(︁ )︁
The correct α value may however well be outside this interval,

like when ( x2 − x1 ) < 0. This is why there is the conditional term
k · 200g .
Many programming languages offer the function atan2(x,y) with two

arguments, which also finds the correct quadrant automatically.
11 A more elegant solution is to use the half-angle formula:11
α ∆y ∆y
(︂ )︂
α = 2· = 2 arctan = 2 arctan .
2 ∆x + s
√︁
∆ x + ∆ x2 + ∆ y2
See figure 3.12.
Example Given point A : x A = 6 800 000 m, yA = 400 000 m, and point C ,

co-ordinates xC = 6 793 000 m, yC = 407 000 m, solve the inverse
geodetic problem for points A and C .
Solution
∆ x = xC − x A = 6 793 000 m − 6 800 000 m = −7000 m,

∆ y = yC − yA = 407 000 m − 400 000 m = +7000 m.
11 This formula too breaks down in the edge case ∆ y = 0 and ∆ x < 0, because then
α = 2 arctan 0 0 is undefined, when the correct solution is α = 200g .
/︁
í Õ ! ¤. û
The similarity co-ordinate transformation 3.6
71
1. The traditional method:
∆y
α AC = arctan + k · 200g =
∆x
= arctan(−1) + k · 200g = −50g + k · 200g .
The correct solution is apparently
α AC = −50g + 200g = 150g .
2. The half-angle formula:
∆y
α AC = 2 arctan √︁ =
∆x + ∆ x2 + ∆ y2
7000
= 2 arctan ⎷ =
−7000 + 7000 2
1
(︂ )︂
= 2 arctan ⎷ = 2 · 75g = 150g .
2−1
And
√︁ ⎷
s AC = ∆ x2 + ∆ y2 = 7000 m · 2 = 9899.495 m.
^ 3.6 The similarity co-ordinate transformation

The similarity12 or Helmert transformation is a transformation between yhden-
two rectangular co-ordinate frames, usually in the plane, or two-dimen- muotoisuus-
muunnos
sionally. It is encountered very often in practical measurement and 12
computation tasks, when bodies of co-ordinate material in two or more
co-ordinate frames need to be combined for joint use. In the general
case we find a sufficient number of common points — often benchmarks kiintopiste
— from the area, the co-ordinates of which are known in both systems,
and carry out an adjustment. The special case in which we have only
two common points, which is minimally sufficient for determining the
transformation, is simpler.
We know the co-ordinates of two benchmarks A and B in both coordi-
nate systems:
( x A , yA ) , ( xB , yB ) , ( u A , v A ) , ( u B , vB ) .
12 In Swedish: likformighetstransformation, in German: Ähnlichkeitstransformation, in

French: similitude.
í Õ ! ¤. û
F IGURE 3.13. Friedrich Robert Helmert (1841–1917) was a great German geode-
sist and a developer of adjustment calculus and its theory. Hum-
^ boldt University Berlin (2017).
In addition, we are given a set of points with co-ordinates only in the

(u, v) system:
(u1 , v1 ) , (u2 , v2 ) , . . . , (u i , v i ) , . . . , (u n , vn ) .
The problem now is to compute a similarity transformation for this point

field:
(u i , v i ) ↦→ ( x i , yi ) , i = 1, . . . , n.
x
u
vA
B
yAB
B
B
−θ
uA
αuv
x AB
s uv
αx y
y
=K
sx
y0 Ou A
v
Ox y x0 y
v
^ F IGURE 3.14. A similarity or Helmert co-ordinate transformation in the plane.
í Õ ! ¤. û
Determining the transformation parameters 3.7
73
See figure 3.14. The transformation is carried out in the following steps:
1. A shift, or translation, of the origin O uv ↦→ O x y , translation parame-
ters ( x0 , y0 ). The components of the origin translation vector t are
(− x0 , − y0 ) in the ( x, y) frame.
2. A rotation of the whole ( u, v) co-ordinate axes frame by an angle kierto
θ . The rotation angle of the axes is positive clockwise, but must
be subtracted from the directions between points when going from
(u, v) to ( x, y).
(In the figure, θ is negative.)
3. A scale transformation of the ( u, v) co-ordinates to the scale of the
( x, y) co-ordinates, by multiplying with the scale factor or scale ratio
K.
The Helmert transformation in the plane is also called a four-parameter
transformation. The parameters are x0 , y0 , θ , and K .
The general form of the Helmert transformation is (figure 3.14):
x = x0 + K cos θ · u + K sin θ · v,
(3.3)
y = y0 − K sin θ · u + K cos θ · v,
in matrix form — see appendix A:
[︄ ]︄ [︄ ]︄ [︄ ]︄[︄ ]︄
x x0 cos θ sin θ u
= +K . (3.4)
y y0 − sin θ cos θ v
^ 3.7 Determining the transformation parameters

Determining the transformation parameters unambiguously requires at
least four “observations” — for example, the total of four co-ordinates of
two points ( x A , yA ), ( xB , yB ). Then we obtain four equations:
x A = x0 + K cos θ · u A + K sin θ · v A ,
yA = y0 − K sin θ · u A + K cos θ · v A ,
xB = x0 + K cos θ · u B + K sin θ · vB ,
yB = y0 − K sin θ · u B + K cos θ · vB ,
again in matric form:

⎡ ⎤ ⎡ ⎤⎡ ⎤
xA 1 0 vA uA x0
⎢ y ⎥ ⎢ 0 1 −u
A vA
⎥⎢ y ⎥
⎢ A ⎥ ⎢ 0
⎥=⎢ ⎥.
⎥⎢ ⎥
⎦ ⎣ K sin θ
⎢ ⎥⎢
⎣ xB ⎦ ⎣ 1 0 vB u B ⎦
yB 0 1 −uB vB K cos θ
í Õ ! ¤. û
x θ
u
K −1
v
t
y
F IGURE 3.15. The stages of the Helmert transformation in the plane in terms
of changes of the co-ordinate axes: axis rotation angle θ , scaling
of axis units K −1 , translation vector of the origin t. Note that
the changes to the origin location, axes directions and axis unit
lengths are precisely the opposite of what happens to point co-
ordinates and numerical direction and distance values between
^ points.
If the old co-ordinates ( u A , v A ), ( u B , vB ) of the same points A and B are

also known, one may from these equations solve four transformation
parameters ( x0 , y0 ) and (K sin θ , K cos θ ) ↦→ (K, θ ) uniquely.
Difference transformation Subtraction yields
[︄ ]︄ [︄ ]︄
x AB xB − x A
= =
yAB yB − yA
[︄ ]︄[︄ ]︄
vB − v A uB − u A K sin θ
= =
− (uB − u A ) vB − v A K cos θ
[︄ ]︄[︄ ]︄
v AB u AB K sin θ
= ,
− u AB v AB K cos θ
with the logical definitions
def def
x AB = xB − x A , yAB = yB − yA ,
def def
u AB = u B − u A , v AB = vB − v A .
Arrange the terms cleverly anew:

[︄ ]︄ [︄ ]︄[︄ ]︄
x AB v AB u AB K sin θ
= =
yAB − u AB v AB K cos θ
í Õ ! ¤. û
Determining the transformation parameters
[︄ ]︄[︄ ]︄
3.7
75
cos θ sin θ u AB
=K . (3.5)
− sin θ cos θ v AB
This equation, the difference transformation, applies for arbitrary
point pairs.
The scale ratio or scale factor is obtained with Pythagoras’ theorem:
√︂
2 2
sxy xAB + yAB
K = s = √︂ .
uv 2 2
uAB + vAB
The rotation angle is13 (positive clockwise, i.e., turning from the x axis kiertokulma
13
to the y axis):
v y
θ = αuv − α x y = arctan uAB − arctan x AB . (3.6)
AB AB
The translation vector is computed starting from the Helmert differ- siirtovektori
ence transformation, equation 3.5. Let us look at the point pair A
and O uv , the origin of the old ( u, v) co-ordinate system. The points
have co-ordinates ( x A , yA ) and ( x0 , y0 ) in the new system, and
co-ordinates ( u A , v A ) and ( u 0 , v0 ) = (0, 0) in the original system.
Then
u AO = u A − u 0 = u A , v AO = v A − v0 = v A ,
and also
x AO = x A − x0 , yAO = yA − y0 .
The difference transformation now gives

[︄ ]︄ [︄ ]︄[︄ ]︄
x A − x0 cos θ sin θ uA
=K ,
yA − y0 − sin θ cos θ vA
so the translation vector is
[︄ ]︄ [︄ ]︄ [︄ ]︄[︄ ]︄
x0 xA cos θ sin θ uA
= −K ,
y0 yA − sin θ cos θ vA
just the translation parameters we are after.
For an arbitrary point ( x, y) the Helmert transformation equations
are now, after all transformation parameters have been solved for:
(3.3)
the same equations 3.3 already given above.
13 Forget for a moment the quadrant problem. In principle, a term k · 200g should be
added, with k a small integer.
í Õ ! ¤. û
Symbolic matric form The Helmert transformation in matric form is

[︄ ]︄ [︄ ]︄ [︄ ]︄[︄ ]︄
= +K . (3.4)
This can be written compactly:
x = x0 + K R u, (3.7)
in which the definitions of the vectors (column matrices) and ma-

trix are
[︄ ]︄ [︄ ]︄ [︄ ]︄ [︄ ]︄
def x def x0 def u def cos θ sin θ
x= , x0 = , u= , R= .
y y0 v − sin θ cos θ
Often one writes K = 1 + m, in which m is the scale distortion.

Usually this number is small and is expressed in the unit ppm
(parts per million).
The transformation equations 3.4 and 3.7 are called the similarity or
Helmert transformation in the plane.
Example
1. Given the co-ordinates of points A and B in the ( u, v) co-
ordinate system:
u A = 0 m, v A = 0 m,
u B = 1500 m, vB = 1500 m,
and in the ( x, y) co-ordinate system:
x A = 2000 m, yA = 3000 m,
xB = 3500.150 m, yB = 4500.150 m.
Assuming that the transformation between the ( u, v) and

( x, y) systems is a Helmert transformation:
[︄ ]︄ [︄ ]︄ [︄ ]︄[︄ ]︄
= +K , (3.4)
calculate its parameters K , θ , x0 , and y0 .

2. Given the co-ordinates of point C in the ( u, v) system:
u C = 1000 m, vC = 2000 m.
Calculate xC and yC .
í Õ ! ¤. û
Datums and datum transformations 3.8
77
Solution
1. We see immediately that
u AB = 1500 m, x AB = 1500.150 m,
v AB = 1500 m, yAB = 1500.150 m.
From this we infer visually, with the help of the difference

transformation 3.5, that K = 1.0001 and θ = 0.
After this, for point A :
x A = x0 + 1.0001 · u A =⇒ x0 = x A − 1.0001 · u A = 2000 m,

yA = y0 + 1.0001 · v A =⇒ y0 = yA − 1.0001 · v A = 3000 m.
2. Calculate
xC = x0 + 1.0001 · u C = 2000 m + 1000.1 m = 3000.1 m,

yC = y0 + 1.0001 · vC = 3000 m + 2000.2 m = 5000.2 m.
^ 3.8 Datums and datum transformations

Geodetic co-ordinates are not just mathematical quantities. Mea-
surements are carried out between points on the terrain, and their
co-ordinates are computed, based on given starting or datum points.
The choice of starting points is always to some extent arbitrary. Every
choice made creates what geodesists call a geodetic datum. In other
words, when geodetic measurements are made on a part of the Earth’s
surface using a certain set of measurement points, and co-ordinates are
conventionally assigned to starting points chosen from this set, we get,
in real life, a solution that represents only a realisation of a certain
reference system.14 14
A co-ordinate reference frame or datum is generally established locally.

When it meets another, similarly established (but based on different
starting points) frame, the co-ordinates of the same points are generally
not the same. For example, in the places where the Finnish and Swedish
14 In English, we call the formal definition a co-ordinate reference system, whereas we

call the realisation on the terrain a co-ordinate reference frame. For example, ETRS
= European Terrestrial Reference System and ETRF = European Terrestrial Reference
Frame. In Finnish, the corresponding terms are gaining traction, too: vertausjärjestelmä
against its realisation or vertauskehys.
í Õ ! ¤. û
precise levelling networks meet at the border in the Torne river valley,
we get two different height values for the same point, which are both
correct.
tasoverkko In the case of horizontal networks we also speak of datums, horizontal
datums: where networks meet at borders, the horizontal co-ordinates
(ϕ, λ) of the same point are generally not precisely the same in both
datums. The differences are, for classical triangulation networks, of the
order of seconds of arc.
For the purpose of transforming the co-ordinates of points in one datum
into co-ordinates of another datum, the literature offers datum transfor-
mation equations.
^ 3.8.1 Example: height network

Let us look at the example of height measurement, levelling. The Finnish
official height system until September 25, 2007, N60, was based on the
height of a certain point. This starting or datum point is a granite pillar
located in the garden of Helsinki astronomical observatory, and a certain
polished surface on that pillar. Here a role is played by historical accident,
by the fact of Helsinki being the Finnish capital. The height value of
the datum point was chosen such, that the heights of points were rather
precisely reckoned from mean sea level in Helsinki harbour at the start
of 1960. Scientifically the choice of Helsinki was arbitrary.
The new Finnish height system, N2000, uses as its starting or datum
pääkiintopiste point the fundamental benchmark PP2000 at the Metsähovi research
station — 40 km to the west of Helsinki — the height value of which was
chosen so that the heights are relative to the Amsterdam mean sea level
NAP, one of the oldest in the world. The choice of Amsterdam is also the
product of history, not science.
The national precise levelling has brought official heights to everywhere
in the country. It is clear that the precision of a calculated point height in
this system will depend on the point’s distance from Helsinki. A height in
Kevo, Northern Lapland, will be clearly more poorly known than a height
in Jyväskylä. And the height of Turku is somewhat imprecise, because
the levelling from Helsinki to Turku was not absolutely precise. On the
other hand, the heights measured for points in the Helsinki area are very
precise, because the datum point, be it Observatory Hill or Metsähovi, is
so nearby.
í Õ ! ¤. û
79
F IGURE 3.16. Fundamental benchmark PP2000 of the N2000 height datum

at the Metsähovi research station. Established by the Finnish
Geodetic Institute, today the Finnish Geospatial Research Insti-
^ tute (FGI) of the National Land Survey of Finland.
Imagine for a moment, that not Helsinki but Turku were the capital
of Finland,15 and that a benchmark in the wall of Turku Cathedral were 15
chosen as the datum point for the Finnish height system. In that case,
all heights of points close to Turku would be very precise, but the points
in the Helsinki area would be similarly imprecise as points in the Turku
area are in the present system, as the levelling between Turku and
Helsinki is somewhat imprecise.
The precision picture depends on the viewpoint, on the chosen datum.
Figure 3.17 shows a levelling network of four points. The height dif-
ferences AB, BC , CD , and D A are given — measured. Furthermore, the
heights above mean sea level of coastal points A and B, measured by a
mareograph or tide gauge are also given.
1. First we adjust the levelling loop, table 3.1a.
15 Not hard to imagine, as this is how it was 1809–1812, and informally before that
during the Swedish imperial era.
í Õ ! ¤. û
+0.072
A D D D
A A
3.443 3.375 3.394
3.443 3.462
−0.925 (+0.016) +0.321 X
2.533 2.514 3.058 2.533 3.077

B C B C B C
+0.548
A datum B datum
Mean error ±10 mm
^ F IGURE 3.17. Alternative vertical datums A and B.
16 2. Use point A as datum point and calculate the heights of the points,16
table 3.1b.
3. Do the same calculation, but now using point B as the datum point,
table 3.1c.
It is seen that in the latter case all calculated heights are greater by
0.019 m. The height differences are of course the same. The difference
^ T ABLE 3.1. Alternative vertical datums A and B.
(a)
Adjusting the levelling loop
Interval Observed Correction Adjusted
AB −0.925 −0.004 −0.929

BC +0.548 −0.004 +0.544
CD +0.321 −0.004 +0.317
DA +0.072 −0.004 +0.068
Closing error +0.016 ↑ 0.000
(b) (c)
Point heights, datum A Point heights, datum B
Point Height Mean error Point Height Mean error
A 3.443 ±0.000 A 3.462 ±0.010
B 2.514 ±0.010 B 2.533 ±0.000
C 3.058 ±0.014 C 3.077 ±0.010
D 3.375 ±0.010 D 3.394 ±0.014
16 The mean errors in the table are made up though realistic-looking.
í Õ ! ¤. û
81
of 0.019 m is precisely the “difference of height differences” of points A
and B between the two methods: (1) levelling plus adjustment, and (2)
mareographs. The difference stems from the uncertainty, or “measure-
ment error”, of the levelling and of the mareograph measurements, and
from the circumstance that the true mean sea surface is not a level or
equipotential surface.
The datum difference between datum A and datum B is 0.019 m.
The datum transformation is
(B ) ( A)
Hi = Hi + 0.019 m,
more generally (︂ )︂
(B ) ( A) (B ) ( A)
Hi = Hi + HA − HA .
Every adjustment produces precision estimates of the computed values,

or mean errors. If we assume that, as a given value, the height value of
the datum point is errorless (mean error zero), the mean errors of the
point heights will grow moving away from the datum point. Table 3.1
contains (invented) mean errors behaving in just this way. They are also
drawn in figure 3.17 as error bars. We see that the precision behaviour of
the network depends on the datum-point choice.
^ 3.8.2 Horizontal co-ordinate datums

As already said, horizontal networks can also have different datums,
horizontal datums. Figure 3.18 depicts one network both in datum AB
and in datum PQ .
◦ The AB datum is created by taking already known approximate
co-ordinate values for points A and B as the formal truth, and
adjusting the whole network without changing these. In this way
one calculates the co-ordinates of the other points, including P and
Q , in the same AB datum.
◦ The PQ datum is again created in the same way, but by fixing the
co-ordinates of points P and Q to prior given approximate values,
and adjusting the network while keeping these fixed. Thus one also
solves the locations of the other points in the network, but now in
the PQ datum.
The figure shows that the point co-ordinates are different in the AB and
PQ datums. The shape of the whole network is nevertheless the same,
independently of whether datum AB or PQ is chosen for calculating
í Õ ! ¤. û
PQ datum
B
A AB datum
F IGURE 3.18. Two different datums of a horizontal network, the AB and PQ

horizontal datums with their starting or datum points. Point
error ellipses and inter-point (relative) error ellipses have been
drawn. Note how the error vanishes in the datum points, the
co-ordinates of which are part of the datum definition and thus
^ assigned values.
the solution. In this case, the transformation between the datums is a

similarity or Helmert transformation.
The prior known co-ordinates for the points A , B, P, and Q originate
verkkotasoitus generally from previous network adjustments, astronomical position
determinations, or are read from a map: they are approximate values of
the co-ordinates. Determining a datum is thus the same as choosing the
datum points: points the approximate co-ordinates of which are taken as
the formal truth in the network adjustment.
It would be a coincidence if the computation of the network in the AB
datum produced the same co-ordinates as computation in the PQ datum.
The differences between the co-ordinates computed in different ways
are comparable in magnitude to the goodness of approximation of the
approximate co-ordinates used. The differences are often so small, that
the transformation parameters are close to zero or unity. In the Helmert
transformation 3.3:
the rotation angle θ is so small, that sin θ ≈ θ and cos θ ≈ 1. Writing

K = 1 + m, with m the scale distortion, m is also a small number, and we
í Õ ! ¤. û
Map projections and height systems. . . 3.9
83
obtain
x = x0 + (1 + m) u + (1 + m) θ v ≈ x0 + u + mu + θ v,
y = y0 − (1 + m) θ u + (1 + m) v ≈ y0 − θ u + v + mv,
as a matric equation
[︄ ]︄ [︄ ]︄ [︄ ]︄ [︄ ]︄[︄ ]︄
x u x0 m θ u
= + + ,
y v y0 −θ m v
an elegant equation in which the second and third terms on the right-
hand side are small, because they contain only the small transformation
parameters x0 , y0 , m, and θ . Thus, the co-ordinate differences x − u and
y − v are also small, as we observed already above.
^ 3.9 Map projections and height systems in a

three-dimensional world
Although the real Earth and her gravity field are three-dimensional
phenomena, which we can describe and handle correctly only in three
dimensions, nevertheless, means of description are very widely used that
are based on “two-plus-one-dimensional” thinking. For this are used map
projections and height systems, which together describe the world by
means of 2D + 1D co-ordinates ( x, y, H ).
Although we have here three co-ordinates, one cannot speak of gen-
uinely three-dimensional co-ordinates because, on the one hand, ( x, y)
and on the other, H , are not comparable.
Among ordinary people — and even among land surveyors — there is
a conceptual model of a “shoebox world”: rectangular, the sides oriented
in the northern and eastern directions, and the height co-ordinate being
simply the distance from the bottom of the shoebox, “sea level”.
It would be easy to be judgemental about this way of thinking. Re-
member, however, that in a small area, the shoebox model is — may
well be — an acceptable approximation. For example, inside cities, plan
maps and rectangular co-ordinates are used without bad repercussions.
The question of the acceptability of this approximation demands careful
analysis.
If the ( x, y, H ) representation is acceptable, then it is simpler, as a
representation of location and height, than the Earth’s true geometry,
true locations with their rectangular geocentric co-ordinates and true
í Õ ! ¤. û
heights with their geopotential numbers. However, in the general case it

is actually a more complicated way of representing location and height:
the complexity of map projections and the gravity field pervades all co-
ordinates thus defined. Errors caused by misconceptions are easy to
make.
Therefore in precise scientific work in geodesy, geocentric, three-
dimensional co-ordinates and geopotential numbers should always
be used. “Plane co-ordinates” — more precisely, map-projection
co-ordinates — and metric heights should always be seen as
derived quantities, on the basis of which no precise computations
should be attempted.
See the following diagram, in which the signs “←→” designate operations
used:
⎧
map projection
⎨(ϕ, λ) ←−−−−−−→ ( x, y)
reference ellipsoid
( X , Y, Z ) ←−−−−−−−−→ (ϕ, λ, h) geoid model (3.8)
⎩ h ←−−−−→ H
Reference ellipsoid A co-ordinate conversion between rectangular and

geodetic co-ordinates is a mathematical, exact operation. The
choice of reference ellipsoid is, however, arbitrary. Today, GRS80
is the standard. In Finland, however, historically the Hayford or
International Ellipsoid of 1924 has also been used.
Map projection A mathematical, exact operation. Many alternatives
are on offer.
Geoid model The height type may be orthometric, normal (or variants
of those two) or dynamic. A geoid model (or similar) is always
needed, see section 4.1.
In diagram 3.8 on the left are the more abstract quantities, whereas on
the right are the more concrete quantities, closer to daily life.
^ 3.10 The time co-ordinate

In geodesy we use, in addition to co-ordinates of place, other co-ordinates
as well. The first of these is time. Time describes the changes happening
in the Earth, the research into which belongs to the field of geodynamics,
on which more in section 18.1. With the measurement precision of modern
geodesy, the Earth lives and changes continuously:
í Õ ! ¤. û
Self-test questions
85
◦ Due to the solid-Earth tide, the ground below our feet moves peri-
odically, twice a day, up and down — even in Finland a couple of
decimetres. We do not notice this of course, because a more stable
reference point is lacking: everything around us moves up and
down with us.
◦ Because of plate tectonics, all continental plates move evenly. The
velocity of motion is of the order of a few centimetres per year, and
can be precisely monitored with GNSS technology.
◦ The rotation of the Earth is irregular. With space geodetic observa-
tion techniques it is possible to follow the variations in the direction
of the Earth’s rotation axis, both relative to the solid Earth — polar
motion — and relative to the celestial sphere — precession and
nutation — and variations in rotation rate — LoD, length of day. vuorokauden
These phenomena together are called “Earth orientation parame- pituus
ters” (EOP).
◦ In Fennoscandia, Canada and elsewhere, the Earth’s surface is
rising slowly after the last ice age, glacial isostatic adjustment
(GIA).
◦ There are other, more local motions too, some of them caused by
human activity.
1. How are map projections classified based on what they distort and
what they preserve?
2. What is the main distinction between “traditional” and “modern”
co-ordinate reference frames?
3. What are the differences between the old KKJ datum and the new
EUREF-FIN datum for the territory of Finland?
4. Name and describe geodetic plane co-ordinates and geodetic polar

co-ordinates. How do you convert between them?
5. Describe the geodetic forward problem in the plane, and its solution.
6. Describe the geodetic inverse problem in the plane, and its solution.
7. Describe the Helmert transformation in the plane. How many free
parameters does the transformation contain? Describe them.
8. What is a datum, and how is one established?
í Õ ! ¤. û
9. Describe briefly the various phenomena that geodynamics studies.
^ Exercise 3–1: Distances

We find the co-ordinates of two points in Finland: the GNSS fundamen-
tal station Metsähovi (METS), and the GNSS station at the Sodankylä
Geophysical Observatory (SODA). The EUREF data centre (ROB, EU-
REF Permanent GNSS Network) offers the following approximate ITRF
co-ordinates:
Station X (m) Y (m) Z (m)
METS 1 2 892 571.00 1 311 843.28 5 512 634.01

SODA 2 2 200 147.00 1 091 638.20 5 866 870.60
From this, we
1. compute the latitude and longitude ϕ and λ, again using the NLS
online service Finnish National Land Survey, Paikkatietoikkuna.
The software asks for 3D Cartesian (rectangular) co-ordinates in
the ETRS89 system, which is not quite the same as approximate
ITRF; we just pretend it is.
2. From these, we compute the distance, on the surface of the GRS80

reference ellipsoid, between the points, solving the inverse geodetic
problem. Use the NGS web-site NGS, Computation utilities to obtain
the distance s 1 .
3. Instead of a reference ellipsoid, we may use a spherical approxi-
mation, with the mean Earth radius being R = 6371.008 km. The
angular distance between METS and SODA can be computed from
17 their latitude and longitude, using the equation17 (cosine rule on
the sphere)
cos ψ = sin ϕ1 sin ϕ2 + cos ϕ1 cos ϕ2 cos(λ2 − λ1 ).
And then, the metric distance along the surface of the sphere is
s 2 = ψR.
jänne-etäisyys 4. Alternatively, the chord distance in spherical approximation:

(︁ 1 )︁
s 3 = 2R sin 2ψ .
í Õ ! ¤. û
Exercise 3–1: Distances
87
5. We return to the NLS web-site (Finnish National Land Survey,
Paikkatietoikkuna) and convert the co-ordinates (ϕ, λ) to ETRS-
TM35FIN, a two-dimensional map projection system.
6. Now, using these two-dimensional map projection co-ordinates ( x, y)

for METS and SODA, we compute the distance between the two
stations using the Pythagoras theorem in two dimensions:
√︂
2 2
s 4 = ( x2 − x1 ) + ( y2 − y1 ) .
7. Now, as the icing on the cake, compute the chord (three-dimensio-

nal) distance, also using Pythagoras, from the original rectangular
co-ordinates:
√︂
2 2 2
s 5 = ( X 2 − X 1 ) + (Y2 − Y1 ) + ( Z2 − Z1 ) .
All these distances are different. Some of the differences are small,
some substantial. Complete the exercise by explaining where all these
differences come from, and draw a figure with all distances marked.
Do all calculations in millimetres rounding accuracy.
17 Smart readers will note that this equation produces, for short distances ψ, a cosine
close to unity, leading to a loss of accurate digits when ψ is recovered by the arccos
function.
It is possible to convert the equation to a “half-angle version” in the following way:
substitute
ψ
cos ψ = 1 − 2 sin2 ,
2
λ2 − λ1
cos (λ2 − λ1 ) = 1 − 2 sin2 ,
2
yielding
ψ λ2 − λ1
(︃ )︃
2 2
1 − 2 sin = sin ϕ1 sin ϕ2 + cos ϕ1 cos ϕ2 1 − 2 sin =
2 2
λ2 − λ1
= (sin ϕ1 sin ϕ2 + cos ϕ1 cos ϕ2 ) − 2 cos ϕ1 cos ϕ2 sin2 =
2
λ2 − λ1
= cos (ϕ2 − ϕ1 ) − 2 cos ϕ1 cos ϕ2 sin2 =
2
ϕ2 − ϕ1 λ2 − λ1
= 1 − 2 sin2 − 2 cos ϕ1 cos ϕ2 sin2 ,
2 2
from which
ψ ϕ2 − ϕ1 λ2 − λ1
sin2 = sin2 + cos ϕ1 cos ϕ2 sin2
, (3.9)
2 2 2
the half-angle version of the spherical cosine rule which is well behaved for points that
are close together (compared to the size of the Earth).
í Õ ! ¤. û
^ Height measurement and the
levelling instrument
4
Lattamiehentie, 01260 Vantaa
Lattamiehentie, 80100 Joensuu
Vaakitsijantie, 90650 Oulu
Lattamiehentie, “Rodman Road”, Vaakitsijantie, “Leveller

Road”, street addresses
^ 4.1 Height, geopotential, and the geoid

Heights express the locations of points in the vertical direction, the
direction of the local gravity vector, the vertical or plumb line. luotiviiva
Intuitively this is based on the naïve “shoebox model” of the Earth’s

figure, where height is the third co-ordinate, the straight, metric distance
from the bottom of the shoebox, sea level.
The shoebox model is also called the “flat Earth approximation”: some-
where below the land surface there is a reference surface, assumed to
be a plane, that coincides with mean sea level. Height is the distance in
metres from this level surface.
In reality, the Earth is not flat and the reference surface is curved,
even undulating. The reference surface is called the geoid. It is a level
or equipotential surface of the Earth’s gravity field, a surface on which
every point has the same value for the geopotential, the potential of the
Earth’s gravity field.
The direction of gravity, the plumb line, is everywhere perpendicular
onto this surface. The distance of a point from this surface, measured
along the plumb line, is called its orthometric height. Thus, orthometric
height has a simple geometric interpretation, and it is of course a metric
quantity.
– 89 –
90 4 H EIGHT MEASUREMENT AND THE LEVELLING INSTRUMENT
^ 4.1.1 The geopotential

Physical geodesy is the branch of geodesy concerned with the gravity field
and gravity potential of the Earth. The geopotential W could be taken
as the fifth “co-ordinate” after the three co-ordinates of place X , Y , Z ,
and time. The geopotential describes the energy level of points in relation
to sea level. This corresponds to the popular conception of “height”. We
have a habit, mostly appropriate in daily life, of expressing height as a
metric quantity. . . but what really interests us is the potential energy
that comes with height. The effect of gravity on our daily activities is
so strong, that “height determination”, the study, determination, and
presentation of the geopotential, forms a large part of practical geodesy
and surveying.
Water “Height”, or potential, represents energy. The energy may be
recovered or stored — hydro-power. The energy may also be de-
structive — floods — for which one has to be prepared. Sewers
have to have a sufficient slope so they will work.
Air The levels of equal pressure in the air follow fairly precisely the
levels of the geopotential. The phenomenon is exploited in baro-
metric height determination. An aircraft also measures its “height”
using an air-pressure sensor.
Traffic Gravity affects the planning of traffic routes. Slopes may not be
too steep, and limited variations in potential, or energy level, along
the route are desirable. In the case of waterways, this happens
automatically in a natural way.
The geopotential is closely related to gravity. Surfaces having the same

value for the geopotential, equipotential surfaces, are what are ordinarily
called “horizontal surfaces”. A freely flowing fluid — sea water, lake
water, air — will settle along an equipotential surface. In the sea, the
hydrostatic pressure is a constant along the equipotential surfaces of
the gravity field, just like, in the atmosphere, barometric pressure — at
suolaisuus least approximately: disturbances are due to salinity and temperature
variations in water and temperature variations in air, and to the currents
these cause.
^ 4.1.2 Metric heights

Heights are “humanised” in a similar manner to what map projections
do for location data. One invents a way in which height “above sea
í Õ ! ¤. û
Height, geopotential, and the geoid 4.1
91
Topographic surface
C Equipotential H
surface
P H∗
Orthometric
Normal H dyn
W Dynamic
Plumb line
Geoid W0
F IGURE 4.1. Different height types map geopotential numbers C in different

^ ways to metric heights (differences exaggerated).
level” can be expressed as a metric quantity, as a height H from some

reference surface, usually mean sea level. The first step is always to
compute the difference between the geopotential values of the point
under consideration and of mean sea level, the geopotential number C of
the point. This number is positive from mean sea level upwards.
Unfortunately, just like with map projections, there is no solution that
is in all respects satisfactory. Something is always distorted.
In the same way as with map projections, there are also different height
types, like
◦ orthometric height H
◦ normal height H ∗
◦ dynamic height H dyn .
They all have their good and bad properties.

Earlier, in the section on geometric co-ordinates (section 2.8), we be-
came acquainted with co-ordinates bound to the reference ellipsoid, of
which one was the height from the reference ellipsoid, h. This co-ordinate
describes the location of a point in a vertical direction and is in a way
also the height of the point. It is nevertheless reckoned from the reference
ellipsoid, a surface that is not physically useable as a reference surface
in daily life. Moreover, it does not describe the energy level with respect
to sea level, like H , H ∗ , and H dyn (and of course geopotential numbers C )
do.
If all this appears at this stage difficult and theoretical, it may be
worthwhile to come back to this section later in the course and read it
again.
í Õ ! ¤. û
Plumb line
P
Shaft
H Topography
Tunnel
Sea level, geoid
Reference ellipsoid
F IGURE 4.2. Orthometric heights are metric distances from the geoid, the water
surface that would form if sea water could freely move under
the topography in an imaginary tunnel network. In that case,
orthometric heights H could be directly measured along the plumb
^ line through a shaft like the one depicted.
^ 4.2 Orthometric height

Orthometric (Greek “correctly measured”) heights H correspond the most
precisely to our concept of “height above sea level”. They are in principle
just metric heights above the geoid. The geoid is that equipotential
surface of the gravity field which on average is on the same level as
mean sea level. In other words, mean sea level continued under the land
masses.
If we could excavate a network of tunnels under the continents (figure
4.2) on the level of the sea surface, the water would spread throughout
the network in such a way, that its surface would be a physical realisation
of the geoid. A point’s orthometric height would be its distance from this
fluid surface. This direct physical interpretation is the reason why many
geophysicists, and many countries — among them Finland until 2007 —
have chosen to use an orthometric height system.
Building a tunnel network like the one depicted is of course not prac-
tical. Inland, the geoid is realised computationally, by carrying out the
í Õ ! ¤. û
Height determination and levelling 4.3
93
Orthometric height X
Height above the H
Plumb line
ellipsoid (e.g., Plumb line
measured by GPS)
h
Plumb line
Centre of
mass
Reference
ellipsoid
Topography
Geoid, height ⊵
from ellipsoid N h=H+N

^ F IGURE 4.3. Important reference surfaces and height concepts.
calculation of a height or levelling network, starting from a chosen coastal

point or set of coastal points. Thus, we disseminate orthometric heights
throughout the country, to everywhere that the levelling network extends
to.
In Finland up until 2007, the official height system or vertical datum
was N60, the zero point of which was Helsinki’s mean sea level at the
start of 1960. N60 heights are to a good approximation orthometric. In
2007, the N2000 height system was taken into use, with the zero level as
the mean sea level according to the official Amsterdam or NAP (Normaal
Amsterdams Peil, Amsterdam Ordnance Datum) datum. N2000 heights
are normal heights, the definition of which differs a little from that of
orthometric heights. For Finland, the difference is of little practical
significance.
^ 4.3 Height determination and levelling
^ 4.3.1 Exotic methods for height determination

The most direct way of measuring height differences is to realise an
í Õ ! ¤. û
20˚ ◦
20 24◦
24˚ 28◦
28˚ 32˚◦
32
20
◦
0 ◦˚
770 700˚
7
19
◦
8 ◦˚
668 688˚
6
24 25
22
21
23
31
28
29
30
20
26
27
◦
6 ◦˚
666 19 666˚
6
17
18
18
◦
4 ◦˚
664 6644˚
18
5
23 24 2
20 212
2
19
◦
2 ◦˚
662 6622˚
18
17
19 16
◦
0 ◦˚
660 600˚
6
15
16
20˚ ◦
20 32˚◦
32
24◦
24˚ 28◦
28˚
2010 Oct 20 13:27:28
F IGURE 4.4. The Finnish geoid model FIN2000 (data © National Land Survey
of Finland). This map shows the heights of the geoid above the
^ geocentric GRS80 reference ellipsoid. Unit m.
equipotential surface of the gravity field by means of a fluid surface. This

is how one can transfer geopotential values from one place to another.
◦ In Denmark and the Netherlands hydrostatic levelling has been
employed, in which a long hose or tube filled with distilled water
is used to transfer the level of the geopotential (the “height”) from
one island to another or between island and mainland. The dis-
í Õ ! ¤. û
Height determination and levelling 4.3
95
tances over which measurements have been done have been tens of
kilometres.
◦ By using water gauges in interior waters, heights may also be vesiasteikko
transferred hydrostatically. As in the water-hose technique, here
the air pressure difference between the terminals must also be
taken into account, as well as the effects of wind and currents. The
method, which works best under an ice cover, has been tested, for
example, in the Netherlands (IJsselmeer, Rijkswaterstaat 1996–97,
Reijnoudt, 1996) and Finland.
◦ Barometers have also been traditionally used for measuring height ilmapuntari
differences. A careful procedure taking into account natural air-
pressure variations due to weather will yield a best-case accuracy of
about one metre. See Heiskanen and Härmälä (1963) pages 84–87.
The geophysical modelling of sea currents has been attempted, for exam-
ple, in the Åland Sea.
Trigonometric traverse levelling must also be mentioned here, see linjavaaitus
figure 6.8 (Takalo, 1995).
A certain hi-tech method for measuring potential differences uses pre-
cise atomic clocks and the slowing of clocks predicted by general relativity
theory. At the time of writing there exist optical lattice clocks, atomic
clocks operating at optical frequencies, which should have the relative
precision of 1 : 1018 required for one-centimetre precision.
Another hi-tech method which is already in use (Gruber et al., 2014) is
the construction of precise, high-resolution geopotential models, which
may be used to calculate a point’s precise geopotential immediately when
satellite positioning has determined the point’s precise geocentric location.
The already completed satellite gravity mission GOCE (2009–2013) is key
to this.
^ 4.3.2 Levelling
The already mentioned levelling technique is the standard method for
determining heights referred to mean sea level. Levelling measures the
height difference between two points, figure 4.5. The distance between
points measured in one measurement set-up is short. By chaining point
intervals into a traverse levelling, height differences between points at
great distances from each other may be determined, and thus heights
mapped over wide areas.
í Õ ! ¤. û
Levelling staffs Horizontal

line of sight
Levelling instrument
back fore
∆ H = back − fore
10
View
20
(Nothing here)
^ F IGURE 4.5. The geometry of levelling.
In Finland, like elsewhere, the levelling network covers the whole coun-
try and provides the opportunity to determine the heights of points in the
height system of the network. There is a hierarchy for levelling networks:
the precise-levelling network, which was measured and maintained in
Finland by the Finnish Geodetic Institute, covers the whole country but
is sparse, the loops of the network being hundreds of kilometres long. The
lower-order levellings by the National Land Survey densify this network,
and local actors — municipalities, builders — connect their own levelling
networks to this system. This brings official heights within the reach of
all users.
korkeuskäyrä The height contours appearing on topographic maps are also in the
maastokartta official system, earlier N60, nowadays N2000.
^ 4.3.3 Creating a height system

The height differences ∆ H provided by geometric traverse levelling may
be added together only within a small area, where local gravity is con-
stant. In larger areas, the height differences ∆ H must first be converted
to geopotential differences ∆C :
∆C = g · ∆ H,
in which g is local gravity. After that, it holds for geopotential differences
í Õ ! ¤. û
The levelling instrument (“level”) 4.4
97
that, around a closed loop,
∑︂
∆ C = 0,
closed loop
although for raw height differences

∑︂
∆ H ̸= 0 !
closed loop
∑︁B
In other words, while the sum of height differences A ∆H depends on
the path chosen from A to B — and is thus not unambiguous — the sum
∑︁B
of potential differences A ∆C is independent of the choice of path. Being
unambiguous, the geopotential is better suited as the basis for the height
system of an area.
^ 4.4 The levelling instrument (“level”)
^ 4.4.1 Construction and function

Levelling (geometric levelling, figure 4.5) depends on a horizontal line of
sight: the optical axis of the levelling instrument’s measuring telescope,
or sight axis, is horizontal. To achieve this, the instrument has a spirit tähtäysakseli
level. Both the spirit level and the telescope are connected to the body of tasain
the instrument.
A traditional levelling instrument (figure 4.6) includes among other
things a measuring telescope, a small circular or bull’s-eye level for ap- rasiatasain
proximate levelling, and a precise tubular level. The instrument also tasaus
has a tripod and footscrews. The optical axis of a well-adjusted level-
ling instrument, the sight axis — the line defined by the crosshairs in hiusviiva-
the eyepiece — is parallel to the horizontal as defined by the level, the ristikko
okulaari
horizon.
At every instrument station, the levelling instrument must be levelled
anew. Many levelling instruments have a separate lifting screw for
precisely levelling the instrument. This is done before every forwards
and every backwards measurement.
The accurate parallellity of the sight axis and the tubular level’s horizon
is achieved using an adjustment screw when checking up the instrument.
^ 4.4.2 Classification
Levelling instruments are classified according to their accuracy, purpose
of use, and construction, in order of increasing accuracy, table 4.1.
í Õ ! ¤. û
Vertical axis
Adjustment screw
Horizon
Tubular level
Sight
axis
Telescope Measure-
ment
Lifting screw
Footscrews
Tripod
^ F IGURE 4.6. Levelling instrument.
We speak of precise levelling, if the precision on a one kilometre double

run is below 0.5 mm. Instruments of this accuracy class have an optical
levelling micrometer, see figure 4.17. The size and magnification of
the measuring telescope grows with accuracy class, from 20× to 40×
magnification. The levelling staff or rod is also chosen to correspond to
the accuracy class, see table 4.2.
^ 4.5 The measuring telescope

A measuring telescope is found in many optical geodetic measuring in-
^ T ABLE 4.1. Classification of levelling instruments.
Instrument type Levelling type
Builder’s level Construction, earthwork levelling

Engineer’s level Engineering, construction levelling
Precise level Base network levelling
High-precision level Precise levelling
í Õ ! ¤. û
The measuring telescope 4.5
99
Object
Eyepiece
Objective D
d f F
Virtual
Real image
Exit pupil image
Entrance pupil
^ F IGURE 4.7. Telescope.
struments, like the theodolite, chapter 5, and the levelling instrument.

A telescope is an optical device presenting an enlarged, and thus more
detailed, image of a remote object. Telescopes for use at night-time also
serve to collect as much light as possible from dim objects, in order to
present a bright image.
The telescope in figure 4.7 has an objective of diameter D , also called
the entrance pupil or aperture, and a focal length F. An eyepiece is used aukko
with a focal length f . If the object is at infinity, the objective first forms
a real image of it in the focal plane, which the eyepiece projects as an
enlarged virtual image to infinity. The magnification equals the ratio
µ= F f.
/︁
All the light leaving the telescope passes through the exit pupil, the
diameter of which is d = D µ . The exit pupil is the image of the entrance
/︁
pupil formed by the eyepiece. For example, for binoculars specified as kiikarit
7 × 50, we have D = 50 mm, µ = 7×, and d ≈ 7 mm. Young people with
a pupil diameter at night of 8 mm can thus place their eyes at the exit
pupils and capture all incoming light.
The tasks of the measuring telescope are to
◦ give a sharp image of the aiming target, the levelling staff
◦ form the sight axis by placing the crosshairs1 which are in the 1
eyepiece, onto the image of the far away levelling staff.
Both tasks demand precise focusing, the movement of lenses along the
1 The crosshairs are nowadays usually engraved onto a glass plate, a reticule. Still in
the 1930s, threads from a spider’s nest web were used! See Bedini (2005a,b).
Crosshairs were apparently invented by the astrometrist William Gascoigne (1612–
1644), who accidentally noticed the suitability of spider silk for this purpose. He was
killed in combat in the English Civil War.
í Õ ! ¤. û
Sight or collimation axis S
Focusing lens Objective

Image plane = crosshairs
Eyepiece
Movement (1)
Movement (2)
F IGURE 4.8. Measuring telescope and focusing. The object of study is to the
right (at a great distance!), the observer’s eye to the left. A mea-
suring telescope is used in many optical geodetic measuring in-
struments, such as the theodolite, chapter 5, and the levelling
^ instrument.
telescope axis, in order to achieve a sharp image of both the object and
the crosshairs. See figure 4.8.
Focusing is done as follows, usually by turning actuator rings or screws
on the telescope:
1. The eyepiece is turned so that the image of the crosshairs becomes
sharp.
2. The focusing element of the instrument is turned so that the image
of the target also appears sharp.
In this case, the image of the target formed by the objective, the focal
2 plane of the eyepiece,2 and the plane of the crosshairs all coincide.
In levelling, commonly equal distances to the front and back staffs are
chosen. If this is not possible because of the terrain, one should focus
carefully at every instrument station before every reading. If not, one
may get parallax: the apparent direction of the telescope’s optical axis
will depend on the position of the observer’s eye in relation to the eyepiece.
Observing through a poorly focused telescope also causes eye fatigue.
2 This only applies if the target is at infinity and the eye of the observer is error-free.
More precisely, the eyepiece + the possible eyeglasses of the observer + the living optics
in their own eye project a sharp image of crosshairs and target image onto the retina.
í Õ ! ¤. û
The tubular level 4.6
101
F IGURE 4.9. Parallax of a measuring telescope. If the image and crosshairs are
not in the same plane, moving the eye with respect to the eyepiece
^ will cause them to move with respect to each other.
One must always focus carefully!
Eyeglasses can be taken off if they are ordinary glasses and not cylindrical
(astigmatism) or prismatic3 (heterophoria), because near- or farsighted- 3
ness can be corrected by focusing the eyepiece.
^ 4.6 The tubular level

The construction of a tubular spirit level is explained in figure 4.10.
The adjustment screw seen in the picture is only used for adjusting the
instrument, rarely in the field. Its purpose is to get the level axis or
horizon L and the sight axis S of the telescope (figure 4.8) precisely
parallel. The level axis is horizontal when the bubble is in the middle, a
precondition for measurement.
The task of the tubular level is to help the observer bring the sight
axis of the levelling instrument into the horizontal plane, perpendicular
to local gravity. With it, the instrument is levelled in the measuring
direction in connection with every reading.
3 Prismatic glasses are relevant only when using binoculars.
í Õ ! ¤. û
Horizon L Adjustment screw
Not levelled Levelled
F IGURE 4.10. Tubular spirit level. Above, the construction, left below, the
bubble seen through the prism system in a coincidence level. The
^ angle α is the sensitivity.
The distance a is the interval between the level’s graduations. Gen-

osa-arvo erally a ∼ 2 mm. The sensitivity of the level is expressed by the angle
α.
In a coincidence level, a system of reflective prisms is used to show the
opposite heads of the bubble side by side, improving the precision of the
levelling achieved.
^ 4.7 Checking and adjusting a levelling instrument
^ 4.7.1 Field check

At certain time intervals one must verify that the sight axis S of the
levelling instrument is parallel with the axis, or horizon, L of the level.
Due to the influence of the environment, every instrument “lives” and
changes, for example, with variations in temperature and air pressure,
and due to handling and wear.
A check is carried out using levelling observations (field check): the
measuring distance ℓ is chosen as 25–50 m, depending on weather condi-
tions: the measurements are best done during overcast weather.
The checking method is based on the circumstance that, when mea-
sured from A , the measurement result, that is the height difference, is
correct, whereas when measured from B, the measurement result con-
tains an error 2ν, in which ν is the error caused by the difference in
í Õ ! ¤. û
Checking and adjusting a levelling instrument 4.7
103
f
f 2′ b
b2
f2
ν 3ν
b′2
f1 ν ν b1
B A
ℓ ℓ ℓ
^ F IGURE 4.11. The geometry of a field check (Kukkamäki method).
direction between the sight axis and level horizon, at staff distance ℓ. We
readily obtain
∆ h A = ( b 1 − ν) − ( f 1 − ν) = b 1 − f 1 ,
∆ h B = ( b 2 − 3ν) − ( f 2 − ν) = b 2 − f 2 − 2ν.
These height differences are identical. From this condition we solve for ν:
b 1 − f 1 = b 2 − f 2 − 2ν =⇒ ν = 12 ( b 2 − f 2 ) − ( b 1 − f 1 ) .
(︁ )︁
The figure shows that
f 2′ = f 2 − ν, b′2 = b 2 − 3ν.
These are readings that can now be calculated. This enables the correc-
tion of the difference in direction between the sight axis and level horizon
in the field, using the adjustment screw meant for just that.
^ 4.7.2 Adjusting a levelling instrument

When, as a result of the field check, it is known that the staff reading is
off by an amount ν, one proceeds as follows to adjust the instrument:
◦ The instrument has an adjustment screw for the tubular level,
which tilts the level with respect to the telescope (figure 4.12a). The
assembly of telescope and tubular-level is attached to the base by a
lifting screw.
1. Level the instrument first approximately using the footscrews,
then precisely using the lifting screw, all the time keeping an
eye on the tubular level.
2. Take the staff reading f .
í Õ ! ¤. û
L
Level
Adjustment screw
Lifting screw
(a)
The tubular level has an adjustment screw
L
Telescope
S
Adjustment screw
X
Base
Lifting screw
(b)
The telescope has an adjustment screw
L
Crosshairs
Lifting screw
(c)
The crosshairs plate can be adjusted
^ F IGURE 4.12. Adjusting the horizon of a levelling instrument.
3. Move, using the lifting screw, to the staff reading f ′ = f − ν,

where ν comes from the field check.
Note! Do not use the footscrews for this, because then the
instrument may also tilt in the transversal direction —
the footscrews are not in line with the sight axis of the
measuring telescope. Their only role is the approximate
levelling of the instrument using the bull’s-eye level.
í Õ ! ¤. û
Self-levelling instrument 4.8
105
4. Now the bubble in the tubular level is no longer in the middle.
Use the adjustment screw for the level to get the bubble in the
middle again. After that, L ∥ S .
◦ The instrument has an adjustment screw for the measuring tele-
scope: the telescope tilts with respect to the level (figure 4.12b). The
assembly of telescope and tubular level is again attached to the
base with a lifting screw.
An equivalent, popular technical solution is an adjustment screw
that shifts the crosshairs plate in the vertical direction within the
image plane (figure 4.12c).
1. Level the instrument.
2. Take a staff reading f .
3. Move, using the telescope’s (or crosshairs glass plate’s) adjust-
ment screw, to the new staff reading f ′ = f − ν.
4. The tubular-level bubble is still in the middle!
^ 4.8 Self-levelling instrument

Self-levelling or automatic levelling instruments use gravity to obtain a
horizontal sight direction.
Early models used gravity to level the whole measuring telescope
according to the pendulum principle (figure 4.13). It is clear that such an
instrument is difficult to use under field conditions due to disturbances
caused by wind and observer proximity.
In today’s instruments, only a prism or mirror guiding the light is used
as a pendulum. It is suspended inside the telescope: a pendulum com-
pensator is built into the telescope. In order to function, the instrument
must already be approximately levelled, for example with the aid of a
bull’s-eye level.
The principle of the self-levelling levelling instrument is shown in
figure 4.14.
The principle of operation of the pendulum compensator is explained
conceptually in figure 4.15, in which the path of the light beam has
been folded open. The figure shows the situation in a co-ordinate frame
attached to the measuring telescope.
A small tilt of the telescope away from the horizontal causes a tilt of the
incoming light beam of α. In order for the image to remain in the same
í Õ ! ¤. û
Weight
F IGURE 4.13. Principle of operation of an early self-levelling levelling instru-

^ ment. An instrument built in this way is not very practical.
place in the image plane of the telescope, the compensator bends the light
beam by an amount 2α, assuming that the distance between the objective
and compensator s is equal to that between the compensator and image
plane, in other words, the compensator is precisely in the middle between
them. A freely suspended mirror turns, relative to the telescope, by an
amount α, and the direction of the reflected beam changes by an amount
2α, just as intended. See Kahmen and Faig (1988) pages 334–336.
The strength of compensator instruments is their ease of use. However,
in the early days there were technical issues, such as magnetism of the
mirror suspension (Kukkamäki and Lehmuskoski, 1984). These problems
Cross-
hairs Optical
Prism axis
Horizontal
Com-
pen-
sator
^ F IGURE 4.14. A modern self-levelling levelling instrument.
í Õ ! ¤. û
Digital levelling instrument
Image
4.9
107
Compensator plane
plane Objective
2α
s s
^ F IGURE 4.15. Principle of operation of the pendulum compensator.
appear to have been solved.
^ 4.9 Digital levelling instrument

Nowadays digital levelling instruments are generally used, as the au-
tomation of measurement that they bring with them saves costs. The
measurements are stored directly into the instrument’s memory, and the
necessary checks are done immediately.
The staff used together with a digital levelling instrument bears a
bar-code that, at least in principle, is no different than that found on viivakoodi
merchandise. Thanks to it, height values can be machine read using the
instrument’s CCD sensor and processor system. As a side result, a crude
staff distance is also obtained, as well as a warning signal if the fore and
back distances are too different.
Unlike a traditional levelling staff, on which the measurement always
uses the edges of no more than two graduation lines, with a digital or
bar-code staff always a whole area is used, of size 30 cm in the case of
the Zeiss (Trimble) DiNi12 instrument. This has both advantages and
disadvantages.
Advantage In the measurements, some sort of average over the edges

of many staff graduations is used. Therefore, the accuracy of
manufacture of the graduations and the accuracy of calibration of
the staff are less critical. The staffs last longer in a useable state.
Disadvantages
◦ The whole interval on the staff that is being used should be
visible. In forested areas, this may cause problems.
◦ The calibration must always be done as a system calibra-
tion: the instrument and staff are calibrated as a “black box”,
together. On the other hand, by combining the calibration
í Õ ! ¤. û
30
20
10
F IGURE 4.16. With a traditional levelling staff, one may also read through
^ foliage. A bar-code staff causes confusion more easily.
of staff graduations and system calibration, one may recon-

struct how the instrument weights the graduations it uses,
and thus get the black box ajar.
The digital method of levelling is in widespread use, even in precise

levelling, and has been the subject of active research (Takalo et al., 2001;
Takalo and Rouhiainen, 2004).
^ 4.10 The levelling staff

The levelling staff or levelling rod is a scale, in Finland a metric one, with
which the height differences between two points are measured using a
jaotus levelling instrument. There are many alternatives for the graduation of
the scale, figure 4.17:
“E” graduation Simplest of all. Its weakness is that a bright white
small square appears a little larger than a dark red small square
4 — the Helmholtz4 brightness illusion. When one interpolates the
millimetres visually, small systematic effects easily occur.
Chessboard graduation Here, the above mentioned weakness has
been corrected.
Line graduation Used in precise levelling. The line interval is 10 or
5 mm.
4 Hermann Ludwig Ferdinand von Helmholtz (1821–1894) was a German physician and
physicist and a student of vision.
í Õ ! ¤. û
The levelling staff 4.10
109
65
04
20 64
03
63
02
10 10 cm 10 10 cm 62
01
5 cm 5 cm
00 00
(a) (b) (c) (d) (e)
“E” Chessboard Precise- Bar-code Levelling micrometer
graduation graduation levelling gradu-
staff ation
F IGURE 4.17. Graduation alternatives for the staff scale: “E” graduation, chess-
board graduation, precise-levelling staff, bar-code graduation. On
the right, an optical levelling micrometer to be placed in front of
^ the objective.
Instruments for precise levelling have an optical levelling microm-

eter, with which a better accuracy is obtained than by visually
interpolating the graduations. The micrometer in front of the tele-
scope objective contains a turnable glass plate, with which one
places the horizontal crosshair on one of the graduation lines. The hiusviiva
reading from the line gives the crude value, the turning scale of
the micrometer plate completes it to a precise reading.
Bar-code graduation To be used with digital levelling instruments. viivakoodi-
jaotus
The staffs are often manufactured from wood, better ones from alu-
minium. In precise levelling, an “invar staff” is used, on which the
graduations are painted on an invar tape, which is mounted on a wooden
or aluminium-alloy frame, and kept under tension by a spring.5 For 5
precise-levelling staffs, a graduation of half a centimetre is used — the
5 The force of the spring is known — 200 N — and its effect on the length of the invar
tape is computable. The coefficient of thermal expansion of invar is close to zero, and
metal, unlike wood, is insensitive to moisture.
í Õ ! ¤. û
^ T ABLE 4.2. Classification of levelling staffs.
Type of staff Length (m) Other
Simple 3–5 Foldable or telescoping, wood or aluminium, two

to four parts, “E” graduation
Base-network levelling staff. 3 Stiff, wooden, chessboard or line graduation, dou-
In German Zweiskalenlatte ble scale or scale on both sides (reversal staff),
bull’s-eye level
Precise-levelling staff 3 Wooden or aluminium frame, invar tape on which
double scale
Bar-code staff 3 Aluminium, invar tape. Used with digital level-
ling instrument
Industrial staff Double-scale staff of extreme precision
Self-calculating staff See subsection 4.11.2
“staff unit” is 5 mm. There are two graduation scales, slightly shifted
with respect to each other, as a double-check on reading mistakes and to
randomise reading errors. Sometimes the scales are on different sides of
kääntölatta the staff: a “reversion staff”.
Upmarket levelling staffs always sport a built-in bull’s-eye level. The

staff must be precisely vertical at the moment of reading!
Staffs are classified according to their purpose of use, see table 4.2,
which describes Central European practice.
A self-calculating staff or area-levelling staff has the following proper-
ties, see figure 4.21:
◦ The graduation increases from above going downwards.
◦ At the lower end is an adjustable foot, which may be pulled out on a
known point, so that the correct fraction of a metre becomes visible.
After that, one goes into the terrain to map any number of point
heights.
In high-precision work, the staffs need to be regularly calibrated, at least
before and after the field season. Digital levelling instruments and the
bar-code staffs they use should be calibrated as a system.
kiintopiste Suitable staff supports may be benchmarks, temporary wooden spikes
hammered into the ground, and the like. Standard staff supports are
kiila depicted in figure 4.18. A levelling spike is used when the soil is soft,
í Õ ! ¤. û
Levelling methods 4.11
111
Mallet
Blow protector
Staff Staff
Handle
(a) (b) (c)

Change plate Spike Rail shoe
F IGURE 4.18. Various temporary levelling-staff supports: change plate (Ger-

^ man: Frosch — “frog” —), levelling spike, rail shoe.
whereas a change plate is used on a hard substrate. A rail shoe is used kilpikonna
when levelling along railways. It may be left in place when a train speeds
past. During the first and second precise levellings of Finland, many
levelling lines ran along railways, but since, more and more, highways
have been used.
^ 4.11 Levelling methods
^ 4.11.1 Traverse levelling

Line or traverse levelling is used in base-network measurements. Its runkomittaus
objective is to bring the official height system close to all users in the
nation, to be used for example as the reference level for detail surveys. kartoitus-
mittaus
The precise-levelling network is the national height base network. It is
korkeusrunko
realised hierarchically by successive network densifications, section 8.3.
The heights are computed for the user community in the national height
system. The heights for the third and last Finnish precise levelling were
still computed in the N60 system; today the national height system is
N2000. All levelling network computations are however first carried out
with geopotential numbers, from which then either orthometric (N60) or
normal (N2000) heights are calculated.
Traverse levelling is done by summing the measurements between
í Õ ! ¤. û
b1 f1
bn fn
b2 f2
^ F IGURE 4.19. Traverse levelling.
6 successive instrument stations (figure 4.19) into a height difference:6
n
∑︂
∆H = (b1 − f 1 ) + (b2 − f 2 ) + · · · + (b n − f n ) = (b i − f i ) .
i =1
◦ The line runs from known point to known point. Sometimes this
piikki is impossible, and we speak of a “spike”. In this case we measure
carefully in both directions: check.
◦ A well-planned levelling network incorporates all measurements
and points into closed loops: check.
◦ In order to minimise the impact of weather- and instrument-related
errors, one makes the fore and back staff distances as equal as
possible: ℓback ≈ ℓfore . The staff distances also may not be too
long, for example in precise levelling, 50 m, however depending on
weather conditions. In overcast weather one can use longer staff
distances; in sunny weather with strong shimmer, staff distances
have to be shortened (Kääriäinen, 1966).
◦ If a levelling line runs along a railway or highway, safety arrange-
ments must be in order.
^ 4.11.2 Area levelling

pintavaaitus Area levelling is explained in figure 4.20. With this method, the height
situation of a whole area is mapped using one known starting point.
Among the points to be measured, there must be at least one other known
point: check.
6 The formula is exact only within a small area where gravity is constant. See section
4.1.
í Õ ! ¤. û
113
f f
f f
f
f Instrument
f b
b f
f f f
Benchmark
F IGURE 4.20. Area levelling. One backwards (b) observation, many forwards
^ ( f ) observations.
Area levelling is an ideal opportunity to use a self-calculating levelling

staff, see figure 4.21. At the starting point, the leg of the staff is pulled
out until in the levelling instrument one sees the correct decimal fraction
of a metre: if the height of the point is known to be 12.75 m, the leg is
pulled out until the observer sees in the telescope the number ⟨ n⟩ .750,
with ⟨ n⟩the number of whole metres. The leg is screwed tight, and the
staff is transferred to the first observation point. The observer must keep
track of the whole metres; the fractions show directly in the telescope.
The results of area levelling are needed and used for
◦ mapping the height situation on a building site before starting to
construct the foundation
◦ creating digital terrain models locally and at high resolution
◦ calculating earthwork volumes to be moved. maamassa
.75 .88
12.88
12.75 New point, fore ( f )
Known, back (b)
F IGURE 4.21. Self-calculating levelling staff. How the leg of the staff is set to
^ the right length.
í Õ ! ¤. û
For the area levelling to be useful, the approximate horizontal locations

of the measured points must have been established by some other means.
Proper checks are important: in addition to the starting point, other
points with known heights should be included in the measurement. An
erroneous starting height would propagate in full to the whole area,
which would be, in sewer construction for example, a fatal and expensive
mistake.
^ 4.11.3 Technical levelling

Installation measurement in industry and construction belongs to the
field of engineering geodesy.
◦ Extreme case: the CERN Large Hadron Collider (LHC) in
Geneva, circumference 27 km, precision on the level of mil-
limetres (Schrock, 2014).
◦ Paper machines, shipyards.
◦ Road construction, bridges, tunnels, railways.
Deformation measurement and monitoring For this purpose, lev-
elling is only one of many methods: deformations are usually
three-dimensional. See section 18.1.
◦ Extreme case: post-glacial land uplift.
◦ Deformations caused by the pumping of natural gas,
petroleum, or drinking or irrigation water, anthropogenic
maan land subsidence. Venice.
vajoaminen
◦ Deformations of dams, reservoirs, bridges, high steel.
vesiallas
◦ Old buildings and archeological artefacts, the tower of Pisa
and the good ship Vasa (Vasamuseet, Deformation Monitor-
ing).
^ 4.11.4 Levelling of profiles and cross-sections

Profiles and cross-sections are measured in connection with construction
projects, especially of roads, railways and canals.
A profile is a longitudinal section of the Earth’s surface, usually along a
7 planned trajectory of a road, railway or waterway.7
maastoon The work starts with setting out the trajectory into the terrain.
merkintä
7 In this case, a parallel line is used.
í Õ ! ¤. û
115
Motor
45◦
Light plane
Pentagon
prism
Laser,
freely
suspended
g
^ F IGURE 4.22. Principle of operation of a laser level.
Benchmark
Line, e.g., a road trajectory
Cross-section
Cross-section
^ F IGURE 4.23. A profile and cross-sections.
í Õ ! ¤. û
Markers are placed at 25, 50 or 100 metre intervals, as well as at

sharp terrain features. The markers are numbered: the number
signs are placed outside the work area. The heights are measured
with a builder’s level.
At least two tie measurements to benchmarks of the height base
network in the area are made, at the start and end points of the
profile. If the use of two benchmarks is not possible, the loop must
be closed as a double-check by measuring back and forth. Closing
errors are corrected in proportion to the levelled distances.
poikkileikkaus A cross-section is a transversal section of the Earth’s surface, perpen-
dicular to the trajectory of the profile. At corners, the angle is split
equally. Cross-sections are typically 20–50 m long. The purpose
of cross-sections is to support the planning work and enable the
calculation of earthwork masses to be moved. The height measure-
ment of the cross-sections is carried out in the same manner as
area levelling. The point density is chosen in accordance with the
terrain and the intended use.
^ 4.11.5 The laser level

lasertaso Often used in area levelling are laser levels, figure 4.22. Laser levels are
instruments stabilised by a pendulum compensator, which projects laser
8 light though a rotating pentagon prism8 into the environment, in order to
form a horizontal plane.

These instruments are handy on building sites, where they realise a
horizontal plane, which the user can make visible with a stick. Spreading
sand, laying floors, or bricklaying a wall along a straight line become
easier. A staff equipped with a suitable sliding receiving device will
directly give the height of the point on which it stands.
A laser level can also usually be quickly reconfigured to produce a
vertical plane of laser light. This is also useful on building sites.
1. What is the reference surface for orthometric heights?
8 Inside the prism are two reflective surfaces with an angle between them of 45◦ . The
angle between the incoming and outgoing rays is thus always 90◦ or 100g .
í Õ ! ¤. û
Exercise 4–1: Heights
117
F IGURE 4.24. Metsähovi research station from the air. Google Maps™. Google
terms of use. Images © 2019 Google, © 2019 Maxar Technologies.
^ Map data © 2019 Google.
2. What is the relationship between the geoid surface and deviations

of the plumb line?
3. How is the sight axis of a levelling instrument (or a theodolite)
defined?
4. How does one focus a measuring telescope?
5. What is parallax, and how does one avoid it?
6. What are the graduation alternatives for levelling staffs?
7. Name three standard levelling-staff supports.
8. In levelling, what is the role of the rodman, the person holding the
staff?
9. Describe the Kukkamäki method of field testing a levelling instru-
ment. What is the purpose of the test?
10. Describe how an automatic or self-levelling levelling instrument
works. Make a drawing!
11. Describe how a laser level works.
í Õ ! ¤. û
^ Exercise 4–1: Heights

Earlier we gave the geographical or geodetic co-ordinates (on the GRS80
reference ellipsoid) of the GPS antenna at Metsähovi research station,
equation 2.5:
ϕ = 60◦ 13′ 2′′.89046, λ = 24◦ 23′ 43′′.13336, h = 94.568 m. (2.5)
1. Use the geoid model given on the UNAVCO web page (UNAVCO,
EGM96 geoid calculator), which is good to ±25 cm on the Finnish
territory, to determine the geoid height N at Metsähovi. Alterna-
tively use Karney, which also gives the newer EGM2008 model.
2. Compute the height H above “sea level”, as represented by the
EGM96 model. The relationship you need is given in figure 4.3.
3. From the air, Metsähovi research station looks like figure 4.24. The
METS GNSS antenna is mounted on the tall steel grid mast to the
right.
The height, in the N2000 system, that is, above sea level, of the
pääkiintopiste fundamental benchmark PP2000, figure 3.16, is 54.4233 m. Assume
that this is also the approximate height of the terrain at the foot of
the steel grid GNSS mast, ±1 m.
Question: how tall is this steel grid mast?
4. Go to Google Maps and measure the latitude and longitude of the
foot of the mast. Compare with the above values for latitude and
longitude. How large are the differences? Discuss.
í Õ ! ¤. û
^ The theodolite
5
Our money’s all spent, to the deuce it went!
The landlord, he looks glum,
On the tap-room wall, in a very bad scrawl,
He has chalked to us a sum.
But a glass we’ll take, ere the grey dawn break,
And then saddle up and away —
Theodolite-tum, theodolite-ti, theodolite-too-ral-ay.
From the folk song “The Old Survey”, around 1905 (The
Institute of Australian Culture, 2012)
T HE THEODOLITE WAS probably invented by Leonard Digges. The in-

vention was published by his son Thomas in his land-surveying textbook
Pantometria in 1571. The origin of the name is unclear. The first theodo-
lites did not yet have a telescope, which was possibly not invented — or
at least attempted to be patented — until 1608 by Hans Lippershey in
the Netherlands.
The theodolite is an angle measurement instrument which measures
horizontal and vertical angles relative to the local horizon (horizontal
plane) and plumb line (direction of gravity). Modern instruments called luotiviiva
electronic tacheometers or total stations also measure distances. Here,
we shall call all of these “theodolites”.
^ 5.1 Horizontal angles and zenith angles

Theodolite measurements are always made inside the Earth’s gravity
field. The vertical axis of the theodolite is levelled to be in the direction
of the local plumb line. Directions and direction differences may be
expressed, in these natural instrument co-ordinates, as horizontal and koje-
koordinaatit
– 119 –
120 5 T HE THEODOLITE
Vertical circle Alidade level
Non-
Tele- standard
scope level
Micro-
scope
Index
Micro-
level
scope
Horizontal circle
F IGURE 5.1. An old-fashioned theodolite. Note the external horizontal and

vertical circles and reading microscopes. Base image Wikimedia
^ Commons, Exploration theodolite.
vertical angles.
Let there be (figure 5.2) a difference in direction, as seen from the
instrument, between points A and B. The points are projected along the
plumb line onto the local horizontal plane of the instrument, yielding
points A ′ and B′ . The difference in direction between points A ′ and B′ ,
the angle α, is the horizontal angle between points A and B. Both the
horizontal directions and horizontal angles are counted positive in the
clockwise direction.
The angles ζ A and ζB are the vertical or zenith angles of points A and
B.
Horizontal angle The angle α formed by the projections O A ′ and OB′

of the rays O A and OB in the horizontal plane.
Zenith angle The angle formed by the plumb line and the ray O A (ζ A )
or OB (ζB ). The zenith angle is always positive.
The plumb line or vertical, the direction of a freely suspended plumb wire,
í Õ ! ¤. û
The axes of a theodolite 5.2
121
+A
ζB Projection
ζA A′
Instrument O
α Hor
izont
al p
lan
e
+B
Projection
B′
Plumb line
^ F IGURE 5.2. Horizontal angle and zenith angle.
is the local direction of the Earth’s gravity vector. The plumb line points
at the centre of mass of the Earth, but only approximately.1 1
^ 5.2 The axes of a theodolite

A theodolite has three axes, figure 5.3:
◦ the vertical axis or standing axis, which must be directed along the
local plumb line or gravity vector
◦ the horizontal axis or trunnion axis, around which the telescope
turns
◦ the sight or collimation axis, the axis of the telescope tube passing tähtäysakseli
though the crosshairs of the eyepiece. hiusviiva-
ristikko
A theodolite has two circles, the horizontal and vertical circle, figure 5.3.
okulaari
Theoretical objective for a well-adjusted theodolite: vaaka- ja
pystykehä
◦ Hz, V and S intersect at the same point.
◦ V ⊥ Hz. If not, the error is called trunnion-axis tilt. tappikaltevuus
◦ S ⊥ Hz (but not S ⊥ V ! Why not?). The error is called collimation
1 Approximately, mainly because the Earth is an ellipsoid of revolution and not a sphere.
The deviation of the plumb line from the direction to the Earth’s centre of mass due to
the flattening is at its largest as much as 11′ at latitudes ±45◦ .
The plumb line is also ever so slightly curved. In addition, there are local, varying
deviations of the plumb line from the surface normal to the ellipsoid of revolution, of an
order of magnitude ranging from a few seconds of arc in even terrain to over a minute
of arc in the mountains.
í Õ ! ¤. û
Objective
Vertical axis
Hz Vertical angle
O
X
S
Horizontal
Sight or axis
X
collimation
Eyepiece
axis
Vertical circle
0.000 gon
Horizontal direction
Horizontal circle
^ F IGURE 5.3. The axes and circles of a theodolite.
error.
◦ Hz and V pass though the centres of the vertical and horizontal
jakokehän circles. If not, we speak of circle eccentricity.
epäkeskisyys
^ 5.3 Construction of a theodolite

See figure 5.4.
^ 5.3.1 Measuring telescope

The measuring telescope of a theodolite is in principle the same as that
of a levelling instrument (figure 4.8): a complicated assembly of lenses,
prisms and mirrors. The telescope turns around the horizontal axis and
pystykehä is joined in its movement by either the vertical circle itself, or by the
index and reading microscope of the vertical circle, depending on the
instrument type.
í Õ ! ¤. û
Construction of a theodolite 5.3
123
Reading optics (microscope)
Vertical circle
Vertical-circle
Collimation level lockscrew
Index fine-motion
screw
Horizontal axis ( Hz)

Vertical-circle fine
motion screw Alidade level
Vertical-circle
microscope
Horizontal-circle
Alidade
microscope
Horizontal-circle fine Vertical Horizontal circle

motion screw axis (V ) Horizontal-circle
X Bull’s-eye level lockscrew
Base
X
Forced Footscrews
centring
Theodolite
footplate
Tripod table
“Tribrach”
Plummet
^ F IGURE 5.4. Theodolite construction.
^ 5.3.2 Reading devices

The theodolite contains reading devices for the horizontal and vertical
circles:
◦ Most often, the images are led by prisms to a microscope eyepiece
beside the eyepiece of the measuring telescope, to facilitate the
work of the observer.
í Õ ! ¤. û
◦ The scale microscope and the optical micrometer are means to

obtaining a greater reading accuracy, see Kahmen and Faig (1988)
pages 66–67.
◦ In a coincidence microscope, readings are taken from two opposite
places on the horizontal or vertical circle. In this case, an optical
micrometer is also always used (Kahmen and Faig, 1988, pages
72–76).
^ 5.3.3 Alidade
The alidade (Arabic al-idhâdah, ruler) is the central part of the theodolite,
which turns around the vertical axis, carrying the telescope with it. It
contains a tubular level, the alidade level, to precisely level the theodolite,
see subsection 5.4.4.
^ 5.3.4 Base
runko The base of the theodolite is the fixed part to which the horizontal circle
is attached, and on which the alidade rests in its bearing.
pakkokeskistys- ◦ The base contains a forced-centring or footscrew device called a
laite tribrach.
◦ The upper part of the base sits forcibly centred in this device: it
may be detached from and reattached to precisely the same place.
◦ The forced-centring device is attached to the tripod head using
2 a large screw.2 In the middle of the tripod head there is a large
round hole and a mechanism to allow enough horizontal motion for
keskistys centring.
tasaus ◦ The footscrews are for levelling the theodolite: the vertical axis of
the theodolite is oriented along the local plumb line.
rasiatasain ◦ A bull’s-eye level is used for crude levelling, while the alidade level
helps in precise levelling.
riippuluoti ◦ The attachment for a string plummet, and an optical or laser plum-
met, are for centring.
^ 5.4 Theodolite handling in the field

As an expensive fine mechanical and optical instrument, a theodolite
should always be treated with due respect:
2 The 5
screw conforms to the geodetic standard of 8 inch, 11 threads per inch.
í Õ ! ¤. û
Theodolite handling in the field 5.4
125
^ F IGURE 5.5. Forced-centring device or plate.
◦ Transport, especially over longer distances, is always done in the

carrying case.
◦ The parasol is for the instrument, not the observer. It also offers
some protection against rain.
◦ The instrument is never aimed directly at the Sun: the glass plate
with the engraved crosshairs would crack, after which the instru-
ment would need to be repaired and re-calibrated. Further damage
occurs if the instrument contains a range-finder. etäisyysmittari
◦ Careful book-keeping is important: write up anything that might

be relevant, like weather conditions (metadata).
^ 5.4.1 Monuments and point descriptions

Unlike a levelling instrument, a theodolite must be placed precisely over
the point to be measured, so that the theodolite measurement refers to
the point.
In figure 5.7 the point Q has been monumented in the terrain, a mea-
surement point with a central mark. Figure 5.6 shows some examples of
monument types in use. maastomerkki
The choice of monument type must be made such that
◦ The point is clearly and uniquely defined. If one intends to do both
precise horizontal positioning and heighting, it is best to use a steel
í Õ ! ¤. û
^ F IGURE 5.6. Various monument types.
bolt with a round head with a small hole in the middle.

◦ The monument withstands the impacts of the weather and environ-
routa ment — frost heaving! A bedrock point is the best. Iron monuments
peruskallio should be protection painted (rust-protection paint; “red lead” is no
lyijymönjä longer recommended for toxicity reasons).
◦ The monument can be easily found — point description!
◦ The point number is marked on the monument or painted (and
chiselled!) next to it.
pistekortti When a point is monumented for future use, a point description must
always be drafted, which helps to find the monument even after decades.
The point description may contain the following information:
◦ The distances of the point from at least three features in the nearby
landscape which are believed to be permanent — trees, building
corners, etc. — measured by tape. In the point description, a sketch
is included of the relative positions of the point and the reference
features.
◦ An approach map including kilometre counts for car drivers, road
signs, description of the landscape, and other useful details.
◦ Precise co-ordinates for hand-held GNSS.
◦ Well-chosen photographs.
^ 5.4.2 Centring and levelling

In order to ensure that angles measured from the point are correct, two
actions need to be performed: centring and levelling.
keskistys Centring In figure 5.7, points Q and O , the monument and the inter-
section point of the theodolite’s axes, must be on the same plumb
line.
tasaus Levelling The vertical axis V of the theodolite must be aligned with
the plumb line. Then, when the theodolite is turned around the
í Õ ! ¤. û
127
Vertical axis V
Hz ⊥ V
S ⊥ Hz
O Horizontal axis Hz
Sight axis S
Plumb
line
g
Q
^ F IGURE 5.7. Theodolite axes. Q is a point monumented in the terrain.
vertical axis, the horizontal axis Hz will maintain a constant angle

with the plumb line: ideally, without trunnion-axis tilt, 100g . The
horizontal axis will thus always be in the horizontal plane.
Centring and levelling are commonly done alternatingly, until the desired tasaus ja
end result is achieved. keskistys
^ 5.4.3 Crude centring

◦ The tripod is placed on the right spot, judging by eye and by chang-
ing the lengths of the legs.
◦ It is also judged by eye that the tripod head is horizontal.
^ 5.4.4 Precise levelling

The precise levelling of the theodolite using the alidade level is done in
the following steps, see figure 5.8:
1. Crude levelling is done with the bull’s-eye level.
2. Level first in the direction of footscrews 1–2, by turning the screws
simultaneously in opposite directions. Take into account, if neces-
sary, the zero error from step 4.
3. Turn the alidade 100g , and also level in this direction using
footscrew 3.
4. Turn the alidade 200g . If the alidade level is adjusted correctly, the
bubble must again be in the middle. If not, move the bubble back
í Õ ! ¤. û
3 3 3
A B
1 B 1 A
1 2 2 2
^ F IGURE 5.8. Precise levelling of a theodolite using the alidade level.
half of the difference, that is to the equilibrium position, using the

footscrews. The zero error of the level is half the shift of the bubble
between steps 3 and 4.
5. Repeat 2–4 until the levelling no longer changes.
Make sure that the bubble can move freely and its end does not attach to
the edge of the level.
If the difference found in step 4 is large, then the alidade level is in
need of adjustment.
^ 5.4.5 Precise centring

The method of centring will depend on the type of plummet used. The
most common is the optical plummet.
No plummet is exact. Moreover, measuring the height of the instrument
over the monumented point is subject to error. Because of this, in special
measurements, when the precision requirement is sub-millimetre, one
uses instrument pillars permanently anchored in the bedrock or on a
deep support (frost heaving!) instead of tripods.
riippuluoti String plummet
◦ Traditional.
◦ Attached in such a way, that levelling does not change the
centring.
◦ Centring: shift the forced-centring device of the theodolite
over the surface of the tripod head in such a way that the tip
of the plummet points at the monumented point. Tighten the
attachment screw of the theodolite. The central hole in the
tripod head gives room to play.
◦ Levelling with the theodolite’s footscrews.
í Õ ! ¤. û
129
g g
Gravity Gravity
Rod
String
Tip
Plummet
200 gon
^ F IGURE 5.9. String plummet and rod plummet.
Problem: sensitive to the wind.

Rod plummet sauvaluoti
◦ A telescoping tube, the upper end of which is attached to the

forced-centring device (through the central hole in the tripod
head), with the lower end placed precisely on the central
mark.
◦ The rod has a bull’s-eye level.
◦ Centring:
1. Move the forced-centring device so that the bull’s-eye
level on the rod is in the middle.
2. Turn the level 200g to the opposite side of the rod: if
the bubble remains in the middle, centring has been
achieved.
3. If not, move the forced-centring device so that the bub-
ble moves half way back to the middle, that is to the
í Õ ! ¤. û
equilibrium position.
The rod plummet comes as part of certain types of tripods (Kern),
which are often used in accurate engineering surveying measure-
ments.
optinen luoti Optical plummet
◦ In the forced-centring device’s side (or in the side of the

theodolite) there is a small telescope with a prism, which
looks straight down. In the focal plane of the telescope’s eye-
piece there are crosshairs or something similar, called an
index. Katso kuva 5.10.
◦ The forced-centring device must have been levelled using the
bull’s-eye level.
◦ Centring and levelling procedure using an optical plummet
(figure 5.12):
1. Using the footscrews of the forced-centring device, place
the index initially on the point marker — meaning that
levelling is lost!
2. By changing the lengths of two of the tripod’s legs, get
the bubble of the bull’s-eye level in the middle again.
The forced-centring device describes a circular movement
around the tip of the third leg, and the place of the index
on the image of the point changes only little.
3. By a parallel shift — loosen the big screw of the forced-
centring device just a little — get the index of the optical
plummet back on the point marker.
Note! Shift, do not turn, because then the levelling will
change!
◦ Because in practice both the levelling 3 and the place of the
index 2 will nevertheless always change a little, repeat proce-
dure 1–3 until the desired end result is achieved. Fortunately
the procedure converges rapidly.
^ 5.4.6 Problem situations

Normally, the centring and levelling procedure as described above will
quickly produce a satisfactory result. In real-life field work, however,
pathological situations are guaranteed to arise, as in figure 5.13. We
í Õ ! ¤. û
131
Level
Shift
(no turn!)
Monumented point
^ F IGURE 5.10. Optical plummet.
leave these as an exercise for the reader.
^ 5.4.7 Forced centring

In precise measurements, a forced-centring device or tribrach is used:
instrument and signal fit into the same device, figure 5.14. In this way tähys
F IGURE 5.11. A benchmark (tube monument) seen though an optical plummet.

^ After focusing, both the target and the crosshairs are sharp.
í Õ ! ¤. û
F IGURE 5.12. An optical plummet and a bull’s-eye level are used at the same
^ time to achieve centring and levelling.
^ F IGURE 5.13. Problem situation.
í Õ ! ¤. û
133
GNSS antenna
Signal
Instrument
Optical plummet
Forced-
centring Lock screw
device
“Tribrach”
Attachment screw
(do not open!)
X
F IGURE 5.14. Principle of forced centring. The instrument and the signal fit
^ into the same device.
one eliminates many of the errors arising from the centring of a tripod.
In really precise measurements, like in engineering geodetic measure-
ments, we use a separate optical plummet. Over short distances, centring
may be the largest error source.
In the measurement, care is taken that every side is measured in both
directions: in a network of three points, the procedure may be according
to figure 5.15. From every instrument station, measure to all (nearby)
points to be equipped with signals.
Forced centring is also useful in situations where we measure on the pakkokeskistys
same point with both the satellite technique (GNSS) and a terrestrial
instrument. The antenna is then attached, through an adapter, to the
forced-centring device, figure 5.14.
í Õ ! ¤. û
^ F IGURE 5.15. Measuring a network using forced centring.
^ 5.4.8 Checking an optical plummet

The most common plummet is the optical plummet — although the laser
plummet is gaining ground. A laser plummet works practically in the
same way as an optical plummet, only the light travels in the opposite
direction (Crawford, 2009).
The weak spot of the optical plummet is that it must be adjusted so
that it looks really straight down when the bubble of the level is in the
middle — the same problem as with a levelling instrument’s sight axis
and tube-level horizon having to be parallel, figure 4.12. This property,
achieved by adjustment, can easily be lost in the handling of the device.
For this reason it must be checked at regular intervals.
One procedure for checking is described in Kahmen and Faig (1988),
on pages 95–96:
1. Mount the forced-centring device — or the whole theodolite, if the
optical plummet is built-in — onto a tripod, level it, and mark the
point on the floor seen in the plummet’s eyepiece on paper taped to
the floor.
2. Mark the outline of the forced-centring plate with pencil or chalk
on the tripod head.
3. Loosen the forced-centring device from the tripod, rotate it 120◦ ,
and place it carefully back into the drawn outline. Level, and mark
on the floor again the point under the crosshairs of the plummet’s
eyepiece.
4. Repeat 3.
5. If the points drawn in steps 1, 3 and 4 on the floor are identical,
then the plummet is in good adjustment. If not, shift the crosshairs
of the eyepiece, using its adjustment screws, to the centre of mass
í Õ ! ¤. û
135
23
F IGURE 5.16. Checking an optical plummet. The plummet should be adjusted

so that it always points to the centre of mass of the three points,
^ the black cross.
of the three points on the floor.

A non-adjusted optical plummet makes the observational material col-
lected with it worthless. A regular check of the state of adjustment, for
example before and after field work, is just as important as measuring
the theodolite instrument height over the point.
^ 5.4.9 Sighting and targeting

◦ As with a levelling instrument, always do the focusing carefully.
◦ At the beginning of the observation work, focus the crosshairs by
adjusting the eyepiece. If you wear ordinary eyeglasses, leave these
off, as explained in section 4.5.
í Õ ! ¤. û
^ F IGURE 5.17. Good signals for horizontal angles.
◦ For every target, focus carefully on the signal using the focusing
screw.
sarjahavainto- ◦ During precise measurements using the method of complete sets,
menetelmä one is not allowed to re-focus during measurement, as then the
collimation will change a little. This presupposes that all targets
are sufficiently far away.
Problem situations:
◦ The signal is thin, for example very far away, and partly covered by
the crosshair line. This is not a problem if the crosshairs look like
figure 5.18a.
◦ One side of the signal is in shadow as sunlight comes from the side.
The asymmetry causes an error called the phase error. See figure
5.18b.
Black-and-white signals are always best.
^ 5.5 Taking readings

In demanding theodolite measurements it is good practice to measure
3 always in both faces,3 face left and face right. The instrument has two
face positions because it has two axes: turning it around both axes by
an amount of 200g will make the measurement telescope point at the
4 same object again. This redundancy4 allows the elimination of a number
kojeasento I of systematic errors. The first position, “face left”, meaning the vertical
3 The inventor of the theodolite with two faces, or “transit”, was presumably William J.
Young of Philadelphia, USA, in 1831.
4 See for theoretical background Wikipedia, 3D rotation group, Topology.
í Õ ! ¤. û
Taking readings 5.5
137
Sunlight
Shadow
(a) (b)
Better crosshairs Phase error
F IGURE 5.18. Targeting. On the left, better crosshairs. Here, both a single and
a double line are available for use. On the right, the effect of
^ asymmetric lightfall (“phase error”).
circle to the left of the telescope as seen by the observer, yields vertical
angles near 100g if the sighting direction is close to the horizontal. The
second position, “face right”, yields values close to 300g . kojeasento II
The traditional method of complete sets, which further reduces a num-
ber of instrument errors, is no longer used:
◦ The measurement of first-order or national base networks has fully runkoverkko
transitioned to using satellite positioning.
◦ Digital angle measurement techniques automatically carry out
a procedure equivalent to the method of complete sets, without
observer intervention.
^ 5.5.1 Graduation circles and the classification of theodolites

The graduation circles of a theodolite are generally made of glass. The jakokehä
diameter is 60–100 mm, for the most precise instruments 250 mm. The
interval for the main scale is 1, 0.5, 0.2, or 0.1 gon, depending on the
precision class of the instrument. The traditional method draws a dense
line pattern using a “graduation machine” into a layer of wax covering jaotuskone
the glass, the lines are etched into the glass using acid, and filled with
dye. This was for a long time a carefully protected Swiss business secret
(Penry and Ingram, 2013).
Theodolites are classified, traditionally and somewhat unofficially,
into “one-minute theodolites”, “one-second theodolites”, and precision
í Õ ! ¤. û
106 107
108 109 110

0 10 20 30 40 50
109.33
106◦ 17′.2
Estimate
Estimate
(a) (b)
Line microscope Scale microscope
^ F IGURE 5.19. Various types of reading microscopes.
theodolites. The boundaries separating these classes are not well defined.
One source (Simonen, 2012) gives > 1 mgon for one-minute theodolites,
0.5 − 1.0 mgon for one-second theodolites, and < 0.5 mgon for precision
theodolites. See also Anon. (1971).
^ 5.5.2 Reading devices and fine-reading methods

The reading device magnifies the image of the scale on the graduation
circle though a microscope. With it, a fine reading is interpolated between
the numbers on the main scale.
Reading microscopes are divided into two types (figure 5.19):
◦ Line microscope:
– Magnification of the main scale and an index line.
– The fine readings are estimated visually, precision is about
jakoväli one tenth of a graduation interval.
◦ Scale microscope:
– An additional scale, the length of which is the same as the
graduation interval of the main scale: generally, main inter-
val is 1 gon, the interval of the additional scale 0.01 gon, 100
jakoviiva graduation lines.
í Õ ! ¤. û
Taking readings 5.5
139
Index line
∆
Image plane
109 110 111
Plan 110.367
plate Estimated digit
D
α
10 20 30 40 50 60
Graduation
^ F IGURE 5.20. Optical micrometer and its reading.
– The graduation line of the main scale is used as the index of

the additional scale.
The line microscope may be equipped with an optical micrometer in order

to achieve a better reading precision, figure 5.20.
When the glass plate is turned by angle α (assumed small), the light
ray shifts in a parallel fashion by an amount5 5
1 1
(︂ )︂ (︂ )︂
∆ = D sin α cos α 1 − n ≈ αD 1 − n ,
in which n is the index of refraction of the glass. The mechanism for taitekerroin
turning the glass plate has a scale showing the value ∆ in the angular
units of the main scale.
The graduation lines of the main scale are double lines, so aligning
with them is easy.
5 Because
⎫
PQ = D sin α,
1 )︂
⎬ (︂
sin α ⎭ =⇒ Q ′ Q = D sin α 1 − ,
PQ ′ = D sin α′ = D n
n
Q′
Q P
′
α
D
α
í Õ ! ¤. û
Eyepiece
Illumination
mirror
Micrometer
Circle
^ F IGURE 5.21. Reading the graduation circle. One circle location.
The nonius or vernier is no longer used in theodolites. In Heiskanen

(1943) it is told that the Portuguese Pedro Núñez (Petrus Nonius) in-
vented a precursor to the nonius as early as in 1542, and the Frenchman
Pierre Vernier today’s nonius in 1631. See also Kahmen and Faig (1988)
page 65.
^ 5.5.3 Reading the circle

◦ One circle location is read, figure 5.21.
◦ In precision theodolites, two opposite circle locations are simulta-
neously read. In this way the eccentricity error of the circle is
cancelled out. See figure 5.22. The crude reading is taken from the
circle, the fine reading from the micrometer: 244 + 0.4 + 0.0617 (in
which 7 is estimated) or 244.4617.
A single circle reading is used in one-minute theodolites, whereas the

method using two opposite readings, using a coincidence microscope, is
usually found in one-second and precision theodolites.
The same methods are used for reading the horizontal and the vertical
circles. Generally the diameter of the vertical circle is less than that of
the horizontal circle, so the reading precision is correspondingly less.
The reading device includes furthermore
it follows that (︂ 1 )︂
∆ = Q ′ Q cos α = D sin α cos α 1 − .
n
í Õ ! ¤. û
Taking readings 5.5
141
Linked
motion of
glass plates
Opposite-sides
circle readings
244.4
45 44 45 44
244 245 244 245
0 10 20 50 60 70 80
0.0617 (7 estimated)
Before micrometer setting. . . . . . and after.
^ F IGURE 5.22. Reading the graduation circle. Two opposite circle locations.
◦ A reading microscope, to which all the readings of the optics are

guided. In the field of view of the microscope, usually both hori-
zontal and vertical readings can be seen. One has to be careful to
choose the correct and wanted numbers. Often, the vertical reading
is marked with the letter V and the horizontal reading with H or
Hz.
◦ An illumination system:
– a mirror that can be turned to guide light to the reading optics
– a light source with a battery to be mounted in the place of the
mirror or in its own socket.
í Õ ! ¤. û
Objective
V
200g
X
Collimation error c = S − S ′
Image
plane Hz
S
200g
S′
Crosshairs
^ F IGURE 5.23. Adjusting the sight axis by shifting the crosshairs.
^ 5.6 Instrumental errors of a theodolite

The instrumental errors of a theodolite are divided into the following
groups:
◦ Axis errors:
– The axes are not perpendicular to each other: collimation error
and trunnion-axis tilt, subsections 5.6.1 and 5.6.2.
– The axes do not intersect at the same point.
◦ Eccentricity errors:
– eccentricity of the graduation circles
– eccentricity of the measuring telescope.
jaotusvirhe ◦ graduation errors of the circles.
In modern theodolites the eccentricity and graduation errors are small.

The instrumental errors cannot in general be corrected, but they can
be determined, in calibration.
^ 5.6.1 Collimation error

The most consequential error is the collimation error, which is easy to
determine and correct. Collimation error means that the angle between
the sight axis S and the horizontal axis Hz is not a right angle: S ̸⊥ Hz.
The sight axis S is realised by the crosshairs in the measuring telescope
í Õ ! ¤. û
Instrumental errors of a theodolite 5.6
143
(more precisely, S is the straight line going through the optical centre of
the objective and the crosshairs, figure 5.23). This is why by shifting the
crosshairs one adjusts to S ⊥ Hz.
◦ If there is no collimation error ( c = 0), the readings of the same
target A in face left and face right are a 1 and a 2 , where a 1 = kojeasennot I
a 2 ± 200g . ja II
◦ If there is collimation error ( c ̸= 0) and the telescope is turned

through from face left to face right (200g around both the horizontal
and the vertical axis), then A will not show under the crosshairs
until the telescope is turned an amount 2 c more.
Assume that object A is approximately in the horizontal plane. Take, in
face left, reading a 1 , and, in face right, reading a 2 .
The correct readings, without collimation error, would be A 1 and A 2 ,
and we would have exactly A 1 = A 2 ± 200g .
In reality we obtain a 1 = A 1 + c and a 2 = A 2 − c. Their difference is
a 1 − a 2 = 2 c ± 200g ,
from which one obtains
c = 12 (a 1 − a 2 ± 200g ) . (5.1)
In this way one may determine c, usually a small number.

Because c is so small, only fractions of a gon are of interest, not the
integer number of gons. Therefore we use the following notation:
[︁ ]︁
a signifies the rounding residue of a, the difference between the
[︁ ]︁
precise value and the value rounded to an integer. So 127.4531 =
[︁ ]︁
0.4531, 16.9850 = −0.0150, and so on. The outcome of the opera-
tion is always between the values −0.5 gon and +0.5 gon.
Then
c = 12 a 1 − a 2 .
[︁ ]︁
This can be calculated from the observation notebook, if the same objects
have been measured in both faces.
[︁∑︁ ∑︁ ]︁ ∑︁[︁ ]︁
a1 − a2 a1 − a2
[ c] = 12 n = .
2n
a1 reading taken in face left
í Õ ! ¤. û
a2 corresponding reading in face right (correspondence means both

the same set and the same direction)
n number of readings (sets × directions).
Correcting collimation error Carry out, in face left, a turning of the

instrument so that the reading is the precomputed “true value”
6 A 1 = a 1 − c. Shift6 the crosshairs (focal plate) to be on top of the
image of the target A . Verify, in face right, that the second “true
value” A 2 = a 2 + c is obtained when the crosshairs are on the target.
A calculation example from an observation notebook (all values in gon):

∑︁4
Face 1 2 3 4 i =1
Left 40.5223 88.6932 119.7601 205.7571 454.7327

Right 240.5217 288.6924 319.7591 5.7564 854.7296
a − a2 −199.9994 −199.9992 −199.9990 200.0007 −399.9969

[︁ 1 ]︁
a1 − a2 0.0006 0.0008 0.0010 0.0007 0.0031
The number of observations per face n = 4, so

3.1 mgon
= 0.3875 mgon = 3cc .875.
[︁ ]︁
c =
8
This is a good field check. In this example, there is hardly any collimation
error.
^ 5.6.2 Trunnion-axis tilt

tappikaltevuus The second axis error is the trunnion-axis tilt t. This error means that
the horizontal axis, or trunnion (Wikipedia, Trunnion) axis, Hz, is not
actually horizontal even after levelling the instrument: it and the vertical
axis V are not perpendicular, V ̸⊥ Hz.
The trunnion-axis tilt may be determined by using a target far away
from the horizontal (ζ ̸= 100g ). Then it holds that
a 1 = A 1 + c + t cos ζ, a 2 = A 2 − ( c + t cos ζ) ,
from which the combined term c + t cos ζ can be determined in the same
way as explained above for c. Separating the errors c and t requires
measurements at two different vertical angles ζ. If one of these is ζ =
100g (cos ζ = 0), we get back the original formula 5.1 for calculating the
collimation error.
6 This is a task for maintenance: the screws are small and hidden.
í Õ ! ¤. û
145
Trunnion-axis tilt
Hz′
Hz
^ F IGURE 5.24. Trunnion-axis tilt.
^ 5.6.3 Zenith-angle measurement and index error

Measuring a zenith angle with a theodolite requires that the index of pystykehän
the vertical circle — the place where the values for the zenith angle are indeksi
being read — is in the horizontal plane. For this purpose a theodolite has
its own fine-motion screw (figure 5.25) with which the frame on which
the index or indices of the vertical circle and the vertical-index level
(collimation level) are mounted together is moved. The frame can be
turned slightly around the horizontal axis. The measuring telescope and
the vertical circle are similarly connected to each other.
Before every zenith-angle measurement it must be ascertained, using
the fine-motion screw of the level and index, that the vertical-circle indices
are really in the horizontal plane. Of course one may not assume that
the vertical-index level has been adjusted so that the index really gives
exactly 100g and 300g precisely when the optical axis of the measuring
telescope is horizontal. This error is called the index error ( i ). It can
be eliminated by measuring in both faces, left and right. One moves
between faces by “plunging the instrument through”: turn the telescope
200g around the vertical axis, and 200g around the horizontal axis. The
í Õ ! ¤. û
Vertical index level or

collimation level
S
0
Index
10
0
30
0
Index
Fine-motion screw for

level and index
Alidade
^ F IGURE 5.25. Observing a zenith angle.
measurement values obtained in both faces are added together:
ζ1 + ζ2 = 400g + 2 i,
from which the index error i follows.

Both observations ζ1 and ζ2 are corrected by an amount − i :
ζ′1 = ζ1 − i, ζ′2 = ζ2 − i.
After this, the condition

ζ′1 + ζ′2 = 400g
holds exactly.
See figure 5.26, which depicts (unlike figure 5.25) a theodolite of which
the vertical circle is read only in one place. In the figure, the sight axis is
in the horizontal plane, so ζ = 100g . The formulas given below, however,
apply generally.
In the left image, the angle ζ is measured and the reading ζ1 = ζ + i
is obtained, in which i is the index error. In the right image, the same
angle ζ is measured, but in face right, and the reading obtained is ζ2 =
(400g − ζ) + i . We obtain:
ζ1 + 400g − ζ2 = ζ + i + ζ − i = 2ζ.
í Õ ! ¤. û
147
V
g
200 Hz
0 200
200g
100
300
100
300
0 −i
200 −i
−i −i
True zenith angle: ζ 100 True zenith angle: ζ 100

Reading: ζ1 98 Reading: ζ2 298
Index error: i = ζ1 − ζ −2 Index error: i = ζ2 + ζ − 400 −2
“True value”: 100 “Inferred index error”: −2

ζ1 − ζ2 + 400 ζ1 + ζ2 − 400
ζ= i=
2 X 2
^ F IGURE 5.26. Index error.
The angle ζ:
ζ = 12 (ζ1 − ζ2 + 400g ) .
The angle i :
ζ1 + ζ2 − 400g = ζ + i − ζ + i =⇒ i = 21 (ζ1 + ζ2 − 400g ) .
The corrected reading is
ζ = ζ1 − i = 400g − ζ2 + i.
Removing the index error Assume the construction in figure 5.25.

1. Aim at a target of which the “true angle” ζ has been calcu-
lated.
2. Turn the index and collimation level together, using the fine-
motion screw, to reading ζ.
3. The bubble of the level will move, adjust the level with its
adjustment screw(s)7 until the bubble is in the middle again. 7
7 The adjustment screws are small and may be somewhat hidden. Adjustment is a
maintenance job.
í Õ ! ¤. û
Check
1. Observe the target in face left (ζ1 ) and immediately after, the
same target in face right (ζ2 ).
2. Calculate the sum ζ1 + ζ2 . If the index error i = 0, ζ1 + ζ2 =
400g . This is also a good field check.
In figure 5.25 there is a manual vertical-circle index (height index). An

automatic height index is more common. The principle is similar to that
used in a self-levelling levelling instrument, and like it, the index must
already be approximately level:
◦ A pendulum compensator.
◦ A liquid compensator using silicone oil. The light to the reading
microscope is either refracted through the liquid or reflected from
its surface. See Kahmen and Faig (1988) pages 394, 395.
◦ In electronic theodolites, this compensation mechanism has been
implemented digitally using a digital tilt meter. So, the index is not
adjusted but rather, the reading is corrected computationally based
on the reading from the tilt meter.
In zenith-angle measurements, instrumental errors are largely the same
as for horizontal angles. Some are eliminated by measuring in both
faces. Measuring complete sets will certainly not diminish the impact of
vertical circle graduation errors, because the vertical circle, unlike the
horizontal circle, cannot be loosened and turned, see section 6.3. It does
allow however for a check on the correctness of the measurements.
The observations in face left and face right should be made as quickly
as possible in succession. For this reason, vertical angles should always
be observed separately, never together with horizontal angles.
^ 5.7 Electronic theodolites

In electronic theodolites all measurements are obtained in numerical
form, facilitating automatic storage, correctness check in connection
with the measurements, and forwarding. The monetary savings in the
measurement activity can be substantial, not just in measurement time
saved, but also in quality gain, when the error-prone reading and manual
writing down of observations is eliminated.
However, though electronic theodolites record the observations on their
own — typically raw data is not even recorded but already pre-processed
í Õ ! ¤. û
Electronic theodolites 5.7
149
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1
0 0 0 0 1 1 1 1 1 1 1 1 0 0 0 0
0 0 1 1 1 1 0 0 0 0 1 1 1 1 0 0
0 1 1 0 0 1 1 0 0 1 1 0 0 1 1 0
^ F IGURE 5.27. A four-bit Gray code.
data — the information to go with the data (“metadata”) must still be

carefully recorded.
In electronic theodolites, the edges of the graduation circles have been
imprinted with different line patterns.
^ 5.7.1 Absolute encoding circles

The Gray code8 is often used, which consists of bit strings in which, at 8
every step, only one bit changes. In figure 5.27 we see an example of a
four-bit Gray code, which has 16 different values.
In realistic applications, more bits are used, for example if one wants
a horizontal-angle resolution of 0.1 gon, there must be 400 · 10 = 4000
different values. This already requires 12 bits.9 9
The code is a black-and-white pattern imprinted on the circle, which a

row of light sensitive diodes, photodiodes, scans, figure 5.28 left. valodiodi
The advantage of the Gray code is that one knows constantly, unam-
biguously, which place on the circle one is reading; in other words, there
is a zero direction marked on the circle. This is why the encoding is called
absolute. In the example case, four photodiode rows side by side are
needed, one for each bit field.
The patterns on the circles are observed electro-optically, in our exam-
ple using a row of photodiodes, nowadays an integrated microelectronic
circuit, like a CCD sensor. The light detector observes the edges between
the black-and-white fields on the circle. Crude, absolute values originate
from the edges of the pattern on the circle; more decimal places are ob-
tained from the interpolating property of the diode array — or imaging
light sensor.
8 Frank Gray (1887–1969) was an American physicist and electronicist, a developer of

television technology. The Gray code is part of a patent awarded to him in 1953.
9 . . . because 212 = 4096 ≳ 4000.
í Õ ! ¤. û
6.7
6
0
1
2
.7
6
.6
7
15
(a) (b)
Absolute encoding circle Incremental encoding circle
^ F IGURE 5.28. An absolute and an incremental encoding circle.
^ 5.7.2 Incremental encoding circles

koodauskehä There is no zero direction marked on the encoding circle. We may only
follow changes in the instrument orientation by counting lines, figure
5.28 right. In the incremental solution, at least two diode detectors are
needed to establish the turning direction: in the depicted example there
are ten. The incremental method of course can only measure direction
differences, that is angles. Crude readings come from counting lines on
the circle, more decimal places are again obtained by interpolation.
^ 5.7.3 Modern automatic instruments

In electronic theodolites, commonly an ATR, automatic target recognition,
system is used. For this purpose, a CCD image sensor has been placed
in the image plane of the measuring telescope. The telescope is aimed
visually at the object. The uncorrected reading comes, as usual, from the
digital encoding-circle system. To the reading is added programmatically,
by means of image processing, a correction from the CCD image.
With electronically readable circles one uses many read-out detectors of
which each reads the code from a different place on the circle, figure 5.29.
í Õ ! ¤. û
Electronic theodolites 5.7
151
Processor
Alidade
Hor. circle
Base
F IGURE 5.29. Electronic readout of the horizontal circle: readings are taken
from all parts of the circle. This replaces the method of complete
^ sets known from optical theodolites.
All readings are collected in a processor, which averages them. In the

average, the periodic and random errors of the circle are eliminated or
essentially reduced. Compare this with the method of complete sets used
with optical theodolites, section 6.3.
With electronic theodolites, locking and unlocking the horizontal circle, vaakakehä
and turning it — and thus applying the method of complete sets by
hand — is unnecessary and impossible. Turning over the telescope and
measuring in both faces is however still necessary, for the same reasons
as with optical instruments (section 5.5).
In some instruments, a rapidly spinning circle is used, drawing on the
same technology used in spinning computer disc drives. In this dynamic
solution, the light detectors transform the line pattern on the circle into
a block signal, figure 5.30. The time shift between two block signals
together with the circle’s spinning velocity ω gives directly the angle α( t)
between the detectors as a function of time — plus an unknown constant,
the “ambiguity”,10 which does not depend on time. For this reason, the kokonaisluku-
tuntematon
10
í Õ ! ¤. û
Processor
α
∆t =
ω
Alidade
Horiz.
circle
Base
F IGURE 5.30. A spinning circle converts the measurement of angles into one of
^ time differences.
spinning circle method is also incremental: it can only measure changes

in the angle.
The advantages of the method are:
◦ It converts angle measurement into electronic time-difference mea-
surement, which can be extremely precise.
◦ The precision with which the graduation lines on the circle are
produced is not critical; the even spinning speed of the circle is.
The range of measurement systems and technical solutions for elec-
tronic theodolites is broad and rapidly developing. Therefore we only
take a closer look at one case.
^ 5.8 Case: Leica robotic tacheometer TCA2003

TCA2003 (Leica, 1997) is a good example of a fairly modern electronic
theodolite or tacheometer. The instrument measures horizontal angles
10 Obviously because to the detector, every line on the circle looks like every other.
í Õ ! ¤. û
Case: Leica robotic tacheometer TCA2003 5.8
153
Cursor bar Clock Battery status Data entry block
7 8 9
Header PÄÄVALIKKO : OHJELMAT 17:25
1 Orientointi 4 5 6
2 Kaarileikkaus
Menu 3 Maastoonmerkintä
X 1 2 3
4 Pisteiden välimatka
Function EXTRA KALIB JÄRAS DATA ASEMA MITT. 0 . +/–

keys
F1 F2 F3 F4 F5 F6 Shift CE
CODE aF... CONT ON
OFF ESC
Light Electronic level Other Arrow keys
^ F IGURE 5.31. Leica TCA2003 robot tacheometer control panel (in Finnish).
and zenith angles as well as slant ranges. The calculation capabilities of vinoetäisyys
its built-in software are quite versatile.
^ F IGURE 5.32. Leica TCA2003 robot tacheometer.
í Õ ! ¤. û
Measurement precision as stated by the manufacturer amounts to 0′′.5

or 0.15 mgon in horizontal and vertical angle measurement, and 1 mm +
1 ppm in distance measurement. The longest measurement distance
under normal conditions is 3.5 km.
The instrument has co-axial ATR (automatic target recognition, subsec-
tion 5.7.3), which also takes part in angle measurement. The CCD sensor
measures the deviations in the horizontal and vertical directions of the
laser beam reflected by the prism, and guides the motors so that the
crosshairs are placed almost over the prism. The small remaining devia-
tion is measured in the CCD image and the angle readings are corrected
correspondingly.
The ATR system can also be programmed to follow a moving target, or
to systematically seek — scan for — a target if it is not in the expected
place.
The instrument has self-calibration facilities, a measurement pro-
pystykehän gramme for determining the vertical-circle index error (subsection 5.6.3),
indeksivirhe the collimation error (sight axis not perpendicular to horizontal axis,
subsection 5.6.1), the trunnion-axis tilt (horizontal and vertical axes not
perpendicular to each other, subsection 5.6.2) as well as the zero point
or constant error of the distance measurement device or range-finder
(subsection 7.4.4).
Many “corrections”, or rather reductions, of distance measurement, like
weather corrections, reduction to the horizontal plane, and even reduction
to the map projection plane (for example using the Gauss–Krüger or UTM
projections) can be done within the instrument itself.
laserluoti The instrument has a laser plummet, which works in the same way as
an optical plummet — the precision is also similar, a little better than
±1 mm — but the light moves in the other direction, from the theodolite
down to the ground. Levelling and centring procedures are otherwise the
same as with an optical plummet.
The instrument has, in addition to a bull’s-eye level, an accurate elec-
tronic level. The display is an LCD and looks the same as a real bull’s-eye
level. It is self-calibrating, and it is not necessary to turn the instrument
by 100g or 200g every time it is levelled (figure 5.8).
The instrument is highly automated and offers a pre-programmed
monitoring measurement of 50 points. This is useful especially in defor-
mation measurements in industry and construction. Two programming
languages are on offer: GSI (Leica) for simple use, and GeoCOM for
í Õ ! ¤. û
Self-test questions
155
advanced use. In addition, the GeoBasic environment allows the develop-
ment of more applications in the PC environment, and their upload to the
instrument.
The exchange of data between the instrument and a computer can be
done in two ways:
◦ Through a serial interface (RS232). This technique is obsolete:
modern devices use a USB port or Bluetooth.
◦ Using a non-volatile PCMCIA memory card. Storage capacity may
be 512 kB–4 MB. The format of the card is the MS-DOS file system
FAT. This solution has also been obsolesced by USB memory sticks.
The format of the observations themselves is the Leica-designed GSI

(Leica) (Geo Serial Interface), documented in the manual.
1. Describe the three axes of a theodolite. Which of them are mutually
perpendicular?
2. Before measuring from a known point, a theodolite must be cen-
tred and levelled. Describe the stages in which this is done, for a
theodolite equipped with an optical or laser plummet.
3. Explain the idea of forced centring. Why is it useful?
4. Reading the circles. What is a scale microscope, what is a coinci-
dence microscope? Make a drawing!
5. Explain Gray codes.
6. What is the difference between absolute and incremental encod-
ings?
7. What is ATR, automatic target recognition?
8. Describe collimation error, trunnion-axis tilt and index error. How
are they determined?
9. A theodolite should never be aimed directly at the Sun. Are there
exceptions to this rule? Google “Roelofs prism”.
í Õ ! ¤. û
^ Angle measurement
6
[. . . ] When preparing a third American edition for the press,
Blunt decided that Bowditch had revised Moore’s work to
such an extent that Bowditch should be named as author.
The title was changed to The New American Practical
Navigator and the book was published in 1802 as a first
edition. Bowditch vowed while writing this edition to “put
down in the book nothing I can’t teach the crew,” and it is
said that every member of his crew including the cook could
take a lunar observation and plot the ship’s position.
Nathaniel Bowditch biography, in The American Practical

Navigator, Bowditch, 2017 edition
^ 6.1 Horizontal angle measurement

In traditional topographic surveying, horizontal angles are measured to maastomittaus
be used in three different contexts:
1. One carries out a triangulation, where horizontal angles and dis- kolmiomittaus
tances, or side lengths, are measured between points of the triangle
network. The triangle sides may be as long as tens of kilometres.
From the observations, the geodetic co-ordinates of the triangle
points are computed, hierarchically in densification stages of the
network. In this way, the first-order triangulation network mea-
sured by the Finnish Geodetic Institute has been densified by the
Finnish National Land Survey.
2. As a further densification of the triangle network, traverse measure- monikulmiojono
ments are carried out. In this method, side lengths and bending
angles are measured between successive points of a traverse (figure
– 157 –
158 6 A NGLE MEASUREMENT
α
α
(a) (b)
Triangulation Radial survey
(c)
Traversing
^ F IGURE 6.1. Use situations for horizontal angle measurement.
6.1c) in order to determine traverse point co-ordinates. To make

calculation possible, the starting and closing points of the traverse
need to be known, for example from a higher-order triangulation.
3. From the points of a traverse, radial surveys, for example, may be
conducted, figures 6.1b and 8.10.
Many of these observation techniques have been replaced by GNSS mea-
surement. Triangulation is no longer done anywhere, while traverse
measurements are done only in situations where GNSS is not useable,
such as underground — tunnels — or in high-rise urban landscapes.
^ 6.2 Intersection and resection

1 Geometries that are commonly used are intersection and resection,1 figure
6.2.
Intersection works as follows: let the known points A and B have co-
ordinates ( x A , yA ) and ( xB , yB ). Let the perpendicular projection
point between these two points of the unknown point C be P, and
1 In Finnish eteen- ja taaksepäin leikkaus, in Swedish avskärning och inskärning, in

Dutch voorwaartse en achterwaartse insnijding, in German Vorwärtsschnitt und Rück-
wärtsschnitt.
í Õ ! ¤. û
Intersection and resection 6.2
159
Unknown
P
C Unknown
θ2
θ1
C
K2
K1
R1 R2
X
θ1
α A
A P β
M
B B
Known Known
(a) (b)
Intersection Resection
^ F IGURE 6.2. Intersection and resection.
its co-ordinates ( xP , yP ). Then
PC = AP tan α = PB tan β,
so
AB = AP + PB = PC cot α + PC cot β = PC (cot α + cot β) .
From this
AB
PC =
cot α + cot β
and
AB cot α AB cot β
AP = , PB = .
cot α + cot β cot α + cot β
Now we use these distances, or equivalently, the coefficients cot α
and cot β, as weights in calculating the co-ordinates of point P as
weighted averages of the co-ordinates of points A and B:
x A cot β + xB cot α yA cot β + yB cot α
xP = , yP = .
cot α + cot β cot α + cot β
After this, the direct calculation formulas for the co-ordinates of
intersection point C are
yB − yA
xC = xP + tan α ( yP − yA ) = xP + ,
cot α + cot β
xB − x A
yC = yP − tan α ( xP − x A ) = yP − .
cot α + cot β
í Õ ! ¤. û
Resection is inherently more difficult, because it is an inverse problem.

The problem was studied as early as the 17th century by Snellius
2 and Laurent Pothenot2 and is often named after them.
Let (figure 6.2b) M be the midpoint of points A and B. If from the
unknown point P the angle θ1 between points A and B has been
measured, then this point is on the circumference of a circle that
passes through points A and B, and seen from the centre point K 1
of which the angle between points A and B is 2 θ1 .
The line segment MK 1 stands perpendicularly on AB: the co-
ordinates ( x M , yM ) of the midpoint M are calculated using the
formulas derived above for the special case α = β. The co-ordinates
of point K 1 are calculated in the following way — remember that
in the role of α we now have 90◦ − θ1 :
xK 1 = x M + 21 cot θ1 ( yB − yA ) = 12 ( x A + xB ) + 12 cot θ1 ( yB − yA ) ,
yK 1 = yM − 12 cot θ1 ( xB − x A ) = 12 ( yA + yB ) − 21 cot θ1 ( xB − x A ) .
The radius of the circle is obtained by the Pythagoras theorem:

√︂
1
R1 = ( xB − x A )2 + ( yB − yA )2 .
2 sin θ1
Now, because there is also another point pair B and C through
which runs a circle of its own, with centre point K 2 and radius R 2 ,
we have a pair of quadratic equations describing two circles:
( x − xK 1 )2 + ( y − yK 1 )2 = R12 , ( x − xK 2 )2 + ( y − yK 2 )2 = R22 .
From this, the co-ordinates ( x, y) of the unknown resection point

P can be solved — at least in principle. There are many ways of
solving this, for example, linearisation with respect to a pair of
approximate values ( x0 , y0 ) with an iterative solution.
These two circles have two intersection points, from which the right
solution must be chosen, and it is not point B. . .
Singularity The precision of resection depends on the geometry
of the points. If points A , B, C , and the unknown point lie
on the same circle, the solution is even impossible: we speak
of a singularity. Then, the circles are identical, and any
point on that circle serves as a solution. We also speak of the
dangerous circle: already close to this geometry, precision
deteriorates ominously.
2 Laurent Pothenot (1650–1732) was a French mathematician and geodesist.
í Õ ! ¤. û
Observation method of complete sets 6.3
161
Note also that the situation is conceptually similar if two distances
R 1 = PK 1 and R 2 = PK 2 are measured to the now known points
K 1 and K 2 . Here, P is the unknown point, co-ordinates ( x, y), and
a singularity occurs when K 1 , K 2 , and P are on a straight line.
When using distance measurement, we do not speak of intersection
and resection, but they are similar situations.
^ 6.3 Observation method of complete sets

The graduation of the horizontal circle of an optical theodolite has been vaakakehä
manufactured with great care, but is always imprecise. Every gon of the
graduation should be the same size, precisely one gon, but in reality is not.
However, the sum of all the gons of the graduation will always be exactly
400 gon. This is why the effect of graduation errors can be diminished by jaotusvirhe
measuring the same angle using different sectors of the horizontal circle,
and taking the average of those. This principle is implemented in the
observation method of complete sets. sarjahavainto-
menetelmä
In this method, traditionally used in the measurement of base networks,
runkoverkko
the same fan of directions is measured multiple times — complete sets —
such that between the sets, the horizontal circle is unlocked3 and turned 3
g/︁
(by roughly 200 n , in which n is the number of complete sets). In this
way one makes use of the different sectors of the horizontal circle, and
systematic errors, for example due to graduation errors of the horizontal
circle, are minimised as one takes the average of the sets.
In this observation method, the instrument’s telescope is also always
turned over — both axes by 200g — so one measures in both faces, face kojeasennot I
left and face right. ja II
The following errors are eliminated by observing in two faces: collima-

tion error and trunnion-axis tilt, axes intersection error, eccentricities of
horizontal circle and measuring telescope. In the method of complete sets,
the impact of horizontal-circle graduation errors is essentially diminished.
Levelling or centring errors are not eliminated! tasaus ja
keskistys
Rejection of observations:
◦ If, in the method of complete sets, the sets do not agree with each
other, they must all be measured again. One may not reject only a
3 The locking screw of the horizontal circle is protected by a lid. In that way, the circle is
not turned by accident during measurement.
í Õ ! ¤. û
63.123 135.643 360.446 32.962

2 sets
3 directions
100 0
2 faces
X 0 I 200 300 I 100 X

1 2
300 200
Turn Turn
Set 1, circle circle Set 2,
281.595 0 gon Horizontal 100 gon 178.913
face left, face left,
face right circle face right
335.640 232.960
263.127 160.442
300 200
3
200 II 0 100 II 300
100 0
81.594 378.917
F IGURE 6.3. Observation method of complete sets. In this case a measurement

of two complete sets of three directions is made in both instrument
^ faces.
single set, because that would distort the statistical properties of

the observational material.
◦ An individual sighting direction may be rejected from all sets.
One does not apply the method of complete sets with an electronic theodo-
lite — or more precisely, the instrument does so itself internally, see
section 5.7.
^ 6.4 Station adjustment of horizontal angles

asematasoitus Station adjustment, the merging of the several complete sets of horizontal
direction measurements into one optimal solution or measurement set, is
an adjustment problem, although a simple one. Ordinarily it was already
solved at the observation-notebook stage.
Modern electronic tacheometers or total stations do not need station
adjustment as they do not collect multiple observation sets in the way
described in section 6.3. We describe it here nevertheless briefly for
historical and methodological interest. There will be a more thorough
discussion of adjustment calculus in chapters 14 and 15.
í Õ ! ¤. û
Station adjustment of horizontal angles 6.4
163
^ 6.4.1 Observation equations
In the following calculation example, four complete sets to three sighting
directions have been measured. The two instrument faces for each series,
face left and face right, have already been merged. In the example, the
number of complete sets is s = 4, the number of directions r = 3, so the
number of observations is n = rs = 12. The number of unknown directions
in the complete-set average is r − 1 = 2. In each complete set there is
one orientation unknown4 also to be estimated. The number of excess 4
observations (redundancy, number of degrees of freedom) is thus
b = rs − ( r − 1 + s) = ( r − 1)( s − 1) .
The observation equations are
ˆ︁j − Ω
θ i j + vi j = α ˆ︁ i , (6.1)
in which i = 1, . . . , 4 — generally i = 1, . . . , s — is the number of the

complete set, j = 1, . . . , 3 — generally j = 1, . . . , r — is the number of the
sighting direction, Ω i the orientation unknown of the horizontal circle for
the complete set, αj the direction unknown of the target (the azimuth of
the measured direction), and θ i j the (raw) direction reading.
We use the “hat” notation for estimators as is standard in statistics,
and we underscore stochastic quantities.
If the vector of observations is
[︂ ]︂T
ℓ= θ 11 θ 12 θ 13 θ 21 θ 22 θ 23 θ 31 θ 32 θ 33 θ 41 θ 42 θ 43
and the vector of unknowns

[︂ ]︂T
x=
ˆ︁ α
ˆ︁1 α
ˆ︁2 α
ˆ︁3 Ω
ˆ︁ 1 Ω
ˆ︁ 2 Ω
ˆ︁ 3 Ω
ˆ︁ 4 ,
then the design matrix is, for the case of four complete sets and three rakennematriisi
4 The orientation unknown captures the reality that a theodolite cannot measure absolute
horizontal directions or azimuths. It can measure only direction differences, that is,
angles. Therefore the unknown azimuth of the zero mark on the instrument’s horizontal
circle for every measured series is added as an unknown to the adjustment problem.
The situation is similar to that in GPS measurement. There, too, all observations from
a receiver contain a common clock-error unknown ∆T , which is why the observable
is called pseudo-range and not range. Similarly we could talk here about “pseudo-
direction observations”. These uninteresting additional unknowns are called nuisance
parameters.
í Õ ! ¤. û
sighting directions,
⎡ ⎤
1 −1
⎢ 1 −1 ⎥
1 −1
⎢ ⎥
⎢ ⎥
⎢ 1 −1
⎢ ⎥
⎥
1 −1
⎢ ⎥
⎢ ⎥
1 −1
⎢ ⎥
A=⎢ ⎥,
⎢ ⎥
⎢ 1 −1 ⎥
1 −1
⎢ ⎥
⎢ ⎥
⎢
⎢ 1 −1 ⎥
⎥
⎢ 1 −1 ⎥
⎢ ⎥
⎣ 1 −1 ⎦
1 −1
bringing the observation equations into the standard form
ℓ + v = A ˆ︁
x,
jäännös- with v the vector of residuals.

virheiden
In order to get a unique solution, one must also fix one direction, or
vektori
linear combination of directions — one cannot solve absolute directions
from relative angle measurements alone. So, in addition is required, say,
α1 = 0. (6.2)
This is totally arbitrary. One could choose just as well α2 = 0, α3 = 0, or

why not α1 + α2 + α3 = 0, which would at least be “democratic”.
In the following we present, without formal proof, how one may calcu-
late the result of a station adjustment, and how the calculation can be
arranged in a simple template. The procedure is statistically optimal.
^ 6.4.2 Angle transformation

First, we carry out the angle transformation: we subtract from every
observation θ i j which is not the first of a complete set, precisely that first
observation of that set θ i1 :
def
θ ′i j = θ i j − θ i1 , i = 1, . . . , s, j = 2, . . . , r.
Subtract the rows of the system of observation equations 6.1 from each
other, yielding
ℓ′ + v′ = A ′ ˆ︁
x
in which
[︂ ]︂T
ℓ′ = θ ′12 θ ′13 θ ′22 θ ′23 θ ′32 θ ′33 θ ′42 θ ′43
í Õ ! ¤. û
165
and
−1 1 0 −1 1
⎡ ⎤ ⎡ ⎤
⎢ −1 1 0 ⎥ ⎢ −1 1 ⎥
−1 1 0 −1 1
⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥
−1 1 0 −1 1 ⎥
⎢ ⎥ ⎢ ⎥
′ ′
A =⎢ ⎥ =⇒ A = ⎢ ⎥,
⎢ ⎥ ⎢
⎢ −1 1 0 ⎥ ⎢ −1 1 ⎥
−1 1 0 −1 1 ⎥
⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢
⎣ −1 1 0 ⎦ ⎣ −1 1 ⎦
−1 1 0 −1 1
if we shorten the vector of unknowns accordingly:

]︁T T
x=[ α
ˆ︁1 α
ˆ︁2 α
[︁
x=
ˆ︁ α
ˆ︁1 α
ˆ︁2 α
ˆ︁3 Ω
ˆ︁ 1 Ω
ˆ︁ 2 Ω
ˆ︁ 3 Ω
ˆ︁ 4 =⇒ ˆ︁ ˆ︁3 ] .
Now we may define the following implicit unknowns:
def
ˆ︁′j = α
α ˆ︁j − α
ˆ︁1 , j = 2, . . . , r,
which corresponds to further simplifying the design matrix as follows:
−1 1 1
⎡ ⎤ ⎡ ⎤
⎢ −1 1 ⎥ ⎢ 1 ⎥
−1 1 1
⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥
−1 1 ⎥ 1
⎢ ⎥ ⎢ ⎥
′ ′′
A =⎢ ⎥ =⇒ A = ⎢
⎢ ⎢ ⎥
−1 1 ⎢ 1
⎥
⎢ ⎥ ⎥
−1 1 ⎥ 1 ⎥
⎢ ⎥ ⎢ ⎥
⎢ ⎢
⎣ −1 1 ⎦ ⎣ 1 ⎦
−1 1 1
if the new unknowns are

[︂ ]︂T
′
x =
ˆ︁ α
ˆ︁′2 α
ˆ︁′3 .
These new unknowns are now in relation to the first direction, they are
thus angle-transformed direction unknowns.
The observation equations are now
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
θ ′12 v′12 1
θ ′13 ⎥ v′13 ⎥ 1 ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢ ⎢ ⎢
θ ′22 v′22
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥ ⎢ 1 ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ [︄ ]︄
⎢ θ ′23 ⎥ ⎢ v′23 ⎥ ⎢ 1 ⎥ α ˆ︁′
2
⎥+⎢ ⎥=⎢
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
θ ′32 v′32 ⎥ α ′
⎢ ⎥
⎥ ⎢ 1
⎥ ˆ︁3
⎢ ⎥ ⎢
⎢ ⎥ ⎢ ⎥ ⎢
⎢
⎢ θ ′33 ⎥ ⎢
⎥ ⎢ v′33 ⎥ ⎢
⎥ ⎢ 1 ⎥
⎥
θ ′42 v′42
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎣ ⎦ ⎣ ⎦ ⎣ 1 ⎦
θ ′43 v′43 1
í Õ ! ¤. û
or
[︄ ]︄ [︄ ]︄ [︄ ]︄[︄ ]︄ [︄ ]︄
θ ′i2 v′i2 1 α
ˆ︁′2 α
ˆ︁′2
+ = = , i = 1, . . . , 4. (6.3)
θ ′i3 v′i3 1 α
ˆ︁′3 α
ˆ︁′3
The optimal solution is the average, see equation 14.6. For the general
case
s
1 ∑︂
ˆ︁′j = s
α θ ′i j , j = 2, . . . , r.
i =1
The adjusted angle transformation is the average over complete

sets of the angle transformations.
Furthermore we may define consistently

s s
def 1
∑︂ def 1 ∑︂
ˆ︁′1 = s
α θ ′i1 = s (θ i1 − θ i1 ) = 0.
i =1 i =1
tasoitus The α′ values are adjusted but still relative directions, not absolute
azimuths.
We have eliminated the orientation unknowns Ω i from the observation
equations, and good riddance. We are however interested in the residuals.
^ 6.4.3 Residuals and degrees of freedom

Now that the unknowns α
ˆ︁′j have been calculated, more precisely esti-
pienimmän mated, with the least-squares method, we can also calculate the residuals.
neliösumman Equation 6.3 gives us residuals of sorts:
menetelmä
jäännösvirhe v′i j = α
ˆ︁′j − θ ′i j , i = 1, . . . , s, j = 2, . . . , r.
These have the problem that direction 1 is not included: if it were, its
residual would always be
v′i1 = α
ˆ︁′1 − θ ′i1 = 0 − 0 = 0.
The proper residuals are obtained by

r
1 ∑︂ ′
v i j = v′i j − r v ik , i = 1, . . . , s, j = 1, . . . , r, (6.4)
k=1
of which the average over directions, including direction 1, for each set
vanishes. Calculation 6.4 shows what happens to the residuals when
í Õ ! ¤. û
167
every complete set i = 1, . . . , s is assigned its own orientation unknown
Ω i , minimising the sum of squared residuals within that set.
Let the mean error σ of the direction observations θ i j be the same for
all, and let the direction observations be uncorrelated. In that case the
variance matrix of the observations is
Σℓℓ = Var {ℓ} = σ2 I s×r ,
in which I s×r is the unit matrix of size ( s × r ) × ( s × r ).

In this case, the quantity (“shifting variate”, Baarda, 1968) siirtosuure
s r
1 T −1 1 ∑︂ ∑︂ 2 1 ∑︂ 2
E = v Σ ℓℓ v = v ij = vi j
σ2 σ2 σ2
i =1 j =1 i, j
is distributed according to the χ2b distribution, the expectancy of which is odotusarvo

b, the number of degrees of freedom, or “excess” observations. See section
15.4.
Earlier we saw that b = ( r − 1)( s − 1), the number of angle transforma-
tions (independent direction measurements) multiplied by the number of
excess complete sets. Thus, E σ2 is distributed according to χ2(r−1)(s−1)
/︁
and its expectancy is ( r − 1)( s − 1). So:

2
∑︁
i, j v i j
{︃ }︃ {︃ }︃
1 ∑︂
E v2i j = ( r − 1)( s − 1) =⇒ σ = E 2
,
σ2 (r − 1)(s − 1)
i, j
and we see that

s ∑︂
r
ˆ︂2 def 1 ∑︂
σ = v2i j
(r − 1)(s − 1)
i =1 j =1
is an unbiased estimator of the variance σ2 of a single observation: harhaton

estimaattori
ˆ︂2 = σ2 .
E σ
{︁ }︁
For the quantity σ we use the name “mean error of unit weight”. It painoyksikön
represents the uncertainty, the mean error, of a typical observation, in keskivirhe
this case, a single, raw direction reading.
We shall return to the subject of adjustment in chapter 14.
^ 6.4.4 Calculation table for station adjustment

A station adjustment consists of the following steps: asematasoitus
í Õ ! ¤. û
1. We fix the arbitrariness in the definition of the directions, like by

fixing the first direction to zero, θ ′i1 = 0; this is called the angle
transformation. We do it by subtracting the first observed direction
reading from the other direction readings in the same complete set.
sarjakeskiarvo 2. We compute the average over all complete sets (“complete-set aver-
age”)
s
1 ∑︂ ′
s θi j, j = 1, . . . , r
i =1
for every sighting direction j .

3. We compute the residuals of the observations relative to these
complete-set averages:
s
1 ∑︂ ′
v′i j = θ ′i j − s θk j, i = 1, . . . , s, j = 1, . . . , r.
k=1
4. We compute the averages of the “first residuals” v′i j over all direc-
tions within every complete set:
r r s
(︄ )︄
1 ∑︂ ′ 1 ∑︂ ′ 1 ∑︂ ′
r v ij = r θ ij − s θk j , i = 1, . . . , s.
j =1 j =1 k=1
5. We compute the final residuals

r
1 ∑︂ ′
v i j = v′i j − r v ik , i = 1, . . . , s, j = 1, . . . , r,
k=1
which corresponds precisely to those residuals we would obtain

if we gave each complete set its own orientation unknown Ω
ˆ︁ i in
addition to the direction unknowns α
ˆ︁j , like in the original system
of observation equations.
The end result is the calculation template presented in table 6.1, which
contains these operations and ought to be self evident.
After filling in the template — in the order: average over complete sets,
′
v , sum per series, v — the mean error of a single direction, or unit weight,
is calculated by first calculating the quadratic sum of the residuals over
all series and directions:
s ∑︂
∑︂ r
vT v = v2i j = vv
[︁ ]︁
i =1 j =1
í Õ ! ¤. û
169
^ T ABLE 6.1. Calculation template table for station adjustment.
Angle transformation (gon) v′ (cc) v (cc)
1. Complete set 0.00 00 0 −3.7

68.84 30 +4 +0.3
209.18 80 +7 +3.3
Complete set sum, average +11 +3.7
2. Complete set 0.00 00 0 −1.3

68.84 25 −1 −2.3
209.18 78 +5 +3.7
Complete set sum, average +4 +1.3
3. Complete set 0.00 00 0 +2.3

68.84 24 −2 +0.3
209.18 68 −5 −2.7
Complete set sum, average -7 −2.3
4. Complete set 0.00 00 0 +2.7

68.84 25 −1 +1.7
209.18 66 −7 −4.3
Complete set sum, average -8 −2.7
Average over complete sets 0.00 00

(“complete-set average”) 68.84 26
209.18 73
(the latter notation was invented by Gauss and is sometimes still used).
Then, we calculate, or estimate5 5
√︂ √︄ √︄
vT v vT v
σ
ˆ︁ = σ
ˆ︂2 = = = 3.8cc .
rs − ( r + s) + 1 (r − 1)(s − 1)
Here
◦ rs is the number of observations ( r directions in s complete sets).
◦ r + s is the number of unknowns ( r direction unknowns α j , s orien-
tation unknowns Ω i ).
◦ The number +1 represents the defect of the problem, the dimen-
sion of the solution space: the circumstance mentioned above, that
5 Note ˆ︂2 = σ2 , however E σ

{︁ }︁ {︁ }︁
that E σ ˆ︁ ̸= σ! So σ
ˆ︁ is not unbiased although σ
ˆ︂2 is. The
reason for this is that the E operator and the square root may not be interchanged: in
{︁⎷ }︁ √︂ {︁ }︁
general, E x ̸= E x .
Counter example: if the probability distribution of x is p(1) = 0.5, p(4) = 0.5, and
{︁⎷ }︁ √︁ ⎷
elsewhere p( x) = 0, we obtain E x = 1.5 but E { x} = 2.5 ≈ 1.58.
í Õ ! ¤. û
adding an amount ∆ to both the directions α j and the orientations

Ω i does not affect the observations in any way that can be observed,
and that thus, the solution is not unique without one extra condi-
6 tion,6 for example, fixing α1 to some value, such as zero.
The number b = rs − ( r + s) + 1 = ( r − 1)( s − 1) is called the number of
degrees of freedom.
After this, the estimate of the mean error of the average taken over all
ˆ︁ ⎷s = 1.9cc .
ˆ︁ s = σ
/︁
complete sets is σ
^ 6.5 Traverse measurement and computation

A traditional method for densifying lower-order base networks has been
jonomittaus traverse measurement or traversing. In spite of the existence of satellite
tosiaikainen positioning, and especially real-time kinematic positioning (RTK), there
continue to be situations where traversing is the best, or even the only,
method of base-network measurement. Such situations include indoors
measurements, underground — mine or tunnel — measurements, and
measurements in high-rise urban landscapes, “urban canyons”, where
GNSS measurement has problems due to the blocking of signal by build-
monitie ings, and spurious signal reflections off buildings, the multipath problem.
7 Going back in history, then, in 1807 Nathaniel Bowditch7 won a compe-
tition for the most appropriate way to compute a traverse. The prize was
ten US dollars (Cooper, 1982, pages 147–150).
The variant of the Bowditch method presented here, a traditional
separate adjustment of directions and co-ordinates, is suboptimal. Yet it
is worth looking at as an example of geodetic computation.
monikulmiojono A traverse is a line of measurement stations, where at every station
one measures directions and distances to the previous and next stations.
The instrument used is generally a total station or tacheometer. The
measurement is usually the last stage in bringing co-ordinates ( x, y) to
reference points in the immediate vicinity of objects to be measured.
There are two types of traverses: open and closed. See figure 6.4. We
6 This extra condition can also be understood as a “pseudo-observation”: then, the

number of observations is rs + 1.
7 Nathaniel Bowditch (1773–1838) was an American mathematician, student of naviga-
tion at sea, and scientific translator and textbook author. One of his writings was “The
American Practical Navigator”, published in 1802. The modern edition of the book, “The
Bowditch”, can be found on the Internet: Bowditch (2017).
í Õ ! ¤. û
Open traverse 6.6
171
Closing point Z Closing
auxiliary
point ZZ
n−1 Z
n−1
Closed traverse
Traverse point
Starting
auxiliary
3 Open traverse
point A A 2
2 3
A
Starting point A
^ F IGURE 6.4. An open and a closed traverse.
always know the co-ordinates ( x, y) of the starting and closing points.

The difference between an open and a closed traverse is in the use of
auxiliary points in the latter. These points, the co-ordinates of which are liitospiste
known, help to orient the traverse correctly.
^ 6.6 Open traverse

Generally, we try to measure a closed traverse, because it gives the
possibility to check and adjust both the measured angles and distances.
There are, however, situations where this is impossible or difficult and
where one has to use an incompletely closed traverse. vaillinaisesti
suljettu
The common case is where we know neither the starting nor the closing
direction. See figure 6.5.
We know
◦ ( x A , yA ), starting point
◦ ( x Z , yZ ), closing point.
We need to compute, or estimate,8 co-ordinates for the traverse points 8

(ˆ︁ yi ), i = 2, . . . , n − 1.
x i , ˆ︁
8 Using again the “hat” notation for estimators, as well as underscoring stochastic
quantities.
í Õ ! ¤. û
x′AZ n
−θ
x AZ Z=n
n−1
α′AZ
α AZ
y
A=1
y′AZ yAZ
^ F IGURE 6.5. An open traverse.
^ 6.6.1 Starting direction

Because we do not know the starting or closing directions, we cannot
adjust directions. Computing the direction α12 is not possible as we lack
a starting direction. We may obtain one in two different ways:
◦ by setting the direction angle of the first side α′12 = 0, figure 6.5, or
◦ (preferred) obtaining somehow, for example by measuring in a map,
likiarvo an approximate value α′12 for α12 .
^ 6.6.2 Computing the traverse

We carry out the calculation of traverse directions in the familiar way:
α12 = α′12 ,
α23 = α12 − 200g + ∠α2 ,
..
.
α i,i+1 = α i−1,i − 200g + ∠α i ,
..
.
αn−1,n = αn−2,n−1 − 200g + ∠αn−1 .
geodeettinen Using these, we calculate co-ordinates for the points using the forward
päätehtävä
í Õ ! ¤. û
Open traverse 6.6
173
geodetic problem recursively:
x′2 = x A + s12 cos α12 , y′2 = yA + s12 sin α12 ,
and so forth:
x′i = x′i−1 + s i−1,i cos α i−1,i , y′i = y′i−1 + s i−1,i sin α i−1,i .
Finally
x′n = x′n−1 + s n−1,n cos αn−1,n , y′n = y′n−1 + s n−1,n sin αn−1,n .
This calculation has been carried out in the correct way, but in a wrongly
oriented co-ordinate frame. The closing errors
def def
w x = x′n − x Z , w y = y′n − yZ
do not tell us anything about measurement errors, but rather about the
approximativeness of the assumed starting direction α′12 . For this reason,
it is not permissible to eliminate the closing errors by adjustment as is
done in a closed traverse.
The whole traverse is rotated by an angle amount θ , and should be
rotated back by a Helmert or similarity transformation.
Because starting point A is the common turning point between both co-
ordinate systems, we may simply calculate the scale ratio K and rotation kiertokulma
angle θ between them:
√︂
∆ xAZ
2
+ ∆ yAZ
2
K = √︂ )︁2 ,
(∆ x′AZ )2 + ∆ y′AZ
(︁
∆ y′AZ ∆ y AZ
θ = arctan − arctan
∆ x′AZ ∆ x AZ
— or with the half-angle formula, with the definition
def y
y // x = √︁ :
x+ x2 + y2
(︂ )︁)︂
∆ y′AZ // ∆ x′AZ − arctan ∆ y AZ // ∆ x AZ
(︁ )︁ (︁
θ = 2 arctan .
Also used is the notation
∆ x AZ = x Z − x A , ∆ yAZ = yZ − yA ,
∆ x′AZ = x′n − x A , ∆ y′AZ = y′n − yA .
í Õ ! ¤. û
Let us construct a transformation:
c = K cos θ , s = K sin θ ,
with the help of which we obtain as the co-ordinates of the traverse points
i = 1, . . . , n:
x i = x A + c ( x′i − x A ) + s y′i − yA ,
(︁ )︁
ˆ︁
yi = yA − s ( x′i − x A ) + c y′i − yA .
(︁ )︁
ˆ︁
In matric form
[︄ ]︄ [︄ ]︄ [︄ ]︄[︄ ]︄
xi
ˆ︁ xA c s x′i − x A
= + .
yi
ˆ︁ yA −s c y′i − yA
Note that in the matrix

[︄ ]︄ [︄ ]︄
c s cos θ sin θ
M= =K
−s c − sin θ cos θ
the direction correction θ is dominated by the assumed starting direction,

when again the scale correction K contains only a correction for the
9 closing error caused by the imprecision of the measurements:9 even if
the starting direction α12 were guessed exactly correctly, nevertheless

(︁ )︁
x′n , y′n would not necessarily coincide precisely with point ( x Z , yZ ). In
the sideways direction however, perpendicular to the traverse, there will
remain observation error, which is not adjusted away, but more precisely
is “absorbed” into θ .
A useful final sanity check is
? ?
x Z = ˆ︁
xn , yZ = ˆ︁
yn .
In addition to this, one should check that K is realistically valued, mean-

ing close enough to the value 1 in view of the known precision of distance
measurement.
^ 6.7 Closed traverse

We know
alkuliitospiste ◦ ( x A , yA ) and ( x A A , yA A ): starting point A and auxiliary start-
ing point A A ,
loppuliitospiste ◦ ( x Z , yZ ) and ( x ZZ , yZZ ): closing point Z and auxiliary closing
point ZZ .
9 . . . in other words, it is a primitive adjustment.
í Õ ! ¤. û
Closed traverse 6.7
175
∠α n
∠αn−1
α′ZZ
( x Z , yZ )
( x A A , yA A ) α ZZ
∠α2 ( x ZZ , yZZ )
X
∠α1
g
α A A − 200
( x A , yA )
F IGURE 6.6. The geometry of a closed traverse. Only angles and starting and
^ closing directions are marked.
See figure 6.6.

We observe
◦ n bending angles ∠α1 , . . . , ∠αn ,
◦ n − 1 distances or side lengths s12 , s23 , . . . , s n−1,n .
We must compute, or estimate
the co-ordinates (ˆ︁ yi ) for the traverse points i = 2, . . . , n − 1. jonopiste
x i , ˆ︁
There are n − 2 new points.
^ 6.7.1 Computing the auxiliary starting and closing directions

The auxiliary starting and closing directions are obtained by solving the alku- ja loppu-
inverse geodetic problem: liitossuunta
geodeettinen
y −y
α A A = arctan x A A − x A + k · 200g , käänteistehtävä
AA A
yZZ − yZ (6.5)
α ZZ = arctan x − x + k · 200g ,
ZZ Z
where k ∈ {0, 1} needs to be chosen so the result is in the correct quadrant.

The approximate Bowditch-like method that we shall use has two
stages. First we process the directions, then the co-ordinates.
^ 6.7.2 Adjustment of directions

Compute the directions or azimuths α A A , α12 , . . . , αn−1,n , αn,n+1 , in which
α A A is known, and αn,n+1 , which is computed from the observations,
corresponds to the above computed auxiliary closing direction α ZZ . As
follows:10 10
í Õ ! ¤. û
α12 = α A A + ∠α1 ,
α23 = α12 − 200g + ∠α2 ,
..
.
α i,i+1 = α i−1,i − 200g + ∠α i , (6.6)
..
.
αn−1,n = αn−2,n−1 − 200g + ∠αn−1 ,
def
α′ZZ = αn,n+1 = αn−1,n − 200g + ∠αn .
At the end we thus obtain the value α′ZZ , which, if all the angle observa-
tions α i were errorless, would be equal to α ZZ computed from co-ordinates
by equation 6.5. In reality though, the direction closing error amounts to
def
wα = α′ ZZ − α ZZ .
This closing error is removed, or adjusted, by dividing it evenly among

all the bending angles, so to every measured bending angle we apply a
correction:
w
δα = − nα ,
after which we again carry out the calculation of angles, producing the
adjusted directions:
α
ˆ︁12 = α A A + (∠α1 + δα) ,
ˆ︁23 = α12 − 200g + (∠α2 + δα) ,
α
..
.
ˆ︁i,i+1 = α i−1,i − 200g + (∠α i + δα) ,
α
..
.
ˆ︁n−1,n = αn−2,n−1 − 200g + (∠αn−1 + δα) ,
α
ˆ︁′ZZ = αn−1,n − 200g + (∠αn + δα) .
α
^ 6.7.3 Co-ordinate adjustment

Using the thusly adjusted directions we compute, for the points 2, . . . , n − 1,
co-ordinates by applying the forward geodetic problem:
◦ Point 2:
∆ x12 = s12 cos α

ˆ︁12 , ∆ y12 = s12 sin α
ˆ︁12 ,
10 For the first bending angle ∠α1 we do not subtract 200 gon! And to keep the directions
α i,i+1 in the interval 0, 400g , one may sometimes have to add or subtract 400 gon.
[︁ )︁
í Õ ! ¤. û
Closed traverse 6.7
177
with the aid of which
x2 = x A + ∆ x12 , y2 = yA + ∆ y12 .
◦ General point i (and also closing point i → n):
∆ x i−1,i = s i−1,i cos α

ˆ︁i−1,i , ∆ y i−1,i = s i−1,i sin α
ˆ︁i−1,i ,
with the aid of which
x i = x i−1 + ∆ x i−1,i = x i−1 + s i−1,i cos α

ˆ︁i−1,i ,
y i = y i−1 + ∆ y i−1,i = y i−1 + s i−1,i sin α
ˆ︁i−1,i .
◦ Generally — remember ( x A , yA ) = ( x1 , y1 ):
k
∑︂ k
∑︂
xk = x A + s i−1,i cos α
ˆ︁i−1,i , y k = yA + s i−1,i sin α
ˆ︁i−1,i .
i =2 i =2
By substitution of k → n we obtain the equations and co-ordinates

(︁ )︁
x n , yn of the closing point. If the observations were errorless, we would
have x n = x Z and yn = yZ , but they are not. The co-ordinate closing errors
are
def def
w x = xn − xZ , w y = yn − yZ .
Closing errors are adjusted by giving a weight coefficient q i−1,i for each
point interval, and the closing errors w x and w y are distributed over the
point intervals in proportion to these weight coefficients.
Small weight coefficient ←−−→ large weight,

large weight coefficient ←−−→ small weight!
We compute the standard correction corresponding to the sum of weight

coefficients, separately for the x and y co-ordinates:
w wy
δ x = − ∑︁n x , δ y = − ∑︁n .
i =2 q i −1,i i =2 q i −1,i
The adjusted co-ordinates become
x i−1 + s i−1,i cos α

x i = ˆ︁
ˆ︁ ˆ︁i−1,i + q i−1,i δ x,
yi−1 + s i−1,i sin α
yi = ˆ︁
ˆ︁ ˆ︁i−1,i + q i−1,i δ y,
or, counting from starting point A ,
k
∑︂ (︁ )︁
xk = ˆ︁
ˆ︁ xA + s i−1,i cos α
ˆ︁i−1,i + q i−1,i δ x ,
i =2
k
∑︂ (︁ )︁
yk = ˆ︁
ˆ︁ yA + s i−1,i sin α
ˆ︁i−1,i + q i−1,i δ y .
i =2
í Õ ! ¤. û
^ T ABLEAU 6.2. Traverse calculation template, separate adjustment of angles

and co-ordinates (“Bowditch method”).
i ∠α i α i,i+1 δα i,i+1 α
ˆ︁i,i+1
A (= 1) 345.3750
1 212.2345 157.6095 +7 157.6102
2 151.4565 109.0660 +15 109.0675
3 221.9823 131.0483 +22 131.0505
4 175.9831 107.0314 +29 107.0343
5 165.3467 72.3781 +37 72.3818
Z (= 5) 72.3818
Closing error wα −0.0037 ↑
i α
ˆ︁i,i+1 s i,i+1 s cos α
ˆ︁ s sin α
ˆ︁ xi yi δxi δ yi xi
ˆ︁ yi
ˆ︁
1 157.6102 502.345 −395.038 +310.315 1000.235 256.256
2 109.0675 487.241 −69.164 +482.307 605.197 566.571 +9 -37 .206 .534
3 131.0505 445.981 −209.001 +393.977 536.033 1048.878 +18 -73 .051 .805
4 107.0343 512.125 −56.472 +509.002 327.032 1442.855 +26 -110 .058 .745
5 72.3818 270.560 1951.857 +35 -146 .595 .711
Z (= 5) 270.595 1951.711
Closing errors w x , w y −0.035 +0.146 ↑ ↑
^ 6.7.4 Calculation template

The whole calculation may be carried out using the template in tableau
6.2, which is readily automated. Let it be given that α A A = 345.3750 gon
and α ZZ = 72.3818 gon, as well as the co-ordinates of starting point A
and closing point Z . The number of stations n = 5. Weighting used:
uniform. The values for the starting and closing points are underlined,
observations are in black, computed values in red, closing errors and
adjustment corrections in blue. This kind of diagram is good enough for
practical work in local base-network measurement.
Some remarks:
◦ The adjustment method described above is approximate, meaning
pienimmän sub-optimal. In a proper least-squares adjustment, the direction
neliösumman and co-ordinate corrections are computed simultaneously, as they
tasoitus
correlate with each other. The results will then be slightly different.
Nevertheless, if the traverse is relatively straight — the sides are
í Õ ! ¤. û
Zenith angles and refraction 6.8
179
roughly in the same direction — the suboptimality will be small,
and a proper adjustment would only lead to small changes in the
final co-ordinates.
On the other hand, it is not desirable that the sides are in precisely
the same direction: that would make finding gross errors more karkea virhe
difficult.
◦ Here, the calculation of the traverse was done in the plane. The
measurements however have been obtained three-dimensionally,
in space. Of course it is assumed that instrument and signal are tähys
precisely and correctly centred and that the instrument is levelled,
and that all relevant reductions to the observations have been done.
This includes especially the slope reduction, equation 7.9. Also,
depending on size and topography of the measurement area, the
height reduction (same equation 7.9) may be necessary.
◦ In computing plane co-ordinates, directions are always referred to
the map north, and distances are also reduced to the map plane,
meaning that the scale distortion caused by the map projection
used, equation 7.10, is accounted for.
◦ Only then can it be said that
– The horizontal angles are plane angles.
– The reduced distances are horizontal distances in the map
plane.
^ 6.8 Zenith angles and refraction

In the measurement of zenith angles, because of the stratification of the kerrostuneisuus
atmosphere, the overwhelmingly greatest source of error is refraction.
On must always be careful concerning weather and the landscape when
choosing time and place, and when processing observations. For example,
hot asphalt in summer can be treacherous.
The effect of refraction does not show in the check ζ1 + ζ2 = 400, subsec-
tion 5.6.3: measuring in both faces does not help. kojeasennot
^ 6.8.1 Refraction coefficient

In geodesy, the habit has formed to characterise refraction by a quantity
named the refraction coefficient, symbol k. This quantity characterises the refraktiokerroin
curvature of the measurement ray or path in the atmosphere compared
í Õ ! ¤. û
− c2
B B
ζ s
hB
A
ζ c1
A s
hA
. . . centr
rt h’s. . . e
To Ea
(a) (b)
Straight geometry Taking curvatures along
F IGURE 6.7. The effects of refraction and Earth curvature on zenith-angle mea-
^ surement.
to the curvature of the Earth:

ray curvature R
k= = , (6.7)
Earth curvature ρ
in which
radius of curvature of the Earth. 1 R is the curvature of the
/︁
R
Earth’s surface. The curvature is the inverse of the radius of
curvature!
radius of curvature of the measurement ray. 1 ρ is the curvature
/︁
ρ
of the measurement ray.
Typical k values in the atmosphere are k L = 0.13 for visible light, and
k M = 0.25 for microwaves, see Kahmen and Faig (1988) page 167. The
curvature of the measurement ray is thus 4–8 times weaker, and the
radius of curvature 4–8 times larger, than the curvature, respectively
radius of curvature, of the Earth’s surface. However, during an atmo-
spheric inversion, exceptionally large k values may occur, even 0.3–0.4,
see Grafarend et al. (1987).
^ 6.8.2 Trigonometric heighting

In figure 6.7 one sees on the left-hand side, how one may determine
the height difference between two points A and B using zenith-angle
measurement. The applicable trigonometric equation is
h B = hA + s cot ζ,
í Õ ! ¤. û
Zenith angles and refraction 6.8
181
in which s is the horizontal distance — the slant range projected onto the vinoetäisyys
horizontal plane — and ζ the measured zenith angle.
When, in reality, both the surface of the Earth and the measurement
ray’s path in the atmosphere are curved, in practice the right-hand side
figure applies, in which, however, the curvatures have been exaggerated.
Based on the figure, one should add to the above equation two correction
terms:
s2
c1 ≈ ,
2R
the correction due to the Earth’s curvature, and
s2 s2
c2 ≈ − = −k ,
2ρ 2R
the correction due to the curvature of the measurement ray — assuming

that, unlike in the figure, the height difference between the points is
small compared to their distance.
Everything together:
s2
h B = hA + s cot ζ + (1 − k) .
2R
If we still take along the height of the theodolite, or instrument, i over

point A and the height of the signal or target t over marker B underneath
it, we obtain
s2
h B = hA + s cot ζ + (1 − k) + i − t, (6.8)
2R
the fundamental equation of trigonometric heighting.
^ 6.8.3 Simultaneous measurement in opposite directions

If we carry out zenith-angle measurements simultaneously at points A
and B, we obtain
s2
h B = hA + s cot ζ A + (1 − k) + iA − i B ,
2R
s2
hA = h B + s cot ζB + (1 − k) + i B − iA .
2R
Here, for the measurement done in A , the instrument height is i A and

the signal height i B , and for the measurement done in B, the instrument
height is i B and the signal height i A . So, we assume the heights of
instrument and signal in the opposite-direction measurements to be the
same at the same point — easy to do with forced centring, subsection pakkokeskistys
5.4.7.
í Õ ! ¤. û
ζ e,2 ζ t,2
ζ t,1 s2
ζ e,1 s1
∆H
^ F IGURE 6.8. Trigonometric levelling traverse.
Rearranging the terms and subtraction yields
h B = hA + 21 s (cot ζ A − cot ζB ) + i A − i B , (6.9)
from which the term describing atmospheric refraction has vanished.

This method for determining height differences between points has
been found to be good and has also been much-used over long distances.
It assumes the distance s to be known with sufficient precision.
This method is used in trigonometric levelling (figure 6.8). The method
can replace traditional levelling, for example, in terrain with great height
variations, where the staff distance of levelling would become short and
takymetri the work laborious. In the method, two total stations and two signal-
reflector assemblies are used, and the observation data is transmitted
by radio modems from one instrument to the other for processing, error
control and storage. Movement to the next point is by car where the
terrain allows (Takalo, 1995).
^ 6.8.4 Ölander’s method

11 As early as in the 1930s, V. R. Ölander11 used in the Finnish primary
triangulation a method for refraction modelling which was based on

approximately simultaneous measurements from each triangulation point
to all neighbouring points. This is thus a method essentially different
from that of simultaneous measurements in opposite directions.
Ölander assigned every triangulation point in the network its own
refraction coefficient as an unknown, which were all solved for by means
verkkotasoitus of network adjustment (Ölander, 1932, see also Grafarend et al., 1987).
11 Victor Rafael Ölander (1897–1973) was a Finnish geodesist and astronomer who
played a central role in the Finnish national primary triangulation.
í Õ ! ¤. û
Heights of instrument and signal 6.9
183
ζ
i
P Terrain t
Q
HP
Geoid, “mean sea surface” HQ
^ F IGURE 6.9. Heights of instrument and signal.
^ 6.9 Heights of instrument and signal

Even though the instrument is levelled and centred on the point, it is
nevertheless always eccentrically set up in the height direction.
The zenith angle to be measured is the angle between the plumb line luotiviiva
and the sight axis of the telescope, figure 6.9. Thus, one must measure tähtäysakseli
the height of the instrument i . Generally the aiming is at a signal on a
tripod, not the point marker (monumented point) itself, so the signal or
target height t must also be measured. So, one must always measure
◦ at the instrument, the height difference between the marker or
monumented point, and the horizontal instrument axis
◦ at the signal, the height difference between the marker (monument)
and the signal target point.
In the depicted situation (figure 6.9), the aiming point is the upper
edge of the white triangle, and it is the height t of that point that must
be measured.
In the height direction, both the instrument and the signal are eccen-
trically set up.
The figure also includes the heights of the monumented points from a
computational reference surface (“sea level”), which the measurements
aim to determine:
◦ height of the instrument from the reference level: HP + i
◦ height of the signal or target from the reference level: HQ + t.
í Õ ! ¤. û
1. What three basic topographic surveying measurement types use
horizontal angle measurement?
2. What is station adjustment? What is an orientation unknown?
3. What is the difference between an open and a closed traverse? How
many closing errors (redundancy, degrees of freedom) are there in
each, which can be used to check for mistakes, or gross errors?
4. Why is traversing not entirely obsolete in spite of the existence of
GNSS?
5. Why is atmospheric refraction a particular concern when measuring

vertical angles, and not so much when measuring horizontal angles?
Could you think of situations where horizontal-angle measurement
could also suffer from refraction?
6. What is the refraction coefficient k?
7. What is trigonometric levelling, and in what situations is it used?
How is the effect of atmospheric refraction eliminated?
8. What was Ölander’s refraction modelling method like?
í Õ ! ¤. û
^ Distance measurement
7
BERGSTRAND, ERIK Ö, observator, Danderyd, f i
Upps 3/7/04 av prof Östen B o Anna Ericsson.
Fil kand 35, fil dr 50, observator Rikets allm kartverk sed
56. – Delt i solförmörkelseexp t Afrika 47. – Utg drsavh om
ljushastigh, utv ny metod f avst:bestämn (geodimetern).
Gift 38 m folkskollär Lisa Torpson f 99, dtr t sem:rektor
Nils T o Klara Svensson.
Erik Östen Bergstrand (1904 – 1987), inventor of the

Geodimeter, Who is Who, Harnesk and Davidsson (1962)
^ 7.1 Mechanical distance measurement

The SI unit of distance, or rather, of length, is the metre, see section 2.1.
Traceability to the metre standard is of central importance in precise jäljitettävyys
distance measurement.
Although today, even short distances are measured electronically or
electro-optically, it is good to understand the oldest1 and technically 1
simplest length measurement method, tape measurement. It continues
to be used in local measurements, when distances are short and the
precision afforded by tape measurement suffices.2 And of course the 2
1 Back in the day, the scale of the Finnish primary triangulation was transferred from
the Nummela baseline to the baselines of the twelve principal sides of the network using
invar wires. Invar is an iron-nickel alloy (64 % iron, 36 % nickel) which has a very small
coefficient of thermal expansion. In the Lapland grade measurement by Maupertuis,
the scale of the triangle network was also obtained mechanically from a baseline —
during winter 1736 on the ice of the Torne river!
2 Careful tape measurement is surprisingly precise!
– 185 –
186 7 D ISTANCE MEASUREMENT
F Chord F
s Arc
^ F IGURE 7.1. Sag correction of measuring tape.
equipment required is inexpensive!

In tape measurement, four corrections must be taken into account.
The tape correction is a tape-specific reduction. It is determined by
calibration in a comparator, the true length of which is known
jäljitettävyys- precisely — thanks to comparison, through the traceability chain,
ketju with the standard metre. The tape correction ∆ℓ0 is now the
difference between the comparator’s true length ℓ0 + ∆ℓc and its
3 length ℓ0 + ∆ℓm measured by the tape to be calibrated:3
∆ℓ0 = ∆ℓc − ∆ℓm .
Here, ℓ0 is the nominal length of the tape, for example ℓ0 = 30 m.

The temperature correction is caused by the thermal expansion of
steel, and thus requires the temperature of the tape to be mea-
sured. The temperature correction is assumed zero for a standard
temperature of t 0 = 20◦ C. If the coefficient of thermal expansion
of steel is α, the temperature correction is
∆ℓt = αℓ0 ( t − t 0 ) .
Here, α is expressed in micrometres per metre and degree. For

example, the coefficient of thermal expansion of a certain steel
−1
alloy is 11.34 µm /m◦ C, or α = 11.34 · 10−6 (◦ C) , because µm /m = 10−6 .
If a measuring tape is 30 metres long and the temperature is 28 ◦ C,
it follows that
∆ℓt = 2.7 mm.
painumis- Sag correction During measurement, the tape is tensioned with a

korjaus
3 Question: why do we use the algebraic sign like this? Why not
∆ℓ0 = ∆ℓm − ∆ℓc ?
í Õ ! ¤. û
Mechanical distance measurement 7.1
187
known force F . No matter how strong this force is, the tape will
always settle into a catenary, in fact a cosinus hyperbolicus (cosh ketjukäyrä
function) figure. We thus need to measure the tensioning force of
the tape, figure 7.1. The theory of the phenomenon is surprisingly
complicated. The end result, the sag correction, the difference in
length between the chord and arc, is jänne
µ2 3
∆ℓs = − ℓ ,
24F 2
proportional to the length of the tape to the third power. In this
equation, µ is the weight of the tape per metre, and F the tension-
ing force of the tape. Alternatively one can measure the sag s in
the middle and use the equation
8 s2
∆ℓs = − .
3ℓ
The slope correction is actually not a correction but rather a reduc-
tion, which is necessary if, instead of the slope distance or slant vinoetäisyys
range, one wishes to obtain the horizontal distance between the
points, the projection of the slant range onto the horizontal plane.
In routine tape measurements, the tape correction, temperature correc-

tion and sag correction are often ignored. Their detailed description with
formulas is found in the literature, for example Kahmen and Faig (1988)
pages 122–130. The slope reduction can however be significant: if it is
given that the height difference of the end points of a distance ℓ is ∆ h,
then the horizontal distance is, according to Pythagoras’ theorem,
√︁
ℓ⊥ = ℓ2 − ∆ h 2 . (7.1)
If we only know the slope angle α, sin α = ∆ h ℓ , between end points, we

/︁
obtain the horizontal distance as follows, figure 7.2:

√︁
ℓ⊥ = ℓ2 − ℓ2 sin2 α = ℓ cos α.
Equation 7.1 may be written with sufficient precision for small slopes
√︃ √︃
)︂2
∆h 1 κ2
)︂2
κ
(︂ (︂
def
ℓ⊥ = ℓ 1− ≈ℓ 1− ≈ ℓ− ℓ = ℓ + ∆ℓ⊥ ,
ℓ 100 2 10 000
in which the slope reduction
ℓκ2
∆ℓ⊥ = − (7.2)
20 000
is presented as a function of the slope percentage κ = 100 ∆ h ℓ⊥ ≈
def /︁
100 ∆ h ℓ .
/︁
í Õ ! ¤. û
ℓ
∆h
α ℓ⊥
^ F IGURE 7.2. slope reduction of slant ranges.
^ 7.2 Electromagnetic radiation

Light, radio waves and many other forms of radiation are examples of
electromagnetic radiation.
In physics, it has been long considered whether visible light is a wave
motion (Huygens) or a stream of particles (Newton). The finding of inter-
ference phenomena resolved the controversy in favour of the wave-motion
theory. This was also decisively helped by the development, by James
4 Clerk Maxwell,4 of the field theory of electromagnetism, which, using
partial differential equations, presents electromagnetic waves as a natu-

ral wave phenomenon occurring in this field. Maxwell even succeeded in
calculating theoretically the propagation speed c, which was close to the
already observed speed of light. . .
vaihe Being a wave motion, electromagnetic radiation has a phase φ. When
we describe a wave motion as the projection of a uniform circular motion
5 onto one dimension5 (figure 7.3), then φ is the angle at the centre of
φ
−→ t
F IGURE 7.3. The phase φ of a wave motion as a function of time t. The name
^ “phase” probably originates from the Moon’s phases.
4 James Clerk Maxwell FRS FRSE (1831–1879) was a Scottish mathematical physicist
who set the theory of the electromagnetic field on a mathematical footing, and also
contributed essentially to theoretical thermodynamics.
5 Equivalently: as the real part of the complex wave function exp( i φ) = cos φ + i sin φ.
í Õ ! ¤. û
Electromagnetic radiation 7.2
189
that circle that measures this uniform motion. The relationship between
phase φ and frequency f is
φ( t) = φ( t 0 ) + 2π f ( t − t 0 ),
in which t is time and t 0 reference time. Clearly, the phase is periodic and
repeats after one cycle, or 2π. Therefore we may always reduce the phase
[︁ )︁
angle to the interval 0, 2π without changing its physical meaning.
Nowadays one can measure the wavelengths and frequencies of the
various forms of electromagnetic radiation very precisely; between them
there is the relationship
λ f = c,
in which f is the frequency and λ the wavelength. The quantity c is the

speed of light (in vacuum), which according to Einstein — or actually
already Maxwell — is a constant of nature. See figure 7.4.
However, quantum theory has made the particle model relevant again.
Light can be described as a stream of particles, photons, the energy of
each of which is
E = hf ,
in which h is Planck’s6 constant. The particle model is especially fertile 6

for high energy levels, the left side of figure 7.4.
The electromagnetic field is a vector field. Therefore electromagnetic
radiation is a transversal wave motion, and may be polarised.7 Figure 7.5 poikittainen
aaltoliike
6 Max Karl Ernst Ludwig Planck (1858–1947) was a German physicist and organiser of 7
German physics. He is remembered for his discovery that the thermal radiation spec-
trum of a black body is caused in a natural way by the quantisation of electromagnetic
radiation.
7 The correct understanding of polarisation was the achievement of Thomas Young
(1773–1829) and Augustin-Jean Fresnel (1788–1827). In studying polarisation, a cer-
tain mineral, clear calcite (CaCO3 ), Iceland spar, had a central role. The crystal is
birefringent (“doubly refracting”) and splits light into two parts according to their direc-
tion of polarisation. Apparently the Vikings used it in navigating by the Sun in overcast
weather. The Dutch Christiaan Huygens (1629–1695), the father of the wave theory of
light, spent ample time on the experimental study of Iceland spar.
Another background story links polarisation to the chemistry of life. Chiral (non-mirror-
symmetric, “handed”) molecules, like sugar, rotate the plane of polarisation of light
passing through them: optical activity. Already Louis Pasteur (1822–1895) was studying
the background of the optical activity of organic tartaric acid, and today we know that
the phenomenon is linked to the chirality of life itself, its “handedness”, like the winding
direction of the helix of the DNA molecule. The polarimeter is a vital tool in the medical
í Õ ! ¤. û
190
7
30 GHz 300 MHz 3 MHz 30 kHz
f
10 nm 1 µm 100 µm 1 cm 1m 100 m 10 km
í
λ= c f
/︁
Sun Earth Cosmic background

Õ ! ¤. û
Pound and
Pro- Elec- H ionisation CO2 GPS L 1 L 2 IBM PC
X Rebka (1959)
ton tron Hydrogen
mass γ mass X-rays Ultraviolet Infrared Microwaves 21 cm Radio
124 MeV 1.24 MeV 12.4 keV 124 eV 1.24 eV 12.4 meV 124 µeV 1.24 µeV
D ISTANCE MEASUREMENT
E = hf
F IGURE 7.4. The electromagnetic radiation spectrum, a physics and astronomy landscape. Note the narrowness of what we can see with our own
^ eyes.
Väisälä interferometry 7.3
191
shows both linearly and circularly polarised radiation. One can choose
two independent directions of polarisation, for example up-down and left-
right, of which all others can be composed by combination. For example,
circularly polarised radiation is obtained by combining two mutually
orthogonal, linearly polarised rays with a phase difference between them
of π 2 . This also works the other way around: by combining clockwise
/︁
and anti-clockwise circularly polarised rays, one obtains again a linearly

polarised ray.
An intermediate form between linear and circular polarisation is repre-
sented by elliptically polarised radiation, the field vector of which turns
along an ellipse-shaped path.
In particle language one might say that the electromagnetic field is
the quantum theoretical wave function of the photon, which thus is a
vector-valued function. The photon is a vector particle, having an intrinsic
angular momentum, or spin, of magnitude h 2π . This spin — which may
/︁
be visualised as an angular-momentum vector — may be oriented along pyörähdys-

the flight direction, or opposite to it, which corresponds to either clockwise momentti
or anti-clockwise circular polarisation. Linearly polarised radiation may
again be described as an equal mix of both spin directions (Wikipedia,
Photon polarization).
^ 7.3 Väisälä interferometry

One classical distance measurement technique which continues to be in
use is Yrjö Väisälä’s white-light interferometry technique, invented as
early as in the 1920s, for the precise measurement of the lengths of long
baselines. See figure 7.6. perusviiva
The method works as follows. White light travels from a source to the
observation instrument along two paths:
◦ directly, by being reflected from the far mirror
◦ by being reflected multiple times — in the figure, four times — back
and forth between the near mirrors.
The light used is white and contains all the different wavelengths that
make up white light. Therefore the coherence length of the light is very
short, only 1.3 µm.
and food industries.

Iceland spar was also used in military equipment, for which reason it was long classified
as a strategic material (Cicala, 2013).
í Õ ! ¤. û
F IGURE 7.5. Polarisation of electromagnetic radiation. The arrows show the

field vector E. On the left, a linearly polarised wave motion; on the
right, a circularly polarised wave motion. Photograph of Iceland
^ spar Wikimedia Commons, A calcite crystal.
This is why the interference fringes will only show up if both paths are,
8 at this accuracy, equally long:8 so, if the distance of the far mirror is a
multiple of the distance between the near mirrors.

This enables the multiplication of a given distance. Say that the dis-
tance between mirrors 0 and 1 is precisely 1 m. Then we can, using
interference, place the far mirror at distance 6 m — precisely. After that,
we take mirror 1 away, and using the same method, using now the mirror
pair 1 and 6 as near mirrors, place the far mirror at 24 m. And so on. . .
In practice, this does not quite work this way. There will always remain
a small difference between the path lengths, which is eliminated and
at the same time measured, using a turnable glass plate of constant
thickness: a compensator plate.
The instrument described is actually an analogue optical correlator.
See figure 7.7. We will go deeper into the correlation method in subsection
12.5.1, in connection with the GPS.
Realising the original distance of one metre between the 0 and 1 mirrors
kvartsimitta is not quite simple, either. This uses a one metre long quartz gauge,9
9
8 This is why laser light will not do! It would produce interference fringes even when
the distances are unequally long.
9 Calibrating those precisely is again a story of its own. . .
í Õ ! ¤. û
Väisälä interferometry 7.3
193
Mirror 0 Compensator plate
Mirror 1 Mirror 4
Light source
Inter-
ference
fringes
3ℓ
ℓ
^ F IGURE 7.6. Väisälä’s interference method.
which is allowed to touch the surface of one mirror. In the air gap between
the other mirror and the end of the gauge — which has been polished to
be slightly convex — Newton rings will now show, an interference pattern.
By observing the rings in sodium light, the width of the air gap can be natriumvalo
determined.
The Väisälä interferometry method is extremely time consuming. Mea-
surement conditions are suitable only very rarely for measuring the
longest distance, 864 m. Moreover, setting up the mirrors, their orien-
tation, and the transfer of their measured places to the underground
permanent markers by projection measurements, is a complex operation
demanding its own time. During measurements, the air temperature
Analogues in geodesy:
◦ Väisälä white-light interferometry
◦ GPS measurement with pseudo-random
codes
◦ VLBI — very long baseline interferometry
Felled in 1915
◦ Plate tectonics, sea-floor magnetic stripes
2000
1980
1940
1880
1900
1920
1960
1
Correlation function
0
^ F IGURE 7.7. Dendrochronology: dating of wood using tree rings.
í Õ ! ¤. û
is continuously read from precision thermometers suspended all along

the line, providing gainful employment and physical exercise to two
measurement assistants.
10 The method has been used for measuring baselines as long as 86410
metres in Nummela, Finland, and in many places around the world.
The precision achievable is at its best ±0.02 mm. The interference mea-
surement itself is more precise still: the bottleneck is the projection
measurement, the transfer of the measurement values from the mirrors
to the underground markers.
^ 7.4 Electronic distance measurement
^ 7.4.1 The speed of light

The speed of light in a vacuum is a constant of nature. Apparently Galileo
(1564–1642) already tried to measure it in 1638 using two lamps and an
assistant: a cover was taken off one lamp, and the assistant on another
hill responded in the same way. Of course the result was useless: the
speed of light would, according to the experiment, be very large, possibly
infinite.
etäisyysmittari The first terrestrial distance measurement devices or range-finders
were developed to determine the speed of light. The prototype of the
11 method is Fizeau’s11 instrument, consisting of a light source, a reflector
hammaspyörä and a rapidly spinning camwheel. If the wheel spins at the right speed,
light leaving through an opening between the teeth will on return be
blocked by a tooth. With a slightly greater or smaller rotation speed,
however, some light will get through. In Fizeau’s tests, the distance
measured was 8.6 km.
Fizeau’s camwheel is a primitive modulator. Nowadays electronic or
electro-optical modulators are used, the task of which is to vary or modu-
late the intensity of outgoing light periodically at a certain frequency.
10 864 = 2 × 2 × 3 × 3 × 4 × 6.
11 Armand Hippolyte Louis Fizeau MIF FRS FRSE (1819–1896) was a French physicist.
He also measured the speed of light in flowing water and found in it an anomaly which
only special relativity could explain. He is one of the 72 French scientists and engineers
whose names were inscribed on the Eiffel Tower, outside and below the first balcony
(Eiffel Tower, 72 names).
í Õ ! ¤. û
Electronic distance measurement 7.4
195
Reflector
8.6 km
Light source
.
Modulation:
Observer Light intensity
↑
−→ Time
F IGURE 7.8. Fizeau’s method for measuring the speed of light. The distance
used for the 1849 measurement was from Montmartre to Mont
^ Valérien, in the municipality of Suresnes, 8 633 m.
^ 7.4.2 Electronic distance measurement instruments

When the superior accuracy of the devices developed to determine the
speed of light became clear, the picture changed. Today, the speed of light
in a vacuum is no longer measured: instead it is a quantity derived from
the definitions of the metre and the second (subsection 2.1.4) which has
the conventionally agreed value of exactly 299 792 458 m s .
/︁
Electronic distance measurement devices or range-finders can be of

three types of construction:
◦ separate: the device is placed into the forced-centring device. This pakkokeskistys-
solution has become rare due to the devices becoming ever smaller. laite
◦ a separate part that is locked onto the theodolite’s measuring tele-
scope. This solution has also become impractical:
– An instrumental tilt correction must be made.
– For example, the classical Distomat range-finder was placed
on top of the theodolite’s telescope, preventing it from being
plunged through to the other face. kojeasento
◦ integrated with the theodolite. One speaks of a coaxial solution: the

light moves in both directions through the theodolite’s measuring
telescope and uses the same optics.
Electronic distance measurement is, based on the frequency range used,
divided into two main types:
í Õ ! ¤. û
Sent phase = 0 Received phase = 0
−→ t
∆ t = t − t0
Start t 0 Stop t
Sent
Received
Clock
F IGURE 7.9. One method of electronic phase measurement: zero phase starts /
^ stops an electronic counter or “clock”.
◦ electromagnetic, using microwaves. This is obsolete as a terrestrial

method — but satellite positioning also uses microwaves!
◦ electro-optical:
– visible light, white: Mekometer
hohtodiodi – laser light or light-emitting diode (LED), visible or near in-
frared. Monochromatic.
Independent of the device type, the measurement takes place by modulat-
ing either light (or infrared) or radio waves (microwaves) with a certain
vaihe-ero frequency, and measuring the phase difference between the outgoing ra-
diation and the radiation returning from the target. Electronic phase
measurement can be very precise, but does not tell us how many whole
wavelengths fit in the distance: the ambiguity problem.
The travel time of the signal is
∆φ 1
(︂ )︂
∆t = +N ,
2π f
in which f is the frequency, ∆φ the measured phase difference in radians,

kokonaisluku- and the integer N the unknown ambiguity. Because the measured phase
tuntematon difference is assumed to lie in the interval 0, 2π , the measurement is
[︁ )︁
not yet enough to fix ∆ t. We can certainly calculate a possible travel time
∆φ 1
∆t = ,
2π f
but it would be a coincidence if it were the true travel time.

Phase measurement can be done through the measurement of time
differences: when the reference signal passes through zero in the positive
í Õ ! ¤. û
197
λ1
λ2
λ3
Sent
Run time t − t 0 −→
Common solution Received
F IGURE 7.10. Ambiguities, or integer unknowns, are resolved by using several

wavelengths. The figure shows signals on three different wave-
lengths and assumes that the phase angles of the transmitted
signals are all zero at the moment of transmission t 0 . The only
possible travel time is the one for which also the calculated phase
angles of all three received signals are identical to their measured
^ values, here also assumed to be zero. See also figure 13.4.
direction, it starts a counter, and when the returning measurement signal

does the same, the counter stops and the value is read out. See figure 7.9.
From the phase measurement ∆φ the distance is computed:
∆φ
(︂ )︂
s = 12 c ∆ t = 21 ( c ∆ t + N λ) = 21 + N λ,
2π
in which λ = c f is the wavelength, and N ∈ Z an unknown number of

/︁
whole wavelengths. Determining the integer unknowns or ambiguities N

is a similar problem as seen in connection with GPS carrier-phase mea- kantoaallon
surements. In a range-finder, several different modulation frequencies vaihe
f i (or, equivalently, wavelengths λ i = c f i ), are used, which have been
/︁
chosen so that only one distance s, and corresponding travel time ∆ t, is

compatible with integer-valued unknowns N i for all wavelengths. See
figure 7.10.
Nowadays there exist small and inexpensive hand-held range-finders,
which work either based on an infrared beam or an acoustic (ultrasound)
beam. They are handy in construction projects, and even real-estate
brokers use them.
^ 7.4.3 Reflectors
In order for electro-optical devices to work, they need a reflector placed
at the target. Typically, a so-called corner-cube prism is used, figure kuutioprisma
7.11. The principle of operation of a corner-cube prism is based on using
í Õ ! ¤. û
−→
z
^ F IGURE 7.11. Corner-cube prism.
three reflective surfaces that are perpendicular to each other, of which

the first inverts the x co-ordinate of the light ray to the − x direction, the
second, the y co-ordinate to − y, and the third, the z co-ordinate into the
− z direction. The end result is a complete inversion of the light ray:
⎡ ⎤ ⎡ ⎤
x −x
⎣ y ⎦ =⇒ ⎣ − y ⎦ .
⎢ ⎥ ⎢ ⎥
z −z
The light ray hitting the prism will be reflected back in precisely the
opposite direction, independently of which direction it came from — as
long as it is within the prism’s opening angle.
Over short distances, reflective stickers can also be used, or the light
be reflected by the target itself, without aids. Then, the accuracy of the
measurement is not necessarily the best possible!
Over longer distances one may use, instead of a single prism, an as-
sembly of three prisms. Over very long distances (tens of kilometres)
one can combine many prisms into a “pack”, figure 7.12. Nowadays such
distances (vectors) are however measured using GNSS.
The prism assembly fits into a forced-centring device.
tähys Warning When using a signal equipped with a prism, one may not place
hiusviiva- the theodolite’s crosshairs on the prism “crosshairs”! It likely is
ristikko not aiming straight, and was never meant to be used for this. Use
the triangular markings on the signal instead, figure 7.13.
^ 7.4.4 Instrumental errors of electronic range-finders

The systematic instrumental error of an electronic range-finder consists
of two parts:
í Õ ! ¤. û
199
^ F IGURE 7.12. A prism pack for measurement over long distances.
◦ the zero-point or constant error

◦ the scale or frequency error.
The constant error is an instrumental constant to be determined by

calibration. The frequency error is a scale error, which is also determined
by calibration: frequency calibration.
The constant or zero-point error is caused by the circumstance that the
electric centre of the instrument is at a different place than its nominal
one. Inside the device, the signal path may contain unknown delays. The
constant error may depend on temperature, may change slowly over time
(“drift” or “creep”), and may change in connection with repairs. For this käynti,
reason, regular calibration is recommended. ryömintä
(a) (b)
Not like this. . . . . . but like this
F IGURE 7.13. Incorrect and correct targeting at a signal equipped with a corner-
^ cube prism.
í Õ ! ¤. û
^ T ABLEAU 7.1. Calculating the constant and frequency error of a range-finder by

linear regression.
In order to calculate — estimate — the constant error a and the frequency error
V from measurements on a calibration baseline, we use the following standard
pair of equations for linear regression:
n n n
(︄ )︄/︄(︄
n n
)︄
∑︂ ∑︂ ∑︂ )︂2
s i ∆s i − ∆s i
∑︂ (︂∑︂
n si
(︁ 2 )︁
V
ˆ︁ = n si − si ,
i =1 i =1 i =1 i =1 i =1
(︄ n n
)︄
1 ∑︂ ∑︂
a
ˆ︁ = ∆s i − V
ˆ︁ si .
n
i =1 i =1
In this, n is the number of calibration distances s i used, at least 2. ∆ s i is

the difference: measured distance minus nominal distance, for each distance
measured.
The reflector also has a zero-point error, and often, the sum of the
zero-point errors of the instrument and reflector together are stated.
◦ The frequency error is often determined in the laboratory using a
precise frequency standard:
f measured − f nominal
V= . (7.3)
f nominal
In this, f measured is the frequency value obtained in calibration mea-

surements in the laboratory, and f nominal is the frequency value
stated by the manufacturer (and acting as the basis for the correc-
tion formulas in the instrument’s firmware).
◦ The constant and frequency errors are determined by calibrating
the instrument on a precise baseline:
s′i = s i − a − Vs i ,
in which
s′ “correct” distance given by the baseline
s distance as measured by the instrument to be calibrated
i number of the measurement point on the baseline, for exam-
ple the number of the pillar.
We write
def
∆ s i = s i − s′i = a + Vs i
í Õ ! ¤. û
Ray propagation in the atmosphere 7.5
201
and solve for a and V by means of linear regression, tableau 7.1.
Of course obtaining a good calibration result requires the use of
a sufficiently long calibration baseline. On a short baseline one
can only determine a with sufficient accuracy. On the other hand,
for many distance measuring devices, the electronic signal can be
taken out and compared with a frequency standard directly, without
using a baseline.
Thus, the instrumental correction for distance measurement is obtained:
s′ = s − a − Vs,
in which
s′ corrected distance
s measured distance
−a instrumental constant or zero-point correction
−V instrumental frequency correction.
The random total error of distance measurement, that is the mean error
of the measurements, also generally depends on the length of the distance
measured. An often-useful formula is
σ = α + β s,
in which α is the random error for zero distance, β the distance-dependent

random error, and σ the mean error of the observations computed from
these. Here it is assumed that systematic errors — like constant error
and frequency error — and the necessary observation reductions have
already been taken care of as corrections.
^ 7.5 Ray propagation in the atmosphere

Light — and other electromagnetic radiation like infrared or radio waves
— travels slower in air, like in other media, than in a vacuum. The
slowing-down effect of the medium is expressed by the index of refraction taitekerroin
n. The definition of the index of refraction is
c
n= ,
c′
in which c′ is the speed of light in air, and c the speed of light in a vacuum,
a constant of nature. Because air is a gas, a low-density medium, the
values of n are always very close to unity. Therefore a definition is also
í Õ ! ¤. û
used that gives the deviation of the index of refraction from unity, in units
of ppm, parts per million:
N = 106 ( n − 1) .
The index of refraction of air for the wavelengths of visible light is, accord-
ing to the following approximate formula accepted by the International
Association of Geodesy IAG at its general assembly of 1999 in Birmingham
UK (Rüeger 1996, page 55 , Anon., 1999):
11.27 K
/︁
273.15 K p hPa
NL = N0 (λ) − e, (7.4)
T 1013.25 hPa T
12 in which12
N0 (λ) index of refraction of the light used (wavelength λ) in dry air

under the following standard temperature and pressure conditions:
T = 273.15 K = 0◦ C, p = 1013.25 hPa, e = 0.0 hPa
T air temperature, unit kelvin (K), i.e., absolute temperature
p air pressure, unit hectopascal (hPa), or millibar (mbar)
e partial pressure of atmospheric water vapour (“absolute humid-
ity”), also in units of hectopascal.
N0 depends only on the wavelength λ of the light used. An approximate

formula for its calculation is
4.8866 µm2 0.0680 µm4

N0 = 287.6155 + + . (7.5)
λ2 λ4
ryhmä- ja vaihe- This is the group index of refraction, which differs from the phase index
taitekerroin of refraction.13 In connection with electronic and electro-optic distance
13
measurement equipment one must use the group index of refraction,
kantoaalto because information travels in the modulations on the carrier wave,
which propagate at group speed.
Example For a helium-neon laser (wavelength λ = 632.8 nm) equation
7.5 yields N0 = 300.243.
12 Yes, the small anomalous water-vapour effect is treated as non-dispersive!

13 In fact, the phase index of refraction is similarly (Anon., 1999):
1.62887µm2 0.01360µm4
N0 = 287.6155 + + .
λ2 λ4
í Õ ! ¤. û
Ray propagation in the atmosphere 7.5
203
The index of refraction for microwaves again is
77.624 K 64.70 K
/︁ /︁
5748 K
(︂ )︂
hPa hPa
NM = ( p − e) + 1+ e. (7.6)
T T T
Unlike the index of refraction for light and infrared, the microwave
index of refraction in the troposphere is not dependent upon wavelength.
However, the propagation of microwaves in the ionosphere is an entirely
different matter relevant to satellite positioning; see subsection 12.7.3.
As an interesting detail, we may still note that, at a temperature of
T = 273.15 K, in the index-of-refraction equation for microwaves 7.6 the
ratio of the effect of e to the effect of p − e (dry-air partial pressure)
is over twenty times larger than in the visible-light index-of-refraction
equation 7.4!
Index p−e e Ratio
Optical 0.296 30 0.25504 0.86075

Microwave 0.284 18 5.22133 18.373
Microwaves are thus very sensitive to water vapour,14 which is one 14

of the drawbacks of this measurement technique. This problem is also
manifest with GPS measurements.
The index of refraction of air affects the measured distance in the
following way:
s − s′ = ( n − 1) s, (7.7)
in which s is the measured distance, and s′ the true distance, which

would have been measured in a vacuum. The refraction correction is now
applied as follows:
s′ = s + ∆ s n ,
in which
∆ s n = − ( n − 1) s = −10−6 N s
is the correction. In the literature, the symbol K ′ is used for this. Because
the correction is so small, it is permissible to use an approximate value likiarvo
for the distance s.
14 This is caused by the non-symmetry and large dipole moment, or chemical polarity, of
water molecules: in the H2 O molecule, the angle between the two O–H bonds is 104◦. 5.
This is also one reason why water is a liquid at room temperature, and such a good
solvent, and why a microwave oven is such a useful device for preparing food. Almost
all other molecules in the atmosphere, like N2 , O2 , CO2 , Ar, and CH4 , are non-polar,
and gases. Wikipedia, Chemical polarity.
í Õ ! ¤. û
^ 7.6 “Curvature corrections”

In distance measurement, the “curvature corrections” are second-order
corrections: they are significant only when the distance of measurement
kartoitus- exceeds many kilometres. Therefore they are inconsequential in detail or
mittaus lower-order base-network surveys.
runkomittaus
Because base-network surveys are nowadays done almost exclusively
X
using GNSS, these corrections are mostly of historical interest, and we
present them only briefly. More detailed expositions are found in the
literature, for example Rüeger (1996, 2002).
refraktiokerroin In the equations for curvature corrections, the refraction coefficient
k, which we already encountered in connection with zenith-angle mea-
surement (subsection 6.8.1), appears again. There are four different
corrections:
The measurement-ray curvature correction The curving of the
measurement ray causes a lengthening of the path. The measured
distance is, for this geometrical reason, longer than the true
distance. The equation for the correction is
s3 2 s
3
∆sρ = − = − k , (7.8)
24ρ 2 24R 2
in which k is the refraction coefficient, R the radius of the Earth,
and s the distance.
The Earth’s surface curvature correction After various reduction
stages, the straight-line distance between the two projection points
on a reference surface is usually obtained. It is, however, the longer
distance over the curved Earth’s surface which is wanted. The
equation for this correction is
s3
∆s R = .
24R 2
The “second velocity correction” Usually, the effect of refraction on
the propagation of the measurement ray is evaluated based on
weather observations — measurements of air pressure, temper-
ature, and humidity — made at both ends of the path. For very
long paths, these measurements are no longer representative of
the whole ray path.
The second velocity correction is a systematic effect caused by the
ray curvature differing from the curvature of the Earth’s surface,
together with the strong vertical gradient of the air pressure —
í Õ ! ¤. û
“Curvature corrections” 7.6
205
B
P
A
n P ̸= 12 ( n A + n B )
F IGURE 7.14. The second velocity correction: on a curved Earth, the indices
of refraction at the end points of the measurement path are not
^ representative.
and thus the index of refraction. When the curvature of the ray
path is smaller than that of the Earth’s surface, at longer distances
the path will “dive” deeper into the Earth’s atmosphere than what
the end points of the path are telling us. See Rüeger (1996) page 81.
The computational formula — the derivation of which is laborious
— is
s3
∆ s n′ = − k (1 − k)
.
12R 2
The literature calls this correction K ′′ .
All three corrections can be combined into one equation:
s3
∆ s ρ Rn′ = ∆ s ρ + ∆ s R + ∆ s n′ = (1 − k)2 .
24R 2
Let us calculate some values, assuming k = 0.2:
s (km) 1 3 10 30 100
∆ s ρ Rn′ 0.66 µm 18 µm 0.66 mm 18 mm 0.66 m
The effect is thus normally very small.

The “terrain correction” The terrain correction of distance mea-
surement (Juhani Kakkuri, personal communication; figure
Isothermal surfaces
A P B
Terrain n P ̸= 21 ( n A + n B )
F IGURE 7.15. The terrain correction of distance measurement. Due to the shape
of the terrain, the indices of refraction at the end points of the
^ measurement path are not representative.
í Õ ! ¤. û
∆sρ B
s
A
∆s R hB
hA
A′ B′
′
s
F IGURE 7.16. Reduction of distance measurement to a reference level, equation

^ 7.9.
7.15; Kakkuri et al., 1981) is caused by the surfaces of equal

temperature within the atmosphere, the isothermal surfaces,
generally following the forms of the terrain. Therefore, in the
same way as in the case of the second velocity correction, the
weather observations at points A and B are non-representative of
the mean value along the measurement path. There is no simple
formula for this phenomenon.
^ 7.7 Geometric reductions
^ 7.7.1 Reduction to a reference surface

Thanks to the above described “curvature corrections”, the original mea-
surement was reduced for the curvature of both the measurement ray
and the Earth’s surface. Thanks to corrections ∆ s n and ∆ s n′ the delays
by the atmosphere were also taken into account. The measurement path
is however still sloping and above the Earth’s surface.
The correction, or reduction, to a chosen reference level corrects both the
slope of the measured distance in space and its height above the chosen
reference surface. The following may be chosen for reference surfaces:
◦ sea level
◦ the surface of a reference ellipsoid
◦ the height of a locally defined reference surface (“zero level”).
The reduction is carried out as follows:
/︁s −)︁(︁∆h
2 2
2
(s′ ) = (︁ /︁ )︁ , (7.9)
1 + hA R 1 + hB R
í Õ ! ¤. û
Geometric reductions 7.7
207
^ T ABLE 7.2. Examples of distance reductions.
s= 100 m 1 km 10 km 100 km
hA = 0 h B = 10 m −0.50 m −51 mm −13 mm −79 mm

hA = 0 h B = 100 m - −5.0 m −0.58 m −0.83 m
hA = 0 h B = 1000 m - - −51 m −13 m
h A = 100 m h B = 100 m −1.6 mm −16 mm −0.16 m −1.6 m

h A = 1000 m h B = 1000 m −16 mm −0.16 m −1.6 m −16 m
in which s is the measured slant range in space, s′ the reduced distance

(meaning the distance between the points A ′ , B′ projected onto the refer-
ence surface, see figure 7.16), hA = A A ′ and h B = BB′ the heights of the
points from the reference surface, ∆ h = hA − h B , and R the (approximate)
radius of curvature of the Earth. The heights of the theodolite and signal
are included in the heights hA and h B .
An approximate equation for when ∆ h ≪ s (and of course hA , h B ≪ R )
is
∆ h2 h + hB
−s A
s′ ≈ s − .
2s 2R
An advantage of this equation is, that it gives separate terms for the
influence of path slope and path mean height.
Table 7.2 gives some examples of corrections s′ − s. It is seen that this
correction can be already substantial also for short distances.
Modern theodolites and total stations are able to compute, in addition takymetri
to device-specific corrections, at least the slope reduction of the measured
range.
^ 7.7.2 Map-projection reduction

A map-projection scale reduction is needed if we wish to have, instead of
the length of an arc reduced to the reference ellipsoid s ell , the length in
the map plane, the projected length s proj . For example, in the case of the
Gauss–Krüger projection, as used in Finland, the approximate reduction
is done as follows:15 15
yA2 + yA yB + yB 2
(︃ )︃
s GK = s ell 1.0 + , (7.10)
6R 2
in which the distance s is between the points A and B, map co-ordinates
( x A , yA ) and ( xB , yB ). Here, y is the raw distance from the central merid-
ian, without the false easting 500 000 m.
í Õ ! ¤. û
This reduction depends on the map projection chosen, and is thus

different for different map projections. For example, the equation for the
Universal Transverse Mercator (UTM) projection is otherwise the same
as 7.10, but the constant — the scale on the central meridian — is 0.9996
instead of 1.0.
1. In what kinds of situations would tape measurement be an sensible
option?
2. Derive the approximate slope correction ∆ℓ⊥ , equation 7.2, ex-
pressed in the slope angle α in degrees. Assume a small angle.
3. What is the relationship between wavelength and frequency for
electromagnetic radiation?
4. What is the relationship between frequency and the energy of a
photon for electromagnetic radiation?
5. What is the polarisation of electromagnetic waves?
6. Describe the correlation process between two identical but random
signals. Why must the signals be random?
7. Describe ambiguity resolution in electronic distance measurement.
8. What are the factors affecting ray propagation in the atmosphere,
15 The derivation starts from the scale of the Mercator projection on the sphere being,
close to the equator,
2
1 1 1 2 1 x
m= ≈ ≈ 1 + 2 ϕ ≈ 1 + 2 ,
cos ϕ 1 − 21 ϕ2 R2
in which x is the metric distance from the equator. For a transversal Mercator projection,
this becomes
y2
m = 1 + 12
,
R2
with y the distance from the central meridian. As clearly the scale distortion is quadratic
in y, it will be, on a straight line on the map, a quadratic function of place along the
line. Then, calculating the mean scale along the line by Simpson integration, equation
10.4, will be exact. As follows ( M is the midpoint between end points A and B):
1 1 (︁ 2 2 2 1 (︁ 2 2 2
m AB = 1 + 12
)︁ )︁
· 6 yA + 4 yM + yB = 1 + yA + ( y A + yB ) + yB =
R2 12R 2
1 (︁ 2 2
)︁ 1 (︁ 2 2
)︁
= 1+ 2 yA + 2 yA yB + 2 yB = 1+ y + yA yB + yB ,
12R 2 6R 2 A
from which the result 7.10 follows.
í Õ ! ¤. û
Self-test questions
209
for visible light and for microwaves?
9. How are range-finders calibrated?
10. How are distance measurements reduced to a reference surface?
def
11. Prove equation 7.8, by defining ψ = s ρ , the angle at the centre of
/︁
the circular-arc beam path, so that
∆ s ρ = ρ 2 sin 12 ψ − ψ .
(︁ )︁
Use the Taylor expansion for the sine function.
í Õ ! ¤. û
^ Base-network and detail-survey
measurement
8
[. . . ] Komitea [Karttakomitea] sitten ehdottaa
toimenpiteenä tarkan karttalaitoksen aikaansaamiseksi,
että perustettaisiin Geodeettinen kommissioni, jonka tulisi
toimittaa: 1) karttalaitoksen pohjaksi tarvittavat
perustavat työt, 2) geodeettiset täytetyöt, 3) topografinen
peruskartta mittakaavassa 1 : 20, 000, Ensimäinen ryhmä
käsittäisi ensiluokan kolmiomittaukset ja
tarkkavaakituksen ja toinen ryhmä pääasiallisesti
alemman luokan kolmiomittaukset. [. . . ]
Finnish Geodetic Institute director Ilmari Bonsdorff (1920)
^ 8.1 Objective and planning of base-network measurement

The task of base-network measurement is to create, through a network runkomittaus
hierarchy, the geometric foundation for mapping the country. For this
purpose a permanent, sufficiently dense and precise benchmark set is kiintopisteistö
created to which the local measurements of the various user groups will
be tied. The co-ordinates of the benchmarks are known in the national
co-ordinate reference frame, and by using them, the locally measured
points and drafted maps will also be in the same frame.
Benchmarks are used both in detail surveys for mapping, and in setting kartoitus-
out plans onto the terrain — “the inverse problem of mapping”. mittaus
maastoon
The planning of base-network measurements starts from taking an
merkintä
inventory of the existing situation and an analysis of needs. The goal is
to build a sufficiently precise and dense set of benchmarks, at a minimum
cost. It pays off, however, to plan it for the future, especially in choosing
the benchmark substrate and in monumentation. Possible future building kiintopiste-
activity is also taken into account in the choice of location, as this may alusta
– 211 –
212 8 B ASE - NETWORK AND DETAIL - SURVEY MEASUREMENT
destroy a point or render it useless by destroying the visibility conditions

for measurement: in the case of theodolite measurement, intervisibility
between points, in the case of satellite measurement, sufficient visibility
of the sky from the point.
rekognosointi Reconnaissance is part of the detailed planning of a network: an on-the-
spot check that the measurements can be carried out as planned. Before
reconnaissance, a “map reconnaissance” is done, in which the situation
is judged from the office. If new points are created, a clear and useable
pistekortti point description must be drafted for every point, with the aid of which
also others can find it.
In addition to precision, attention must also be paid to reliability.
karkea virhe Reliability means that possible gross measurement errors are noticed
with the greatest possible ease, and that the effect on the end result — the
co-ordinate solution — of the greatest possible gross error that remains
undetected is as small as possible. To this end, there must be sufficient
redundancy in the network: the measurement plan should always contain
enough extra measurements above and beyond the required minimum.
kolmiomittaus- Nowadays, instead of the traditional solution — a triangulation net-
verkko work densified by traverses — GNSS networks are commonly measured.
monikulmiojono
Those, too, need to be designed right, meaning hierarchically, and the
measurements must be planned so that the accuracy and point-density
objectives are achieved in an economical way.
The following alternative methods are on offer for base-network mea-
surement:
◦ satellite positioning (GNSS)
takymetri ◦ traditional terrestrial measurement using total stations
ilmakolmiointi ◦ photogrammetric aerotriangulation.
The choice is dictated by the purpose of use — accuracy requirements

and size of and ease of movement within the area — as well as by the
visibility conditions at the points.
Measurement technology for base networks has, with satellite posi-
tioning, undergone a revolutionary change. Traditionally, base-network
measurements were done for the horizontal using triangulation and
traversing, and using precise levelling and lower-order levelling for mea-
suring heights. Nowadays, satellite positioning is always used if at all
possible. There are however situations where traditional techniques hold
their own, like tunnel and mine surveying where the sky cannot be seen.
í Õ ! ¤. û
Guidance and standards 8.2
213
A good introduction to the subject is Salmenperä (1998), on which this
presentation is partly based.
^ 8.2 Guidance and standards

Many different standards and guidance documents exist to ensure the ohjeistus
quality and effectiveness of geodetic measurements. We present here
only those with a clear official status.
Important guidance was provided by the Zoning Survey Guide (“Kaa- Kaavoitus-
voitusmittausohjeet”) of the Finnish National Land Survey (Anon., 2003). mittausohjeet
The guidance concerns base-network measurement, detail survey, aerial
mapping, and the drafting of a zoning base map, as well as documenting kaavan
the work. pohjakartta
Base-network measurement, co-ordinate reference systems, map projec-

tions, and zoning surveying has of late been the subject of guidance by
JUHTA, the Advisory Committee on Information Management in Public
Administration, who publish the series JHS, Recommendations for Infor-
mation Management in Public Administration. In 2013 it was decided
that from then on, the guidance concerning zoning surveying would be
published in the JHS series.
Of the guidance documents relating to these subjects, we may mention
the following on-line publications, unfortunately only in Finnish:
◦ JHS 196: EUREF-FIN co-ordinates in Finland (JUHTA, 2016a).
◦ JHS 197: EUREF-FIN co-ordinate systems, related transformations
and map-sheet division for ETRS89 (JUHTA, 2016b).
◦ JHS 163: Finland’s height system N2000 (JUHTA, 2010).
◦ JHS 178: Interface for geographic information services for local
authorities. This document defines an interface called kuntaGML1 1
(a variant of Geographic Mark-up Language) (JUHTA, 2012a).
◦ JHS 184: Point measurement in the EUREF-FIN co-ordinate reference
system (JUHTA, 2012b).
◦ JHS 185: Composing base maps for the city. This replaces in part the
earlier Zoning Survey Guide and the Zoning Base Map Guide (“Kaa-
van pohjakartta”) by the Finnish National Land Survey (JUHTA,
2014).
Terminology work also represents important standardisation work in the sanastotyö
1 The name translates as “municipalityGML”.
í Õ ! ¤. û
^ T ABLE 8.1. Methods for base-network measurement.
Areal Order Traditional Modern

extent Old New methods methods
Global - - - GNSS, VLBI, satellite
laser, DORIS
1000 km - E1 (Stellar GNSS, continuously
triangulation) operating (FinnRefTM )
100 km I E1, E1b, E2 First-order GNSS, EUREF-FIN
triangulation densification
10 km II, III E3 Lower-order GNSS, static
triangulation
1 km IV, V E4 Traverses, aero- GNSS, aerotriangulation,
triangulation RTK with caution
Finnish-language area. Let us mention the Vocabulary of Geoinformatics

(Sanastokeskus TSK, 2018) drafted in collaboration by the National Land
Survey and the Finnish Terminology Centre TSK.
Guidance and standardisation is always a work in progress.
^ 8.3 Network hierarchy and classification

During the past quarter of a century, almost all base-network measure-
ments have already been done using the satellite positioning technique
— with the exception of precise levelling, for technical reasons (the geoid
problem). Table 8.1 catalogues the technologies used back then and now
in connection with base-network measurement.
The table also illustrates well the concept of network hierarchy: more
local networks are always tied to more extensive ones, which serve as
the former’s “formal truth”. The order of operations is always from the
large to the small: first the most extensive networks are measured, which
are then densified with measurements in a smaller area. In this way, a
kiintopisteistö benchmark set is obtained that
◦ Covers the whole country.

◦ Is dense enough: construction requires starting points sufficiently
nearby, at most a few hundred metres from the project area and
from each other.
◦ Is of homogeneous quality.
í Õ ! ¤. û
Network hierarchy and classification 8.3
215
(a)
A hierarchical level is forgotten in between
(b)
Use of two different hierarchy paths in the same measurement
F IGURE 8.1. The significance of network hierarchy, and mistakes often made.
^ Not like this. . . but like this.
The hierarchical method is meant to prevent the unpleasant situation

from occurring in which neighbouring points have co-ordinates deter-
mined for them along different paths, so that the relative location preci-
sion between them is weak.
The new measurement technologies mentioned in the table will be
discussed in later chapters.
The planning of fundamental geodetic works in Finland started already
before independence, and the work started immediately after (Bonsdorff,
1920). The network of the first-order triangulation mentioned in the kolmiomittaus
table comprises 364 points and covers the whole territory of Finland. The
network was measured by the Finnish Geodetic Institute during 1919–
1987. Lower-order triangulations and traverses were measured by the jonomittaus
then National Board of Survey, today the National Land Survey. Local
í Õ ! ¤. û
◦
20◦ 24◦ 28◦ 32 ◦
72 ◦ 72
◦
70 ◦ 70
Kevo
Kilpisjärvi
Hetta ◦
68 ◦ Savukoski 68
Sodankylä
◦
66 ◦ Tornio 66
Kuusamo
Oulu
Pyhäjoki ◦
64 ◦ Kivetty 64
Vaasa
Romuvaara
Joensuu ◦
62 ◦ 62
Orivesi
Mikkeli
Rauma
Finnström Metsähovi Virolahti ◦

60 ◦ Tuorla Masala 60
Degerby
20 ◦ 28◦ 32◦
24◦
F IGURE 8.2. The Finnish continuously operating GNSS network FinnRef™, sta-
tus 2018. The stations collect GNSS measurements continuously at
a rate of one measurement event per second. The data centre is at
the Finnish National Land Survey’s Geospatial Research Institute
^ FGI in Masala, Kirkkonummi, 30 km west of Helsinki.
measurements were carried out by many players, such as municipalities.

In a similar way (JUHTA, 2012b) the points of orders E1 and E1b were
measured by the Finnish Geodetic Institute, whereas the orders E2
and E3 were measured by the National Land Survey and the Finnish
Maritime Administration. E4 and the use-point orders E5 and E6 are
measured by municipalities.
í Õ ! ¤. û
Network hierarchy and classification 8.3
217
20 ◦
24◦
Pettuvuori
Pässilänvuori Isolehtomäki
Nummikangas
Kaskinen 1029
62 ◦ Vaaterinmäki
62◦
Samminmaja
Siikaneva Puolanka
Isoviita Mutala
Pohjalahti
Hevosoja
Aulanko
Väättäinen Porlammi
Karhunoja
Rokokallio Kymi
Perniö Härkäpää
60 ◦ Degerby Bredberg Helsinki

Dragsfjärd 60◦
Kökar
Hanko
20 ◦
24◦
F IGURE 8.3. The Finnish EUREF-FIN first-stage densification network, detail.

These points, together with the FinnRef™ points, form the E1
^ order.
Today, the highest level in the Finnish national GNSS network hierarchy
is formed by the continuously operating GNSS network FinnRef™. Earlier
on it consisted of 13 stations, in 2012–2013 a renovation was carried
out, bringing it up to 20 stations. The observations are collected by the
Finnish National Land Survey’s Geospatial Research Institute FGI, the
former Finnish Geodetic Institute. In 1996–1999 the FGI carried out a
two-stage EUREF-FIN densification using the static GPS measurement
technique, comprising in total some 450 points. The network of the first
stage consists of 100 points, see figure 8.3 and JUHTA (2016a). It was
measured in 1996–1997. Together with the continuously operating GNSS
network FinnRef, it forms the modern order I, or E1. Together they
determine the EUREF-FIN co-ordinate reference frame. The other EUREF-
FIN densification phase, which was measured in 1998–1999, comprises
350 points, and was designed to offer easier-to-reach points for practical
measurements. Its order is E1b.
The National Land Survey has already for many years carried out base-
í Õ ! ¤. û
network measurements using the static GNSS measurement technique:

there are some 2000 points of order E2, the measurements of which are
tied directly to points of orders E1 and E1b. In recent years, kinematic
tosiaikainen GNSS measurement (RTK, real-time kinematic) has become common in
lower-order base-network measurements, although its suitability for this
has been credibly questioned. The technique should not be used without
adequate reliability safeguards (JUHTA, 2012b).
A good description of the co-ordinate solutions used in Finland and the
relationships between them is given in Häkli et al. (2009).
^ 8.4 The terrain, the ellipsoid and the map plane

Geodetic measurement networks, like a triangulation network or a tra-
verse, are in fact three-dimensional networks, figure 8.4. A logical idea
tasoitus is then to also carry out the computation of the network, the adjust-
ment, three-dimensionally: the point co-ordinates are written three-
dimensionally in the form of rectangular co-ordinates, and every ob-
servable is described as a function of those co-ordinates of the points
between which the measurement takes place. This is how one obtains
the observation equations upon which the adjustment of the network is
based.
verkkotasoitus Three-dimensional network adjustment is a tempting thought, mostly
because of the simplicity of the underlying idea. The formation of obser-
vation equations is nevertheless complicated, as the measurements are
koje- done, at every point, in instrument co-ordinates; co-ordinates in which
koordinaatit the z axis points upwards along the local plumb line. The direction of the
luotiviiva plumb line, which can be measured by astronomical means, is different

at each point, as seen in figure 8.4.
This means that at least in the observation equations for the horizontal
angles (azimuths) and zenith angles, the direction of the plumb line must
vinoetäisyys be along. This makes these equations seriously complicated. For slant
ranges, on the other hand, the observation equation is simple.
Terrestrial geodetic measurements are always made close to the physi-
cal surface of the Earth, generally between points located on the surface.
Thus we may call the network geometry “quasi-two-dimensional”. It
would also seem to make sense to try to carry out the computations in two
dimensions, on a suitably chosen, mathematically simple computation
or reference surface close to the Earth’s surface. The Earth’s physical
í Õ ! ¤. û
The terrain, the ellipsoid and the map plane
Horizontal and vertical

8.4
219
Local angle measurements
plumb line
Vertical
Horizontal
Triangulation network Traverse
F IGURE 8.4. Geodetic networks are three-dimensional. A triangulation network

^ and a traverse in space.
surface with its mountains and depths is however too craggy to serve as
a computation surface.
More suitable computation surfaces are the reference ellipsoid, or —
in a small area — the map projection plane, figure 8.5. In preparation
for computation, the observations are reduced to this computation or
reference surface.
^ 8.4.1 Adjustment on the reference ellipsoid

The reference ellipsoid coincides rather well with the Earth’s surface, and
as a simple mathematical surface it may serve as a computation surface.
Table 8.2 shows the magnitude of the differences between the physical
surface of the Earth and the reference ellipsoid.2 And let us remark still, 2
that humankind lives close to the surface of the solid Earth on land, but
close to the sea surface on the seas: the impression given by the table
exaggerates the thickness of the human living space.
For comparison, the difference between the equatorial and polar radii of
the GRS80 reference ellipsoid is already 21.4 km or 0.336 %. A “reference
sphere” would be a clearly poorer approximation.
The reference ellipsoid was widely used as a computation surface as
early as in the 19th century, before the existence of computing machinery
and satellite positioning. The mathematics needed is complicated, but
the method is more intuitive: terrestrial geodetic networks are on the
2 For comparison: the mountain Olympus Mons on the planet Mars is 22 km above its
surroundings, 0.65 % of the radius of Mars. Gravity on Mars is only one-third of that on
Earth.
í Õ ! ¤. û
B′
A
C
A′ C′
(a)
Measurement in the terrain
x
Z
B′′
B′
A′ C′
Y
A ′′
X C ′′
y
(b) (c)
Projected on the reference ellipsoid In the map projection plane,
“straight lines” are curved
F IGURE 8.5. Use of a reference ellipsoid and a map projection plane as reference
^ surfaces when mapping the Earth.
Earth’s surface, close to the reference ellipsoid, and the local plumb lines
along which the vertical axes of measurement instruments are aligned
^ T ABLE 8.2. Goodness of approximation by the reference ellipsoid: separation

between the physical surface of the Earth and the reference ellipsoid,
both in kilometres and in proportion to the Earth’s radius.
Unit km %
Highest (Mount Everest) +8.8 +0.138

Deepest (Mariana Trench) −11 −0.17
Land mean height +0.84 +0.013
Sea mean depth −3.8 −0.06
Sea surface (geoid) ±0.1 ±0.0016
í Õ ! ¤. û
The terrain, the ellipsoid and the map plane 8.4
221
are close to the normals to the reference ellipsoid surface.
Nowadays base networks are measured using satellite technology, and runkoverkko
the traditional method has passed into history. GNSS networks are always
adjusted truly three-dimensionally.
^ 8.4.2 Adjustment in the map plane

The adjustment of small, local networks, such as traverses, can be done
without significant error in the map projection plane.
A map projection is always applied in such a way that, first, for a
terrain point, geodetic co-ordinates ϕ and λ are calculated on the surface
of the reference ellipsoid. This is how the projection onto the surface of
the ellipsoid is done. Then, the points on the ellipsoid are projected onto
the map plane.
Of course the curved surface of the ellipsoid cannot be mapped onto
the plane without error. The objects projected are distorted: directions,
distances, surface areas, may all be wrong in the map plane. The map
projection is chosen so, that some aspects that are considered important,
are not distorted. Some other aspects then are distorted, sometimes badly.
For example, a conformal projection maps angles and distance ratios kulmatarkka
correctly, but, as the classical Mercator projection demonstrates, it can
depict surface areas spectacularly wrongly.
In conformal projections, small objects are nevertheless mapped with
their correct shapes: their scales and absolute orientations may be wrong,
but their shapes are correct. For larger objects, the “straight” lines pro-
jected from the ellipsoid are curved in the map plane. In the map plane,
directions may be different from those on the surface of the ellipsoid3 — 3
although in conformal projections the angles are identical.
Expressing and solving the network adjustment problem in the map
projection plane is relatively simple, however it presupposes that
◦ the distance measurements are reduced, first to the reference sur-
face, then to the map projection plane (so the scale reduction of the
map projection has been done).
◦ The map projection is conformal, so the measured horizontal angles
are directly useable without reduction. Projections used for general
maps are conformal, like the Gauss–Krüger and UTM projections
3 In the classical Mercator projection they are however identical, a valued property in
navigation at sea.
í Õ ! ¤. û
ζ̄B
Plumb line B
ζB
B ζ̄ A
s
ζA A hB
hB
A
hA
hA C s proj
C
α α
Projection pla
ne
F IGURE 8.6. Transferring the geometry for adjustment of a small network to

^ the map projection plane.
used in Finland. In Finland, map co-ordinates can thus be used

directly in network adjustment.
◦ The zenith angles between two measurement points A and B have
been measured in both directions, and the angles to be used are
1 g
averages of sorts of the measurements: ζ A = 2 ζ A + 200 − ζB ,
( )
1 g
ζB = 2 ζB + 200 − ζ A .
( ) Then one may calculate heights from the
reference level according to rectangular geometry (see subsection
6.8.3).
◦ Known points, like the starting and closing points as well as the
liitospiste auxiliary points of a traverse, have known co-ordinates in the map
projection plane.
A visual explanation of this approach is presented in figure 8.6.
^ 8.5 Detail survey

Detail survey (Kahmen and Faig, 1988, pages 285–303) is the stage of
the measurement process, based on base-network measurements, that
serves to map details in the terrain. It is the most laborious stage of
the whole mapping project. A detail survey consists of collecting the
data and processing it into the desired end product: a map or a digital
paikkatieto geospatial data set. In the processing stage of detail surveying, the results
of the base-network measurement come along, assuring the geometric
correctness of the result.
In the following we describe in more detail four classical terres-
sidoslinja- trial methods: right-angle survey, tie-in survey, radial survey, and
mittaus free-stationing survey.
A popular detail survey method is also real-time kinematic (RTK) satel-
í Õ ! ¤. û
Detail survey 8.5
223
50 gon
Ranging
Double
rod
pentagon
Field prism (2×)
sketch
Level
Right-angle
pentagon
prism
Tape measure
^ F IGURE 8.7. Tools of the right-angle survey method.
lite positioning, which is often competitive. Its use is however difficult,

for example, in a high-rise urban landscapes, “urban canyons”, and im-
possible underground.
A new method for the detail survey of extended areas is the mobile
mapping system. In it, a terrestrial laser scanner integrated with a
GNSS receiver, an inertial measurement unit, and a cluster of cameras,
is mounted on a car and driven around the area to be mapped. The
amount of observational data produced is huge, and the processing work
is demanding. The strength of the method is that manual field work is
largely eliminated: only the signalisation of known points remains. A
weakness is that occluded patches easily remain in the data. The results katvealue
may be presented not only as a traditional map, but also as visual scenery
like in Google Street View™.
^ 8.5.1 Right-angle survey (prism surveying)

A measuring tape, a double pentagon prism, ranging rods4 to mark out linjaseiväs
4
mapping lines, and drawing paper are needed for this: see figure 8.7.
The measurement is carried out according to figure 8.8. A and B are
known points, often traverse points from a lowest-order survey. The right jonopiste
angles are created using a double pentagon prism: when one stands on
4 German fluchtstab, French, Dutch jalon.
í Õ ! ¤. û
x
x
Tape measure 8.0
5
6
4.3
11.8
2
4.3
3
19.1
8.0 P
8.84
12.2
i
25.9
8
6.21
5
B
5
bi bP
aP
5.54
ai
17.9
B
12.6
3.87
3
9.35
6
8.20
A Right-angle A
pentagon
prism
y y
(a) (b)
Method Example
^ F IGURE 8.8. Right-angle survey.
the line AB, both end points (or rather, the ranging rods set up on them)
show on top of each other in a device containing two pentagonal prisms.
One of them looks by a right angle (100g ) to the right, the other by the
same angular amount to the left, in the opposite direction.
The distances b may not be longer than one tape length (50 m).
On the right-hand side of the figure is shown how a building is mea-
sured using the right-angle method.
One should always take pains to make sure that there is sufficient
redundancy to check for mistakes. In this example, the wall measures of
the house could be measured.
eksteriööri The measurements with their number values are written on a field
sketch, preferably neatly and systematically, in a way that will also be
intelligible to others besides the drafter at the moment of drawing.
^ 8.5.2 Tie-in survey

Often, a method is used in which the measurement is carried out only with
sidoslinja- distance measurements in several densification steps (“Tie-in survey”,
mittaus Kahmen and Faig, 1988) and figure 8.9. In this example a cross-measure
(dashed line) of the parcel and the wall measures of the house were
í Õ ! ¤. û
Detail survey 8.5
225
32.2
12
46.
0
20.3
B
9
37.41
0.0
96
20.
.62
9.64
5.13
6
18.5
50
19.89
0.0
16.70
30.75
6
17.8 1
4.50 14.7
1
13.4
6.50
5.19 4.51
0.0 0.0 D 0.0

C 0.0
47.75
29.38
33.90
21.88
0.0
F IGURE 8.9. Tie-in survey. A cross-measure and the wall measures of the house
^ serve as checks.
measured as checks. In this way one finds out in the field if there is an
error in the measurements. A crude check can be done graphically.
More often, a mixed form is used in which the tie-in method comple-
ments the right-angle survey method.
^ 8.5.3 Radial survey

Radial survey is explained in figure 8.10. Determination of the location
in the plane of the points i = 1, 2, . . . , n is done by means of measuring
the angles θ i and the (slant) ranges s i . In the figure, the example point is
i = 2. From a point with known co-ordinates — typically a traverse point
— A , the horizontal angle θ2 between another known auxiliary point B,
and the point to be determined, point 2, is measured. The orientation
direction α AB is calculated by solving the inverse geodetic problem from geodeettinen
the given location co-ordinates of points A and B. käänteistehtävä
í Õ ! ¤. û
North
Tacheometer
B
α AB B
1
X A
A θ2
Prism pole
s2 2
4 2
3
(a) (b)
Measurement geometry Work mode in perspective
^ F IGURE 8.10. The radial survey method.
After this
x i = x A + s i cos (α AB + θ i ) = s i sin ζ i cos (α AB + θ i ) ,

yi = yA + s i sin (α AB + θ i ) = s i sin ζ i cos (α AB + θ i ) .
Here, the measured slant range has been reduced to the horizontal
def
distance s i = s i sin ζ i , in which ζ i is the vertical or zenith angle, which is
also measured.
Radial survey thus includes trigonometric height determination: as the
instrument measures both the zenith angle ζ i and the slant range s i , one
obtains, in addition to the horizontal distance s i , the third co-ordinate
z i = z A + s i cos ζ i .
The height of either point A or point B must be known as the starting

value for the height computation. If the height of point B is given, it is not
even necessary to measure the instrument height of the tacheometer over
marker A , because the height of the prism pole will not change during
measurement.
The equipment used is an electronic tacheometer or total station
equipped with suitable software. The instrument is chosen based on
the accuracy objective of the mapping to be undertaken.
Radial surveying is a numerical mapping method in which the map
product is generated computationally based on the measurement values
í Õ ! ¤. û
Detail survey 8.5
227
θ i , ζ i , and s i using their metadata. As the measurements are collected
electronically, this is inexpensive and readily automated when the num-
ber of points to be measured is large and accuracy requirements high.
The radial surveying method is especially practicable in urban areas,
because the setting out of mapping lines required in the right-angle
method can be cumbersome due to traffic. Objects of which both plane
and height information are required — special technical measurements,
building and utility-line surveys — and busy work sites are also suitable
for radial survey.
^ 8.5.4 Free-stationing survey

In free-stationing survey — German: freie Standpunktwahl — the tache-
ometer is placed on a freely chosen point in the terrain, subject only
to good visibility to the points to be measured as well as at least two,
preferably three to four, points of which the co-ordinates are known,
normally base-network points. The advantage of the method is, that the
instrument does not have to be precisely centred over a known point
marker or monument: the need for centring, and measuring instrument keskistys
height, goes away.
The method has become widespread with the availability of electronic
tacheometers and increased computing power. In principle, however,
the method could be used with a theodolite and measuring tape. See
figure 8.11. Note that there is no point marker (monument) under the
tacheometer!
Let the co-ordinates of the points A ( x A , yA ) and B ( xB , yB ) be known.
The instrument is set up on the unmarked point O . Measurements are
made to the points A and B and the unknown points i = 1, 2, 3, . . . , n (the
example point in the figure is i = 2):
◦ horizontal directions θ
◦ distances s.
Ignoring heights for now, the measurement yields, in local or instrument
co-ordinates ( u, v) :
uA = sA, v A = 0,
u B = s B cos (θB − θ A ) , vB = s B sin (θB − θ A ) ,
u i = s i cos (θ i − θ A ) , v i = s i sin (θ i − θ A ) .
So we use the forward geodetic problem to compute ( u, v) co-ordinates for geodeettinen

päätehtävä
í Õ ! ¤. û
x u
4 A Tacheometer
sA O A
1
O
θ2
sB s2
3 Prism pole
2 (Nothing)
B 2
y
B
v
(a) (b)
Measurement geometry Work mode in perspective
^ F IGURE 8.11. The free-stationing survey method.
all points A , B, 1, 2, . . . , n in a co-ordinate frame of which the u axis is

O A.
As we know, we can solve for the unknown parameters of the Helmert
transformation if at minimum, the co-ordinates of two points, A and B,
are known in both frames.
Using the Helmert transformation (section 3.6) we now transform the
instrumental co-ordinates ( u i , v i ) into terrain co-ordinates ( x i , yi ).
^ 8.6 Carrying out a detail survey

maastomittaus Detail surveys can be carried out as topographic surveys, in which case
they often cover limited areas. The instrument to use then is the elec-
tronic tacheometer. Alternatives are GNSS-RTK — the real-time kinematic
method — or aerial mapping, which however may not always be suitable
on their own due to lack of visibility in the terrain. In local measurements,
traditional prism and tape measurement may be considered, but is used
rarely nowadays due to its low productivity.
^ 8.6.1 Data to be collected, mode of operation

From every measurement station, measure three-dimensionally, for every
point to be measured, the horizontal direction (θ ), zenith angle (ζ), and
slant range ( s). The instrument calculates the topocentric rectangular
co-ordinates ( x, y, z) and performs simple checks itself. When the whole
object has been mapped, a draft printout is made at the site office, and
an overall quality check is carried out.
í Õ ! ¤. û
Carrying out a detail survey 8.6
229
The following data is collected:
General information
◦ site, date, time, weather, observer
◦ identifying codes as needed, see subsection 8.6.2.
From each station point
◦ station point (number, type), instrument height
◦ auxiliary points (number, type), horizontal direction, zenith
angle, distance, prism height
◦ survey points (number, type), horizontal direction, zenith
angle, distance, prism height.
Work phases
1. Choose terrain points based on the terrain and purpose of
use of the measurement. For example, if forming a precise
terrain model of an uneven area, points need to be collected
at sufficient density.
2. Carry out the measurement: collection and pre-processing.
3. Process the material.
4. Present and archive the result. The result is a report on
the measurement work, containing, among other things, a
description of methods used, measurement conditions, point
co-ordinates and their estimated accuracy, a draft map, and
possibly calculations of areas or volumes or other relevant
measurement results.
The working mode is fully digital.
Equipment and software
The processing capacity of total stations is sufficient for many uses.
However,
◦ More and more, one sees a standard tablet or similar, loaded
with versatile software, control the total station wirelessly.
◦ The tablet should be ruggedised for terrain use.
◦ Software guides the whole observation workchain in the field.
◦ Software enables collection, testing, processing and reporting
in the field.
In the computations for topographic surveying, the separate phases of
í Õ ! ¤. û
Terrain −→
Data collection
Total station recording
⏐
↙ ↓
←− −→
Processing in
Processing in Printout of
data logger
micro computer observation
Worksite ⏐
book
⏐ ↘ ↓
↓
x, y
Storage (USB Map printout

stick, DVD)
⏐
↓
http:// ...
−→ −→
Stored data
Server
Office ↘ ↘ Move on-line
x, y
To data base Production

output
X
F IGURE 8.12. Workflow diagram of detail survey. The figure shows some ar-
chaic technologies such as pen plotters and diskettes that time
^ has left behind although the work stages persist.
base-network measurement and mapping should also be distinguished.

Both have their own routines.
^ 8.6.2 Encoding field data

maastotieto- A topographic information system is a special case of a geographic informa-
järjestelmä tion system. It serves the efficient collection of topographic information
by geodetic means for further processing, and thus differs from general
geographic information systems.
As an extreme example of a topographic information system we can
mention the proposal by the Finnish Ministry of Agriculture and Forestry
í Õ ! ¤. û
231
for a national topographic information system (Karlsson, 2015). It is
defined as consisting of the following components (§ 3.1.2):
◦ The national co-ordinate and height system [probably intended are
the horizontal and vertical co-ordinate reference frames].
◦ The topographic database and information management systems maastotieto-
related to it. kanta
◦ All information corpora and information products collected or pro-

duced for the maintenance of the topographic database, and services
maintained in order to obtain suggestions for its maintenance.
◦ All those information and service products that the administrator of
the topographic information system produces from the topographic
database in order to bring it on-line and improve its usability.
Of course most topographic information systems will not be on the na-
tional level.
Topographic survey data is collected in numerical form, and also has, maastomittaus-
besides point measurement data, metadata, data describing data. tieto
The concept of metadata may be illustrated by the example of a topo- maastokartta

graphic map: much more has been depicted on the map than just the
points measured. Points form objects, linear (roads, streets, waterways,
. . . ), area-shaped objects (parcel boundaries, buildings, forests and fields,
. . . ) or three-dimensional (hills and valleys, terrain forms). Everything
is depicted in different ways on the map, and the method of depiction
is documented in the map’s legend.5 The legend is thus in a way the 5
metadata of an ordinary paper map.
Documenting the measurements already in the measurement phase
requires that, at the same time, the metadata is also recorded: does this
point belong to a parcel boundary, is it the edge of a road, is it a tree (and
which species), or is it just a height point in the terrain from which height korkeuskäyrä
contours or earthwork volumes will be calculated? To this end, encoding maamassa
methods or catalogues have been developed, making the transfer of data
easier and as automatic as possible. See for example OGC, Catalogue
Service.
Some aspects must be encoded in the terrain while doing the topo-
graphic survey:
◦ The point numbers (labels) of points measured in the terrain. These
5 Legenda, Latin: what can be read.
í Õ ! ¤. û
Topographic survey
⏐
⏐
↓
Project processing:
◦ point encoding
◦ co-ordinate calculation
◦ volume calculation
⏐
⏐
↓
⏟ ⏞⏞ ⏟
Choice of points to store:
Storage in
◦ transformation (Helmert)
project data base
◦ recoding
⏐
⏐
↓
Storage in general data base
F IGURE 8.13. The encoding process for topographic data. The information may
be stored into a general data base, on a case-by-case basis, for
^ later use.
can also be generated (semi-)automatically.

◦ Point classification codes.
◦ Classification of other objects consisting of points, like lines and
areas.
yksilöivä ◦ Certain identifying attributes.
ominaisuustieto
In connection with computation and data-base entry, the encoding may
be supplemented in some respects, for example with topology data. The
encoding carried out in the terrain is also not final because
◦ Not all terrain points will be entered into general data bases.
◦ A project-specific encoding is not suitable for general use.
Often, project-specific data is not entered into a general data base at

all, but the area is mapped again when the work is done — “as-built”
mapping. The idea is that one maps the finished situation, so no confusion
can arise between what was implemented, and what was only planned
but, after a differing implementation, never re-measured.
See chart 8.13.
í Õ ! ¤. û
233
F IGURE 8.14. Attribute data of objects in multiple layers. With a geographic

information system, spatial data can be efficiently combined,
^ analysed, and refined.
Old encodings in widespread use are based on the classification of the

Zoning Survey Guide, aimed at producing a map. The main starting point
is shared use of geospatial information by the various players in the field.
Geographic information technology creates possibilities for this and
many other efficient uses, such as
◦ The locations of objects are given in the same common co-ordinate
frame. Earlier this was KKJ and its map projections, see subsections
3.2.1 and 3.3.1. Today, it is always EUREF-FIN and its various map
projection co-ordinate frames, see subsections 3.2.3 and 3.3.3. The
transformations between the co-ordinate frames are known. This
facilitates the combination of different objects, producing added
value.
^ T ABLE 8.3. Classification of topographic data.
Data on nature Cultural data Location data Attribute data

Topsoil type Properties Co-ordinate data Identifying a.
Earth’s Buildings, structures (where) Locating a.
surface forms Street and Geometry data Timestamp a.
Soil, bedrock utility networks (what shape) Descriptive a.
Vegetation Zoning Topology data
Waters Street names, (relations with
local names neighbours)
í Õ ! ¤. û
◦ Many methods and tools are on offer for combining, analysing and
refining data from many sources.
ominaisuustieto – Often, various attribute data are presented on different data
layers of a digital map, which can be processed together using
various operators: map algebra, figure 8.14.
– The information carried by objects can be sorted and classified
according to different attributes, for example
∗ all GNSS points on a map sheet
∗ the drilling points in square ( xa , ya ) − ( xb , yb )
kaivonkansi ∗ the manhole covers of the municipal sewer network.
◦ Spatial data may be visualised and, in this way, also made available
to people who are not mapping professionals.
Topographic data can be classified by content into two main information
types: nature data and cultural data (Salmenperä, 1998, sivut 83–84).
Another way of classifying topographic data is as either location data, or
attribute data. See table 8.3.
Many cultural attributes are invisible in the terrain, like place names,
kaavoitus ownership, parcel boundaries, zoning, historical details, and so forth.
1. What is the task of base-network measurement?
2. How does network hierarchy work? Why is it important? What
could go wrong if it is not done properly?
3. What is reconnaissance? What are the requirements of a good point
description?
4. What are commonly used computation or reference surfaces for
geodetic network computation?
5. What terrestrial methods are available for detail surveys? Explain
the strengths and weaknesses.
6. Explain why aerial photogrammetry cannot be the only method for
executing detail surveys.
7. Explain why GPS — for example real-time kinematic GPS, RTK —
cannot be the only method for executing detail surveys.
8. What is metadata, and why is it important? Give an example of
metadata.
í Õ ! ¤. û
^ Construction surveying
9Bergensbanen
Video: Norsk Rikskringkasting

Song: Everloving – Moby
^ 9.1 Zoning plans and setting out

The types of zoning plan in use are regulated by law, in Finland the kaava
Maankäyttö- ja rakennuslaki (“Land Use and Building Act”) 1999/132
(Ministry of the Environment, Legislation on land use and building).
They are the local detailed plan and the local master plan.1 Both are asemakaava
approved by the municipality. In addition, there are still regional plans, yleiskaava
1
which are plans at a higher level.
A local master plan gives the outlines of a land-use plan for a municipal
area. The plans comprise zoning base maps, the scale of which varies kaavan
between 1 : 20 000 − 1 : 10 000 and 1 : 5000 − 1 : 4000. pohjakartta
According to the new Zoning Survey Guide (JUHTA, 2014), three mea- Kaavoitus-
surement classes are defined. Every measurement class has a corre- mittausohjeet
sponding recommended scale for the zoning base map. Digital map
products have no actual scale, but the precision at which map material
is collected needs to be in accordance with the recommended scale. The
co-ordinate and height reference frames are EUREF-FIN and N2000, and
Gauss–Krüger is used as the map projection: ETRS-GK n, where n is the
longitude of the municipality as an integer number.
1. The first measurement class includes local detailed plan areas that
1 Internationally the nomenclature is highly variable: general or comprehensive plan is

also used.
– 235 –
236 9 C ONSTRUCTION SURVEYING
are built-up areas where land is extremely valuable, and where

there is a local detailed plan with a binding parcel division, or a
building prohibition aimed at drafting such a plan. The base-map
scale of the local detailed plan is 1 : 500 or 1 : 1000. In maps that are
intended to be used as part of a municipal geographic information
system and used in technical planning requiring great precision,
the still more precise measurement class 1e may be used.
2. The second measurement class includes local detailed plan areas
that are built-up areas in which the local detailed plan to be drafted
does not require a binding parcel division. The scale of the base
maps is 1 : 1000 or 1 : 2000.
ranta-asema- 3. The third measurement class includes shore detailed plan areas and
kaava lake-shore and sea-shore areas, as well as other such areas where
land is clearly more valuable than land in agricultural and forestry
haja-asutusalue use, such as scattered settlement areas. The base maps are drawn
at a scale of 1 : 2000, in special cases 1 : 4000 or 1 : 5000.
maastoon The setting out onto the terrain of the boundary markers and location
merkintä of buildings must be done according to need before actual construction
starts. The accuracy of the setting out and the method of carrying out the
works is strictly regulated, in Finland in the Zoning Survey Guide. The
runkoverkko base network of the zoned area may have to be brought up to standard.
takymetri In setting out, an automatic total station, real-time kinematic (RTK)
tosiaikainen GNSS positioning, or any other sufficiently accurate measurement tech-
nology is used. The co-ordinates of both the known points and the points
to be set out in the terrain are entered into the instrument’s memory. The
radial-survey or free stationing methods may be used. The instrument
merkitsemis- pre-calculates setting-out measures. Where to place the instrument may
mitat be flexibly decided in the terrain: point intervisibility is not always clear
from the map.
^ 9.2 Setting out and infrastructure

kaavoitus In connection with zoning, the infrastructure of the zoned area is planned:
◦ In the zoning plan, a certain intended use is assigned to an area.

◦ The formation of property organises the land ownership situation
rasite and boundaries as well as easements, like rights of way.
◦ Planning and construction implement the intended use as stated in
the zoning plan, and the area is taken into use.
í Õ ! ¤. û
Setting out and infrastructure 9.2
237
Mapping
⏐
⏐
↓
Zoning −−−→ Test piling

⏐
⏐
↓
Property formation −−−→ Setting out parcel boundary points

⏐
⏐
↓
Construction planning
⏐ ⎧
⏐
↓ ⎪
⎪
⎪ Setting out building locations
⎪
⎨
Construction −−−→ Streets and other infrastructure
⎪
⎪
⎪
⎩ Construction oversight inspections
⎪
^ F IGURE 9.1. Setting out onto the terrain, process description.
The measurements related to the building of municipal technology or

infrastructure — streets, roads, street furniture, utility lines, cables,
among many other things — form their own sub-field of measurement.
Zoning-plan calculation:
◦ The sketch of the plan, a graphical presentation, is presented in
numerical form.
◦ The plan is interpreted as circular arcs and straight line segments jana
starting from known elements, figure 9.2. In the example, first a
polyline A 1 , . . . , A 7 is computed, and to this, the circular arcs K 1 , murtoviiva
. . . , K 7 are fitted.
The zoning-plan boundaries: the boundaries of blocks of houses,
parcels, spaces, general traffic areas, recreational areas as well as
areas to be built on, are to be set out onto the terrain.
◦ Co-ordinates are calculated for the objects to be set out onto the
terrain.
See figure 9.3. Using the inverse geodetic problem, directions and dis- geodeettinen
tances are computed to the points to be set out, reckoned from the location käänteistehtävä
of the instrument. When using the radial-survey method, the place of the
instrument is chosen to be a known point.
The radial-survey method (subsection 8.5.3) in setting out works as
follows:
í Õ ! ¤. û
B
A2 T21 T12
T23 A1
D
T32
R2 R1
A3 K7 K1
R3
T34 R7 T71
K2
K3
T43 R4 A6 A7
K4 T67
K5 R5
A4 T56 R6
E
T45
K6 A
A5
^ F IGURE 9.2. Zoning-plan calculation — an example.
1. The setting-out measures, angle θ i and distance s i , are calculated

using the co-ordinates of points P i , A , and B.
tähys 2. The instrument is set up on point A and aimed at signal B, both
Pi Existing
points
B New
si
points
.
10 m
θi
R = 12 m
R = 30 m
8m
(a) (b)
Radial measurement Application to a road area
^ F IGURE 9.3. Setting out onto the terrain using the radial survey method.
í Õ ! ¤. û
Straight lines, circular arcs, rounding of corners 9.3
239
C P C P1 P2 P3 P4
D D
bC bC b1
bP A b2 b3
A b4
bD bD
aC aC
a1
a −→ B a2 B
aP a3
a4
aD aD
(a) (b)
Principle Several points
^ F IGURE 9.4. Straight setting-out method.
known. The reading on the horizontal circle is set to zero. vaakakehä
3. The telescope is turned until the reading is θ i .

4. The reflective prism is, within the field of view of the telescope,
moved so that its distance reading becomes s i .
5. A pile or stake is driven, a precise marker, on the location of which, paalu
for example, a boundary marker is built.
The free-stationing and right-angle survey methods are also being used.
^ 9.3 Straight lines, circular arcs, rounding of corners
^ 9.3.1 Setting out a straight line into the terrain

The setting-out measures (a P , b P ) of unknown point P from the given
straight line AB must be determined. If we know that point P is on
the straight line CD , we may derive the setting-out measures of point P
directly from the setting-out measures of points C and D : (a C , b C ) and
(a D , b D ), and the position of point P on the straight line CD , the distance
CP :
CP CP
aP = aC + (a − a C ) , bP = bC + (b − bC ) .
CD D CD D
This is the straight method of setting out, for example with a measuring
tape and a right-angle or double pentagon prism.
Alternatively, co-ordinates are used. For example, if the co-ordinates of
points A , B, C and D are already known, then the co-ordinates of point P
are readily computed from the distance CP , and from those, the setting-
out measures (a P , b P ) of point P with respect to line AB are calculated.
í Õ ! ¤. û
T2
t
r
A
θ α
B
K
Pi
θi
ai bi t
r
T1
^ F IGURE 9.5. Rounding of corners with a circular arc.
Note that these setting-out measures are nothing but rectangular co-
ordinates, and calculating them amounts to a co-ordinate transformation
in the plane.
The result is easily generalised to the case in which there are several
points P i on the straight line CD , for all of which setting-out measures
(a i , b i ) are calculated.
^ 9.3.2 Circular arc

Circular arcs are used widely in planning, due to their simplicity. A
circular arc is defined by four parameters, figure 9.5:
◦ angle α between the tangents
◦ half the arc centre angle, θ
◦ arc radius r
◦ tangent length t.
There are two dependencies between these parameters:
α
θ = 100g − ⇐⇒ α = 200g − 2θ ,
2
t = r tan θ ⇐⇒ r = t cot θ .
Setting out the arc onto the terrain is done as follows:

1. Normally the calculation of the straight lines has already deter-
mined the intersection point A of the two tangents, and the angle α.
2. Choose one further parameter: the radius r , and calculate the
others using the equations given above.
í Õ ! ¤. û
Straight lines, circular arcs, rounding of corners 9.3
241
C
T2
ℓ2
D t2 T1 ℓ1
A1
r2
ℓ1 K2
ℓ2
θ2 K 1 F
τ A T2
K2 θ1 ℓ1
r1 K1
E
ℓ1 t1 ℓ2
T1
T3
B A2 ℓ2
(a) (b)
Compound curve Reversed curve
F IGURE 9.6. Rounding of corners with a compound curve. The reversed curve is
^ a special case of the compound curve.
3. Measure the distance t from A along the tangent, yielding the

tangent points T1 and T2 .
4. From both of these determine, with double pentagon prism and
measuring tape, the centre point of the arc K (redundancy!)
5. On line K A now mark the arc point B.
6. From centre point K mark as many arc points as needed, using the
distance r . One example point P i is marked in the figure.
7. The rectangular setting-out measures of point P i : (a i , b i ), are also
readily obtained.
The rounding of corners is commonly done using circular arcs. Combining
straight lines and circular arcs is very common in local detailed zoning
maps, as in building plans in general. The lines are connected according
to the situation using various conditions, which ensures continuity and
the appearance of smooth curvature. Here we present a few examples.
^ 9.3.3 Compound curve

An example case is the connection of two straight lines with two (or more)
circular arcs bending in the same direction, a compound curve, figure korikäyrä
9.6a, Jamal (2017). So, at its simplest we have two straight lines and
two circular arcs, that at their junctions — there are three of them — are
parallel.
The situation of figure 9.6a has the following parameters:
◦ the lengths of the tangents t 1 = AT1 , t 2 = AT2
í Õ ! ¤. û
◦ the radii of the arcs r 1 , r 2

◦ the intersection angle of the tangents τ
◦ the central angles of the arcs θ1 , θ2 .
There are multiple dependencies between parameters:
◦ In triangle △ ADE one sees immediately that ∠ ADE = θ2 , because
K 2 F ⊥ DE and K 2 T2 ⊥ AC , and similarly ∠DE A = θ1 . Therefore
τ + θ1 + θ2 = 200g .
Furthermore T1 E = EF = ℓ1 and T2 D = DF = ℓ2 .
◦ ℓ1 = r 1 tan 12 θ1 , ℓ2 = r 2 tan 12 θ2 .
◦ By the sine rule
AD AE DE t −ℓ t −ℓ ℓ +ℓ
= = =⇒ 2 2 = 1 1 = 1 2
sin θ1 sin θ2 sin τ sin θ1 sin θ2 sin τ
and by substitution
t 2 − r 2 tan 12 θ2 t 1 − r 1 tan 12 θ1 r 1 tan 12 θ1 + r 2 tan 21 θ2
= = .
sin θ1 sin θ2 sin τ
Thus we can calculate all seven parameters catalogued above, if given
◦ two of the three angles τ, θ1 , θ2 , and
◦ two of the four lengths r 1 , r 2 , t 1 , t 2 .
In the case depicted in the figure, the setting out onto the terrain is done
as follows:
1. Measure the distances t 1 − ℓ1 and t 2 − ℓ2 from A along the tangents,
yielding the points E and D .
2. The tangent intersection angles of the individual circular arcs at
points D and E are T1 EF = 200g − θ1 and T2 DF = 200g − θ2 .
3. After this, the setting out is done separately for circular arcs 1 and
2 in the way already explained above.
^ 9.3.4 Reversed curve

The alternative case where the circular arcs bend in opposite directions
(but otherwise the situation is similar to that of the compound curve) is
the S-curve, also reversed curve, figure 9.6b, Jamal (2017). What makes
the situation a little cumbersome is the possibility that the intersection
2 point A of the straight lines2 T1 A 1 and T3 A 2 may not exist, namely if the
lines are parallel.
2 The notation used in the S-curve figure does not directly match that in the compound
curve figure.
í Õ ! ¤. û
Transition curve 9.4
243
L
Clothoid I
P1
Clothoid II
R P2
R
Curvature (1/R )
P3 P2
P1 Distance (L) P3
^ F IGURE 9.7. Principle of the clothoid.
^ 9.4 Transition curve

A transition curve, Euler spiral (Wikipedia, Euler spiral) or clothoid is
used in the planning of railways and fast motorways. In fast traffic, not
only the centre line of the road, but also its curvature3 must be continuous, 3
for the following reasons:
◦ For example, the control of the movement of an articulated lorry
through the steering wheel is slow.
◦ The surface of the road or railway is tilted in the sideways direc-
tion against the centrifugal force. This transversal tilt, which is poikittais-
proportional to the curvature of the road, may change only slowly kallistuma
and continuously along the length of the road.
For these reasons, a combination of straight lines and circular arcs is
unsuitable.
The equation of the clothoid is
RL = A2 ,
3 Curvature is the inverse of the radius of curvature!
í Õ ! ¤. û
in which A is the parameter of the clothoid, L is the distance along the

clothoid, that is, along the road, and R is the local radius of curvature.
As can be seen, the radius of curvature changes as a continuous function
of the distance
A2
R= .
L
4 At constant speed, the centrifugal force4 F is inversely proportional to
the radius of curvature:
v2 v2
= 2 L, F= (9.1)
R A
so, assuming that the design speed of the road, v, is a constant, the
centrifugal force, and thus also the transversal tilt of the road surface
needed, is a linear function of distance travelled. This explains the
suitability of the clothoid curve as the shape of motorway and railway
curves.
Of course, care must be taken that the clothoid satisfies the continuity
condition for the radius of curvature at the start and end points, with
another clothoid, with a circular arc, or with a straight line ( A = R = ∞).
See figure 9.7, in which a straight line links to clothoid I (P1 P2 ), which
links to clothoid II (P2 P3 ), which again continues as a straight line from
point P3 onward. The train arrives at point P1 upright. In the interval
P1 P2 it tilts sideways at a fixed rate, and arrives at its maximum tilt
angle at point P2 . In the interval P2 P3 the tilt diminishes linearly, and
at point P3 the train is again upright, and continues its journey straight
ahead.
If the speed of the train is equal to the design speed v according to
equation 9.1, the resultant of gravity and centrifugal force is always
perpendicular to the floor of the train, and the passengers do not notice
anything.
The clothoid is also used in planning motorways, although there, the
true speeds of vehicles will vary.
^ 9.5 Road and street surveying

Roadbuilding comes with the following measurement and calculation
tasks:
maastomittaus ◦ Conducting a topographic survey in order to draft the road-plan
4 . . . more precisely, the pseudo-force per unit of mass experienced by the travellers in the
vehicle, which acts however from their viewpoint in precisely the same way as gravity.
í Õ ! ¤. û
Construction surveying 9.6
245
base map, for example using aerial mapping (photogrammetry), as
a basis for road planning.
◦ Calculating the trajectories in co-ordinates and setting-out mea-
sures.
◦ Calculating the co-ordinates, setting-out measures, if necessary
earthwork masses and other volumes, for the road structures and maamassa
special structures (bridges, tunnels, underpasses among others).
◦ Carrying out on-site measurements for building the road, setting
out.
◦ After project completion, usually an “as-built” survey is carried out.
The same work phases also occur in other technical measurements related
to large-scale construction projects.
^ 9.6 Construction surveying

Construction surveying measurements include the installation measure-
ments of buildings, parts of buildings, bridges, tunnels, reservoir dams, vesiallas
industrial machines and similar structures.
◦ The measurement starts from the base network. First, a project
measurement base is created, a sufficient set of horizontal and ver-
tical benchmarks in the project area. Base-network points are used kiintopisteistö
as benchmarks. Densification points or “use points” are created as
needed, in a hierarchical fashion.
◦ The actual measurements are carried out from the use points. The
measurements are done separately as horizontal and height mea-
surements, and the point sets are also partly separate.
◦ Both national and international standards have been created to
guide the measurements.
^ 9.6.1 Setting out a building location onto the terrain

When the building permit has been obtained, the builder can apply to the
authorities for a decision to set out the building location onto the terrain. paalutus,
The procedure has three objectives: maastoon
merkintä
◦ Setting out the building location onto the terrain, verifying that no
part of the building is too close to the parcel boundary.
◦ Checking the correctness of the wall measures of the building,
important information for the builder.
í Õ ! ¤. û
◦ Verifying the correctness of the building’s height location.
The land survey authorities only measure with a view to assigning where
to build, unless agreed otherwise. The precision and number of points
is not enough to actually start building. The builder can continue the
survey work in connection with the actual building work.
jonopiste Either a traverse point or a boundary marker is chosen as a site refer-
ence point. If there are not enough of those nearby, or not of the required
precision, the first task is measuring new base points. Map co-ordinates
( x, y) for these are calculated and they are marked on the site plan.
Wall measures, cross-measures, distances from boundaries, and dis-
tances from pre-existing buildings may be used as control measures.
linjavaaitus The height location of a building is obtained by traverse levelling, which
korkeuskiinto- runs from one general height benchmark to another. Close to the building
piste site, at least two height base points are created for the later work, unless
nearby there are already enough general points. The correct height of
a building is critical for functioning sewers and, in low-lying locations,
flood safety.
^ 9.6.2 Location review of a building

sijainti- In location review it is verified that the building is in the correct place and
katselmus at the correct height. The review is carried out when the foundation of
the building is completed. After acceptance, construction may continue.
^ 9.7 Other measurements
^ 9.7.1 Technical measurements, deformation measurement

Technical measurements, or engineering geodesy, form their own spe-
cialised discipline. This also includes precise deformation measurements:
◦ Deformation monitoring, during construction and afterwards over
the object’s life cycle. Objects: reservoir dams, tunnels, bridges,
other large structures, skyscrapers, and so on.
◦ Nowadays often monitoring measurement using automated equip-
ment.
◦ Industrial measurements indoors or outdoors, installation measure-
ments of large machines, paper machines, shipyards.
◦ Engineering surveying.
í Õ ! ¤. û
Other measurements 9.7
247
◦ Tunnel and mine surveying. Tight constraints on the measurement
geometry, as well as the non-functioning of GNSS, are characteristic
of these measurements.
In these measurements, traceability of the quantities measured is central. jäljitettävyys
Careful (system) calibration and good metrological practice are of great
importance.
^ 9.7.2 Machine guidance of working machines

GNSS technology is used widely, in addition to more traditional positioning
technologies, for guiding working machines in real time. The reliability
requirements are obviously tough if an expensive working machine is be-
ing guided to lay, for example, asphalt on a motorway. Machine downtime
is costly, errors even more so.
Terrestrial real-time guidance is usually used in the construction of
reservoir dams, bridges, tunnels, and other infrastructure. This kind
of technology was used in the construction phases of the Dutch Easter myrskyvuoksi-
Scheldt storm-surge barrier (figure 9.8) and the Danish Great Belt and suoja
Iso-Belt
Sound bridge-tunnel solutions, as in many similar projects.
Juutinrauma
Shipping containers and cranes in harbours are also positioned in real
time using GNSS technology in order to improve efficiency (Pitkä, 2009).
Agricultural and forestry machines may be guided by real-time GNSS
(“precision farming”), so seedstock, fertiliser, and pesticides can be ad- täsmäviljely
ministered with precision according to very localised needs.
^ 9.7.3 Underground utility-line mapping

Only part of the underground utility lines, like telephone, data and
electric power cables as well as sewage, water and city heating pipes,
have been mapped satisfactorily, mostly on maps prepared by various
municipal institutions, of which there may be many per municipality.
When utility-line surveys are this decentralised, the geodetic quality of johtokartoitus
the maps will vary.
Cities generally use the presentational and preparation practice of
the 1974 utility-lines mapping standard SFS 3161. The standard was johtokartta-
renewed in 1996. standardi
Utility-line maps are used for many needs: in zoning for planning tech-
nical maintenance and the network, in connection with construction work,
for maintaining the lines by the owner institution, and in connection with
í Õ ! ¤. û
F IGURE 9.8. The AGA/Minilir infrared tracking device — French military

technology! — in use on the building site of the Dutch Easter
Scheldt storm-surge barrier, 1980. © 2016 Nicolàs de Hilster, (De
^ Hilster), figure 4, with permission.
managing crisis situations and damages. The scale is most commonly

1 : 500.
Mapping methods:
kiintopiste ◦ The measurement should be based on general benchmarks
available in the area, so that the result is obtained in the same
system. If necessary, a network densification is carried out,
meaning new benchmarks are created following the Zoning
Survey Guide.
◦ The same measurement methods are used as more generally
kartoitus- in detail survey, section 8.5.
mittaus
◦ The mapping of new lines is done during construction, when
the lines are still visible.
kaivonkansi ◦ The visible parts of old lines (manhole covers, distribution
cabinets) are mapped. The underground parts can sometimes
maatutka be located with a metal detector. Ground-penetrating radar
(GPR) and electrical resistivity tomography (ERT) have also
í Õ ! ¤. û
Self-test questions
249
been used.
Underground utility-lines marking service: The lines are marked johtojen
on the terrain for the builder, in order to avoid costly damage näyttöpalvelu
caused by excavation work. The service is usually offered by the
owner of the line, for example a power company.
1. What is zoning and why is it necessary?
2. How many measurement classes are used in Finland? How are they
defined? What scales are the zoning maps on for each class?
3. Describe the various methods of setting out objects onto the terrain.
4. What is a compound curve? A reversed curve?
5. How many independent parameters uniquely determine a com-
pound curve?
6. Explain why clothoids are used for fast roads and railways.
7. Describe the Finnish practice of “location review”. At what stage of
construction does it take place?
8. What types of objects may deform in ways that can be monitored by
precise measurements?
9. How are underground utility lines mapped?
í Õ ! ¤. û
^ Digital terrain models and volume
calculation
10
Measuring the volume of a wine barrel with a measuring

rod, inspiration for Johannes Kepler’s “wine-barrel formula”,
or Simpson’s rule (image Deutsche Fotothek, 1523)
C ONSTRUCTION IN THE built environment and its planning, as well as

the technical measurements made in those contexts, make extensive use
of digital height and terrain models.
The terms digital height model, DHM, digital elevation model, DEM,
or digital terrain model, DTM, refer to a file consisting of points on the
Earth’s surface, which describes, more or less well, the forms of this
surface. High-resolution terrain models are expensive to produce over
large areas, but are nevertheless available for many countries. Terrain
models for Finland are produced by the National Land Survey.
– 251 –
252 10 D IGITAL TERRAIN MODELS AND VOLUME CALCULATION
Information about the Earth’s surface and its forms can be obtained
maastomittaus using topographic surveying measurements, photogrammetrically from
aerial imagery, and by scanning from aircraft or satellites. The scanner
synteettinen may be laser based or can be a microwave radar, a SAR, or synthetic-
aukko aperture radar, which interferometrically achieves a very high resolution.
Global terrain models

1. The older model GTOPO30 (US Geological Survey, original
1996, USGS, GTOPO30). The resolution is 30′′ , about one
kilometre. The model was updated with data from the SRTM,
see item 4 below. The model contains no sea-depth data.
2. The GLOBE model (Global One-km Base Elevation project
(NOAA and many others, GLOBE Task Team and others, 1999).
The resolution is also 30′′ ≈ 1 km. The current model version
contains no sea-depth data.
3. The ETOPO1 model, which supersedes the older ETOPO5
and ETOPO2 models (NOAA et al. 2008). The resolution is
1′ , about 1.8 km. The model contains depth data in addition
to elevation data.
4. The Shuttle Radar Topography Mission imaged the Earth’s
topography between latitudes 60◦ N and 56◦ S. The flight took
place in February 2000. The resolution is one second of arc
on the Earth’s surface, about 30 m. In 2014, all SRTM data
was declared public. The data does not include sea depths.
ilmalaser- The national terrain model New technologies, like airborne laser
keilaus scanning, have already been operational for many years and are
in widespread use. In Finland, the National Land Survey has also
for a long time been scanning various areas in order to build a
new, precise national terrain model. There are two models: model
KM10 is currently complete at a spatial resolution of 10 m, Finnish
National Land Survey, Elevation model 10 m, and model KM2 is
partially complete at a resolution of 2 m, Finnish National Land
Survey, Elevation model 2 m. The KM2 model is planned to be
completed by 2020. Based on the INSPIRE directive (INSPIRE
Knowledge Base), the data is free of charge.
í Õ ! ¤. û
Terrain-model measurement, construction, presentation 10.1
253
◦
20
20˚
◦
24
24˚
◦ 28◦
28˚
32˚
32 ◦
2 ◦˚
772 722˚
7
◦
0 ◦˚
770 700˚
7
◦
8 ◦˚
668 688˚
◦
6 ◦˚
666 666˚
6
◦
4 ◦˚
664 644˚
6
622˚ ◦
2 ◦˚
662 6
◦
600˚
6
0 ◦˚
660
20˚ ◦ 32˚ ◦
20 24˚
24◦ 28˚
28◦ 32
2010 Oct 21 10:08:11
F IGURE 10.1. The global terrain model ETOPO2 version 2 on the Finnish terri-
^ tory.
^ 10.1 Terrain-model measurement, construction, presentation
^ 10.1.1 Measurement geometry

◦ The points to be measured do not generally form a regular pattern,
and may be freely chosen within the constraints of the measurement
technique used.
◦ The random-point method is a statistical sampling method. The otanta
sampling density may be higher where terrain forms are more
í Õ ! ¤. û
(a) (b)
Connect the measured points. Interpolate the heights to the
“Triangulation” nodes of a regular grid.
“Gridding”
F IGURE 10.2. Presentation of terrain models: triangulated network or point-

^ grid.
variable.
◦ The point density is higher near break lines: linear features in the
terrain at which the terrain slope changes.
^ 10.1.2 Measurement technologies

maastomittaus Geodetic topographic surveying This is a low-productivity tech-
nique that often only complements other methods.
Photogrammetry Points are measured in a stereo model formed from
two aerial photographs, nowadays often automatically by correlat-
ing the digital images. However, ground control points (GCPs) and
katvealue occluded areas are surveyed geodetically.
ilmalaser- Airborne laser scanning This technology collects huge numbers of
keilaus three-dimensional terrain points in the form of a point cloud, from
which the terrain surface can be extracted by suitable processing.
From the point data measured, either a triangulated network or a regular

point-grid is generated.
í Õ ! ¤. û
Use of terrain models 10.2
255
^ 10.1.3 Model presentation
With terrain models, different forms of presentation are used:1 1
Point-grid presentation This agrees well with the way computers op-
erate: the handling of large amounts of data is also straightforward
and easy.
A regular grid may be square, rectangular, or more complicated,
like a hexa tile pattern (“beehive”) or a triangular pattern.
Triangulated-network presentation Here, points representative of
the terrain forms are chosen and connected by lines into a cover of
triangles. A well-known mathematical triangulation technique is
Delaunay triangulation, which gives beautiful triangles of which Delaunay’n
the sides are as equal as possible. kolmiointi
A triangulated network type of terrain presentation is more diffi-

cult to manipulate, but it is also able to present difficult terrain
forms, like sharp edges, better than a grid presentation, using a
smaller number of points. Moreover, if the resolution of the ter-
rain model varies by area, the triangulation presentation is better,
because the sizes of the triangles vary with resolution.
In the literature, the method is referred to as TIN: triangulated
irregular network.
CAD (computer-aided design) software, in use in planning offices, knows

how to use digital terrain models and how to display them in many
different ways; for example as a perspective image. Building plans are
also in digital form and can be combined with this.
Let us also mention in this connection the multi-resolution “tiling”
methods which are based on the discrete wavelet transform (DWT, aalloke-
Wikipedia, Discrete wavelet transform) and are meant for the interactive muunnos
presentation of materials, for example Fraser et al. (2013). The image
format JPEG 2000, as well as Google Earth™, are also based on this
technique. This form of presentation enables rapid changes in viewing
location and zooming. It is even more suited than Delaunay triangulation
for presenting materials of highly variable resolution interactively.
^ 10.2 Use of terrain models

Terrain models are used for, among other things:
1 In image processing one speaks in analogous fashion of pixel and vector graphics.
í Õ ! ¤. û
orto-oikaisu ◦ The ortho-rectification of aerial photographs: the removal of pro-

jection errors of aerial imaging in the production of orthophoto
maps.
korkeuskäyrä ◦ The calculation of the height contours to be shown on a map.
◦ The planning of a traffic route (road, street, waterway, electric
power line, . . . ), to minimise (under other constraints, like maxi-
mum slope or minimum radius of curvature):
maamassa – earthwork volumes to be moved, section 10.4
– the difference between soil to be removed and soil to be added
– consumption of fuel, travel time by a typical vehicle using the
route.
◦ The creation and visualisation of three-dimensional landscape mod-
els, in support of planning and public debate over the plans.
◦ The resolution of visibility issues, like in connection with the place-
ment of cell-phone or radio masts.
◦ The planning of ski pistes.
◦ The support of a military application: automatic navigation in low
flight of cruise missiles, but also jet fighters, hiding from radar
“inside” the terrain.
◦ The calculation of the gravity effect of terrain masses (terrain cor-
rection) in gravity-field and geoid determination.
◦ The provision of the lower boundary in numerical weather predic-
tion (NWP) and climatic general circulation models (GCM).
◦ The creation of realistic landscapes for video games and flight
simulators.
◦ Many others.
Strictly speaking, a digital elevation model (DEM) describes the heights,

latvusto not only of the terrain, but also of buildings, the forest canopy and the
like, whereas a digital terrain model (DTM) describes only the heights of
the terrain itself. Nevertheless the words are often used as synonyms.
^ 10.3 Calculating surface areas

Calculating surface areas is discussed in Kahmen and Faig (1988) in
section 8.6.
merkitsemis- A handy way of calculating is to use setting-out measures with respect
mitat
í Õ ! ¤. û
Calculating surface areas 10.3
257
II
−b3 a3 − +
3
2
a2 b2
III I
a4 1
4 −b4 a1 b1
b −→
↑
⏐a
⏐
^ F IGURE 10.3. The use of setting-out measures in calculating surface areas.
to a baseline. In figure 10.3, the setting-out measures form trapezoids.

Area I is calculated as follows:
A I = 12 (a 2 − a 1 )( b 1 + b 2 ) ,
and area III as follows:
A III = 21 (a 4 − a 3 )( b 3 + b 4 ) ,
in which one should pay attention to the algebraic signs. etumerkki

Area II is also obtained in a similar way, although it is the difference
between two surface areas. It is nevertheless formally a trapezoid:
A II = 12 (a 3 − a 2 )( b 2 + b 3 ) .
All equations can be made compatible by agreeing, for example, that the
b values are positive on the right side seen when moving in the direction
of a, and negative on the left side. Also, the a indices are followed in
numerical order, in the example case, anti-clockwise. Then we obtain for
the total area, by summation,
A = A I + A II + A III + · · ·
and all the algebraic signs, also those of the small cancelling triangles,
are automatically correct.
If we have the use of co-ordinates, there are other ways of calculating
surface areas. The total surface area is obtained as a sum of trapezoids
(the i index is circular: n + 1 is the same as 1):
n
∑︂
A = + 21 ( x i+1 − x i )( yi+1 + yi ) , (10.1)
i =1
í Õ ! ¤. û
x The surface area is x The surface area is x

III+IV I+II+V II+III+IV I+V
– –
4 4
5(= 0) 3 5(= 0)
3
IV
←IV→ ←−III−→ r dθ
1(= 6) III r
x1 2 ←V→ II
x2 r1 2
V II ←−I−→ Integration
←−−I−−→ r2 dθ
y θ12 y y
y1 y2
(a) (b) (c)
In rectangular co-ordinates In polar co-ordinates Continuous integration
^ F IGURE 10.4. Calculating surface area.
and by interchanging x and y:

n
∑︂
A = − 12 ( yi+1 − yi )( x i+1 + x i ) . (10.2)
i =1
From equation 10.1 we obtain

n
∑︂ n
∑︂
A = − 12 x i ( yi+1 + yi ) + 12 x i+1 ( yi+1 + yi ) ,
i =1 i =1
and by re-numbering the second term — as the i index is circular:

n
∑︂ n
∑︂ n
∑︂
A = − 12 x i ( yi+1 + yi ) + 1
2 x i ( yi + yi−1 ) = 1
2 x i ( yi−1 − yi+1 ) .
i =1 i =1 i =1
Similarly equation 10.2 yields

n
∑︂
A = 12 yi ( x i+1 − x i−1 ) .
i =1
These equations are known as the shoelace formulas (Wikipedia, Shoelace

formula).
If equations 10.1 and 10.2 are added together and divided by two, we
obtain
n
∑︂
1
A= 2
( x i+1 yi − yi+1 x i ) .
i =1
í Õ ! ¤. û
Calculating surface areas 10.3
259
F IGURE 10.5. A polar planimeter from 1908 (Wikipedia, Planimeter). The

surface area is measured by drawing along the outer edge of the
^ figure.
This equation calculates the surface of a polygon as the sum of triangles

extending from the origin. It can be shown — a graphical proof is given
in figure 10.6 — that the surface area of such a triangle (example in the
figure) is
A i,i+1 = 21 ( x i+1 yi − yi+1 x i ) = 21 r i r i+1 sin θ i,i+1 . (10.3)
This planimeter equation 10.3 is the principle of operation of the polar

planimeter.2 Of course the equation can also be used directly numerically, 2
if the figure is given in polar co-ordinates:
n
∑︂
1
A= 2 r i r i+1 sin θ i,i+1 .
i =1
y
1
x2 y1 − x1 y2
1
(
2 x2 y1 − x1 y2
)
2
r1
(x1 y2 ) θ12
r2
x
^ F IGURE 10.6. Graphical proof of the planimeter equation.
2 The polar planimeter integrates mechanically the expression

˛
1
2 r 2 (θ ) d θ ,
which is the surface area of the closed figure.
í Õ ! ¤. û
^ 10.4 Volume calculations

maamassa The calculation of earthwork masses is explained in Kahmen and Faig
(1988) sections 14.2 and 14.3. Often, the case is about calculating amounts
of gravel, sand and similar building materials to be moved in connection
with, for example, road-building.
The means of measurement or determination are
pintavaaitus ◦ Area levelling or use of a laser level. Only small projects, labour-
intensive.
”pekkaniska™” ◦ Photogrammetry from the air, or terrestrially from an aerial work
platform.
laserkeilaus ◦ Laser scanning, from the air or terrestrial — high productivity.
◦ Height contours from a map. Computationally, the method is simi-
lar to surface-area calculation, section 10.3.
◦ A digital terrain model.
^ 10.4.1 Simpson’s rule and quadrature

neliöinti A handy method for the numerical integration, or quadrature, of earth-
3 work volume from profile data is Simpson’s rule,3 Kahmen and Faig
(1988) subsection 14.2.1. The equation is
V = 16 (ω1 + 4 ωm + ω2 ) ℓ, (10.4)
in which ω1 , ω2 are the surface areas of the end-point cross-sections, ωm

is the surface area of the midpoint cross-section, and ℓ is the length of
the whole object.
Simpson’s rule can be proven as follows. Let the function to be inte-
grated be f ( x), and let us have at our disposal function values at the
points (−∆ x, 0, ∆ x):
f −1 = f (−∆ x), f 0 = f (0), f 1 = f (∆ x ).
Approximate the function f by a fourth degree polynomial:
f˜︁( x) = a + bx + cx2 + dx3 + ex4 .
3 Thomas Simpson (1710–1761) was the son of an English weaver and an autodidact
mathematician, fellow of the Royal Society. He actually did not invent Simpson’s rule,
although it appeared in his textbook: the rule was already known to Johannes Kepler.
í Õ ! ¤. û
Volume calculations 10.4
261
ω2
ωm
ω1
^ F IGURE 10.7. Simpson’s integration rule in volume calculation.
The integral over the polynomial is

ˆ +∆ x
+∆ x
f˜︁( x) dx = ax + 12 bx2 + 13 cx3 + 41 dx4 + 15 ex5 ⃓−∆ x =
(︁ )︁⃓
−∆ x
= 2a ∆ x + 23 c ∆ x3 + 25 e ∆ x5 . (10.5)
Write as well
f˜︁−1 = a − b ∆ x + c ∆ x2 − d ∆ x3 + e ∆ x4 ,
f˜︁0 = a,
f˜︁1 = a + b ∆ x + c ∆ x2 + d ∆ x3 + e ∆ x4 ,
so that the linear combination
I = p −1 f˜︁−1 + p 0 f˜︁0 + p 1 f˜︁1 = a ( p −1 + p 0 + p 1 ) +

+ b ∆ x + d ∆ x3 (− p −1 + p 1 ) + c ∆ x2 + e ∆ x4 ( p −1 + p 1 ) . (10.6)
(︁ )︁ (︁ )︁
Comparing equations 10.5 and 10.6 shows that, to get I as close as

possible to the integral 10.5, we must choose
p −1 + p 0 + p 1 = 2 ∆ x, − p −1 + p 1 = 0, p −1 + p 1 = 32 ∆ x,
which yields
p −1 = p 1 = 13 ∆ x, p 0 = 43 ∆ x.
Substituting this into equation 10.6 yields
I = 2a ∆ x + 23 c ∆ x3 + 32 e ∆ x5 , (10.7)
and the difference with the integral is

ˆ +∆ x
4
f˜︁( x) dx − I = − 15 e ∆ x5 ,
−∆ x
í Õ ! ¤. û
1 1
X X
0 0
− 21 π 0 1
2π − 12 π 0 1
2π
F IGURE 10.8. Alternatives for quadrature. Left, a mathematically well-behaved

function (cos x, black), Simpson (red) works best. Right, a “jagged”
function. The trapezoid rule (blue) works just as well. Realistic
^ terrain is between these extremes.
a fifth-degree function of the point spacing ∆ x. This means that by

choosing ∆ x small enough, we can get the expression 10.7 very quickly
close to the true value of the polynomial. We may write
I = p −1 f˜︁−1 + p 0 f˜︁0 + p 1 f˜︁1 = 13 f˜︁−1 ∆ x + 43 f˜︁0 ∆ x + 13 f˜︁1 ∆ x =

[︂ ]︂
1
= 6 f −1 + 4 f 0 + f 1 · 2 ∆ x.
˜︁ ˜︁ ˜︁
def def def

Into this we substitute the true values f −1 = ω1 , f 0 = ωm , f 1 = ω2 of
the function, as well as ℓ = 2 ∆ x, and obtain Simpson’s rule 10.4. If the
function f to be integrated is not pathological — but terrain forms could
well be pathological! — then Simpson’s rule will also converge quickly to
it.
Note that Simpson’s rule is exact not only for quadratic, but even for
cubic polynomials.
^ 10.4.2 Alternative quadrature rules

neliöintisääntö Often, a simpler quadrature rule
V = 21 (ω1 + ω2 ) ℓ
(the “trapezoid rule”) works, or even
V = ωm ℓ
(the “rectangle rule”). They do not, however, converge as beautifully as

Simpson’s rule: for both, the error is proportional to the point spacing, or
object length, cubed, ∆ x3 = ℓ3 .
í Õ ! ¤. û
Self-test questions
263
2
ω 4 2
3 1 ω
3 1
h2
h2
h1
h1 h4
h3
h3 ω ω
^ F IGURE 10.9. Volume calculation from digital terrain models.
If the accuracy of Simpson’s rule is insufficient, because the object is too

jagged or “pathological”, one may achieve a better accuracy by dividing
the object into slices, applying some simpler quadrature rule to each of
those, and summing the contributions obtained.
Volume calculation for triangulated-network or point-grid type terrain
models is depicted in figure 10.9. In the case of the triangulated-network
model, the volume of a surface element is evaluated using the equation
h1 + h2 + h3
V =ω .
3
In the case of a point-grid model, the equation to be used is
h1 + h2 + h3 + h4
V =ω .
4
Generalisation: the area of a surface element ω multiplied by the average
height calculated from n corner points
n
1 ∑︂
def
h= n hi.
i =1
These equations are approximate but often sufficient.
1. Which observation techniques are available for collecting terrain
point information useable for the construction of terrain models?
í Õ ! ¤. û
2. Which are the two main techniques for the presentation of terrain
models?
3. Discuss applications of terrain models.
4. Explain how the surface area of a parcel may be determined from
setting-out measures of its boundary.
5. Explain how a polar planimeter works.
6. Explain how the quadrature of volumes from profile data using
Simpson’s rule works.
7. What was the mission concept of the Shuttle Radar Topography
Mission (SRTM)?
í Õ ! ¤. û
^ The third dimension
11 Here as he walked by
on the 16th of October 1843
Sir William Rowan Hamilton
in a flash of genius discovered
the fundamental formula for
quaternion multiplication
i 2 = j 2 = k2 = ijk = −1
& cut it on a stone of this bridge.
Inscription on Broom Bridge, Dublin
^ 11.1 Geocentric co-ordinate reference systems

In modern geodesy, the measurement methods of satellite and space
geodesy are integral parts of the global geodetic observing system (IUGG,
GGOS). Unlike traditional geodetic measurement methods, which carry
out their measurements on, or close to, the Earth’s surface, these mea-
surements are genuinely three-dimensional, and they also require the
use of three dimensions in computations involving them. In addition, the
platforms of, at least, satellite measurements orbit the Earth, meaning
that the centre of mass of the Earth becomes naturally the origin of the
co-ordinate frame used. This is why, in satellite geodesy, we use geocen-
tric, three-dimensional co-ordinate reference systems. See figure 11.1.
Geocentric co-ordinates are often denoted by capital letters, like X , Y , Z .
Geocentric The origin of the system is in the centre of mass of the
Earth, and the Z axis is directed along the rotation axis of the
Earth.
There are two types of geocentric systems:
– 265 –
266 11 T HE THIRD DIMENSION
Z
Rotational motion
Greenwich
Y
Y′
Vernal
equinox X′
X
Greenwich sidereal time θ
F IGURE 11.1. The inertial or celestial ( X ′ , Y ′, Z ) and the terrestrial, co-rotating,

^ or ECEF ( X , Y, Z ) co-ordinate reference systems.
Inertial or celestial There is no rotational motion. The direc-

tions of the axes are fixed with respect to the stars.
kevät- The X axis points (usually) to the vernal or spring equinox,
päiväntasaus the “Greenwich of the sky”.
Terrestrial or co-rotating Also Earth-centred, Earth-fixed, or
ECEF. The directions of the axes are fixed with respect to
the solid, rotating Earth.
The X axis points in the direction of the Greenwich merid-
ian.
kiertokulma Between the inertial and the terrestrial systems there is a rotation angle
Greenwichin called Greenwich sidereal time. It changes rapidly with time, at the same
tähtiaika angular rate as the rotation of the Earth with respect to the stars.
Right-handed co-ordinate frame A corkscrew which progresses in

the positive x direction, turns from the y-axis direction to the
z-axis direction, figure 11.2.
A geocentric system is right-handed if the X axis points in the direction

of the intersection of the planes of the Greenwich meridian and the
equator, the Z axis to the North Pole, and the Y axis to 90◦ east.
Geocentric co-ordinates may be rectangular ( X , Y, Z ), spherical, geode-
í Õ ! ¤. û
Geocentric co-ordinate reference systems 11.1
267
z
F IGURE 11.2. A right-handed co-ordinate frame. A y → z corkscrew (Wikimedia

Commons, Corkscrew) progresses in the x direction — as does a
z → x corkscrew in the y direction, and an x → y corkscrew in the
^ z direction.
tic or geographical, or ellipsoidal co-ordinates. The following relations

exist between them (figure 11.3):
⎡ ⎤ ⎡ ⎤
X cos φ cos λ
⎣ Y ⎦ = r ⎣ cos φ sin λ ⎦ ,
⎢ ⎥ ⎢ ⎥
Z sin φ
in which (φ, λ, r ) , the distance r from the geocentre and the geocentric
latitude and longitude φ and λ, are spherical co-ordinates, and
⎡ ⎤
( N (ϕ) + h) cos ϕ cos λ
⎡ ⎤
X
⎥ ⎢ ( N (ϕ) + h) cos ϕ sin λ ⎥ ⎥,
⎣ Y ⎦=⎢
⎢
⎣ (︂(︁ /︁ )︁ )︂ ⎦
2
Z b a2 N (ϕ) + h sin ϕ
in which (ϕ, λ, h), the height h from the reference ellipsoid and the geode-
tic latitude and longitude ϕ and λ, are geodetic, also called geographical,
co-ordinates. The quantities a and b are the semi-major and semi-minor iso- ja
axes of the Earth ellipsoid; in other words, the equatorial and polar radii. pikkuakselin
puolikkaat
The transversal radius of curvature is
poikittais-
a2 kaarevuussäde
N (ϕ) = √︁ .
a2 cos2 ϕ + b2 sin2 ϕ
The third geocentric co-ordinate type, ellipsoidal co-ordinates, are some-

times used in scientific work, but they have no practical significance in
land-surveying; see Heiskanen and Moritz (1967) pages 39–45.
í Õ ! ¤. û
h
r
φ ϕ X /Y
O
N (ϕ)
F IGURE 11.3. Geocentric (φ) and geodetic (ϕ) latitude and transversal radius of
^ curvature N.
1 The advantage of rectangular1 co-ordinates is that calculations are easy

with them. For example, the distance s between two points ( X 1 , Y1 , Z1 )
2 and ( X 2 , Y2 , Z2 ) is simply2
√︂
2 2 2
s 12 = ( X 2 − X 1 ) + (Y2 − Y1 ) + ( Z2 − Z1 ) .
If the points are given in the form (ϕ1 , λ1 , h 1 ), (ϕ2 , λ2 , h 2 ), the correspond-
ing equation will not be quite as simple!
^ 11.2 Topocentric co-ordinates

In practical measurement work, often local or topocentric, three-dimen-
3 sional co-ordinates3 are used, in which the origin is the location of mea-
koje- surement itself, the instrument (instrument co-oordinates). It is natural

koordinaatit to use spherical co-ordinates (α, ζ, s), in which s is the slant range from
vinoetäisyys
the instrument, α is the azimuth or horizontal direction angle, and ζ is
tähys the zenith angle. From these, the rectangular co-ordinates of the signal
1 Rectangular co-ordinates are often called Cartesian, after René Descartes. Strictly
speaking Cartesian co-ordinates have straight co-ordinate lines, whereas rectangular
co-ordinates could also be curvilinear. In fact, both spherical and geodetic co-ordinates
are rectangular in this more general sense.
2 Of course this is the straight distance in space, often passing though the solid body of
the Earth. Usually we are more interested in the distance over the Earth’s surface.
3 Greek topos = place, like utopia = non-existent place.
í Õ ! ¤. û
Topocentric co-ordinates 11.2
269
Z z S
S
x z
s
y
I x ζ
α
y
Y I
O
F IGURE 11.4. The topocentric or instrument co-ordinate frame, as well as the

geocentric co-ordinate frame. The instrument is I, the geocentre
O and the measured location or signal, S. Note that this co-
^ ordinate frame is left-handed.
or target are easily calculated:

⎡ ⎤ ⎡ ⎤
x sin ζ cos α
⎣ y ⎦ = s ⎣ sin ζ sin α ⎦ .
⎢ ⎥ ⎢ ⎥
z cos ζ
Total stations or electronic tacheometers can give precisely these in-

strument co-ordinates, either in spherical4 (α, ζ, s) or rectangular ( x, y, z) 4
form. Conventionally we write topocentric co-ordinates in lower case.
Figure 11.4 depicts both the co-ordinate axes ( x, y, z) of the topocentric
system, and the axes ( X , Y, Z ) of the geocentric system. In this figure,
the signal S may be a point to be measured on the Earth’s surface, but
also a satellite orbiting the Earth. In any case, the measurements are
always obtained topocentrically first, with respect to the plane of the local
horizon (grey circle) of the observation point I .
4 Strictly speaking only gyrotheodolites can provide the absolute azimuth α. For an
ordinary instrument, the unknown azimuth of the zero on the horizontal circle needs
to be determined separately, typically in a network adjustment or by an astronomical
azimuth determination (section 11.6).
í Õ ! ¤. û
The transformation between these two rectangular three-dimensional

yhden- systems is a three-dimensional similarity of Helmert transformation, on
muotoisuus- which more in the following.
muunnos
^ 11.3 Three-dimensional transformations

A two-dimensional similarity or Helmert transformation is, if the turn
from the X axis to the Y axis has the same direction as a positive angle
α: [︄ ]︄ [︄ ]︄[︄ ]︄
X′ cos α sin α X − X0
=K ,
Y′ − sin α cos α Y − Y0
[︂ ]︂T
in which α is the rotation angle, K the scale ratio, and X 0 Y0 are
the co-ordinates of the origin of the new system, written in the old one.
The corresponding three-dimensional transformation equation is ob-
tained by adding the Z axis and keeping its direction fixed:
⎡ ⎤ ⎡ ⎤⎡ ⎤
X′ cos α3 sin α3 0 X − X0
⎣ Y ⎦ = K ⎣ − sin α3 cos α3 0 ⎦ ⎣ Y − Y0 ⎦ .
⎢ ′ ⎥ ⎢ ⎥⎢ ⎥
Z′ 0 0 1 Z − Z0
kiertomatriisi The size 3 × 3 rotation matrix visible in the equation may be called R 3 (α3 ).
In the same way as around the Z axis, rotations may also take place
around the Y or X axis. In that case, we obtain analogously the rotation
matrices
⎡ ⎤ ⎡ ⎤
1 0 0 cos α2 0 − sin α2
R 1 (α1 ) = ⎣ 0 cos α1 sin α1 ⎦ , R 2 (α2 ) = ⎣ 0 1 0 ⎦.
⎢ ⎥ ⎢ ⎥
0 − sin α1 cos α1 sin α2 0 cos α2
The general similarity or Helmert transformation containing all three

kierto, siirto rotations (and three translations, and a scaling) can now be expressed in
the following compact form:
′ (︁ )︁
R = K R R − R0 . (11.1)
Here we write the vectors as column vectors of their components or co-

ordinates of location, as follows:
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
X′ X X0
′
R = ⎣ Y ′ ⎦, R = ⎣ Y ⎦, R0 = ⎣ Y0 ⎦ .
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
Z′ Z Z0
í Õ ! ¤. û
Transformation in the case of small rotation angles 11.4
271
The overbar signals that we are dealing with column vectors of component
values, not the vectors in space themselves. It may be left off when the
context is clear.5 5
Furthermore, the rotation kierto
R = R 3 (α3 ) R 2 (α2 ) R 1 (α1 )
is the combination, or “chaining”, of three rotations. Equation 11.1 is

called the (three-dimensional) Helmert or similarity transformation. The
elements of the size 3 × 3 matrix R representing the rotation of the co-
ordinate frame are complicated trigonometric expressions in the angles
α1 , α2 , and α3 , and we do not derive them here.
^ 11.4 Transformation in the case of small rotation angles

Often, the axes of two co-ordinate frames are very close to each other. In
that case, the rotation angles are small and one may make the approxi-
mation that sin α ≈ α and cos α ≈ 1. Then, the equations become simpler.
If, in addition, it may be assumed that the scale ratio K is close to unity,
one may write
K = 1 + m,
in which m, the scale distortion, is small.

In addition
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
1 0 0 1 0 −α2 1 α3 0
R1 ≈ ⎣ 0 1 α1 ⎦ , R 2 ≈ ⎣ 0 1 0 ⎦ , R 3 ≈ ⎣ −α3 1 0 ⎦.
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
0 −α1 1 α2 0 1 0 0 1
If all α i are small, one may furthermore assume that all α i α j ≈ 0.
5 The vectors themselves are
R = k ( X i + Y j + Z k) = R0 + k′ ( X ′ i′ + Y ′ j′ + Z ′ k′ ) , R0 = k ( X 0 i + Y0 j + Z0 k) ,
{︁ }︁ {︁ }︁
in which i, j, k and i′ , j′ , k′ are orthonormal bases of the space, for the old and the
new co-ordinate frame respectively. Also K = k k′ , the ratio of the scale deformations
/︁
of the two frames. For a pure rotation this becomes
R = X i + Y j + Z k = X ′ i′ + Y ′ j′ + Z ′ k′ .
í Õ ! ¤. û
We obtain
⎡ ⎤
1 α3 −α2
R = R 3 (α3 ) R 2 (α2 ) R 1 (α1 ) = ⎣ −α3 1 α1 ⎦ = I + ∆R,
⎢ ⎥
α2 −α1 1
in which I is the 3 × 3 unit matrix, and

⎡ ⎤
0 α3 −α2
∆R = ⎣ −α3 0 α1 ⎦
⎢ ⎥
α2 −α1 0
is a skew-symmetric (antisymmetric) matrix: ∆R T = −∆R .

Now we obtain from equation 11.1:
′
R = (1 + m)( I + ∆R ) R − R0 ≈ ( I + mI + ∆R ) R − R0
(︁ )︁ (︁ )︁
′
=⇒ R − R = (mI + ∆R ) R − ( I + mI + ∆R ) R0 .
siirto If we now also assume the translations R0 to be small, then from this
follows the co-ordinate correction equation
⎡ ⎤
m α3 −α2
′
R − R ≈ ( mI + ∆R ) R − R0 = ⎣ −α3 m α1 ⎦ R − R0 , (11.2)
⎢ ⎥
α2 −α1 m
′ ′
in which m, α1 , α2 , α3 , R0 , and R − R are all small (but R and R are
large).
The form 11.2 is the “small” form of the general Helmert transforma-
tion, between two realisations of co-ordinate reference systems that are
close together, like, for example, the different realisations of ITRS, the In-
ternational Terrestrial Reference System. In that case, the rotation angles
α i are of the order of a fraction of a second of arc, and the translation
vector R0 is under 10 cm.
^ 11.5 The transformation between two reference ellipsoids

A classical case is the transformation between two geodetic datums which
are defined on two differently non-geocentric reference ellipsoids, for
example in Europe between the Hayford ellipsoid of the ED50 datum
and Eastern Europe’s Krasovsky ellipsoid. A brute-force method is then
to first convert geodetic co-ordinates (ϕ, λ, h) to rectangular ( X , Y, Z ),
carry out a three-dimensional Helmert transformation between the two
í Õ ! ¤. û
The transformation between two reference ellipsoids 11.5
273
datums, and convert the result back into geodetic co-ordinates in the
other datum, (ϕ′ , λ′ , h′ ).
If the difference between the two datums is small and consists only of
a shift of the reference ellipsoid’s centre, there is a simple relationship
between the centre shift and the change in geodetic co-ordinates. See
figure 11.5.
Let the location vector of a point from the centre of one ellipsoid be
R1 , and from the other, R2 , and the difference, in rectangular geocentric
co-ordinates,6 6
⎡ ⎤ ⎡ ⎤
X2 − X1 ∆X
∆R = R2 − R1 = ⎣ Y2 − Y1 ⎦ = ⎣ ∆Y ⎦ .
⎢ ⎥ ⎢ ⎥
Z2 − Z1 ∆Z
Around the point, we define local topocentric co-ordinates ( x, y, z) on a

{︁ }︁
unit-vector, or orthonormal, basis N, E, U (”north, east, up”). At the ortonormaali
location of the point, the principal radii of curvature of the reference kanta
ellipsoid are M (ϕ), the meridional radius of curvature, and N (ϕ), the
transversal radius of curvature. Now the topocentric co-ordinate differ-
ences are
∆ x = M (ϕ) ∆ϕ, ∆ y = N (ϕ) cos ϕ ∆λ, ∆ z = ∆ h,
and
∆R = N ∆ x + E ∆ y + U ∆ z = N M (ϕ) ∆ϕ + E N (ϕ) cos ϕ ∆λ + U ∆ h.
In matrix notation, this is

⎡ ⎤ ⎡ ⎤
∆X [︂ ]︂ M (ϕ) ∆ϕ
∆R = ⎣ ∆Y ⎦ = N E U ⎣ N (ϕ) cos ϕ ∆λ ⎦ =
⎢ ⎥ ⎢ ⎥
∆Z ∆h
⎡ ⎤⎡ ⎤
NX E X UX M (ϕ) ∆ϕ
= ⎣ NY EY UY ⎦ ⎣ N (ϕ) cos ϕ ∆λ ⎦ ,
⎢ ⎥⎢ ⎥
NZ EZ UZ ∆h
in which the matrix is orthogonal, in fact a rotation matrix:

⎡ ⎤ ⎡ ⎤
NX E X UX − sin ϕ cos λ − sin λ cos ϕ cos λ
R = ⎣ NY EY UY ⎦ = ⎣ − sin ϕ sin λ cos λ cos ϕ sin λ ⎦ .
⎢ ⎥ ⎢ ⎥
NZ EZ UZ cos ϕ 0 sin ϕ
í Õ ! ¤. û
Mer
id ian
N ∆x U ∆z
∆λ
∆h
N cos ϕ E∆y
M
Parallel R
ure
vat
N
cur
re
tu
s of
a
u rv
diu fc
so
l ra
u
a di
∆ϕ lr
ona
a
rs
ridi
e
sv
an
Me
Tr
ϕ
cos ϕ ∆λ
−
90 ◦
F IGURE 11.5. The differential connection between rectangular (N, E, U) co-

^ ordinates and geodetic co-ordinates on the reference ellipsoid.
The columns of the matrix are the geocentric components of the vectors
N, E, and U.
Inverting an orthogonal matrix is easy: R −1 = R T , or
⎡ ⎤ ⎡ ⎤⎡ ⎤
M (ϕ) ∆ϕ − sin ϕ cos λ − sin ϕ sin λ cos ϕ ∆X
⎣ N (ϕ) cos ϕ ∆λ ⎦ = ⎣ − sin λ cos λ 0 ⎦ ⎣ ∆Y ⎦ .
⎢ ⎥ ⎢ ⎥⎢ ⎥
∆h cos ϕ cos λ cos ϕ sin λ sin ϕ ∆Z

Thus we may easily calculate what are the effects of shifting the centre
of the reference ellipsoid on geodetic co-ordinates (ϕ, λ, h), evaluated on
the ellipsoid:
⎡ ⎤ ⎡ ⎤ ⎡ (︁ )︁−1 ⎤ ⎡ ⎤
ϕ2 ϕ1 M (ϕ) 0 0 X2 − X1
)︁−1 T
⎣ λ2 ⎦ = ⎣ λ1 ⎦ + ⎣
(︁
0 N (ϕ) cos ϕ 0 ⎦ R ⎣ Y2 − Y1 ⎦ .
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
h2 h1 0 0 1 Z2 − Z1
6 This is also the vector connecting the centres of the two reference ellipsoids!
í Õ ! ¤. û
Laplace azimuth measurements 11.6
275
X
Ellipsoidal normals
ϕ1 , λ1 ϕ2 , λ2
ξ1 , η 1 Plumb line Φ, Λ
ξ2 , η 2
h1 H
Geoid
N1
h2
N2 Datum
transformation
E ll Ell
ips ips
oid oid
2 1
F IGURE 11.6. The effect of a datum transformation (shift of the centre of the
reference ellipsoid) on geodetic latitude and longitude ϕ and λ,
deviations of the plumb line ξ and η, geoid heights N, and the
^ heights h of points from the reference ellipsoid.
And when the geodetic co-ordinates (ϕ, λ, h) change, then the deviations luotiviivan
of the plumb line and the heights of the geoid also change, their definitions poikkeama
being
ξ = Φ − ϕ, η = (Λ − λ) cos ϕ, N = h − H, (11.3)
in which (Φ, Λ) are astronomically determined latitude and longitude,

(ξ, η) are deviations of the plumb line in the north and east directions, h
is the height above the ellipsoid, and H the height above sea level, that
is, the geoid, while N is the geoid height reckoned from the reference
ellipsoid. From this is obtained directly
⎡ ⎤ ⎡ ⎤⎡ ⎤
− M (ϕ) ∆ξ − sin ϕ cos λ − sin ϕ sin λ cos ϕ ∆X
⎣ − N (ϕ) ∆η ⎦ = ⎣ − sin λ cos λ 0 ⎦ ⎣ ∆Y ⎦ ,
⎢ ⎥ ⎢ ⎥⎢ ⎥
∆N cos ϕ cos λ cos ϕ sin λ sin ϕ ∆Z

because Φ, Λ, and H can be calculated directly from measurements,
without any use of a reference ellipsoid. See figure 11.6.
^ 11.6 Laplace azimuth measurements

A geodetic network computed on the reference ellipsoid is oriented by
astronomical observations. Two of the three orientation degrees of free-
í Õ ! ¤. û
Plumb line Ellipsoidal normal
Horizon
Ellipsoid
tangent
plane Geoid
Reference ellipsoid
F IGURE 11.7. The deviation of the local plumb line from the normal on the ref-
erence ellipsoid surface. The general situation in a point neither
^ on the reference ellipsoid nor on the geoid.
dom are fixed by making the direction of the ellipsoidal normal in the
datum point or points equal to the astronomically determined direction
luotiviiva of the plumb line. That leaves a third degree of freedom, the network’s
orientation with respect to the local north.
The directions measured in the network are projected onto the plane of
the local horizon. If one ignores the local deviations of the plumb line —
meaning that one assumes that the local plane of the horizon is parallel
to the local tangent plane to the reference ellipsoid — one may say that
the sighting azimuths are projected onto the reference ellipsoid. This
assumption is, however, not correct. The local horizon is perpendicular to
local gravity, the direction of which differs a little from that of the normal
to the reference ellipsoid. The phenomenon is precisely the plumb-line
deviation, see figure 11.7.
The deviations of the plumb line are ξ in the south-north direction, and
η in the west-east direction. Their equations 11.3 were already given:
ξ = Φ − ϕ, η = (Λ − λ) cos Φ,
with (Φ, Λ) astronomical latitude and longitude. ϕ and λ, geographical

latitude and longitude, are geodetically computed co-ordinates such as can
7 be found on maps, computed with respect to a certain reference ellipsoid.7
Let the astronomical azimuth (absolute direction) measured with re-

spect to the plane of the local horizon be A , and the same sighting mea-
7 So this means that the deviations of the plumb line will depend on the reference
ellipsoid chosen. The choice of the local reference ellipsoid or datum was often made so
that the sum of the squares of plumb-line deviations was minimised over the area of
interest. In other words, so the ellipsoid would fit as well as possible to a level surface
of the very local gravity field, the geoid.
í Õ ! ¤. û
Traditional “2D+1D” co-ordinates 11.7
277
sured with respect to the reference ellipsoid, that is the geodetic azimuth,
α.
Then we may say that
A − α = η tan Φ + (ξ sin α − η cos α) cot ζ =

= (Λ − λ) sin Φ + (Φ − φ) sin α − (Λ − λ) cos α cos Φ cot ζ, (11.4)
(︁ )︁
in which ζ is the zenith angle. Equation 11.4 is called the Laplace azimuth
equation. If the sighting direction is in the horizontal plane, then cot ζ = 0
and the correction above is a constant for the observation site, as the
dependence on azimuth α vanishes.
Figure 11.8 explains where both terms come from:
◦ The first term η tan Φ comes from the projection of the direction
of the celestial pole onto the local horizon being different from
the projection onto the tangent plane to the reference ellipsoid. It
depends on the height of the celestial pole, or astronomical latitude
Φ.
◦ The second term (ξ sin α − η cos α) cot ζ comes from the difference in
projections of the signal direction onto the local horizon and onto
the tangent plane of the ellipsoid. It depends on the zenith angle ζ
of the signal, and vanishes if ζ = 100g .
^ 11.7 Traditional “2D+1D” co-ordinates

Co-ordinate frames in which horizontal location and height are given
separately have long been in widespread use. An example of this is
the KKJ system, the Map Grid Co-ordinate System, which was in use
in Finland but is now obsolete, and the height system N60. KKJ gives
horizontal co-ordinates in the Gauss–Krüger projection on the Hayford
ellipsoid, also known as the International Ellipsoid of 1924. The co-
ordinates are based on the ED50 (European Datum 1950) system, which
was created in 1950 by a joint adjustment of the triangulation networks yhteistasoitus
of all Western European countries. This is a traditional, non-geocentric kolmiomittaus-
verkko
datum.
The N60 system gives orthometric heights, meaning heights from the
geoid, not the reference ellipsoid. The geoid (see section 16.4) is an undu-
lating reference surface similar to mean sea level. With the International
Ellipsoid, one must use the Bomford8 geoid model, which back in the day 8
í Õ ! ¤. û
Zenith
Ellipsoidal Plumb Polaris

normal line
n
Horizo
Φ
Observer
A
α
Ell Signal
tan ipsoid
gen
t pl
ane
X
(a)
Part related to pole elevation Φ
Hic sunt leones
Zenith
Polaris
Ellipsoidal
normal Ellips
oid
tange
o n nt pla
Signal H oriz n e
Plumb
line
ζ
Observer
A
α
(b)
Part related to signal elevation 100g − ζ
F IGURE 11.8. The Laplace phenomenon: the effect of the plumb-line deviation
^ on the azimuth.
í Õ ! ¤. û
Traditional “2D+1D” co-ordinates 11.7
279
was determined in connection with the ED50 project.
Stating the three-dimensional location of a point in the form ( x, y, H ),
in which ( x, y) is a KKJ co-ordinate pair, and H an orthometric N60
height, is problematic: the connection with the systems used by satellite
positioning is complicated. Transforming the co-ordinates ( x, y, H ) into
geocentric ( X , Y, Z ) co-ordinates involves the following steps:
1. The KKJ co-ordinates ( x, y) have already undergone a two-dimen-
sional Helmert transformation, aimed at achieving an approximate
compatibility with the still older VVJ or “Helsinki System”. This
transformation is documented in Ollikainen (1993). The inverse of
this transformation needs to be applied: ( x, y) =⇒ ( x′ , y′ ).
2. The Gauss–Krüger map projection is applied in the inverse direc-
tion, ( x′ , y′ ) =⇒ (ϕ, λ), on the International or Hayford ellipsoid.
3. In order to transform the orthometric height H into a height h from
the reference ellipsoid, we need a geoid model that is compatible
with the reference ellipsoid used, thus Bomford’s geoid model. The
geoid height N = h − H is needed in every point.
4. The geodetic co-ordinates (ϕ, λ, h) are to be transformed into rect-
angular ( X ′ , Y ′, Z ′ ) co-ordinates. These three-dimensional co-ordi-
nates are still in the European Datum 1950 system, which is not
(precisely) geocentric.
5. As the ED50 datum is non-geocentric, we still need a three-dimen-
sional Helmert transformation to arrive at precisely geocentric
co-ordinates ( X , Y, Z ), see section 11.8 for details.
KKJ has already been replaced by new map projection systems, which
are based either on the Gauss–Krüger or the UTM (Universal Transverse
Mercator) projection on the GRS80 reference ellipsoid, and by the EUREF-
FIN datum, the ETRS89 (European Terrestrial Reference System 1989)
system’s Finnish national realisation. The new N2000 system is used as
the height system, and its connection to the ellipsoidal heights is given by
the geoid model FIN2005N00, see Bilker-Koivula and Ollikainen (2009).
The above points 2–4 continue to apply, albeit with other names.
8 Brigadier Guy Bomford (1899–1996) was a gifted British geodesist and student of geoid
determination.
í Õ ! ¤. û
^ T ABLE 11.1. Transformation parameters from Ollikainen (1993) for EUREF89

→ ED50.
Parameter Value Precision Unit
∆X 93.477 ±3.345 m
∆Y 103.453 ±5.534 m
∆Z 123.431 ±2.736 m
′′
ex −0.246 ±0.168
′′
ey 0.109 ±0.106
′′
ez 0.068 ±0.112
m −2.062 ±0.417 ppm
^ 11.8 Case: the transformation between ED50 and EUREF89

This is a Helmert transformation of type 11.1, to an accurately geocentric
system. Because the Hayford ellipsoid on which ED50 is based is not
(accurately) geocentric, the origin of the co-ordinate frame must be shifted
from the centre of the reference ellipsoid to the centre of mass of the Earth.
The shifts (translations) are of the order of a hundred metres, and the
rotations and the scale change are also substantial, see Ollikainen (1993)
page 15 and table 2 on page 13:
⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤
X 1 ez −e y X ∆X
⎣ Y ⎦ = (1 + m) ⎣ − e z 1 ex ⎦ · ⎣ Y ⎦ + ⎣ ∆Y ⎦ ,
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
Z
ED50
ey −e x 1 Z
EUREF89
∆Z
for which table 11.1 gives the transformation parameters according to

Matti Ollikainen’s solution for the territory of Finland.
ED50 (European Datum 1950) is the traditional European datum on
which KKJ is based. It was created well before the satellite age. As can
be seen from the table, it is non-geocentric. EUREF89 is a modern, GNSS-
based European co-ordinate reference frame. The precision figures given
in the table are large because co-ordinates determined in the traditional
way over a large area are just not very precise.
More recent information on matters of co-ordinates and transforma-
tions for the Finnish territory can be found in Häkli et al. (2009).
^ 11.9 Case: the transformation between ITRF and ETRF

A three-dimensional, satellite-based, thus geocentric co-ordinate refer-
ence frame called EUREF-FIN is in use in Finland. It is the national
í Õ ! ¤. û
Self-test questions
281
^ T ABLE 11.2. Transformation parameter values ITRF2005 → ETRF2005 from
Boucher and Altamimi (2007) tables 3 and 4.
Parameter Value (cm) Parameter Value (10−3 ′′/a)
T1 5.6 ̇1
R 0.054
T2 4.8 ̇2
R 0.518
T3 −3.7 ̇3
R −0.781
realisation of ETRS89, the European Terrestrial Reference System, on the

Finnish territory.
All geodetic satellite measurements, however, give a position solution in
the same frame as that in which the orbital elements of the GPS satellites rata-alkiot
are given, like ITRF2005. Then, the following transformation is needed to
the corresponding ETRS89 realisation, ETRF2005:
⎡ ⎤ ⎡ ⎤ ⎡ ⎤ETRF2005
X X T1
⎣ Y ⎦ ( t) = ⎣ Y ⎦ ( t) + ⎣ T 2 ⎦ +
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
Z Z T3
ETRF2005 ITRF2005 ITRF2005
⎡ ⎤ETRF2005 ⎡ ⎤
0 −Ṙ3 Ṙ2 X
⎢ ̇ ̇1 ⎥
+ ⎣ R3 0 −R × ( t − 1989.0) × ⎣ Y ⎦ ( t) ,
⎢ ⎥
⎦
−Ṙ2 Ṙ1 0 Z
ITRF2005 ITRF2005
in which the dot on the R parameters (Newton’s dot notation) indicates
derivation with respect to time. The R ̇ parameters in this equation
include the tectonic motion of the Eurasian plate.
The parameter values for the equation are found in the instructions
written by the EUREF subcommission, and are presented in table 11.2.
As can be seen, the transformation parameters in this case are many
orders of magnitude smaller than in the earlier-described transforma-
tion between EUREF-FIN and ED50. Both co-ordinate reference frames,
ETRF2005 and ITRF2005, are geocentric on the centimetre level.
1. Which are the two main types of geocentric co-ordinate reference
systems?
2. What is sidereal time and what does it describe? tähtiaika
3. How many transformation parameters does a three-dimensional

Helmert transformation have?
í Õ ! ¤. û
Ellipsoidal
λ=0
normal Plumb
Λ=0
line
Plumb-line
deviation
Λ=0
λ=0
Geoid
Reference
ellipsoid
^ F IGURE 11.9. Greenwich geometry: longitude is a direction, not a place.
4. Name the parameters of the three-dimensional Helmert transfor-

mation.
5. Describe the Laplace azimuth measurement and what it is used for.
^ Exercise 11–1: Greenwich: explain this

People have been taking their inexpensive hand-held GNSS receivers —
mobile phones, even — to the Greenwich or zero meridian, figure 2.11,
finding that they do not show zero longitude.
How is that possible?
There is a lot of good explanation for this on the Internet, including one
Journal of Geodesy article from 2015 — and a lot of plain old tabloid non-
sense. Do not buy the nonsense. In this exercise, provide an explanation
in your own words, showing that you “get” it.
Imagine that you’re called up by a journalist who has heard about this,
and wants your take on it as a geodesist. Your explanation — an elevator
speech — should be so lucid that she “gets” it, and what’s more, that when
she gets home and tells her husband what she has learned, he “gets” it
too. . .
í Õ ! ¤. û
^ The Global Positioning System (GPS)
12
Wouldn’t it be noteworthy that Transit 5B-5 (launched
1964) is the oldest active satellite in space? It never went
into operation because the data up/download link system
failed after deployment into LEO. But since that day it
transmits its doppler carrier and marker signals on 136.650
MHz, of course the batteries failed many years ago and so
the electronics only function when the satellite is in
sunlight. But it still manages to boot up most days, after
more than 50 years of operation, and can be received using
simple handheld scanners. It’s considered an admirable
“space monument” by sat enthusiasts.
176.0.30.145, November 29, 2015. Wikipedia, Transit 5B-5

— oldest active satellite
I N LAND - SURVEYING, THE ROLE of the GPS has during the last two or
three decades grown to be dominant, both in Finland and worldwide. The
professional literature, especially in the English language, is extensive.
Poutanen (2017) is a significant work in Finnish. A good basic work in
English is Hofmann-Wellenhof et al. (2001).
Here we will concentrate on the GPS, which has been fully operational
for a long time. In parallel to this system, which is operated by the US
military authorities, similar systems by other countries have in recent
years appeared on the scene. The Russian GLONASS deserves a special
mention: after a time of decay it has now grown back to operationality.
Satellites from both systems are today routinely used together in land-
surveying.
The Europeans have developed their own Galileo system, and the
Chinese their BeiDou or “Compass” system. Satellites of both systems
– 283 –
284 12 T HE G LOBAL P OSITIONING S YSTEM (GPS)
are functioning in orbit and both systems are close to full operationality.
These systems are, in the aggregate, called “GNSS” or Global Navigation
Satellite Systems. However, the relative simplicity of the way the original
GPS works makes it a suitable model for teaching the basics, and therefore
we shall concentrate on it.
The GPS is originally a navigation system. It was not the first radio
navigation system using satellites: an earlier satellite navigation system
was the Transit system, or NNSS (Navy Navigation Satellite System),
unofficially the “Doppler positioning system”. This system, which was op-
erational 1964–1996, consisted of five satellites orbiting the Earth in low
orbits. Geodetically useful positioning required several satellite passages
overhead, amounting in practice to at least 24 hours of observations.
See figure 12.3 for a visual illustration of the NNSS.
The GPS satellites are in much higher orbits, and anywhere on Earth,
at almost any time, at least six of them are “visible”. More commonly,
1 the number of visible satellites ranges from eight to over ten.1 Therefore
GPS positioning can be done almost instantly, within a few seconds or

minutes.
^ 12.1 Radio navigation and hyperbolic systems

Of the older terrestrial radio navigation systems for seafarers may be
mentioned the Decca system, closed down in 2000, which is an example
of a hyperbolic system. Other systems worth mentioning are Loran-C and
2 Omega, both no longer in use.2
kantoaalto Decca transmitted non-modulated carrier waves in the frequency band

70–129 kHz, corresponding to wavelengths 2.3–4.3 km. Its use required
at least three transmitters, one “master” and at least two “slave” or
auxiliary stations. The transmissions of the stations were precisely syn-
chronized, though they were all transmitting on different wavelengths.
The term “hyperbolic system” is based on the circumstance that a vessel
without a clock synchronised with the transmitters can observe only the
1 Positioning instruments making use of more than one satellite system, like GPS and
GLONASS, “see” even more satellites, even in poor-visibility locations like the centres of
big cities.
2 Some countries, including the US, however, are considering bringing Loran-C back
to life under the name “eLoran” (“Enhanced Loran”), because, using high-power long-
wavelength radio transmitters, it is less susceptible to interference compared to GPS.
í Õ ! ¤. û
Radio navigation and hyperbolic systems 12.1
285
Auxiliary
s red
RED GREEN
Red lane
s master
s green
Master Green lane

Auxiliary
Harbour
^ F IGURE 12.1. The Decca system.
difference in reception times between the waves from two transmitters.

See figure 12.1.
On the map are drawn, in two different colours, hyperbolas, curves
of which the points have a difference in distance from the master and a
certain auxiliary that is constant. For example, for the red hyperbolas we
^ F IGURE 12.2. A Decca receiver. Wikimedia Commons, Decca Navigator Mk 12.
í Õ ! ¤. û
have
s red − s master = constant
and for the green hyperbolas
s green − s master = constant.
Every curve has its own constant. Unfortunately this constant cannot be
3 observed, because all carrier waves look the same.3
kujalaskuri Therefore, an incremental measurement method is used. Each lane

counter of the Decca instrument has two hands, one for counting the lanes,
the other for the fraction within a lane. The hands move together, just
like the hands of a clock.
The ship must set the lane counters to the correct starting values at a
known location, such as the port of departure. After that, during cruising,
they follow the development over time of the lane values: every lane hand
vaihe-ero on the phase-difference dial tracks how many full turns the phase hand
has made — see figures 12.1, 12.2. It is a precondition that the radio
tukiasema connection with the base stations stays uninterrupted.4
4
At any point in time, one may use the two5 lane numbers and residual
5 [︁ )︁
phase differences — fractional numbers ∈ 0, 2π — to read one’s own
location on a sea chart, on which the hyperbolas are pre-drawn.
Like the GPS, the Navy Navigation Satellite System NNSS was also
a hyperbolic system: the hyperbolas were formed with two successive
satellite locations in space, for times t 0 and t 0 + 2 min, as the focal points.
The difference between the distances to the two locations is obtained by
integrating the Doppler shift of the received frequency over this time
span, yielding the difference in cycles. The equation is
ˆ t 0 +2 min
c
s( t 0 ) − s( t 0 + 2 min) = ( f ′ − f ) dt,
f t0
in which f is the nominal frequency of the transmission, f ′ the received

frequency, and c the speed of light. This requires the nominal transmis-
sion frequency to be accurate: the NNSS system was also the first satellite
system that served the dissemination of precise time.
pyörähdys- Every hyperboloid of revolution in space intersects the Earth’s surface
hyperboloidi
3 In connection with the GPS this is called the ambiguity problem.
4 This is referred to as a kinematic method. Absolute location is not measured: instead
the user follows continuously how they are moving in relation to a known starting point.
5 Actually Decca used three colours: red, green and purple.
í Õ ! ¤. û
The GPS satellite 12.2
287
t0 t 0 + 2 min
s( t 0 )
s( t 0 + 2 min)
^ F IGURE 12.3. The NNSS Transit system.
in a curve. The orbits of the Transit satellites are polar, meaning that
these curves are running in the general west-east direction. A fix could be
obtained already from a single overpass, and using multiple, both north-
and southgoing, overpasses in a least-squares adjustment6 improved the pienimmän
precision of the determination of both latitude and longitude. neliösumman
tasoitus
6
^ 12.2 The GPS satellite
Characteristic of the GPS, being originally a military positioning system,
is that the satellites are active and the users passive. So, the positioning
instruments used by the users, the GPS receivers, are quiet, whereas the
satellites contain radio transmitters. A GPS satellite is in a way a flying
Decca base station.
A GPS satellite is big, a contraption appearing as in figure 12.4a. It

contains, among others, the following components:
◦ A precise atomic clock, either a caesium, rubidium, or hydrogen-
maser clock. This clock synchronises all signals (carrier waves and
modulations) which the satellite transmits. Because the clocks have
6 At the time of deployment of the NNSS, the computing capacity required was still hard
to come by, especially in a form factor suitable for submarine use. A special-purpose
computer was designed for this, weighing a quarter of a ton, but small enough — and
with rounded corners! — to go through the hatch of a nuclear submarine, Wikipedia,
The AN/UYK-1 (TRW-130)!
í Õ ! ¤. û
(a) (b)
GPS (Block II) GLONASS
(c) (d)
Galileo BeiDou
^ F IGURE 12.4. Positioning satellites.
a limited lifespan, each satellite includes spare clocks.

◦ Radio transmitters. The antennas of the satellite point all the time
to the Earth. Transmission power is significant only within a direc-
tional cone that contains the Earth as a whole; the total power is
about 50 W. Two carrier-wave frequencies are used, 1575.42 MHz
(L 1 ) and 1227.60 MHz (L 2 ), which enables the elimination of the
effect of the ionosphere. The various codes that are used in posi-
tioning, as well as a code containing orbit and other information for
the users, are modulated onto the carrier waves.
í Õ ! ¤. û
The GPS segments 12.3
289
◦ Communication channels. The satellite receives data from the
GPS control centre. Besides control commands, this data contains
information on the orbits, clock corrections, “health”, etc., of all
GPS satellites. The information is stored in the satellite’s memory
and transmitted forward to users as a message modulated onto the
carrier wave.
◦ Solar panels produce the power needed by the equipment. The
amount of power has grown from 400 W (Block I) to almost three
kilowatts (Block IIF). At times, the satellite moves through the
Earth’s shadow; for this, there are batteries.
◦ Small rocket engines or thrusters for controlling attitude and orbit,
as well as a stock of propellant (monopropellant hydrazine). Be- ajoaine
cause of orbit perturbations, orbit corrections are needed at regular ratahäiriö
intervals.
◦ The satellites are stabilised on three axes: the antennas are point-
ing to the Earth, the solar panels to the Sun. For stabilisation,
reaction wheels (“flywheels”) are used.
As, during the lifetime of the GPS, the field of electronics has seen huge
developments, there are several satellite generations: “Block I”, “Block
II/IIA”, “Block IIR”, “Block IIR-M” and “Block IIF”, see Misra and Enge
(2010). The first Block-IIIA satellite was launched in 2018.
Satellites in operation today are all Block II or higher. The masses of
the satellites are 845 kg (Block I), 1660 kg (Block II) and 2269 kg (Block
III). The design lifetime of the satellites — limited by the propellant
stock for orbit maintenance, the diminishing power of the solar cells and
batteries, and the development of defects in clocks and electronics in the
Earth’s outer radiation belt — is 4.5 (Block I), 7.5 (Block II) or today
even 12–15 years. The satellites have regularly exceeded their design
lifetimes.
^ 12.3 The GPS segments

The GPS consists of three segments, figure 12.5:
Space segment The satellites themselves. avaruuslohko
Control segment Command centre, telemetry and control, tracking valvontalohko

stations, orbit determination, time synchronisation.
The main command centre, or Master Control Station, of the GPS
is located in Colorado Springs. There are some two dozen orbit
í Õ ! ¤. û
Control segment:
◦ Command centre
◦ Time synchronisation
◦ Orbit tracking
Space segment:
◦ Frequencies L 1 , L 2
◦ Precise time
◦ Orbit predictions
User segment:
◦ Reception of
satellite signal
◦ On land, on sea,
and in the air
^ F IGURE 12.5. The three segments of the GPS.
tracking and control stations around the globe. Like the whole
GPS, the control segment also resides under the US Department of
Defense, more precisely the US Air Force.
Every control and tracking station is equipped appropriately with,
among other things, a GPS receiver using a precise caesium clock.
Once every 24 hours, new orbital data — “broadcast ephemeris”
— and correction information for the satellite’s atomic clock is
uploaded to the satellites. The satellites include these orbit and
clock data into the radio signal they transmit, to be used by all
users.
käyttäjälohko User segment All users, on land, at sea and in the air — and more and
more also in space, in low Earth orbits — with their receivers.
í Õ ! ¤. û
The GPS constellation 12.4
291
^ 12.4 The GPS constellation
The design constellation of the GPS consists of 24 satellites and three
“active spares”, which can be moved into place immediately when an
active satellite breaks down. The satellites are in six different orbital
planes, with four satellites in each plane. In reality there are over 30
satellites operating today.
The height of the orbits from the Earth’s surface is 20 200 km. The or-
bital period around the Earth is 11h 58m , so, in the time it takes the Earth
to rotate once around her axis, 23h 56m , the satellites are “seen” again
in the same spots in the sky as the previous day. The GPS measurement
geometry repeats every day four minutes earlier, because the length of the
day in the mean solar time used by our clocks is four minutes longer than
the rotation period of the Earth.
The tilt of the orbital plane, or inclination, is7 i = 55◦ . Because of this, 7
the geometry of the GPS constellation is not very strong at high northern
latitudes, from where the satellites are seen mostly in the southern sky.
The system has currently such complete coverage that at least four
satellites are “visible” (meaning that they are above an elevation angle of korkeuskulma
10◦ ) anywhere on Earth at any time. The number of visible satellites is
almost always, and usually substantially, larger than this.
^ 12.5 Codes in the GPS signal

Two pseudo-random codes are modulated onto the carrier waves trans-
mitted by the GPS satellites on two different frequencies: the C/A code and
the P code. In addition, there is still the navigation message — containing
“broadcast ephemeris” and almanac data — which is also modulated onto
the carrier waves.8 See table 12.1. 8
The modulation technique used is phase modulation:9 the phase of the vaihe-
modulaatio
7 The
9
Block I satellites had a different inclination, i = 63◦ . None of those satellites are
working today.
8 In connection with GPS modernisation, frequency L 5 , 1176.45 MHz, has been added to
the signal. It is meant to be used in Safety of Life (SoL) critical operations, like aviation
and rescue services. In addition, new civilian and military codes have also been added
to the L 1 and L 2 frequencies.
9 Other modulation types in existence are amplitude modulation — where the strength
of the carrier, the amplitude, is varied in the rhythm of the signal — and frequency
modulation — where the frequency of the carrier is made to vary. Common radio
í Õ ! ¤. û
F IGURE 12.6. The original GPS constellation. The orbits and locations of satel-
^ lites are realistic with respect to the Earth.
carrier is turned “upside down”, meaning it is shifted by an angle of 180

degrees, or π, when the code switches between zero (0) and one (1).
Because of the modulation, the signal transmitted by the GPS satellites
^ T ABLE 12.1. Codes included in the GPS signal.
Modulation Carrier
Name Explanation frequency Repeat period wave
Coarse / Acquisition, 1.023 Mb 1 ms L1
/︁
C/A s
code Civilian Access
P code Precise / Protected 10.23 Mb 1 week L1, L2
/︁
s
P(Y) Combination of P code Same Same

and secret W code
- Navigation message 50 bits 12.5 minutes L1, L2
/︁
s
stations use amplitude modulation, and FM stations frequency modulation.
í Õ ! ¤. û
Codes in the GPS signal 12.5
293
Principle: Carrier wave. . .
. . . modulated
Phase inversion Modulation

(180◦ phase shift) = code
+1 = 0 +1 = 0
−1 = 1
^ F IGURE 12.7. The principle of phase modulation.
is pretty broadband. The effective bandwidth is several times the bit

frequency of the P code, that is several tens of MHz.
When already the transmission bandwidth requirement for a single
satellite is this large, one might think that the bandwidth demand of the
whole constellation would be huge. This is however not so: all satellites
use the same carrier-wave frequencies L 1 and L 2 . The receiver is able to kantoaalto-
separate the signals of different satellites from each other with the aid of taajuus
their different pseudo-random codes (C/A code and P code). Every satellite
has its own code, or “fingerprint”, in the same way that in navigation
at sea, every lighthouse has its own flashing sequence. The technical
solution goes by the name CDMA, or “code division multiple access”. koodijako-
kanavointi
The navigation message is a bit stream containing, besides precise or-
bital information on the broadcasting satellite, crude orbital information
— an almanac — on all satellites. Locking on to one satellite signal is
sufficient to receive the approximate orbital information for all satellites.
The pseudo-random codes are generated according to a documented
mathematical recipe (Wikipedia, Gold code). The codes thus are not
genuinely random: they can be exactly reconstructed using the same
recipe. They however behave statistically like genuinely random bit
sequences.
^ 12.5.1 The “tree rings” of GPS

The GPS receiver’s antenna on the Earth’s surface receives signals from
all the satellites that are in the sky at the moment of observation. This
whole “soup” travels along the cable to the receiver’s electronics.10 Here, 10
the first two tasks are performed:
10 In the pre-amplifier of the antenna, the analogue signal is processed to bring the
carrier-wave frequency down to a much lower value, without affecting the modulation
(“downbanding” or “heterodyning”). This prevents crosstalk of the amplified signal back
í Õ ! ¤. û
Code on received satellite signal
1 0 1 0 0 1 0 1 1 0 1
Same code (“replica”) generated by receiver
1 0 1 0 0 1 0 1 1 0 1
−→ t ∆t Time delay from correlation
F IGURE 12.8. The correlation method for determining the travel time ∆ t of the
^ GPS signal.
1. Separate the signals of different satellites from each other using

11 their individual pseudo-random code or “fingerprint”.11
2. Determine the travel time of the signal of each satellite from satel-
^ T ABLEAU 12.2. How does dendrochronology work?
The method works as follows: the laboratory holds a reference sequence of tree
rings, which has been built from partially overlapping tree-ring sequences. Rainy
years show as thick, dry years as narrow rings.a In the reference sequence, the
true, absolute year number of every ring is known.
Building a reference sequence is challenging. After doing so, however, the age
of any wooden object can be determined by comparing its tree rings with the
reference sequence, until the place is found where they match (correlate). The
method works because the succession of wet and dry years is largely random.
Similar methods are in use in many fields of science: dating and correlation of
ice drilling cores or geological deposits, correlation of magnetisation stripes on
the sea floor, and others.
a However, at the tree line in the mountains or the Arctic, tree-ring thickness is con-
trolled mostly by temperature. This circumstance has been used for reconstructing
paleotemperature time series. It can be said that the tree-ring width is a proxy for
temperature.
into the antenna, and makes further processing easier, like the digitisation of the signal
in an A-D (analogue-to-digital) converter.
11 The pseudo-random codes of the different satellites have been carefully designed to be
orthogonal to each other, meaning the true signal of one satellite correlates as weakly
as possible with the replica signal of another satellite generated in the receiver, even
with the correct time shift.
í Õ ! ¤. û
Codes in the GPS signal 12.5
295
lite to receiver.
Both tasks are carried out by a correlation method which seeks the time
shift ∆ t between the signal received from the satellite and a signal of the
same form, a “replica”, generated by the receiver, which would make them
identical. The replica sequence is shifted in time alongside the received
one until a strong correlation is found, a similarity or correspondence of
the pattern. The time difference obtained is in essence the GPS observable.
A suitable metaphor for the correlation method employed is the use
of year rings to date wooden objects, or dendrochronology, which was vuosilusto
already briefly explained in figure 7.7 and on which more in the text
tableau 12.2.
In a way, the comparison of the received GPS code with the replica
code generated by the receiver is also “dating”: the “age” of the signal
travelling from the satellite to Earth is determined. . .
When the correlation processing yields ∆ t = t rcv − t xmit , multiplying it
by the propagation speed c of the signal will give the pseudo-range to pseudoetäisyys
the satellite, the basic observable of GPS measurement. It is called a
pseudo-range, because it contains more than just the geometric distance,
among other things the clock offsets. We will return to this presently. kellopoikkeama
^ 12.5.2 C/A code and P code

Because the one-millisecond C/A code consists only of 1023 bits, one must
look at only 1023 alternative time shifts ∆ t. This goes very quickly. Each
satellite has its own “personal” C/A code, therefore, initially N × 1023
comparisons have to be made, where N is the number of satellites.
The C/A code can be used to determine the pseudo-range only modulo
300 kilometres, because the code repeats every millisecond, the time in
which the signal travels 300 kilometres. See figure 12.9. This is good
enough if the approximate location of the receiver is already known with
this accuracy. The state of modulation of the C/A code switches, if it
switches, at intervals of one microsecond (“chip rate”), the time in which
the radio signal travels 293 metres. The accuracy of measurement when
using the C/A code is better than this: if the receiver electronics can
measure the phase of the modulation with an accuracy of 1 % of a full modulaation
cycle, then the measurement accuracy is ±3 m. vaihe
Better accuracy is offered by the P code. It, too, is a pseudo-random

code, but its length is no less than 267 days. Every satellite uses its own,
í Õ ! ¤. û
296
12
Navigation code 50 Hz
20 460×
í
C/A code 1.023 MHz = 293 m Repeat 1 msec = 1023 “chips” = 300 km
T HE G LOBAL P OSITIONING S YSTEM (GPS)

Õ ! ¤. û
10×
P code 10.23 MHz = 29 m Correlation accuracy 1 % = ±0.3 m
154 × /120×
Carrier wave 19 / 24 cm = 1.6 / 1.2 GHz Phase measurement accuracy 1 % = ±2 mm
^ F IGURE 12.9. The various frequencies and effective wavelengths of the GPS signal.
GPS receivers 12.6
297
satellite-specific, one-week long subinterval of this period. The receiver
must also in this case be able to generate a “replica” of the code. Because,
however, the quantity ∆ t has already been obtained, using the C/A code,
to microsecond accuracy, only decimals more precise than this need to be
looked at.
The P code “chip rate” or bit frequency is ten times faster than that
of the C/A code, 10.23 Mb s , which corresponds to a travelled distance of
/︁
29 m.12 Again assuming a modulation phase measurement accuracy of 12

1 %, this corresponds to an accuracy of the pseudo-range observation of
±30 cm.
The P code is kept from civilian users by encryption. The encryption
is done by modulating, on top of the P code, a W code, the generating
algorithm for which has not been published.
Both the P code and the C/A code are thus modulated on the L 1 carrier
frequency. Distinguishing between them has been made possible by using
phase quadrature: while the P code is modulated with phase shifts 0 (bit
value 0) and π (bit value 1), the C/A code is similarly modulated with
phase shifts + 12 π and − 12 π, or +90◦ and −90◦ . One speaks of “in-phase” suora vaihe
and “quadrature” modulations.
The phase angles of the navigation message are the same as those of
the C/A code. This is not a problem as their frequencies are so different:
twenty repeats of the whole C/A code fit within one bit of the navigation
message, duration 20 ms! See figure 12.9.
^ 12.6 GPS receivers

The receivers intended for precise geodetic work are always dual- kaksitaajuus-
frequency instruments that can measure the carrier phase of the GPS koje
signal. Unlike with inexpensive hand-held devices, the antenna is kantoaallon
vaihe
usually separate and connected to the receiver by a coax cable. The
weak satellite signal is already amplified inside the antenna by a
pre-amplifier.13 esivahvistin
13
12 This is the “effective wavelength” of the P code. It is calculated as follows:

c
λeff = ,
f
in which f is the “chip rate”, a frequency-like quantity, unit s−1 or Hz, and c is the speed
of light. So if c = 300 000 000 m and f = 10 230 000 s−1 , it follows that λeff = 29.3 m.
/︁
s
í Õ ! ¤. û
WGS84 0255 GMT:21:14:27

LAT 61:22.3152N PDOP: 001
LON 25:48.4748E HDOP: 00
ALT 29.00 M VDOP: 01
◦
COG 230.40 Tr TDOP: 01
c SOG 0.30 Km/h SVS : 08 e
FOM 1 AGE : 000
Press ^v to toggle page 2
[-] NO WEST SOUTH < MENU > NORTH EAST YES [+]
0 1 2 3 4 5 6 7 8 9
^ F IGURE 12.10. Control panel of the Ashtech Z-12.
pakkokeskistys- The antenna may be mounted, using a standard forced-centring device,

laite onto a geodetic tripod. In real-time mapping surveys however, a measur-
tosiaikainen
ing pole or staff is used, with the antenna screwed onto the top. Attached
to the pole is the GPS receiver (if separate) with peripherals. The antenna
5
has a standard 8 inch hole at the bottom with a similar standard screw
thread (11 threads per inch) as in most geodetic instruments.
With every receiver type comes its own antenna type; the antennas
and receivers of different manufacturers are not generally electrically
intercompatible. However, the choke-ring antenna model (figure 12.11) is
available for precise geodetic work from many different manufacturers.
Choke rings lessen the problem of reflection of radio waves from the
monitie Earth’s surface and other surfaces, or multipath.
The electric centre of the antenna, the point where, in the geometric
interpretation, the radio waves are apparently received, is not the same as
the antenna’s official reference point (ARP). It is not even unambiguously
raja- defined, but depends somewhat on the cut-off elevation angle used for
korkeuskulma the observations, see figure 12.12. As a metaphor, one may think of the
apparent place of a fish under water, which also depends on the angle
vaihekeskipiste of view. One speaks of the variation of the antenna’s phase centre, see
Poutanen (2017) section 7.4 or Hofmann-Wellenhof et al. (2001) section
13 The DC voltage feed required by the pre-amplifier comes from the receiver, also
through the coax cable. This may complicate or prevent the mixing of receivers and
antennas from different brands.
í Õ ! ¤. û
GPS receivers 12.6
299
Hood (plastic) Crossed dipoles
“Choke rings”
Preamplifier
Coax antenna cable

Tripod attachment (screw thread)
F IGURE 12.11. A choke-ring GPS antenna for precise geodetic work. The an-
tennas provided by manufacturers are simpler and especially
^ smaller and lighter.
6.5.
When measurements are carried out over a relatively small area, using
only one type of antenna, this variation of the phase centre vanishes
from the end result, and from the position difference vectors computed
between different points in the network. If, however, antenna types are
mixed, or extensive networks measured — hundreds or thousands of
kilometres across — one ought to calibrate the phase delay patterns of vaiheen
the antennas, which are fairly complicated functions of both the elevation kulkuviive
angle η and the azimuth direction α. The calibration in which this phase
delay pattern ∆φ(η, α) is determined can be carried out in the laboratory
using an artificial GPS signal source, or as a field calibration in which two
antenna types are always compared to each other. Field calibration is
thus always relative, referring to some agreed reference type of antenna.
In highly precise geodynamic deformation measurements it is nowa-
days the practice to calibrate, not just antenna types, but individual
antennas.
The radio waves transmitted by GPS satellites are clockwise circularly
polarised. Upon reflection, the polarisation direction reverses to anti-
clockwise. The antenna — in the example in figure 12.11, a cross dipole
— is built so that it transfers only the clockwise polarised signal on to the
receiver. In this way, the harm caused by reflections is minimised.
Technological development goes into the direction of greater integration.
í Õ ! ¤. û
Cut-off angle B B
A
Cut-off angle A
“Electric
X centre”
Apparent A
Apparent B
Crossed dipoles True
centre Antenna place
reference point
^ F IGURE 12.12. The electric centre of an antenna is not a self-evident thing!
Today’s GPS positioning instruments including those for geodetic use

are so small that they are integrated with the antenna. Because the
instruments are very autonomous, they no longer even have a proper
display screen.
ohjelmisto- Another development is the onwards march of software-defined re-
pohjainen ceivers: today’s personal computers are beginning to be powerful enough
vastaanotin
to do the digital processing work that in today’s receivers is done on the
hardware level. Then one only needs, in addition to a general-purpose PC,
a “dumb” analogue radio device with antenna (Lázaro, 2012).
^ 12.7 Observables of GPS

A geodetic GPS receiver stores the observations it makes into its memory
as a long table containing many numbers. The numbers in the table
represent distances between the receiver and the different satellites
observed. It is easy to understand why the number count becomes so
large: if, for example, the time between measurements is 30 seconds, and
five satellites are visible in the local sky, and with a single-frequency
receiver we observe both the C/A code and the P code, then the number of
^ T ABLEAU 12.3. Start of a RINEX file. The device collects five observation types:
carrier-phase observations and P code observations on both frequencies L 1 and
L 2 , as well as C/A code observations on frequency L 1 . The observations are
stored at intervals of 30 seconds. There are eleven satellites on the first epoch, 1.
January 2000 0 : 00 : 00, and also on the second epoch, 0 : 00 : 30. They are all
GPS satellites (G). The observation station is DGAR, Diego Garcia in the Indian
Ocean (Vine, 2011). →
í Õ ! ¤. û
Observables of GPS 12.7
301
2.00 OBSERVATION DATA G (GPS) RINEX VERSION / TYPE
teqc 1999Oct8 gpsops 20000103 21:13:37UTCPGM / RUN BY / DATE
OSF1 V4.0 564|Alpha|cc 4.4.18.4|=+|=| COMMENT
DGAR MARKER NAME
30802M001 MARKER NUMBER
GNOG JPL OBSERVER / AGENCY
T341U AOA SNR-8000 ACT 3.3.32.3 REC # / TYPE / VERS
250 AOAD/M_T ANT # / TYPE
1916269.8405 6029977.3167 -801720.2273 APPROX POSITION XYZ
0.0814 0.0000 0.0000 ANTENNA: DELTA H/E/N
1 1 WAVELENGTH FACT L1/2
5 L1 L2 P1 P2 C1 # / TYPES OF OBSERV
30.0000 INTERVAL
COMMENT
This data is provided as a public service by NASA/JPL. COMMENT
No warranty is expressed or implied regarding suitability COMMENT
for use. For further information, contact: COMMENT
Dave Stowers, NASA/JPL m/s 238-600 COMMENT
4800 Oak Grove Drive, Pasadena CA 91109 USA COMMENT
COMMENT
2000 1 1 0 0 0.0000000 GPS TIME OF FIRST OBS
END OF HEADER
00 1 1 0 0 0.0000000 0 11G21G23G17G30G 1G31G29G22G15G25G 3
-6078127.503 4 -4736200.435 4 23397694.178 23397698.378 23397695.030
-2556364.753 4 -1991958.142 4 24025055.814 24025059.840 24025056.373
-8073501.747 5 -6291024.522 5 22565280.025 22565283.337 22565280.587
18247234.140 4 14218628.480 4 24610505.696 24610508.842 24610505.143
-1299479.831 4 -1012581.476 4 24824108.761 24824113.289 24824108.748
-5233446.124 4 -4077998.775 4 24175634.461 24175638.438 24175635.537
16878293.604 4 13151917.927 4 24427189.279 24427193.024 24427188.034
-13489828.171 5 -10511530.918 5 22792735.726 22792739.295 22792736.451
-4494062.929 4 -3501865.147 4 23961699.555 23961704.148 23961699.834
-21099958.763 9 -16441519.960 9 20331187.861 20331190.408 20331187.808
-15215098.740 5 -11855903.290 5 22202394.742 22202398.143 22202394.948
00 1 1 0 0 30.0000000 0 11G21G23G17G30G 1G31G29G22G15G25G 3
-6132427.986 4 -4778512.477 4 23387361.455 23387365.218 23387361.703
-2586441.342 4 -2015394.448 4 24019332.425 24019336.115 24019332.872
-7990741.587 5 -6226536.097 5 22581028.707 22581032.399 22581029.091
18274808.415 4 14240114.880 4 24615752.673 24615756.082 24615752.310
-1317133.094 4 -1026337.267 4 24820749.907 24820754.243 24820749.907
-5259685.471 4 -4098444.960 4 24170641.011 24170645.465 24170641.809
16938081.982 4 13198506.304 4 24438566.644 24438570.267 24438567.293
-13548930.874 5 -10557584.968 5 22781488.710 22781492.153 22781489.545
-4600217.585 4 -3584583.008 4 23941499.061 23941503.148 23941500.039
-21083529.873 9 -16428718.216 9 20334314.137 20334316.731 20334314.115
-15210926.086 5 -11852651.873 5 22203188.835 22203192.175 22203189.231
í Õ ! ¤. û
observations collected within one minute already amounts to 60 30 · 5 ·

(︁ /︁ )︁
2 = 20. In an hour, this means 1200 observations; if these are stored into
14 the memory as ordinary double precision real numbers14 (eight bytes per
number), we need 9.4 kilobytes of storage space. In 24 hours, the storage

requirement would then be 225 kB.
An often-used format in international GNSS data exchange is RINEX,
Receiver-Independent EXchange format, Gurtner and Estey (2007). This
is a text format, human-readable, in which observational data from most
geodetic receiver types can be transferred, read, and processed, in a
way that is independent from the manufacturer of the receiver. See the
example in tableau 12.3 from the Diego Garcia station (Vine, 2011) in the
Indian Ocean.
^ 12.7.1 Pseudo-ranges as observables

etuliite Why does one speak of pseudo-ranges? The prefix “pseudo” is because
the observable’s value is affected by, in addition to the geometric distance
between satellite and receiver, the clock errors or offsets ∆ t and ∆T , as
kulkuviive well as the propagation delays caused by the ionospheric and tropospheric
15 media.15 Thus the observation equation for pseudo-range is obtained:
p = ρ + c (∆T − ∆ t) + d ion + d trop , (12.1)
in which
p pseudo-range
ρ natural (geometric) distance. According to Pythagoras
√︂
ρ= ( x − X )2 + ( y − Y )2 + ( z − Z )2 ,
in which
[︂ ]︂T
x y z location of the satellite in space
[︂ ]︂T
X Y Z location of the receiver in space
c speed of light in a vacuum
∆T offset of receiver clock from GPS time (clock correction = −∆T )
14 The “real” numbers of computers are in fact rational numbers. . .

15 More precisely: the effects of free electrons and neutral molecules, respectively, on
the propagation of radio waves. Therefore, the stratosphere also causes “tropospheric
delay”.
í Õ ! ¤. û
303
x, y, z, ∆ t
Broadcast
ephemeris
d ion
d trop
X , Y , Z, ∆T
Unknowns
X
^ F IGURE 12.13. Pseudo-range observation.
∆t offset of satellite clock from GPS time

d ion “ionospheric” — meaning caused by the ionised atmosphere —
propagation delay
d trop “tropospheric” — meaning caused by the neutral atmosphere —
delay.
The clock offset ∆ t of the satellite is included in the broadcast ephemeris

transmitted by the satellites. The receiver clock offset ∆T on the other
hand remains unknown, and it will have to be estimated as one unknown
together with the co-ordinates. Thus there are four unknowns for every
GPS receiver: three co-ordinates X , Y , and Z , and the clock offset ∆T —
if we forget for a moment about the atmospheric unknowns d ion and d trop .
Four pseudo-range observations, to four different satellites, are enough to
determine four unknowns. See figure 12.13.
^ 12.7.2 The carrier wave’s phase angle as observable

The “chip rate” of the C/A code, the number of bits transmitted per sec-
ond, is 1.023 MHz, corresponding to a “wavelength” of 293 m, when the
corresponding number for the P code, 10.23 MHz, means a wavelength
of 29.3 m. If, instead of a modulation, we use the carrier wave itself, the
relevant wavelength becomes 19 cm (L 1 ) or 24.4 cm (L 2 ), a distance that
is two orders of magnitude shorter. Geodetic GPS positioning, where dual-
frequency receivers observe the phase of the carrier waves transmitted vaihe
by the GPS satellites, is based on this. Electronic phase measurement is
relatively easy and precise, but, as the market is small and specialised,
í Õ ! ¤. û
^ T ABLE 12.4. Properties of GPS carrier waves. “Factor” denotes the multiple of
the base frequency (10.23 MHz).
Carrier wave Frequency (MHz) Wavelength (cm) Factor
L1 1575.42 19.0 154×

L2 1227.60 24.4 120×
the prices of the devices are nevertheless as high as for other geodetic
instruments, upwards of thousands of euros.
As is also the case with electronic range-finders (distance measurement
devices), phase measurement is always afflicted by an ambiguity prob-
lem. Measured phase values φ may be expressed as values within the
[︁ )︁
single-cycle interval 0, 2π , and the corresponding pseudo-range between
satellite and receiver can be determined only “modulo an integer number
of wavelengths”. If a certain pseudo-range value P is compatible with a
measurement done on wavelength λ, then the pseudo-range values P + λ,
P − λ, P + 2λ, P − 2λ, . . . are also compatible.
The observation equation for the carrier phase is, as a phase difference
16 angle,16 in radian units
def ρ + c (∆T − ∆ t) + D ion + D trop

φ = φ + 2π N = 2π
λ
or as a distance, unit metres
φ φ
(︂ )︂
def
P =λ =λ + N = ρ + c (∆T − ∆ t) + D ion + D trop . (12.2)
2π 2π
The left-hand side of the observation equation is either the phase angle
def
φ = φ + 2π N or the equivalent pseudo-range P = P + λ N , see figure 12.14.
Both contain the correct integer number N of wavelengths making up
the range between satellite and receiver. The symbol φ without overbar,
on the other hand, designates the “raw” phase difference measurement.
[︁ )︁
It is often assumed that φ ∈ 0, 2π at least at the moment of signal
acquisition.
Other symbols used in the equation:
D ion the propagation delay of the carrier wave caused by the ionosphere
(which is actually negative, D ion = − d ion )
16 The measured quantity is actually the phase of the received radio wave subtracted
from that of the receiver’s reference oscillator, representing the delay in transit from
satellite to receiver. In the difference, the phase angle of the received wave enters with
a negative algebraic sign.
í Õ ! ¤. û
305
φ
φ (︁ φ )︁
P = λ 2π = λ 2π + N
F IGURE 12.14. Measurement of the phase of the GPS signal’s carrier wave. The
[︁ )︁
measured phase difference angle is φ, initially φ ∈ 0, 2π , the
metric pseudo-range, including whole wavelengths, is P, and
^ the ambiguity is N, in this example N = 2.
D trop the delay caused by the troposphere, D trop = d trop

λ the wavelength of the received carrier wave according to table 12.4,
including Doppler shift
N the integer unknown or ambiguity. kokonaisluku-
(︃ √︂ )︃ tuntematon
2 2 2
The variables ρ = ( x − X ) + ( y − Y ) + ( z − Z ) , ∆ t and ∆T are the
same as those for code pseudo-range, see equation 12.1. There are again
four geodetic, real-valued unknowns: X , Y , Z , and ∆T . But now the
integer values N , one for each observation, also need to be determined in
addition to the geodetic unknowns. This may seem impossible, but there
are ways of doing it; see section 13.3.
In equations 12.1 and 12.2 we have used different symbols for the
ionospheric and tropospheric delays of the carrier wave, than for the
corresponding delays of code measurement, because they are different.
One says that the ionosphere is dispersive for radio waves: the speed of
propagation depends on the frequency, or equivalently, on the wavelength.
The travel speed of the code modulations is the group velocity, which
is always less than the speed of light in a vacuum. The travel speed of
the carrier phase is the phase velocity. In a dispersive medium these two
speeds differ from each other.
^ 12.7.3 Effects of the ionosphere and troposphere

The ionosphere is a dispersive medium to radio waves: different frequen-
cies propagate at different velocities. As a consequence of dispersion,
í Õ ! ¤. û
(Propagation speed in vacuum)
Group Phase
Time
F IGURE 12.15. The propagation of a wave packet in a dispersive medium, phase

and group velocity. The carrier wave travels with phase velocity,
the modulations — also the pseudo-random codes of the GPS
^ signal — travel at group velocity.
the phase and group velocities of propagation are different. For phase
vaihetaite- propagation, the index of refraction is
kerroin √︄
f p2 c2 c4 c6
np = 1− ≈ 1+ + + +··· (12.3)
f2 f2 f4 f6
In this equation the constants c i and the plasma frequency f p depend on

the electron density, n e , of the ionosphere. A good approximate formula,
which explains 99.9 % of the whole ionospheric propagation effect, is
(Seeber, 1989):
C
np = 1 − , C = 40.3 n e m3 s2 ,
/︁
f2
in which the electron density n e is expressed in electrons per m3 . Typical

numbers are 108 –1012 m−3 . Electron density varies between day and
night — greater in the daytime — with the season — greater in summer
— with solar activity, and of course with the latitude and height of the
location.
ryhmä- The group propagation velocity, more precisely the group index of
taitekerroin refraction, is obtained as the derivative with respect to the frequency17 of
17
17 When the propagation velocity is c′ = c n , in which c is the speed of light in a vacuum,
/︁
í Õ ! ¤. û
307
np f ,
(︁ )︁
d np f C
ng = = 1+ 2, C like above.
df f
Because the index of refraction of the ionosphere is dependent upon the

frequency f , it is possible to eliminate the effect of the ionosphere by
combining measurements made on two different frequencies. This is the
fundamental reason why the GPS uses two different frequencies L 1 and
L2.
If we form the linear combination of code observations
def f 12 p 1 − f 22 p 2
p3 =
f 12 − f 22
and correspondingly
def f 12 n g,1 − f 22 n g,2
n g,3 = ,
f 12 − f 22
we obtain
f 12 1 + C f 12 − f 22 1 + C f 22
(︁ /︁ )︁ (︁ /︁ )︁
f 12 − f 22
n g,3 = = = 1,
f 12 − f 22 f 12 − f 22
from which it is seen that the effect of the ionosphere has vanished.18 18
The troposphere — more precisely, the neutral atmosphere, which also

includes the stratosphere, and neutral fractions of higher layers — on the
other hand is not dispersive to radio waves. Its speciality is, however, a
strong dependence on water vapour content of the index of refraction. The
equation, the same equation 7.6 that also applies for electronic distance
measurement, is (Rüeger, 1996, 2002):
77.624 K 64.70 K
/︁ /︁
5748 K
(︂ )︂
6 hPa hPa
NM = 10 ·( n M − 1) = ( p − e )+ 1+ e,
T T T
in which n M is the index of refraction for microwaves. Here, the unit of p taitekerroin
and e is hPa (hectopascal), or millibar. T is the absolute temperature in
kelvin.
it follows that the phase propagation velocity is greater than the speed of light. The
carrier phase cannot however carry information, so the directionality of time according
to thermodynamics is preserved. . . if it were possible to move information faster than
light, it would, according to special relativity, also be possible to carry information back
in time!
18 The result and its proof are the same if we take, instead of the group index of refraction,
the phase one, n p .
í Õ ! ¤. û
GPS receiver
Y
Origin, ρ
centre of mass
of the Earth r
X
GPS satellite
^ F IGURE 12.16. Geometry of GPS positioning.
The coefficient of the partial pressure of water vapour e in this equation

is at ambient temperature as much as 17 times the coefficient for dry-air
partial pressure p − e!
Because both the ionosphere and the troposphere affect the propagation
of GPS radio waves, GPS measurements can be used for both ionospheric
and tropospheric research. Meteorologists and climate researchers are
very interested in this; see section 18.7.
^ 12.8 GPS measurement geometry

See figure 12.16. Let the position vector of satellite S in a geocentric
frame be r = rS , the position vector of observation station A in the same
frame be R = R A , and let the distance between the two be ρ = ρ SA .
Then the following vector equation applies:
rS = R A + eSA ρ SA ,
in which eSA is the direction (unit vector) to satellite S seen from observa-
tion station A . The task of GPS positioning is to compute R A when ρ SA is
given to sufficiently many satellites S .
The Pythagoras theorem gives
⃦ √︂ 2 2 2
ρ SA
⃦ S
= ⃦r − R A ⃦ = ( xS − X A ) + ( yS − Y A ) + ( z S − Z A ) , (12.4)
í Õ ! ¤. û
Measurement geometry and sensitivity of observations 12.9
309
in which
r S = x S i + y S j + z S k, R A = X A i + YA j + Z A k
are the geocentric position vectors of satellite S and observation station

A . The unit vectors i, j, and k are aligned with the co-ordinate axes:
{︁ }︁
i, j, k forms an orthonormal basis in Euclidean space. ortonormaali
kanta
Often one writes the vectors as column vectors of their components, the
co-ordinates of location:
⎡ ⎤ ⎡ ⎤
xS XA
def def
rS = ⎣ yS ⎦ , R A = ⎣ YA ⎦ .
⎢ ⎥ ⎢ ⎥
zS ZA
From the observations, however, no true distances ρ are obtained, but

pseudo-ranges p, equation 12.1, or P , equation 12.2.
The atmospheric propagation delays d ion , d trop , D ion , and D trop must
also be taken into account in some way. The alternatives are
◦ Elimination from the observation equations — as described in sub-
section 12.7.3 for the ionospheric effects d ion and D ion .
◦ Computation using a good, externally provided atmospheric model.
◦ Modelling of the atmospheric effect using unknown parameters, to
be estimated from the same observation equations — as we will
describe for the tropospheric effects d trop and D trop in subsection
18.7.1.
^ 12.9 Measurement geometry and sensitivity of observations

Above it was shown that the observables of GPS are pseudo-ranges, the
observation equation of which looks like this:
p = ρ + c (∆T − ∆ t) + d ion + d trop .
Let us leave out the effect of the atmosphere, and assume also that the
satellite’s orbit — and thus, the momentaneous position vector of the
satellite in space as computed from the orbit and clock time — and the
satellite’s clock offset ∆ t are known, in other words already taken into
account:
p = ρ + c ∆T,
in which ρ is the geometric distance between satellite and receiver, and

∆T is the clock offset of the receiver.
í Õ ! ¤. û
We can write equation 12.4 more simply:

√︂
2 2 2
ρ SA = ρ = (x − X ) + ( y − Y ) + (z − Z) ,
[︂ ]︂T
in which x y z is the known position vector of the satellite in
space, computed from the orbit information, also called the ephemeris.
[︂ ]︂T
X Y Z is the receiver position vector. Now
√︂
2 2 2
p = ( x − X ) + ( y − Y ) + ( z − Z ) + c ∆T,
in which there are four unknowns, X , Y , Z , and ∆T .

In what follows, we shall use instead the expression c ∆T as the un-
known, as it has the same dimension, length, as the co-ordinate un-
knowns.
To solve four unknowns, it is sufficient to have observations to four
satellites. If the number of useable satellites is larger, we have redun-
tasoitustehtävä dancy and an adjustment problem.
häiriö Question In what way do small “perturbances” in the location of the

receiver impact a certain measurement p?
Answer Look at the place of the satellite in the sky. See figure 12.17.
Let the satellite’s direction vector as seen from the observation
site be e. This is a unit vector: its length is
√︂
∥e∥ = e21 + e22 + e23 = 1.
If the satellite’s place in the sky is azimuth α, elevation η, then

def
e = e1 · N + e2 · E + e3 · U
and the column vector of components is

⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤
e1 cos α sin ζ cos α cos η x− X
def ⎢ ⎥ 1⎢
e = ⎣ e 2 ⎦ = ⎣ sin α sin ζ ⎦ = ⎣ sin α cos η ⎦ = ρ ⎣ y − Y ⎦ .
⎥ ⎢ ⎥ ⎢ ⎥
e3 cos ζ sin η z−Z

{︁ }︁
Here, N, E, U is the orthonormal basis of unit vectors in the
local-horizon system (“north, east, up”). η = 90◦ − ζ is the satellite’s
elevation angle above the horizon.
We carry out a sensitivity analysis. In what way do small co-
ordinate shifts ∆ X , ∆Y , or ∆ Z in the location of the observation
site influence the observable p?
í Õ ! ¤. û
311
Zenith GPS satellite
“Celestial e3 · U
sphere” e
ζ N
W η
e1 · N
α
Observation site
e2 · E
E
F IGURE 12.17. The geometry between a GPS satellite and an observation site.
{︁ }︁
N, E, U is the local-horizon orthonormal basis (“north, east,
^ up”).
◦ If the co-ordinate shift influences highly, then the observa-

tion p will help in determining the unknown co-ordinate in
question.
◦ If the co-ordinate shift does not influence at all, then the
unknown in question cannot be determined using observation
p.
◦ The greater the sensitivity of the observations, the better the
precision of the solution of the unknowns.
Intuitively observation p is most affected, in a ratio of one-on-one, by
a shift in observation site location along the satellite’s direction
vector e. Shifts in observation site location perpendicular to this
direction vector have no effect at all.
Equation:
“Effect” = − ∆R · e = − (∆ X e 1 + ∆Y e 2 + ∆ Z e 3 ) .
⟨︁ ⟩︁
(12.5)
This intuitive result may also be derived more formally by lineari-

sation. See tableau 12.5 and section 14.6.
í Õ ! ¤. û
^ T ABLEAU 12.5. A more rigorous derivation of the influence formula 12.5 by

means of linearisation.
[︁ ]︁T
Choose for the observation station X Y Z an approximate location
]︁T
, and an approximate clock offset ∆T = 0. Then we can con-
[︁
X0 Y0 Z0
struct an approximate pseudo-range observation
√︂
p0 = ( x − X 0 )2 + ( y − Y0 )2 + ( z − Z0 )2 .
Carry out a Taylor series expansion in the neighbourhood of the approximate

location, around the value p 0 . The first, linear terms yield
∂p ∂p ∂p
p ≈ p0 + ( X − X 0) + (Y − Y0 ) + ( Z − Z0 ) + c∆T
∂X ∂Y ∂Z
∂p ∂p ∂p
=⇒ ∆ p = p − p 0 ≈ ∆X + ∆Y + ∆ Z + c∆T. (12.6)
∂X ∂Y ∂Z
Here, the coefficients, partial derivatives, are obtained as follows:
∂p x− X ∂p y−Y ∂p z−Z
=− , =− , =− . (12.7)
∂X ρ ∂Y ρ ∂Z ρ
The values for the coefficients are evaluated at the approximate location
[︁ ]︁T
X0 Y0 Z0 instead of at the true but unknown location. This suffices,
because in equation 12.6 the values (∆ p, ∆ X , ∆Y , ∆ Z ) are small differences
between the true ( p, X , Y, Z ) and approximate values ( p 0 , X 0 , Y0 , Z0 ).
The partial derivatives 12.7 are direction cosines, the direction, apart from the
algebraic sign, to the satellite as seen from the observation station, projected
onto the co-ordinate axes X , Y , and Z.
^ 12.9.1 DOP quantities and observation equations

The measure of the quality of the measurement geometry of the GPS
satellites in the local sky is DOP, dilution of precision. Using the above
geometric sensitivity analysis, we may calculate various variants of DOP.
A larger DOP number means a poorer measurement geometry!
DOP characterises the quality of the satellite geometry: how much worse
or better one may get the co-ordinates resolved from “standard quality”
measurements due to the poorer or better geometry of the satellites on
offer in the sky. This is valuable information in the planning phase of
measurement work.
One rule of thumb is that the measurement geometry is acceptable if
GDOP < 7 . . . 10, depending on the intended use.
í Õ ! ¤. û
313
If we now describe the “variation” of the co-ordinates of the observation
[︂ ]︂ T
site by a (small) difference quantity ∆R = ∆ X ∆Y ∆ Z , and the
“variation” of the receiver clock by an (also small) difference quantity
c ∆T , we may express the dependence of the observable p( i) on these
altogether four unknowns like this:
−∆ X
⎡ ⎤
[︂ ]︂ ⎢ −∆Y ⎥
( i) ( i) ( i) ( i)
∆p = e1 e2 e3 1 ⎢ ⎥=
⎢ ⎥
⎣ −∆ Z ⎦
−∆ X
⎡ ⎤
c ∆T
[︂ ]︂ ⎢ −∆Y ⎥
= cos α i cos η i sin α i cos η i sin η i 1 ⎥,
⎢ ⎥
⎣ −∆ Z
⎢
⎦
c ∆T
in which α i and η i are the azimuth and elevation of satellite i in the local
sky. This way of writing is called linearisation.
This equation may be understood as an observation equation. If the
equation is written symbolically, as is the practice in geodesy, in the
form19 19
ℓ + v = Aˆ︁
x, (12.8)
x, the estimators of the

then the elements of the vector of unknowns ˆ︁
unknowns, are −∆ X
ˆ︁ , −∆Yˆ︁ , −∆ Z
ˆ︁, and c ∆ˆ︂
T , the vector of observations ℓ
is made up of values ∆ p( i) , and the design matrix is rakennematriisi
⎡ ⎤
cos α1 cos η1 sin α1 cos η1 sin η1 1
⎢ cos α2 cos η2 sin α2 cos η2 sin η2 1 ⎥
⎢ ⎥
⎢ .. .. .. .. ⎥
⎢ . . . . ⎥
A=⎢ ⎥. (12.9)
⎢ ⎥
⎢ cos α cos η i
i
sin α i cos η i i
sin η 1 ⎥
.. .. .. .. ⎥
⎢ ⎥
. . . . ⎦
⎢
⎣
cos αn cos ηn sin αn cos ηn sin ηn 1
This is in fact the linearised version of the original observation equation

12.1.
This design matrix contains everything we know about the GPS mea-
surement geometry.20 From this, all DOP quantities may be calculated, 20
19 The residuals v are needed to reconcile the observations ℓ, which contain measurement
uncertainty, with each other when there are more observations than unknowns. See
section 14.4.
20 As does also a sky plot, for example Borre (2009).
í Õ ! ¤. û
^ T ABLE 12.6. Variants of the DOP quantity.
Acronym Name Quantity characterised
GDOP Geometric DOP Place and time

PDOP Position DOP Place
HDOP Horizontal DOP Horizontal location
VDOP Vertical DOP Height
TDOP Time DOP Time
without using a single real observation — it suffices that the places of the
satellites in the sky can be computed. The size of the matrix is n × 4: n
rows and four columns, where n is the number of satellites available for
use.
rekognosointi The situation is the same as in the case of reconnaissance of a terres-
trial geodetic network: the quality of the network can already be judged
based on point locations and planned measurement geometry, before
even a single measurement has been carried out. This is a great tool for
planning.
We will discuss more about observation equations and least-squares
adjustment in section 14.4. Here we do not even try to compute a least-
squares solution. We only look into the precision of the four unknowns
−∆ X , −∆Y , −∆ Z , and c ∆T to be computed!
We assume, for this computation, that all observations are equally pre-
cise — their precision may be assumed 1 — and that they are statistically
independent of each other. Then, the following simple calculation is valid.
It gives a picture of the role of the GPS measurement geometry in the final
precision of the measurement results. Other factors, like the technical
capability of the receiver and antenna used, duration of measurement
and the atmosphere, can be looked at separately.
^ 12.9.2 Error ellipsoids for presentation of measurement precision

From the design matrix A we may construct the normal matrix, or weight
matrix of the unknowns, as follows:
def
N = Pxx = A T A. (12.10)
Expressed in satellite positions in the sky, given as azimuths α i and

elevation angles η i , the result is according to equation 12.11. The summa-
∑︁
tion is understood to be over the azimuths and elevations all satellites,
α i , η i , i = 1, . . . , n.
í Õ ! ¤. û
Measurement geometry and sensitivity of observations
∑︁ 2
12.9
315
cos α cos2 η sin α cos α cos2 η cos α sin η cos η cos α cos η
⎡ ∑︁ ∑︁ ∑︁ ⎤
2
∑︁ 2 2
sin α cos α cos η sin α cos η sin α sin η cos η sin α cos η
⎢ ∑︁ ∑︁ ∑︁ ⎥
Pxx = ⎢ ⎥.
⎢ ⎥
∑︁ 2
cos α sin η cos η sin α sin η cos η sin η sin η
∑︁ ∑︁ ∑︁
⎣ ⎦
cos α cos η sin α cos η sin η
∑︁ ∑︁ ∑︁
n
(12.11)
−1 def
The inverse of this weight matrix Pxx , Q xx = Pxx , is the weight-
coefficient matrix:
⎡ ⎤
q xx qxy q xz q xt
⎢ q
⎢ yx q yy q yz q yt
⎥
Q xx = ⎢ ⎥.
⎥
⎣ q zx q z y q zz q zt ⎦
q tx qty q tz q tt
This matrix, like the weight matrix Pxx or the design matrix A , still
describes exclusively the geometry of the measurement site and satellites,
and nothing else.
Now, the variance matrix of the vector of unknowns, or solution,
[︂ ]︂T
x=
ˆ︁ −∆ X
ˆ︁ −∆Yˆ︁ −∆ Z
ˆ︁ c∆ˆ︂
T
is ⎡ ⎤
q xx qxy q xz q xt
⎢ ⎥
2
⎢ q yx q yy q yz q yt
2⎢
Σxx = σ Q xx = σ ⎢
⎥
⎥.
⎣ q zx q z y q zz q zt
⎥
⎦
q tx qty q tz q tt
The constant σ2 is called the (a priori) variance of unit weight. Its square
root, the mean error of unit weight σ, is the mean error, assumed constant, painoyksikön
of a single observable, meaning one pseudo-range. keskivirhe
The variance matrix of the co-ordinate solution is a 3 × 3 element sized

submatrix Σrr of the variance matrix of the unknowns:
⎡ ⎤
q xx qxy q xz
Σrr = σ2 ⎣ q yx q yy q yz ⎦ . (12.12)
⎢ ⎥
q zx q z y q zz
The DOP quantities are calculated directly from the weight-coefficient

matrix Q xx :
√︁ √︁ ⎷
PDOP = q xx + q yy + q zz , HDOP = q xx + q yy , VDOP = q zz .
í Õ ! ¤. û
The time t comes along in the quantities TDOP and GDOP. Remember,
however, that the unknown to be estimated is c ∆
ˆ︂T rather than ∆
ˆ︂T , and
q tt corresponds to this quantity:
⎷ √︁ √︁
TDOP = q tt , GDOP = q xx + q yy + q zz + q tt = PDOP2 + TDOP2 .
The mean errors of the co-ordinates are obtained as the square roots of
the diagonal elements of the variance matrix 12.12:
⎷ ⎷ ⎷
σ X = σ q xx , σY = σ q yy , σ Z = σ q zz .
Also
⎷
σ c∆T = σ q tt .
The familiar point mean error in the plane is now related directly to the
HDOP quantity:
precision instrumental technology, etc. geometry

⏞ ⏟⏟ ⏞ def √︂ ⏞ ⏞ ⏟⏟ ⏞
σP σ2X + σ2Y σ
⏞ ⏟⏟
= = · HDOP .
The co-ordinate variance matrix can be graphically represented by a

three-dimensional error ellipsoid. The error ellipsoid around a measure-
ment point visualises the uncertainty of the location based on the above
definitions. A similar DOP ellipsoid is obtained by leaving off the constant
σ2 : it has the same shape as the error ellipsoid, but it has no metric size,
as the elements of the matrix Q xx , like those of the design matrix A , are
dimensionless.
In the general case it is not so simple to calculate the parameters of the
ellipsoid from the matrix elements. Let us look at a simpler special case,
which nevertheless is practically relevant. If the measurement geometry
is symmetric, meaning that the satellites and their elevation angles are
evenly distributed by azimuth, around the horizon, it follows that the DOP
ellipsoid will be oriented along the co-ordinate axes: q x y = q x y = q yz = 0
(and q xx = q yy !), and the matrix is
⎡ ⎤
q xx
Σrr = σ2 ⎣ q yy ⎦.
⎢ ⎥
q zz
In this special case, the axes of the DOP ellipsoid are pointing along
the local co-ordinate axes, and the longest axis points in the vertical
direction. In a practical measurement situation, the variance matrix
í Õ ! ¤. û
317
σ σ
DOP ellipsoid
Z
X Error ellipsoid 1
Error ellipsoid 2
F IGURE 12.18. The connection between the DOP ellipsoid and error ellipsoid,
and the mean error of unit weight σ. The DOP ellipsoid only
characterises the effect of the geometry, whereas the error el-
lipsoid also depends on the precision of measurement, meaning
^ the device type.
of the co-ordinates is often close to this. The longest axis of the error
ellipsoid is almost always close to the vertical, which tells us that the
height is more weakly determined than the horizontal location.21 In this 21
⎷
q zz
⎷
q yy
Z
⎷
q xx
F IGURE 12.19. The DOP ellipsoid of GPS positioning, assuming its principal axes
^ are in the same directions as the co-ordinate axes.
21 Reasons for this are:
í Õ ! ¤. û
case the co-ordinate mean errors are

⎷
σ X = σY = σ · 21 2 · HDOP, σ Z = σ · VDOP,
akselin puolikas based on the above definitions. The lengths of the semi-axes of the DOP
⎷ ⎷ ⎷
ellipsoid are q xx , q yy and q zz . In the symmetric case, we thus have
q xx = q yy = 12 HDOP2
in the horizontal plane, and always
q zz = VDOP2
in the vertical direction.
^ 12.9.3 DOP and measurement planning

In general it can be said that the measurement geometry is better if HDOP
and VDOP, thus also PDOP, are smaller. This again requires that
◦ A sufficient number of satellites is above the horizon and observable
from the measurement site.
◦ They are positioned evenly around the sky.
This of course works only if one has a sufficiently free view from the
measurement site up to the sky. In practice, there will always be obstacles,
which should be mapped on a horizon plot for planning the measurement.
Many planning programmes let the user draw a horizon plot and take it
into account when calculating DOP.
The situation improves if the receivers to be used can make use of both
the GPS and GLONASS systems, meaning more satellites are seen, and
good measurement geometry is more easily achieved.
Of course, the GPS positioning geometry is only one factor among others.
Other factors affecting measurement precision include
◦ Only satellites above the horizon contribute to the determination of height, so

this is extrapolation. In the horizontal plane again, there are satellites in the
east and in the west, in the north and in the south, which contribute to the
positioning: interpolation.
◦ Moreover, because of the same non-symmetry, the vertical location unknown Z
and the clock unknown ∆T “compete” for the same information when they are
estimated together from the same observation data.
◦ The uncertainty in the signal delays caused by the atmosphere also affects mostly
in the vertical direction, whereas — also due to the horizontal stratification of
the atmosphere — the situation in the horizontal plane is more symmetric.
í Õ ! ¤. û
319
◦ The capabilities of receiver and antenna.
◦ Local disturbances, like multipath: reflections off the ground and
objects nearby.
◦ The activity of the Sun, ionospheric conditions.
◦ The measurement mode: static or kinematic, absolute (for example
“precise point positioning”, PPP) or relative.
◦ In relative GPS measurement, the distance between measurement
points or from the base station. Geodetic measurements are nearly
always relative and are carried out as network measurements.
◦ In static GPS measurement, the duration of measurement, the num-
ber of measurement epochs. Geodetic base-network measurements runkomittaus
are always static, despite the method being time-consuming, be-
cause of its robustness. Only in local measurements, like detail kartoitus-
surveys, is the faster kinematic technique (RTK, real-time kine- mittaus
matic) used.
^ 12.9.4 Example 1: an azimuthally symmetric geometry

Assume that the satellites are evenly distributed around the sky, according
to azimuth α, for every elevation angle η. Then, in the above weight
matrix 12.11:
◦ Almost all non-diagonal elements vanish, because they contain
either sin α or cos α or even sin α cos α. Only sin η does not vanish.
∑︁
◦
n
∑︂ n
∑︂ n
∑︂ n
∑︂
cos2 α i cos2 η i = 12 cos2 η i , sin2 α i cos2 η i = 12 cos2 η i .
i =1 i =1 i =1 i =1
Therefore, the weight or normal matrix Pxx = A T A becomes an almost

diagonal matrix that would be relatively easy to invert — although we
will not even try:
Pxx = A T A =
n
⎡ ⎤
1 2 i
cos η
∑︁
⎢ 2 0 0 0 ⎥
i =1
n
⎢ ⎥
1
cos2 η i
⎢ ∑︁ ⎥
⎢
⎢ 0 2 0 0 ⎥
⎥
i =1 ⎥.
=⎢ n n
sin2 η i i ⎥
sin η ⎥
⎢ ∑︁ ∑︁ ⎥
⎢
⎢ 0 0
i =1 i =1
n
⎢ ⎥
sin η i
⎣ ∑︁ ⎦
0 0 n
i =1
í Õ ! ¤. û
Transform now observation equation 12.8 and design matrix 12.9 in the
following way:
ℓ + v = A ΛΛ−1ˆ︁
x= A x,
˜︁˜︁
in which
−∆ X −∆ X
⎡ ⎤ ⎡ ⎤
ˆ︁ ˆ︁
⎢ −∆Yˆ︁ ⎥ ⎢ −∆Y
ˆ︁ ⎥
x=⎢ ⎥ = Λ−1ˆ︁
x=⎢ (12.13)
⎢ ⎥ ⎢ ⎥
⎣ −∆ Z −∆ Z
˜︁ ⎥
ˆ︁ ⎦ ⎣ ˆ︁ ⎦
(︁ 1 ∑︁
c∆ −∆ Z
ˆ︁ + c ∆ˆ︂
)︁(︁ )︁
˜︂T n sin η T
and
sin η1 − n1
⎡ ⎤
cos α1 cos η1 sin α1 cos η1 sin η 1
∑︁
⎢ cos α2 cos η2 sin α2 cos η2 sin η2 − n1 sin η 1 ⎥

⎢ ∑︁ ⎥
⎢ .. .. .. .. ⎥
⎢ . . . . ⎥
A = AΛ = ⎢ ⎥,
˜︁ ⎢ ⎥
⎢ cos α cos η i
i 1 ∑︁
sin α i cos η i i
sin η − n sin η 1 ⎥
.. .. .. .. ⎥
⎢ ⎥
. . . . ⎦
⎢
⎣
1 ∑︁
cos αn cos ηn sin αn cos ηn sin η − n n
sin η 1
22 because22
⎡ ⎤ ⎡ ⎤
1 0 0 0 1 0 0 0
⎢ 0 1 0 0 ⎥ ⎢ 0 1 0 0 ⎥
Λ=⎢ ⎥, Λ−1 = ⎢ ⎥.
⎢ ⎥ ⎢ ⎥
⎣ 0 0 1 0 ⎦ ⎣ 0 0 1 0 ⎦
0 0 − n1 sin η 1 0 0 + n1 sin η 1
∑︁ ∑︁
Now we get a clean diagonal matrix:
˜︁T A
P˜︁xx = A ˜︁ =
n
⎡ ⎤
1 2 i
cos η
∑︁
2 0 0 0 ⎥
i =1
⎢
n
⎢ ⎥
1 2 i
cos η
⎢ ∑︁ ⎥
⎢ 0 2 0 0 ⎥
=⎢ i =1 ⎥.
⎢ ⎥
n
(︃ n
)︃2
1 ∑︁
⎢ ⎥
i
sin η − n sin η j
∑︁
⎢
⎢ 0 0 0 ⎥
⎥
⎣ i =1 j =1 ⎦
0 0 0 n
Let us write the equation for the error or visual ellipsoid of P˜︁xx :
xT P˜︁xx˜︁
˜︁ x = 1,
22 Verify that ΛΛ−1 = Λ−1 Λ = I !
í Õ ! ¤. û
321
in which ˜︁
x is as in equation 12.13. The result is
2
ˆ︁ 2 + p 22 ∆Yˆ︁ 2 + p˜︁33 ∆ Z
p 11 ∆ X ˆ︁2 + p 44 ∆˜︂
T =
2
∆Xˆ︁ 2 ∆Yˆ︁ 2 ∆ Z
ˆ︁2 ∆˜︂
T
= q + q + + = 1,
xx yy q
˜︁zz q
˜︁tt
in which are immediately seen the elements of the variance matrix of the
unknowns
−1
)︁−1
= AT A
(︁
Q xx = Pxx
according to the definition:
2
q xx = q yy = ∑︁ ,
i cos η
2 i
from which
√︁ 2
HDOP = q xx + q yy = √︁∑︁ .
i cos η
2 i
Similarly, after a little reorganising,23 23
1 1 n
q
˜︁zz = = = ∑︁ ,
˜︁33 ∑︁ (︁sin η i − 1 ∑︁ sin η j )︁2 n
p 2 i
sin η − ( i sin η i )
2
∑︁
i n j i
and VDOP is its square root.24 24
^ 12.9.5 Example 2: a singular case

Look again at design matrix A , expression 12.9, and write it into the form
⎡ (1) (1) (1) ⎤ ⎡ (︁ )︁T ⎤
e1 e2 e3 1 e(1) 1
⎢ (2) (2) (2) )︁T
1 ⎥ ⎢ e(2)
⎥ ⎢ (︁ ⎥
⎢ e1 e2 e3 1 ⎥
A=⎢
⎢ .. .. ⎥ = ⎢
⎥ ⎢
.. .. ⎥ ,
⎥
⎣ . . ⎦ ⎣ . . ⎦
( n) ( n) ( n) (︁ (n) )︁T
e1 e2 e3 1 e 1
23 Because
(︄ )︄2
∑︂
i 1 ∑︂
sin η − sin η j =
n constant constant
i j ⏟ ⏞⏞ ⏟ ⏞⏞ ⏟
(︄ )︄⏟ n
(︄ )︄2
∑︂ ∑︂ 2 ∑︂ ∑︂ 1 ∑︂
= sin2 η i − sin η i sin η j + sin η j =
n n2
i i j i =1 j
(︄ )︄(︄ )︄ (︄ )︄2
∑︂
2 i2 ∑︂
i
∑︂
i 1 ∑︂
i
= sin η − sin η sin η + sin η =
n n
i i i i
(︄ )︄2
∑︂
2 i 1 ∑︂
i
= sin η − sin η .
n
i i
24 If η is constant — so, all η i , i = 1, . . . , n are the same — then the denominator vanishes!
So, the determination of height by GPS requires that there are satellites on different
elevations in the sky.
í Õ ! ¤. û
⎷
1 − a2 e2
⎷
1 − a 2 e1
Difference
observation ae0
∆R
Observer
^ F IGURE 12.20. The circle singularity or “dangerous circle” for GPS.
assuming there are n satellites. If the satellites are all on the same circle,
then the unit direction vector in the direction of satellite i is
( i) ( i) ( i)
e( i ) = e 1 i + e 2 j + e 3 k = a e0 + b ( i ) e1 + c ( i ) e2
(︁ )︁2 (︁ i )︁2
in which the values b( i) and c( i) satisfy the condition b i + c = 1 − a2 ,
for all satellites i = 1, . . . , n. Here, ae0 is the vector from the observer to
the centre of the circle. Thus there are only three independent vectors
e( i) when four are needed.
See figure 12.20. The situation is also geometrically clear: if the obser-
vation site is shifted along the direction of the vector e0 , any difference
between the pseudo-ranges from two different satellites will remain un-
changed. What is happening here is that the receiver’s clock unknown ∆T
and the component of the observation site’s location in the e0 direction
(in other words the projection onto the e0 direction) cannot be separated
from each other in this geometry.
This is the circle singularity or “dangerous circle” in the case of GPS
taaksepäin positioning: compare section 6.2. GPS positioning is in fact a three-
leikkaus dimensional resection!
^ 12.9.6 Calculation example for DOP quantities

See figure 12.21.
Let one satellite be in the zenith (η = 90◦ ) and three satellites at an
elevation angle η = 30◦ at azimuths α = 0◦ , 120◦ , 240◦ . Compute first
í Õ ! ¤. û
323
1
4
N 3
120◦ 120◦ 30◦

W E
120◦
^ F IGURE 12.21. Calculation example of DOP quantities.
design matrix A according to equation 12.9. It is given that
α1 = 0◦ , η 1 = 90◦ ,
α2 = 0◦ , η 2 = 30◦ ,
α3 = 120◦ , η 3 = 30◦ ,
α4 = −120◦ , η 4 = 30◦ .
Numerical values are obtained by remembering that
sin(90◦ ) = 1, cos(90◦ ) = 0,
⎷
sin(30◦ ) = 12 , cos(30◦ ) = 21 3,
⎷
sin(120◦ ) = − sin(−120◦ ) = 12 3, cos(120◦ ) = cos(−120◦ ) = − 12 .
The result is
⎡ ⎤ ⎡ ⎤
0 0 1 1 0 0 0 1
⎢ 1⎷ 1 ⎥ ⎢ 1⎷ 1
2 3 0 2 1 ⎥ ⎢ 2 3 0 1 ⎥
⎢ ⎥
2
A=⎢ ⎷ ⎷ = ⎢ 1⎷ ⎥.
⎥
1 1 1 2 1
⎣ − 4 3 34 1
(︁ )︁
⎢ −2 · 2 3 2 3 2 1 ⎦
⎥
2 1 ⎦
⎷ ⎷ 1⎷
⎣
2 1
− 12 · 12 3 − 21 3 3 − 34 1
(︁ )︁
2 1 − 4 2 1
Next we compute the weight matrix of the unknowns, or normal matrix,

equation 12.10:
⎡ 9 ⎤
8 0 0 0
⎢ 0 9 0 0 ⎥
Pxx = A T A = ⎢ 8
⎥.
⎢ ⎥
⎣ 0 0 34 32 ⎦
3
0 0 2 4
í Õ ! ¤. û
Inverting this matrix would yield Q xx , the weight-coefficient matrix of

the unknowns. Here, we invert the matrix only partially:
8
⎡ ⎤
0 0 0
⎡ ⎤
9 q xx
8
⎢ 0
⎢ ⎥ ⎢
−1 9 0 0 ⎥ ⎢ q yy ⎥
Q xx = Pxx =⎢ ⎥=⎢ ⎥.
[︄ ]︄−1 ⎥
3 3 q zz q zt
⎣ 0 0
⎢ ⎥ ⎣ ⎦
4 2 ⎦
3 q tz q tt
0 0 2 4
(12.14)
From this we read directly that the weight coefficients of the co-ordinates
X and Y are
q xx = q yy = 89 = 0.888 . . . ,
and thus √︂
16
= 34 = 1.333 . . . .
√︁
HDOP = q xx + q yy = 9
From equation 12.14 we see that the Z

ˆ︁ co-ordinate and the clock unknown
c∆
ˆ︂T are “entangled” with each other ( q zt ̸= 0) and calculating their weight
25 coefficients is not attempted here.25
^ 12.10 Orbits of the GPS satellites

The orbit of a GPS satellite in the Earth’s gravitational field is approxi-
mately an ellipse satisfying Kepler’s laws. In practice, it is almost a circle,
26 the radius26 of which is 26 560 km and the orbital period 11h 58m . The tilt
angle of the orbital planes with respect to the equator, the inclination, is
i = 55◦ , meaning that at the latitude of Finland, the GPS satellites will
25 In fact, the sub-matrix to be inverted is

[︄ ]︄−1 [︄ ]︄
3 3 16
4 2 3 −2
3
= ,
2 4 −2 1
and thus
8
⎡ ⎤
9
⎢ 8 ⎥
Q xx = ⎢ 9
⎥.
⎢ ⎥
16
⎣ 3 −2 ⎦
−2 1
This is now the weight-coefficient matrix (and, up to the variance of unit weight, the
]︁T
c∆
[︁
variance matrix) of the vector of unknowns −X
ˆ︁ −Y
ˆ︁ −Z
ˆ︁ ˆ︂T .
In this result ˆ︁ and ∆ˆ︂
is also seen, how Z T “compete” for the same information:
√︂ VDOP =
⎷
√︂
16 4
q zz = 3 ≈ 2.309, when, without the clock unknown, it would be VDOP = 3 ≈ 1.155.
26 So: the distance from the Earth’s surface is about 26 560 km − 6378 km = 20 182 km,
using 6378 km for the Earth’s radius.
í Õ ! ¤. û
Orbits of the GPS satellites 12.10
325
N
North pole
Zenith
E W
F IGURE 12.22. The six orbital planes of GPS satellites in the Helsinki sky. GPS
satellites will never be found inside the oval around the zenith,
although they are also visible in the northern sky, though very
^ close to the horizon.
never pass through the zenith. However, due to their great height, the
satellites are also visible “over the North Pole” in the northern half of the
sky, albeit very low. See figure 12.22.
We need six orbital elements to describe a satellite orbit. As orbital rata-alkiot
elements we may choose, for example, the three co-ordinates of position
r( t 0 ) = x( t 0 ) · i + y( t 0 ) · j + z( t 0 ) · k
and the three components of velocity
dr ⃓
⃓
= ṙ ( t 0 ) = ẋ ( t 0 ) · i + ẏ ( t 0 ) · j + ż ( t 0 ) · k
dt t= t0
⃓
at a certain time t 0 , using Newton’s dot notation for the time derivative.27 27
{︁ }︁
The vectors i, j, k form an orthonormal basis. Position and velocity,
r ( t ) = x ( t ) · i + y ( t ) · j + z ( t ) · k, ṙ ( t) = ẋ ( t) · i + ẏ ( t) · j + ż ( t) · k,
can be calculated from these equations for some later moment t, by

just calculating, step by small step, forward in time, correcting both the
27 This dot notation for the derivative of time, fluxion, was introduced by Newton in
1665.
í Õ ! ¤. û
Motion in Earth’s gravitational field
New position r( t)
New velocity ṙ ( t) ṙ ( t 0 ) Velocity
r( t 0 ) Position
F IGURE 12.23. Satellite orbital motion described by position and velocity vec-
^ tors.
velocity using the gravitation equation and the position using the velocity,
figure 12.23. We know the attraction field of the Earth as an equation:
the acceleration caused by the attraction is computable when we know
the position in space of the satellite.
The geometry of a satellite orbit is normally described using the six
28 Kepler orbital elements,28 Ω, i , ω, a, e, and ν, see figure 17.14. More
details are given in Poutanen (2017) section 5.1, Hofmann-Wellenhof
et al. (2001) subsection 4.2.1, and in section 17.7. There is a one-on-
one correspondence between the Kepler elements and the position and
velocity vector representation described above:
Kepler: Ω, i, ω, a, e, ν ←→ r, ṙ .
{︁ }︁ {︁ }︁
This means that, from the given Kepler elements, we may calculate
the position of the satellite in space, as well as its velocity. All GPS
computation software packages know how to do this.
28 So every satellite has six Kepler orbital elements that describe the shape, size and
orientation of the orbit of that satellite, as well as the location of the satellite in its orbit,
in space.
í Õ ! ¤. û
327
^ 12.10.1 Navigation message transmitted by the satellites
As already noted, all GPS satellites transmit a navigation message mod-
ulated upon the carrier wave of their radio signal. The modulation
frequency of the navigation message is 50 Hz, so every second contains 50
bits. The whole navigation message consists of 25 “packets” (frames), each
of which contains 1500 bits and transmission of which lasts 30 seconds.
Thus, the total length is 37 500 bits, and the duration of the transmission
is 12.5 minutes.
When a GPS receiver is switched on for the first time, the search for
satellites starts. Immediately when the first satellite is “caught” (lock-on),
the reading of the navigation message starts. Lock-on may easily last for
several minutes, especially if the approximate location fed to the receiver
is seriously wrong, or the instrument has been transported between
continents. After that, however, finding the other satellites generally
proceeds apace.
The navigation message is regularly uploaded to the satellites by the
control segment. After that, the information is, as part of the signal of the
GPS satellites, available to all users of the GPS. The navigation message
consists of three parts.
◦ Information related to time keeping, like the clock corrections of the
satellites, the “health” information on the satellites, the quality of
the positioning signal and orbital data transmitted by the satellite,
and the freshness of the navigation message.
◦ The satellite’s orbital information (broadcast ephemeris), dissem-
inated by the satellite by radio. These ephemeris are computed
by the US military authorities and are based on observation data
continuously produced by a global network of tracking stations. The
orbital elements of all satellites computed from the observations are
uploaded to the satellites by the control segment, typically once per
24 hours. They are then transmitted from the satellite’s memory,
modulated onto the radio signal, as a bit stream to all users.
The orbital elements are the Kepler elements augmented by coef-
ficients used to calculate the perturbations caused by the Earth’s
flattening, both secular (growing linearly with time, in the mean keskiliike
motion, orbital inclination29 and right ascension of the ascending 29
node) and short-period (half the satellite orbital period), in the or- rektaskensio
29 Not actually secular: rather, long-period.
í Õ ! ¤. û
nouseva solmu bital inclination, the radius and the angle between ascending node
and satellite (the argument of latitude). There are nine coefficients,
which for GPS satellites need to be taken into account. The origin of
the theory used is the classical article by Yoshihide Kozai (Kozai,
1959).
Broadcast ephemeris are used in navigation applications and in real-
time positioning. It is also practical to use them in GPS surveying
and relative positioning in relatively small areas.
From the ephemeris, every satellite’s position in space at the mo-
ment of observation is computed, so that they may be used as
“beacons” for the determination of the location of the ground sta-
tion. From the ephemeris, also the velocity of the satellite is com-
30 puted.30 More is said about the computing methods used in chapter
5 of Poutanen (2017) and in chapter 4 of Hofmann-Wellenhof et al.
(2001).
31 ◦ The almanac for all satellites.31 The purpose of the almanac is to
provide approximate orbital elements for all satellites, sufficient
for planning measurement campaigns and helping the receiver find
satellites. The almanac is valid for many weeks. The almanac also
contains a crude global ionosphere model.
^ 12.10.2 Precise ephemeris

Precise ephemeris, precise orbital elements, may be obtained by the user
some time afterwards directly from the Internet. They are computed and
distributed by a service called the International GNSS Service (IGS), see
section 12.11. A standard format, SP3 (“Standard Product 3”) is used
for data distribution, originally designed by the US National Geodetic
Survey. See tableau 12.7.
The SP3 ephemeris file contains the orbital data in the form of three-
dimensional component vectors of position and velocity,
⎡ ⎤ ⎡ ⎤
x( t i ) ẋ ( t i )
⎣ y( t i ) ⎦ , ⎣ ẏ ( t i ) ⎦ ,
⎢ ⎥ ⎢ ⎥
z( t i ) ż ( t i )
30 Knowing the velocity of the satellite would not be necessary for this, but is needed to
calculate the Doppler shift of the signal frequency. The receiver must know the Doppler
shift of every satellite in order to lock on, and remain locked on, to the satellite signal.
31 Almanac is apparently not a word of Arabic origin, although it looks like it is.
í Õ ! ¤. û
329
^ T ABLEAU 12.7. Precise ephemeris in the original SP3 format. Satellite numbers,
position vectors, velocity vectors, clock correction and clock drift, date and time,
etc. Start of example file © US National Geodetic Survey.
#aV1994 12 17 0 0 0.00000000 96 d ITR92 FIT NGS

## 779 518400.00000000 900.00000000 49703 0.0000000000000
+ 25 1 2 4 5 6 7 9 12 14 15 16 17 18 19 20 21 22
+ 23 24 25 26 27 28 29 31 0 0 0 0 0 0 0 0 0
+ 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
+ 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
+ 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
++ 7 6 5 5 5 5 5 5 5 6 5 5 5 5 6 5 5
++ 5 5 6 5 5 5 5 5 0 0 0 0 0 0 0 0 0
++ 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
++ 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
++ 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
%c cc cc ccc ccc cccc cccc cccc cccc ccccc ccccc ccccc ccccc
%c cc cc ccc ccc cccc cccc cccc cccc ccccc ccccc ccccc ccccc
%f 0.0000000 0.000000000 0.00000000000 0.000000000000000
%f 0.0000000 0.000000000 0.00000000000 0.000000000000000
%i 0 0 0 0 0 0 0 0 0
%i 0 0 0 0 0 0 0 0 0
/* CCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCCC
* 1994 12 17 0 0 0.00000000
P 1 16258.524750 -3529.015750 -20611.427050 -62.540600
V 1 -6560.373522 25605.954994 -9460.427179 -0.024236
P 2 -21998.652100 -8922.093550 -12229.824050 -131.326200
V 2 -9852.750736 -12435.176313 25738.634180 -0.029422
P 4 -26019.547600 4809.810900 -2508.578200 3.544600
V 4 2559.038002 -3340.527442 -31621.490838 0.016744
*
*
*
P 29 -1638.431050 -24391.479200 10455.312650 3.690300
V 29 5754.005457 -12065.761570 -27707.056273 0.003537
P 31 6265.255800 -25687.986950 -753.359000 70.830800
V 31 3053.344058 -63.091750 31910.454757 0.033749
* 1994 12 17 0 15 0.00000000
P 1 15716.820135 -1169.850490 -21281.578766 -62.542746
V 1 -5439.955846 26738.341429 -5409.793390 -0.023226
P 2 -22813.261065 -9927.616864 -9816.490189 -131.328686
V 2 -8178.974330 -9924.329320 27813.754308 -0.025238
*
*
í Õ ! ¤. û
X
−180◦ −150◦ −120◦ −90◦ −60◦ −30◦ 0◦ 30◦ 60◦ 90◦ 120◦ 150◦ 180◦ ◦
90◦ 90
60◦ 60◦
30◦ 30◦
X
◦ ◦
0 0
−30◦ −30◦
−60◦ −60◦
−90 −90◦
−180◦ −150◦ −120◦ −90◦ −60 ◦ ◦
−30 0 ◦
30 ◦
60 ◦
90◦
120 ◦ ◦
150 180◦
GMT 2018 Mar 17 21:37:35
^ F IGURE 12.24. The tracking stations of the IGS, situation in 2018 (data © IGS).
for epochs t i . The velocity vector is optional. The data is tabulated at

time intervals of 15 minutes: t i+1 − t i = 15m . From these, the position r ( t)
and velocity ṙ ( t) are interpolated to the moment of measurement t by
32 Lagrange32 interpolation. In addition, the ephemeris contain information
characterising the behaviour of the satellite clock and the accuracy of the
orbital information, among other things.
The current version of the format is SP3-c, which also allows the distri-
bution of orbital information on GLONASS satellites. It is a textual format
and human-readable.
The best-known source from as early as 1992 has been the International
GNSS Service. The precise orbits produced by them are published on the
Internet a couple of weeks after the time of observation.
In addition to precise ephemeris, “rapid orbits” are also being produced.
These are almost as precise as precise ephemeris, but are turned out
faster.
The newest “ultra-rapid” solutions are satellite orbital predictions
twenty-four hours into the future, which can thus be used in real-time
applications.
32 Joseph-Louis Lagrange (1736–1813) was a French mathematician, astronomer, devel-

oper of classical mechanics, one of the 72 names on the Eiffel Tower, Eiffel Tower, 72
names.
í Õ ! ¤. û
The International GNSS Service IGS 12.11
331
Precise ephemeris are, unlike broadcast ephemeris, disseminated over
the Internet and not via GPS satellites. The organisations producing them
are independent from the United States military authorities.
Unlike broadcast ephemeris, precise ephemeris are very close to the
true orbits of the satellites, where they actually were at that moment.
Broadcast ephemeris are predictions and therefore less accurate.
^ 12.11 The International GNSS Service IGS

The IAG (International Association of Geodesy) established the Interna-
tional GNSS Service (IGS) in 1990, and it became an official service of
the IAG in 1994. The main purpose of the service is to produce precise
orbital ephemeris in support of geodynamics research, like the study of
the motions of the solid Earth. However, its products are used much more
broadly, in many fields of geophysics.
The activities of the IGS are led by a Central Bureau, currently at the
JPL (Jet Propulsion Laboratory) in the United States. In 2015 the IGS
used observations made globally by some 500 GNSS stations to compute
its orbital ephemeris. The number has grown only slowly over recent
years.
The computation as such is carried out by seven different computing
centres, and the orbital data is available for use a couple of weeks after the
time of measurement. The computed data also include the clock correction
parameters for the satellites. The Earth’s orientation parameters (EOP)
polar motion and variations in the length of the day (LoD) are published vuorokauden
separately. See IGS Central Bureau. pituus
1. Explain how a hyperbolic positioning system like Decca functions.
2. Why does GPS broadcast on two different carrier frequencies?
3. How does the densest part of the Earth’s atmosphere, the tropo-
sphere, affect the propagation of GPS radio waves? Why are meteo-
rologists interested in this?
4. How is it possible that all GPS satellites broadcast on the same
frequencies? How does the receiver separate out the signals from
the different satellites?
í Õ ! ¤. û
^ T ABLEAU 12.8. DOP calculation script.
% Dilution of precision (DOP):

% Part 1. Run the program. What does the error message tell you?
% What can you say about the value of VDOP?
% Places of satellites, azimuth A and elevation h in the sky (below).
% Part 2. Change the elevation of the first satellite 30 -> 60 degrees.
% Run again. Note down the HDOP and VDOP values.
% Why is the run now successful?
A1 = 0; h1 = 30;
A2 = 90; h2 = 30;
A3 = 180; h3 = 30;
A4 = 270; h4 = 30;
conv = pi/180;
% Design matrix A (below)
% Part 3. Modify the program to include a fifth satellite, place
% in the sky A5 = 0, h5 = 45.
A = [cos(A1*conv)*cos(h1*conv), sin(A1*conv)*cos(h1*conv), sin(h1*conv), 1;
cos(A2*conv)*cos(h2*conv), sin(A2*conv)*cos(h2*conv), sin(h2*conv), 1;
cos(A3*conv)*cos(h3*conv), sin(A3*conv)*cos(h3*conv), sin(h3*conv), 1;
cos(A4*conv)*cos(h4*conv), sin(A4*conv)*cos(h4*conv), sin(h4*conv), 1];
N = A’*A;
Ninv = inv(N)
HDOP = sqrt(Ninv(1,1) + Ninv(2,2));
VDOP = sqrt(Ninv(3,3));
% Part 4. Add to the program the evaluation and output of PDOP.
% PDOP = Position dilution of precision. See lecture notes
% for definition.
fprintf(1, ’HDOP = %20.10f\n’, HDOP);
fprintf(1, ’VDOP = %20.10f\n\n’, VDOP);
% Part 5. Play around with the five satellite places in the sky,
% in order to minimise PDOP.
5. Describe how the pseudo-random codes modulated on the GPS car-

rier wave make it possible to measure pseudo-ranges. Where does
correlation come in?
6. What is multipath, and what methods are available to minimise its
influence?
7. Explain the concept of DOP, dilution of precision.
8. Why are geodesists interested in measuring the phase of the car-
rier wave of the GPS signal, even though it is harder than code
measurement?
9. What are broadcast ephemeris, who generates them, and how does
í Õ ! ¤. û
Exercise 12–1: Calculation of DOP quantities
333
the user acquire them?
10. What are precise ephemeris, who generates them, and how does
the user acquire them?
11. What are in tableau 12.3 the units of the observables L1, L2, P1,
P2, and C1? Do not look up the RINEX format definition!
^ Exercise 12–1: Calculation of DOP quantities

In this exercise you are going to write code to calculate the various
DOP quantities of a GPS measurement geometry, when the places of the
satellites in the sky, their azimuths α and elevation angles η, are given.
You may use your preferred rapid prototyping language: MATLAB™,
GNU Octave, Scilab (Scilab Enterprises), R (The R Project for Statistical
Computing), even Excel™. And do not bother with inputting from files,
just put the satellite positions into the source.
1. For an arbitrary number of satellites, write, or adapt, code to build
the design matrix A and the normal or weight matrix Pxx = A T A .
One could also build, from the vector of observations ℓ, the right-
hand side of the system of normal equations, A T ℓ, but we do not
need it here. The beauty of DOP is that we can use it in measurement
planning, before any real observations are available.
2. Make your software output to the screen the various DOP quantities.
3. Now assume you have five satellites, one more than the minimum to
make positioning possible. Play with the satellite positions (α i , η i ),
i = 1, . . . , 5, in order to minimise one of the DOPs, for example the
PDOP. What is the best geometry of five satellites you find?
4. You cheated, didn’t you? You cannot observe GPS satellites that are
below the horizon. So, introduce the constraint η > 5◦ .
5. After PDOP, try to minimise HDOP, and VDOP.
What did you learn?
í Õ ! ¤. û
^ Processing GPS observations
13
[. . . ] everything is related to everything else, but near
things are more related than distant things. [. . . ]
First law of geography, by Waldo R. Tobler (1970)
^ 13.1 Forming difference observations

In geodetic GPS measurement and GPS surveying applications, commonly
one wishes to measure the difference in location between two points, like
a point pair in a local geodetic network. The distance between the points
may be in the order of 100–1000 km. This is a much shorter distance
than that to the GPS satellites, which orbit at a height of some 20 000 km.
See figure 13.1.
Sodankylä
α
Helsinki
^ F IGURE 13.1. The “common-mode” error assumption.
– 335 –
336 13 P ROCESSING GPS OBSERVATIONS
Seen from the satellite, the angle α separating the observation sites
is very small, in the example case (Helsinki and Sodankylä) only 2◦ !
For this reason, many errors will be partly common to the two sites,
similar and approximately equal in magnitude. The effect of the satellite
kellopoikkeama clock error or clock offset is even identical at both observation sites.
The effect of orbit error is, due to the geometry, roughly the same; the
errors caused by ionosphere and troposphere are also similar, due both
1 to the similarity in geometry and the long-range spatial correlation1 of
atmospheric conditions.
On the other hand, however, one should remember that the difference
luotiviiva in directions of the plumb line between Helsinki and Sodankylä is already
korkeuskulma 7◦ , so the difference in elevation angle of a satellite above the local horizon
may amount to this much because of that alone.
Based on the “common-mode” error assumption, we form differences
between observations from two sites to one satellite. In these differences,
many errors vanish entirely or are materially reduced. Forming the
difference is straightforward: subtract two simultaneously made raw
observations from each other, each lifted from an observation file looking
like tableau 12.3.
The differences can be single — either between two receivers or between
two satellites, in which cases one can use the visually appropriate symbols
∆ or ∇ — double, or triple, between successive measurement epochs,
symbol δ. See figure 13.2.
The influence of forming the various difference types on the magnitude
of errors — the interesting thing here! — has been catalogued in table
13.1.
^ 13.1.1 Single differences

We explain with equations how a single difference is calculated from
original observations. We shall see how some systematic errors are
eliminated altogether while others are substantially reduced.
One receiver (observer) A , two satellites S, T
def def
p ST T S
A = pA − pA,
ST
PA T
= PA S
− PA .
1 By this is meant that conditions change only slowly with place. Helsinki and Sodankylä
lie practically in the same climate zone, and if there is a high or low pressure zone over
Northern Europe, it will undoubtedly affect both places. The synoptic scale (Wikipedia,
Synoptic scale in meteorology) of weather phenomena is of the order of 1000 km.
í Õ ! ¤. û
Forming difference observations 13.1
337
Satellite S Satellite T
∇
∆
Receiver A Receiver B
(a) (b)
Single difference Single difference between
between receivers satellites
Epoch 2
Epoch 1
∆∇
δ∆∇
(c) (d)
Double difference Triple difference
F IGURE 13.2. Forming difference observations of various types and symbols

^ used for them.
Write the observation equation 12.1 for pseudo-range in this ex-

tended notation:
p SA = ρ SA + c ∆T A − ∆ tS + d ion,
S S
(︁ )︁
A + d trop, A ,
p TA = ρ TA + c ∆T A − ∆ tT + d ion,
T T
(︁ )︁
A + d trop, A .
It has been taken into account that, of the clock offsets, ∆ t is

satellite-specific, ∆T again is observer-, that is, receiver-, specific.
Subtraction yields the difference quantity
p ST ST
A = ρ A − c ∆t
ST ST
+ d ion, ST
A + d trop, A ,
í Õ ! ¤. û
^ T ABLE 13.1. Effect of forming difference observations from GPS observations on

the magnitude of various errors.
Type of difference
Error source ∇ST ∆ AB ∆ AB ∇ST δ12 ∆ AB ∇ST
Satellite orbit {r, ṙ } - ↓ ↓ ↓↓

Satellite clock ∆ t - 0 0 0
Receiver clock ∆T 0 - 0 0
Ionosphere d ion , D ion - ↓ ↓ ↓↓
Troposphere d trop , D trop - ↓ ↓ ↓↓
Ambiguities N - - - 0∗
↓ The error is reduced substantially, especially for short distances between

measurement points.
↓↓ The error is diminished even more strongly.
0 The error is completely eliminated.
0∗ The error is eliminated, unless there is a cycle slip.
in which the definitions apply

def def
ρ ST T S
A = ρA − ρA,
ST
d ion, T S
A = d ion, A − d ion, A ,
def def
∆ tST = ∆ tT − ∆ tS , ST
d trop, T S
A = d trop, A − d trop, A .
Here, the clock offset of receiver A , ∆T A , has vanished, because,

being a receiver property, it is the same for different satellites
and cancels out in calculating the difference observation between
satellites.
This is important in practice, because receiver clocks are usually
käynti based on inexpensive quartz oscillators, the drift of which is so
large that it needs to be taken into account.
A similar observation equation derived from equation 12.2 also
kantoaallon applies for the raw carrier-phase observable:
vaihe
ST φ ST ST
PA =λ = P A − λN A =
2π
= ρ ST
A − c∆ t
ST ST
+ D ion, ST ST
A + D trop, A − λ N A ,
[︁ )︁
in which we may assume φ ∈ 0, 2π , and
ST def
T S
NA = NA − NA
kokonaisluku- is the difference of the ambiguities between satellites S and T .

tuntematon Without a cycle slip occurring, it will be constant in time.
vaihekatko
í Õ ! ¤. û
Forming difference observations 13.1
339
Two receivers A, B, one satellite S
def def
p SAB = p SB − p SA , S
P AB = PBS − P A
S
. (13.1)
Here, in the same way, the satellite clock offset ∆ t drops out: it is
a property of the satellite, not the receiver, and vanishes when one
calculates the difference quantity between two different receivers
with the same satellite.
Moreover, the influence of orbit errors, ionosphere and troposphere
is reduced substantially: for short distances AB between receivers
it holds that
⃓ S ⃓ def ⃓ S
⃓ρ ⃓ = ⃓ρ − ρ S ⃓ ≪ ⃓ρ S ⃓ ≈ ⃓ρ S ⃓
⃓ ⃓ ⃓ ⃓ ⃓
AB B A A B
and for AB → 0:
S def S S
d ion, AB = d ion, B − d ion, A → 0,
S def
S S
d trop, AB = d trop, B − d trop, A → 0,
S def
S S
D ion, AB = D ion, B − D ion, A → 0,
S def
S S
D trop, AB = D trop, B − D trop, A → 0,
because
◦ The measurement geometry is almost the same at point A as
at point B, see figure 13.1.
◦ Atmospheric conditions do not change much between points
A and B: the measurement rays traverse nearly the same air
mass.
◦ The elevation angle in the sky of satellite S seen from point
A is nearly the same as that seen from point B.
^ 13.1.2 Other difference quantities

In the same way, by combining the operations described above, we may
also calculate double and triple differences. The formulas look compli-
cated but the process is straightforward, see the summary in tableau
13.2. The equations in the tableau are directly derived from the original
observation equations 12.1 and 12.2 by addition and subtraction.
Double differences are much-used in geodetic software packages for
processing static GPS network measurements. The double differences
are constructed in the office from measurements simultaneously collected
in several locations from common satellites.
í Õ ! ¤. û
^ T ABLEAU 13.2. Summary of GPS observables and difference quantities. The

notations d = d ion + d trop , D = D ion + D trop are used in the equations. Note the
consistent use of super- and subscripts.
Carrier phase φ,
Pseudo-range p equivalent pseudo-range P
p SA = ρ SA + c(∆T A − ∆ tS ) + d SA S
PA = ρ SA + c(∆T A − ∆ tS ) + D SA − λ N A
S
Single difference, between satellites:

p ST
A = ρ ST
A − c ∆ t ST
+ d ST
A
ST
PA A − c∆ t
= ρ ST ST
+ D ST ST
A − λN A
Single difference, between receivers:

p SAB = ρ SAB + c∆T AB + d SAB S
P AB = ρ SAB + c∆T AB + D SAB − λ N AB
S
Double difference:
p ST
AB = ρ ST ST
AB + d AB
ST
P AB = ρ ST ST ST
AB + D AB − λ N AB
Triple difference:
δ12 p ST
AB = δ12 ρ ST ST
AB + δ12 d AB
ST
δ12 P AB = δ12 ρ ST ST
AB + δ12 D AB − λ · (cycle slips)
Triple differences, again, calculated by subtracting double differences

for successive time epochs from each other, are almost exclusively used
for cleaning up carrier-phase observations, as they are uniquely able to
detect “cycle slips”: sudden changes in integer ambiguities caused by
interruption of the radio connection between satellite and receiver.
^ 13.2 Relative (static) GPS

The difference in location, or vector, R AB = RB − R A between two observa-
tion sites A and B close to each other may be solved more precisely than
the absolute location of either site R A or RB with respect to the centre
of mass of the Earth. The reason for this is the cancellation or partial
cancellation from simultaneous observations of various error sources,
the impacts of which are similar in both points. Difference observations
p AB = p B − p A , P AB = PB − P A between the two observation sites, in which
this cancellation already occurs, are used for computation. As shown
earlier in figure 13.1, the places of the satellites in the sky are nearly the
same seen from observation sites A and B, and the atmospheres above A
and B are also likely rather similar.
How much, for example, does the orbit error of satellites S and T
affect the determination of the vector AB? See figure 13.3. An order-of-
magnitude rule of thumb says that the positioning error δ caused by orbit
í Õ ! ¤. û
Relative (static) GPS 13.2
341
S ∆ T
A d δ
B
^ F IGURE 13.3. Double difference, short distance between GPS receivers.
error in point B relative to point A is
d
δ ≈ s ∆,
in which ∆ is the assumed orbit error. This is only a crude estimate. We

know that s ? 20 000 km.
The values given in table 13.3 for the orbit error, 1 m and 2 cm, corre-
spond to the accuracies of today’s broadcast2 and precise ephemeris. The 2
conclusion is that
^ T ABLE 13.3. Approximate relation between orbit error, length of vector, and
positioning error.
Vector length d (km) Orbit error ∆ (m) Positioning error δ (mm)
1 1 0.05
10 1 0.5
100 1 5
1000 1 50
1 0.02 0.001
10 0.02 0.01
100 0.02 0.1
1000 0.02 1
2 This is a crude estimate. The quality of broadcast ephemeris has improved since the
early days of GPS, slowly but surely. Other GNSS systems, like GLONASS, Galileo and
BeiDou, perform at about the same level, or perhaps a little less well (Montenbruck
et al., 2015).
í Õ ! ¤. û
Possible λ1 solution Common solution

λ1
λ2
2s
s
Possible λ2 solution
Sent
Received
F IGURE 13.4. One-dimensional ambiguity resolution in the case of a distance

measurement instrument. See also figure 7.10. Depicted is the
situation where the transmitted phases cross zero in the positive
direction. It is assumed that at that moment the received phases
^ are all measured to be zero too, see figure 7.9.
In GPS surveying work in a small area (> 100 km) the satellite
orbits may be assumed known.
ST
In geodetic work, first one computes double differences P AB from ob-
servations at points A and B. As we are dealing with carrier-phase
ST
observations, the ambiguities or integer unknowns N AB must first be
resolved. After that, a vector
⎡ ⎤ ⎡ ⎤
X AB XB − X A
R AB = ⎣ Y AB ⎦ = ⎣ YB − Y A ⎦
⎢ ⎥ ⎢ ⎥
Z AB ZB − Z A
between the points is computed from the observations. This is where the
term relative GPS (or GNSS) measurement comes from.
A generalisation of this vector solution is the measurement and pro-
cessing of observations from a number of points, a geodetic network.
^ 13.3 Fixing ambiguities

Resolving the integer unknowns or ambiguities is a precondition for using
kantoaalto GPS carrier-wave observations. There are several methods for this.
◦ Distance measurement equipment resolves the integers by measur-

ing at several different wavelengths. Figure 13.4, as earlier on
í Õ ! ¤. û
Fixing ambiguities 13.3
343
2 3 n
io 1b
ot Mo
4 M tio
n
1 2−3 3− 2a
1−
2 4 X 1b 1a−2a 2a−
1a − → →2
1a b 2b
Common
solution Common
solution
Search area
(e.g., code
positioning) Search area
(a) (b)
Three-dimensional Three-dimensional plus time
F IGURE 13.5. Various ambiguity-resolution methods used in GPS computation.

The coloured line bundles — actually cut-outs of bundles of hy-
perbolas, or in three dimensions, hyperboloids of revolution —
indicate the sets of relative receiver position solutions that are
compatible with double-difference observations obtained using
either a pair of satellites, or a single satellite at two different
^ times.
figure 7.10, shows how this puzzle can be solved.

A GPS satellite transmits on two frequencies L 1 and L 2 . When there
are two frequencies, they can be combined in a way which makes
it easier to resolve the ambiguities over short distances. Calculate
the phase difference φw = φ1 − φ2 between the phase measurements
at L 1 and L 2 . It is like using a carrier wave with a frequency
f w = f 1 − f 2 = 347.82 MHz, and a corresponding “wavelength” λw =
c f = 86 cm. This method is called wide-laning.
/︁
leveäkuja-
The wide-lane solution only works over short distances, because ratkaisu
otherwise the unknown difference in ionospheric influence between
the two measurement sites grows too large.
The code observation already yields a pseudo-range at the metre
precision level, after which the wide-laning method yields the am-
biguities, and the phase difference φw yields an “ambiguity-free” vaihe-ero
pseudo-range at a precision level of centimetres. pseudoetäisyys
◦ Use many satellites. At the moment, over 30 GPS satellites orbit
í Õ ! ¤. û
the Earth. At almost any moment, 6–10 of these are in the local
sky above the horizon. This is a more complicated method, because,
unlike in distance measurement, the geometry is three-dimensional.
Efficient algorithms for this exist.
◦ Use the same satellites for a longer time. Because the GPS measure-
ment geometry has time to change, we obtain more conditions.
See figure 13.5.

In recent years, other global positioning systems have appeared besides
GPS. The Russian GLONASS — in which every satellite has its own carrier
taajuusjako- transmission frequency, frequency division multiple access (FDMA), which
kanavointi complicates ambiguity resolution3 — is operating, after a long period
kokonaisluku-
of decay, again with a full constellation of 24 satellites, the European
tuntemattomien
ratkaisu Galileo system is approaching completion, and the Chinese are coming
3 with their Beidou-3 system. Joint use of the systems in the same receiver
is technically complicated but promises a very fast and reliable resolution
ohjelmisto- of ambiguities. Relief may come from “software-defined GNSS receivers”,
pohjainen in which all processing work after the antenna and analogue electronics
GNSS-
is implemented digitally in software on an off-the-shelf computer.
vastaanotin
^ 13.4 Real-time positioning

jälkilaskenta The static method described above is based on post-processing. For geode-
tic use, this is usually unproblematic. The use of precise satellite orbits —
essential if one wants geodetic precision for long vectors — also imposes a
certain waiting time: a couple of weeks in the case of precise ephemeris.
Sometimes, however, we need the co-ordinates of new points immedi-
ately, and there are also situations in which this would be useful or handy.
tosiaikainen Then, we speak of real-time4 positioning. Navigation is a broad field of
4
application.
In precise GPS positioning, real-timeness can be implemented by trans-
ferring the observations made at known point A on the fly, for example
by radio, to unknown point B for joint processing.
3 In the new GLONASS-K satellites, CDMA, code division multiple access, like in GPS, is
also being offered.
4 The formal definition of real-timeness is a guaranteed latency. It may be long, as
long as it is guaranteed. The latency or response time is the time that elapses from a
measurement event to the availability for use of the measurement values.
í Õ ! ¤. û
Real-time positioning 13.4
345
^ 13.4.1 Differential GPS (DGPS)
Differential GPS is a real-time positioning method based on code observa-
tions, which makes use of a reference or base station. It is thus a relative tukiasema
measurement between base station A and moving receiver, or rover, B.
As always with measurements from two receivers close to each other,
the error sources are the same or nearly the same in both receivers: the
orbit errors and clock offsets of the satellite, as well as the effect of the
atmosphere, cancel out in the difference measurement between A and B,
either completely or nearly so.
In the DGPS method, it is not the raw observations at point A that
are transferred to point B — that would be an unnecessary amount of
information to transfer. Instead, the difference between the measured and
the computed pseudo-range is first calculated from the observations. For
(0)
every observation p A , the geometric distance ρ A between observation
site A and satellite S can be computed. This “reference distance” is based
on the known location of point A and the orbital data transmitted by the
satellite. Then, one obtains the pseudo-range offset for each satellite in pseudoetäisyys-
the sky: poikkeama
def (0)
d pA = pA − ρA .
The orbital data from the same satellite are also available to the moving liikkuva
(0)
receiver B, which may itself compute from these the same ρ A — after all, vastaanotin
the location of point A is known. So, the information content of the offsets
d pA is the same as that of the full measurements pA , and the offsets may
replace them in the dissemination.
Using pseudo-range offsets has the following advantages:
◦ The numerical values are much smaller. The offsets were of the
order of ±100 m back when selective availability (SA, an artificial
reduction in accuracy of the disseminated orbital and clock informa-
tion) was still on. When SA was switched off in 2000, the magnitude
of the offsets dropped to the level of ±5 m. Both orders of magni-
tude are fractions of the size of the observables themselves, tens of
thousands of kilometres.
◦ The values change more slowly. They crawl over the course of hours
in a way which looks random. Extrapolation over several seconds or
minutes into the future works better than with raw observations.
For both reasons, the communications bandwidth needed is much less,
and the following channels are sufficient:
í Õ ! ¤. û
Without corr.
Computed place
d pB
d pA
With corr. Corrections True
B
place A
^ F IGURE 13.6. Principle of operation of the DGPS method, somewhat simplified.
◦ Mobile telephony. Modern network data connections (3G, 4G) are

fast indeed. The “mobile Internet”.
◦ A radio modem (short distances).
◦ For navigation at sea, long-wave radio.
The values of the offsets also vary slowly as functions of place. Therefore
we may, if the distance AB is suitably short, 100–1000 km, write with
sufficient accuracy
d p B ≈ d pA + c (∆TB − ∆TA ) ,
in which ∆T A and ∆TB are the receiver clock offsets.

Instead of the original, single observations, we construct difference
observations p ST
B between two satellites S and T :
}︄
d p SB ≈ d p SA + c (∆TB − ∆T A )
=⇒ d p ST T S T S
B = d pB − d pB ≈ d p A − d p A ,
d pT T
B ≈ d p A + c (∆T B − ∆T A )
a difference between disseminated quantities, from which the clock off-

sets of both receivers are eliminated. Now, we may calculate the single
í Õ ! ¤. û
347
difference in point B of the ranges to the two satellites:
T (0) S (0) ST (0)
ρB − ρB = ρB = p ST ST
B − d pB ,
a purely geometric quantity between the location of point B and the

locations in space of the satellites S and T computed from the orbital
data, all known to the user. From these, the unknown location of point B
may be solved for, without any influence from the satellites’ orbit errors or
clock offsets. Three difference observations, four satellites, are sufficient.
By differential GPS positioning and navigation (“DGPS”) is commonly
meant precisely this real-time method using the modulations of the GPS
signal, the C/A code and P code.
^ 13.4.2 Kinematic real-time positioning, “RTK”

If we use carrier-phase measurements in real time, we speak of RTK,
real-time kinematic, measurement. With this technique, precision is
much better, even if only over short distances. Unlike DGPS, this is an
incremental method: the measurement must start from a point with
known co-ordinates, and preferably also close on a known point — just in
case. The co-ordinates of the points measured in between are obtained in
relation to these known points.
See figure 13.7. When using double differences, between two satellites
and the base station and rover, the only unknowns are the rover’s three
co-ordinates X , Y , and Z , its location in three-dimensional space R3 .
All possible locations of the rover that are compatible with the double-
difference observation now form a bundle of hyperboloids of revolution. pyörähdys-
The distance separating the hyperboloids corresponds to one wavelength hyperboloidi
of the observable. In the figure, the hyperboloids are drawn in cross-
section as curves of different colours, corresponding to two different
satellite pairs. As always with GPS measurement, the minimum number
of satellites is four.
On known point A is installed a permanent (or semi-permanent) GPS
reference or base station. Let the momentaneous location of the rover be
B. The double differences are
ST
P AB = PBT − P A
T
− PBS + P A
S
.
Substituting equation 12.2 into this yields
ST
P AB = ρ ST ST ST ST
AB + D ion, AB + D trop, AB − λ · N AB .
í Õ ! ¤. û
rence equi-
e-diffe surfa
Doubl ces
k
a lin
Dat
Rover B
Base station A
Known Known
end point starting point
F IGURE 13.7. Principle of operation of the RTK method, simplified to two di-
mensions. The hyperbolas are the rover locations which are com-
patible with phase double differences corresponding to a round
number of whole wavelengths. The dashed hyperbola segments
again are locations compatible with the observed double differ-
^ ences. Compare this with the Decca figure 12.1!
Let us forget the atmospheric terms for a moment:
ST
P AB = ρ ST ST
AB − λ · N AB .
5 If the rover’s antenna is set up on a point with known co-ordinates5 B, all

co-ordinates are known:
◦ The co-ordinates of satellites S and T can be computed from the
orbital ephemeris — even from broadcast ephemeris, if the distance
AB is not very long.
◦ The co-ordinates of stations A and B are assumed known.
Therefore, the geometric double difference for that moment, ρ ST

AB ( t 0 ), is
6 computable. After that, the ambiguities6 may also be resolved from the
5 Or the co-ordinates of the rover’s starting point are determined from observations
before starting to move, the on-the-fly method. See the next subsection.
í Õ ! ¤. û
349
observations:
ST
ST P AB ( t0 ) − ρ ST
AB ( t 0 )
N AB =− .
λ
After this, we take the rover away from the known point and go and
measure. We move in the terrain and measure a suitable number of
unknown points — but in such a way that the connection to the satellites
ST
is not interrupted. Then namely the values of the ambiguities N AB do
not change either, and we may straightforwardly calculate the geometric
double difference
ρ ST ST ST
AB ( t P ) = P AB ( t P ) + λ · N AB
ST
from the measurements P AB ( t P ) at terrain point P . This is the essence
of kinematic GPS positioning. The correct place is obtained immediately,
even with millimetre-precision: of course, only in relation to the reference
station, not absolutely. Therefore, the precise geodetic determination of
the location of the reference or base station is essential.
RTK works best over short distances, from hundreds of metres to tens of
kilometres. The real-time nature requires use of a data communications tosiaikaisuus
link between base station and rover. The possible data communications
solutions are in principle the same as in the case of DGPS.
^ 13.4.3 RTK, technical considerations

Above it was assumed that the moving receiver, the rover, starts from a
point with co-ordinates known to geodetic precision. Such a point may
however also be created “on the fly”, by staying on the starting point for so lennossa
long that the satellite geometry changes and the resolution, the fixing to
integer values, of the ambiguities becomes successful. The link with the
base station has to be open, and the signal connection with those satellites
that are also visible from the base station must be uninterrupted.
Four satellites are generally sufficient for solving the location with GPS.
The RTK technique’s “on-the-fly” initialisation, however, requires at least
a fifth satellite, in order to resolve the ambiguities. Real-time quality
control of the measurements also requires this. Without redundancy,
6 These values must thus be integers or close to integers, in which case they may be
rounded, resulting in the “fix” solution. If they are not near integer values, they may
not be rounded. Then one obtains the, weaker, “float” solution. This may happen if, for
example, there is an error in the given co-ordinates of A or B. Or if the effect of the
atmosphere is too strong after all, or the distance AB too long.
í Õ ! ¤. û
ST
the ambiguities N AB would be freely chooseable, and one could compute,
from the three values ρ ST
AB thus obtained, a fantasy vector solution R AB ,
without any contradictions being generated!
The more satellites are available, the quicker the ambiguity resolution
will proceed. This is why instruments have appeared on the market that
are able to use simultaneously the signals from both the GPS and, for
example, the GLONASS satellites.
Over short distances — less than a kilometre — the number of satel-
lites effectively doubles, because the difference between the L 1 and L 2
leveäkuja frequencies, the “wide lane”, can be used, the effective wavelength of
which is 86 cm.
Today’s RTK instruments are able to intelligently use many known
points around a measurement area. Before and after the survey, these
points are visited, and the instrument forms, using the known and mea-
sured point locations, a local transformation formula. Using the formula,
all measured points are transformed to the same system in which the
known points have been given. This is a handy but also dangerous prop-
erty: the accuracy of the transformed co-ordinates cannot be better than
the interior accuracy of this local system. If it is, for example, the old KKJ
system based on traditional measurement techniques (subsection 3.2.1),
one loses the major advantage of GPS surveying, its superior geometric
accuracy!
^ 13.4.4 Network-mode real-time services

In recent years, many base-station networks have been built, in order to
offer both differential GPS (DGPS) and real-time kinematic (RTK) support
services. The main problem with using a single base station is that the
corrections disseminated by it are good close to the base station, but
deteriorate quickly as the distance of the rover from the base station
grows.
It is intuitively clear that the corrections change only slowly, and
almost linearly, with place. The correction for the satellite clock offset
is even constant. The satellites are so high that even over an area the
size of Europe the geometry is almost the same in different parts of the
continent. Moreover, the atmosphere is usually rather similar everywhere
within a small area. This opens up the possibility of interpolation, and
tukiasema- the base-station networks do precisely this.
verkko
Realistic alternatives for disseminating correction signals:
í Õ ! ¤. û
SBAS systems 13.5
351
◦ Geostationary communications satellites. Their advantage is the
homogeneous coverage of large areas, their disadvantage, the satel-
lites’ low elevation angle at Finnish latitudes.
◦ Using the mobile Internet through the mobile telephony network.
The Internet is not real-time, but in practice fast enough for it to
often not matter. Advantages are
– an easy way to charge for the service
– the possibility to supply, to the location of the receiver, tailored
corrections — the “virtual base station” idea virtuaali-
tukiasema
– a large data transfer capacity at little cost.
Using the services in network mode also requires software support. The
geometric aspect of interpolating the corrections is easy; the problem is
formed by the modelling of the propagation delay by the atmosphere. In kulkuviive
order to achieve good accuracy, the base-station network used for the
model computation must be sufficiently dense. This complex problem
field is the subject of active research, for example Koivula et al. (2018).
^ 13.5 SBAS, satellite based augmentation systems

The term wide-area differential GPS (WADGPS) is also used for this type
of system. These systems use geostationary satellites to disseminate dif-
ferential corrections for GPS positioning. Globally, three intercompatible
systems exist:
◦ WAAS (Wide Area Augmentation System, USA)
◦ EGNOS (European Geostationary Navigation Overlay System)
◦ MSAS (Multi-functional Satellite Augmentation System, Japan).
These are already in widespread use, although their development contin-

ues. The services are continent-wide and are based on the simultaneous
use of many GPS base stations in network mode. In this way, accurate
differential corrections are obtained for the area covered by the base
stations. The signal structure and frequencies used are the same as for
the GPS satellites, which makes it relatively easy to modify an existing
GPS receiver design for SBAS use.
A core application is GPS integrity monitoring, which sounds an alarm eheyden

if the quality of positioning cannot be guaranteed. This is important valvonta
in safety critical applications — SoL, “Safety of Life” — like positioning
aircraft during approach and landing.
í Õ ! ¤. û
MSAS
36 000 km
WAAS EGNOS
^ F IGURE 13.8. Satellite-based augmentation systems (SBAS).
^ 13.6 Real-time satellite positioning support services in

Finland
^ 13.6.1 The DGNSS service of the Finnish Transport Agency

This service, Väylä, Radionavigaatiopalvelut, operated by the Finnish
Transport Infrastructure Agency (earlier the Finnish Transport Agency,
still earlier the Finnish Maritime Administration) sends correction mes-
sages by radio on long wavelengths, frequencies 287.5–314.5 kHz. The
service covers the sea areas of the Baltic Sea, and in Finland also the lake
area of Saimaa. The user community consists of seafarers. The service is
free of charge.
^ 13.6.2 Trimnet VRS

Geotrim, Trimnet is the real-time kinematic (RTK) network service main-
tained by Geotrim Oy, originally planned for use by the Finnish National
Land Survey. The technology used is called network RTK, more precisely,
virtual reference station RTK (VRS-RTK), and is based on generating, for
every user, a computational “virtual” base station close to them. Correc-
tion data for this base station are generated in the RTCM-SC104 format, a
standard format (“RTCM format”) useable by devices of any brand.
The network covers all of Finland with over a hundred base stations
(as of 2019) and also supports the use of GLONASS. The corrections are
disseminated commercially over the mobile Internet: a cluster of servers
is located at Geotrim headquarters in Vantaa, to which the user can log
in over the Internet. The corrections obtained are tailored for each user
í Õ ! ¤. û
TheGDGPS system 13.7
353
separately. The precision obtainable is of the order of a centimetre in the
horizontal plane, a little poorer vertically.
^ 13.6.3 HxGN SmartNet

HxGN SmartNet (Hexagon Geosystems, HxGN SmartNet), operated and
maintained by Hexagon Oy, is a real-time kinematic (RTK) base-station
network offering corrections as a commercial service over the mobile
Internet. In Finland there are currently (2017) over a hundred base
stations.
^ 13.6.4 The experimental service of the National Land Survey

This service of the Finnish National Land Survey’s Geospatial Research
Institute, Finnish National Land Survey, About positioning services,
which is at the moment experimental and free of charge but which re-
quires registration, offers both differential GNSS and network RTK. The
base stations are the twenty new FinnRef stations.
^ 13.6.5 Archive data service

These network RTK service providers, in Finland and abroad, usually
also supply under agreement archived data in the RINEX7 format for 7
post-processing. Typically, the data is thinned out, from a data rate of
one Hz — one measurement event per second from all visible satellites —
1
to, say, 30 Hz, one measurement event every 30 seconds.
^ 13.7 The GDGPS system

JPL, The Global Differential GPS System. This is a global real-time DGPS
service using the Internet as its distribution channel. The system was
developed and is being operated by NASA’s Jet Propulsion Laboratory
(JPL). As part of the service, it offers an online precise point positioning
(PPP) service using RINEX files uploaded by the user.
7 RINEX, Receiver-INdependent EXchange format, is a practical, device independent

text format that also a human being can read. Gurtner and Estey (2007). There is a
conversion program for every receiver brand. See tableau 12.3.
í Õ ! ¤. û
1. What is the “common-mode” error assumption?
2. What kinds of difference observations are there, and what error
sources do they eliminate or reduce?
3. What is differential GPS (DGPS)?
4. What is real-time kinematic (RTK) positioning?
5. For what purpose have SBAS (satellite-based augmentation systems)
been built?
6. What is RINEX?
^ Exercise 13–1: Geodetic GPS positioning

The task is to determine a precise vector between two stations by pro-
cessing “in the cloud”. GNSS data from many hundreds of continuously
operating stations are available online, as are services for using these
data for geodetic position computation. We are going to exploit this for a
test computation.
1. Get the RINEX data files from the Sodankylä (SODA) and Met-
sähovi (METS) stations; see figure 13.1. As sources you may use
the web-site of the FGI (Finnish National Land Survey, RINEX ser-
vice), which however requires registration, or the SOPAC web-site
(SOPAC, GPS Explorer) of the Scripps Institution of Oceanography
in San Diego, California, or the EUREF data centre (ROB, EUREF
Permanent GNSS Network) at the Royal Observatory in Brussels,
8 Belgium. Download data for a single day8 — the same for both
stations — of which the size should not exceed 5 MB.
2. The data downloaded may be Hatanaka compressed (SOPAC,
Hatanaka file compression). This is a compression technique
specifically for RINEX data, exploiting the similarity of the data
for successive epochs. You need to get the conversion program
CRX2RNX as a binary for your operating system, and convert the
data into uncompressed RINEX. The latter is human-readable and
even clear. The Hatanaka compressed file is also human-readable,
and the compression technique used is clearly visible.
Remember that the data may also be compressed by a standard
8 Suggestion: your last year’s birthday!
í Õ ! ¤. û
Exercise 13–1: Geodetic GPS positioning
355
method, like zip or gzip or Unix compress (*.Z) . . . which you first
have to expand. On Windows, the utility 7-zip may be useful.
Also, on Windows, you may run into the line-ending problem: Unix
text files end their lines with a line-feed (LF) only, Windows uses a
carriage-return (CR) followed by a line-feed (LF). The text editor
Notepad++ may be useful.
3. Now, you can upload your RINEX files to the cloud. There are two
alternatives — choose one for this exercise:
◦ AUSPOS – Online GPS Processing Service, an Australian
government service. It tends to be a little slow: often the
result comes overnight, in the form of an extensive report.
Publicly available data from nearby IGS stations is used in the
positioning computation, as shown in a map.
◦ GDGPS APPS, a US government service run by the Jet Propul-
sion Laboratory JPL — the headquarters of the International
GNSS Service IGS (subsection 12.11) — and yes, they too use
data from the IGS network (figure 12.24) for fixing the refer-
ence frame in the computations.
You need to upload METS and SODA separately, one file at a
time.
4. The AUSPOS results come by email.
(a) Read the results carefully. What other stations were included
in the computation, and in what reference frame is the result
expressed?
(b) The geocentric Cartesian (rectangular) co-ordinate solution X ,
Y , Z.
(c) The geodetic co-ordinates, and their precisions (“Positional
Uncertainty”).
(d) Other interesting stuff. How is the tropospheric propagation
delay modelled? Yes, they estimate dry and wet tropospheric
zenith propagation delays as well as horizontal gradients!
(e) Note the use of a geoid model, for obtaining heights over mean
sea level. How good do you think it is?
(f) Ambiguity resolution.
5. The APPS results appear online.
For APPS, go to the summary file (*.sum) and look up the following
í Õ ! ¤. û
things:
(a) The geocentric Cartesian co-ordinate solution X , Y , Z , and the
co-ordinate uncertainties (“sigmas”). How does this precision
concept differ from that of AUSPOS?
(b) The geodetic co-ordinates Lat, East_Lon and Height, and their
sigmas. Compare the height sigma with the others.
(c) The other interesting headers. How is the troposphere mod-
elled here? Compare with AUSPOS.
í Õ ! ¤. û
^ Adjustment calculus in geodesy
14
[. . . ] Mais comment vous décrire mon admiration et mon
étonnement, en voïant se metamorphoser mon
correspondant estimé M. Leblanc en cette illustre
personnage, qui donne un exemple aussi brillant de ce que
j’aurois peine de croire. Le goût pour les sciences abstraites
en général et surtoût pour les mysteres des nombres est fort
rare : on ne s’en étonne pas ; les charmes enchanteurs de
cette sublime science ne se decelent dans toute leur beauté
qu’à ceux qui ont le courage de l’approfondir. Mais
lorsqu’une personne de ce sexe, qui, par nos mœurs et par
nos préjugés, doit rencontrer infiniment plus d’obstacles et
de difficultés, que les hommes, à se familiariser avec ses
recherches epineuses, sait neansmoins franchir ces entraves
et penétrer ce qu’elles ont de plus caché, il faut sans doute,
qu’elle ait le plus noble courage, des talens tout à fait
extraordinaires, le génie supérieur. [. . . ]
Letter from Gauss to Sophie Germain, 1807 (Friedelmeyer,

2014). See also Wikipedia, Sophie Germain.
^ 14.1 Why adjustment?

In geodesy, as in science in general, we know that all measurements
are wrong. Therefore, we collect always more measurements than the
strict minimum, so as to be able to judge at least somewhat realistically
the uncertainties in the measurement results. This practice is called
redundancy.
A good example is measurements in a geodetic network. In a triangu- kolmiomittaus
lation network, figure 14.1, directions from the triangle points to other
– 357 –
358 14 A DJUSTMENT CALCULUS IN GEODESY
α1
α2
2 α3
^ F IGURE 14.1. Triangulation network, triangle.
triangle points are measured.

From the direction measurements to neighbouring points made at
points 1, 2, and 3, angles α1 , α2 , and α3 are calculated. For the angles,
1 the triangle condition1 applies:
α1 + α2 + α3 = 180◦ .
The triangle condition makes the following checks possible:

1. The measurements from which the angles α1 , α2 , and α3 were cal-
karkea virhe culated do not contain gross errors. For example, if we obtain for
the sum of angles α1 + α2 + α3 = 173◦. 6742, we may immediately
conclude that there must be at least one gross error in the mea-
surement set, because the accuracies of the instruments used are
fractions of a degree.
2. The amount by which the sum obtained differs from the theoretical
value of 180◦ allows us to infer the precision of the measurement.
For example, if we obtain as the sum α1 + α2 + α3 = 179◦. 9958, then
def
the closing error ∆ = α1 +α2 +α3 −180◦ = −0◦. 0042, and the inference
is that the precision of the measurement method used is several
thousandths of a degree.
If we do more than the minimum number of measurements in this way,
we need a method for removing the, small, contradictions between these
measurements, reconciling them with each other. A brute-force trick
1 On the curved surface of the Earth, the sum is not exactly 180◦ but a little larger, the
spherical excess. This is an example of non-Euclidean geometry.
í Õ ! ¤. û
The average 14.2
359
would be to just throw measured value α3 away, and compute a replace-
ment value, guaranteed to be compatible, of α3 = 180◦ − α1 − α2 . However,
one may justifiably ask, why α3 rather than α1 or α2 ? Such arbitrari-
ness is unacceptable, and we should not just throw away the valuable
information contained in the observation. The right solution is network verkkotasoitus
adjustment.
In the simple triangle case we divide the closing error equally among
the angles: the adjusted angle values will be democratically
ˆ︁1 = α1 − 13 ∆,
α ˆ︁2 = α2 − 31 ∆,
α ˆ︁3 = α3 − 31 ∆,
α
after which α
ˆ︁1 + α
ˆ︁2 + α
ˆ︁3 = 180◦ exactly.
If we however know that, say, the angle α3 was measured twice (and
the value α3 is the average of these measurements) but the angles α1 and
α2 only once using the same instrument, we may take this into account
by adjusting in the following way:
ˆ︁1 = α1 − 25 ∆,
α ˆ︁2 = α2 − 52 ∆,
α ˆ︁3 = α3 − 51 ∆,
α
in which still ∆ = α1 + α2 + α3 − 180◦ . In this case one speaks of weighting

the measurements. The measurement α3 is given a double weight — and
thus a halved correction — compared to the measurements α1 and α2 .
A large weight means a small correction, and vice versa, figure 14.2.
The weight ratios are 1 : 1 : 2, the correction ratios are the reverse 2 : 2 : 1,
and the sum of the correction ratios is 5. This is how the above correction
coefficients (“weight coefficients”) 25 , 25 , and 1
5 are obtained.
The adjustment of a realistic, complex triangulation network (or any kolmiomittaus-
geodetic network) is mathematically much more complicated, but this is verkko
the basic idea.
It is commonly assumed that the random measurement errors of the
observations are distributed according to the bell curve named after C. F.
Gauss (figure 2.5), in other words, that they are normally distributed. At
least in that case, the theoretically best solution is given by least-squares
adjustment. The simplest example of this is computing the average. pienimmän
neliösumman
tasoitus
^ 14.2 The average
Assume that the same quantity has been observed n times — so, we have
a stochastic quantity ℓ — the observation values being ℓ i , i = 1, . . . , n. The
observations have a statistical expectancy µ, and they all have the same odotusarvo
í Õ ! ¤. û
Small correction
Large correction
Large weight
Small weight
F IGURE 14.2. Metaphor: a large weight means a small correction and vice
^ versa.
3 standard deviation or mean error2 σ. The average of the observations is3

2
n
1 1 ∑︂
ℓ = n (ℓ1 + ℓ2 + · · · + ℓn ) = n ℓi .
i =1
One can show that, as an estimator of the expectancy µ, this is the “best
possible” linear combination of observations. One can also show that this
jäännösvirhe linear combination minimises the sum of squares of the residuals4
4
def
vi = ℓ − ℓi , i = 1, . . . , n,
n
∑︂
v2i = v21 + v22 + · · · + v2n = min.
i =1
pienimmän This property is the origin of the term “least-squares method”.
neliösumman
We may also estimate the standard deviation or mean error of a single
menetelmä
otoskeskivirhe
observation, σ, by the equation for the sample mean error
⌜
⃓ n
⃓ 1 ∑︂
σ
ˆ︁ = ⎷ v2i .
n−1
i =1
2 Expressed
{︁ }︁
more theoretically, if the expectancy operator is E · , we may write µ =
2
E {ℓ} and σ2 = E (ℓ − µ) .
{︁ }︁
3 Here, the values ℓ i are written as stochastic, because the formation of the average may
be repeated, to form different realisations of the stochastic quantity ℓ.
4 The residual ℓ − ℓ i of an observation is not the same as (the opposite of) the error ℓ i − µ
of that observation! The residual is computable from the observations, the error is not.
í Õ ! ¤. û
Linear regression 14.3
361
y
Linear regression by least squares:

Σ is minimised! x
^ F IGURE 14.3. The idea of linear regression.
From this again, the quality, or uncertainty, measure for the average ℓ, its
ˆ︁ ⎷n . This value,
ˆ︁ n = σ
/︁
mean error (standard deviation) estimate follows, σ
which describes the uncertainty of the average, thus becomes smaller the
longer the series of measurement values is, in other words, the larger n.
If we know the mean error σ of a single observation a priori, we may
/︁⎷
also directly use the equation σn = σ n , the result of which is not an
estimate but a computed value.
^ 14.3 Linear regression

In linear regression, two parameters a and b are estimated when obser-
vations y that depend linearly on the argument x are given:
yi + vi = a
ˆ︁ + bx
ˆ︁ i .
Here, v i is again the residual of observation i . Parameters a and b

describe a straight line that runs as well as possible — meaning, with as
(︁ )︁
small residuals as possible — through the measured “point cloud” x i , y i .
See figures 14.3 and 14.4.
Linear regression is a least-squares method, figure 14.3. The sum of
the squared residuals is minimised.
The least-squares solution is
∑︁ (︁ )︁ ∑︁ ∑︁ ∑︁ ∑︁
n xy − x y y−b
ˆ︁ x
b
ˆ︁ = , a
ˆ︁ = n ,
( x 2 ) − ( x )2
∑︁ ∑︁
n
in which the compact notation is used
∑︂ n
∑︂
def
(·) = (· ) ,
i =1
í Õ ! ¤. û
y
Observation i
yi
Residual v i
arctan b
a
x
xi
^ F IGURE 14.4. Linear regression, definitions of quantities.
(︁ )︁
a summation over all n points, or co-ordinate pairs, x i , y i . We can write
the solution even more neatly in terms of averages:
⟨︁ ⟩︁ ⟨︁ ⟩︁
x y − ⟨ x⟩ y ⟨︁ ⟩︁
b
ˆ︁ = , a ˆ︁ ⟨ x⟩ ,
ˆ︁ = y − b
2
⟨ x2 ⟩ − ⟨ x⟩
if the average is written as
n
1 ∑︂
def
⟨·⟩ = n (·)
i =1
The “hat notation” (a,

ˆ︁ b
ˆ︁) is an often-used way to designate estimators.
^ 14.4 Theory of least-squares adjustment
^ 14.4.1 Calculating the solution from the observations

pienimmän Presumably the first to use the method of least-squares was Carl
neliösumman Friedrich Gauss, although Adrien-Marie Legendre5 has also been claimed
menetelmä
5
to be the inventor of the method. Gauss also carried out extensive
6
6 geodetic network computations in Hannover using his method.
In astronomy, the first application of the method was to compute the
orbits of asteroids and comets from observations. This, and the adjust-
ment of geodetic networks, were special cases of situations that occur all
the time in the life of an observer:
5 Adrien-Marie Legendre (1752–1833) was a French mathematician, one of the 72 names

on the Eiffel Tower (Eiffel Tower, 72 names).
6 The numerical work of the network adjustment was carried out by an army of manual,
human computers under Gauss’ command. Back then, a “computer” was a human being!
í Õ ! ¤. û
Theory of least-squares adjustment 14.4
363
◦ We have available a body of observations, and we wish to compute
from it certain interesting unknowns, in a way which
– Treats all observations as equally valuable.
– Makes the deviations of the computed values for the unknowns
of interest from their “true values” as small as possible.
◦ In addition, it would still be desirable that
– Any gross errors still hiding out in the observations are found
and removed.
The parametric form of the least-squares adjustment method, which is
based on the formation of observation equations, is used for this.
^ 14.4.2 The observation equations

Forming the observation equations is done as follows. We write all
observations as linear7 functions of all unknowns: 7
ℓ1 + v1 = a 11 ˆ︁
x1 + a 12 ˆ︁
x2 + · · · + a 1m ˆ︁
xm ,
ℓ2 + v2 = a 21 ˆ︁
x1 + a 22 ˆ︁
x2 + · · · + a 2m ˆ︁
xm ,
.. (14.1)
.
ℓn + v n = a n1 ˆ︁
x1 + a n2 ˆ︁
x2 + · · · + a nm ˆ︁
xm ,
if there are n observations ℓ i , n residuals v i , and m unknowns ˆ︁

x j.
The system of equations can be conveniently written in the form of a
matric equation
ℓ + v = Aˆ︁
x, (14.2)
in which
⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤
ℓ1 v1 a 11 a 12 ··· a 1m x1
ˆ︁
ℓ2 v2 ⎢ a a 22 · · · a 2m x2 ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎥ , A = ⎢ .21
⎢ ⎥ ⎢ ⎥ ⎥ ⎢ ˆ︁
ℓ=⎢ ..
⎥, v = ⎢ .. ⎢ . .. .. .. ⎥ x = ⎢ .. ⎥ .
⎥ , ˆ︁ ⎢ ⎥
⎢
⎣ .
⎥
⎦
⎢
⎣ .
⎥
⎦ ⎣ . . . . ⎦ ⎣ . ⎦
ℓn vn a n1 a n2 · · · a nm xm
ˆ︁
The matrix A is rectangular: n > m, it is taller than it is wide. There
are more observations, that is equations, than there are unknowns: re-
dundancy. The observations ℓ, the unknowns ˆ︁
x and the residuals v are jäännös-
abstract vectors, elements of an abstract vector space: virheiden
vektori
ℓ, v ∈ R n , x ∈ Rm .
ˆ︁
7 Often, the observation equations of real life are not linear. Then, linearisation is
usually possible. See section 14.6.
í Õ ! ¤. û
Often, one may assume that all observations ℓ i , i = 1, . . . , n have the

same mean error σ (and a similarly shaped statistical distribution), and
that the observations are statistically independent of each other, which
also means that they do not intercorrelate. This assumption is referred
to as i.i.d. — “independent, identically distributed”.
^ 14.4.3 The normal equations

pienimmän From the matric equation 14.2 we compute the least-squares solution by
neliösumman first multiplying from the left with the matrix A T , the transpose of A :
ratkaisu
A T Aˆ︁
x = A T ℓ + A T v.
8 Set8
A T v = 0,
x:
yielding for the least-squares solution ˆ︁
A T A ˆ︁
x = A T ℓ.
(︁ )︁
(14.3)
x: the coeffi-
This is a system of m equations and m unknowns in vector ˆ︁
cient matrix A T A is square.
^ 14.4.4 Solving the normal equations

The solution, or estimator, is obtained in for example the following way:
)︁−1 T
x = AT A
(︁
ˆ︁ A ℓ, (14.4)
assuming that the matrix A T A can actually be inverted, that it is not

singular.
The equations 14.3 are known as the normal equations.
In the case that all observations have the same mean error σ (and thus
the same variance σ2 ), and that they do not intercorrelate, the solution
14.4 is optimal in the least-squares sense.
Of course, solving the system of equations by traditional means, without
x, being the unknowns,
matrices, to find the elements of the solution vector ˆ︁
8 If def
one writes η = A ξ, the following holds:
η · v = ( A ξ) · v = ξ T A T v = 0
⟨︁ ⟩︁ ⟨︁ ⟩︁
for an arbitrary vector ξ; we say that the sub-space of vectors A ξ of the space of observ-
ables (the “solution space”, spanned by the columns of the matrix A ) is perpendicular
upon the sub-space of residuals. This is where the term “normal equations” comes from.
í Õ ! ¤. û
Theory of least-squares adjustment 14.4
365
is also a readily useable method: write the normal equations 14.3 in the
following way, which is well-suited to computer coding:
m
(︄ n )︄ (︄ n
)︄
∑︂ ∑︂ ∑︂
a jk a ji xi =
ˆ︁ a jk ℓ j , k = 1, . . . , m.
i =1 j =1 j =1
This is a system of m linear equations in m unknowns ˆ︁

x i , for the solution
of which numerical standard methods and software libraries are on offer.
The greatest challenge is usually finding suitable observation equations
14.1 in a concrete measurement situation.
The solutions for the average and linear regression presented above,
sections 14.2 and 14.3, are special cases of the general adjustment solu- tasoitus
tion, as we shall show.
^ 14.4.5 Assessing the precision

When one computes the least-squares solution with the equation
)︁−1 T
x = AT A
(︁
ˆ︁ A ℓ,
one can also, with the propagation law of variances, obtain the precision varianssien
x.
of estimator ˆ︁ kasautumislaki
We reiterate the assumption that the variance matrix of the observa-

tions ℓ is Var ℓ = σ2 I , so the observations do not correlate with each
{︁ }︁
other and are all equally precise.

If we define the linear operator
def )︁−1
L = AT A AT,
(︁
we obtain, based on the linear dependence,
x} = Σxx = L Var{ℓ} LT =
Var{ˆ︁
)︁−1 )︁−1 )︁−1
= AT A A T · σ2 I · A A T A = σ2 A T A
(︁ (︁ (︁
. (14.5)
−1
This interesting result tells us, that the matric quantity ( A T A ) repre-
sents the propagation of the mean error σ of the observations into the
x.
variances of the end result of the adjustment ˆ︁
def
The matrix N = Pxx = A T A is called the weight matrix of the unknowns
def −1
or normal matrix, and its inverse, the matrix Q xx = ( A T A ) , is called
the weight-coefficient matrix of the unknowns (Baarda, 1981).
í Õ ! ¤. û
^ 14.5 Examples of the least-squares method

Both average and linear regression are great examples of least-squares
adjustment methods: practical and still relatively simple. In the following,
we go through them step by step, showing how they are special cases of
the general least-squares method.
^ 14.5.1 The average as least-squares adjustment

Observe the same quantity x directly n times:
x1 + v1 = ˆ︁
x
x2 + v2 = ˆ︁
x
.. (14.6)
.
x n + v n = ˆ︁
x
x is the estimator
Here, x1 , x2 , . . . , x n are individual observation values, ˆ︁
of the unknown quantity x, and v1 , v2 , . . . , v n are the residuals of the
observations.
The secret to formulating a suitable adjustment procedure is: find the
standard form of the system of observation equations,
ℓ + v = Aˆ︁
x.
In the present case, success means choosing

⎡ ⎤ ⎡ ⎤ ⎡ ⎤⎫n times
x1 v1 1 ⎪
⎪
⎥⎪
x2 v2 1
⎢ ⎥ ⎢ ⎥ ⎢ ⎪
⎢ ⎥ ⎢ ⎥ ⎢ ⎥⎬ [︂ ]︂
ℓ=⎢ ..
⎥, v=⎢ ..
⎥, A=⎢ ..
⎥ , x=
ˆ︁ x .
ˆ︁
. . .
⎢ ⎥ ⎢ ⎥ ⎢ ⎥⎪
⎣ ⎦ ⎣ ⎦ ⎣ ⎦⎪
⎪
⎪
xn vn 1
⎭
Verify that this really agrees with equations 14.6.

The normal equations are now
N b
⏟⏞⏞⏟ ⏟⏞⏞⏟
A T A ˆ︁
x = ATℓ,
in which the normal matrix N is

⎡ ⎤⎫n times
n times 1 ⎪
⎪
⎥⎪
1
⏟[︂ ⏞⏞]︂⏟ ⎢ ⎪
T
⎢ ⎥⎬ [︂ ]︂
N=A A= 1 1 ··· 1 ⎢ ..
⎥ = n
.
⎢ ⎥⎪
⎣ ⎦⎪
⎪
⎪
1
⎭
í Õ ! ¤. û
Examples of the least-squares method 14.5
367
and the right-hand side vector
⎡ ⎤
n times x1
⎥ [︄ n ]︄
x2
⏟[︂ ⏞⏞
]︂⏟ ⎢
⎢ ⎥ ∑︂
b = ATℓ = 1 1 ··· 1 ⎢ ..
⎥= xi .
.
⎢ ⎥
⎣ ⎦ i =1
xn
So, the solution is obtained as follows:
b b
N ⏟[︄ ⏞⏞ ]︄⏟ N −1
⏟ ⏞⏞ ⏟
⏟[︂ ⏞⏞ ]︂⏟ n ⏟⏞⏞⏟n
∑︂ 1 ∑︂
n x=
ˆ︁ xi =⇒ ˆ︁
x= n xi ,
i =1 i =1
the classical equation for the average!
Deriving the precision, or mean error, of the solution is not hard either.
The assumption that the mean errors of the observations x i , i = 1, . . . ,
n, are all the same σ and that the observations are uncorrelated — a
precondition for using the equation for the average — means that the
variance matrix of the observation vector ℓ is
⎡ ⎤
1
1
Var ℓ = σ2 I = σ2 ⎢
{︁ }︁ ⎢ ⎥
.. ⎥,
⎣ . ⎦
1
a matrix of size n × n. Then, according to equation 14.5, the variance

matrix of the vector of unknowns9 is 9
)︁−1 1
x} = σ2 A T A = n σ2 ,
(︁
Var{ˆ︁
and the mean error of the unknown is the square root of this:
σ
σx = ⎷ .
n
^ 14.5.2 Linear regression as least-squares adjustment

Write the observation equations of linear regression into the form
yi + vi = a
ˆ︁ + bx
ˆ︁ i ,
(︁ )︁
in which every pair x i , y i , i = 1, . . . , n is one observation, and the
coefficients a
ˆ︁ and b
ˆ︁ of the straight line to be fitted are to be determined.
If we write the vector of observations, the vector of residuals, the vector
of unknowns, and again (essential!) the design matrix as rakennematriisi
9 The x has here only one element, as the “matrix” Var{ˆ︁

“vector” ˆ︁ x} has, too.
í Õ ! ¤. û
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
y1 v1 1 x1
[︄ ]︄
y2 v2 1 x2 ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥ a ⎢
ℓ=⎢
ˆ︁
..
⎥, v=⎢ ..
⎥, x=
ˆ︁ , A=⎢ .. .. ⎥ ,
⎥
⎢
⎣ .
⎥
⎦
⎢
⎣ .
⎥
⎦ b
ˆ︁ ⎢
⎣ . . ⎦
yn vn 1 xn
one may write this system into the standard form of observation equa-
10 tions10
ℓ + v = Aˆ︁
x.
In the normal equations
N b
⏟⏞⏞⏟ ⏟⏞⏞⏟
A T A ˆ︁
x = ATℓ
the normal matrix is
⎡ ⎤
1 x1 ⎡
∑︁n
⎤
[︄ ]︄ ⎢
1 x2
⎥ n xi
1 1 ··· 1 ⎢ ⎥ ⎢ i =1
⎥
N = AT A = ⎢ .. ..
⎥=⎢ n n
⎥
x1 x2 · · · xn ⎢ ⎣ . .
⎥ ⎣ ∑︁
⎦ xi
∑︁
x2i
⎦
1 xn i =1 i =1
and the right-hand side vector

⎡ ⎤
y1 ⎡ n
∑︁
⎤
⎥ ⎢ i=1 y i ⎥
[︄ ]︄ ⎢
y2
⎥
T 1 1 ··· 1 ⎢
b= A ℓ= ⎢ ..
⎥=⎢ n ⎥.
x1 x2 · · · xn ⎢ ⎣ .
⎥ ⎣ ∑︁
⎦ xi yi
⎦
yn i =1
From this, the solution is obtained by inversion of the above 2 × 2 ma-

trix. However, this may also be done simply by elimination and back-
substitution. The system of equations is
n n
(︄ )︄
∑︂ ∑︂
n·a
ˆ︁ + xi ·b
ˆ︁ = yi ,
(︄ n
)︄ (︄ in=1 )︄ i =1
n
∑︂ ∑︂ ∑︂
xi ·a
ˆ︁ + x2i ·b
ˆ︁ = xi yi .
i =1 i =1 i =1
Subtract the first equation from the second after multiplication by the
1 ∑︁ n
factor n i =1 x i , yielding
n n n n n
(︄ (︃∑︂ )︃2 )︄
∑︂ 1 ∑︂ 1 ∑︂ ∑︂
x2i − n xi ·b
ˆ︁ = xi yi − n xi yi ,
i =1 i =1 i =1 i =1 i =1
10 The use of the letter x may be confusing here, and has nothing to do with the x i !
í Õ ! ¤. û
Examples of the least-squares method 14.5
369
^ T ABLE 14.1. Measurement results for linear regression.
∑︁5
i→1 2 3 4 5 i =1
xi 1.51 2.44 3.34 4.41 5.05 16.75

yi 2.32 3.12 3.57 3.93 4.15 17.09
x2i 2.28 5.95 11.16 19.45 25.50 64.34
xi yi 3.50 7.61 11.92 17.33 20.96 61.32
from which
n n n
(︄ )︄/︄(︄
∑︂ ∑︂ ∑︂ n
∑︂ n
(︃∑︂ )︃2 )︄
b
ˆ︁ = n xi yi − xi yi n x2i − xi .
i =1 i =1 i =1 i =1 i =1
Now a
ˆ︁ is obtained by back-substitution:
n n
(︄ (︃∑︂ )︃ )︄
1 ∑︂
a
ˆ︁ = n yi − xi ·b
ˆ︁ .
i =1 i =1
The expressions found are equivalent to those given in section 14.3.
^ 14.5.3 Calculation example of linear regression

The measurement results are given in table 14.1. A graphical presenta-
tion of the measurements is given in figure 14.5.
The needed sums are precomputed in the table:
n
∑︂ n
∑︂ n
∑︂ n
∑︂
xi , yi , x2i , xi yi .
i =1 i =1 i =1 i =1
y
y = 1.76 + 0.495 x
4
Mean error of observations

1
x
1 2 3 4 5
^ F IGURE 14.5. Calculation example of linear regression.
í Õ ! ¤. û
From this
5 · 61.32 − 16.75 · 17.09 20.3425

b
ˆ︁ = = = 0.495,
2 41.1375
(︂5 · 64.34 − 16.75 )︂
1
a
ˆ︁ = 17.09 − 16.75 · b
ˆ︁ = 1.76.
5
This solution has been plotted into figure 14.5.

The normal matrix, or weight matrix of the unknowns, is obtained as
follows:
n
⎡ ⎤
∑︁
⎢ n xi ⎥ [︄ ]︄
i =1 5.00 16.75
N = Pxx = A T A = ⎢ n n
⎥=
⎣ ∑︁
xi
∑︁
x2i
⎦ 16.75 64.34
i =1 i =1
and its inverse is

[︄ ]︄
)︁−1 1.5640 −0.4072
Q xx = A T A
(︁
= ,
−0.4072 0.1215
the weight-coefficient matrix. The variance matrix

[︄ ]︄ [︄ ]︄
σa2 σab 2 2 1.5640 −0.4072
Σxx = = σ Q xx = σ
σab σb2 −0.4072 0.1215
is obtained from this, from which are obtained the mean errors σa and σb
as the square roots of the diagonal elements:
ˆ︁ = 1.76 ± 1.25 σ,
a
ˆ︁ = 0.495 ± 0.349 σ.
b
Here, σ is the a priori (given in advance) mean error of a single y value,

painoyksikön the mean error of unit weight.
keskivirhe
^ 14.6 Linearisation of geodetic models

In geodesy, as more generally in science, relationships are found between
two quantities that behave non-linearly. Examples of this are the rela-
tionship between observables and unknowns, or between co-ordinates in
two different co-ordinate frames.
Many theories, however, like the least-squares adjustment method,
are based on linear equations, the mathematics of which is essentially
virheiden simpler. The law of propagation of errors (variances) also applies only to
kasautumislaki linear relationships between quantities.
í Õ ! ¤. û
Linearisation of geodetic models 14.6
371
In practice, a formally non-linear relationship, for example between
point co-ordinates and the measured direction to a point, is often al-
most linear within the uncertainty area of the point location. Geodetic
measurements are exceptionally precise: the location uncertainty of a
point may be mere centimetres when the distance between points can
be hundreds of metres or kilometres. In that case, instead of looking at
the original quantities, one can look at the relationship between small
variations or differences in these quantities — which will be almost linear.
This will be demonstrated using a Taylor series expansion.
^ 14.6.1 The scalar case

Normally if we have two quantities between which exists a functional
relationship
y = f ( x),
we may linearise it by choosing an approximate value x0 and expanding likiarvo
the function into a Taylor series in the neighbourhood of the approximate
value. We obtain
df ⃓
⃓
y = f ( x0 ) + ( x − x0 ) + · · ·
dx x= x0
⃓
or
y − y0 ≈ a ( x − x0 ) , (14.7)
in which
df ⃓
⃓
def def
y0 = f ( x0 ), a= .
dx x= x0
⃓
This may be written as
∆ y = a ∆ x,
which is often abbreviated to
y = ax,
as long as it is remembered that now x and y are “linearised quantities”,

def def
difference quantities ∆ x = x − x0 and ∆ y = y − y0 reckoned from the
approximate values x0 and y0 .
^ 14.6.2 The vector case

If there are two vectorial quantities,
⎡ ⎤ ⎡ ⎤
x1 y1
x2 y2
⎢ ⎥ ⎢ ⎥
⎥ ∈ Rn , ⎥ ∈ Rm ,
⎢ ⎥ ⎢ ⎥
x=⎢ .. y=⎢ ..
. .
⎢ ⎥ ⎢ ⎥
⎣ ⎦ ⎣ ⎦
xn ym
í Õ ! ¤. û
y y = f ( x)
Linearization
interval
y0
y = y0 + a ( x − x0 )
x
x0
^ F IGURE 14.6. One-dimensional mapping and linearisation.
between which exists a functional relationship
y = F (x) = F ( x1 , x2 , . . . , xn )
or
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
y1 F1 (x) F1 ( x1 , x2 , . . . , xn )
y2 ⎥ ⎢ F2 (x) ⎥ ⎢ F2 ( x1 , x2 , . . . , xn )
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢ ⎥
⎢ ..
⎥=⎢ ..
⎥=⎢ ..
⎥,
. . .
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎣ ⎦ ⎣ ⎦ ⎣ ⎦
ym F m (x) F m ( x1 , x2 , . . . , xn )
likiarvojen it becomes complicated. We can choose a vector of approximate values

vektori ⎡ (0) ⎤
x1
⎢ (0) ⎥
def
⎢ x2 ⎥
x0 = ⎢ ⎢ ..
⎥,
⎣ .
⎥
⎦
(0)
xn
def
and a corresponding vector of approximate values y0 = F (x0 ), after which
again
∂F ( x1 , x2 , . . . , xn ) ⃓⃓
⃓ (︂ )︂
(0)
y = y0 + x1 − x1 +
∂ x1 ⃓
x=x 0
∂F ( x1 , x2 , . . . , xn ) ⃓⃓
⃓ (︂ )︂
(0)
+ x2 − x2 +
∂ x2 ⃓
x=x 0
..
.
∂F ( x1 , x2 , . . . , xn ) ⃓⃓
⃓ (︂ )︂
(0)
+ xn − xn +···
∂ xn ⃓
x=x 0
í Õ ! ¤. û
373
or
∂F i ( x1 , x2 , . . . , xn ) ⃓⃓
⃓ (︂ )︂
(0) (0)
yi = yi + x1 − x1 +
∂ x1 ⃓
x=x
0
∂F i ( x1 , x2 , . . . , xn ) ⃓⃓
⃓ (︂ )︂
(0)
+ x2 − x2 +
∂ x2 ⃓
x=x
0
..
.
∂F i ( x1 , x2 , . . . , xn ) ⃓⃓
⃓ (︂ )︂
(0)
+ xn − xn +··· , i = 1, . . . , m.
∂ xn ⃓
x=x
0
In this equation there are m different rows, and in every row there are
n different (linear) terms. As a summary of this system of equations, we
write the following matric equation:
y = y0 + A (x − x0 ) + · · · ,
in which the matrix A is

⎡ ∂ ∂ ∂ ⎤⃓
F F ··· F ⃓
∂ x1 1 ∂ x2 1 ∂ xn 1 ⃓
∂ ∂ ∂
⎢ ⎥⃓
⎢ ⎥⃓
F F ··· F
∂ x1 2 ∂ x2 2 ∂ xn 2
⎢ ⎥⃓
A=⎢ ⎥⃓ .
⎢ .. .. .. ⎥⃓
⎢ . . . ⎥⃓
∂ ∂ ∂
⎣ ⎦⃓
⃓
F F ··· F
∂ x1 m ∂ x2 m ∂ xn m
⃓
x=x0
This matrix is the matrix of Jacobi11 of the vector mapping F : Rn → 11

Rm between the two abstract vector spaces Rn and Rm . The matrix vektorikuvaus
describes locally, in a neighbourhood of point x = x0 , the way in which
small “perturbances” in vector x propagate into vector y: häiriö
∆y = y − y0 ≈ A (x − x0 ) = A ∆x,
def def
with definitions ∆x = x − x0 and ∆y = y − y0 . So, the map between the kuvaus
difference quantities ∆x and ∆y is locally linear. This is referred to as
linearisation.
In the general case, m ̸= n. In the special case m = n we may think that
the mapping F has an inverse mapping G = F −1 , for which käänteiskuvaus
∆x = G (∆y).
Locally, in the neighbourhood of the approximate point x0 , we may say likipiste
11 Carl Gustav Jacob Jacobi, 1804–1851, was a Jewish German mathematician, professor
at the University of Königsberg 1827–1842.
í Õ ! ¤. û
x2 y2
F : R2 ↦→ R2
x1 y1
^ F IGURE 14.7. A two-dimensional mapping.
of this: if the matrix A is singular, meaning its determinant det A = 0,

in a neighbourhood of point x0 , this means that the mapping F does not
in that neighbourhood have an inverse mapping. This again means that
there may be many (in fact, infinitely many) different values ∆x having
the same image ∆y = F (∆x). On the other hand, if det A ̸= 0 in the point
x0 , such an inverse mapping does (in a sufficiently small neighbourhood
of approximate point y0 ) exist.
Interpretation The determinant det A describes the way in which vol-

umes are mapped under the vector mapping F .
If, for example, n = m = 2, it describes how the surface area of a
small square in the Rn space is mapped to the surface area of a
suunnikas parallelogram in the Rm space. The determinant is the ratio of
these two surface areas.
If n = m = 3, the determinant similarly equals the ratio of the
volumes of a small cube in Rn space and of its corresponding
suuntaissärmiö parallelepiped in Rm space.
If the ratio is zero, then apparently the square is “squeezed” to
a line segment, and the cube to a parallelogram, and the map is
thus singular.
^ 14.6.3 Linearisation of observation equations

Let us consider as an example the functional relationship between un-
knowns x and observables ℓ, which in a realistic observation geometry
is rarely linear. We have to linearise: let the non-linear observation
í Õ ! ¤. û
375
equations be
ℓ + v = F (ˆ︁
x), (14.8)
in which F (·) is a multidimensional non-linear “observation function”.

The equations are linearised by expanding them into a Taylor series
around roughly estimated solution co-ordinates (“approximate values”),
and using only the first-degree terms of the series. If the approximate
co-ordinates used are not good enough, we end up iterating.
Choose the vector of approximate values x0 and compatibly the vector
ℓ0 so that the following applies for them:
ℓ0 = F (x0 ). (14.9)
So, if the number of unknowns is m and the number of observations n:

(0) (0) (0) (0)
ℓi = F i ( x1 , x2 , . . . , xm ), i = 1, . . . , n.
This is subtracted from equation 14.8, and we do a Taylor series expan-

sion, retaining only the linear terms:
(︂ )︂
(0) (0) (0) (0)
ℓi − ℓi + v i = F i (ˆ︁
x1 , ˆ︁ xm ) − F i ( x1 , x2 , . . . , xm ) ≈
x2 , . . . , ˆ︁
m ⃓
∂F i ⃓⃓
(︂ )︂
(0)
∑︂
≈ xj − xj .
∂x j ⃓
ˆ︁
j =1 x=x0
Call
⃓
∂F i ⃓⃓
def
Ai j = , i = 1, . . . , n, j = 1, . . . , m, (14.10)
∂ x j ⃓x=x
0
the elements of the second-order design matrix. The matrix itself is then rakennematriisi
⎡ ∂ ∂ ∂ ⎤⃓
F F ··· F
⃓
∂ x1 1 ∂ x2 1 ∂ xm 1 ⃓
∂ ∂ ∂
⎢ ⎥⃓
⎢ ⎥⃓
F F ··· F
∂ x1 2 ∂ x2 2 ∂ xm 2
⎢ ⎥⃓
A=⎢ ⎥⃓ .
⎢ .. .. ..⎥⃓
⎢ . . .⎥⃓
∂ ∂ ∂
⎣ ⎦⃓
⃓
F F ··· Fn ⃓
∂ x1 n ∂ x2 n ∂ xm (0) (0)
x1 = x1 , x2 = x2 , ..., xm = xm
(0)
If we call
def def
∆ℓ = (ℓ − ℓ0 ) = ℓ − F (x0 ) , ∆ˆ︁
(︁ )︁
x = (ˆ︁
x − x0 )
(“replacement” or “linearised” observables and unknowns), we obtain for

the linearised observation equations
∆ℓ + v = A ∆ˆ︁
x. (14.11)
í Õ ! ¤. û
The least-squares solution to be computed here minimises the sum of

−1
squares of residuals vT Q ℓℓ v, wherefrom the name comes. The matrix Q ℓℓ
12 is the weight-coefficient matrix12 characterising the relative precision and
possible statistical interdependence (correlation) of the observations.

From equation 14.11, often the ∆ are left off for the sake of writing
convenience. The ∆ quantities are typically much smaller than the “whole”
quantities. Therefore the numerics work well even if the elements of the
matrix A are not exact.
Equation 14.9 however must always be calculated precisely, and with a
sufficient number of decimals.
^ 14.7 Propagation of variances

If the stochastic quantity y is a linear function of the stochastic quantity
x:
y = λ x,
we may also write
σy = λσx ,
in which σx and σy are the mean errors of quantities x and y. We may

also write
{︁ }︁
E y = E {λ x } = λ E { x }
odotusarvojen (“propagation of expectancies”), in which E {·} is the expectancy operator.
kasautumislaki The expectancy is a linear operator.
If we define the variance as follows:

{︂(︁ )︁2 }︂
def
Var{ x} = σx2 =E x − E { x} ,
it follows that
σy2 = λ2 σx2 .
This is the law of propagation of variances for simple stochastic quanti-

ties.
If the stochastic quantities
⎡ ⎤ ⎡ ⎤
x1 y1
x2 y2
⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥
x=⎢ ..
⎥, y=⎢ ..
⎥
. .
⎢ ⎥ ⎢ ⎥
⎣ ⎦ ⎣ ⎦
xn ym
12 One writes Var ℓ = Σℓℓ = σ2 Q ℓℓ , in which Σℓℓ is the variance matrix of the observa-
{︁ }︁
tions, and σ the mean error of unit weight.
í Õ ! ¤. û
Example: the forward geodetic problem 14.8
377
have several components — in other words, they are abstract vectorial
quantities — it holds, if y = Λx, that
E y = Λ E {x}
{︁ }︁
(14.12)
and
Var y = Λ Var{x}ΛT ,
{︁ }︁
(14.13)
in which now Λ and the variances are matrices.

⎡ ⎤
λ11 λ12 ··· λ1n
⎢ λ21 λ22 λ2n
⎢ ⎥
···
Λ=⎢
⎥
⎢ .. .. ..
⎥
⎣ . . .
⎥
⎦
λm1 λm2 · · · λmn
is an m × n size matrix.
σx21
⎡ ⎤
σ x1 x2 · · · σ x1 xn
..
⎢ σ x2 x1 σx22
⎢ ⎥
.
Var{x} = Σxx = ⎢
⎥
⎢ .. ..
⎥
..
.
⎥
⎣ . . ⎦
σ x n x1 ··· ··· σx2n
is a square matrix of size n × n, and
σy21
⎡ ⎤
σ y1 y2 · · · σ y1 ym
..
⎢ σ y2 y1 σy22
⎢ ⎥
.
Var y = Σyy = ⎢
{︁ }︁ ⎥
⎢ .. ..
⎥
..
.
⎥
⎣ . . ⎦
σ ym y1 ··· ··· σy2m
is a square matrix of size m × m. Here, the variances are

{︂(︁ )︁2 }︂
σx2i = Var x i = E
{︁ }︁
xi − E{xi }
and the covariances

{︁ }︁ {︂(︁ )︁(︁ )︁}︂
σ x i x j = Cov x i , x j = E xi − E{xi } x j − E{x j } ,
and similarly for the components of y.

Equation 14.13 is called the general law of propagation of variances. It varianssien
is a generalisation of equation 2.4, already derived in subsection 2.4.4, kasautumislaki
for an arbitrary number of variables. The linearity property assumed
here may be obtained by linearisation if needed, as discussed earlier.
í Õ ! ¤. û
σx sσα
Q
s
α σs
σy
y
P
^ F IGURE 14.8. Quantities related to the error ellipse.
^ 14.8 Example: the forward geodetic problem
As an application of the law of propagation of variances, we present the

geodeettinen forward geodetic problem in the plane, in which the known uncertainties
päätehtävä of direction and distance measurement are propagated into co-ordinate
uncertainties of an unknown point.

Given measured quantities s and α as well as the co-ordinates ( xP , yP )
of the starting point P , the problem is determining the co-ordinates of
the unknown point Q , figure 14.8:
xQ = xP + s · cos α, yQ = yP + s · sin α.
The problem is solved in the following way. Choose approximate values

s(0) , α(0) :
s = s(0) + ∆ s, α = α(0) + ∆α,
and write a Taylor series expansion:
∂ cos α ⃓
⃓
xQ ≈ xP + s(0) cos α(0) + ∆ s cos α(0) + s(0) ∆α =
∂α α=α(0)
⃓
∆x
x(0) ⏟ ⏞⏞ [︄ ]︄⏟
∆s
⏟ ⏞⏞ ⏟ [︂ ]︂
= xP + s(0) cos α(0) + cos α(0) − s(0) sin α(0) ,
∆α
and in the same way
∆y
y(0) ⏟ ⏞⏞ [︄ ]︄⏟
∆s
⏟ ⏞⏞ ⏟ [︂ ]︂
yQ ≈ yP + s(0) sin α(0) + sin α(0) s(0) cos α(0) .
∆α
í Õ ! ¤. û
379
Now, dropping, but remembering, the approximation labels (0), and
turning the vectors x and y into random or stochastic quantities:
[︄ ]︄ [︄ ]︄
def ∆x def ∆s
y= , x= ,
∆y ∆α
and
[︄ ]︄ [︄ ]︄
cos α − s sin α σ2s 0
Λ= , Var{x} = ,
sin α s cos α 0 σ2α
the above equations may be written compactly as
y = Λ x.
The variance matrix is

[︄ ]︄
σx2 σx y
Var{y} = = Λ Var{x}ΛT =
σx y σy2
[︄ ]︄[︄ ]︄[︄ ]︄
cos α − s sin α σs2 0 cos α sin α
=
sin α s cos α 0 σα2 − s sin α s cos α
and the elements are
σx2 = σs2 cos2 α + σα2 s2 sin2 α,

σy2 = σs2 sin2 α + σα2 s2 cos2 α,
σ x y = cos α sin α σs2 − s2 σα2 ,
(︁ )︁
computed using the law 14.13 of propagation of variances.13 13
By substitution of
xQ − xP yQ − yP
cos α = s , sin α = s ,
13 If we express the variance of direction α in gon, we may substitute into all equations
)︃2
σα [g]
(︃
σα2 = ,
ρ
in which ρ is the size of a radian in the degree unit in question, in this case ρ =
63.661 977 236 758. Similarly if one uses seconds of arc: then
)︃2
σα [′′ ]
(︃
σα2 = ,
ρ
in which now ρ = 57.295 779 513 × 60 × 60 = 206 264.806 247.
í Õ ! ¤. û
an alternative form is still obtained:

(︂ x − x )︂2
Q P 2
σx2 = Var{∆ x} = s σs2 + ( yQ − yP ) σα2 ,
{︁ }︁ (︂ yQ − yP )︂2 2 2
σy2 = Var ∆ y = s σs + ( xQ − xP ) σα2 ,
(︃(︂ )︃
σs
)︂2
σ x y = Cov ∆ x, ∆ y = − σα2
{︁ }︁
s ( xQ − xP )( yQ − yP ) . (14.14)
This is how mean errors of the observations σs , σα are converted into

co-ordinate mean errors σx , σy . The precision is affected both by the
observational precisions σs , σα and by the geometry s, α.
^ 14.8.1 The error ellipse and its eigenvalue problem

The error ellipse or uncertainty ellipse is the statistical area of uncertainty
of the location solution ( x, y) of a point. This can be used in statistical
testing.
As a measure of point precision, there exists a suitable quantity that is
independent of the directions of the co-ordinate axes. The error ellipse
is really a visual representation of the point variance matrix. The vari-
ance matrix of the co-ordinates ( xQ , yQ ) of point Q in figure 14.8 — or
equivalently, of the linearised co-ordinates (∆ x, ∆ y) — may be written as
{︄[︄ ]︄}︄ [︄ }︁ ]︄ [︄ ]︄
∆x Var{∆ x} Cov ∆ x, ∆ y σx2
{︁
σx y
Σ = Var = = .
∆y Cov ∆ x, ∆ y Var{∆ y} σy2
{︁ }︁
σx y
The invariants of this matrix are its eigenvalues and -vectors: the so-
ominaisarvo- lutions of the eigenvalue problem (Σ − λ I ) x = 0, (λ i , x i ), i = 1, 2. If we
tehtävä rotate the co-ordinate axes so, that they are oriented along the main axes
of the ellipse, we obtain

[︄ ]︄
s2 σα2
Σ=
σs2
and obviously λ1 = s2 σα2 and λ2 = σs2 .

More generally, one solves the determinant equation
[︄ ]︄ [︄ ]︄
Σ11 − λ Σ12 σx2 − λ σx y
det(Σ − I λ) = det = det = 0.
Σ21 Σ22 − λ σx y σy2 − λ
This is called the characteristic polynomial:
det(Σ − I λ) = σx2 − λ σy2 − λ − σ2x y = 0,

(︁ )︁(︁ )︁
í Õ ! ¤. û
381
so
λ2 − σx2 + σy2 λ + σx2 σy2 − σx2y = 0,
(︁ )︁ (︁ )︁
(14.15)
a standard-issue quadratic equation. The solutions or eigenvalues are ominaisarvo

(︃ )︃
1
)︁ √︂(︁ )︁2
σx2 + σy2
(︁ (︁ )︁
λ1,2 = 2 ± σx2 + σy2 − 4 σx2 σy2 − σx2y =
(︃ )︃
1
)︁ √︂(︁ )︁2
σx2 + σy2
(︁
= 2 ± σx2 − σy2 + 4σx2y =
)︁ √︂ 1 (︁ )︁2
= 12 σx2 + σy2 ±
(︁
4 σx2 − σy2 + σx2y , (14.16)
⎷
and the semi-major and semi-minor axes of the error ellipse are λ1 and iso- ja
⎷ pikkuakselin
λ2 .
puolikkaat
The directions of the axes may also be determined: look at the linear
combination of co-ordinates
z(θ ) = x sin θ + y cos θ ,
which is a function of the direction angle θ .

The propagation law of variances yields
Var{ z} = σx2 sin2 θ + σy2 cos2 θ + 2σ x y sin θ cos θ .
The axis directions of the visual ellipse are stationary values of this
function of direction angle θ ,
d
Var{ z} = 0.
dθ
By differentiation
2 sin θ cos θ σx2 − σy2 + 2 cos2 θ − sin2 θ σ x y = 0

(︁ )︁ (︁ )︁
=⇒ sin 2θ σx2 − σy2 + 2 cos 2θ · σ x y = 0,

(︁ )︁
and
2σ x y
(︃ )︃
1
θ= 2 arctan − 2 + k · 100g =
σx − σy2
−σ x y
= arctan (︁ √︂ (︁ )︁2 + k · 100g ,
1 1
)︁
2 σx − σy + 4 σx − σy + σx2y
2 2 2 2
using the half-angle formula for the arc tangent.14 14
One obtains still from equation 14.16:
λ1 + λ2 = Var{ x} + Var y = σx2 + σy2

{︁ }︁
(14.17)
í Õ ! ¤. û
15 and15
λ1 λ2 = det Σ = σx2 σy2 − σx2y . (14.18)
The quantities 14.17 and 14.18 are invariants — so, always the same,
no matter how the co-ordinate axes are oriented — and quantity 14.17,
called the point variance of point P , σP2 , is a particularly suitable measure
of point precision:
σP2 = σx2 + σy2 .
The point mean error σP is the square root of this point variance.
^ 14.9 Observables and observation equations in practice

Here we present the observables of classical geodesy: measurements
vinoetäisyys of horizontal directions, zenith angles and slant ranges, in the form of
linearised observation equations. The GPS observables presented already
above, equations 12.1 and 12.2, are special cases of slant-range measure-
ment.
All the observables presented are of observations between two points,
from one point to the other. The observation equations describe the
dependence of the observables on the unknowns, in this case the three-
dimensional co-ordinates of the points.
We only present the observation equations of single observations: in
a real-life situation one must combine the observation equations for
all observations, and all unknowns, into a system of equations before
proceeding to solution.
^ 14.9.1 Slant-range measurement

tähys Let the co-ordinates of instrument and signal be ( x1 , y1 , z1 ) and ( x2 , y2 , z2 ).
Then, the functional model, equation 14.8, is
√︂
2 2 2
s = ( x2 − x1 ) + ( y2 − y1 ) + ( z2 − z1 ) . (14.19)
14 This is how one avoids division by zero in the edge case σx2 = σy2 .
15 Because the equation 14.15 can be written into the form
(λ − λ1 )(λ − λ2 ) = λ2 − (λ1 + λ2 ) λ + λ1 λ2 = 0,
from which λ1 + λ2 = σx2 + σy2 , the trace of the matrix Σ, and λ1 λ2 = σx2 σy2 − σx2y , its
determinant.
í Õ ! ¤. û
Observables and observation equations in practice
[︂ ]︂T
14.9
383
Assume first that point 1, co-ordinates x1 y1 z1 , is known. Then,
[︂ ]︂T
unknowns are only the co-ordinates of point 2, x2 y2 z2 .
(0) (0) (0)
Choose approximate values x2 , y2 , z2 for these, and consistently
√︃(︂ )︂2 (︂ )︂2 (︂ )︂2
def (0) (0) (0)
s(0) = x2 − x1 + y2 − y1 + z2 − z1 .
Subtraction yields
√︂
(0) 2 2 2
∆s = s − s = ( x2 − x1 ) + ( y2 − y1 ) + ( z2 − z1 ) −
√︃(︂ )︂2 (︂ )︂2 (︂ )︂2
(0) (0) (0)
− x2 − x1 + y2 − y1 + z2 − z1 ≈
∆ x2 ∆ y2 ∆ z2
⏟(︂ ⏞⏞ )︂⏟ ⏟ ⏞⏞ )︂⏟ ⏟(︂ ⏞⏞ )︂⏟
∂s ∂s ∂s
(︂
(0) (0) (0)
≈ x − x2 + y − y2 + z − z2 ,
∂ x2 2 ∂ y2 2 ∂ z2 2
a truncated Taylor expansion.

Here,
[︂ the
]︂ abstract
[︂ “vector”
]︂ ℓ of the observations has only one element,
ℓ= ∆ s = s − s(0) , and design matrix A (equation 14.10) consists of
the partial derivatives of this one observable with respect to all unknowns
x2 , y2 , z2 :
∂s ∂s ∂s
[︂ ]︂
A= .
∂ x2 ∂ y2 ∂ z2
The elements are derived by differentiating equation 14.19:
∂s x −x ∂s y −y ∂s z −z
= 2 s 1, = 2 s 1, = 2 s 1.
∂ x2 ∂ y2 ∂ z2
The numerical result is obtained by substituting the approximate values:

(0) (0) (0)
[︃ ]︃
A≈ x2 − x1 y2 − y1 z2 − z1 .
s(0) s(0) s(0)
Using the approximate values of the co-ordinates is allowed if they are
“good enough”, so close to reality that the linearisation is valid.
The elements of this matrix A are the components of a unit vector! By
using the approximate values for the horizontal direction angle (azimuth)
α and the zenith angle ζ, we may write
[︂ ]︂
A≈ cos α(0) sin ζ(0) sin α(0) sin ζ(0) cos ζ(0) ,
or symbolically
A = eT12 ,
í Õ ! ¤. û
in which e12 is the direction vector between points 1 and 2 (as an abstract
vector of components), for which holds
∥e12 ∥ = 1.
The more general case in which both points are unknown is discussed
next. The vector of unknowns is formed
[︂ ]︂T
def (0) (0) (0) (0) (0) (0)
x= x1 − x1 y1 − y1 z1 − z1 x2 − x2 y2 − y2 z2 − z2 ,
[︂ ]︂ [︂ ]︂
and the “vector” of observations is again ℓ = ∆s = s − s(0) . Design
matrix A , equation 14.10, is
∂s ∂s ∂s ∂s ∂s ∂s
[︂ ]︂
A= .
∂ x1 ∂ y1 ∂ z1 ∂ x2 ∂ y2 ∂ z2
Based on the previous, the end result is
(0) (0) (0) (0) (0) (0)

[︃ ]︃ [︂ ]︂
A ≈ x12 y12 z12 x12 y12 z12 = −eT12 eT12 ,
− (0) − (0) − (0)
s s s s(0) s(0) s(0)
in which we used the abbreviations
(0) def (0) (0) (0) def (0) (0) (0) def (0) (0)
x12 = x2 − x1 , y12 = y2 − y1 , z12 = z2 − z1 .
The linearised observation equation is now

[︄ ]︄
[︂ ]︂ [︂ ]︂ ∆r1
s − s(0) = −eT12 eT12
∆r2
in which the linearised co-ordinates

⎡ ⎤ ⎡ (0) ⎤
∆xi xi − xi
∆r i = ⎣ ∆ yi ⎦ = ⎣ yi − yi(0) ⎦ ,
def
i = 1, 2
⎢ ⎥ ⎢ ⎥
(0)
∆zi zi − zi
are the unknowns, six in total.

Write this in stochastic form, with residual v:
x
ℓ
ˆ︁
v A ⏟ ⏞⏞ ]︄⏟
[︄
⏟[︂ ⏞⏞ ]︂⏟ ⏟[︂ ⏞⏞ ]︂⏟ ⏟[︂ ⏞⏞ ]︂⏟ ∆ˆ︁r1
∆s + v = −eT12 eT12 ,
∆r2
ˆ︁
in which
(0) (0) (0) (0) (0) (0)
[︃ ]︃
T
e12 = x2 − x1 y2 − y1 z2 − z1 =
s(0) s(0) s(0)
í Õ ! ¤. û
Observables and observation equations in practice 14.9
385
x
∆x 2
∂α
∆x
∂x
α
∂α
∆y
∂y
∆y y
1
F IGURE 14.9. The geometry of azimuth measurement (seen from above) and
^ elements of the design matrix.
]︂(0) ]︂(0)
x2 − x1 y2 − y1 z2 − z1 x12 y12 z12
[︂ [︂
= = ,
s s s s s s
⎡ ⎤ ⎡ (0) ⎤
∆ˆ︁
xi xi − xi
ˆ︁
def (0)
∆ˆ︁
r i = ⎣ ∆ ˆ︁ yi − yi ⎦ ,
yi ⎦ = ⎣ ˆ︁ i = 1, 2,
⎢ ⎥ ⎢ ⎥
(0)
∆ˆ︁
zi zi − zi
ˆ︁
def
∆ s = s − s(0) .
Here, the superscript (0) always identifies an approximate value. For

approximate values, the functional model applies exactly:
√︃(︂
(0) 2 (0) 2 (0) 2
)︂ (︂ )︂ (︂ )︂
(0)
(︁ (0)
)︁ (0) (0) (0)
s =f x = x2 − x1 + y2 − y1 + z2 − z1 .
^ 14.9.2 Azimuth measurement

If the azimuth or horizontal direction α is measured between unknown
points 1 and 2:
y −y
(︂ )︂
α = arctan x2 − x1 + k · 200g ,
2 1
then the vector of unknowns can be formed as

[︂ ]︂T
(0) (0) (0) (0)
x= x1 − x1 y1 − y1 x2 − x2 y2 − y2 .
The design matrix is obtained again using the chain rule, as follows,
abbreviating x12 = x2 − x1 , y12 = y2 − y1 . See figure 14.9 for the geometric
í Õ ! ¤. û
logic:
∂ arctan y12 x12 ∂ arctan y12 x12 ∂ y12 x12 ∂ x12

(︁ /︁ )︁ (︁ /︁ )︁ (︁ /︁ )︁
= · · =
∂ x1 ∂ y12 x ∂ x12 ∂ x1
(︁ /︁ )︁
12
(︃ )︃
1 y12 y12 y12 sin α
= (︁ /︁ )︁2 · − 2 · (−1) = = = ρ .
1 + y12 x12 x12 2 2
x12 + y12 ρ2
The whole matrix is now
sin α cos α sin α cos α (0)

[︂ ]︂
A≈ + ρ − ρ − ρ + ρ =
[︂ y − y x −x y −y x2 − x1 ]︂(0)
2 1
= + − 2 2 1 − 2 2 1 + ,
ρ2 ρ ρ ρ2
√︂
2 2
in which ρ = ( x2 − x1 ) + ( y2 − y1 ) , the distance between instrument
and signal, projected onto the horizontal plane.
From this, the linearised observation equation is obtained:
x
ˆ︁
⏟ ⏞⏞ ⎤⏟
⎡
⏟⎢ ∆ˆ︁
A x1
ℓ v
⏟ ⏞⏞
⎢ ∆ ˆ︁
⎥
⏟⏞⏞⏟ ⏟⏞⏞⏟ [︂ y2 − y1 x2 − x1 y2 − y1 x2 − x1 ]︂ ( 0) y1 ⎥
∆α + v = +
ρ2
−
ρ2
−
ρ2
+
ρ2
⎢
⎢ ∆ˆ︁
⎥,
⎣ x ⎥
2 ⎦
∆ ˆ︁
y2
in which the deltas are formed in the usual way by subtracting approxi-
mate values from observed or estimated values:
(0) (0)
(0)
∆ˆ︁ x1 − x1 , ∆ ˆ︁
x1 = ˆ︁ y1 = ˆ︁
y1 − y1 ,
∆α = α − α ,
(0) (0)
∆ˆ︁ x2 − x2 , ∆ ˆ︁
x2 = ˆ︁ y2 = ˆ︁
y2 − y2 .
^ 14.9.3 Horizontal-angle measurement

In practice, azimuth measurement is only possible using a gyrotheodolite,
an expensive specialty instrument popular in underground measure-
ments. Ordinary theodolites are not able to measure absolute horizontal
directions but only direction differences, also known as horizontal angles.
One way of deriving an observation equation for this observation type is
by adding an orientation unknown to the vector of unknowns, as already
done in subsection 6.4.1: in that case, instrument station 1 still has, in
addition to the three co-ordinates x1 , y1 , and z1 , a fourth unknown Ω1 ,
and the observation equation is
y −y
θ = arctan x2 − x1 − Ω1 + k · 200g .
2 1
í Õ ! ¤. û
Observables and observation equations in practice 14.9
387
Zenith
z
∆z
∂ζ
∆z
∂z
ζ
∂ζ x
∆ρ
∂ρ
α ∆ρ ρ
^ F IGURE 14.10. The geometry of zenith-angle measurement.
Linearisation is done as in the preceding subsection, with the result
∆ℓ v
⏟⏞⏞⏟ ⏟⏞⏞⏟ x
ˆ︁
∆θ + v =
⏟ ⏞⏞ ⎤⏟
⎡
∆ˆ︁
x1
A
⏟ ⏞⏞ ⏟⎢ ∆ ˆ︁
⎢
y1 ⎥
⎥
[︂ y2 − y1 x2 − x1 y2 − y1 x −x ]︂ (0 )
+ 2 2 1
⎢ ⎥
= + − −1 − ⎢ ∆Ω
ˆ︁ 1 ⎥,
ρ2 ρ2 ρ2 ρ ⎢ ⎥
∆ x
⎢ ⎥
⎣ ˆ︁2 ⎦
∆ ˆ︁
y2
(0)
in which ∆Ω
ˆ︁ 1 = Ω
ˆ︁ 1 − Ω . This unknown represents the unknown az-
1
imuth of the zero mark on the horizontal circle of the instrument while it vaakakehä
stands on point 1.
^ 14.9.4 Zenith-angle measurement

The zenith angle between points 1 and 2 is measured,
z2 − z1 z −z
ζ = arccos s = 100g − arctan 2 ρ 1 ,
√︂
2 2
in which again ρ = ( x2 − x1 ) + ( y2 − y1 ) .
Write [︂ ]︂
A= − A 12 A 12
í Õ ! ¤. û
in which (chain rule)

]︂(0)
∂ζ ∂ρ ∂ζ ∂ρ ∂ζ
[︂
A 12 = · · =
∂ρ ∂ x2 ∂ρ ∂ y2 ∂ z2
[︂ ]︂(0)
1 1 1
=
s cos ζ · cos α s cos ζ · sin α − s sin ζ =
[︂
z2 − z1 x2 − x1 z2 − z1 y2 − y1 ρ ]︂(0)
= · ρ · ρ − . (14.20)
s2 s2 s2
Here, the partial derivatives of ζ have been computed in cylindrical co-
def
ordinates (ρ , α, z), with the definition z = z2 − z1 . In these co-ordinates,
ζ = ζ(ρ , z), so ζ does not depend on the azimuth, and we obtain (figure
14.10):
∂ζ 1 1 z −z ∂ζ 1 1ρ
= cos ζ = s 2 s 1 , = − s sin ζ = − s s .
∂ρ s ∂ z2
Now, the linearised observation equation is
x
ˆ︁
A
ℓ
⏟ ⏞⏞ ]︄⏟
v ⏟[︂ ⏞⏞ [︄
]︂⏟ ∆ˆ︁
⏟⏞⏞⏟ ⏟⏞⏞⏟ r1
∆ζ + v = − A 12 A 12 ,
∆r2
ˆ︁
in which A 12 as above, equation 14.20, and

⎡ ⎤ ⎡ (0) ⎤
∆ˆ︁
xi xi − xi
ˆ︁
(0)
∆ˆ︁
r i = ⎣ ∆ ˆ︁ yi − yi ⎦ ,
yi ⎦ = ⎣ ˆ︁ i = 1, 2.
⎢ ⎥ ⎢ ⎥
(0)
∆ˆ︁
zi zi − zi
ˆ︁
^ 14.10 Tacheometer measurement

We can bring the results of the previous four subsections together in
describing tacheometer measurement, the simultaneous measurement of
a horizontal direction, a vertical angle and a distance:
⎡ ⎤
∆ˆ︁
x1
∆ ˆ︁
⎢ ⎥
⎢ y1 ⎥
⎡ ⎤ ⎡ ⎤ ⎡ ⎤⎢ ⎥
s vs [3×3] 0 [3×3]
⏟⏞⏞⏟ ⎥ ⎢
⎢ ∆ˆ︁
z1 ⎥
⎥
⎥ ⎢ ⏟ ⏞⏞ ⏟
θ
⎣ ⎦ ⎣ θ ⎦ = ⎣ − A 12
+ v −1 A 12 ⎦ ⎢ ∆Ω ⎥,
⎢ ⎥ ⎢ ⎢ ˆ︁ 1 ⎥
∆ˆ︁
⎢ ⎥
ζ vζ 0 ⎢
⎢ x2 ⎥
⎥
∆ ˆ︁
y2
⎢ ⎥
⎣ ⎦
∆ˆ︁
z2
with ⎡ x2 − x1 y2 − y1 z2 − z1 ⎤(0)
s s s
⎢ y2 − y1 x2 − x1 ⎥
A 12 = ⎢
⎢ − + 0 ⎥
ρ2 ρ2 ⎥
z2 − z1 x2 − x1 z2 − z1 y2 − y1 ρ
⎣ ⎦
ρ ρ −
s2 s2 s2
í Õ ! ¤. û
Helmert transformation in the plane 14.11
389
with all the above definitions and results.
^ 14.11 Helmert transformation in the plane
^ 14.11.1 Theory
In the plane, if more than two points are given in both co-ordinate frames,
it is possible to derive the unknown parameters of a Helmert transfor-
mation between the two frames. We start from equation 3.4 for a single
point (note the notation change):
[︄ ]︄ [︄ ]︄ [︄ ]︄[︄ ]︄
x′ x0 cos θ sin θ x
= +K .
y′ y0 − sin θ cos θ y
def
Writing K = 1 + m, with m the scale distortion, assumed to be small, this
becomes, for also small rotation angles θ : kiertokulma
[︄ ]︄ [︄ ]︄ [︄ ]︄[︄ ]︄
x′ x0 1 θ x
≈ + (1 + m) ≈
y′
y0 −θ 1 y
[︄ ]︄ [︄ ]︄ [︄ ]︄[︄ ]︄ [︄ ]︄[︄ ]︄
x0 x m 0 x 0 θ x
≈ + + + ,
y0 y 0 m y −θ 0 y
from which [︄ ]︄ [︄ ]︄ [︄ ]︄
x′ x x0 + mx + θ y
− ≈ ,
y′ y y0 + m y − θ x
which can be rearranged into
x
⏟ ⏞⏞ ⏟
ℓ A ⎡ ⎤
⏟
[︄ ⏞⏞ ]︄⏟ ⏟
[︄ ⏞⏞ ]︄⏟ x0
′
x −x x(0) y(0)
1 0
⎢ y ⎥
⎢ 0 ⎥
= ⎥.
′
0 1 y(0) − x(0)
⎢
y −y ⎣ m ⎦
θ
This is recognised as a set of two observation equations, with the ob-
servation vector, the vector of unknowns, and the design matrix being,
respectively, ℓ, x, and A, and in which x(0) , y(0) are approximate val-
(︁ )︁
ues16 of the co-ordinates ( x, y) of the point. 16
For multiple points given in the two frames:

⎡ ⎤
x0
(0) (0)
[︄ ]︄ [︄ ]︄
x′i − x i 1 0 xi yi ⎢ y ⎥
⎢ 0 ⎥
= (0) (0) ⎥,
yi′ − yi
⎢
0 1 yi −xi ⎣ m ⎦
θ
16 Use of approximate values is allowed here, as both m and θ are assumed to be small.
í Õ ! ¤. û
with i the point number. We see that, if the number of available points n
exceeds 2, there will be redundancy: more observation equations — 2 n —
than there are unknowns — 4. Written stochastically, we now have, with
residuals,
x
ˆ︁
⏟ ⏞⏞ ⏟
ℓ v A ⎡ ⎤
⏟ ⏞⏞ ]︄⏟ [︄
⏟ ⏞⏞ ]︄⏟ ⏟[︄ ⏞⏞ ]︄⏟ x0
ˆ︁
(0) (0)
[︄
x′i − x i v2i−1 1 0 xi yi ⎢ y0
ˆ︁ ⎥
+ = ⎥, i = 1, . . . , n.
⎢ ⎥
y′i − y i v2i (0) (0) ⎢
m
0 1 yi −xi ⎣ ˆ︁ ⎦
θ
ˆ︁ (14.21)
^ 14.11.2 Propagation of uncertainty

With more than two points common between the ( x′ , y′ ) and ( x, y) frames,
it will be possible not only to derive the Helmert transformation param-
eters, but also how the point uncertainties propagate into those of the
transformation parameters. See for example Kahmen and Faig (1988),
pages 253–255.
We start from the Helmert transformation equation rewritten as ob-
servation
(︂ )︂equations, 14.21. It is assumed that the co-ordinate values
(0) (0)
x i , yi available for the points are good enough, and that the scale
distortion m and the rotation angle θ of the axes are both small. This is
typically the case as many, also local, co-ordinate reference frames are
approximately of the correct scale and oriented correctly to the north.
The above observation equations have 2 n rows, where n is the number
of points: i = 1, . . . , n. The design matrix A looks like
(0) (0)
[︄ ]︄
1 0 xi yi
A= (0) (0)
.
0 1 yi −xi
If we assume that the co-ordinates of the given points are all equally pre-
cise, we may derive the normal matrix, or weight matrix of the unknowns,
∑︁
Pxx as follows (the symbol means summation over all given points, all
sub- and superscripts dropped):
⎡ ∑︁ ∑︁ ⎤
n 0 x y
∑︁ ∑︁
⎢ 0
T n y − x ⎥
Pxx = A A = ⎢ ∑︁ ⎥.
⎢ ⎥
( x2 + y2 )
∑︁ ∑︁
⎣ x y 0 ⎦
( x + y2 )
2
∑︁ ∑︁ ∑︁
y − x 0
The significance of this is that the variance matrix of the unknowns
í Õ ! ¤. û
Helmert transformation in the plane
[︂ ]︂T
14.11
391
x=
ˆ︁ x0 ˆ︁
ˆ︁ ˆ︁ θˆ︁
y0 m is precisely
σx20 σ x0 y0 σ x0 m σ x0 θ
⎡ ⎤
⎢ σ y0 x0 σy20 σ y0 m σ y0 θ
{︁ }︁ def ⎢ ⎥
x =⎢
Var ˆ︁ ⎥ = σ2 Q xx = σ2 Pxx
−1
,
⎥
⎣ σmx0 σm y0 σm 2
σ mθ ⎦
σ θ x0 σθ y0 σθ m σθ2
−1
in which Q xx = Pxx is the weight-coefficient matrix of the unknowns and
σ is the mean error of unit weight, in this case, the assumed precision of
one point co-ordinate. Here, the diagonal elements σx20 = Var ˆ︁
{︁ }︁
x0 and so
{︁ }︁
on are variances, σ x0 y0 = Cov ˆ︁
x0 , ˆ︁
y0 and so on, covariances.
∑︁ ∑︁
If now x= y = 0, in other words, ( x, y) are centre-of-mass co-
ordinates, we obtain
⎡ ⎤
n 0 0 0
⎢ 0 n 0 0 ⎥
Pxx = ⎢
⎢ ⎥
∑︁ 2
( x + y2 )
⎥
⎣ 0 0 0 ⎦
( x + y2 )
2
∑︁
0 0 0
−1
and, with Q xx = Pxx ,
⎧⎡ ⎤⎫
⎪
⎪ x0 ⎪
ˆ︁ ⎪
⎪ ⎪
⎨⎢ y0 ⎥ ⎬
⎥ = σ2 Q xx =
{︁ }︁ ˆ︁
x = Var ⎢
Var ˆ︁
⎢ ⎥
⎪
⎪ ⎣ mˆ︁ ⎦⎪
⎪
⎪ ⎪
θ
⎩ ˆ︁ ⎭
1 n
⎡ /︁ ⎤
0 0 0
1 n
⎢ 0 /︁
0 0 ⎥
= σ2 ⎢ ⎥.
⎢ ⎥
1
/︁∑︁
⎣ 0 0 ( x2 + y2 ) 0 ⎦
1 ∑︁
/︁
0 0 0 ( x 2 + y2 )
Now, the variances of the transformation parameters are
σ2
x0 = σx20 = σ2 (Q xx )11 = n ,
{︁ }︁
Var ˆ︁
σ2
y0 = σy20 = σ2 (Q xx )22 = n ,
{︁ }︁
Var ˆ︁
/︂
2 ∑︁ (︁ 2
2
= σ2 (Q xx )33 = σ
{︁ }︁
ˆ︁ = σm x + y2 ,
)︁
Var m
/︂∑︁ (︁
2
Var θˆ︁ = σθ2 = σ2 (Q xx )44 = σ
{︁ }︁
x 2 + y2 .
)︁
Moreover, the parameters do not statistically correlate with each other:

the covariances between them vanish.
Now, the actual observation equations to be solved are 14.21, in which
the quantities v2i−1 , v2i are the important residuals, containing valuable
quality-control information.
í Õ ! ¤. û
^ T ABLE 14.2. Point set given in two different co-ordinate frames. The points
are in the municipality of Porvoo, the co-ordinate frames are ETRS-
GK27 and KKJ zone 3.
Point x′ y′ x y
10 6697976.388 27427023.033 6698108.117 3427192.039

36 6700867.976 27428566.410 6700999.695 3428735.464
714 6701212.704 27424871.393 6701344.461 3425040.439
717 6696502.735 27428579.172 6696634.432 3428748.182
17_vara 6697821.437 27424568.639 6697953.182 3424737.628
35_vara 6709919.416 27434433.390 6710051.148 3434602.545
2061 6694497.478 27432539.402 6694629.164 3432708.409
2062 6704170.468 27432163.151 6704302.189 3432332.245
2063 6703595.075 27426736.815 6703726.833 3426905.879
1. What is redundancy?
2. Why are observations weighted in an adjustment?
3. Why are the residuals of an adjustment of interest?
4. What is the mean error of unit weight?
5. What is the difference between variance matrix and weight-
coefficient matrix?
6. What is a better-known name for the matrix of Jacobi resulting
from the linearisation of observation equations?
7. What is the propagation law of variances for a stochastic vector
quantity?
8. How many independent invariants has the variance matrix of a
point location in the plane? What are they (for example)?
9. What are the parameters of a two-dimensional or plane Helmert
transformation? What does the variance matrix of the least-squares
solution of the plane Helmert transformation parameters look like
in centre-of-mass co-ordinates?
^ Exercise 14–1: Helmert transformation parameter estimation

A set of points is given in two different co-ordinate frames, see table 14.2
(Porvoo). With this data, do the following operations:
17 1. Calculate (solve) the four Helmert transformation parameters17 x0 ,
í Õ ! ¤. û
Exercise 14–1: Helmert transformation parameter estimation
393
y0 , m, and θ . An octave script is provided for the purpose, tableau
14.3.
2. Calculate the parameters x0′ , y0′ , m′ , and θ ′ of the inverse transfor-
mation — so, swap columns x′ and y′ with columns x and y.
3. Verify that the scales K = m + 1 and K ′ = m′ + 1 are each other’s
inverses: K ′ = 1 K and thus m′ ≈ − m, and that the rotation param- kiertoparametri
/︁
eters θ are each other’s opposites: θ ′ = −θ .

4. Look at the residuals. How precise were the input co-ordinates of
the points?
5. How many observations are there (hint: the number of observations
equals the number of residuals)? How many unknowns? What is the
number of degrees of freedom b (the difference between the number
of observations and the number of unknowns, the redundancy)?
√︂ ∑︁
def 1 2n 2
6. Estimate σ
ˆ︁ = b i =1 v i , in which b is the number of degrees of
freedom, v i the residuals, and 2 n the number of observations. The
estimated quantity σ is called the mean error of unit weight, the
typical precision of a point co-ordinate.18 18
7. Repeat the calculation after changing one x value by adding one

metre. Note the effect on the residuals. Would you be able to
identify this co-ordinate as the one containing the gross error? Also
recompute σ
ˆ︁. What do you see?
17 We leave off the stochastic underlines here, as these parameter values are just single
realisations of the corresponding stochastic variables.
18 More correctly, this only holds true if one of the two co-ordinate sets given, ( x′ , y′ )
and ( x, y), is assumed to be exact. Otherwise it is the typical precision of a co-ordinate
difference.
í Õ ! ¤. û
^ T ABLEAU 14.3. Calculation script for Helmert transformation. Note that in this
code, no reduction to center-of-mass co-ordinates takes place. The co-ordinate
convention is the geodetic one: x points north, y east. Unlike in the text, however,
the rotation angle θ is assumed positive in the anti-clockwise direction.
format long
% Data input:
npts = 9;
pts = [10, 36, 714, 717, 17, 35, 2061, 2062, 2063];
x2 = [6697976.388, 6700867.976, 6701212.704, 6696502.735, 6697821.437,...
6709919.416, 6694497.478, 6704170.468, 6703595.075];
y2 = [27427023.030, 27428566.410, 27424871.393, 27428579.172, 27424568.639,...
27434433.390, 27432539.402, 27432163.151, 27426736.815];
x1 = [6698108.117, 6700999.695, 6701344.461, 6696634.432, 6697953.182,...
6710051.148, 6694629.164, 6704302.189, 6703726.833];
y1 = [3427192.039, 3428735.464, 3425040.439, 3428748.182, 3424737.628,...
3434602.545, 3432708.409, 3432332.245, 3426905.879];
% Observable vector:
yoffset = (27-3)*1E6;
ell = [x2’-x1’;y2’-y1’-yoffset];
% Construct design matrix:

A = zeros(2*npts,4);
for i = 1:npts
A(i,:) = [1, 0, x1(i), -y1(i)];
A(i+npts,:) = [0, 1, y1(i), x1(i)];
end
% Solve for unknowns:

x = inv(A’*A)*A’*ell;
x0 = x(1);
y0 = x(2) + yoffset;
scale = 1.0 + x(3);
rot = 57.296*x(4);
printf(’\nUnknowns:\n\n’);
printf(’x0: %20.10f\ny0: %20.10f\nscale :%20.10f\nrot: %20.10f\n\n’,...
x0, y0, scale, rot);
% Residuals:
v = A*x - ell;
printf(’Residuals:\n\n’);
for i = 1:npts
printf(’%5i %10.3f %10.3f\n’, pts(i), v(i), v(i+npts));
end
í Õ ! ¤. û
^ Statistical methods in geodesy
15
[. . . ] There has not been a single date in the history of the
theory of gravitation when a modern significance test would
not have rejected all laws and left us with no law.
Nevertheless the law did lead to improvement for centuries,
[. . . ]
Harold Jeffreys, 1939 (Jeffreys, 1998, page 391)
I N THIS CHAPTER we present two interrelated subjects:

◦ Statistical testing, in the context of the validation of and outlier poikkeava
detection in geodetic network measurements. havaintoarvo
◦ The reliability of geodetic networks.
The framework of hypothesis testing with null and alternative hypotheses

is adopted. We also show how this framework may be used, for example
for geodetic deformation analysis.
We would be amiss in not pointing out that hypothesis testing is not
the appropriate framework for settling all scientific disputes. There are
other, often more appropriate, techniques, like the Akaike information
criterion (Burnham and Anderson, 2013) and Bayesian approaches. The
above quote from Harold Jeffreys1 is apposite. 1
^ 15.1 The method of least-squares

Explaining the method of least-squares is simplest if one assumes that pienimmän
all observables are stochastic quantities that are normally distributed neliösumman
menetelmä
(figure 2.5), both individually and together: they form a multivariate
1 Sir Harold Jeffreys FRS (1891–1989) was a British mathematician, statistician, geo-
physicist, and astronomer, an influential advocate of Bayesian statistics.
– 395 –
396 15 S TATISTICAL METHODS IN GEODESY
normal distribution. If normally distributed observables are statistically

independent of each other — for example, if they were produced by
independent measurement processes — this is automatically the case.
The method of least-squares as a way to minimise the random errors in
estimated quantities such as point co-ordinates has been presented above.
parametrinen In chapter 14, the parametric adjustment method was presented, in which
tasoitus observations are expressed as functions of the unknowns. The alterna-
ehtotasoitus tive, adjustment by conditions, is suitable, for example, for computing

monikulmiojono traverses.
Here, the parametric method will be discussed in more detail.
Let the observations, as a vector ℓ, be linear functions of the unknowns
x:
ℓ A x n
⏟⎡ ⏞⏞ ⎤⏟ ⏟
⎡ ⏞⏞ ⎤⏟ ⏟⎡ ⏞⏞ ⎤⏟ ⏟⎡ ⏞⏞ ⎤⏟
ℓ1 a 11 a 12 · · · a 1m x1 n1
ℓ2 ⎥ ⎢ a 21 a 22 · · · a 2m x2 n2
⎢ ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥
⎢ ⎥⎢ ⎥ ⎢ ⎥
⎢ ..
⎥=⎢ . .. ..
⎥⎢ ..
⎥+⎢ ..
⎥.
⎥ ⎢ .
. ⎦ ⎣ . . . . .
⎢ ⎥⎢ ⎥ ⎢ ⎥
⎣ ⎦⎣ ⎦ ⎣ ⎦
ℓn a n1 a n2 · · · a nm xm nn
[ n] [ n× m] [ m] [ n]
Here, n is the number of observations, m the number of unknowns. The

observations, elements of the vector ℓ, are stochastic quantities. Assume
that they are normally distributed around the “true” value of the observed
quantity. Then, the elements of the vector of observation errors n are also
2 normally distributed.2
pienimmän In this, rather general, case we may compute the least-squares solution
neliösumman in the following way:
ratkaisu
−1
)︁−1 (︁ T −1 )︁
x = A T Q ℓℓ
(︁
ˆ︁ A A Q ℓℓ ℓ , (15.1)
[ m] −1 [ m× n] [ n]
[ m× m]
in which Q ℓℓ is the weight-coefficient matrix of the observations, size

[ n × n ]: ⎡ ⎤
q 11 q 12 ··· q 1n
⎢ q21 q22 · · · q2n
⎢ ⎥
⎥
Q ℓℓ = ⎢
⎢ .. .. .. ..
⎥.
⎣ . . . .
⎥
⎦
q n1 q n2 · · · q nn
2 Often, they are also assumed to be statistically independent from each other, meaning
that their random variations happen independently of each other. However, neither the
x nor those of the vector of residuals v will be statistically
elements of solution vector ˆ︁
independent of each other.
í Õ ! ¤. û
The residuals from the adjustment 15.2
397
This matrix is related to the variance matrix of the observations as
follows:
⎡ ⎤
σ12 σ12 · · · σ1n
⎢ σ21 σ22 · · · σ2n
⎢ ⎥
{︁ }︁ def
Var ℓ = Σℓℓ = ⎢ ⎥ = σ2 Q ℓℓ ,
⎥
⎢ .. .. ..
.
..
⎣ . . .
⎥
⎦
σn1 σn2 ··· σn2
in which
{︃(︂ }︃
{︁ }︁)︂2
σi2 = σ2 q ii ,
{︁ }︁
= Var ℓ i = E ℓi − E ℓi
{︃(︂ }︃
{︁ }︁)︂(︂ {︁ }︁)︂
= σ2 q i j .
{︁ }︁
σ i j = Cov ℓ i , ℓ j = E ℓi − E ℓi ℓj − E ℓj
Here, σ is the mean error of unit weight. painoyksikön

keskivirhe
The variance matrix of the solution is obtained though propagation of
varianssien
variances. Let kasautumislaki
x = L ℓ,
ˆ︁
in which
def −1
)︁−1 −1
L = A T Q ℓℓ A T Q ℓℓ
(︁
A .
Then, based on equation 15.1:
−1
)︁−1 −1 −1 −1
)︁−1
Q xx = LQ ℓℓ LT = A T Q ℓℓ A T Q ℓℓ A A T Q ℓℓ
(︁ (︁
A · Q ℓℓ · Q ℓℓ A =
−1
)︁−1
= A T Q ℓℓ
(︁
A
x,
by suitable elimination. So, the variance matrix of the solution vector ˆ︁
Σxx = σ2 Q xx , is obtained in any case as a side product of computing the
solution, equation 15.1!
^ 15.2 The residuals from the adjustment

The least-squares estimators of the observations ℓˆ︁ and unknowns ˆ︁
x are
connected to each other through the functional model
ℓˆ︁ = Aˆ︁
x,
and from the original observations
ℓ = Ax + n
one computes the residuals:3 jäännösvirhe

3
def
v = ℓˆ︁− ℓ = Aˆ︁
x − ℓ = A (ˆ︁
x − x) − n.
í Õ ! ¤. û
Residuals are central in the quality control of geodetic network

solutions.
◦ The size of the residuals tells something about the contradictions

karkea virhe present in the network solution, possible gross errors, or even model
errors.
◦ The size of the residual of a certain observation can indicate
whether there may be a gross error hiding in this observation.
◦ The network must be reliable: there has to be redundancy, an
over-determination by the observational material. For example, all
kinds of closing errors offer possibilities for checking.
With little or no redundancy, the residuals may well be small or
even zero, but this means nothing!
An often-used form of the observation equations is
ℓ + v = Aˆ︁
x.
pienimmän The residuals v of a least-squares adjustment have four nice properties,

neliösumman here given without proof:
tasoitus
1. The quadratic form
def −1
E = vT Pℓℓ v = vT Q ℓℓ v = σ2 vT Σ−ℓℓ1 v ,
(︁ )︁
the weighted sum of the squares of the residuals, is minimised —

this is what the methods of least-squares got its name from. In fact,
the square root of this quantity is the norm of the vector of residuals
4 v, or its length, in the Q ℓℓ metric, which is thus minimised:4 E =
2
∥v∥Q .
3 Thevector v of residuals is not the same as the vector of observation errors, or “noise”,
n! The residual is the difference between the original observation and the adjusted
observation: in other words, a correction. However, not even an adjusted observation —
or unknown — is the “truth”. The truth is not precisely knowable; it is only approximable
at best, and the values of the elements of the vector n, unlike the values of the elements
of the vector v, cannot be computed.
4 One could eliminate the weight matrix altogether by applying a co-ordinate transfor-
mation in the vector space of observations: do a Cholesky decomposition Pℓℓ = ΓΓT ,
def
vT˜︁
resulting in E = vT Pℓℓ v = vT ΓΓT v = ˜︁ v = ΓT v. This is automatically achieved
v, with ˜︁
def
ℓ = ΓT ℓ. This is the straightforward way of reducing
by redefining the observables as ˜︁
the general least-squares problem 15.1 to the simpler unweighted one 14.3. The new
ℓ do not correlate with each other and have identical mean errors.
observables ˜︁
í Õ ! ¤. û
The residuals from the adjustment 15.2
399
Ob-
“Null space”
ℓ ser-
va-
tion
space
ℓ3
||v|| minimal
ℓ1 ℓ2 v
a2
ℓ optimal
ˆ︁
O ℓˆ︁′ sub-opt.
a1 “Parameter space”
F IGURE 15.1. Least-squares adjustment as an orthogonal projection. For vi-

sualisation, it is assumed that the observation space has three
dimensions, the parameter space two, and thus the number of
^ degrees of freedom, the dimensionality of the null space, is one.
2. The variance Σλλ of an arbitrary linear combination λ

ˆ︁ = Λˆ︁
x of the
⎷
x (and its mean error Σλλ ) is minimised.
unknowns ˆ︁
3. The adjusted observables ℓˆ︁ and the residuals v are mutually or- tasoitettu
thogonal in the Q ℓℓ metric: if the scalar product is defined as havaintosuure
⟨︁ ⟩︁ def −1
a·b Q
= aT Q ℓℓ b:
T −1 −1
xT A T Q ℓℓ
⟨︁ ⟩︁ ⟨︁ ⟩︁
ℓˆ︁· v Q
x·v
= Aˆ︁ Q
= ( Aˆ︁
x) Q ℓℓ v = ˆ︁ v = 0,
because
−1
aTi Q ℓℓ v = 0, i = 1, . . . , m,
in other words, the vector of residuals is orthogonal to all columns

a i of the design matrix A . rakennematriisi
Figure 15.1 gives a geometrical interpretation: the unknowns are
those coefficients in the linear combination of the columns of the A
matrix that minimise the norm of the vector of residuals.
x and the
4. The covariance matrix between the vector of unknowns ˆ︁
vector of residuals v vanishes: Σxv = σ2 Q xv = 0. So, they do not jäännös-
correlate with each other. virheiden
vektori
Because
ℓ = Aˆ︁
x − v,
í Õ ! ¤. û
it follows, based on the law of propagation of variances and the above-

mentioned property 4, that
Q ℓℓ = AQ xx A T + Q vv =⇒ Q vv = Q ℓℓ − AQ xx A T , (15.2)
a useful equation for computing the weight-coefficient matrix Q vv — and

the variance matrix Σvv = σ2 Q vv — of the residuals.
^ 15.3 Testing, hypotheses for testing

The observational material may contain gross errors. In a real-life ad-
justment calculus we must be able to say, based on our knowledge of the
statistical distribution of the observations,
◦ Something about the possible occurrence of gross errors.
◦ How large gross errors would have to be in order to be noticed and
removed.
◦ The propagation of gross errors of this magnitude into the un-
knowns of interest.
Finding gross errors belongs to the field of statistical testing.
Gross errors that are found can be handled in two ways:
◦ They are removed from the observation set, and the measurements
in question are repeated. After the fact, this is laborious and costly,
wherefore at least part of statistical testing is done already in the
field.
◦ They are simply left out. This assumes that the measurement was
planned redundantly from the start: so many measurements have
been made that one can afford to leave a (small) fraction of them
out.
Statistical testing always requires the formulation of hypotheses. One of
the hypotheses is always the
Null hypothesis All the measurements in the network are correct,
there are no gross errors in them. This hypothesis is designated
by the symbol H0 .
In addition, there must always be at least one
Alternative hypothesis The network contains some gross error, or
some combination of gross errors, or a specific gross error. This
hypothesis is designated by the symbol Ha .
í Õ ! ¤. û
Testing, hypotheses for testing 15.3
401
^ T ABLEAU 15.1. The planning and measurement process.
Network planning
⏐
Reliability
⏐
Precision ↓
How much uncertainty If a gross error occurs, how
−→←−
will there be in the ⏐ well can we detect it in
co-ordinates? testing?
⏐
↓
Network geometry
and quality metrics
⏐
⏐
↓
Network measurement
⏐
Overall validation
⏐
Testing the observations
↓
Are the results sensible?
−→←− For each observation: is this
Are the models used valid? ⏐
Are there gross errors?
⏐
↓ one observation OK?
End result
◦ Adjusted network
◦ Precision
◦ Reliability
◦ Confidence that methods
used are pretty OK
Generally we wish to know, or make a judgement on, two matters:

◦ Are there generally any gross errors left in this observation set?
◦ Is this specific observation in error?
Tableau 15.1 shows the role of testing in the whole planning and
measurement process.
These questions will be discussed separately in the following sections.
í Õ ! ¤. û
^ 15.4 Overall validation
^ 15.4.1 The χ2 distribution

Firstly we choose the following alternative hypothesis:
Ha : somewhere in the measurement material (we do not yet

know where) there is a gross error.
This kind of hypothesis can be tested using the χ2 test. The method, and
the tables belonging to it, can be found in statistics textbooks and on the
Internet. The quantity to be tested is the length of the vector of residuals
in the Σℓℓ or Q ℓℓ metric, its norm, squared:
1 2 def T −1 1 (︁ T −1 )︁ def 1 2
E = ∥v∥Σ = v Σ ℓℓ v = v Q ℓℓ v = 2 ∥v∥Q . (15.3)
σ 2 σ 2 σ
painoyksikön This quantity — note the scaling with the variance of unit weight — is
varianssi distributed according to the χ2 distribution: the χ2 distribution with
n− m
n − m degrees of freedom, figure 15.2. The number of degrees of freedom
is the difference between the number of observations and the number of
unknowns, also known as the redundancy b = n − m.
Conceptually, a stochastic quantity with the χ2b distribution, for b
degrees of freedom, is obtained as the sum of the squares of b independent,
odotusarvo standard-normally distributed — that is, having an expectancy zero
and a mean error one — stochastic quantities n i , i = 1, . . . , b. As the
expectancy of one such square equals the variance of the standard normal
distribution, being one, it follows that the expectancy
b
∑︂ b
∑︂ b
∑︂
χ2b n2i
{︁ }︁ {︁ }︁ {︁ }︁
E = E = Var n i = 1 = b.
i =1 i =1 i =1
Yet another perspective is that the quantity
def 2 ℓℓE vT Σ−1 v

σ
ˆ︂2 =
n−m =σ n−m
harhaton has the expectancy E σ ˆ︂2 = σ2 and is thus an unbiased estimator of σ2 . It

{︁ }︁
estimaattori is called the a posteriori variance of unit weight. See also subsection 6.4.3.
/︂
The ratio σ2 σ2 is expected to be close to unity if the null hypothesis is
ˆ︂
valid.
^ 15.4.2 The overall test

By testing the above quantity 15.3, one may infer whether the measure-
í Õ ! ¤. û
Overall validation 15.4
403
0.2
χ2 distribution for
4 degrees of freedom
non-central χ2 ,
non-centrality parameter λ = 3
Testing bound
0.1
αχ2
βχ2
0.0
0 2 4 6 8 10 12 14 16 18 20
^ F IGURE 15.2. The χ2 distribution with four degrees of freedom.
ment material contains some gross error or not,5 without yet stating in 5
which observation it might be found.
MATLAB contains ready routines for applying the χ2 method.
The quantity6 E σ2 is distributed according to χ2n−m only in the case 6
/︁
that the material contains no gross errors, that is, the null hypothesis H0
applies. Then, the expectancy of the testing variate is
1 ⃓
{︂ ⃓ }︂
= E χ2n−m = n − m.
{︁ }︁
E E ⃓ H0
σ2
Assume now instead however, that the observations contain one or more
gross errors, taken together ∇ℓ: the alternative hypothesis Ha applies.
The effect of this error vector on the residuals is ∇v. In this case, the
distribution of the quantity E σ2 is the non-central χ2 , in figure 15.2 the
/︁
red curve.
5 The χ2 test cannot distinguish between actual gross errors and the possible unsuit-
ability of the functional model ℓ = A x applied to the material. If it happens that the χ2
test rejects the null hypothesis, but all observations appear to be okay, it might be that
there is a problem with the functional model: some systematic effect may have been
overlooked.
6 In Baarda’s terminology: “shifting variate”. Willem Baarda (1917–2005) was a pioneer
of modern adjustment theory and statistical geodesy.
í Õ ! ¤. û
^ 15.4.3 Reliability
The interesting question is now how large the effect of ∇ℓ on ∇v is going
to be. We may hope that it will be large, because then, the network is
7 reliable.7 Generally the whole of ∇ℓ does not propagate into ∇v; the
tasoitus adjustment conveys part into the vector of unknowns, as a gross-error

8 effect,8 of magnitude ∇x. See also section 15.9.
In this situation, the vector of residuals is v + ∇v. Written out into

terms, the testing variate becomes
1
E = ∥v + ∇v∥2Σ = vT Σ−ℓℓ1 v + vT Σ−ℓℓ1 ∇v + ∇vT Σ−ℓℓ1 v + ∇vT Σ−ℓℓ1 ∇v.
σ 2
The expectancy of the testing variate becomes
1 ⃓
{︂ ⃓ }︂
E E ⃓ Ha =
σ2
1
{︁ T }︁ −1 1 T −1
= E v T Σ− Σℓℓ ∇v + ∇vT Σ−
ℓℓ E v + ∇v Σℓℓ ∇v =
{︁ }︁ {︁ }︁
ℓℓ v + E v
= E χ2n−m + 0 + 0 + ∇vT Σ−1

{︁ }︁
ℓℓ ∇v = ( n − m) + λ,
def
in which λ =∇vT Σ−1
ℓℓ ∇v is called the non-centrality parameter of the χ
2
distribution. It describes how far the effect of the assumed gross error on
the residuals, ∇v, extends outside the uncertainty area of the observations
as described by the matrix Σℓℓ .
A quadratic quantity is always positive. Therefore, the χ2 test is
one-sided, unlike the later presented test for the normal distribution.
So, because E is a quadratic quantity, every gross error — and even
systematic errors, that is, errors in the functional model used — will
tend to increase it. Each and every error tends to make χ2 larger, and
makes noticing the error more likely. This makes the χ2 test such a useful
overall test.
In fact, the χ2 test validates a lot more than just the observations. It
assures that
◦ The observation set does not contain any (large) gross errors.
◦ The functional model used (the observation equations) is valid with
sufficient accuracy.
◦ The assumed mean errors of the observations (and the possible
assumption of non-correlatedness) are realistic.
7 This is called interior reliability.

8 This is called exterior reliability. A small effect means a good exterior reliability.
í Õ ! ¤. û
Locating gross errors 15.5
405
^ 15.5 Locating gross errors
If we have inferred that the material presumably contains one or more
gross errors, we next want to find out which observations are under sus-
picion. Let us assume for simplicity that a gross error occurs only in one
observation, although there might well be errors in several observations
simultaneously.
The simplest way to search for gross errors, or rather, to undertake
outlier detection, is to look at the residuals. poikkeava
havaintoarvo
Let the vector of residuals be
[︂ ]︂T
v= v1 v2 · · · v i · · · v n−1 v n .
The element v i is the residual of observation number i , that is the obser-

vation ℓ i . The element’s variance is
σ2v i = σ2 Q vv
[︁ ]︁
ii
,
and its mean error σv i is the square root of this.

Assume that the residuals v i are normally distributed. Then, we may
test every observation i = 1, . . . , n:
⃓ ⃓
⃓v ⃓ > 1.96 σv =⇒ ℓ is probably in error
i i i
⃓ ⃓
⃓v ⃓ ≤ 1.96 σv =⇒ ℓ is presumably correct.
i i i
This two-sided test based on the normal distribution uses a significance merkitsevyys-
level of 95 %: Even if there is no gross error, there nevertheless is a taso
probability of 100 % − 95 % = 5 % that, based on the test, observation ℓ i
will be rejected.
Table 15.2 gives the rejection bounds for different significance levels in
the two-sided test based on the standard normal distribution.
The method described here works correctly only if the observations do
not correlate with each other, so that the matrix Q ℓℓ is a diagonal matrix.
If it is not, the literature offers an adapted9 testing method called data 9
snooping (Baarda, 1968).
9 The def−1
trick is simply that, instead of the residuals v, weighted residuals w = Q ℓℓ v and
their variances are used.
The logic is, that if we search for a gross error in observation number i , we look for
the linear combination of residuals in which the error shows clearest. We calculate the
orthogonal projection of v (in the Q ℓℓ metric) on the direction of the assumed gross error
í Õ ! ¤. û
^ T ABLE 15.2. Rejection bounds h for significance levels α2 in a two-sided test

based on the standard normal distribution, having mean error
σ = 1 and expectancy µ = 0. See figure 15.4.
1 − α2 , % α2 , % h
5 95 1.96
2.5 97.5 2.24
1 99 2.57
0.1 99.9 3.29
^ 15.6 Calculation example: linear regression

Let us return to the linear regression example already used in subsection
14.5.3, see table 15.3.
Recall that the least-squares solution found was
ˆ︁ = 1.76 ± 1.25 σ,
a
ˆ︁ = 0.495 ± 0.349 σ.
b
We compute the (︂ function)︂ values aˆ︁ + bx i of the fitted line, as well as its
ˆ︁
∑︁n
residuals v i = aˆ︁ + bx
ˆ︁ i − y . The condition
i i =1 v i = 0 is a good sanity
check.
^ T ABLE 15.3. Example of linear regression.
∑︁5
i→ 1 2 3 4 5 i =1
xi 1.51 2.44 3.34 4.41 5.05 16.75

yi 2.32 3.12 3.57 3.93 4.15 17.09
a
ˆ︁ + bx
ˆ︁ i 2.51 2.97 3.41 3.94 4.26
vi +0.19 −0.15 −0.16 +0.01 +0.11 0.00
v2i 0.0361 0.0225 0.0256 0.0001 0.0121 0.0964
[︂ ]︂T
def
ei = 0 0 ··· 1 ··· 0 0 (where the element “1” is in place i ):
def −1
w i = ⟨e i · v⟩Q = eTi Q ℓℓ v.
Together, the components w i form the vector w and they are optimally suited for
discerning gross errors, or “outliers”.
In the test, also the mean error σw i of every individual w i is needed, to be computed
from the diagonal elements of the matrix (equation 15.2)
def
−1 −1 −1 −1
Q ℓℓ − AQ xx A T Q ℓℓ
(︁ )︁
Q ww = Q ℓℓ Q vv Q ℓℓ = Q ℓℓ ,
just as when testing using v.
í Õ ! ¤. û
Calculation example: linear regression 15.6
407
^ T ABLE 15.4. Values of the cumulative χ23 distribution. αχ2 is the significance
level of the χ2 test.
´x ´∞
0 χ23 ( x′ ) dx′ x χ23 ( x′ ) dx′
(︁ )︁ (︁ )︁
x αχ2 1 − αχ2
4.642 0.80 0.20

6.251 0.90 0.10
7.815 0.95 0.05
9.837 0.98 0.02
11.345 0.99 0.01
12.838 0.995 0.005
14.796 0.998 0.002
16.266 0.999 0.001
If the observations y i have a variance matrix Σℓℓ = σ2 I , then the

“shifting variate” to be tested is siirtosuure
1
E = vT Σ−ℓℓ1 v,
σ2
in which v is the vector formed by the residuals v i . We obtain

n n
1 1 ∑︂ 2 ∑︂
E = v i =⇒ E = v2i .
σ2 σ2
i =1 i =1
If it is given a priori that σ = ±0.15, it follows that
1 0.0964
E= = 4.28.
σ 2 0.0225
The quantity E σ2 is distributed according to χ23 : there are n = 5

/︁
observations and m = 2 unknowns (a and b), so the number of degrees of

freedom (redundancy) is n − m = 3. According to table 15.4, the probability
that under the null hypothesis χ23 > 4.642 is 20 %, so the value 4.28 is
fully acceptable, at least on a significance level of 80%.
Next, the individual residuals are tested. Compute first the weight-
coefficient matrix of the vector of residuals using equation 15.2:
Q vv = Q ℓℓ − AQ xx A T ,
in which Q ℓℓ = I , [︄ ]︄
1.5640 −0.4072
Q xx =
−0.4072 0.1215
í Õ ! ¤. û
^ T ABLE 15.5. Example of linear regression. Computing the residuals, their

variance-covariance matrix, and normalised residuals.
i→ 1 2 3 4 5
vi +0.19 −0.15 −0.16 +0.01 +0.11
+0.3887 −0.4032 −0.2019 +0.0375 +0.1807

−0.4032 +0.6998 −0.2006 −0.0821 −0.0112
Q vv −0.2019 −0.2006 +0.8007 −0.1978 −0.1969
+0.0375 −0.0821 −0.1978 +0.6646 −0.4178
+0.1807 −0.0112 −0.1969 −0.4178 +0.4502
σv i 0.0935 0.1255 0.1342 0.1223 0.1006

⃓ ⃓/︂
⃓v ⃓
i σv i 2.03 1.20 1.19 0.08 1.09
was already computed in subsection 14.5.3, and

⎡ ⎤
⎡ ⎤ 1 1.51
1 x1
⎥ ⎢ 1 2.44
⎢ ⎥
1 x2
⎢ ⎥
⎢ ⎥ ⎢ ⎥
A=⎢ .. ..
⎥ = ⎢ 1 3.34 ⎥.
. .
⎢ ⎥ ⎢ ⎥
⎣ ⎦ ⎢ 1 4.41 ⎥
⎣ ⎦
1 xn
1 5.05
After careful calculation (MATLAB!), we obtain the matrix Q vv , shown in
table 15.5. Of this matrix, it is mostly the diagonal elements that are
interesting:
√︂ ⎷
σv1 = σ [Q vv ]11 = 0.15 · 0.3887 = 0.15 · 0.623 = 0.0935,
√︂
σv2 = σ [Q vv ]22 = 0.1255,
10 and so on. (Remember that σ = 0.15.) See table 15.5.10
As can be seen, all observations are acceptable, with the exception of y1 ,

which, on the 95 % significance level, is barely rejected (rejection bound
1.96). However, already on a significance level of 97.5 %, it too is accepted.
^ 15.7 Adding a gross error

karkea virhe Next, we add to the observed value y a simulated gross error +1.0.
3
10 Note how the mean errors of the residuals are systematically smaller than the mean
errors of the observations σ = ±0.15, especially close to the edges! With a large number
of points, this phenomenon vanishes and we may write Q vv ≈ Q ℓℓ .
This is often done in any case. Then, gross errors in the edge points will not be noticed
sufficiently well.
í Õ ! ¤. û
Significance level of the test 15.8
409
^ T ABLE 15.6. Example of linear regression. A simulated gross error in point 3:
original data, linear regression, residuals, testing.
∑︁5
i→ 1 2 3 4 5 i =1
xi 1.51 2.44 3.34 4.41 5.05 16.75

yi 2.32 3.12 4.57 3.93 4.15 18.09
x2i 2.28 5.95 11.16 19.45 25.50 64.34

xi yi 3.50 7.61 15.26 17.33 20.96 64.66
a
ˆ︁ + bx
ˆ︁ i 2.71 3.17 3.62 4.14 4.46
vi +0.39 +0.05 −0.95 +0.21 +0.31 0.01
v2i 0.1521 0.0025 0.9025 0.0441 0.0961 1.1973
σv i 0.0935 0.1255 0.1342 0.1223 0.1006
⃓ ⃓/︂
⃓v ⃓
i σv i 4.17 0.40 7.08 1.72 3.08
Rejection? * ** *
Now, as the least-squares solution we obtain the result of table 15.6:
5 · 64.66 − 16.75 · 18.09 20.2925

b
ˆ︁ = = = 0.493,
2 41.1375
(︂5 · 64.34 − 16.75 )︂
ˆ︁ = 15 18.09 − 16.75 · b
a ˆ︁ = 1.97.
In table 15.6 the σv i values have not changed.

Compute the “shifting variate”
n
1 1 ∑︂ 2 1.1973
E= 2 vi = = 53.21!
σ 2 σ 0.0225
i =1
There is something very wrong here. . .

Now look at table 15.6. The largest testing value by far, 7.08, is seen for
the erroneous observation 3. But observations 1 and 5 are also rejected
at even a 99 % significance level! For this reason, one should proceed
carefully. Based on the test, one should reject only one observation at a
time, after which the whole least-squares computation should be repeated. pienimmän
neliösumman
laskenta
^ 15.8 Significance level of the test
^ 15.8.1 Choice of rejection bound

When we test a certain alternative hypothesis against the null hypothesis
using an assumedly normally distributed testing quantity or variate, one
must choose a suitable rejection bound. If the variate to be tested exceeds
í Õ ! ¤. û
y = 1.97 + 0.493x
3
Mean error of observations

1
x
1 2 3 4 5 6
F IGURE 15.3. Example of linear regression, observation 3 contains a simulated

^ gross error.
this bound, H0 is rejected and Ha accepted. Choosing the rejection bound

is an important strategic decision.
See figure 15.4. In the figure, the rejection bound chosen is h = 2.5 σ:
if the testing variate exceeds 2.5 times its own mean error σ, the null
hypothesis H0 is rejected and the alternative hypothesis Ha is accepted.
Now, the strategy may lead to two types of error:
ensimmäisen ◦ The null is rejected although it is valid. This is called an error of the
lajin virhe first kind.11 The probability of this error happening is the size of the
11
vertically hatched (blue) area. In the case of normal distribution,
it amounts to 1 − α = 1.24 % (two-sided), if the rejection bound is
Gross error k
σ
β
Rejection bound h = 2.5 σ 1−α
^ F IGURE 15.4. Statistical testing based on normal distribution.
11 Also rejection error.
í Õ ! ¤. û
Significance level of the test 15.8
411
^ T ABLE 15.7. Rejection bound h and probability of rejection in a test in the case
of normal distribution. α1 is the significance level of the one-sided,
α2 of the two-sided test.
h σ
/︁
1 − α1 , % 1 − α2 , %
2.0 2.28 4.56

2.5 0.62 1.24
3.0 0.13 0.27
3.5 0.02 0.05
h = 2.5 σ, see table 15.7. The quantity α is called the significance

level of the test.
◦ The null is accepted although there is a gross error: H0 is false,
the alternative hypothesis Ha is true. This is an error of the second toisen lajin
kind.12 The probability of it happening depends on the size k of the virhe
12
gross error, more precisely, on the size of the normalised difference
( k − h) σ .
/︁
Its complement, the probability of rejection, is called the power β13 erotuskyky
13
of the test. In figure 15.4 it is the size of the pink area.
Choosing the testing strategy, choosing h, is thus always a compromise.
It depends on the relative costs of errors of the first and second kind —
including non-monetary, such as reputational, “costs”. h = 3σ is often
used — the “three-sigma rule”.
^ T ABLE 15.8. Assumed size k of gross error and corresponding power β of the
test. Normal distribution and rejection bound h = 2.5 σ assumed.
k σ ( k − h) σ
/︁ /︁
β, %
3.0 0.5 69.1

3.5 1.0 84.1
4.0 1.5 93.3
4.5 2.0 97.7
5.0 2.5 99.4
5.5 3.0 99.9
6.0 3.5 99.98
12 Also acceptance error.

13 So, the probability of an error of the second kind, if there is indeed a gross error in the
observation, is 1 − β, or 100 % − β.
í Õ ! ¤. û
^ 15.8.2 Harmonisation of overall and per-observation tests

In testing the body of observations, there is a link between the overall
validation test and the per-observation tests! The link is through the
significance levels. See figure 15.5: if the significance level of the χ2 test
is αχ2 and that of the test for a single observation is α, the connection is
αχ2 = αn−m ,
14 with n − m the number of degrees of freedom.14 In other words, the joint

probability that all observations individually pass their tests must be the
same as the probability of passing the overall validation. Only on that
condition may it be expected that, if the common χ2 test finds something
“rotten”, the tests for the individual observations will also point to the
“guilty” observation.
Example If αχ2 = 95 % with ten degrees of freedom, it follows that
⎷
10
α= αχ2 =
√︁
n− m 0.95 ≈ 0.99489 = 99.489 %,
some ten times closer to 100 %.
After removing or correcting the “guilty” observation, the testing proce-

15 dure is repeated, until the χ2 test is passed.15
^ 15.9 Reliability
^ 15.9.1 Principle
The reliability of a measurement network is the property that gross errors
are found easily, and are found even if they are relatively small.
Reliability corresponds to the network being “strong”. It is however not
the same kind of strength as when the network is precise.
See figure 15.6. From points A , B, and C are measured the directions
to a fourth point. Error ellipses for three different cases are drawn:
I when the point is far from the points A and B
II when the point is in a location where the directions to points A
and B are perpendicular to each other, and
14 This procedure is similar to the well-known Bonferroni correction, Wikipedia, Bonfer-

roni correction.
15 If the test still fails, perhaps the other models used should be checked, such as the
assumed precisions of the observations and so on.
í Õ ! ¤. û
Reliability 15.9
413
Obs. 2
3
Obs. 1
a
Overall acceptance area
are
Ac
c
e
ep
anc
ta
nc
ept
e
ar
Acc
ea
1 2
an ce area
Accept
Obs. 3
Common
acceptance
area of
observations
F IGURE 15.5. Harmonisation of the significance levels of the overall validation

^ and per-observation tests.
III when the point is between points A and B.
As can be seen, the most precise result is obtained in case III. The error
ellipse is the smallest.
However, reliability is poor (non-existent) in case III. If the measure-
ment made from point C contains a gross error (dashed line), we still
C III
II
I
^ F IGURE 15.6. An example of reliability.
í Õ ! ¤. û
x
B
σρ
ρ
/︁⎷
σρ 2
P
A y
O C
^ F IGURE 15.7. Another example of reliability.
obtain in case III a seemingly good — precise — but erroneous result. See
the dashed error ellipse.
In cases I and II, a gross error in the observation from C will produce
a contradictory result. It is not possible to find a location for the target
point that is compatible with the direction measurements from all three
points. This is a good thing, because it enables the detection of the gross
error. The network is then called reliable.
In network planning, attention must be paid to reliability, of course in
addition to precision. The network must be designed with appropriate
redundancy: it must contain measurements that check each other. Com-
mon sense helps a great deal here. There are mathematical and software
tools for evaluating the reliability of a network. One must always ask,
“what if this or that observation were in error. . . would I notice?”
^ 15.9.2 Another example

In this example, figure 15.7, the observation points A , B, and C are
located on the edge of a circle, and point P, the direction to which is being
measured, is located near the origin (centre point of the circle) O .
The observation equations are obtained by looking at the geometry:
ℓ v A
⏟
⎡ ⏞⏞ ⎤⏟ ⎡
⏟ ⏞⏞ ⎤⏟ ⏟ ⎡ ⏞⏞
⎤⏟⏟ ⏞⏞x
ˆ︁
θ AP − θ AO v1 −1 0 [︄ ]︄⏟
⎥ 1⎢ xP
⎣ θ BP − θBO
⎥ ˆ︁
⎦ + ⎣ v2 ⎦ = ρ ⎣ 0 −1 ⎦ .
⎢ ⎥ ⎢
yP
θ CP − θCO
ˆ︁
v3 1 0
í Õ ! ¤. û
Reliability 15.9
415
Symbolically
ℓ + v = Aˆ︁
x,
in which
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
ℓ1 θ AP − θ AO −1 0
1⎢
ℓ = ⎣ ℓ2 ⎦ = ⎣ θ BP − θBO ⎦ , A=ρ⎣ 0 −1 ⎦ .
⎢ ⎥ ⎢ ⎥ ⎥
ℓ3 θ CP − θCO 1 0
The least-squares solution is

[︄ ]︄−1 [︄ ]︄ [︄ ]︄
1
)︁−1 2 0 ℓ3 − ℓ1 2 ℓ3 − ℓ1
( )
x = AT A ATℓ = ρ ·
(︁
=ρ· .
−ℓ2 −ℓ2
ˆ︁
0 1
From this are obtained the residuals
− 12 (ℓ3 − ℓ1 ) − 21 ℓ1 − 12 ℓ3
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
ℓ1
x−ℓ = ⎣
v = Aˆ︁ ℓ2 ⎦ − ⎣ ℓ2 ⎦ = ⎣ 0 ⎦. (15.4)
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
1
2 ℓ3 − ℓ1
( ) ℓ3 − 12 ℓ1 − 12 ℓ3
Note 1 As can be seen, the observation ℓ2 has vanished from the residu-
als! If ℓ2 = θ BP − θBO contains a gross error, we are never going to
notice it as an overly large residual.
Note 2 From the residuals it cannot be seen whether a gross error comes
from observation ℓ1 or observation ℓ3 . In the residuals, their
coefficients are identical.
We may also write equation 15.4 as

)︁−1 T
x − ℓ = A AT A
(︁
v = Aˆ︁ A ℓ − ℓ = − R ℓ,
with the redundancy matrix

1 1
⎡ ⎤
2 0 2
def )︁−1
R = I − A AT A AT = ⎣ 0 0 0 ⎦ .
(︁ ⎢ ⎥
1 1
2 0 2
Each diagonal element of the redundancy matrix is a rough-and-ready

measure for how well the geometry controls for a gross error in the
1
corresponding observation. R 11 = R 33 = 2 tells us that observations
ℓ1 and ℓ3 are somewhat controlled, but R 22 = 0 tells us that ℓ2 is not
controlled at all. A sensible requirement is that all R ii ? 0.5.
Next, compute the shifting variate
1
E = vT Σ−ℓℓ1 v.
σ2
í Õ ! ¤. û
Here, Σℓℓ is the variance matrix of the observations. Assume that the
observations do not correlate with each other and that their mean error
is σ. Then ⎡ ⎤
1
Σℓℓ = σ2 ⎣ 1 ⎦.
1
We obtain ( H0 is the null hypothesis):
3
1 ⃓ 1 ∑︂ 2 1
⃓
E H 0 = vi = (ℓ1 + ℓ3 )2 .
σ 2 σ 2 2σ 2
⃓
i =1
Because the mean errors of both ℓ1 and ℓ3 are σ and they do not correlate,
⎷
the mean error of the sum ℓ1 +ℓ3 is σ 2 and its variance 2σ2 . The number
of degrees of freedom is 1 and the variate E σ2 is distributed according
/︁
to χ21 , as should be the case according to theory.
By comparing the value E σ2 computed from the observations with

/︁
the values from the χ21 table, one can test whether the observations
might contain a gross error. If all observations are free of gross
errors, the expectancy of E σ2 is 1.
/︁
Nevertheless, as already pointed out above, we cannot observe any gross

16 errors in ℓ at all. We say16 that the measurement geometry is reliable for
2
observations ℓ1 and ℓ3 , but unreliable for observation ℓ2 . If observation
ℓ2 contained a gross error of size ∇, it would slip in its entirety into the
17 co-ordinate yP as an error −ρ · ∇! We also say17 that the measurement
ˆ︁
yP , but reliable for unknown ˆ︁
geometry is unreliable for unknown ˆ︁ xP .
A sufficiently large gross error ∇ in observations ℓ1 or ℓ3 would again
be detected as an overly large value for the shifting variate (alternative
hypothesis H a ):
1 ⃓ 1
⃓
E ⃓ Ha = 2 (ℓ1 + ℓ3 + ∇)2 ,
σ 2 2σ
2
the expectancy of which is 1 + 12 ∇ σ , for a non-centrality parameter
(︁ /︁ )︁
(︁ /︁ )︁2
λ = 12 ∇ σ , see figure 15.2. If ∇ ≫ σ, this would be detected with
considerable confidence.
Note Reliability has nothing to do with precision! The precision of the
16 This is called interior reliability.

17 Exterior reliability.
í Õ ! ¤. û
Deformation analysis 15.10
417
[︂ ]︂T
x=
unknowns ˆ︁ xP ˆ︁
ˆ︁ yP is described by their variance matrix
[︄ ]︄
1
)︁−1 0
x = σ2 A T Q ℓℓ
−1
= σ2 ρ 2 2
{︁ }︁ (︁
Var ˆ︁ A ,
0 1
⎷
xP is 12 σρ 2, and that of ˆ︁
so, the mean error of ˆ︁ yP is σρ , and they
are uncorrelated with each other. See the error ellipse in figure
15.7.
However, a good mean error gives no solace if the co-ordinate
yP contains a gross error. . .
solution ˆ︁
^ 15.9.3 The meaning of redundancy

Even though the reliability of a measurement network is good, we may
still ask whether it is easy to identify the observation in which the gross
error has occurred. It this is not easy, we end up measuring all suspect
observations again, or throwing them out. This is not good.
From the viewpoints of both good reliability and identifiability of gross
errors, the degree of redundancy of a geodetic measurement network
should not be too low. If the number of observations is n and the number
of unknowns m, then the number of conditions, or degrees of freedom, is
n − m. The degree of redundancy is then ( n − m) n . This is often stated
/︁
as a percentage. For example, linear regression of a straight line through

3
five points: n = 5, m = 2, so a degree of redundancy of 5 = 60 %. On
the other hand, a levelling line of ten points between two known points:
1
n = 11, m = 10, the degree of redundancy being 11 = 9 % — weak, but
unfortunately common. By measuring in both directions we obtain n = 22,
12
m = 10, so a degree of redundancy of 22 ≈ 55 %, which is already good. A
good rule of thumb is that a degree of redundancy of 50 % is desirable.
^ 15.10 Deformation analysis

Deformation analysis is one practical application of statistical testing.
The null hypothesis H0 in these tests is, that no observable deformation
has happened. There may be many different alternative hypotheses Ha ,
from the hypothesis that some unspecified deformation took place, to
many concrete hypotheses about the precise nature of the deformation
sought.
Deformation analysis is also an application that involves the time di-
mension: measurements collected in two or more measurement epochs
í Õ ! ¤. û
are compared. The deformations studied may be natural, like deforma-

tions in the Earth’s crust brought about by tectonic movements or by
varying glacial loading; or they may be brought about by human activity,
maan like the subsidence caused by mineral extraction (petroleum, natural gas,
vajoaminen irrigation water, . . . ). The object of study may be the Earth’s crust in
vesiallas an area, or a building or other built structure like a reservoir dam. The
possibilities are very broad.
Deformation analysis is discussed in the textbooks by Cooper (1987)
pages 331–352 and Vaníček and Krakiwsky (1986) pages 611–659.
^ 15.10.1 Height deformation analysis

In one dimension, height deformation analysis studies vertical movement,
for example using levelling. In the simplest case, the same levelling line
or network of n points has been measured twice:
H i ( t 1 ), H i ( t 2 ), i = 1, . . . , n,
and the variance matrices of the heights, Σ( t 1 ) and Σ( t 2 ), are also avail-
able.
Clearly, comparison is possible only if both measurements are first
reduced to the same reference or datum point. We choose the first network
point, point 1, as the datum point:
(1) (1)
H1 ( t 1 ) = H1 ( t 2 ) ( = some agreed value) .
After this, the variance matrices for both measurement times or epochs
are only of size ( n − 1) × ( n − 1), because now point 1 is known and no
longer has (co-)variances.
⎡ (1) (1) (1) ⎤
σ22 ( t 1 ) σ23 ( t 1 ) · · · σ2n ( t 1 )
⎢ (1)
⎢ σ32 ( t1 ) σ(331) ( t1 ) · · · σ(3n
1) ⎥
(1) ( t1 )
Σ
⎥
( t1 ) = ⎢
⎢
.. .. .. ..
⎥,
. . . .
⎥
⎣ ⎦
(1) (1) (1)
σn2 ( t 1 ) σn3 ( t 1 ) · · · σnn ( t 1 )
and the same for Σ(1) ( t 2 ). Here

⎫
(1) {︁ (1) }︁
σ ii ( t ℓ ) = Var H i ( t ℓ ) , ⎬
ℓ = 1, 2; i, j = 2, . . . , n.
(1) (1)
{︁ (1) }︁
σ i j ( t ℓ ) = Cov H i ( t ℓ ), H j ( t ℓ ) ,⎭
í Õ ! ¤. û
419
(1)
σ44 ( t 2 )
3
(1)
∆H 4
t2 4
(1)
1 t1 σ44 ( t 1 )
F IGURE 15.8. Height deformation monitoring network for the epochs t 1 (red)
^ and t 2 (blue). Realistic error bars.
Now, calculate the height displacements between the two measurement

epochs and their joint variance matrix, assuming that the measurements
made at times t 1 and t 2 are statistically independent of each other:
(1) def (1) (1)
∆H i = H i ( t 2 ) − H i ( t 1 ), i = 2, . . . , n,
(1)
Σ∆H∆H = Σ(1) ( t 1 ) + Σ(1) ( t 2 ).
After this it is intuitively clear — provided that both sets of height mea-
surements are multi-normally distributed — that the following quantity,
the shifting variate, has the χ2n−1 distribution:
)︂T (︂ )︂−1
1
(︂
(1)
E = ∆H(1) Σ∆H∆H ∆H(1) ,
σ2
in which ⎡ (1) (1) ⎤ ⎡ (1) ⎤
H 2 ( t2 ) − H 2 ( t1 ) ∆H 2
⎢ (1)
⎢ H 3 ( t2 ) − H (31) ( t1 ) ⎥ ⎢ ∆H (31)
⎥ ⎢ ⎥
(1) def
∆H = ⎢
⎥
⎥=⎢
.. ..
⎥
. .
⎢ ⎥ ⎢ ⎥
⎣ ⎦ ⎣ ⎦
(1) (1) (1)
H n ( t2 ) − H n ( t1 ) ∆H n
is the (abstract) vector of height differences.
Statistical testing for deformation is based on this variate E .
^ 15.10.2 Horizontal deformation analysis

In two dimensions we proceed in the same way as in the one-dimensional
case, except that
1. It is tempting to write the plane co-ordinates as complex numbers.
2. There are now two datum points, the co-ordinates of which are
considered identical between the two epochs.
í Õ ! ¤. û
So, if there are n points, the size of the variance matrix is now ( n − 2) ×
(n − 2). The variance matrix is now also complex valued, and Hermitian:
liittoluku its transpose is its complex conjugate.
18 The testing variate is again the shifting variate18
( AB) −1
)︂† (︂
1
(︂ )︂
E = d( AB) Σdd d( AB) ,
σ2
in which d is the complex vector of all co-ordinate differences, or displace-

ment vector:
def
d( AB) =
(︂ )︂
( AB) ( AB) ( AB) ( AB)
⎡ ⎤
( t2 ) − x3 ( t1 ) + i y3 ( t2 ) − y3 ( t1 )
⎡ ( AB) ⎤
x ∆z3
⎢ 3( ) (︂ )︂
⎥ ⎢ ∆z( AB)
⎥ ⎢
⎢ x AB ( t ) − x( AB) ( t ) + i y( AB) ( t ) − y( AB) ( t ) ⎥
2 1 2 1 4
=⎢ 4 4 4 4
⎥
⎢
⎥=⎢
⎥ ⎢ ⎥,
.. ..
.
⎥
.
⎢ ⎥ ⎣ ⎦
⎣ ⎦
( AB)
( AB) ( AB)
x n ( t 2 ) − x n ( t 1 ) + i y(nAB) ( t 2 ) − y(nAB) ( t 1 ) ∆zn
(︁ )︁
with
( AB) ( AB) ( AB)

∆z i = zi ( t2 ) − zi
( t1 ) =
(︂ )︂ (︂ )︂
( AB) ( AB)
= xi ( t2 ) − x i ( t1 ) + i y(i AB) ( t2 ) − y(i AB) ( t1 ) , i = 3, . . . , n.
AB is the chosen datum or starting point for both epochs t 1 and t 2 .

The other points are numbered 3, 4, . . . , n. The symbol † signifies both
19 transposition and complex conjugate, the Hermitian19 conjugate:
def T
A† = AT = A .
^ 15.10.3 Example
Let the adjusted co-ordinates x i ( t 1 ), i = 3, . . . , 6 of the deformation
20 network from the first measurement epoch be given20 in table 15.9a,
18 Warning: in Cooper’s book (Cooper, 1987, page 335) there is a mistake under equation
(9.52): the correct equation is (inverse, not transpose):
ˆ︁ t Q −1 d
Ω=d ˆ︁ .
d
19 Charles Hermite (1822–1901) FRS FRSE was a French mathematician and number
theorist.
í Õ ! ¤. û
421
II
∆z4
I
4
B
t2
6
3 t1
F IGURE 15.9. Horizontal deformation monitoring network for the epochs t 1 (red)
and t 2 (blue). Realistic error ellipses. The alternative hypothesis
^ that domain II moves with respect to domain I is also indicated.
and the co-ordinates of the second measurement epoch x i ( t 2 ), i = 3, . . . , 6

be given in table 15.9b.
Compute the inter-epoch differences vector d, table 15.9c.
Using real numbers, with the definition
⎡ ⎤ ⎡ ⎤
x3 ( t 2 ) − x3 ( t 1 ) ∆ x3
y3 ( t 2 ) − y3 ( t 1 ) ⎥ ∆ y3 ⎥
⎢ ⎥ ⎢ ⎥
⎢ ⎢
⎢ ⎥ ⎢ ⎥
⎢
⎢ x4 ( t 2 ) − x4 ( t 1 ) ⎥ ⎢
⎥ ⎢ ∆ x4 ⎥
⎥
∆ y4
⎢ ⎥ ⎢ ⎥
def
⎢ y4 ( t 2 ) − y4 ( t 1 ) ⎥ ⎢ ⎥
d = x( t2 ) − x( t1 ) = ⎢ ⎥=⎢ ⎥,
x5 ( t 2 ) − x5 ( t 1 ) ∆ x5
⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥
∆ y5
⎢ ⎥ ⎢ ⎥
⎢
⎢ y5 ( t 2 ) − y5 ( t 1 ) ⎥ ⎢
⎥ ⎢
⎥
⎥
x6 ( t 2 ) − x6 ( t 1 ) ∆ x6
⎢ ⎥ ⎢ ⎥
⎣ ⎦ ⎣ ⎦
y6 ( t 2 ) − y6 ( t 1 ) ∆ y6
20 These are only the co-ordinates of the points to be tested. They are assumed to be
connected, for both epochs, to the same two unnamed datum points — say, points 1 and
2, or A and B — outside the area, which are assumed to be motionless.
í Õ ! ¤. û
^ T ABLE 15.9. Horizontal deformation analysis, co-ordinates. Unit metre.
(a) (b) (c)

Epoch 1 Epoch 2 Displacements
i x i ( t1 ) yi ( t1 ) i x i ( t2 ) yi ( t2 ) i ∆xi ∆ yi
3 1234.123 2134.453 3 1234.159 2134.448 3 +0.036 −0.005

4 1681.045 2507.487 4 1681.123 2507.516 4 +0.078 +0.029
5 755.495 2623.456 5 755.507 2623.487 5 +0.012 +0.031
6 1248.865 3051.775 6 1248.951 3051.807 6 +0.086 +0.032
we find by computation
6 (︂
∑︂ )︁2 (︁ )︁2 )︂
T
= 0.017771 m2 .
(︁
d d= x i ( t2 ) − x i ( t1 ) + yi ( t2 ) − yi ( t1 )
i =3
Similarly with complex numbers, with the definition
∆z3
⎡ ⎤ ⎡ ⎤
z3 ( t 2 ) − z3 ( t 1 )
⎢ z (t ) − z (t ) ⎥ ⎢ ∆z ⎥
def 2 4 1
d=⎢ 4 ⎥=⎢ 4
⎢ ⎥ ⎢ ⎥
⎦ ⎣ ∆z5
⎥
⎣ z5 ( t2 ) − z5 ( t1 ) ⎦
z6 ( t 2 ) − z6 ( t 1 ) ∆z6
we obtain in the same way by computation

6
∑︂
†
z i ( t 2 ) − z i ( t 1 ) = 0.017771 m2 .
(︁ )︁(︁ )︁
d d= zi ( t2 ) − zi ( t1 )
i =3
def def
Here, z i = x i + i y i , and z i = x i − i y i is its complex conjugate.
Let the precisions (mean co-ordinate errors) of the co-ordinates x i ( t 1 )
and y i ( t 1 ) measured at the first epoch be σ1 = ±5 cm, and the precisions
of the co-ordinates x i ( t 2 ) and y i ( t 2 ) of the second epoch σ2 = ±1 cm —
for every point, and furthermore the co-ordinates are assumed to be
21 uncorrelated.21 The variance matrices of the co-ordinate vectors are thus
Σ1 = σ12 I, Σ2 = σ22 I.
We compute the mean error σ∆ of a single co-ordinate displacement

∆ x i = x i ( t 2 ) − x i ( t 1 ), or, equivalently, ∆ y i = y i ( t 2 ) − y i ( t 1 ). Propagation of
21 This is obviously unrealistic: in real networks, the point error grows with the distance
from the datum points, and the co-ordinate errors within each epoch are strongly
correlated.
í Õ ! ¤. û
423
variances yields
σ∆2 = σ12 + σ22 = (25 + 1) cm2 = 26 cm2 .
Now, the variance matrix of the co-ordinate displacements is
Σdd = Σ1 + Σ2 = σ∆2 I,
⎷
with σ∆ = 26 cm = 5.1 cm = 0.051 m.
Now, we choose for the mean error of unit weight just this value:
def
σ = σ∆ . Then we may also write
Σdd = σ∆2 Q dd = σ2 Q dd = σ2 I,
so the weight-coefficient matrix is the unit matrix.

Compute the deformation’s testing variate, the shifting variate:
−1
1 T −1 dT Q dd d dT d
E = d Σ dd d = = 2.
σ2 σ2 σ
Here, d = x( t 2 ) − x( t 1 ) is the displacement vector, the abstract vector of
co-ordinate differences between the epochs. Because we assume that both
co-ordinate sets are given in the same, common datum, the definition
points of which nevertheless do not belong to the set 3–6, we may assume
that all co-ordinates are free. In that case, the number of degrees of
freedom is b = 2 n = 8, where n is the number of points. The variance
matrix of the components of the displacement vector, or vector of co-
ordinate differences, d, is σ2 I . We obtain
1 1 (︁ T )︁ 0.017771 m2
E = d d = = 6.835.
σ2 0.0026 m2 0.0026 m2
Question The quantity E σ2 is distributed according to the χ28 distribu-
/︁
tion. If the limit value of this distribution for a significance level

of 95 % is 15.51 (see Cooper (1987) page 355), has a deformation
probably taken place in this case?
Answer No, it has not. 6.835 < 15.51.
Question If, however, the assumed precisions were σ1 = σ2 = ±1cm,
would then a deformation have probably taken place, at a signifi-
cance level of 95 %?
Answer Yes, it would. σ2 = (1 + 1) cm2 = 0.0002 m2 and
1 1 (︁ T )︁ 0.017771 m2
E= d d = = 88.9 > 15.51.
σ 2 0.0002 m 2 0.0002 m2
í Õ ! ¤. û
1. What is the relationship, and difference, between the variance ma-
trix Σℓℓ and the weight-coefficient matrix Q ℓℓ of the observations?
2. What is the relationship between the a priori variance of unit
weight σ2 and the a posteriori one σ
ˆ︂2 ?
3. What are errors of the first kind and errors of the second kind?
4. What is the power of a statistical test?
5. What is the relationship between the significance level αχ2 of the
overall validation test and the significance level α of the outlier test
on the individual observations? Why?
6. What is redundancy, and why is it important?
7. What is interior and what is exterior reliability?
8. What are the steps in planning and measuring a geodetic network?
9. What is the Hermitian conjugate of a matrix?
10. Heathrow airport, UK, receives about 35 million incoming inter-
national passengers per year. A fancy new system is proposed to
be installed that, by analysing the behaviour of people from closed-
circuit video, can “flag” them as potential terrorists. The rate of
“false positives”, or errors of the first kind, is 1 − α = 1 %. The rate of
errors of the second kind, false negatives or justified but not-called
alarms, 1 − β, is believed to be small, less than 50 %.
The background to this is that since 1970 there have been some
4000 deaths due to terrorism in the UK.
How would you handle the passengers flagged by the system, and
why?
(a) Have them all killed.
(b) Arrest and investigate them.
(c) Send them back to where they came from.
(d) Have a chat with them and informally look into their back-
grounds before doing anything.
(e) That system is worthless.
í Õ ! ¤. û
^ Gravity in geodesy
16
Inmiddels was op 6 september 1943 ook het contact met de
‘goede’ Utrechtse hoogleraren tot stand gekomen en werden
in nauwe samenwerking met hen hervormingsplannen voor
het onderwijs (b.v. betreffende het studium generale)
uitgewerkt. Naast prof. dr. J. Jongbloed was vooral prof. dr.
ir. F. A. Vening Meinesz, te wiens huize in Amersfoort vele
vergaderingen werden gehouden, hierbij een stuwende
kracht. In verband met diens studiereizen per onderzeeër
werd het contact aangeduid als K18. Vermeldenswaard is,
dat tot en met 19 september 1944 33 vergaderingen met de
hoogleraren werden belegd.
Reference work De ondergrondse pers (“The Underground

Press”) 1940 – 1945 (Winkel and de Vries, 1989, in Dutch)
^ 16.1 Measuring gravity

Gravitation is a fundamental force of the universe. It is an attraction
acting between all celestial bodies. According to Newton’s law of gravita-
tion, the force is proportional to the masses of both bodies, and inversely
proportional to the square of the distance between the bodies.
Galileo Galilei (1564–1642) was the first to show experimentally, that
all bodies fall equally fast: their acceleration in free fall is the same,
independently of their mass m.
This may be understood in this way: that as mass m grows, gravita-
tional force F grows, but also the inertia of the body m, appearing in the
equation F = ma, grows, and the acceleration a remains unchanged. One
says that the “heavy mass” of the body is the same as its “inertial mass”. hidas massa
However, a broader way to look at this, which makes more theoretical
– 425 –
426 16 G RAVITY IN GEODESY
sense and is more general, is, that there is no difference in principle

between gravity and all kinds of pseudo-forces — like centrifugal force —
caused by non-uniform motion. The Hungarian baron and geophysicist
1 Loránd Eötvös1 carried already out many very precise tests to investigate
if any difference could be found between “heavy” and “inertial” mass,

for example in connection with material composition. The answer was
2 consistently “no”.2
From this, Albert Einstein (1879–1955) concluded logically that gravity

is a geometric property of space-time connected with its curvature, and
developed his famous field equations of general relativity, which link
together the curvature tensor of space-time and the stress-energy tensor
of the matter contained in it. Locally, for example inside a small, closed
elevator, it is not possible to know if the gravity sensed inside the elevator
is the result of the Earth’s attraction, or of the acceleration produced by
a rocket engine in the floor of the elevator!
The measurement and study of the acceleration of gravity is a special-
ism called gravimetry.
Gravity is the acceleration of free fall, and is expressed in the SI unit
m s2 . On the Earth’s surface, gravity is about 9.8 m s2 . In gravimetry, how-
/︁ /︁
ever, smaller units are needed, like the milligal (mGal) and the microgal
(µGal). In very precise work, even the nanogal (nGal) is encountered.
As a fraction of
Unit In SI units ( m s2 ) gravity (roughly!)
/︁
mGal 10−5 10−6

µGal 10−8 10−9
nGal 10−11 10−12
Instruments called gravimeters have been built for measuring gravity.

An ordinary field gravimeter is in principle just an extremely sensi-
tive spring balance (figure 16.2). Its measurement precision may be
0.01–0.1 mGal. Furthermore, there exist ballistic gravimeters, which
measure interferometrically the acceleration of a falling body; figure 16.1.
Ballistic gravimeters are absolute. Field or spring gravimeters are not
käynti absolute, and they have a drift. This means that the measurement values
1 Loránd baron Eötvös de Vásárosnamény (1848–1919) was a Hungarian geophysicist

and student of gravity.
2 Eötvös’ tests have been repeated with even much greater accuracy. The answer did not
change.
í Õ ! ¤. û
Measuring gravity 16.1
427
Vacuum pump system
Cage transporter
Prism protective cage
Falling prism
“Superspring”
g
Reference prism
Semi-transparent mirror
Laser
Mirror
Interferometer
^ F IGURE 16.1. An absolute or ballistic gravimeter.
produced by the same acceleration of gravity change slowly over time.

Therefore, field measurements are planned to always start from a known
point and end on a known point. The measurement values are adjusted tasoitus
between the end points in proportion to time according to the drift thus
determined.
δ(ε)
)
F (ε
α F (ε) cos(α + δ + ε)
F IGURE 16.2. Principle of operation of a relative or spring gravimeter: increas-

^ ing sensitivity through a diagonal solution, astatisation.
í Õ ! ¤. û
h( x, y)
s
s +
+
–
s
+ x
i y
F IGURE 16.3. The terrain height h( x, y) depicted by height contours, and height
gradients (arrows). “+”, “−”, local maximum, minimum, “s” saddle
^ point. On the right, the terrain in perspective.
^ 16.2 Gravity and geopotential
^ 16.2.1 The gradient of a scalar field

Figure 16.3 shows how the forms of the terrain are depicted on a map by
korkeuskäyrä height contours. In the figure we could have used any scalar function of
two variables as an example, instead of the terrain height h( x, y).
3 The figure shows the gradient of the height field, the vector field3
∂ h( x, y) ∂ h( x, y) def
v( x, y) = i+ j = ∂ x h( x, y) i + ∂ y h( x, y) j = grad h = ∇ h
∂x ∂y
as arrows. Here, i and j are unit vectors in the x and y co-ordinate
directions. This is a vector valued field, the value of which at every point
( x, y) consists of two components, the partial derivatives of the height
field with respect to the co-ordinates, in this point, in Euler notation ∂ x h
and ∂ y h.
The gradient vector describes the slope of the Earth’s surface: the
steeper the slope of the Earth’s surface is, the longer the gradient vector.
And the direction of the vector is of course the direction in which the
4 terrain is sloping.4
The gradient is always perpendicular to the height contour, which is a

set of points having the same height value, an equi-value curve. Along it,
height is constant.
3 The name of the symbol ∇ is nabla. This may be an old Greek word for a Phoenician
harp, the shape of which it resembles.
4 Actually the arrows are drawn in the “wrong” direction, the direction into which the
Earth’s surface is going down. Thus they describe the vector field −v.
í Õ ! ¤. û
Gravity and geopotential 16.2
429
W ( x, y, z)
F IGURE 16.4. A geopotential table. Tables like this can be found in science
museums. The surface of the table describes the gravity potential
of the Earth, albeit only in two dimensions. The arrows again
depict the gradient of the geopotential, the slope of the table
surface.
On the geopotential table, a glass marble can be made to orbit
around the “Earth” in an elliptical Kepler orbit, if the figure of
the surface agrees sufficiently realistically with Newton’s law of
^ gravitation.
Like the height field h( x, y), we may also visualise the geopotential
W ( x, y, z) in three-dimensional space, with “height contours” or equipo-
tential surfaces, and a three-dimensional gradient. On the equipotential
surfaces, the value of the geopotential is constant.
Figure 16.4 is a similar visualisation of the geopotential by the curved
surface of a table. The distance of the surface from the floor corresponds to
the geopotential, representing the energy level of an object on the surface.
The circles drawn on the table visualise the equipotential surfaces (in
reality three-dimensional) of the Earth’s gravity field, and the curves
radiating outwards from the Earth visualise the “lines of force” along voimaviiva
which the gradient vector of the potential — in other words, the gravity
vector — everywhere points.
^ 16.2.2 Normal gravity and disturbing potential

Gravity consists of two parts:
◦ the attraction by the Earth’s masses
◦ the centrifugal force, a pseudoforce caused by the Earth’s rotation.
The contribution of centrifugal force to all of gravity is about −0.3 % on
the equator, the same order of magnitude as the difference in the gravity
between equator and poles.
í Õ ! ¤. û
The gravity field of the Earth contains all kinds of irregular variations
from place to place. Most of the gravity field can however be described
pyörähdys- as the field of an ellipsoid of revolution. This mathematically defined,
ellipsoidi regular model field, in which the flattening and rotational motion of the
Earth have been taken into account, is called the normal field.
The lines of force and equipotential surfaces of the normal gravity field
are depicted in figure 16.5. The reference ellipsoid is one equipotential
surface of the normal gravity field, in the same way as the geoid (section
16.4) is an equipotential surface of the true gravity field.
The potential of the normal gravity field, the normal potential, is
written with the symbol U ( x, y, z). Normal gravity itself is the gradient
of this potential.
The gravity vector is the gradient of the geopotential W ( x, y, z):
∂W ∂W ∂W
g = ∇W = grad W = i+ j+ k = ∂ x W i + ∂ y W j + ∂ z W k,
∂x ∂y ∂z
in which i, j, and k are the unit vectors in the x, y, and z directions.

In the same way, the normal gravity vector
∂U ∂U ∂U
γ = ∇U = grad U = i+ j+ k = ∂ x U i + ∂ yU j + ∂ z U k
∂x ∂y ∂z
is also the gradient of the normal gravity potential U .

By subtracting the normal potential from the true gravity potential,
häiriö- the disturbing potential is obtained:
potentiaali
def
T = W − U.
def
The strength of normal gravity is designated by the symbol γ = ∥γ∥, in
def
the same way as the strength of true gravity g = ∥g∥. Because the two
vectors have nearly identical directions, straight down, we may also write
∂W ∂U
g≈− = −∂h W, γ≈− = −∂hU.
∂h ∂h
Normal gravity can be calculated exactly if we know, for a point P , the

geodetic latitude ϕP and the height from the reference ellipsoid h P :
γP = γ(ϕP , h P ).
Normal gravity, like true gravity, diminishes quickly when moving

upwards. The rate of diminishing is about 0.3 mGal for every metre. The
dependence on latitude is much weaker.
í Õ ! ¤. û
Gravity and geopotential 16.2
431
^ T ABLEAU 16.1. Normal potential and normal gravity according to GRS80
(Heikkinen 1981, simplified). In the equations, 9.78 . . . is normal gravity it-
self, and 0.000 003 0877 . . . the vertical gradient of normal gravity on the equator
on the surface of the reference ellipsoid. Units m, m s2 and m2 s2 .
/︁ /︁
U = 62 636 860.8500 +
+ −9.780 326 77 − 0.051 630 75 sin2 ϕ − 0.000 227 61 sin4 ϕ − 0.000 001 23 sin6 ϕ h +
(︁ )︁
+ +0.015 438 99 − 0.000 021 95 sin2 ϕ − 0.000 000 10 sin4 ϕ · 10−4 · h2 +

(︁ )︁
+ −0.000 024 22 + 0.000 000 07 sin2 ϕ · 10−8 · h3 + 0.000 000 04 · 10−12 · h4 + · · · ,

(︁ )︁
∂U
= −9.780 326 77 − 0.051 630 75 sin2 ϕ − 0.000 227 61 sin4 ϕ − 0.000 001 23 sin6 ϕ +
∂h
+ +0.030 877 98 − 0.000 043 90 sin2 ϕ − 0.000 000 20 sin4 ϕ · 10−4 · h +
(︁ )︁
+ −0.000 072 65 + 0.000 000 21 sin2 ϕ · 10−8 · h2 + 0.000 000 15 · 10−12 · h3 + · · · .

(︁ )︁
X
Ellipsoidal plumb lines
Normal
or normals
gravity
γ
γ
Equipotential surfaces of
γ the normal gravity field
X
Reference ellipsoid,
flattening exaggerated
^ F IGURE 16.5. The normal gravity field of the Earth.
í Õ ! ¤. û
Geopotential field Equipotential surface

Normal potential Line of force (plumb line)
Reference ellipsoid Geoid
F IGURE 16.6. Equipotential surfaces and lines of force of the geopotential and
^ the normal potential.
^ 16.2.3 Distance between equipotential surfaces

Because the normal gravity field is meant to be an idealised represen-
tation of the true gravity field, the equipotential surfaces of both for
the same potential value W = constant and U = constant, for the same
constant value, run close to each other. See figure 16.7. The points P and
luotiviiva Q lie on the same plumb line: P lies on the surface W = WP of the W field,
whereas Q lies on the corresponding surface U = UQ = WP of the U field.
So
WP = UQ
and linearisation with respect to height h yields

∂U ⃓
⃓
UP ≈ UQ + ζ ⃓ = UQ − ζ γP ,
∂h P
in which ζ is the distance separating points P and Q .
Subtraction yields the disturbing potential
def T
TP = WP − UP = UQ − UP = ζ γP =⇒ ζ = γ P . (16.1)
P
5 Equation 16.1 is the famous Bruns5 equation. The quantity ζ is called the
korkeus- height anomaly (of point P ). It is the distance between an equipotential

anomalia
5 Ernst Heinrich Bruns (1848–1919) was a gifted mathematician and astronomer, whose
greatest achievements were related to the theory of the Earth’s gravity field.
í Õ ! ¤. û
Gravity anomalies 16.3
433
Ellipsoidal normal Plumb line
(ϕP , λP ) (Φ P , Λ P )
WP (= UQ ) P
UP ζ
Q
UQ
H
h
H∗
Geoid
W0 (= U0 )
N
Reference ellipsoid
U0
^ F IGURE 16.7. Equipotential surfaces of true and normal gravity field.
surface of the Earth’s gravity field and the corresponding surface of the
normal field, the surface having the same potential. The Bruns equation
links this distance directly to the disturbing potential.
When point P is located on the geoid, we have WP = W0 , and Q is
located on the reference ellipsoid, so UQ = U0 = W0 . In this case we use,
instead of the notation ζ, the designation N, the geoid undulation, the
geoid height, the distance of the geoid from the reference ellipsoid. The
Bruns equation is in this case
T
N = γ0 ,
0
in which the values of both T0 and γ0 are evaluated on the geoid. In

practice, N ≈ ζ, except in the mountains. At sea level, N = ζ exactly.
^ 16.3 Gravity anomalies

In practice, the height h of a point from the reference ellipsoid can be
obtained empirically only using satellite positioning.6 This is why one 6
writes, using h P = HP + N ≈ HP∗ + ζ, see figures 16.7 and 16.8:
∂T ⃓
⃓
− ⃓ = gP − γ(ϕP , hP ),
∂h P
6 Nowadays, thanks to GNSS, it is easier to obtain the quantity

def
δ gP = gP − γ(ϕP , hP ),
which is called the gravity disturbance.
í Õ ! ¤. û
Plumb line (ΦP , ΛP )
gP γP
γQ
^ F IGURE 16.8. True and normal gravity vectors.
in which (Taylor series expansion)

∂γ ⃓
⃓
γ(ϕP , hP ) ≈ γ(ϕP , HP∗ ) + ⃓ · ζ,
∂h P
so, with the Bruns equation 16.1, we obtain
∂T ⃓ ∂γ ⃓ ∂γ ⃓ TP
⃓ ⃓ ⃓
− ⃓ ≈ gP − γ(ϕP , HP∗ ) − ⃓ · ζ = gP − γ(ϕP , HP∗ ) − ⃓ · .
∂h P ∂h P ∂ h P γP
From this
∂T ⃓ ∂γ ⃓ 1
⃓ ⃓
gP − γ(ϕP , HP∗ ) = −
⃓ + T . (16.2)
∂ h P ∂ h P γP P
⃓
This expression is called the gravity anomaly, with the definition
def
∆ g P = gP − γ(ϕP , HP∗ ). (16.3)
The gravity anomaly ∆ g P can be computed if two quantities have been

7 measured:7
◦ the gravity value g P at point P gravimetrically

◦ the height HP ≈ HP∗ of the point from the geoid, or “above sea level”.
Most often — and before the satellite era, always — the height of a
gravity measurement point is determined by reading from a map, from a
ilmapuntari photogrammetric stereo model, by using a barometer, or from traverse
linjavaaitus levelling. In all cases, one obtains just the height HP above sea level. The
precision of measurement of the heights varies from several centimetres
to around a metre. This uncertainty propagates straight into the anomaly
values ∆ g, by the vertical gradient of normal gravity ∂h γ ≈ −0.3 mGal m.
/︁
The gravity anomaly ∆ g is an empirical quantity that can be

calculated from measurements made on the Earth’s surface.
ilma-anomalia The most commonly used gravity anomaly is the free-air anomaly ∆ g,
7 For the latitude ϕP , an approximate value is good enough.
í Õ ! ¤. û
The gravimetric geoid 16.4
435
the definition of which was given as 16.3. Gravity anomaly values vary
in the interval ±100 mGal, more rarely, for example in the mountains,
±200 mGal. Within Finland, the range of variation is ±60 mGal. Gravity
anomalies and their variations from place to place describe the irregular-
ities of the interior mass distribution of the Earth, and are therefore of
geophysical and geological interest.
It may be mentioned as a curiosity, that the Dutch researcher Felix A.
Vening Meinesz8 found, south of the island of Java, at the Java (today 8
Sunda) deep-sea trench, a large deficiency in gravity. We know today that syvänmeren
deep-sea trenches are those places on the Earth’s surface where, as a hauta
part of plate tectonics, the Earth’s oceanic crust “dives” into the Earth’s
mantle where it will be geologically recycled: subduction, see figure 18.7. alityöntö
^ 16.4 The gravimetric geoid

From equation 16.2 above together with definition 16.3 is obtained
∂T ∂γ 1 ∂h γ
∆g = − + T = −∂h T + γ T, (16.4)
∂h ∂h γ
which is called the fundamental equation of physical geodesy, Heiskanen fysikaalisen

and Moritz (1967) equation 2-148. So: geodesian
perusyhtälö
The gravity anomaly ∆ g is a linear combination of disturbing

potential T and its vertical derivative of location ∂h T .
The quantity ∂h γ ≈ −0.3 mGal is the, already mentioned, vertical gradi-

/︁
m
ent of normal gravity.
The geoid, the equipotential surface of the Earth’s gravity field that
describes the figure of the whole field (the “mathematical figure of the
Earth”, Gauss), may be determined by gravimetric means, starting from
the fundamental equation of physical geodesy 16.4. Let us assume the
Earth to be a sphere. Then
GM
γ=
R2
and by differentiation
∂γ ∂γ GM
= = −2 3 ,
∂ h ∂R R
8 Felix Andries Vening Meinesz (1887–1966) was a Dutch geophysicist, geodesist and
gravimetrist. He wrote together with V. A. Heiskanen the textbook The Earth and its
Gravity Field (1958).
í Õ ! ¤. û
g
N
Mass excess
Mass
deficit
−N
F IGURE 16.9. Relationship between variations in the Earth’s gravity and those
^ in geoid height.
from which follows

∂T 2
∆g = − − T,
∂h R
an equation valid on the surface of a spherical Earth.
It is intuitively clear that there is some kind of link between variations
of gravity and variations of the geopotential on the Earth’s surface. Both
are caused by the uneven distribution of masses inside the Earth. As
figure 16.9 shows, the excess masses inside the Earth will cause both
an excess in gravity (the level surfaces of the geopotential will be closer
together) and a rising of the geoid above the surface of the reference
ellipsoid, whereas, on the other hand, mass deficiencies inside the Earth
will lead to both a shortfall in gravity and a depression of the geoid below
the reference ellipsoid.
The relationship between gravity anomalies ∆ g and geoid heights N is
however not simple. The quantities are connected by an integral equa-
9 tion, the Stokes equation. George Gabriel Stokes9 derived the following
í Õ ! ¤. û
The gravimetric geoid 16.4
437
N S (ψ)
Earth’s
centre
∆ g dσ
ψ Moving in-
tegration
point
Earth’s
centre
^ F IGURE 16.10. The geometry of the Stokes integral equation.
classical integral equation10 in 1849: 10
¨
R
N= S (ψ) ∆ g d σ. (16.5)
4πγ
With the equation, we may compute geoid heights from the global field
of gravity anomalies. In the equation, R is the radius of the spherical
Earth, γ mean gravity on the Earth’s surface, and S (ψ) is the Stokes
function, also called the Stokes kernel. It depends only on the geocentric
angular distance ψ between the evaluation point for the geoid height N
and the measurement point for the anomaly ∆ g seen from the centre of
the Earth:
1 /︁ ψ ψ ψ
(︂ )︂
S (ψ) = − 6 sin + 1 − 5 cos ψ − 3 cos ψ ln sin + sin2 .
sin ψ 2 2 2 2
Look more closely at equation 16.5. The quantity d σ is the solid- avaruuskulma
angle element, a surface element on a sphere of unit radius, in spherical
co-ordinates d σ = cos φ dφ dλ.
Open up the equation in the following way:
ˆ 2π ˆ + π 2
/︁
R
S ψ(φ, λ, φ′ , λ′ ) ∆ g(φ′ , λ′ ) cos φ′ dφ′ dλ′ .
(︁ )︁
N (φ, λ) =
4πγ 0 −π 2
/︁
9 George Gabriel Stokes (1819–1903) was a gifted English mathematician, physicist and
geophysicist.
10 Deriving the equation is difficult and uses the fundamental equation of physical
geodesy 16.4 as a boundary condition for solving the Laplace field equation in the space
exterior to the Earth. See Heiskanen and Moritz (1967) chapter 2.
í Õ ! ¤. û
In this, (φ, λ) are the co-ordinates — strictly speaking, the geocentric

latitude and longitude — of the point at which the geoid height N is being
computed. The co-ordinates (φ′ , λ′ ) are again the latitude and longitude of
the point at which the gravity anomaly ∆ g is given. This point traverses
the whole Earth’s surface with the computation of the double integral.
The angle ψ is the angular distance between these two points as seen
11 from the centre of the Earth.11
We see here that computing even a single value N requires ∆ g values

from everywhere on the Earth’s surface, in order to evaluate the above
integral completely. Divide the Earth’s surface into cells, or blocks, of size
1◦ × 1◦ — in total 360 × 180 = 64 800 of them — and calculate the value of
the integral numerically as a sum
360 ∑︂
)︂2 ∑︂ +90
R π
(︂
S ψ(φ, λ, φ′ , λ′ ) ∆ g(φ′ , λ′ ) cos φ′ , (16.6)
(︁ )︁
N (φ, λ) =
4πγ 180
i =1 j =−89
in which λ′ = i ◦ − 0◦. 5 and φ′ = j ◦ − 0◦. 5. Over the whole of the Earth’s

surface, there are 360 × 180 = 64 800 different values of N to be calculated,
if the desired resolution of the geoid model is also 1◦ × 1◦ .
Now one also understands why close international collaboration is so
essential to studying the Earth’s gravity field!
In practice, the greatest influence on the computed geoid height is by
local gravity anomalies, the values of anomalies in the immediate vicinity
of the point of evaluation. The function S (ψ) is for small values of ψ
approximately
2
S (ψ) ≈ ψ .
More remote areas also have an effect, but to take them into account, it
suffices to use a lower-resolution global geopotential model produced by
satellite geodesy.
Globally, the geoid deviates from the reference ellipsoid by about
±100 m. The global mean sea surface in its turn follows the geoid, be-
cause it is an equipotential or equilibrium surface. The mean sea surface
11 The equation for calculating the angular distance is
cos ψ = sin φ sin φ′ + cos φ cos φ′ cos(λ′ − λ),
or more precisely for small angles ψ, the half-angle formula 3.9:
ψ φ′ − φ λ′ − λ
sin2 = sin2 + cos φ cos φ′ sin2 .
2 2 2
í Õ ! ¤. û
The gravity field and heights 16.5
439
−180◦ −90◦ 0◦ 90◦ 180◦
90◦ 90◦
45◦ 45◦
0◦ 0◦
−45◦ −45◦
−90◦ −90◦
◦ ◦ ◦ ◦
−180◦ −90 0 90 180
F IGURE 16.11. The global geoid model EGM2008. Geoid heights reckoned from
the GRS80 reference ellipsoid range from −107 m (blue) to +86 m
^ (red). © 2013 US National Geospatial-Intelligence Agency.
deviates from the geoid by at most ±2 m. The permanent part of the

deviation is called the sea-surface topography, see section 1.6. In addition, meritopografia
there are deviations varying in time, like the phenomenon of the tides
and the deviations caused by winds and air-pressure variations.
^ 16.5 The gravity field and heights
^ 16.5.1 Geopotential and gradient

In figure 16.3 we see how one can depict the forms of the terrain on a
map by means of height contours.
The most natural of all measures of height, the geopotential, is not a
metric height. It is a measure of energy: it describes the level of potential
energy of a test body, a unit mass, in the gravity field of the Earth. This
makes it a geophysically sensible quantity.
The geopotential is connected to gravity in this way, that the gravity
vector, g, is the derivative of place, or gradient, of the geopotential W , see
figure 16.12:
∂W ∂W ∂W
g = ∇W = grad W = i+ j+ k, (16.7)
∂x ∂y ∂z
í Õ ! ¤. û
Equipotential surfaces W =constant
Point P W = W0 − 3∆W
W = W0 − 2∆W
W = W0 − ∆W
∂W W = W0
−
∂z
∂W
k y − ∂y
∂W Gravity vector
j
∂x
∂W ∂W ∂W
O g= i+ j+ k
i ∂x ∂y ∂z
x
F IGURE 16.12. The gravity vector is the gradient of the geopotential, the deriva-
^ tive with respect to the three co-ordinates of location.
{︁ }︁
in which i, j, k are unit vectors in the directions of the axes of the
ortonormaali ( x, y, z) co-ordinate frame, forming an orthonormal basis.
kanta
Because of this, the local gravity vector has the following properties:
◦ It is always perpendicular to equipotential surfaces — also to sea
level!
◦ The closer to each other the equipotential surfaces are, the longer
it is.
The gravity field is a conservative field. This means that, when one
transports a test mass around a closed path, no net work is done. In a
conservative force field, the force vector can always be expressed as the
gradient of a potential, in the way depicted in figure 16.12.
The difference in potential between points A and B is now the same as
the work done by a unit mass moving from point A to point B, and the
following integral applies ( s is the path length along AB):
ˆ B ˆ B ˆ B (︂
dW ∂W ∂W ∂W
)︂
WB − WA = dW = ds = dx + dy + dz =
A A ds A ∂x ∂y ∂z
ˆ B ⟨︁ ˆ B ⟨︁
⟩︁ ⟩︁
= grad W · d x = g · d x . (16.8)
A A
From this it can be seen that the work is the scalar product of the gravity
í Õ ! ¤. û
441
X
ˆ B 4
∑︂
WB − WA = ⟨g · dx⟩ ≈ ⟨g i · ∆x i ⟩.
A i =1
WB
B ∆x4 W = constant
g4 ∆x3 ∆x2
∆x1
g3
g2
A
WA g1
^ F IGURE 16.13. The path integral of work.
vector g and the path vector element along the direction of the path
def
d x = i dx + j dy + k dz.
In practice, often instead of the geopotential itself W , its difference
def
C = − (W − W0 ) with the geopotential of mean sea level (or some other
suitable reference surface) W0 is used. This potential difference, which
grows in the upward direction, is called the geopotential number,12 and 12
the above integral equation becomes
ˆ B ⟨︁ ⟩︁
CB − C A = − g · dx .
A
In the case of a closed path, we have

˛
⟨︁ ⟩︁
g · d x = 0.
Geopotential numbers are calculated from the measurement results of a

levelling extending over the country. All metric heights of terrain points,
like orthometric height, are calculated from the geopotential numbers of
the points.
^ 16.5.2 Geopotential unit, GPU

As we saw in section 16.1, gravity is expressed in the SI unit m s2 . We
/︁
again use the geopotential unit, or GPU, as the measurement unit for the
geopotential. The SI unit of geopotential is m2 s2 : distance × force mass
/︁ /︁
12 German geopotentielle Koten, French cotes géopotentielles.
í Õ ! ¤. û
P
WP
g
∆ H3 ∆ H3′
g ∆ H2 H ∆ H2′
∆ H1 ∆ H1′
O W0
Geoid
F IGURE 16.14. Heights and equipotential surfaces. The stronger the gravity
g = ∥g∥ is — always perpendicular to the equipotential surfaces
^ — the closer together are the equipotential surfaces.
= distance × acceleration, unit m × m s2 . In the gravity field of the Earth

/︁
close to the surface, where the acceleration of gravity is g ≈ 9.8 m s2 , a

/︁
height difference of one metre corresponds to a potential difference of

about 9.8 m2 s2 .
/︁
Define
def
1 GPU = 10 m2 s2 ,
/︁
then a height difference of one metre corresponds to a potential difference

of about 0.98 GPU, and similarly a potential difference of 1 GPU corre-
sponds to a height difference of about 1.02 m.
Thus one can, thanks to the fortuitous circumstance that g is close to
10 m s2 , express geopotential differences in a unit that is slightly more
/︁
intuitive than the corresponding SI unit!
^ 16.5.3 Orthometric heights

Orthometric heights H were introduced already. Let us take a closer look.
In figure 16.14 the orthometric height of point P is H . In this simple
example, it is the sum of three height differences:
H = ∆ H1′ + ∆ H2′ + ∆ H3′ , (16.9)
in which the ∆ H ′i , i = 1, 2, 3 are the separations between the equipotential

surfaces on the plumb line of the point.
Levelling, however, yields the height differences ∆ H1 , ∆ H2 , and ∆ H3
on the terrain, on the Earth’s surface between point and coast. In this
í Õ ! ¤. û
443
case a levelling has been carried out from coastal point O , the height of
which is assumed to be zero. Now,
H ̸= ∆ H1 + ∆ H2 + ∆ H3 !
Height differences obtained from levelling may not just be added

together to get the height of a point.
This tells us that height, though a metric quantity, is not a very nicely
behaved quantity.
For this reason, in scientific work we always use, instead of metric
heights, the already-presented geopotential numbers C = − (W − W0 ) .
We may simply write (note that the units also match):
∆/︁C = g · ∆H
[ m2 s2 ] [ m s2 ] [m]
/︁
in which ∆C is the geopotential difference (between two arbitrary points),

the work that needs to be done in order to transport one unit mass over
the height difference ∆ H between the points.
If in point O it holds that W = W0 , it follows that C O = 0. Then, in the
example case of figure 16.14, the geopotential number of point P is
C = ∆C 1 + ∆C 2 + ∆C 3 = g 1 ∆ H 1 + g 2 ∆ H 2 + g 3 ∆ H 3 , (16.10)
which is computable if, along the levelling line, local gravity g has also
been measured.
However, the following also holds:13 13
C = g′1 ∆ H1′ + g′2 ∆ H2′ + g′3 ∆ H3′ ,
in which the gi′ , i = 1, 2, 3 are gravity values inside the rock, on the plumb
line of point P .
Define the mean gravity along the plumb line by the following equa-
tion:14 14
13 In the general case the equation is

ˆ H
C= g( z) dz,
0
in which z is the arc length measured along the plumb line.
í Õ ! ¤. û
def g′1 ∆ H1′ + g′2 ∆ H2′ + g′3 ∆ H3′ C

g= = .
∆ H1 + ∆ H2 + ∆ H3
′ ′ ′ H
It follows that
C
C = gH ⇐⇒ H = ,
g
the classical definition of orthometric heights. The equation tells that the
amount of work needed to move a unit mass from the geoid to point P, is
force × distance:
work force distance
C = g · H .
We are left with the problem of determining g, the mean of gravity along
the plumb line. Measuring values g′i inside the Earth’s crust is usually
impossible. . . . Therefore, in practice the determination is based on the
value g P measured on the Earth’s surface, by assuming that gravity
grows going downwards, inside the Earth’s crust, according to a certain
15 formula.15 In this way, an approximate value for the orthometric height
likiarvo is obtained, the accuracy of which, at least for the Finnish territory, is
totally adequate.
Orthometric height is but one way of building a metric height system.
There are other ways, like normal height and dynamic height. All are
heights “above sea level”, but the methods of definition and calculation
are slightly different. And all three have their own pluses and minuses.
Orthometric heights are not without their problems. The tunnel net-
work of figure 4.2 does not exist, and measuring gravity inside the rock
— along the local plumb line — is normally not possible. In practice,
orthometric heights are determined with the aid of levelling, starting
from the coast, along the Earth’s surface. If we want to calculate precise
orthometric heights from levelling, we unfortunately need detailed data
on
◦ the density of the rock below the height point
◦ the forms of the terrain around the height point: a terrain model.
So, even if orthometric heights are physically elegant, their precise de-
termination may in practice be troublesome. Scientifically one says that
orthometric heights are not hypothesis-free. The hypotheses required are
14 The general equation is

ˆ H
C 1
g= = g( z) dz.
H H 0
15 For example, Poincaré-Prey reduction, see Heiskanen and Moritz (1967) section 4-2.
í Õ ! ¤. û
445
precisely the density of the Earth’s crust and the terrain’s local forms.
In practical computation, the effect of the terrain is often omitted, and
density values are taken from geological maps. The error thus made is
usually small.
^ 16.5.4 Normal heights

Normal heights H ∗ are, simply stated, orthometric heights computed
from geopotential numbers C as if the true gravity field of the Earth were
a regular mathematical model field based on an ellipsoid of revolution: a
normal gravity field. Therefore, no information is needed related to the
true, complicated gravity field. Normal heights are computed easily and
precisely without any knowledge of local rock density or terrain model.
The equation for normal height is
C
H∗ = ,
γ
in which γ is the mean of normal gravity, calculated again along the
plumb line of the point.16 16
However, unlike orthometric heights, normal heights have no direct

physical interpretation.
On most of the Finnish territory, the differences between orthometric
and normal heights are of the order of millimetres. In the mountains
globally, they can easily be several decimetres.
In many countries — among others Russia, Sweden, and nowadays also
Finland — normal heights are used instead of orthometric heights. Their
precise calculation is easier. Whereas orthometric heights are interpreted
as heights from the geoid, normal heights are reckoned from a similar
surface called the quasi-geoid. On the sea, this surface coincides with
the geoid and thus also with mean sea level, but under the land, and
especially under the mountains, it differs from the geoid. The earlier
proposed “tunnel-network metaphor”, letting in sea water under the
continents, is not appropriate for the quasi-geoid.
^ 16.5.5 Dynamic heights

Dynamic heights are rarely used. They are calculated simply by dividing
the geopotential number C by the normal gravity at latitude 45◦ at zero
height, γ45 , which is a constant:
C
H dyn = γ .
45
16 Now, the averaging is done over the range of normal heights (0, H ∗ ).
í Õ ! ¤. û
^ T ABLE 16.2. Properties of various height types.
Correctness Hypothesis
Height type Metric Energetic freeness Equation
Geopotential number −− + + C
H= C g
/︁
Orthometric + − −
H∗ = C γ
/︁
Normal − − +
H dyn = C γ45
/︁
Dynamic −− + +
^ 16.5.6 Properties of different height types

It is common among all height types, that a metric height is obtained by
dividing the geopotential number by some suitable gravity value: the SI
unit of geopotential numbers is m2 s2 , the unit of acceleration of gravity
/︁
2
and the unit of metric height is indeed ( m s2 )
/︁ /︁ /︁
is m s2 , ( m s2 ) = m, as it
/︁
should be.
Independently of height type, all metric heights are computed from
the energy level of the point, the aforementioned geopotential number
C . The only “heights” that can be measured and computed precisely
are geopotential numbers C = − (W − W0 ). All other heights are derived
quantities. Besides precision, some useful properties are always lost in
their computation, like when projecting a curved surface onto a flat one.
The user of height values desires from practical heights a number of
good properties, familiar from geometric heights within a small area:
oikeellisuus Metric correctness Metric correctness means that, if there are two
points P and Q straight above each other, and the distance between
them is 1 m, then HP − HQ is also precisely 1 m. Only orthometric
heights have this property. The metric correctness of dynamic
heights is especially weak.
The closer to the true mean gravity along the plumb line the
expression in the denominator of the formula is, the better the
metric correctness of the resulting height type will be.
Energetic correctness This means that water always flows “down”
in the sense of the height type in question. Of the three types
mentioned, only dynamic heights are energetically correct — by
virtue of their direct proportionality to geopotential numbers C .
Exact calculability, dependence on uncertain hypotheses
Normal heights and dynamic heights may be precisely cal-
í Õ ! ¤. û
Bouguer anomalies 16.6
447
culated based on theory. However, the normal field, or reference
ellipsoid, chosen for the computations needs to be stated.
Orthometric heights require knowledge of both the true gravity
field and the form and density of the topography. Especially the
assumed density is an uncertain hypothesis. In practice, however,
the uncertainty caused by these factors is fairly small.
^ 16.6 Bouguer anomalies

Earlier we noted that free-air anomalies — equation 16.3 — tell us
something about the interior mass distribution of the Earth. However,
the whole topography underneath and around a point also affects the
free-air anomaly ∆ g of the point. The forms of the topography above sea
level are visible and usually well-known. Therefore it would seem logical
to remove computationally the effect of the terrain forms from the free-air
anomalies, in order to obtain a quantity that tells us only about the mass
distribution of the Earth below sea level.
This is how the Bouguer anomaly is obtained:
∆ g B = ∆ g FA − g top ,
def
in which ∆ g FA = ∆ g is the free-air anomaly, ∆ g B the Bouguer anomaly
and g top the vertical component of the attraction of the topography acting
at the point.
Bouguer anomalies may be calculated precisely or approximately. In
the first case we use a numerical model of the topography, a digital terrain
model (DTM). We also use a density model for the Earth’s crust if one
exists. In the approximate calculation, we only take into account the
effect of the Bouguer plate, as a simple closed formula:
g top = 2πGρ d,
in which
g top attraction of the plate, only in the vertical direction
G Newton’s universal gravitational constant, see section 1.2
ρ density of the matter of the plate
d thickness of the plate, assumed to be equal to the local height H of
the terrain.
í Õ ! ¤. û
20◦ 24◦ 28◦ 32◦ 20◦ 24◦ 28◦ 32◦
64◦ 64◦ 64◦
62◦ 62◦ 62◦
60◦ 60◦ 60◦
20◦ 24◦ 28◦ 32◦ 20◦ 24◦ 28◦ 32◦
F IGURE 16.15. Free-air and Bouguer anomalies for Southern Finland com-
puted from the EGM2008 geopotential model. Data © Bureau
Gravimétrique International (BGI) / International Association
^ of Geodesy.
The plate is assumed to be flat and to extend to infinity in all directions.

If the density is ρ = 2.67 g cm3 = 2670 kg m3 , we obtain
/︁ /︁
g top = 0.1119 d,
in which d is in metres and g top in milligals. The milligal unit was

explained at the beginning of the chapter.
^ 16.7 Astronomical position determination

The local plumb line or direction of the local gravity vector stands per-
pendicular upon equipotential surfaces. Determining the direction of
the plumb line in an absolute sense has been possible by traditional
astronomical means.
One speaks of astronomical position determination because the first
practical application of the method was determining an unknown position,
especially at sea. Later, the method was used to study geophysically
interesting variations (deviations) of the direction of the plumb line on
land. The already mentioned Pierre Bouguer noticed in South America,
like George Everest in India, that in the vicinities of mountain ranges
the plumb line is deflected towards the mountain range, and interpreted
this correctly as being caused by the mountains’ own gravitation.
í Õ ! ¤. û
Astronomical position determination 16.7
449
Plumb-line deviations
Mountain Geoid
Earth’s crust
“Root”
Earth’s mantle
^ F IGURE 16.16. The root of a mountain range and its effect on the plumb line.
The attempt to estimate the effect of the mass of the mountains, how-
ever, produced a result that was much larger than the actually observed
plumb-line deviations. The reason for this is today known to be isostatic luotiviivan
compensation: under the mountains there is a root consisting of lighter poikkeama
rock that keeps the mountain “afloat” on the plastically deforming Earth’s
mantle.
Various instruments are used for astronomical position determination,
such as the meridian circle, the astrolabe, and the zenith tube.
The optical axis of an astrolabe points always upwards by a fixed angle.
The optical axis of a zenith tube again always points upwards vertically, to
the zenith, under an elevation angle η = 90◦ . Therefore, by observing the korkeuskulma
passage through the zenith of stars of which the declination δ is known,
the astronomical latitude Φ of the location is obtained by Φ = δ. The
time of transit through the meridian is measured at the same time, from meridiaanin
which is obtained the astronomical longitude Λ — because the zenith läpikulku
direction lies in the plane of the meridian. A zenith tube is in a way both
a meridian circle and an astrolabe.
In preparation for observations, a star programme is drafted, a list
of stars that will transit the meridian, a certain elevation circle,17 or korkeuskulma-
the zenith. In case of the astrolabe, one should take care that the stars piiri
17
are distributed evenly around the whole horizon. In this way, a precise
17 In Arabic almucantar.
í Õ ! ¤. û
1
2η
^ F IGURE 16.17. A levelling instrument converted to astrolabe.
determination of both Φ and Λ is achieved, and the impact of atmospheric

refraction minimised.
An astrolabe may be built easily from a levelling instrument, by adding
a sixty-degree angle prism in front of the objective; see figure 16.17. The
most precise (Danjon) astrolabes again use a mercury mirror together
with a sixty-degree prism.
^ 16.8 Measuring the gravity gradient

We have several times mentioned the potential of gravity, which is a
measure of the potential energy (energy content of location) of a test mass
inside the gravity field. The acceleration vector of gravity, or free fall,
is defined as the gradient of this geopotential W, its rate of change with
place:
g = grad W = ∇W =
∂W ∂W ∂W
= gx i + g y j + gz k = i+ j+ k = ∂ x W i + ∂ y W j + ∂ z W k,
∂x ∂y ∂z
where the grad operator is
∂· ∂· ∂·
grad · = ∇· = i + j + k = ∂ x (·) i + ∂ y (·) j + ∂ z (·) k.
∂x ∂y ∂z
{︁ }︁
Here i, j, k is an orthogonal triad of unit vectors, or orthonormal basis,
oriented along the three axes of the ( x, y, z) co-ordinate frame.
The acceleration vector of gravity is thus the gradient of the geopoten-
tial. This vectorial quantity is location-dependent. We know that gravity
grows going downwards, at least in free air. Going up, gravity diminishes
by some 0.3 mGal for every metre in height.
í Õ ! ¤. û
Measuring the gravity gradient 16.8
451
F IGURE 16.18. The gravitational-gradient or tidal force field: gravitation varies

with place. The lower diagram gives the visual ellipsoid of the
gravitational gradient tensor: it depicts the force field inside a
^ freely falling object, for example a satellite.
Gravity as a vector varies in a more complicated way in the vicinity of

masses, if one has the use of sufficiently accurate measurement devices.
We speak of the gravity-gradient tensor, or Eötvös tensor. Written in both
the traditional and the more compact Euler notation:
∂2 ∂2 ∂2
⎡ ⎤
∂ x∂ y ∂ x∂ z
⎡ ⎤
⎢ ∂ x2 ⎥ ∂ xx ∂ x y ∂ xz
def
⎢ ∂2 ∂2 ∂2 ⎥
M=⎢ ⎣ ∂ yx ∂ yy ∂ yz ⎦ W.
⎥W = ⎢ ⎥
⎢ ∂ y∂ x ∂ y2 ∂ y∂ z ⎥
∂2 ∂2 ∂2 ∂ zx ∂z y ∂ zz
⎣ ⎦
∂ z∂ x ∂ z∂ y ∂ z2
In a topocentric co-ordinate frame, where x points north, y east and z up,

this matrix has the approximate form
⎡ ⎤
−0.15
M≈⎣ m, (16.11)
⎥ mGal/︁
−0.15
⎢
⎦
0. 3
in which
∂2 ∂
∂ zz W = W= gz = 0.3 mGal m
/︁
∂z 2 ∂z
is truly the “free-air” standard gravity gradient, the vertical gradient of
normal gravity: if we write according to Newton (note that the vector g
í Õ ! ¤. û
is pointing down whereas the z co-ordinate grows going up, that is the
reason for the minus sign):
GM
gz = − ,
(R + z)2
we obtain by taking the derivative
∂ GM ∂ (R + z) 2 gz
≈ 3 · 10−6 m s2 m = 0.3 mGal m .
/︁ /︁
gz = 2 =−
/︁
∂z (R + z)3 ∂z (R + z)
The quantities ∂ xx W and ∂ yy W again describe the curvatures of the
equipotential surfaces in the x and y directions, in the following, geomet-
rically intuitive way:
∂2 g ∂2 g
∂ xx W = W = −r , ∂ yy W = W = −r ,
∂x 2 x ∂y2 y
in which rx and r y are the radii of curvature in the x and y directions.

Substituting rx = r y = R ≈ 6378 km (try!) gives
∂ xx W = ∂ yy W = −1.5 · 10−6 m
/︁ /︁
2
= −0.15 mGal m.
/︁
s m
The already mentioned Hungarian researcher Baron Loránd Eötvös did

many ingenious experiments (Eötvös, 1998) in order to measure the
torsiovaaka components of the gravity-gradient or Eötvös tensor with the torsion
balance he built. The method continues to be in use in geophysical
research, because the gravity gradient as an observable is very sensitive
to local variations in the density of the Earth’s crust.
In the general case, we can evaluate the Eötvös tensor by performing
the partial differentiations above. We do so for a central force field:
GM
W= r ,
⎷
where r = X 2 + Y 2 + Z 2 is the distance from the geocentre. The co-
ordinates ( X , Y , Z ) are now geocentric. We obtain
⎡ ⎤
∂X
⎥ GM
[︂ ]︂
M= ∂X ∂Y ∂Z ⎣ ∂Y ⎦ r =
⎢
∂Z
⎛ ⎡ ⎤⎞
X
⎜ GM ⎢
[︂ ]︂
= ∂X ∂Y ∂Z ⎝− 3 ⎣ Y ⎦⎠ =
⎥⎟
r
Z
⎡ ⎤
3 X 2 − r2 3X Y 3X Z
GM ⎢
= 3Y X 3Y 2 − r 2 3YZ ⎦.
⎥
r5
⎣
3Z X 3 ZY 3 Z 2 − r2
í Õ ! ¤. û
Measuring the gravity gradient 16.8
453
The Eötvös tensor is the matrix of partial derivatives of the gravity
acceleration vector ⎡ ⎤
X
GM ⎢
g=− ⎣ Y ⎦
⎥
r3
Z
with respect to place, as can be seen in the above equation:
⎛ ⎡ ⎤⎞
X
⎜ GM ⎢
[︂ ]︂ [︂ ]︂
M= ∂X ∂Y ∂Z ⎝− 3 ⎣ Y ⎦⎠ = ∂X ∂Y ∂Z g.
⎥⎟
r
Z
In honour of Eötvös we use as the unit of gravity gradient the eötvös,

symbol E:
1 E = 10−9 m = 10−4 mGal
/︁ /︁
m.
2
/︁
s m
The tensor 16.11 given above close to the Earth’s surface is now, again in
local topocentric ( x, y, z) co-ordinates:
⎡ ⎤
−1500
M≈⎣ −1500 ⎦ E.
⎢ ⎥
3000
In every case
∂2 W ∂2 W ∂2 W
+ + = ∂ xx W + ∂ yy W + ∂ zz W = 0.
∂ x2 ∂ y2 ∂ z2
This condition, the Laplace field equation,18 applies more generally for a 18
gravity potential in vacuum.19 19
The gravitational-gradient fields of the Sun and Moon are known on

the Earth’s surface as the tidal force field, which causes the phenomenon vuoroveden
of the ocean tides, with a dominant period, due to the Earth’s rotation, of voimakenttä
twice a day.
18 Pierre-Simon Laplace (1749–1827) was a French mathematician, physicist and as-

tronomer, one of the 72 names on the Eiffel Tower, and surely the most gifted of them
all (Eiffel Tower, 72 names).
19 Here we have not considered, besides the atmosphere, the centrifugal force of the
Earth’s rotation, which causes the Laplace field equation to not be valid in a co-rotating
co-ordinate frame, even in a vacuum. The correction term can, however, be precisely
calculated. And for the disturbing potential T = W − U the Laplace field equation is
again exactly true.
í Õ ! ¤. û
1. What are the differences between an absolute and a relative
gravimeter?
2. By how many milligals is gravity less at aircraft height (10 km) than
on the ground below? How much is the difference in percentage
points of total gravity?
3. The same questions for the International Space Station, mean
height 400 km.
4. In videos from the International Space Station, the people there
appear to be weightless, so there is no gravity. Yet the station orbits
at a height of some 400 km, where the Earth’s gravitation is still
strong. How is this possible?
5. What is the geopotential, and what is its relationship with gravity?
6. How is a normal gravity field defined?
7. In what situation would you use Bouguer anomalies rather than
free-air anomalies?
8. What are deviations of the plumb line, and how are they deter-
mined?
9. What causes the deviations of the plumb line close to mountain
ranges, and why are they smaller than theoretically expected?
10. What quantity does the unit eötvös (E) describe?
^ Exercise 16–1: Gravimetric geoid computation

Consider the computational arrangement of section 16.4, specifically
calculation equation 16.6:
360 ∑︂
)︂2 ∑︂ +90
R π
(︂
S ψ(φ, λ, φ′ , λ′ ) ∆ g(φ′ , λ′ ) cos φ′ , (16.6)
(︁ )︁
N (φ, λ) =
4πγ 180
i =1 j =−89
in which λ′ = i ◦ − 0◦. 5 and φ′ = j ◦ − 0◦. 5.

1. How many elementary operations — a multiplication, possibly fol-
lowed by an addition or subtraction — would you have to perform
in total to calculate a global geoid model at resolution 5◦ × 5◦ using
gravity anomaly data given at 1◦ × 1◦ resolution?
You may assume the Stokes function S (ψ), the cosines of latitude
φ′ and the anomaly values to have been precomputed. A rough
í Õ ! ¤. û
Exercise 16–1: Gravimetric geoid computation
455
estimate, say within ten percent, suffices for estimating the length
in elementary operations of a computer run doing this computation.
2. Interchange the order of the two loops, and move cos φ′ to outside
the inner loop. How does your answer change?
í Õ ! ¤. û
^ Space geodesy
17
1462 Zamenhof, provisional designation 1938 CA, is a
carbonaceous Themistian asteroid from the outer regions of
the asteroid belt, approximately 27 kilometers in diameter.
It was discovered on 6 February 1938, by Finnish
astronomer Yrjö Väisälä at the Iso-Heikkilä Observatory in
Finland. The asteroid was named after L. L. Zamenhof, the
creator of Esperanto. It is a recognized
Zamenhof–Esperanto object.
Wikipedia, 1462 Zamenhof
^ 17.1 Earth rotation, orbital motion, sidereal time

The Earth orbits the Sun in 365.25 days. She also rotates around her own
axis in a day. Figure 17.1 shows both the physical situation, the orbit of
planet Earth around the Sun, and the apparent situation as seen from
Earth, the annual path of the Sun along the zodiac or ecliptic. eläinrata
During a civil day, 24 hours, the Earth turns once around her axis with siviili-
respect to the mean Sun. With respect to the stars, however, the rotation vuorokausi
period is a little shorter: when the Earth has, in a year, turned 365.25
times around her axis with respect to the Sun, she has, with respect to
the stars, completed 366.25 rotations. Thus, the duration of one rotation
is
365.25 h
24 = 23h 56m 4s .
366.25
The name of this period is a sidereal day. tähtivuorokausi
The rotation of the Earth with respect to the stars — or, equivalently,
the apparent rotation of the stars with respect to the Earth — is measured
by an angle called sidereal time. It is calculated from clock time and tähtiaika
– 457 –
458 17 S PACE GEODESY
Celestial
North Pole
Autumnal equinox
True Earth orbit Spring

Summer
Winter
Winter
Sum-
sol-
mer
stice
solstice
Autumn
X X
Apparent
solar
path
Ecliptic or zodiac
Celestial equator
Vernal equinox (“Aries”)
Hic sunt dracones
F IGURE 17.1. The orbit of the Earth around the Sun, and the apparent path
of the Sun across the celestial sphere. Season names are boreal,
^ referring to the Northern Hemisphere.
calendar date using tables drafted for this purpose. In astronomical

tähtikello observatories, sidereal clocks are also used that show sidereal time, and
run about 1 365.25 part, or 0.27 %, faster than ordinary clocks.
/︁
Due to the annual motion of the Earth, the constellations that are
visible in the evening hours shift slowly forwards along with the season.
Every season has its own distinctive constellations visible in the evening
twilight: in winter, Orion, the stars Sirius and Procyon, in summer, the
constellations of the Lyre, the Swan and the Eagle.
Seen from Earth, the Sun travels along a yearly path, the zodiac (“ring
1 of beasts”, Wikipedia, Zodiac, Name) or ecliptic.1 At the beginning of
1 The name “ecliptic” originates from the fact that this is where solar and lunar eclipses
happen. Of course, because both types of eclipse require the Sun, Moon and Earth to be
í Õ ! ¤. û
Earth rotation, orbital motion, sidereal time 17.1
459
41
α
β
Ecl
Aries iptic
η Pisces
α
γ
Equator
Time of Hipparchus
Equinox today
F IGURE 17.2. The vernal equinox and its movement, the precession of the
^ equinoxes.
spring, the Sun moves from the Southern Hemisphere to the Northern
one, at a point called the vernal or spring equinox. In the time of the kevät-
ancient Greeks, this point was in the constellation of the Ram, hence the päiväntasaus
traditional name “First Point of Aries” — though, due to precession, it is
nowadays located in the constellation of Pisces, the Fishes. In the same
way, back then, the Sun was at the time of the summer solstice in the
constellation of Cancer, the Crab, and at the time of the winter solstice in
the constellation of Capricorn (a mythological goat-like creature), and the
constellations gave their names to the tropics. Due to precession, today’s kääntöpiiri
solstices happen in the constellations of Taurus (the Bull) and Sagittarius
(the Archer) . . .
The absolute orientation of the whole globe with respect to the stars
is described by Greenwich sidereal time (GAST, Greenwich Apparent Greenwichin
Sidereal Time). Local sidereal time (LAST, Local Apparent Sidereal Time) näennäinen
tähtiaika
is obtained using astronomical longitude:
LAST = GAST + Λ,
in which Λ is the longitude of the site reckoned east, and of course

converted into time units (15◦ → 1h , 1◦ → 4m , etc.).
on the same straight line.
í Õ ! ¤. û
Apparent rotation X
of the celestial sphere
Upper culmination
Zenith
North pole Meridian
h
Equator
δ
Apparent
diurnal
motion
Observer
E
N
h
S
W
Local horizon
^ F IGURE 17.3. Hour angle h and declination δ on the celestial sphere.
^ 17.2 Heavenly and Earthly co-ordinates

rektaskensio Celestial co-ordinates are right ascension α and declination δ. The right
ascension is a longitude, measured however eastward from the vernal
equinox, the right ascension of which is thus 0.
In the local sky, however, we use the co-ordinates hour angle h and
declination δ. See figure 17.3. The pair ( h, δ) can be directly computed
korkeuskulma from the azimuth and elevation, if the local astronomical latitude Φ is
known.
During a sidereal day, a star moves apparently from east to west,
and passes through the meridian plane two times: the upper and lower
culmination. The lower culmination remains unseen if the declination
of the star is too small compared to the local latitude. Of the southern
constellations, even the upper culminations remain unseen by us!
The following relationship exists between sidereal time θ (= LAST), the
hour angle h, and the right ascension α of a star:
h = θ − α.
í Õ ! ¤. û
Heavenly and Earthly co-ordinates 17.2
461
If we write for Greenwich sidereal time (GAST), θ0 and for the eastern Greenwichin
astronomical longitude of the observation site, Λ, we obtain tähtiaika
h = θ0 + Λ − α.
If, of the four quantities, three are given, the fourth can be calculated.
We speak of
1. time determination, if the unknown is θ0 . Back in time this was a
service of astronomy to society, the maintenance of civil time. Today,
with clocks so much more precise, it is about monitoring the Earth’s
rotation.
2. longitude determination of the site, if the unknown is Λ. This was
critical for navigation at sea (Sobel, 1998).
3. determination of the right ascension of a star, if the unknown is α.
This is how star catalogues are constructed.
In these cases, the measurement is generally done at the moment when
h = 0: the meridian transit. Cases 2 and 3 require the use of a precise meridiaanin
clock; case 1 requires knowledge of one’s own longitude and the right läpikulku
ascension of the star used.
The meridian circle is often used to precisely time the transit of a star
through the meridian. A meridian circle is a telescope having only a
horizontal axis, which has been built fixed into the east–west direction,
in such a way that the sight axis of the telescope will always be in the tähtäysakseli
local meridian plane. In the eyepiece, one observes how the star moves okulaari
underneath the crosshair. The precise moment is recorded electrically hiusviiva
together with time signals.
A graduated circle is attached to the axis, which allows the reading jakokehä
of the elevation or height angle η at the moment of transit. From this,
the declination δ of the star can be calculated: between it, the latitude of
the site Φ, and the elevation angle η exists a relationship (for the case of
upper culmination on the Northern Hemisphere)
η + ∆η refr = (90◦ − Φ) + δ,
in which ∆η refr is the correction for atmospheric refraction. The refraction

correction must be made carefully based on local measurements of air
pressure, temperature and humidity aimed at determining the index of taitekerroin
refraction of the local air.
By combining methods 1–3 above one can build catalogues of both the
places of stars (α, δ) and of observation stations (Φ, Λ), and at the same
í Õ ! ¤. û
Balloon
*
* Background
Flash stars
*
Schmidt-Väisälä
telescope (camera)
Flash light
*
Photographic plate
* * δ
* X
Image
* measurement
F IGURE 17.4. Yrjö Väisälä’s stellar triangulation. Equipment used and princi-
ple of photography and direction determination. (a) Equipment
^ used in the measurement.
time follow the progress of sidereal time — in other words, the rotation of
the Earth — θ0 . All this must be done in a consistent way.
Sobel (1998) is a fine book on the role of time in position determination
at sea.
^ 17.3 Väisälä’s stellar triangulation

This method was invented by Yrjö Väisälä during the Second World War,
ilmatorjunta- while watching anti-aircraft shells exploding over Turku. In his 1946
kranaatti article “Maan toinen kuu” (“A second moon of the Earth”, Väisälä, 1946)
he described how targets high up in the sky could be photographed from

different places on the Earth’s surface, and thus a geodetic network built
even connecting points between which there is no direct line of sight.
The method uses meteorological sounding balloons carrying powerful
flashtubes. The flash train is photographed against the stellar back-
ground, the images are developed and the places of the flashes among the
stars are measured and calculated. Thus the momentaneous direction
vector between photography site and balloon is obtained, in astronomical
co-ordinates (Kakkuri, 1973). This balloon method has been tried in a
production setting in Finland and Hungary (Czobor and Németh, 1981).
The side lengths of the network were several hundreds of kilometres.
í Õ ! ¤. û
Väisälä’s stellar triangulation 17.3
463
*
* * * * *
*
*
**
S * * T
δ
*
δ
α
Q
α
P
F IGURE 17.5. Yrjö Väisälä’s stellar triangulation. Equipment used and prin-
ciple of photography and direction determination. (b) Measure-
^ ment geometry.
In short, from two observation sites A and B on the Earth’s surface two
high signals S and T visible in the sky are photographed against korkea tähys
the stellar background. Powerful catadioptric Schmidt2 -Väisälä 2
telescopes are used as cameras (Wikipedia, Schmidt camera). A
glass image plate covered by light-sensitive emulsion is placed into
the image plane of every camera.
After development, the image plates3 are measured to extract, in an 3
2 Bernhard Schmidt (1879–1935) was a telescope construction and design genius born in
Estonia.
3 Back then, glass plates were used instead of films for their geometric stability. The
locations of the flashes are measured in plate co-ordinates, as are the locations of the
fixed stars captured on the plate. Using the known places of the fixed stars, the places
of the flashes are solved in celestial co-ordinates (α, δ).
í Õ ! ¤. û
accurate measurement device, the locations, in “plate co-ordinates”, of

both the signals — the flashes — and chosen background stars. Because
the places of fixed stars in terms of celestial co-ordinates, right ascension
(α) and declination (δ), are already known, we may derive transformation
tähys equations for all plates, which will also yield the places of the signals or
flashes in the sky (α, δ), seen from both observation sites.
These “places” on the celestial sphere are in reality directions in three-
dimensional space. From the directions are formed planes P = S AB and
Q = T AB. The intersection line of the planes is AB. Thus the direction
vector from A to B has been obtained in three-dimensional space.
isoympyrä In fact, the plane P is, on the celestial sphere, the same as the great
circle through the image points of flash S taken in points A and B, and Q
is the great circle through the images of flash T . The intersection point
tähtikolmio- of the great circles is the direction in space of connecting line AB: stellar
mittaus triangulation is direction measurement using heavenly auxiliary points.
When a sufficient number of directions between ground stations has

verkkotasoitus been collected, a network adjustment can be carried out. Transforming the
original directions from the celestial co-ordinate frame to one co-rotating
with the Earth, using a model of the Earth’s rotation (“Greenwich sidereal
time”) requires that the flashes are accurately time-tagged.
Väisälä’s idea was, as may already be inferred from the name of his ar-
ticle, to use, instead of stratospheric balloons, artificial satellites orbiting
the Earth. This method has also been tried: the active geodetic satellite
ANNA 1B (“Army, Navy, NASA, Air Force”) from 1962 was equipped with
powerful flashtubes. On the other hand, the passive satellite PAGEOS
(Passive Geodetic Earth Orbiting Satellite) from 1966 was a balloon, over
30 metres in diameter, made of very thin, aluminised Mylar® film, in a
4000 km high orbit. The satellite was clearly visible to the naked eye,
until it disintegrated in the 1970s, undoubtedly due to the corrosive effect
4 of the Sun’s ultraviolet rays.4
5 Using a global network of fifteen massive Baker–Nunn5 cameras, it
4 DuPont write in the document Mylar® polyester film — Safe Handling: “Mylar® is not
recommended for applications requiring prolonged exposure to direct sunlight due to
degradation when exposed to ultraviolet rays [. . . ]” (DuPont Teijin Films).
5 The Baker–Nunn camera was a variant of the Schmidt camera, which was optimised
for observing satellites. It used 55 mm broad Cinemascope motion-picture film, and it
had a three-axis mount capable of fast tracking. The aperture was 50 cm, the weight
of the whole instrument, 3.5 tonnes. The inventors were Harvard astronomer-optician
í Õ ! ¤. û
Variations in the Earth’s rotation 17.4
465
(a) (b)
PAGEOS satellite, test inflation (NASA) Baker–Nunn camera (NASA)
^ F IGURE 17.6. Satellite geodesy from the photographic archives of NASA.
became thus possible to build the first intercontinental geodetic triangu- kolmiomittaus-
lation networks, which however, with the advent of so much more precise verkko
methods (GNSS, VLBI, satellite laser-ranging), have honourably faded into laser-
scientific history. etäisyysmittaus
^ 17.4 Variations in the Earth’s rotation

The rotation of the Earth varies in both speed and direction.
The momentaneous rotation axis of the Earth, or equivalently, the pole,
moves with respect to the solid Earth’s crust, a movement called polar
motion. It includes two circular motions:
◦ An annual motion, period 365 days. This is a forced motion, mainly
James Baker (1914–2005) and mechanical engineer Joseph Nunn (1905–1968). The
satellite tracking network consisted of fifteen stations. Boller and Chivens, Baker Nunn,
image archive, built twelve cameras for the Smithsonian Astrophysical Observatory
and three for the US Air Force.
At least three Baker–Nunn cameras are or were spending their golden years as astro-
nomical sky-survey cameras (Carter et al., 1992; Fors et al., 2013; Mondal et al., 2009;
see also BakerNunn.org).
í Õ ! ¤. û
2000
1995
1990
ich 1985
Greenw
1980
0.8 1975
0.6
X 1970
0.4
y 0.2 −0.5
0 0 X
−0.2 0.5
1 x
F IGURE 17.7. Polar motion for the period 1970–2000, unit second of arc. Source:
IAG International Earth Rotation and Reference Systems Service
^ IERS.
caused by the atmosphere.

6 ◦ The “Chandler6 wobble”, period about 435 days.
Chandlerin ◦ In addition to these, there has been observed a slow drift of the pole,
huojunta
which is related to changes in mass distribution of the solid Earth,
like the post-glacial land uplift.
The amplitude or radius of both circular motions is about 0′′.1–0′′.2, on
the Earth’s surface about 3–6 m.
The Chandler wobble is theoretically understood: it is free nutation,
already predicted for a flattened, rigid Earth by Leonhard Euler. The
motion is damped, due to the Earth not being rigid but (elastically and
plastically) deformable. The fact that the wobble continues requires an
explanation. The explanation is that variations in the pressure exerted
by oceans and atmosphere are its driving force (Gross, 2000).
vuorokauden The rate of the Earth’s rotation (length of day, LoD) also varies. This
pituus phenomenon is monitored in a similar way to polar motion. It is closely
ilmakehän associated with variations in the angular momentum of the Earth’s at-
pyörähdys- mosphere, a quantity that numerical weather models (NWP, numerical
momentti
weather prediction) can calculate very well.
Monitoring of polar motion and length of day, historically and today:
◦ Measuring the variations in latitude by astronomical means. Use
of this method started in 1899 (the International Latitude Service,
6 Seth Carlo Chandler, Jr. (1846–1913) was an American astronomer.
í Õ ! ¤. û
Variations in the Earth’s rotation 17.4
467
Reference pole (CIO) Real (instantaneous) pole
90◦ − Φ A
90◦ − ΦB
D
Plumb line B
F Polar motion
(exaggerated)
A
90◦ − ΦC
Earth’s centre of mass
E
C
F IGURE 17.8. Polar motion causes variations in the latitudes of observation sta-
^ tions, with the help of which the phenomenon may be monitored.
from 1987 the International Polar Motion Service), using six “In-
ternational Latitude Observatories” (Misuzawa, Japan; Charjui,
Turkestan, later Kitab, Uzbekistan; Carloforte, Italy; Gaithersburg,
Ukiah, Cincinnati, USA) which are located on different continents
at the same latitude, 39◦ 08′ around the globe. As the instrument
used at the observatories was a zenith tube, this allowed them to all
use the same selection of stars, enabling a uniform data analysis.
Monitoring of the rate of rotation of the Earth has also been tra-
ditionally done by astronomical means. Civil time was originally
defined by means of the Earth’s rotation: the Earth herself was
used as a clock. The instrument used was the meridian circle (for
example in Greenwich, but also at the Helsinki observatory), with
which the transit of a star through the meridian was observed. Back
then, timekeeping was of vital importance to navigation at sea.
◦ Using positioning satellites, first the Transit system, nowadays
í Õ ! ¤. û
GNSS. The GNSS method is today more accurate than the astronomi-
cal method.
pitkäkanta- ◦ With VLBI, very long baseline interferometry, which provides the
interferometria vectors between observation stations in an inertial or celestial sys-
tem as a function of time. From this, one may compute the momen-
taneous direction of the Earth’s rotation axis in the same system,
and even the momentaneous orientation of the whole Earth. Ac-
curacy is even better than for the GNSS method, a fraction of a
millisecond of arc.
The origin of calculation of the polar motion, CIO, Conventional Inter-
national Origin, was initially the mean place of the pole over the years
1900–1905. The currently used origin of calculation is close to this.
Polar motion and length-of-day variations, together with nutation and
precession, are called Earth orientation parameters, EOP. Their monitor-
ing and publication is a task of the IERS (International Earth Rotation
and Reference Systems Service). In precise geodetic work, the EOP must
be taken into account! The correction information needed can be found
on the Internet.
It was not until the 1930s that it was noticed that the rotation of the
Earth is slightly irregular, and a more regular time scale was looked for.
The first attempt was ET, ephemeris time, based on the orbital motion
of the planets, especially the Moon. When sufficiently accurate atomic
clocks appeared, they were taken into use, creating atomic time (TAI),
also suited for demanding scientific use in which a time scale is needed
that is strictly uniform.
Today’s civil time is UTC, Universal Time Co-ordinated. One of its
design objectives is to follow everywhere on Earth the cycle of daylight
7 dominating daily life. Therefore it must follow with sufficient precision7
the variations in the rotation of the Earth. UTC, which is based on TAI,
follows the variation in the Earth’s rotation with an error of at most
0.9 s. To this end, twice a year, at the end of December and at the end
karkaussekunti of June, UTC executes, if needed, a leap second or seconds (Wikipedia,
Leap second). The difference UTC – TAI, an integer number of seconds, is
tabulated in the almanacs.
7 In recent years there has been a public discussion in the time field on whether the leap
seconds are worth the trouble they are causing, when the time of the time-zone system
tied to UTC agrees with local solar time anyway to no better than half an hour or so.
í Õ ! ¤. û
Precession and nutation of the Earth 17.5
469
GPS time differs from both UTC and TAI. Like TAI, it is uniform and does
not ever execute leap seconds. In 1980, when GPS timekeeping started,
GPS time was identical with UTC. For this reason
GPS = TAI − 19 s
and the difference GPS – UTC is a varying number of seconds, included in

the GPS navigation message.
^ 17.5 Precession and nutation of the Earth

The attraction of the Sun and Moon, together with the flattening of the
Earth, causes a slow motion of the Earth’s rotation axis in space. This
conical motion is called precession. The word “precession” originally
denoted the earlier and earlier occurrence of the equinoxes, the times päiväntasaus
of the year when day and night are equally long. This phenomenon,
observed first by Hipparchus, brings on, over the centuries, the vernal
and autumnal equinox earlier and earlier:8 it shifts the place of the Sun 8
in the sky amidst the stars at the moment of equinox backwards along
the zodiac.9 See figure 17.2. 9
In fact, the rotation axis of the Earth turns in some 25 800 years around
an axis that stands perpendicular on the plane of the Earth’s orbit. This
plane, the apparent plane of the Sun’s orbit as seen from Earth, is also
called the zodiac or ecliptic, see above.
Precession is not the same as polar motion. Polar motion is the motion
of the Earth’s rotation axis with respect to the solid Earth. Its magnitude
is under a second of arc, on the Earth’s surface a few metres.
Besides precession, the Earth’s rotation axis also goes through a small
periodic motion called nutation. Its main period is 18 years and it is
caused by periodic changes in the orbit of the Moon which take effect
through the Moon’s attraction. The phenomenon can be precisely com-
puted and is found in almanacs, just like the precession.
The precession makes the rotation axis of the Earth — and thus the
place of the celestial pole among the stars — slowly describe a large
8 Though not earlier on the civil calendar, based as that is on the rhythm of the seasons!
9 Since then, the word has been taken into use to denote the turning motion of the axes
of other spinning objects, like spinning tops, gyroscopes, and atomic nuclei in a magnetic
field. The latter phenomenon, Larmor precession, is fundamental to magnetic resonance
imaging (MRI).
í Õ ! ¤. û
α Cyg (Deneb)
Lyra
α Lyr (Vega) Ursa

α UMi
Minor
(Polaris)
Draco
α UMi
Precession
25 800 a (Polaris)
Ecliptic
pole Nutation (18 a)
Ecliptic plane
24 h
Moon
Sun
F IGURE 17.9. The Earth’s precession. On the left is shown how, due to preces-
sion, the celestial pole describes a circle among the stars over a
^ period of some 25 800 years.
circle. The nutation again causes small “wobbles” on this regular motion.
Whereas today, the North Star (Polaris, α UMi) is close to the celestial
pole, it will in the remote future be Vega (α Lyr), as it also was in the
remote past.
Precession and nutation are, unlike polar motion, motions of the Earth’s
rotation axis with respect to the stars.
^ 17.6 Space weather

The Sun is a star, which produces in its interior nuclear power by “burn-
ing” hydrogen to helium. In the centre of the Sun, the temperature is
about 15 million kelvins. The thermal energy in the solar core travels
10 very slowly to the surface through many gas layers, first as radiation,10
closer to the surface carried by convection currents, and finally again

leaving the visible surface or photosphere as radiation, light, of which a
vanishingly small fraction arrives also on Earth.
The outer layer of the Sun, of a thickness of about 27 % of the solar
radius, is in a permanent state of convective “bubbling”. Because of this,
10 Thermal radiation, in the X-ray range due to the high temperatures.
í Õ ! ¤. û
Space weather 17.6
471
F IGURE 17.10. The corona of the Sun during a total eclipse (image processed),
Adler (2017). The patterns formed by the magnetic field lines in
^ the corona and the solar wind escaping to space are visible.
energy is continuously being sent up to the Sun’s highest layer, made up

of thin plasma, the corona, heating this layer to millions of degrees.11 11
The precise form of this energy stream — acoustic, magnetic, . . . — is the
subject of active investigation.
The corona, which is visible only during a total Solar eclipse, is so hot
that it leaks continuously to space. This plasma or particle flow can be
observed near the Earth as the solar wind.
Inside the convection layer of the Sun “lives” a complicated magnetic
field, somewhat in the same way as in the liquid outer core of the Earth.
The Sun does not rotate as a solid body: the higher-latitude zones lag
clearly behind the equatorial zone. This, together with the magnetic
field, acts as a complicated natural dynamo. Like the Earth’s magnetic
field, the Sun’s magnetic field can also flip its direction, which happens
periodically, approximately every 11 years.
The Sun’s gas is an ionised plasma, which conducts electric currents
almost like a superconductor. Therefore, the gas and the magnetic field suprajohdin
11 For comparison, the temperature of the Sun’s visual surface, or photosphere, is “only”
approximately 5700 K.
í Õ ! ¤. û
F IGURE 17.11. Sunspots and their magnetic field lines in UV light above the so-
lar surface. Sunspots always occur in pairs, as a magnetic north
and a magnetic south pole. NASA Solar Dynamics Observatory,
^ Atmospheric Imaging Assembly.
are entangled in each other (“frozen-in”) in an inseparable way. The

subject is studied by the discipline of magnetohydrodynamics (MHD), to
which a short introduction is given in appendix B.
auringonpilkku During times of an active Sun, sunspots are seen on its surface, areas
of high magnetic field strength where the field of the convection layer
breaks out to the surface. The spots are born and always show up in pairs,
magnetic north and south poles. The magnetism in the spots inhibits the
natural convection, and thus prevents energy from reaching the solar
surface. In the centres of sunspots, temperatures are even a couple of
12 thousand degrees lower than on the solar surface on average,12 about
3000–4000 K.
The magnetic field of the sunspots extends into the space above the
spots, and affects the motion of the plasma there. With special imaging
12 However, even a single large sunspot transferred to the night sky would still outshine
the full Moon!
í Õ ! ¤. û
Satellite orbital motion 17.7
473
Auroral rings
Magnetopause
Magneto-
sphere
Geo-
magnetic
tail
Solar wind
Radiation belts
Shock front
^ F IGURE 17.12. Space weather, the magnetosphere and aurorae.
equipment,13 the field’s lines of force can be observed as bright swirls 13

of gas. The topology of the field may suddenly change (reconnection), voimaviiva
causing the energy released from the field to throw the hot plasma into
space. During the solar flare, an excess of ultraviolet radiation and X-rays
is generated, causing extra ionisation in the lowest layers of the Earth’s
ionosphere. This adds to the absorption of short-wave radio, disrupting
radio traffic.
About a day later, the plasma itself arrives in the neighbourhood of
the Earth, and its interaction with the Earth’s own magnetic field and
ionosphere causes aurorae. These, too, affect radio traffic and the quality revontuli
of GPS observations. During an eruption, it may be impossible to collect
useable GPS measurements. Generally during a solar maximum, the
quality of GPS observations is poorer than during a quiet Sun.
The Earth’s radiation belts also consist of hot plasma, fast, electrically
charged particles, which the Earth’s magnetic field keeps contained in a
“magnetic bottle” (Wikipedia, Magnetosphere).
^ 17.7 Satellite orbital motion

In the same way as the Earth orbits the Sun, the orbital motion of
satellites around the Earth also follows the laws of Kepler.14 We already 14
touched upon the matter regarding GPS, section 12.10. Kepler’s laws are:
13 Beautiful imagery and videos are found: SOHO images; SOHO MPEG movies.
14 Johannes Kepler (1571–1630) was a German astronomer, mathematician and mystic.
í Õ ! ¤. û
∆t
d2 d1
∆t
∆t
F2 F1 F1
(a) (b)
Drawing an ellipse Kepler’s second law
F IGURE 17.13. ← An ellipse is the set of points for which the sum of the dis-
tances d 1 + d 2 from two focal points F1 and F2 is constant. This
property — a consequence of the ellipse being a conic section —
is most easily proven using Dandelin spheres (Wikipedia, Dan-
delin spheres). It also means that if you place a lamp in one
focus of an elliptical mirror, an image of the lamp will appear in
the other focus. This is what gave the focal points their name.
→ According to Kepler’s second law, the radius vector of a planet
^ sweeps in the same time ∆ t over an always same-sized area.
I The satellite moves around the Earth in an elliptic orbit in a plane.

polttopiste The centre of mass of the Earth is located in one of the focal points
of the orbital ellipse.
II The radius vector between the satellite and the centre of mass
of the Earth always sweeps in the same amount of time over the
pintalaki same surface area (law of areas).
III The squares of the periods of different satellites stand in the same
isoakselin ratios as the cubes of the semi-major axes of their orbital ellipses.
puolikas See below, equation 17.1.
rata-alkiot A satellite orbit is described by six Kepler orbital elements, figure 17.14.
The orbital elements are described in more detail in appendix C. The
angle θ0 is Greenwich sidereal time, which describes the orientation of
the Earth with respect to the stars.
Kepler’s laws apply only approximately, especially for low orbits. The
uneven distribution of the Earth’s masses, and especially her flattening,
ratahäiriö cause orbit perturbations. These are exploited for studying the internal
í Õ ! ¤. û
Choosing a satellite orbit 17.8
475
E
Pole
ν
Z a
Apogee Perigee
b ω
X
Perigee
ea
Y
ν ω Ascending node
Satellite
i
Ω
Earth
θ0 rotation
̇
Ω
X
X X′ (Greenwich)
Vernal equinox
Apogee
^ F IGURE 17.14. Kepler’s orbital elements.
mass distribution of the Earth. This is how satellite geodesy has become
an essential tool for studying the solid Earth.
^ 17.8 Choosing a satellite orbit

The tilt of a satellite orbital plane relative to the equatorial plane, or its
inclination, is an important parameter from the viewpoint of the intended
use of the satellite. The inclination is in practice the same as the greatest
possible northern or southern latitude over which the satellite can fly. If,
for example, it is given that the inclination of some satellite is 55◦ , one
may conclude that the satellite will never come down over Finland.15 15
15 This is assuming that the satellite is not capable of aerodynamic flight upon atmo-
spheric entry.
í Õ ! ¤. û
^ T ABLE 17.1. Kepler’s third law for Earth satellites.
Height (km) Period Remark
0 84m 29s Schuler period

400 92m 34s
800 100m 52s
20 183 11h 58m GPS
35 785 23h 56m Geostationary
376 603 27d 07h Moon
From the viewpoint of low-flying weather and remote-sensing satellites,

the importance of the inclination is that the area that a satellite can
properly map lies approximately between these maximum latitudes.
This limitation does not apply for high-flying satellites. For example,
geostationary satellites may perfectly well map the Nordic area. The
imaging angle is not, however, good.
The choice of the height of a satellite orbit is made using Kepler’s laws
of orbital motion. Kepler’s third law says:
GMP 2 = 4π2 a3 , (17.1)
in which a = a e + h is the semi-major axis of the satellite orbit, the mean

distance from the centre of the Earth. The quantity h again is called the
mean height of the satellite. P is the orbital period, the time taken to go
around the Earth.
The perigee and apogee heights from the Earth’s surface are formally
calculated as follows:
h P = (1 − e ) a − a e , h A = (1 + e) a − a e =⇒ hA − hP = 2 ea.
In this, a e is the equatorial radius of the Earth, according to the GRS80

reference ellipsoid, 6378 137 m.
^ 17.9 Satellite orbital precession, Sun-synchronous orbit

The figure of the Earth affects satellite orbital motion. For example,
dynaaminen the quantity J2 , the dynamic flattening, the value of which is J2 =
litistyneisyys 1082.6267 · 10−6 , is the largest of those many spherical-harmonic co-
pallofunktio-
efficients that together describe the figure of the Earth and that affect
kerroin
satellite motion. In the case of J2 , one important effect is, that the satel-
lite orbital plane turns at a certain rate around the rotation axis of the
í Õ ! ¤. û
Satellite orbital precession, Sun-synchronous orbit 17.9
Sun’s (apparent)
477
diurnal motion
Satellite
Attraction caused orbital motion X
by Earth’s flattening
Diurnal motion of satellite

orbit’s ascending node
^ F IGURE 17.15. Sun-synchronous orbit.
Earth, a phenomenon called nodal precession. Because of this, if the ratatason

satellite flies over the same place day after day, it will do so every day a prekessio
number of minutes earlier. The rate of precession, for a circular orbit of
radius a, is described by the equation
√︃
̇ = dΩ = − 3
)︂2
GM ae
(︂
Ω 2 a J2 cos i,
dt a3
in which a e is the equatorial radius of the Earth and i the inclination
angle of the orbit with respect to the equator. Substituting numerical
values yields
dΩ ]︁ cos i
= −1.318 94 · 1018 m3.5 s−1 · 3.5 .
[︁
dt a
If, as a calculation example, we substitute into this as the satellite height
h = 800 km =⇒ a = 6378 137 m + 800 000 m = 7178 137 m,
we obtain
dΩ
= − 1.331 02 · 10−6 rad s−1 · cos i = −6◦. 589 day−1 · cos i.
[︁ ]︁ [︁ ]︁
(17.2)
dt
For practical reasons (solar panels!) we often choose the satellite orbit so,
that the orbital plane turns with the apparent annual motion of the Sun:
360◦
= 0◦. 9856 day−1 .
[︁ ]︁
365.25 days
If the inclination of the orbital plane is chosen in the range 96◦ –102◦ ,
depending on the height, the dynamic flattening J2 of the Earth will
cause just the suitable rotation of the orbital plane (“no-shadow orbit”,
Sun-synchronous orbit).
í Õ ! ¤. û
Height (km) Critical inclination
500 97◦. 4
750 98◦. 4
1000 99◦. 5
1500 102◦. 0
1. A sidereal day is 23h 56m 4s . Why is it shorter than a civil day, 24 h?
2. What is the ecliptic? How did it get its name?
3. What is the vernal equinox?
4. What mechanism causes the four seasons?
5. What is precession, and what is nutation? What causes them?
6. How did the Tropics of Cancer and of Capricorn get their names?
7. What geometrical quantity does Greenwich Apparent Sidereal Time
(GAST) represent?
8. Describe the co-ordinates on the celestial sphere, right ascension
and declination.
9. What is the hour angle of a celestial object? How is it related to its
right ascension?
10. What do we mean by the upper and lower culmination of a star?
Can they both always be observed (assuming clear skies)?
11. What is a meridian transit, and how is it observed?
12. Explain Väisälä’s stellar triangulation.
13. Describe the components of polar motion and variations in length
of day. How are they observed, historically and today?
14. How are precession and nutation different from polar motion?
15. What is a plasma?
16. What is the solar wind? Is it a sub- or supersonic flow? Why?
17. What is a Carrington event? Ask Google.
18. Could the wreck of the International Space Station (assuming it is
not brought down in a controlled way) ever impact Finnish territory?
Why / why not?
19. How does one engineer a Sun-synchronous orbit? Why is it useful?
í Õ ! ¤. û
^ Geodesy and geophysics
18
Satellites have revolutionized oceanography. This is not so
much because of the instrument packages (remarkable as
they are) but the ability to sample adequately, and to
sample globally (two different things). Take the US–French
altimetry mission TOPEX/POSEIDON that sampled the
topography of the sea surface at about 7 km intervals to an
astonishing precision of one-inch [. . . ]. When you go over
the list of accomplishments, you find that what really made
the difference was the sampling. I consider this the most
successful ocean experiment of all times. [. . . ]
Walter Munk (2002)a
a Walter Heinrich Munk (1917–2019) was an influential American

physical oceanographer.
^ 18.1 Geodynamics
Geodynamics is the field of study within geophysics that studies the
motions taking place in the Earth, like plate tectonics, post-glacial land
uplift, and other motions of the Earth’s crust, local, global, natural or
human-caused. The study of the Earth’s rotation is also normally included
with geodynamics.
Of the observation techniques suitable for geodynamic research, we
have already discussed satellite positioning. All GNSS methods — GPS,
GLONASS, BeiDou and Galileo — are suitable for precise geodynamics
measurements.
However, in scientific research we always want to use as many as
possible, as independent as possible techniques for the study of the same
– 479 –
480 18 G EODESY AND GEOPHYSICS
phenomenon, in as versatile a fashion as at all possible. Therefore we

use also some very precise, but also very expensive, research methods in
geodynamics research:
laser- ◦ satellite laser-ranging
etäisyysmittaus
◦ laser-ranging to the Moon
pitkäkanta- ◦ VLBI, very long baseline interferometry with radio telescopes
interferometria
◦ DORIS, Doppler Orbitography and Radiopositioning Integrated by
Satellite (The IAG’s International DORIS Service), a French satellite
orbit determination system that serves geodetic and geodynamic
research.
The satellite laser-ranging and lunar laser-ranging techniques are very
similar. Laser satellites are heavy, massive spheres manufactured from
köyhdytetty high-density materials like bronze or depleted uranium, studded with
uraani reflective corner-cube prisms. The Apollo astronauts left several reflec-
kuutioprisma
tive panels covered in prisms on the Moon serving the same purpose.
The Soviet vehicles deposited on the lunar surface in unmanned flights,
Lunokhod 1 and 2, also carry reflective prisms.
The distance to satellites or to the Moon is determined by measuring
the two-way travel time of a light pulse, multiplying it by the speed of
light, and dividing by two. The influence of the atmosphere on signal
propagation must be taken into account. Unlike GNSS, the laser-ranging
technique measures real distances, without clock unknowns, not pseudo-
ranges. Due to this, the global laser-station network is somewhat stronger
geometrically than the global GNSS network. For this reason it gives a
valuable boost to the global monitoring network’s geometric strength.
VLBI, very long baseline interferometry, uses the radio signals of remote
1 radio sources, quasars. Quasars are so far away and compact1 that in
practice they are point sources. Therefore the wave form of the noise-like
radio signal that they transmit will be precisely the same, no matter
where on Earth it is received.
VLBI observations are carefully planned. Many radio telescopes around
the globe participate in the campaigns. All participants execute an
agreed programme, in which the same objects are observed simultane-
1 The dominant theory is that quasars are supermassive black holes in the cores of
remote, thus young, galaxies. The radiation is generated when the hole sucks up matter
from its surroundings. Our own Milky Way galaxy also has a black hole at its centre,
Sagittarius A*, which however radiates much more weakly.
í Õ ! ¤. û
Geodynamics 18.1
481
LAGEOS 1
LAGEOS 2
60 cm X
a e i
LAGEOS 1 12 273 km 0.004 44 109◦. 84
LAGEOS 2 12 273 km 0.013 77 52◦. 68
X
F IGURE 18.1. A LAGEOS satellite. The diameter of the satellite is 60 cm, it

consists of an exterior aluminium sphere to which are attached
426 corner-cube prism reflectors (see subsection 7.4.3), and an
interior sphere made of bronze. The mass of the satellite is about
400 kg. The LAGEOS 1 satellite also carries a map designed by
Carl Sagan, describing the current locations of the continents,
as a message to future, possibly alien, finders (Wikipedia, The
LAGEOS time capsule). On the right, the orbits of the LAGEOS 1
^ and 2 satellites.
ously within a certain frequency band.

There are various solutions and generations of instrumentation for stor-
ing the signal. “Mark II” used common-or-garden VHS video tapes. The
signal was converted by special equipment into a video signal and stored
to tape. Today, the recording is often done digitally onto a hard-drive ”mäyräkoira”
pack, and direct transfer over the Internet, using the UDP/IP protocol, is
also becoming popular.
The recordings are read and correlated with each other, finding the time
offset that makes the correspondence between the sequences apparent,
just like in comparing the year rings of two pieces of wood (figure 7.7). vuosilusto
This correlation is a compute-intensive process which earlier required
specialised equipment. As a result of the correlation process, we obtain
the difference in reception times ∆ t at the two radio telescopes, which is
thus the observable of this technology. Its corresponding distance c ∆ t
may also be interpreted as the projection of the vector R AB connecting
the radio telescopes A and B on the direction vector to the quasar, see
figure 18.2.
í Õ ! ¤. û
t
fron
ve
Wa Quasar
t
c∆
Atomic
clock
Tape A R AB
Tape B
Correlator
Atomic “White noise”

c∆ t clock
^ F IGURE 18.2. Principle of operation of very long baseline interferometry.
In order to work, like the famous Väisälä interferometry the method

requires that the signal is random, that is, white noise: the coherence
length has to be short. It is the same with GPS code measurement, for
which the noise is generated artificially, as pseudo-random code.
By carrying out measurements to many different quasars, one may
solve for the vector R AB between radio telescopes A and B in a geocentric
— either inertial or celestial, tied to the stars, or co-rotating with the Earth
— co-ordinate reference frame. Just like satellite laser observations, VLBI
observations also strengthen the global geodynamic monitoring networks.
Geodetic VLBI observations are also carried out in Finland, with the
Metsähovi radio telescope, figure 18.3.
One essential difference between, on the one hand, VLBI observations,
and on the other, both laser and GNSS observations, is that the centre of
mass of the Earth does not come along in the VLBI observation equations.
Quasars are at such huge distances, that any small change in the location
of the centre of mass of the Earth would affect the observations by the
radio telescopes at both ends of the VLBI vector in precisely the same way,
and the effect on the end result would be zero.
On the other hand, the momentaneous direction of the rotation axis of
the Earth is present in the measurement geometry, and VLBI has become,
besides GNSS, a favourite means of monitoring the Earth’s rotation, both
í Õ ! ¤. û
Plate tectonics 18.2
483
F IGURE 18.3. The radio telescope of the National Land Survey’s Finnish
Geospatial Research Institute (FGI) at the Metsähovi research
station, used for geodetic very long baseline interferometric (VLBI)
observations. The telescope, aperture 13.2 m, belongs to the inter-
^ national VLBI Global Observing System (VGOS).
polar motion and variations in the length of the day. vuorokauden

pituus
Close to the Metsähovi research station, three kilometres distant in
Sjökulla, shielded by the landscape from the radio telescope, is a DORIS
beacon containing an active radio transmitter. The French DORIS (Doppler
Orbitography by Radiopositioning Integrated on Satellite) is an unusual
system: the observation stations are active and the satellites passively
collect their signals. An advantage of this solution is the centralisation
of data collection. Around the world there are 60 DORIS stations, with
stations on all major continental plates.
^ 18.2 Plate tectonics

The German Alfred Wegener2 proposed as early as in 1912 that the 2
continents, fragments of the Earth’s crust, were moving slowly over the
soft interior of the Earth, like ice floes on the sea. As evidence he offered
í Õ ! ¤. û
Mid-
Atlantic
Ridge
South America
and Africa
150M years
before present
F IGURE 18.4. Alfred Wegener’s continental drift theory and the Mid-Atlantic
^ Ridge.
the shapes of the continental coast lines, which fit together remarkably
well (figure 18.4) and rock types and fossils, which are often very similar
in corresponding coastal locations.
Nobody believed him at the time. It did not help matters that seis-
poikittainen mology showed the Earth’s mantle to be hard as steel: transversal —
aaltoliike sideways oscillating — waves, S waves, only travel in solids, and they
travel very well through the Earth’s mantle.

Not until the 1960s did the theory receive more support, especially
based on research into the Earth’s magnetic field. Back then, it was
already possible to measure the local magnetic field from an aircraft,
military technology developed for detecting submarines. The minute
variations in the field were mapped, and parallel magnetisation stripes
appeared everywhere on the maps, see figure 18.6. The stripes run in
the direction of the Mid-Atlantic Ridge, or similar ridges in other oceans.
Their pattern is the same everywhere, even if sometimes broader, some-
times narrower — like the tree rings from the same period in different
trees, tableau 12.2.
2 Alfred Lothar Wegener (1880–1930) was a German meteorologist, geophysicist and

Greenland explorer, where he perished at the age of fifty and was buried into the
continental ice sheet.
í Õ ! ¤. û
485
Magnetic
field
Inner
Mantle,
core,
solid, Outer
solid
plastic core,
liquid
S
waves
P
Crust
waves
F IGURE 18.5. The internal structure of the Earth. Convection currents in

the outer core generate the electric currents that maintain
the Earth’s magnetic field — the natural-dynamo theory. The
core consists fairly certainly of an iron-nickel alloy with lighter-
^ element impurities.
The theoretical explanation is that new Earth crust — sea floor — is merenpohja
being formed all the time at the Mid-Atlantic Ridge: hot, liquid magma Atlantin
rises up, cools down and solidifies, and the iron-ore particles in the magma keskiselänne
turn themselves permanently along the direction of the magnetic field at
that point in time.
At this moment, the Earth’s northern magnetic pole — in Canada — is
physics-wise a south pole, S ; the southern magnetic pole, near Tasmania,
is an N type pole. These roles, however, swap at irregular intervals, of
the order of a million years. These alternating directions are recorded
into the sea floor in the same way that sound is recorded on the magnetic
tape of a tape recorder.
The Earth’s internal structure according to our best current knowledge
í Õ ! ¤. û
Sea-floor
motion
Forming new
sea floor
S
Palaeomagnetic
zones N
Reversal of Earth’s magnetic direction

N α β
Time Radiometric
Magneti-
dating
(before pres.) sation
(Holmes)
S γ
^ F IGURE 18.6. Palaeomagnetism and sea-floor spreading.
is presented in figure 18.5.

3 At the same time, the British geologist Arthur Holmes3 developed
radiometrinen radiometric dating, which exploits the uniform decay rate of radioactive
iänmääritys isotopes of various elements occurring in the Earth’s crust as a geological
clock (Lewis, 2000). Among the useful decay processes are potassium 40
→ argon 40, half-life 1.25 billion years, uranium 238 → lead 206, 4.47
billion years, and uranium 235 → lead 207, 710 million years.
By measuring the concentrations of the decay product and the original
isotope, one may infer how much time has elapsed since the solidification
of the rock or mineral. The ages determined in this way for different
places on the sea floor agree very well with the pattern of the magnetic
stripes: the longer the distance from the central ridge, the older the sea
floor.
We say that the continental plates move, but the plates include sea
floor as well. We speak of the Eurasian plate, although it also contains
the whole North Atlantic sea floor north of the Azores and east of the Mid-
Atlantic Ridge. What is happening is a very slow convective motion, which
carries the plates of the Earth’s crust along with it. The rates of motion
vary from a few centimetres per year to as much as ten centimetres. The
precise form of the convective pattern is still unclear: does it take place
3 Arthur Holmes FRS FRSE (1890–1965) was a British geologist, a pioneer of radiometric
dating and our understanding of the mechanisms of plate tectonics.
í Õ ! ¤. û
487
Mid-oceanic ridge
Plate motion
Deep-sea trench Conrad
Crust discontinuity
Sea X
Mohorovičić
Lithosphere discontinuity
Subduction Bottom of
Convection
lithosphere
Asthenosphere
Benioff zone
Mantle 660 km
discontinuity
^ F IGURE 18.7. Global plate tectonics.
throughout the mantle, or in two layers?

The thermal energy that maintains the convection originates from two
sources: radioactive decay, and “primordial heat”, the ongoing cooling alkulämpö
down and stratification after the formation of the Earth. Both parts are kerrostuminen
believed to be of similar magnitude (Wikipedia, Earth’s internal heat
budget). The heat produced by the Earth’s core is entirely primordial: the
solid inner core continues to grow at the expense of the liquid outer core.
The convection theory is presented in figure 18.7, and more details on
the mechanism of continental motion in figure 18.8. At the Mid-Atlantic
Ridge (and the Mid-Pacific, Mid-Indian etc. ridges) new sea-floor crust
is being formed out of the magma rising from the mantle (however, as
already said, the Earth’s mantle itself is not liquid). At the edges of the
oceans again are found deep-sea trenches, under which the oceanic crust syvänmeren
dives down into the mantle (subduction). hauta
alityöntö
In precise geodetic work, the continental-plate motion must be taken
into account. When computing the velocity of motion, the NUVEL models
may be used, which tabulate the motion of each plate as a rotation around kiertoliike
a given pole at a given rotation rate. The NUVEL-1A model (DeMets et al.,
1994) has been computed from geological research material. The results
í Õ ! ¤. û
Diverging boundary Transform fault Converging boundary
Shield-volcano Mid- Island arc

Graben
archipelago oceanic (Japan)
(East Africa)
(Hawaii) ridge Stratovolcano
Su
bd
uc
tio
n
Mantle
“Hot spot”
Core
F IGURE 18.8. Mechanisms of plate tectonics, plate boundaries and volcanism.

The thickness of the Earth’s crust is exaggerated. The “deep hot
^ spot” hypothesis is not generally accepted among geophysicists.
agree, within their measurement uncertainties, with the results obtained

by GPS and other methods.
Over the years, improved plate-motion models have been computed,
see UNAVCO, plate-motion calculator.
^ 18.3 Glacial isostatic adjustment (GIA)

In many Arctic areas, the Earth’s surface has been rising slowly after
the end of the last ice age: glacial isostatic adjustment (GIA). Outside
maan the Arctic area, the phenomenon takes the form of subsidence in the
vajoaminen periglacial bulge. And the changes in sea level associated with the
pullistuma
varying volume of land ice have a global impact.
When modelling the response of the solid Earth to varying ice load, we
need to consider the following modes of response:
í Õ ! ¤. û
Glacial isostatic adjustment (GIA) 18.3
489
Elastic The object responds immediately, and returns to its original
shape immediately when the load vanishes. For example, a spring,
a tennis ball.
Plastic The object responds slowly and continuously. It does not return
to its original shape when the load vanishes. For example, syrup,
modelling clay. The resistance offered to deformation by a plastic
substance is characterised by a property called viscosity.
Viscosity The more solid a substance, the higher the value of its
viscosity. Unit: Pa s (pascal second) or N s m2 .
/︁
The physical character of post-glacial land uplift is one of plastic rebound.

The phenomenon is studied because it offers a possibility to determine
viscosity values for the Earth’s mantle. The results obtained point to the
following structure:
1. the “lid” or lithosphere, thickness 50–100 km, which has a high (in
practice, infinite) viscosity and responds purely elastically4 4
2. under the lithosphere, the asthenosphere, a layer with a relatively

low viscosity, 1020 –1021 Pa s, thickness several hundred km
3. under the asthenosphere, the lower mantle, which has a relatively
high viscosity, order 1022 Pa s or even higher.
The numerical values given above are highly uncertain and may well
change as a result of ongoing research.
For comparison, the viscosity of running water at a temperature of 20◦ C
is 0.001 Pa s; that of liquid sodium — used as coolant in breeding reactors natrium
— 0.0007 Pa s at its melting point 98◦ C. The viscosity of pitch, again, was hyötöreaktori
determined in a famous, and still ongoing, experiment (Edgeworth et al., piki
1984) as (2.3 ± 0.5) · 108 Pa s.
Example: the Fennoscandian land uplift

The land is rising in Finland, as it is in Sweden, Norway, Denmark,
Scotland and Canada. All these vertical movements are caused
by the melting, after the last ice age, of continental glaciers or ice mannerjäätikkö
sheets, some 11 000 years ago. In Fennoscandia a continental ice
sheet, of a thickness of up to two-three kilometres, pressed the
Earth’s crust down. The phenomenon is called isostatic adjustment:
although both the Earth’s crust and the mantle consist of solid
4 . . . until the loading becomes too heavy and the rock fractures: it is frangible.
í Õ ! ¤. û
Continental ice sheet (melted ≈11 000 yrs ago)
Sweden Finland
Atlantic Baltic Sea
Norway
Lithosphere
(viscosity ≈100 km
“infinite”) X
Asthenosphere
(viscosity ≲ 1021 Pa s)
660 km
interface
Lower mantle
(viscosity ? 1022 Pa s)
F IGURE 18.9. Post-glacial land uplift in Fennoscandia. The figure gives a ver-
tical cross-section of the area; in fact, the phenomenon is three-
dimensional. Topography and ice-sheet thickness are exagger-
^ ated.
matter, they nevertheless give way slowly and plastically under

great loads.
The post-glacial land uplift in Fennoscandia is taking place in an
area in the shape of an ellipse, the centre of which, the land-uplift
Merenkurkku maximum, is located at the narrow (“Kvarken”) of the Gulf of Both-
nia, on the Swedish side, where the uplift amounts to over 9 mm a .
/︁
Going outwards from here, the rate diminishes in all directions.

At the Finnish south-eastern border, only 3 mm remains. The
/︁
a
zero land-uplift line runs through Denmark and Northern Poland,
curving on the Lithuanian and Russian territories to the north-
east. Outside, that is to the south of, the line, the land is slowly
subsiding.
Fennoscandia is not the only area in the world where the Earth’s crust is
rising as a result of the termination of the last ice age. We speak of GIA,
í Õ ! ¤. û
Glacial isostatic adjustment (GIA) 18.3
491
0◦ ◦
10 ◦ 20◦ 30
70 ◦ 70
◦
65 ◦ ◦
65
60 ◦ ◦
60
55 ◦ ◦
55
0◦
◦
10 ◦ 30
20◦
F IGURE 18.10. Horizontal and vertical motions in Fennoscandia as determined

by the BIFROST project. The horizontal arrows and vertical bars
^ are not on the same scale. Source Lidberg et al. (2009).
glacial isostatic adjustment. Other similar extended areas are:

◦ The Laurentide (Northern Canada) land uplift area. Much more
extensive than the Fennoscandian one.
◦ The West Siberian land uplift area. Little studied (Ehlers et al.,
2015).
When the Earth’s crust rises, material flows slowly inwards in the as-
thenosphere towards the uplift centre, into the void under the crust that
would otherwise be created. The land uplift is not only a vertical motion:
horizontal motions also occur, as GNSS monitoring over extended periods
has shown.
Techniques for studying and measuring the post-glacial land uplift
include
◦ Regularly repeated (at intervals of decades) precise levelling.
◦ Monitoring sea level with respect to the Earth’s crust by using
í Õ ! ¤. û
mareographs, also called tide gauges, at the coast.

◦ Monitoring changes in gravity, for example the Nordic gravity
profiles (Mäkinen et al., 2010). The changes in gravity have two
causes:
– The point of measurement shifts with the land uplift away
from the Earth’s centre, and gravity diminishes.
– Due to the flow of mass under the Earth’s crust, the amount of
mass under the point increases and gravity grows stronger.
The observed change in gravity is the net result of both effects.
◦ GNSS monitoring in three dimensions. This activity started in the
1990s, when sufficiently precise GNSS receivers and processing
methods became available.
^ 18.4 Local geodynamics
^ 18.4.1 Anthropogenic motions of the Earth’s crust

We humans, by our activities, often cause motions of the Earth’s crust,
vesiallas anthropogenic motions. For example, building a reservoir dam and the
filling of its reservoir causes an additional loading of the local Earth’s
crust, and may even cause tiny earthquakes.
In Venezuela, the pumping of oil in the Caracas region has caused a
very noticeable local subsidence of the land. There are many other similar
areas in the world. In the Netherlands, in Groningen, the pumping
of natural gas has caused a subsidence, for which farmers are being
financially compensated. For this reason, the motions are measured with
geodetic precision, at regular intervals.
The pumping of drinking and irrigation water from porous layers —
aquifers — can also cause subsidence of the land by as much as metres.
This is a very common problem everywhere, but especially in developing
countries (Wikipedia, Groundwater-related subsidence).
^ 18.4.2 Natural motions of the Earth’s crust

Local movements of the Earth’s crust often happen close to geological
murtovyöhyke faults, like the San Andreas fault in California, the boundary between
two tectonic plates, where the plates move slowly with respect to each
other. When the movement of the plates gets stuck, stresses inside the
í Õ ! ¤. û
Deformation monitoring 18.5
493
Earth’s crust build up, and may after years be released destructively in
the form of an earthquake.
Various geodetic methods are used nowadays to monitor motions of the
Earth’s crust. At the mid-ocean ridges, like in Iceland, too, all kinds of valtameren
geodetic measurement activity takes place in the service of geophysical keskiselänne
research.
The continental plates are not necessarily completely rigid and of one
piece; rather, inside them there are also all kinds of faults along which
tectonic movements have taken place — and surely still take place. In
addition to observing microseismicity, one could try to observe these with
geodetic monitoring techniques (Ahola, 2001).
Variations in sea level, like the tides, may be reflected in the level of
the Earth’s crust in coastal areas. One speaks of ocean and atmospheric
loading. The motion caused by this loading may even be a couple of
centimetres, but peters out quickly going in-land. Only in recent years has
it been possible to measure this motion using GNSS, but the uncertainties
are large. This tidal loading is also visible in long gravimetric monitoring
time series. It is one way of studying the local elastic properties of the
solid Earth.
The effect of the atmosphere, mostly variations in air pressure, should
also be visible in this way. The phenomenon is, however, very weak
and hard to observe with confidence. The problem with gravimetric
techniques again is the difficulty of separating the effect of loading from
the many other effects of the atmosphere on the measurement device and
its surroundings.
^ 18.5 Deformation monitoring

Both traditional (total-station measurement, precise levelling) and mod- takymetri
ern techniques like GNSS monitoring can be used to monitor local defor-
mations in the Earth’s crust.
In earthquake-prone areas (for example Japan, California) GNSS mon-
itoring networks made up of many hundreds of continuously operating
receivers have been built. In Japan, the Sendai earthquake of 2011 was
recorded by 1200 stations covering the country, as shown in the video.
However, more traditional techniques, like monitoring with automatic
total stations, are also used. Collecting and processing the materials is
done in real time.
í Õ ! ¤. û
Detecting deformations from the observational material is a similar

karkea virhe task as detecting gross errors: the same kind of statistical testing (section
15.3) may be used, with a suitably chosen alternative hypothesis Ha .
The SAR (synthetic-aperture radar) technique from satellites (Field-
synteettinen ing, 2017) is also used for deformation monitoring. “Synthetic aperture”
aukko means that the radar images taken during different satellite overpasses
are computationally combined in such a way that a “virtual objective”

the size of the distance between overpasses is created. The method is
interferometric (InSAR). After suitable processing, the deformations show
up in the images as interference fringes.
^ 18.6 Studying the Earth’s gravity field from orbit

During the decade 2000–2010, three satellite missions were launched to
study the fine structure of the Earth’s gravity field or geopotential; in
other words, to draft a global geoid map.
^ 18.6.1 CHAMP
CHAMP (Challenging Minisatellite Payload for Geophysical Research
and Applications, CHAMP Mission) was launched from the Plesetsk
cosmodrome in Russia on 15 July 2000. The orbital height of CHAMP
was initially only 450 km, which dropped to 350 km during the flight (for
comparison, the orbital height of the GPS satellites is 20 000 km).
CHAMP carried a GPS receiver for precise orbit determination of the
satellite. From GPS data, one may compute the precise location x( t) of the
satellite as a function of time. From this, one may calculate the geometric
acceleration a( t) by differentiation:
d2
a( t) = x( t).
dt2
kiihtyvyys- The satellite also carried an accelerometer, which served to measure the
mittari satellite accelerations caused by the atmosphere’s aerodynamic forces
acting on the satellite. After eliminating these, what is left is only the
accelerations caused by the Earth’s gravitational field, from which a
precise geopotential or geoid model may be computed.
The data collected by CHAMP has been used to compute global geopoten-
tial models. The geographical resolution of the models has been modest,
of an order of 1000 km. The satellite returned into the atmosphere on 19
September 2010 and burned up after 58 277 orbits.
í Õ ! ¤. û
Studying the Earth’s gravity field from orbit 18.6
495
“Synthetic
SAR radar satellite antenna”
s
as
erp
Ov
Target
area
(a)
The geometry of SAR imaging
(b)
Example of an InSAR image
F IGURE 18.11. InSAR image. Earthquake of 17 August 1999 in Izmit, Turkey.

The deformation interval between interference fringes of the
same colour is some 7 cm of horizontal motion. The thin red
^ lines are faults. NASA / JPL-Caltech.
í Õ ! ¤. û
Magnetometer GPS-3
GPS-2
beam GPS-4
GPS-1
GPS antenna
Acceleration vector
Nom
i nal
orbi
t
“CHAMP” True
orbit
Solar
Earth internal mass
cells
density variations
(example)
F IGURE 18.12. Determining the Earth’s gravitational field by tracking the orbit
^ of a low-flying satellite using a GPS receiver.
^ 18.6.2 GRACE
GRACE (Gravity Recovery And Climate Experiment, GRACE Mission)
measured the temporal changes in the Earth’s gravity field, very precisely,
at a time resolution of about a month, at a fairly crude spatial resolution.
These temporal changes are mostly caused by movements of the “blue
film”, the atmosphere and hydrosphere. The quantity measured is also
merenpohjan called the sea-floor pressure, perhaps somewhat surprisingly. The expla-
paine nation is, that the quantity is proportional to all of the mass contained in
a column of air and water.

GRACE was a satellite pair: the satellites (“Tom and Jerry”) flew in a
tandem formation at some 450 km height, at an average separation of
220 km. A microwave link measured changes in the distance between the
satellite at an accuracy of 1 µm s . Both satellites also carried sensitive
/︁
accelerometers to measure and eliminate atmospheric drag.

The measurement system was so sensitive, that changes in a water
layer of even a millimetre’s thickness could be noticed, if they extended
over an area the size of a continent, some 500 km.
The successful launch took place in 2001. The data has been a treasure
for hydrologists (GRACE Mission, hydrology). It was not until October
2017 when one of the satellites developed a malfunction that measure-
ments came to a close. In the following months, the satellites re-entered
the atmosphere.
A GRACE Follow-On Mission (GRACE-FO) was launched in 2018.
í Õ ! ¤. û
Studying the Earth’s gravity field from orbit 18.6
497
Difference between line-of-sight accelerations
Satel
lite 1 lite 2
Satel
Distance 220 km
Precise ranging, wavelength 1.5 cm
m
ht 450 k
Heig
F IGURE 18.13. Basic idea of the GRACE satellite pair: measuring the tiny tem-
poral changes in the Earth’s gravity field using SST, satellite-
to-satellite tracking. The satellites measure mass shifts in the
Earth’s “blue film” — atmosphere, hydrosphere — variations of
^ “total sea-floor pressure” ↓.
^ 18.6.3 GOCE
GOCE (Gravity Field and Steady State Ocean Circulation Explorer) was
the most ambitious of the satellite missions. The satellite was launched
from Plesetsk, Russia, on 17 March 2009. The orbital height was only
250 km, and the satellite carried a rocket engine (an ionic engine) and a
stock of propellant (xenon) for orbit maintenance against atmospheric ajoaine
drag. The GOCE payload contained a gravitational gradiometer, an instru-
ment for precisely measuring components of the gradient of the Earth’s
attraction, the dependence of the attraction on the three co-ordinates
of place. The gradiometer consisted of several extremely sensitive ac-
celerometers mounted on a frame.
GOCE has worked well. However, in July 2010 a serious malfunction
occurred in the telemetry link, which was repaired in August. The
mission ended on the 21th of October 2013, and on the 11th of November,
the satellite returned into the atmosphere and was seen burning up over
the Falkland Islands.
It has been theoretically established that gradiometry is the best way
to measure the very local features of the gravity field, better than orbital
tracking with GNSS. The smallest details in the geoid map that can be
í Õ ! ¤. û
− Measurement of
GOCE satellite acceleration
1 differences
5 −
4 X
3
6
Gradiometer
Accelerometer
Unknown
density variations
F IGURE 18.14. Determining the Earth’s gravitational field with the gravita-
^ tional gradiometer on-board the GOCE satellite.
seen in the GOCE material are 100 km in diameter, their precision being
as good as ±2 cm.
One important application for a global geoid map this precise is map-
ping the deviations of the mean sea surface from the geoid, an equipoten-
tial surface, with similar precision. The true location in space of sea level
is obtained by satellite radar altimetry, also at a precision level of several
centimetres. This difference in level between the mean sea surface and
meritopografia an equipotential surface, the sea-surface topography, may be inverted
into a map of ocean currents. The theory behind this is explained in
subsection 18.10.2. See figure 18.15.
The name of the GOCE mission was inspired by this possibility.
í Õ ! ¤. û
Atmospheric research and GNSS 18.7
499
F IGURE 18.15. A sea-surface topography map produced by the GOCE mission

© European Space Agency. Unit cm. Ocean surface currents
drawn on top, NOAA / Rick Lumpkin (NOAA, Ocean currents).
^ Compare with figure 18.21!
^ 18.7 Atmospheric research and GNSS
^ 18.7.1 Water-vapour values from a GNSS network

The atmosphere affects the propagation of radio waves, and thus of
the GNSS signal. As can be seen from the refractive-index equation for taitekerroin
microwaves 7.6:
77.624 K 64.70 K
/︁ /︁
5748 K
(︂ )︂
hPa hPa
NM = 106 · ( n M − 1) = ( p − e) + 1+ e,
T T T
both the total air pressure p and the partial pressure of water vapour,
or absolute humidity, e affect propagation, but in different ways. If we
substitute T ≈ 285 K (approximately 12 ◦ C), we obtain
NM = 0.27 ( p − e) + 4.8 e.
Write this into the form
NM = ap dry + b p wet ,
in which p dry = p − e is the total of the partial pressures of the dry

constituents of the atmosphere, and p wet = e is the partial pressure of
í Õ ! ¤. û
water vapour. The coefficients a and b describe the influences of dry

air (mostly nitrogen, oxygen, argon, and carbon dioxide) and of water
vapour. The values a ≈ 0.27 hPa−1 and b ≈ 4.8 hPa−1 depend somewhat
on temperature. However, b is some 18 times a: water vapour affects the
propagation of the GNSS signal some 18 times more strongly than dry air.
poolisuus Behind this phenomenon is the chemical polarity of the water molecule;
see section 7.5.
The partial pressure of water vapour is a less-used way to describe
how much water vapour there is in the atmosphere. The more-used
way is relative humidity in percentages, the amount of water vapour
kyllästyminen compared to the amount at saturation, when there is as much water
vapour in the air as there can be at a certain temperature, before it starts
condensing out as liquid. The saturation partial pressure e sat is given in
5 fair approximation by the equation of Clausius5 and Clapeyron:6
6 (︃ )︂)︃
1 1
(︂
e sat (T ) = e sat (T0 ) exp −4895 K · − .
T T0
When we know that the temperature T0 = 100 ◦ C = 373.15 K is the boiling
point of water: e sat (T0 ) = 1 atmosphere = 1013.25 hPa, it follows that
(︃ )︂)︃
1 1
(︂
e sat (T ) = 1013.25 hPa · exp −4895 K · − . (18.1)
T 373.15 K
The lower the satellite is in the local sky, the longer the path through
the atmosphere. If the zenith angle is ζ, we may describe the signal
˜︁ and b
kulkuviive propagation delay in the atmosphere by the following equation (a ˜︁
are the integrals of the above-mentioned coefficients a and b over the full
height of the atmosphere, divided by 106 ):
ap
˜︁ dry + b
˜︁ p wet a˜︁( p − e) + be
˜︁ d
d trop = = = zenith . (18.2)
cos ζ cos ζ cos ζ
verkkotasoitus In a GNSS network adjustment, we may solve the station values d zenith ,
zeniitin zenith propagation delays, as unknowns, by substituting equation 18.2
kulkuviive into the GNSS observation equations 12.1 — and with the identity D
trop =
d trop into equations 12.2. (In a small area, one can only resolve zenith-
delay differences between stations.)
If we also measure, at the same station, the total air pressure with a
ilmapuntari barometer:
5 Rudolf Julius Emanuel Clausius (1822–1888) was a German physicist and mathemati-
cian, one of the founders of thermodynamics.
6 Benoît Paul Émile Clapeyron (1799–1864) was a French engineer and physicist, one of
the 72 names on the Eiffel Tower (Eiffel Tower, 72 names).
í Õ ! ¤. û
Atmospheric research and GNSS 18.7
501
140
Partial water-vapour pressure (hPa) 120

100%
100
80
Rel. humidity 60 %
60
40
Dew point
20%
20
0
0 10 20 30 40 50
Temperature ( ◦ C)
F IGURE 18.16. Saturation partial pressure e sat and partial pressures e of water
vapour at various temperatures and relative humidities accord-
ing to equation 18.1. The red path shows how the dew point
(30◦ C) follows from temperature (40◦ C) and relative humidity
(60 %). This means that temperature and dew point together
can be used to determine relative humidity. Many hygrometers
— devices for measuring the humidity of air — are based on this
^ principle.
p = p dry + p wet ,
GNSS-1 GNSS-2 GNSS-2 GNSS-3
GNSS-3
GNSS-4
“Total zenith delay”
^ F IGURE 18.17. Use of GNSS for studying the troposphere.
í Õ ! ¤. û
we may solve separately for p dry and p wet .

The latter quantity, converted to amount of matter in an air column,
is also called the integrated water-vapour content or total precipitable
water-vapour content (converted to liquid water, unit mm). It is a quantity
that weather and climate researchers are very much interested in!
^ 18.7.2 GNSS radio occultation or “limbsounding”

Using low-flying satellites, it is possible to measure how the radio signal
from a GNSS satellite is slowed down as the lowest point of the ray dives
deeper and deeper into the atmosphere. This technique is called GNSS
limbsounding or radio occultation. The first satellite to exploit this tech-
nique was GPS/MET. The Danish Ørsted satellite and the aforementioned
CHAMP also used the technique.
The technique requires having a GNSS receiver on board the satellite.

As there are already some 30 satellites in the GPS alone, and the low-orbit
satellite goes around the Earth once every 1.5 hours, a quite substantial
amount of information is collected every 24 hours.
The technique is important because it allows the determination of
temperatures high up in the atmosphere, above the tropopause, where
there is not much water vapour left. This is the layer in which climatic
7 global warming would become most visible.7 The ozone layer is located at
the top of the measurement range, and processes taking place there also
affect the temperature.
The technique is the following: the delay caused by the atmosphere is
proportional to air density, which is a piecewise exponential function of
8 height:8
H − H0
(︂ )︂
ρ ( H ) = ρ ( H0 ) exp − ,
S ( t, k, g)
skaalakorkeus in which S ( t, k, g) is the scale height, a function of air temperature t, air
9 composition k, and gravity g. If we assume that k9 and g are known, we
may calculate temperature t from scale height S .
7 Climate models predict surprisingly, and observations confirm, that, while tempera-
tures at the Earth’s surface and in the troposphere go up, a compensating lowering of
temperatures is expected for the stratosphere. The cause is enhanced radiative cooling
by increasing carbon dioxide concentrations.
8 It is a little more complicated than this, as the GNSS signal travels through many
atmospheric layers, not only the layer of closest approach to the Earth’s surface.
9 The water vapour content at these heights is practically zero due to the low tempera-
ture.
í Õ ! ¤. û
Long-term variations in Earth rotation axis and orbit 18.8
503
D True
Delay
C
B Without atmosphere
A
Time
GNSS satellite
orbit
ow
e in l
ellit
Sat
A
B
C
D
^ F IGURE 18.18. How the GNSS radio occultation technique works.
^ 18.7.3 Ionosphere sounding

From GPS measurements at two frequencies, L 1 and L 2 , we may compute
ionospheric models, in which the ionospheric electron density is mapped,
as a function of place (ϕ, λ, h) and time. From this can then be calculated
TEC, the total electron content, integrated along the path between satellite
and receiver.
These models have been computed, nearly in real time, using the
global GNSS network, since 1998 by the IGS, the International GNSS
Service. A data format, IONEX, Ionosphere Map Exchange Format, has
been developed in support of this activity.
^ 18.8 Long-term variations in Earth rotation axis and orbit

Milanković10 proposed as early as in 1941 a hypothesis that variations in 10
the insolation — the radiative power coming from the Sun — at the edges
of the large continental ice sheets cause their growth and retreat, and
that these variations would be caused by astronomical factors. The work
of Milanković is based on earlier work by, among others, James Croll,11 11
to explain the ice ages and interglacials.
10 Milutin Milanković (1879–1958) was a Serbian polymath, engineer and climatologist.
í Õ ! ¤. û
60
40
Insolation / temperature
20
−20
−40
−60
−800 000 −600 000 −400 000 −200 000 0
Year (before 1950)
F IGURE 18.19. Milanković cycles over the past 800 000 years on both hemi-
spheres. Red, theoretically computed summer insolation — from
July 21 to August 20 — at 65◦ northern latitude (outside the
atmosphere). Blue, Antarctic temperature estimate, computed
using the deuterium isotope. Arbitrary scales. Sources: Laskar
et al. (2004); Jouzel and Masson-Delmotte (2007), see IMCCE,
^ Milanković app and the Science article Jouzel et al. (2007).
^ 18.8.1 Tilt of the rotation axis

Besides the precessional motion, the obliquity, or tilt, of the Earth’s axis
of rotation with respect to the ecliptic also varies slowly. Currently (epoch
2000.0) the tilt angle is 23◦ 26′ 21′′ and is slowly diminishing, by about
0′′.47 per year.
Actually the variation is periodic, between 22◦. 1 and 24◦. 5, with a period
ratahäiriö of some 41 000 years. This variation is related to orbit perturbations
caused by the planets on the Earth’s orbital plane.
^ 18.8.2 Climatological precession

Astronomical precession is the earlier described phenomenon, that the
rotation axis of the Earth turns around in a conical figure in a period
isoakseli of about 25 800 years. In addition, the major axis of the Earth’s orbital
11 James Croll FRS (1821–1890) was a Scottish autodidact physicist, astronomer and
climatologist.
í Õ ! ¤. û
Land-ice research and climate change 18.9
505
ellipse also rotates in the orbital plane, the period being 112 000 years
(apsidal precession). This motion is caused by orbit perturbations by the
other planets.
The net result of both processes is climatological precession, the period
of which is )︂−1
1 1
(︂
+ a = 21 000 a.
112 000 25 800
This is the time period in which the season when the Earth is closest
to the Sun cycles through the calendar. Currently the Earth is closest
to the Sun at the beginning of January. Only 11 000 years ago, at the
termination of the last ice age, the Earth was closest to the Sun in July.
The result was 7 % more solar radiation power for the Arctic summers.
The direct impacts of climatological precession are opposite in the
northern and southern polar areas, because the boreal (northern) summer
is the austral (southern) winter, and the austral summer is the boreal
winter.
^ 18.8.3 Variation of the orbital eccentricity

The eccentricity of the Earth’s orbit is also variable, with important
periods being ∼ 100 000 and 413 000 years, between the values 0 and
0.068. This variation, caused by orbit perturbations from the other
planets, modulates the effect of climatological precession.
^ 18.9 Land-ice research and climate change

Much geophysical research is aimed at continental ice sheets, where,
thanks to polar amplification — the enhanced warming of the polar
regions — the warming of our planet’s climate is first becoming visible.
A continental ice sheet is a glacier: it grows above the snow line from
falling snow that slowly compacts, first into firn and then into ice, and
flows slowly and plastically to the coasts, where it loses mass through
melting and formation (“calving”) of icebergs. poikastuminen
The thicknesses of both the Greenland and the Antarctic ice sheet have
been monitored by satellite techniques, among them radar altimetry.
Radar mapping from orbit (SAR, synthetic-aperture radar) has also been
used for this. Observing the small changes of course requires long time
series, and this is still a shortcoming of the satellite methods.
GRACE (subsection 18.6.2) has also been used to study the change in
mass of these continental ice sheets. This change represents directly the X
í Õ ! ¤. û
True orbit
Calculated orbit
h sat ℓ
Sea-surface
Footprint
topography H
Geoid
Reference ellipsoid
Geoid height N
Sea surface Mean sea surface
^ F IGURE 18.20. The measurement geometry of satellite radar altimetry.
contribution of these ice sheets to sea-level rise.

The GNSS technique has also been used for studying both crustal and
ice motions, both in Greenland (Khan et al., 2010) and in Antarctica,
for example in the surroundings of the Finnish Antarctic base Aboa in
Dronning Maud Land (Koivula and Mäkinen, 2003).
^ 18.10 Geodetic oceanography
^ 18.10.1 Satellite altimetry and the geoid

It is possible to measure the location of the momentaneous sea surface
in a geocentric system from a satellite. The instrument used is a radar
altimeter (Rummel and Sansò, 1993) and its precision of measurement is
a few centimetres. See figure 18.20.
A satellite altimetric radar works in this way: it sends a short mi-
crowave pulse straight downwards, where it is reflected back from the
sea surface. The reflection is received and the shape of the return pulse
is analysed. The reflection does not just come from the point straight
underneath the satellite, but from a whole area called the footprint. This
í Õ ! ¤. û
Geodetic oceanography 18.10
507
F IGURE 18.21. Theoretical link between sea-surface topography and ocean cur-
^ rents.
footprint may have a diameter of several kilometres, more if the sea state aallokko
includes high waves.
Based on the analysis of the pulse shape, the distance between the
satellite and the sea surface is inferred. If the satellite orbit is known
— nowadays altimetric satellites always carry a GNSS receiver — the
momentaneous, geocentric location of the sea surface may be computed.
Successive measurement points form arcs running either from south to ratakaari
north or from north to south.
^ 18.10.2 Ocean currents and sea-surface topography

Mean sea level aligns pretty nicely with the geoid, an equipotential
surface of the Earth’s gravity field. If sea water were in a “state of rest”,
an equilibrium state like in communicating vessels, the mean sea surface yhtyvät astiat
would be exactly the same as the geoid. This is, however, not the case.
The causes of this sea-surface topography are
◦ local variations in temperature and salinity suolaisuus
◦ variations in air pressure from place to place, to which the sea

surface responds like an “inverted barometer” (IB) ylösalainen
ilmapuntari
◦ The drag of winds with the sea surface, pushing water masses in
certain directions
◦ the Coriolis force acting on ocean currents. coriolisvoima
Ocean currents respond to the rotation of the Earth though the Coriolis12 X
12
í Õ ! ¤. û
13 force, in the same way as air currents (Buys Ballot’s13 law): currents
try to turn to the right in the Northern Hemisphere, to the left in the
Southern one. This causes a transversal tilt of ocean currents, which is
proportional to the flow velocity and the sine of latitude, sin ϕ. We speak
of geostrophic flow.
The magnitude of the sea-surface topography is globally ±1 m. The
Baltic Sea also has a sea-surface topography, a tilt of the sea surface, as
pohjukka a result of which the bottoms of the Gulfs of Finland and Bothnia are
about 20–30 cm higher than the western parts of the Baltic Sea, due to
the salinity gradient. There is also a height difference across the Danish
straits, caused by the push of predominantly westerly winds.
Earlier on, some satellite missions were mentioned that were intended
to determine the precise location in space of the geoid. The sea-surface
topography on the oceans, the difference between mean sea level and
the geoid, is caused mostly by ocean currents. The rotation of the Earth
causes currents to tilt in the transversal direction. Using this observable
phenomenon, it is possible to theoretically compute the volume of the
current, the amount of water transported, and from this in turn the
thermal energy carried along with the water. This is an essential matter
for climate research. This was one objective of the ambitious GOCE project
(subsection 18.6.3).
In figure 18.21 we see the relationship between ocean currents and
sea-surface topography. We may also describe this in equations, the
geostrophic equations:
∂H ω ∂H ω
= +2v y γ sin ϕ, = −2v x γ sin ϕ, (18.3)
∂x ∂y
in which H ( x, y) is the sea-surface topography, ω the rotation rate of

the Earth, γ gravity, x and y plane co-ordinates in the east and north
directions, and v x and v y are the east and north components of the flow
velocity. Because the equations contain sin ϕ, the effect is opposite in the
Northern and Southern Hemispheres, and in the immediate vicinity of
the equator, it vanishes.
By simple partial differentiation, we may infer from the observed
12 Gaspard-Gustave de Coriolis (1792–1843) was a French mathematician, one of the 72

names on the Eiffel Tower (Eiffel Tower, 72 names).
13 Christoph Hendrik Diederik Buys Ballot (1817–1890) was a Dutch chemist and
meteorologist.
í Õ ! ¤. û
Geodetic oceanography 18.10
509
sea-surface topography H ( x, y) to the flow pattern v( x, y), in which v =
v x i + v y j is the current velocity vector in the ( x, y) plane.
^ 18.10.3 Mareographs, satellite altimetry, sea-level rise

Starting in 1992, the French-American satellite TOPEX/Poseidon and
its follow-up satellites Jason-1, Jason-2 and Jason-3 (NASA JPL,
TOPEX/Poseidon, Jason) have been doing pioneering work monitoring
the rise of the global sea level. Traditionally, the sea level has been
monitored using mareographs or tide gauges; the rise of the sea level over
the whole 20th century has been estimated to have been of the order of
1.3–2 mm a , or 13–20 cm over the whole century.14
/︁ 14
A mareograph measures the rise in the sea level with respect to the
Earth’s crust on which it has been built. A long tube connects the instru-
ment with the open sea, in order to dampen wave motion.
This method has two problems:
◦ The Earth’s crust itself may move. It is mostly the post-glacial
land uplift, more commonly called GIA (glacial isostatic adjustment)
that is important here: it affects Fennoscandia and Canada most
strongly, but its effects extend to all of Europe and North America.
Outside the land-uplift area there is a broad zone in which the land
is slowly subsiding, the “periglacial bulge”. This phenomenon must
be carefully modelled and removed from the mareograph data, if
one wishes to compute global sea-level rise precisely. Nowadays
there are continuously operating GNSS stations co-located with
many mareographs, with the help of which the vertical land motion
may be empirically determined. Unfortunately these time series
are still short.
◦ There are only mareographs on the coasts, and there, only at in-
dividual points. There are large ocean areas where there is not a
single mareograph nearby. And the further one goes back in time,
the weaker the situation becomes. In the Southern Hemisphere,
the situation was really weak before around 1950.
Altimetric satellites can measure the location of the sea surface geocen-
trically — using an on-board GNSS positioning device — and everywhere
underneath their orbits. For example, the inclination angle of the orbit
14 A new analysis (Dangendorf et al., 2017) proposes that the sea-level rise before 1990
was substantially slower, only some 1.1 ± 0.3 mm a .
/︁
í Õ ! ¤. û
of a satellite like TOPEX/Poseidon is 66◦ , and the whole ocean surface be-
tween latitudes 66◦ S and 66◦ N is being mapped at an interval of some ten
ratatason days. When the plane of the satellite orbit precesses, with respect to the
prekessio Sun, once in 117 days, we obtain the long-term trend in the global mean
sea level with at least this temporal resolution. During the latest two
decades, the mean sea level has risen some 3.2 ± 0.4 mm a , clearly more
/︁
than during the 20th century on average (Sea Level Research Group).
The reason for the acceleration of the rise is undoubtedly the warming of
the climate.
^ 18.10.4 Height systems, mareographs, sea-surface topography

Mareographs may also be used to observe the sea-surface topography, but
only on the coast. This presupposes that the mareographs are connected
to the same height system. In Finland, the matter has been organised so
that the 13 mareographs on the Finnish coast operated by the Finnish In-
stitute of Marine Research, nowadays part of the Finnish Meteorological
Institute, are connected to the national precise-levelling network, which
the Finnish Geodetic Institute, today the Finnish National Land Survey’s
Geospatial Research Institute FGI, maintains. The necessary connecting
levellings are carried out by the geodesists at a few years’ intervals.
The Finnish mareographs measure continuously the level of the lo-
cal sea surface with respect to the local Earth’s crust. The measure-
ment values are transferred electronically to the headquarters of the
Finnish Meteorological Institute in Helsinki, where they are processed
and archived.
Near five mareographs, pillars have been erected by the Finnish Geode-
tic Institute in order to carry out regular GNSS measurements. In this
way one may also obtain the local sea surface in a geocentric reference
frame by means of satellite positioning.
Use of GNSS buoys is also becoming more common: these obtain the
geocentric location of the momentaneous sea surface, far away from the
coast, too. The sea state may also be monitored.
1. Which geodetic observation techniques useful for geodynamics do
you know?
2. What is the evidence showing that the Earth’s outer core is liquid?
í Õ ! ¤. û
Self-test questions
511
Kemi
Tide gauge
18
Seasat Furu-
altimetry ögrund Oulu
satellite 20 Raahe
track Ratan
Mean sea surface 22 Pietarsaari

(Vermeer, 1983) Vaasa
Spikarna 19
Kaskinen
Mäntyluoto
24
Rauma
23 19 Turku Hamina
Forsmark Helsinki Kronstadt
20
21 Degerby
Stockholm Hanko Shepelevo
Ristna Suurupi 16 15
Marviken Landsort
27
Ölands norra udde Visby
Kungsholms- 20 Skulte
22
33 fort
Liepaja
24
Klaipeda
25
Władysławowo
26
Ustka
28
F IGURE 18.22. Tide gauges (mareographs) of the Baltic Sea, some ground
tracks of the Seasat satellite back in 1978, and contours of a
solution for the mean sea surface computed from the satellite’s
^ data.
3. Describe the polar motion and its monitoring.

4. In what way were the mission objectives of the GRACE and GOCE
satellite missions different?
5. How can a network of GNSS stations be used to determine the
water-vapour content of the atmosphere above it?
6. What is the atmospheric scale height, and how can GNSS radio
occultation (“limbsounding”) be used to determine it?
í Õ ! ¤. û
512 G EODESY AND GEOPHYSICS
F IGURE 18.23. “Schematic of the Earth’s orbital changes (Milankovitch cycles)

that drive the ice age cycles. ‘T’ denotes changes in the tilt (or
obliquity) of the Earth’s axis, ‘E’ denotes changes in the eccentric-
ity of the orbit (due to variations in the minor axis of the ellipse),
and ‘P’ denotes precession, that is, changes in the direction of the
axis tilt at a given point of the orbit.”
^ From IPCC (2007), Frequently Asked Question 6.1, page 449.
7. What are the three changes in the Earth’s rotation axis and orbital
motion causing long-term climatic variations — the glacial cycle —
according to the theory of Milanković?
8. See figure 18.23. After what you learned about Kepler’s laws of
orbital motion, what is wrong with this picture (and caption text)?
9. According to a newspaper article, Associated Press (1988), military
aircraft that crash-landed in Greenland during the Second World
War were found 46 years later 260 feet deep in the ice. Based on
your knowledge of how glaciers work, explain how they ended up
this deep. What is the annual snowfall rate at the site of the crash?
10. What is the Coriolis force?
11. The geostrophic equations are
∂H ω ∂H ω
= +2v y γ sin ϕ, = −2v x γ sin ϕ. (18.3)
∂x ∂y
Explain each occurring symbol.
12. What was and is the mission objective of the TOPEX/Poseidon satel-
lite and its follow-up satellites, the Jasons?
í ¤. û
^ Properties of matrices
Matrices are used very widely in science to describe multidimensional

A
phenomena, like co-ordinate transformations and mappings from one
vector space to another. The matrix is written out as a two-dimensional
table of numbers between parentheses or square brackets. The symbolic
notation for a matrix is an italic1 capital letter, like A , B, I . 1
Formally, matrices are a type of number, like real numbers and complex
numbers. Many operations on numbers, like addition and subtraction,
also exist for matrices: same-sized matrices may be added simply by
adding together the corresponding elements. Other operations, like
multiplication, do not always work between matrices: they require that
the sizes of the matrices, the numbers of their rows and columns, are
suitable. To wit, the number of columns in the left matrix must be the
same as the number or rows in the right matrix, for multiplication to be
even possible.
^ A.1 Adding and subtracting matrices

Same-sized matrices are added by adding together the corresponding
elements:
C = A+B
means
ci j = ai j + bi j
for every pair i , j , i = 1, . . . , n, j = 1, . . . , m, if the matrices A , B, and C

are of size n × m. For example, in the following way ( n = m = 2):
[︄ ]︄ [︄ ]︄ [︄ ]︄ [︄ ]︄
c 11 c 12 a 11 a 12 b 11 b 12 a 11 + b 11 a 12 + b 12
= + = ,
c 21 c 22 a 21 a 22 b 21 b 22 a 21 + b 21 a 22 + b 22
1 Alternatively, many authors use bold capital letters.
– 513 –
514 A P ROPERTIES OF MATRICES
with the numerical example
C A B
⏟[︄ ⏞⏞ ]︄⏟ ⏟ ⏞⏞ ]︄⏟ ⏟[︄ ⏞⏞ ]︄⏟
[︄
5 2 3 0 2 2
= + ,
1 4 0 2 1 2
in which c 11 = a 11 + b 11 , and so on. The same applies for subtraction.
^ A.1.1 Multiplying a matrix with a constant

This amounts to multiplying every element with this constant. For
example, [︄ ]︄ [︄ ]︄
a 11 a 12 ka 11 ka 12
k = .
a 21 a 22 ka 21 ka 22
^ A.2 Matrices and vectors

A matrix with only one row is called a row vector; a matrix with only one
column, a column vector. For example,
⎡ ⎤
[︂ ]︂ c1
r= r1 r2 r3 r4 , c = ⎣ c2 ⎦ .
⎢ ⎥
c3
Vectors are commonly denoted by small letters.
^ A.3 The unit matrix

In the same way that for numbers there exists a zero element 0, for which
holds ∀a ∈ R : a + 0 = 0 + a = a, and a unit element 1 with the property
∀a ∈ R : a · 1 = 1 · a = a, these also exist for matrices. The unit matrix is
simply ⎡ ⎤ ⎡ ⎤
1 0 ··· 0 1
⎢ 0 1 ··· 0 ⎥ ⎢ 1 ⎥
⎣ ... ... . . . ... ⎦ = ⎣
I =⎢ ..
⎥ ⎢ ⎥.
. ⎦
0 0 ··· 1 1
It is always square. Every number of dimensions has its own unit matrix,
like ⎡ ⎤
1
I3 = ⎣ 1 ⎦
1
is the unit matrix for three dimensions.
í Õ ! ¤. û
Matrix multiplication A.4
515
^ A.4 Matrix multiplication
Multiplication of matrices is done following the simple scheme “row times
column”. If
C = A · B,
this means that for every i and k:

n
∑︂
c ik = a i j b jk ,
j =1
meaning that the matrix

⎡ ⎤
c 11 ··· c 1k ··· c 1n
⎢ . .. .. .. ⎥
⎢ .. . . · . ⎥
⎢ ⎥
C=⎢ c i1 · · · [ c ik ] · · · c in
⎢ ⎥
⎥
⎢ .. .. .. .. .. ⎥
⎢
⎣ . . . . .
⎥
⎦
c n1 · · · c nk ··· c nn
is obtained by multiplying every row of matrix

⎡ ⎤
a 11 ··· a1 j ··· a 1n
⎢ . .. .. .. ⎥
⎢ .. . · . . ⎥
⎢ ⎥
A = ⎢ [a i1 ] → [a i j ] → [a in ]
⎢ ⎥
⎥
⎢ . .. ..
.. ..
⎥
⎢ .. . . . .
⎥
⎣ ⎦
a n1 ··· an j ··· a nn
with every column of matrix

⎡ ⎤
b 11 · · · [ b 1k ] · · · b 1n
⎢ . .. .. .. ⎥
⎢ .. . ↓ . . ⎥
⎢ ⎥
B = ⎢ b j1 ··· [b jk ] · · · b jn ⎥ .
⎢ ⎥
⎢ . .. ⎥
⎢ .. .. ..
⎣ . ↓ . . ⎦
⎥
b n1 · · · [ b nk ] · · · b nn
So, one sums c ik = a i1 b 1k + a i2 b 2k + · · · + a in b nk , “row i times column k”,

for each element of the matrix C . In the above formula we have marked
one row, with row number i , in the matrix A , and one column, with
column number k, in the matrix B. Multiplication of this row with this
column thus yields the element of C called c ik .
Here it is assumed, for the sake of simplicity, that all matrices are
square, n × n rows and columns. This is not inevitable; however, the
number of columns of A and the number of rows in B must be the same
for multiplication to be possible.
í Õ ! ¤. û
Exercise Verify that for an arbitrary square matrix A it holds that

A · I = I · A = A.
^ A.5 The transpose

The transpose of matrix A , the matrix A T , is defined as
A AT
⏟
⎡ ⏞⏞ ⎤⏟T ⏟⎡ ⏞⏞ ⎤⏟
a 11 a 12 ··· a 1m a 11 a 21 ··· a n1
⎢ a 21 a 22 · · · a 2m ⎥ ⎢ a 12 a 22 a n2
⎢ ⎥ ⎢ ⎥
··· ⎥
⎢ . .. ..
⎥ =⎢ . .. ..
⎥,
⎢ . .. ⎥ ⎢ . ..
⎣ . . . . ⎦ ⎣ . . . .
⎥
⎦
a n1 a n2 · · · a nm a 1m a 2m · · · a nm
[ n× m] [ m× n]
or more compactly
AT
(︁ )︁
ij = ( A ) ji , i = 1, . . . , m, j = 1, . . . , n.
The transpose A T of matrix A is thus obtained by interchanging rows

and columns, or, if m = n and we have a square matrix, by “mirroring” all
elements though the main diagonal. Trivially, if the transposition is done
twice, one obtains back the original matrix:
)︁T
AT
(︁
= A.
The transpose of a vector changes a row vector into a column vector and
vice versa:
⎡ ⎤T ⎡ ⎤
a1 a1
a2 a2
⎢ ⎥ [︂ ]︂ [︂ ]︂T ⎢ ⎥
⎢ ⎥ ⎢ ⎥
⎢ ..
⎥ = a1 a2 · · · a n , a1 a2 · · · a n =⎢ ..
⎥.
. .
⎢ ⎥ ⎢ ⎥
⎣ ⎦ ⎣ ⎦
an an
The latter notation is often used in running text to save paper when
presenting a column vector.
^ A.6 The inverse matrix

The inverse matrix N −1 of a square matrix N is defined as the matrix K ,
for which K · N = N · K = I , in which I is the unit matrix already defined
above. A property of the unit matrix again is I · N = N · I = N for all
square matrices N of the same size. The role of the unit matrix is the
same as for real numbers the role of the number 1. For a real number
í Õ ! ¤. û
The inverse matrix A.6
517
a again, the concept corresponding to the inverse matrix is the number
1 a.
/︁
There exist good numerical algorithms for inverting a matrix. Inversion

is however heavy for large matrices: the computational work required
is proportional to the matrix size n to the third power! So, inverting a
size 20 × 20 matrix takes about 8 times the computation time needed for
inverting a size 10 × 10 matrix.
Manual inversion is only readily possible for matrices of size up to 3 × 3
— and is not worth the bother (Reagan, 2021).
For example, if [︄ ]︄
3 1
A= ,
2 4
its inverse matrix is [︄ ]︄
−1 0.4 −0.1
A = ,
−0.2 0.3
which is easily verified:
[︄ ]︄ [︄ ]︄
3 · 0.4 − 1 · 0.2 −3 · 0.1 + 1 · 0.3 1 0
A · A −1 = = = I,
2 · 0.4 − 4 · 0.2 −2 · 0.1 + 4 · 0.3 0 1
and similarly A −1 · A = I.
A valuable tool in matrix computations is the MATLAB™ software which
is in widespread use within the science community. For home users, there
are similar open-source offerings like octave or scilab.
In practice, the computation is based on solving systems of linear
equations: if X = A −1 is the inverse of the matrix A , it holds, for the case
of a size 2 × 2 matrix, that
A X = I,
or [︄ ]︄[︄ ]︄ [︄ ]︄
a 11 a 12 x11 x12 1 0
= ,
a 21 a 22 x21 x22 0 1
or
}︄
a 11 x11 + a 12 x21 = 1
=⇒ x11 , x21 ,
a 21 x11 + a 22 x21 = 0
}︄
a 11 x12 + a 12 x22 = 0
=⇒ x12 , x22 .
a 21 x12 + a 22 x22 = 1
If the matrix is of size n × n, similarly n systems of equations are created,

with in each, n equations in n unknowns.
í Õ ! ¤. û
Singularity Just like the division 1 a does not work for all numbers a
/︁
— specifically not for a = 0 — also the computation of the inverse

matrix does not work for all matrices. A simple example of a
non-invertible matrix is
[︃ ]︃
1 1
A= ,
1 1
and the solution equations above become conflicting:
x11 + x21 = both 1 and 0,

x12 + x22 = the same.
Such a matrix is called singular.
^ A.7 The determinant of a matrix

The determinant of a square matrix is a single number that characterises
the nature of the mapping that a matrix represents, from one vector space
into another of the same number of dimensions: how a small volume
element in one vector space is mapped onto the other one, the ratio of
the source and destination volumes. Also, if the determinant vanishes,
that means that the mapping has no inverse mapping and the matrix is
singular.
The determinant is computed recursively:
⎡ ⎤
a 11 a 12 ··· a 1n
⎢ a 21 a 22 · · · a 2n
⎢ ⎥
⎥
det ⎢
⎢ .. .. .. ..
⎥=
⎣ . . . .
⎥
⎦
a n1 a n2 · · · a nn
⎡ ⎤ ⎡ ⎤
a 22 a 23 ··· a 2n a 21 a 23 ··· a 2n
⎢ a 32 a 33 · · · a 3n ⎢ a a 33 · · · a 3n
⎢ ⎥ ⎢ ⎥
⎥ − a 12 det⎢ .31
⎥ ⎥
= a 11 det⎢
⎢ .. .. .. .. ⎢ . .. .. ..
⎥+···
⎣ . . . .
⎥
⎦ ⎣ . . . .
⎥
⎦
a n2 a n3 · · · a nn a n1 a n3 · · · a nn
⎡ ⎤
a 21 a 22 ··· a 2(n−1)
⎢ a 31 a 32 · · · a 3(n−1)
⎢ ⎥
n−1
⎥
· · · + (−1) a 1n det ⎢
⎢ .. .. .. ..
⎥.
⎣ . . . .
⎥
⎦
a n1 a n2 · · · a n(n−1)
The rule is: for every element of the first row, multiply with the deter-
minant of the matrix formed by deleting the first row and the column of
í Õ ! ¤. û
Vectorial products A.8
519
that element.2 And alternate the algebraic sign: a plus sign for odd- j a 1 j , 2etumerkki
a minus sign for even- j a 1 j .
A typical property of the determinant is that is is fully antisymmetric:
exchange, in the matrix A , any two columns, or any two rows, with each
other, and the determinant of the new matrix A ′ will be det A ′ = − det A .
^ A.8 Vectorial products

The scalar product of two vectors a and b is defined as
⟨︁ ⟩︁ def
a · b = ∥a∥∥b∥ cos α,
in which α is the angle between the vectors.

{︁ }︁
Let e1 , e2 , . . . , en be an orthonormal basis, that is, a set of vectors ortonormaali
spanning an n-dimensional vector space, for which kanta
⎧
⟨︁ ⟩︁ ⎨1 if i = j,
ei · e j =
⎩0 if i =
̸ j.
Write
a = a 1 e1 + a 2 e2 + · · · + a n e n , b = b 1 e1 + b 2 e2 + · · · + b n en ,
and
⎡ ⎤ ⎡ ⎤
a1 b1
a2 b2
⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥
a=⎢ ..
⎥, b=⎢ ..
⎥.
. .
⎢ ⎥ ⎢ ⎥
⎣ ⎦ ⎣ ⎦
an bn
Now, the scalar product of the vectors is calculated as a matrix product:

⟩︁ def T
a · b = aT b = b a = a 1 b 1 + a 2 b 2 + · · · + a n b n .
⟨︁
It is seen that the scalar product is symmetric:

⟨︁ ⟩︁ ⟨︁ ⟩︁
a·b = b·a .
The vectorial product of vectors is more complicated. In two dimensions

it is a scalar:
⟨︁ ⟩︁ def
a × b = ∥a∥∥b∥ sin α,
2 This determinant is called the cofactor of that matrix element.
í Õ ! ¤. û
with α the oriented angle between the two vectors.

In three dimensions the vectorial product is itself a vector, of length
⃦⟨︁ ⟩︁⃦
⃦ a × b ⃦ = ∥a∥∥b∥ sin α,
which is perpendicular to both vectors a and b:

⟨︁ ⟩︁ ⟨︁ ⟩︁ ⟨︂⟨︁ ⟩︁ ⟩︂ ⟨︂⟨︁ ⟩︁ ⟩︂
a × b ⊥ a, a × b ⊥ b, ⇐⇒ a×b ·a = a × b · b = 0.
The vectorial product is antisymmetric:

⟨︁ ⟩︁ ⟨︁ ⟩︁
a×b = − b×a ,
⟨︁ ⟩︁
and thus a × a = 0. It needs an orientation, a corkscrew rule: if you turn
⟨︁ ⟩︁
a corkscrew from a to b, it will move forward in the direction of a × b .
{︁ }︁
For the orthonormal basis in three dimensions e1 , e2 , e3 this means
⟨︁ ⟩︁ ⟨︁ ⟩︁ ⟨︁ ⟩︁
e1 × e1 = e2 × e2 = e3 × e3 = 0,
⟨︁ ⟩︁ ⟨︁ ⟩︁ ⟨︁ ⟩︁
e1 × e2 = e3 , e2 × e3 = e1 , e3 × e1 = e2 ,
assuming that the basis vectors span a right-handed co-ordinate frame.

Calculation is done using determinants:
⎡ ⎤
e1 e2 e3
⟨︁ ⟩︁
a × b = det ⎣ a 1 a 2 a 3 ⎦ =
⎢ ⎥
b1 b2 b3
[︄ ]︄ [︄ ]︄ [︄ ]︄
a2 a3 a3 a1 a1 a2
= e1 det + e2 det + e3 det =
b2 b3 b3 b1 b1 b2
= (a 2 b 3 − a 3 b 2 ) e1 + (a 3 b 1 − a 1 b 3 ) e2 + (a 1 b 2 − a 2 b 1 ) e3 .
í Õ ! ¤. û
^ A short introduction to
magnetohydrodynamics
^ B.1 Plasma
A plasma is a gas in which the atoms have been stripped of part of
B
their electrons: ionisation. The negatively charged electrons are moving
around freely within the plasma, among the remaining, positively charged
atom remnants, or ions. In space, plasmas are commonly hot, with
temperatures varying from thousands to billions of degrees.
Plasmas in space are so extended and rarefied, that they behave like ohut
superconductors: electric currents run through them for a long time suprajohdin
without noticeable weakening. For this reason, a magnetic field also
cannot move though a plasma: matter and field lines move hand-in-hand.
This is called frozen-in magnetism or Alfvén’s theorem. If a plasma
is compressed, the magnetic field contained in it will grow stronger:
similarly if the plasma is stretched and rolled up like puff pastry, as is lehtitaikina
happening inside the Sun’s convection layer. This is how natural dynamos
work.
^ B.2 Maxwell’s equations

Maxwell’s field equations are1 1
⟨︁ ⟩︁ ∂E ⟨︁ ⟩︁ ∂B
∇ × B = µ 0 j + ϵ0 µ 0 , ∇×E = − ,
∂t ∂t (B.1)
⟨︁ ⟩︁ ⟨︁ ⟩︁ ρ
∇ · B = 0, ∇·E = ϵ .
0
(E electric field vector, B magnetic field vector, j vector of electric current
density, ρ density of electric charge, ϵ0 , µ0 constants of nature.)
1 This compact form of the equations we owe to Oliver Heaviside (1850–1925). Heaviside
was a telegraph researcher who developed the theory of electric circuits. He also
proposed the existence of the ionosphere as an explanation for the long range — beyond
the horizon — of radio waves.
– 521 –
522 B A SHORT INTRODUCTION TO MAGNETOHYDRODYNAMICS
The partial differential operators ∇× and ∇· are known in vector calcu-

roottori lus as curl and divergence operators, curl and div. They are defined as
follows:
⎡ ⎤
i j k
⟨︁ ⟩︁ def ⎢ ∂ ∂ ∂ ⎥
curl E = ∇ × E = det ⎢
⎣ ∂x
⎥=
∂y ∂z ⎦
Ex Ey Ez
∂E z ∂E y ∂E y ∂E x
(︃ )︃ (︃ )︃
∂E x ∂E z
(︂ )︂
= − i+ − j+ − k,
∂y ∂z ∂z ∂x ∂x ∂y
⟨︁ ⟩︁ def ∂E x ∂E y ∂E z
div E = ∇ · E = + + ,
∂x ∂y ∂z
{︁ }︁
ortonormaali in which i, j, k is the orthonormal basis of the ( x, y, z) co-ordinate frame,
kanta and
E = E x i + E y j + E z k.
^ B.3 “Frozen-in” magnetic field

In a superconductor there can be no net electric field integrated around a
loop — as it would cause an infinite electric current! So
˛
⟨︁ ⟩︁
E · t ds = 0,
∂S
in which t is the tangent vector of loop ∂S . The Stokes loop integral
theorem (Wikipedia, Kelvin–Stokes theorem) — also called the Kelvin–
2 Stokes theorem because Lord Kelvin2 apparently discovered it — says
⟨︁ ⟩︁
that this integral is the same as the surface integral of the curl ∇ × E
of E over the surface S , the edge of which is ∂S . That integral must thus
also vanish: ¨ ⟨︂ ⟩︂ ˛
⟨︁ ⟩︁ ⟨︁ ⟩︁
∇ × E · n dS = E · t ds = 0.
S ∂S
In this, n is the normal on the surface S . From the Maxwell–Faraday
equation B.1 follows now
¨ ⟨︂ ¨ ¨ ⟨︂
∂B ∂ ⟨︁ ∂n
⟩︂ ⟩︁ ⟩︂
· n dS = 0 =⇒ B · n dS = B· dS.
S ∂t S ∂t S ∂t
∂n
⟨︂ ⟩︂
Here, the integrand on the right, B · , vanishes if the loop surface S
∂t
∂n
is chosen to be fixed, as then = 0. Then, if also ∂S is constant:
∂t
¨ ¨
∂ ⟨︁ ⟩︁ ∂ ⟨︁ ⟩︁
B · n dS = 0 =⇒ B · n dS = 0.
S ∂t ∂t S
2 Sir William Thomson (1824–1907) FRS FRSE (Lord Kelvin) was a British physicist,
engineer and inventor. He was ennobled in 1866 for his work on the trans-Atlantic
telegraph cable.
í Õ ! ¤. û
History of the field
˜ ⟨︁ ⟩︁
B.4
523
The integral S B · n dS is the magnetic flux through the loop surface S , vuo
the total number of field lines passing though the loop, which thus must
be constant in time and cannot change. This is why a superconducting
sphere floats, or levitates, on a magnetic field: the field cannot penetrate
the sphere.
In a moving plasma, the assumption that ∂S is constant no longer
holds. We assert without proof (Wikipedia, Alfvén’s theorem) that the
above applies also in a plasma for an arbitrary loop connected with the
plasma: the plasma and the field are moving hand-in-hand — though the
plasma can freely flow along the field lines. If the plasma is compressed,
the magnetic field strength increases.3 3
^ B.4 History of the field

The term “plasma” was invented by Nobel laureate Irving Langmuir4 4
(1927). The scientific discipline studying plasmas is called magnetohydro-
dynamics (MHD). It has been estimated that, in the universe, over 99 % of
normal — not “dark” — matter is plasma.
A well-known magnetohydrodynamicist was the Swedish Nobel laure-
ate Hannes Alfvén, 1908–1995. He described in 1942 how in a supercon-
ductor the magnetic field lines move along with the matter.
1. In a vacuum, both electric charge ρ = 0 and electric current j = 0.
The first two Maxwell equations B.1 become
⟨︁ ⟩︁ ∂E ⟨︁ ⟩︁ ∂B
∇ × B = ϵ0 µ 0 , ∇×E = − . (B.2)
∂t ∂t
Letting the operator ∇× act from the left followed by substitution

yields
⟨︁ ⟩︁
⟨︂ ⟨︁ ⟩︁⟩︂ ∂ ∇×E ∂2 B
∇× ∇×B = ϵ0 µ 0 = −ϵ0 µ0 2 . (B.3)
∂t ∂t
3A spectacular example of this is the magnetic field of a neutron star or pulsar

(Wikipedia, Pulsar), the collapsed, still rapidly spinning core of a star.
4 Irving Langmuir (1881–1957) was an American chemist and physicist known for his
research on matter.
í Õ ! ¤. û
524 B A SHORT INTRODUCTION TO MAGNETOHYDRODYNAMICS
We now use a theorem from vector calculus,

⟨︂ ⟨︁ ⟩︁⟩︂ ⟨︁ ⟩︁ ⟨︁ ⟩︁
a× b×c = b a·c −c a·b ,
which in our case means

⟨︂ ⟨︁ ⟩︁⟩︂ ⟨︁ ⟩︁ ⟨︁ ⟩︁
∇× ∇×B = ∇ ∇ · B − ∇ · ∇ B.
⟨︁ ⟩︁
Here, according to the third Maxwell equation B.1, ∇ · B = 0, so
(︃ )︃
∂2 ∂2 ∂2
⟨︂ ⟨︁ ⟩︁⟩︂ ⟨︁ ⟩︁
∇× ∇×B = − ∇·∇ B = − + + B,
∂ x2 ∂ y2 ∂ z2
and, with equation B.3:

(︃ )︃
∂2 ∂2 ∂2 ∂2
+ 2 + 2 − ϵ0 µ 0 2 B = 0. (B.4)
∂x 2 ∂y ∂z ∂t
(a) Show by substitution that B( u) is a solution to this equation

if u = x − ct or u = x + ct. Use the chain rule.
(b) What is the value of parameter c?
(c) What other solutions can you think of?
(d) Now let equations B.2 be acted upon by the operator ∇× and
substitute the other way around to find a similar equation for
E.
(e) Alternatively, assume B = B( x, t) and decompose equation B.4:
(︃ )︃
∂2 ∂2
− ϵ0 µ 0 2 B=0
∂x 2 ∂t
∂ ⎷ ∂ ∂ ⎷ ∂
(︂ )︂(︂ )︂
=⇒ − ϵ0 µ 0 + ϵ0 µ 0 B = 0,
∂x ∂t ∂x ∂t
to which the separate equations
∂ ⎷ ∂ ∂ ⎷ ∂
(︂ )︂ (︂ )︂
− ϵ0 µ 0 B = 0, + ϵ0 µ 0 B=0
∂x ∂t ∂x ∂t
lead to the same solutions.

(f) [Advanced] Still alternatively, try separation of variables:
write
B( x, t) = B0 · X ( x) · T ( t),
and derive separate differential equations for the scalar func-

tions X ( x) and T ( t). Don’t be confused by the vector constant
B0 .
í Õ ! ¤. û
^ The Kepler orbital elements for
satellites
C
^ C.1 Angular elements describing the orientation in space of
the orbit
Ω The right ascension, or astronomical longitude, of the ascending rektaskensio
node. The zero point of this longitude is the place among the
stars where the plane of the ecliptic and the plane of the equator
intersect, and where the Sun crosses the equator at the beginning
of spring: the vernal equinox point. kevät-
päiväntasaus
i The inclination, the tilt angle of the orbital plane with respect
to the equatorial plane. For GPS satellites i = 55◦ . This is also
the highest latitude (north and south) where the satellite moves
through the zenith.
ω The argument of perigee. The angular distance, seen from the
Earth’s centre, between the ascending node and the perigee of the nouseva solmu
satellite orbit.
^ C.2 Elements describing the orbit’s size and shape

a The semi-major axis of the satellite orbit, the “mean radius” of the isoakselin
orbit. The mean height of the satellite is obtained from this by puolikas
subtracting the Earth’s radius.
⎷ /︂
e a2 − b2 a , in which
The eccentricity of the satellite orbit, e =
b is the semi-minor axis. This describes how much the Earth’s pikkuakselin
centre is “to the side” from the centre of the orbital ellipse, in other puolikas
words, how large is the height difference between the perigee (the
lowest point of the orbit) and apogee (the highest point).
– 525 –
526 C T HE K EPLER ORBITAL ELEMENTS FOR SATELLITES
^ C.3 Elements describing the satellite’s place in its orbit, its

“time table”
There are three alternatives:

luonnollinen ν( t) true anomaly. The direction angle of the satellite in the orbital
anomalia plane as seen from the Earth’s centre, reckoned from the perigee
in the direction of motion.
E ( t) eccentric anomaly. The direction angle of the satellite in the orbital
plane as seen from the centre of the ellipse, after the orbital ellipse
has been inflated into a circle.
M ( t) mean anomaly. The time that has elapsed since the last passage
through the perigee, in units of the satellite’s orbital period P .
Between these exist the following relationships, which may be used to

convert them into each other:
(︁ 1 )︁ √︃
tan ν( t ) 1+ e
(︁ 12 = , E ( t) = M ( t) + e sin E ( t).
1− e
)︁
tan 2 E ( t)
Study figure 17.14. E and ν are geometric quantities. The mean anomaly
M , on the other hand, is a mere measure of time, scaled to the period P of
the satellite, and referenced to its time of passage τ through the perigee:
def t−τ
M ( t) = 2π .
P
1. From the definition of inclination: “This is also the highest latitude
(north and south) where the satellite moves through the zenith.”
Does this precisely apply on a flattened Earth?
í Õ ! ¤. û
Bibliography
ABCDEFGHIJKLMNOPRSTUVW
A
Michael S. Adler. Wikimedia Commons, Total solar eclipse of 2017, 2017. URL
https://commons.wikimedia.org/wiki/File:Total_Solar_Eclipse_8-21-17.jpg. ©
2017 (CC BY-SA 4.0). Accessed 19th May, 2019. 471
Joel Ahola. Nuottavaaran siirroksen deformaatiotutkimus GPS-mittausten

avulla. Master’s thesis, Helsinki University of Technology, Department of
Surveying, Espoo, February 2001. 493
Associated Press. World War II planes found in Greenland in ice 260 feet deep.
The New York Times (archive), 4th August 1988. URL
https://www.nytimes.com/1988/08/04/us/world-war-ii-planes-found-in-
greenland-in-ice-260-feet-deep.html. Accessed 12th April, 2019. 512
AUSPOS – Online GPS Processing Service. Geoscience Australia. URL http:

//www.ga.gov.au/scientific-topics/positioning-navigation/geodesy/auspos.
Accessed 10th April, 2019. 355
Willem Baarda. A testing procedure for use in geodetic networks, volume 2

number 5 of Publications on geodesy, new series. Netherlands Geodetic
Commission, Delft, 1968. URL https://www.ncgeo.nl/downloads/09Baarda.pdf.
Accessed 14th May, 2019. 167, 405
Willem Baarda. S-transformations and criterion matrices, volume 5 number 1 of

Publications on geodesy, new series. Netherlands Geodetic Commission, Delft,
second revised edition, 1981. URL
https://www.ncgeo.nl/downloads/18Baarda.pdf. Accessed 2nd October, 2021.
365
BakerNunn.org. URL http://bakernunn.org. Accessed 19th May, 2019. 465
– 527 –
528 ABCDEFGHIJKLMNOPRSTUVW B IBLIOGRAPHY
Silvio A. Bedini. Along came a spider – Spinning silk for cross-hairs. The search
for cross-hairs for scientific instrumentation, Part 1. The American Surveyor,
March/April 2005a. URL
https://archive.amerisurv.com/PDF/TheAmericanSurveyor_BediniAlongCa
meASpider-SpinningSilkForCross-HairsPart1_March-April2005.pdf.
Accessed 14th May, 2019. 99
Silvio A. Bedini. Along came a spider – Spinning silk for cross-hairs. The search
for cross-hairs for scientific instrumentation, Part 2. The American Surveyor,
May 2005b. URL https://archive.amerisurv.com/PDF/TheAmericanSurveyo
r_BediniAlongCameASpiderPart2_May2005.pdf. Accessed 14th May, 2019. 99
Mirjam Bilker-Koivula and Matti Ollikainen. Suomen geoidimallit ja niiden

käyttäminen korkeuden muunnoksissa. Research note (in Finnish) 29,
Finnish Geodetic Institute, 2009. URL https://www.maanmittauslaitos.fi/site
s/maanmittauslaitos.fi/files/fgi/GLtiedote29.pdf. Accessed 14th May, 2019. 279
BIPM, SI base units. Bureau International des Poids et Mesures. URL

https://www.bipm.org/en/measurement-units/si-base-units. Accessed 14th
May, 2019. 24
Boller and Chivens, Baker Nunn. URL

https://bollerandchivens.com/?page_id=1455. Accessed 14th May, 2019. 465
Ilmari Bonsdorff. Geodeettisen laitoksen vuosikertomus ajalta heinäkuu 1918 –

joulukuu 1919. Annual report, Finnish Geodetic Institute, 1920. © 1920
Geodeettinen laitos. 211, 215
Kai Borre. GPS Easy Suite II — A MATLAB Companion. Inside GNSS,

March – April 2009. URL
http://www.insidegnss.com/auto/marapr09-borre.pdf. Accessed 7th April, 2019.
313
Claude Boucher and Zuheir Altamimi. Memo : Specifications for reference

frame fixing in the analysis of a EUREF GPS campaign, 2007. URL
http://users.auth.gr/kvek/20070327-MEMO-ver6.pdf. Accessed 11th March,
2019. 281
Nathaniel Bowditch, LL. D. American Practical Navigator. National

Geospatial-Intelligence Agency, Springfield, Virginia, 2017. URL
https://msi.nga.mil/NGAPortal/MSI.portal?_nfpb=true&_pageLabel=msi_p
ortal_page_62&pubCode=0002. Accessed 14th May, 2019. 157, 170
Kenneth P. Burnham and David R. Anderson. Model Selection and Multimodel

Inference: A Practical Information-theoretic Approach. Springer Science and
Business Media, New York, 2013. ISBN 978-1-4757-2917-7. 395
í ¤. û
B IBLIOGRAPHY ABCDEFGHIJKLMNOPRSTUVW
529
C
Bradley Darren Carter, Michael C. B. Ashley, Yin-Sheng Sun, and John W. V.

Storey. Redesigning a Baker-Nunn Camera for CCD imaging. Proceedings
Astronomical Society of Australia, (1):74–76, 1992. URL
https://newt.phys.unsw.edu.au/~mcba/pubs/carter92.pdf. Accessed 14th May,
2019. 465
CHAMP Mission. CHAMP – CHAllenging Minisatellite Payload. Deutsches

Geoforschungszentrum, Helmholtz-Zentrum Potsdam. URL
https://www.gfz-potsdam.de/champ/. Accessed 12th April, 2019. 494
Roger Cicala. Iceland spar: The rock that discovered optics. Lensrentals, 11th
December 2013. URL https://www.lensrentals.com/blog/2013/12/iceland-spar-
the-rock-that-discovered-optics/. Accessed 8th April, 2019. 191
Michael Alan Ralph Cooper. Modern Theodolites and Levels. Granada

Publishing Ltd, London, second edition, 1982. ISBN 978-0-2461-1502-7. 170
Michael Alan Ralph Cooper. Control Surveys in Civil Engineering. Collins,

London, 1987. ISBN 978-0-8939-7272-1. 418, 420, 423
Nicholas Crane. Mercator – The Man who Mapped the Planet. Weidenfeld and
Nicholson, London, 2002. ISBN 978-0-8050-6624-1. 3
Wesley G. Crawford, RPLS. Back to basics: Quick setup with a laser plummet.
Point of Beginning, 1st January 2009. URL
https://www.pobonline.com/articles/93629-back-to-basics-quick-setup-with-
a-laser-plummet. Accessed 14th May, 2019. 134
Arpád Czobor and Zsuzsanna Németh. Balloon-borne beacons for stellar

triangulation. Advances in Space Research, 1(11):29–34, 1981. URL
https://doi.org/10.1016/0273-1177(81)90445-2. Accessed 14th May, 2019. 462
Sönke Dangendorf, Marta Marcos, Guy Wöppelmann, Clinton P. Conrad,

Thomas Frederikse, and Riccardo Riva. Reassessment of 20th century global
mean sea level rise. Proceedings of the National Academy of Sciences, 114(23):
5946–5951, 2017. URL https://doi.org/10.1073/pnas.1616007114. Accessed
14th May, 2019. 509
Albert Dastre. Questions scientifiques - L’Heure légale, volume 4e période, tome

148. 1898. URL https://fr.wikisource.org/wiki/Questions_scientifiques_-
_L%E2%80%99Heure_l%C3%A9gale/01. Accessed 4th March, 2020. 53
Charles DeMets, Richard G. Gordon, Donald F. Argus, and Seth Stein. Effect of
recent revisions to the geomagnetic reversal time scale on estimates of
í ¤. û
current plate motions. Geophysical Research Letters, 21(20):2191–2194, 1994.

URL https://doi.org/10.1029/94GL02118. Accessed 14th May, 2019. 487
Deutsche Fotothek. Rechenbrett, Vermessung des Inhalts eines Fasses, 1523.

URL http://www.deutschefotothek.de/documents/obj/88963434. Accessed 18th
November, 2019. 251
John Louis Emil Dreyer. The Well of Eratosthenes. The Observatory, 37:
352–353, 1914. URL http:
//articles.adsabs.harvard.edu//full/1914Obs....37..352D/0000352.000.html.
DuPont Teijin Films. Mylar polyester film, Safety in Handling and Use. URL
https://web.archive.org/web/20201027150459/https://stenbacka.fi/wp-
content/uploads/sites/3/2016/07/mylar_a_perustieto.pdf. Accessed 19th
November, 2021. 464
Valerie J. Easton and John H. McColl. STEPS Statistics Glossary v1.1. URL
http://www.stats.gla.ac.uk/steps/glossary/probability_distributions.html.
R. Edgeworth, B. J. Dalton, and Thomas Parnell. The pitch drop experiment.

European Journal of Physics, pages 198–200, 1984. URL
https://smp.uq.edu.au/files/109/outreach-pitch-drop-abstract.pdf. Accessed
12th August, 2019. 489
Jürgen Ehlers, Philip Hughes, and Philip L. Gibbard. The Ice Age.
Wiley-Blackwell, 2015. ISBN 978-1-118-50781-0. 491
Eiffel Tower, 72 names. List of the 72 names on the Eiffel Tower. URL
https://en.wikipedia.org/wiki/List_of_the_72_names_on_the_Eiffel_Tower.
Accessed 7th April, 2019. 194, 330, 362, 453, 500, 508
Anon. Elements of surveying. Headquarters, Department of the Army, 1971.

URL https://play.google.com/books/reader?id=564XAAAAYAAJ. Accessed
14th March, 2019. 138
Loránd Eötvös. Three Fundamental Papers of Loránd Eötvös. Loránd Eötvös

Geophysical Institute of Hungary, 1998. ISBN 963-7135-02-2. Ed. Zoltán
Szabó. 452
EUREF. EUREF, Reference Frame Sub Commission for Europe. URL

http://www.euref.eu/. Accessed 14th April, 2019. 57
í ¤. û
531
F
Eric Fielding. Introduction to SAR Interferometry, 2017. URL https:

//appliedsciences.nasa.gov/sites/default/files/Session4-SAR-English_0.pdf.
Accessed 22nd September, 2021. 494
Finnish National Land Survey, About positioning services. URL

https://www.maanmittauslaitos.fi/en/maps-and-spatial-data/expert-
users/positioning-services. Accessed 10th April, 2019. 353
Finnish National Land Survey, Elevation model 10 m. URL

users/product-descriptions/elevation-model-10-m. Accessed 9th April, 2019.
252
Finnish National Land Survey, Elevation model 2 m. URL

users/product-descriptions/elevation-model-2-m. Accessed 9th April, 2019.
252
Finnish National Land Survey, Paikkatietoikkuna. URL

https://kartta.paikkatietoikkuna.fi/?lang=en. Accessed 29th September, 2021.
86, 87
Finnish National Land Survey, RINEX service. URL

https://www.maanmittauslaitos.fi/en/maps-and-spatial-data/positioning-
services/rinex-palvelu. Accessed 10th April, 2019. 354
Octavi Fors, Jorge Núñez, José Luis Muiños, Francisco Javier Montojo, Roberto
Baena-Gallé, Jaime Boloix, Ricardo Morcillo, María Teresa Merino, Elwood C.
Downey, and Michael J. Mazur. Telescope Fabra ROA Montsec: A new robotic
wide field Baker-Nunn facility. Publications of the Astronomical Society of the
Pacific, 125(927):522–538, May 2013. URL https://doi.org/10.1086/670941.
465
Clemens Franz. Wikimedia Commons, Der Meridianstein in Hammerfest,

Norwegen, 2005. URL
https://commons.wikimedia.org/wiki/File:Hammerfest_Meridianstein.jpg. ©
2005 Clemens Franz (CC BY-SA 3.0). Accessed 13th April, 2019. 10
Joshua Fraser, Anoop Haridas, Guna Seetharaman, Raghuveer M. Rao, and

Kannappan Palaniappan. KOLAM: a cross-platform architecture for scalable
visualization and tracking in wide-area imagery. Proceedings of the SPIE,
8747, 2013. URL https://doi.org/10.1117/12.2018162. Accessed 17th
September, 2021. 255
Jean-Pierre Friedelmeyer. Du côté des lettres (2) : une lettre de Sophie Germain
à Carl Friedrich Gauss (20 février 1807), et la réponse de celui-ci (30 avril
í ¤. û
1807), 2014. URL https://images.math.cnrs.fr/Du-cote-des-lettres-une-lettre-

de-Sophie-Germain-a-Carl-Friedrich-Gauss-20. Accessed 14th May, 2019.
357
GDGPS APPS. Automatic Precise Positioning Service of the Global Differential

GPS System. URL https://apps.gdgps.net/apps_file_upload.php. Accessed
10th April, 2019. 355
Geotrim, Trimnet. URL https://geotrim.fi/palvelut/trimnet-vrs/. Accessed 1st

March, 2020. 352
GLOBE Task Team and others. The Global Land One-km Base Elevation
Project, 1999. URL https://www.ngdc.noaa.gov/mgg/topo/globe.html. Accessed
9th April, 2019. 252
GRACE Follow-On Mission. Jet Propulsion Laboratory. URL

https://gracefo.jpl.nasa.gov/mission/overview/. Accessed 12th April, 2019. 496
GRACE Mission. Measuring Earth’s Surface Mass and Water Changes. Jet
Propulsion Laboratory. URL https://grace.jpl.nasa.gov/. Accessed 12th April,
2019. 496
GRACE Mission, hydrology. NASA. URL

https://commons.wikimedia.org/wiki/File:
Global_Gravity_Anomaly_Animation_over_LAND.gif. Accessed 12th April,
2019. 496
Erik W. Grafarend, Horst Kremers, Juhani Kakkuri, and Martin Vermeer.

Modelling vertical refraction in the SW Finland Triangular Network
TAGNET adjustment. Report 87:2, Finnish Geodetic Institute, Helsinki,
1987. 180, 182
Erik W. Grafarend, Rey-Jer You, and Rainer Syffus. Map Projections:

Cartographic Information Systems. Springer Verlag, second edition, 2014.
ISBN 978-3642364938. 55
Richard S. Gross. The excitation of the Chandler wobble. Geophysical Research

Letters, 27:2329–2332, 2000. URL https://doi.org/10.1029/2000GL011450.
Thomas Gruber, Reiner Rummel, Johannes Ihde, Gunter Liebsch, Axel Rülke,
Uwe Schäfer, Michael Sideris, Elena Rangelova, and Philip Woodworth.
Scientific Roadmap towards Height System Unification with GOCE. Slide
show, GOCE 5th International GOCE User Workshop, Paris, 2014. URL
https://slideplayer.com/slide/10319716/. Accessed 17th September, 2021. 95
í ¤. û
533
GUM. Evaluation of measurement data – Guide to the expression of
uncertainty in measurement. Report JGCM 100:2008, Working Group 1,
Joint Committee for Guides in Metrology, 2008. URL https:
//www.bipm.org/utils/common/documents/jcgm/JCGM_100_2008_E.pdf.
Accessed 17th August, 2019. 30
Werner Gurtner and Lou Estey. RINEX, The Receiver Independent Exchange
Format v. 3.00. 28th November 2007. URL
https://epic.awi.de/id/eprint/29985/1/Gur2007a.pdf. Accessed 9th April, 2019.
302, 353
Pasi Häkli, Jyrki Puupponen, Hannu Koivula, and Markku Poutanen. Suomen
geodeettiset koordinaatistot ja niiden väliset muunnokset. Research note (in
Finnish) 30, Finnish Geodetic Institute, Masala, 2009. URL https://www.ma
anmittauslaitos.fi/sites/maanmittauslaitos.fi/files/fgi/GLtiedote30.pdf.
Accessed 14th May, 2019. 57, 218, 280
Jonathan Hardis. Why is a second defined as "the duration of 9,192,631,770

periods of the radiation corresponding to the transition between the two
hyperfine levels of the ground state of the caesium 133 atom."? Why not
9,192,632,000 or other numbers? Quora answer, 2018. URL
https://www.quora.com/Why-is-a-second-defined-as-the-duration-of-9-192-
631-770-periods-of-the-radiation-corresponding-to-the-transition-between-
the-two-hyperfine-levels-of-the-ground-state-of-the-caesium-133-atom-
Why-not-9-192-632-000-or-other-numbers. Accessed 8th April, 2019. 26
Paul Harnesk and Åke Davidsson. Vem är Vem? Stor-Stockholm 1962. Project
Runeberg, 1962. URL http://runeberg.org/vemarvem/sthlm62/0152.html.
Accessed 17th November, 2019. 185
Markku Heikkinen. Solving the shape of the Earth by using digital density
models. Report 81:2, Finnish Geodetic Institute, Helsinki, 1981. 431
Veikko A. Heiskanen. Kenttämittaus ja kartoitus. Werner Söderström Oy,

second, expanded edition, 1943. 140
Veikko A. Heiskanen and Seppo Härmälä. Maastomittaus ja kartoitus. Werner

Söderström Oy, fifth, unchanged edition, 1963. 16, 18, 95
Weikko A. Heiskanen and Helmut Moritz. Physical Geodesy. W. H. Freeman

and Company, San Francisco, 1967. i, 13, 267, 435, 437, 444
Weikko A. Heiskanen and Felix A. Vening Meinesz. The Earth and its Gravity
Field. McGraw-Hill, New York, 1958. 435
í ¤. û
Hexagon Geosystems, HxGN SmartNet. URL https:///www.hxgnsmartnet.com.

Accessed 4th March, 2020. 353
Nicolàs de Hilster. Geodetic instruments. URL

http://dehilster.info/geodetic_instruments/1980_sat-sagem_AGA-Minilir.php.
© Nicolàs de Hilster. Accessed 9th April, 2019. 248
Bernard Hofmann-Wellenhof, Herbert Lichtenegger, and James Collins. GPS

Theory and Practice. Springer-Verlag, Vienna, fifth corrected edition, 2001.
ISBN 978-3-2118-3534-0. i, 283, 298, 326, 328
Humboldt University Berlin. Friedrich Robert Helmert, 2017. URL https://www.

researchgate.net/publication/318994932_Friedrich_Robert_Helmert_found
er_of_modern_geodesy_on_the_occasion_of_the_centenary_of_his_death. ©
2017 Humboldt-Universität zu Berlin, Universitätsbibliothek (CC BY 3.0).
Anon. IAG Resolutions adopted at the XXIIth General Assembly in

Birmingham, 1999. URL
https://iag.dgfi.tum.de/fileadmin/IAG-docs/IAG_Resolutions_1999.pdf.
Resolution 3. Accessed 14th May, 2019. 202
IGS Central Bureau. URL

https://igs.org/governance-management#central-bureau. Accessed 17th
IMCCE, Milanković app. IMCCE, Paris Observatory. URL

http://vo.imcce.fr/insola/earth/online/earth/online/index.php. Accessed 12th
April, 2019. 504
INSPIRE Knowledge Base. European Commission. URL

https://inspire.ec.europa.eu. Accessed 9th April, 2019. 252
The Institute of Australian Culture. The Old Survey [song, 1905], 2012. URL
http://www.australianculture.org/the-old-survey-1905/. Accessed 19th August,
2019. 119
International DORIS Service. URL https://ids-doris.org/. Accessed 12th April,

2019. 480
IPCC. Climate Change 2007: The physical science basis. Contribution of

Working Group I to the Fourth Assessment Report of the Intergovernmental
Panel on Climate Change. In Susan Solomon, Dahe Qin, Martin R. Manning,
Zhenlin Chen, Melinda Marquis, Kirsten B. Averyt, Melinda M. B. Tignor,
and Henry L. Miller, editors, Fourth Assessment Report of the
Intergovernmental Panel on Climate Change. Cambridge Univ. Press,
Cambridge UK, 2007. URL https://www.ipcc.ch/report/ar4/wg1/. 512
í ¤. û
535
IUGG, GGOS. Global Geodetic Observing System. URL https://ggos.org.
Haseeb Jamal. Compound curves in engineering survey, 2017. URL

https://www.aboutcivil.org/compound-curves-in-engineering-surveying.html.
Accessed 14th May, 2019. 241, 242
Harold Jeffreys. The Theory of Probability. Oxford Classic Texts in the Physical
Sciences. Oxford University Press, 1998. ISBN 978-0-1915-8967-6. Originally
published in © 1939. 395
Jean Jouzel and Valerie Masson-Delmotte. EPICA Dome C Ice Core 800KYr
deuterium data and temperature estimates. IGBP PAGES/World Data Center
for Paleoclimatology Data Contribution Series # 2007-091. NOAA/NCDC
Paleoclimatology Program, Boulder CO, USA., 2007. URL
https://doi.pangaea.de/10.1594/PANGAEA.683655. Accessed 11th March,
2019. 504
Jean Jouzel, Valerie Masson-Delmotte, Olivier Cattani, Gabrielle B. Dreyfus,

Sonia Falourd, Georg Hoffmann, Bénédicte Minster, Julius Nouet, Jean-Marc
Barnola, Jérôme Chappellaz, Hubertus Fischer, Jean-Charles Gallet, Sigfus
Johnsen, Markus Leuenberger, Laetitia Loulergue, Dieter Luethi,
Frederic Parrenin Hans Oerter, Grant M. Raisbeck, Dominique Raynaud,
Adrian Schilt, Jakob Schwander, Enricomaria Selmo, Roland Souchez, Renato
Spahni, Bernhard Stauffer, Jørgen Peder Steffensen, Barbara Stenni,
Thomas F. Stocker, Jean-Louis Tison, Martin Werner, and Eric W. Wolff.
Orbital and millennial Antarctic climate variability over the past 800,000
years. Science, 317(5839):793–796, 2007. URL
https://doi.org/10.1126/science.1141038. Accessed 4th March, 2020. 504
JPL, The Global Differential GPS System. URL https://www.gdgps.net.

JUHTA. JHS 163 Finland’s height system N2000. JHS Recommendation,

Advisory Committee on Information Management in Public Administration,
2010. URL https://www.suomidigi.fi/ohjeet-ja-tuki/jhs-suositukset/jhs-163-
suomen-korkeusjarjestelma-n2000. 213
JUHTA. JHS 178 Interface of geographic information services for local

authorities. JHS Recommendation, Advisory Committee on Information
Management in Public Administration, 2012a. URL
https://www.suomidigi.fi/ohjeet-ja-tuki/jhs-suositukset/jhs-178-kunnan-
paikkatietopalvelurajapinta-vanhentunut. 213
í ¤. û
JUHTA. JHS 184 Point measurement in the EUREF-FIN coordinate reference

system. JHS Recommendation, Advisory Committee on Information
Management in Public Administration, 2012b. URL
https://www.suomidigi.fi/ohjeet-ja-tuki/jhs-suositukset/jhs-184-
kiintopistemittaus-euref-fin-koordinaattijarjestelmassa. 213, 216, 218
JUHTA. JHS 185 Composing base maps for the city. JHS Recommendation,
2014. URL https://www.suomidigi.fi/ohjeet-ja-tuki/jhs-suositukset/jhs-185-
asemakaavan-pohjakartan-laatiminen. 213, 235
JUHTA. JHS 196 EUREF-FIN coordinates in Finland. JHS Recommendation,

2016a. URL https://www.suomidigi.fi/ohjeet-ja-tuki/jhs-suositukset/jhs-196-
euref-fin-jarjestelman-mukaiset-koordinaatit-suomessa. 213, 217
JUHTA. JHS 197 EUREF-FIN coordinate systems, related conversions and

map sheet distribution. JHS Recommendation, Advisory Committee on
Information Management in Public Administration, 2016b. URL
https://www.suomidigi.fi/ohjeet-ja-tuki/jhs-suositukset/jhs-197-euref-fin-
koordinaattijarjestelmat-niihin-liittyvat-muunnokset-ja-karttalehtijako. 61,
63, 213
Erkki Kääriäinen. The second levelling of Finland in 1935-1955. Publication 61,

Finnish Geodetic Institute, Helsinki, 1966. 112
Anon. Kaavoitusmittausohjeet 2003. Publication 94, Finnish National Land

Survey, 31st January, 2003. URL https://www.maanmittauslaitos.fi/sites/ma
anmittauslaitos.fi/files/old/kaavoitusmittausohjeet_2003_0.pdf. Accessed 22nd
Heribert Kahmen and Wolfgang Faig. Surveying. Walter de Gruyter, Berlin –

New York, 1988. ISBN 978-3-1100-8303-3. i, 27, 106, 124, 134, 140, 148, 180,
187, 222, 224, 256, 260, 390
Juhani Kakkuri. Stellar Triangulation with Balloon-borne Beacons and

Satellites. PhD thesis, University of Helsinki, 1973. 462
Juhani Kakkuri, Teuvo Parm, Vidal Ashkenazi, and S. A. Crane. Accuracy

analysis of the Finnish laser geodimeter traverse. In Proc., Int. Symp. on
Geodetic Networks and Computations, volume 5, pages 85–93, Munich, 1981.
Verlag der Bayerischen Akademie der Wissenschaften. 206
Ulla Kallio. Tasoituslasku. TKY 587. Otatieto, third edition, 1998. ISBN
978-9-5167-2267-5. i
í ¤. û
537
Kari-Pekka Karlsson. Maastotietojärjestelmä 2030 -selvitys. Final report,
Ministry of Agriculture and Forestry, 2015. URL
http://mmm.fi/documents/1410837/1801204/Maastotietojarjestelma_2030_-
selvitys_Loppuraportti_28.2.2015.pdf. Accessed 24th November, 2021. 231
Charles Karney. Online Geoid Calculator. URL

https://geographiclib.sourceforge.io/cgi-bin/GeoidEval. Accessed 14th April,
2019. 118
Shfaqat Abbas Khan, Lin Liu, John Wahr, Ian Howat, Ian Joughin, Tonie van
Dam, and Kevin Fleming. GPS measurements of crustal uplift near
Jakobshavn Isbræ due to glacial ice mass loss. Journal of Geophysical
Research: Solid Earth, 115(B9):B09405, 2010. URL
http://dx.doi.org/10.1029/2010JB007490. Accessed 14th May, 2019. 506
Hannu Koivula and Jaakko Mäkinen. Geodetic activities at Finnish Antarctic

research station Aboa. In Fifth International Antarctic Geodesy Symposium
AGS’03, Lviv, Ukraine, Sept. 15–17, 2003. URL
https://www.researchgate.net/publication/237467751_GEODETIC_ACTIVIT
IES_AT_FINNISH_ANTARCTIC_RESEARCH_STATION_ABOA. Accessed
14th May, 2019. 506
Hannu Koivula, Jaakko Kuokkanen, Simo Marila, Sonja Lahtinen, and Tuukka
Mattila. Assessment of sparse GNSS network for network RTK. Journal of
Geodetic Science, 8(1):136–144, 2018. URL
https://doi.org/10.1515/jogs-2018-0014. Accessed 15th August, 2019. 351
Yoshihide Kozai. The motion of a close Earth satellite. Astronomical Journal,

64(1274):367–377, 1959. URL https://doi.org/10.1086/107957. Accessed 14th
May, 2019. 328
T. J. Kukkamäki and Pekka Lehmuskoski. Influence of the earth magnetic field

on Zeiss Ni 002 levels. Report 84:1, Finnish Geodetic Institute, Helsinki,
1984. 106
Jacques Laskar, Philippe Robutel, Frédéric Joutel, Mickael Gastineau,

Alexandre C. M. Correia, and Benjamin Levrard. A long term numerical
solution for the insolation quantities of the Earth. Astronomy and
Astrophysics, 428:261–285, 2004. URL
https://doi.org/10.1051/0004-6361:20041335. Accessed 14th May, 2019. 504
Javier Arribas Lázaro. GNSS Array-based Acquisition: Theory and

Implementation. PhD thesis, Universitat Politécnica de Catalunya, 2012.
URL https://theses.eurasip.org/theses/449/gnss-array-based-acquisition-
theory-and/download/. Accessed 14th May, 2019. 300
í ¤. û
Rachel Leah. Rapper B.o.B. starts GoFundMe to prove earth is flat. Salon, 25th
September, 2017. URL https://www.salon.com/2017/09/25/b-o-b-flat-earth/.
Leica. TPS - System 1000 – Quick Start. Leica Geosystems AG, Heerbrugg,
Switzerland, 1997. URL http:
//www.engineeringsurveyor.com/software/leica/tps2000_quick_start_en.pdf.
Accessed 1st December, 2021. 152
Cherry L. E. Lewis. The Dating Game. Cambridge University Press, 2000.

ISBN 978-1-1391-9713-7. 486
Martin Lidberg, Jan M. Johansson, Hans-Georg Scherneck, and Glenn A. Milne.

Recent results based on continuous GPS observations of the GIA process in
Fennoscandia from BIFROST. Journal of Geodynamics, 50(1):8–18, July 2009.
URL https://doi.org/10.1016/j.jog.2009.11.010. Accessed 14th May, 2019. 491
Jaakko Mäkinen, Andreas Engfeldt, Linda Engman, Bjørn Geir Harsson, Tõnis
Oja, Sven Rekkedal, Knut Røthing, Paavo Rouhiainen, Hannu Ruotsalainen,
H. Skatt, Gabriel Strykowski, Heikki Virtanen, Karin Wieczerkowski, and
Detlef Wolf. The Fennoscandian Land Uplift Gravity Lines: Comparison of
observed gravity change with observed vertical motion and with GIA models.
Report, Nordiska Kommissionen för Geodesi, 2010. URL
https://www.nordicgeodeticcommission.com/wp-content/uploads/2014/10/1-
Makinen_et_al_land_uplift_gravity_lines.pdf. Accessed 11th March, 2019.
492
Stephen Malys, John H. Seago, Nikolaos K. Pavlis, P. Kenneth Seidelmann, and

George H. Kaplan. Why the Greenwich meridian moved. Journal of Geodesy,
89(12):1263–1272, 2015. URL https://doi.org/10.1007/s00190-015-0844-y.
Accessed 14th May, 2019.
Stephen Malys, Robert Wong, and Scott A. True. The WGS84 Terrestrial
Reference Frame in 2016. Presented at eleventh meeting of the International
Committee on GNSS, November 2016, Sochi, Russia, 2016. URL
https://www.unoosa.org/pdf/icg/2016/icg11/wgd/02wgd.pdf. Accessed 17th
September, 2021. 58
Pierre Louis Moreau de Maupertuis. La Figure de la Terre, Determinée par les

observations. . . . Jean Catuffe, Amsterdam, 1738. URL
https://www.doria.fi/bitstream/handle/10024/69470/001.pdf. Accessed 7th
April, 2019. 1, 8
Ministry of the Environment, Legislation on land use and building. URL

https://ym.fi/en/land-use-and-building-act. Accessed 17th September, 2021.
235
í ¤. û
539
Patrap Misra and Per Enge. Global Positioning System – Signals,
Measurements, and Performance. Ganga-Jamuna Press, revised second
edition, 2010. ISBN 978-0-9709-5442-8. 289
Soumen Mondal, K. G. Gupta, Sneh Lata, Biman J. Medhi, Tarun Bangia, T. S.

Kumar, Shobhit Yadav, and S. K. Singh. Development of ARIES Baker-Nunn
camera to a wide-field Imaging Telescope with CCD. arXiv, 2009. URL
https://arxiv.org/pdf/0905.0361.pdf. Accessed 14th May, 2019. 465
Oliver Montenbruck, Peter Steigenberger, and André Hauschild. Broadcast

versus precise ephemerides: A multi-GNSS perspective. GPS Solutions, 19(2):
321–333, 2015. URL https://doi.org/10.1007/s10291-014-0390-8. Accessed
14th May, 2019. 341
Walter Heinrich Munk. The U.S. Commission on Ocean Policy, Testimony, 12th
April, 2002. URL https://govinfo.library.unt.edu/oceancommission/meetings/a
pr18_19_02/munk_statement.pdf. Accessed 14th May, 2019. 479
NASA JPL, TOPEX/Poseidon, Jason. URL https://sealevel.jpl.nasa.gov/.

NASA Solar Dynamics Observatory, Atmospheric Imaging Assembly. URL

https://sdo.gsfc.nasa.gov/assets/img/browse/2013/07/15/20130715_001424_4
096_0171.jpg. Accessed 19th May, 2019. 472
NGS, Computation utilities. URL

https://geodesy.noaa.gov/TOOLS/Inv_Fwd/Inv_Fwd.html. Accessed 14th
September, 2020. 68, 86
NOAA et al. ETOPO1 Global Relief Model, 2008. URL

http://dx.doi.org/10.7289/V5C8276M. Accessed 14th May, 2019. 252
NOAA, Ocean currents. How does the ocean affect climate and weather on land?
URL https://oceanexplorer.noaa.gov/facts/climate.html. © NOAA. Accessed
12th November, 2019. 499
GNU Octave. URL https://www.gnu.org/software/octave/. Accessed 13th April,

2019. 333
OGC, Catalogue Service. Open Geospatial Consortium. URL

https://www.ogc.org/standards/cat. Accessed 9th April, 2019. 231
Victor Rafael Ölander. Trigonometriset korkeusmittaukset pitkillä

välimatkoilla ja niiden tarkkuus. Maanmittaus, 1932. 182
í ¤. û
Matti Ollikainen. GPS-koordinaattien muuntaminen kartastokoordinaateiksi.

Research note (in Finnish) 8, Finnish Geodetic Institute, Helsinki, 1993. 279,
280
Boudewijn Overgaauw, Boudewijn A. C. Ambrosius, and Karel F. Wakker.

Analysis of the EUREF-89 GPS data from the SLR/VLBI sites. Bulletin
Géodésique, 68:19–28, 1994. URL https://doi.org/10.1007/BF00806749.
Teuvo Parm. Kansallisen koordinaattijärjestelmän luominen Suomessa.

Maanmittaus, 63(1), 1988. 56
Jerry Penry and David Lee Ingram. The American theodolite. The American
Surveyor, 10(9), 2013. URL
https://archive.amerisurv.com/PDF/TheAmericanSurveyor_PenryIngram-
TheAmericanTheodolite_Vol10No9.pdf. Accessed 22nd September, 2021. 137
William B. Penzes. Time Line for the Definition of the Meter, undated. URL
https://www.nist.gov/system/files/documents/pml/div683/museum-
timeline.pdf. Accessed 17th September, 2021. 26
Mika Pitkä. Sataman tehokkuuden osatekijät – tarkastelussa lastauksen ja

purun automatisointi. Bacherlor’s thesis, Kymenlaakso University of Applied
Sciences, 2009. 247
Porvoo. Koordinaatti- ja korkeusjärjestelmä. URL

https://www.porvoo.fi/koordinaatti-ja-korkeusjarjestelma. © City of Porvoo.
Robert Pound and Glen A. Rebka, Jr. Gravitational red-shift in nuclear

resonance. Physical Review Letters, 3(9):439–441, 1959. URL
https://doi.org/10.1103/PhysRevLett.3.439. Accessed 14th May, 2019. 190
Markku Poutanen. GPS-paikanmääritys. Ursa, Helsinki, 1998. ISBN

951-9269-89-4. Ursan julkaisuja 64. i
Markku Poutanen. Satelliittipaikannus. Ursa, Helsinki, 2017. ISBN

978-9-5259-8541-2. i, 283, 298, 326, 328
Jyrki Puupponen and Jaakko Järvinen. Historical video documentation of

triangulation measurements in Finland. In Integrating Generations, FIG
Working Week, Stockholm, Sweden, 14th – 19th June, 2008. URL
https://www.fig.net/resources/proceedings/fig_proceedings/fig2008/papers/hs
05/hs05_02_puupponen_jarvinen_2934.pdf. Accessed 8th April, 2019. 57
í ¤. û
541
R
RAKLI ry. Kiinteistöalan yhteiskunnallinen ja kansantaloudellinen merkitys.

Report, Finnish Association of Building Owners and Construction Clients
RAKLI, 2014. URL https://kti.fi/wp-content/uploads/Kiinteist%C3%B6alan-
yhteiskunnallinen-ja-kansantaloudellinen-merkitys-2014.pdf. Accessed 17th
September, 2021. 16
Reginald Reagan. One Matrix, Inverted, 26th July 2021. URL

https://raginrayguns.tumblr.com/post/657824795686207489/david-salsburg-
the-lady-tasting-tea-during. Accessed 17th September, 2021. 517
Jan Reijnoudt. Het hele IJsselmeer waterpas, 1996. URL

https://www.digibron.nl/viewer/collectie/Digibron/id/tag:RD.nl,19960213:
newsml_9f3bdeec998e02ec9e49004d5e149c1661831c48735609a683bd383e.
ROB, EUREF Permanent GNSS Network. Royal Observatory of Belgium. URL

http://www.epncb.oma.be/_networkdata/stationlist.php. Accessed 10th April,
2019. 86, 354
Rovaniemi, The Degree Measurement Expedition. The degree measurements by

de Maupertuis in the Tornionlaakso valley 1736–1737. City of Rovaniemi.
URL http://lapinkavijat.rovaniemi.fi/maupertuis/index_eng.html. Accessed
16th May, 2019. 8
The R Project for Statistical Computing. R Foundation. URL

https://www.r-project.org/. Accessed 13th April, 2019. 333
Jean M. Rüeger. Electronic Distance Measurement – An Introduction.

Springer-Verlag, Berlin Heidelberg, fourth edition, 1996. 202, 204, 205, 307
Jean M. Rüeger. Refractive indices of light, infrared and radio waves in the
atmosphere. UNISURV report S 68, School of Surveying and Spatial
Information Systems, University of New South Wales, 2002. URL https://ww
w.sage.unsw.edu.au/sites/sage/files/SAGE_collection/SpecialSeries/s68.pdf.
Accessed 4th October, 2021. 204, 307
Reiner Rummel and Fernando Sansò, editors. Satellite Altimetry in Geodesy

and Oceanography, volume 50 of Lecture Notes in Earth Sciences, 1993.
Springer-Verlag. 506
Hannu Salmenperä. Runko- ja kartoitusmittaukset. Report 4, Tampere

University of Technology, Rakennustekniikan osasto, Geoinformatiikan
laboratorio, Tampere, 1998. i, 213, 234
í ¤. û
Sanastokeskus TSK. Geoinformatiikan sanasto. TSK 51, Finnish National

Land Survey, 2018. URL
http://www.tsk.fi/tiedostot/pdf/GeoinformatiikanSanasto.pdf. Accessed 14th
May, 2019. 214
Gavin Schrock. CERN. xyHt magazine, 2014. URL

https://www.xyht.com/civiltransportation/cern-2/. Accessed 2nd July, 2019.
114
Scilab Enterprises. URL https://www.scilab.org/. Accessed 13th April, 2019. 333
Sea Level Research Group. University of Colorado. URL

https://sealevel.colorado.edu/. Accessed 12th April, 2019. 510
Günter Seeber. Satellitengeodäsie. Grundlagen, Methoden und Anwendungen.

De Gruyter, Berlin New York, 1989. ISBN 978-3-1101-0082-2. 306
Shuttle Radar Topography Mission, 2014. URL https://www2.jpl.nasa.gov/srtm/.

Juho Simonen. Maalaserkeilaus infrarakentamisen mittauksissa. Master’s

thesis, Aalto-yliopisto, 2012. 138
Dava Sobel. Longitude: The True Story of a Lone Genius Who Solved the
Greatest Scientific Problem of His Time. Walker and Company, New York,
1998. ISBN 978-1-8570-2571-2. Paperback. 461, 462
SOHO images. ESA and NASA. URL

https://sohowww.nascom.nasa.gov/data/realtime-images.html. Accessed 5th
March, 2020. 473
SOHO MPEG movies. ESA and NASA. URL

https://sohowww.nascom.nasa.gov/data/realtime/mpeg/. Accessed 5th March,
2020. 473
SOPAC, GPS Explorer. Scripps Orbit and Permanent Array Center. URL
http://sopac.ucsd.edu/dataBrowser.shtml. Accessed 5th March, 2020. 354
SOPAC, Hatanaka file compression. Scripps Orbit and Permanent Array Center.
URL http://sopac.ucsd.edu/hatanaka.shtml. Accessed 5th March, 2020. 354
Govert Strang van Hees. Stokes’ formula using fast Fourier techniques.
Manuscripta geodaetica, 15:235–239, 1990. URL
https://doi.org/10.1007/978-1-4612-3104-2_47. Condensed version. Accessed
14th May, 2019. 14
í ¤. û
543
T
Mikko Takalo. Computer-supported Finnish EDM height traversing. Part I:

technical description. Surveying Science in Finland, 13(2):56–99, 1995. 95,
182
Mikko Takalo and Paavo Rouhiainen. On system calibration of digital level. In

Ingenieurvermessung 2004, 14th International Conference on Engineering
Surveying, Zürich, 15th – 19th March, 2004. URL
https://www.researchgate.net/publication/254398461_Development_of_a_S
ystem_Calibration_Comparator_for_Digital_Levels_in_Finland. Accessed
22nd September, 2021. 108
Mikko Takalo, Paavo Rouhiainen, Pekka Lehmuskoski, and Veikko Saaranen.

On calibration of Zeiss DiNi12. In FIG Working Week 2001, Seoul, Korea,
6th – 11th May, 2001. FIG Commission 5. URL https://www.researchgate.net
/publication/237495166_ON_CALIBRATION_OF_ZEISS_DINI12. Accessed
12th March, 2019. 108
Martti Tikka. Käytännön geodesia III. Mittausmenetelmät. Otakustantamo,

Espoo, 1985. i
Martti Tikka. Käytännön geodesia I. Mittaustekniikan perusteet ja

rakennusteknilliset sovellutukset. Otatieto, Helsinki, third revised edition,
1991. i
Waldo R. Tobler. A Computer Movie Simulating Urban Growth in the Detroit

Region. Economic Geography, 46, Supplement: Proceedings. International
Geographical Union. Commission on Quantitative Methods, 1970. URL
https://www.jstor.org/stable/143141. Accessed 22nd September, 2021. 335
Wolfgang Torge. Geodesy. Walter de Gruyter, Berlin – New York, third

completely revised and extended edition, 2001. ISBN 978-3-11-087995-7. i, 3,
14, 15
UNAVCO, EGM96 geoid calculator. URL

https://www.unavco.org/software/geodetic-utilities/geoid-height-
calculator/geoid-height-calculator.html. Accessed 21st September, 2021. 118
UNAVCO, plate-motion calculator. URL

https://www.unavco.org/software/geodetic-utilities/plate-motion-
calculator/plate-motion-calculator.html/models. Accessed 5th March, 2020.
488
USGS, GTOPO30. URL https://www.usgs.gov/centers/eros/science/usgs-eros-

archive-digital-elevation-global-30-arc-second-elevation-gtopo30?qt-
í ¤. û
science_center_objects=0#qt-science_center_objects. Accessed 9th April, 2019.

252
Yrjö Väisälä. Maan toinen kuu. Tähtitaivas, 6, 1946. 462
Peter Vaníček and Edward Krakiwsky. Geodesy – The Concepts. Elsevier

Science Publishers, Amsterdam, second edition, 1986. ISBN
978-0-4448-6149-8. i, 418
Vasamuseet, Deformation Monitoring. URL https:

//www.vasamuseet.se/en/research/support-vasa/deformation-monitoring.
Väylä, Radionavigaatiopalvelut. URL

https://vayla.fi/palveluntuottajat/ammattimerenkulku/liikkuminen-
vesivaylilla/radionavigaatiopalvelut. Accessed 22nd September, 2021. 352
Martin Vermeer. A new SEASAT altimetric geoid for the Baltic. Report 83:4,
Finnish Geodetic Institute, Helsinki, 1983. 511
David Vine. Island of Shame: The Secret History of the U.S. Military Base on
Diego Garcia. Princeton University Press, revised edition, 2011. ISBN
978-0-6911-4983-7. URL https:
//press.princeton.edu/books/paperback/9780691149837/island-of-shame.
Accessed 22nd September, 2021. 300, 302
Paul Wessel, W. H. F. Smith, Remco Scharroo, J. Luis, and F. Wobbe. Generic

Mapping Tools: Improved Version Released. EOS Trans. AGU, 94(45):
409–410, 2013. URL http://dx.doi.org/10.1002/2013EO450001. iii
Wikimedia Commons, A calcite crystal. A calcite crystal displays the double

refractive properties while sitting on a sheet of graph paper. URL
https://commons.wikimedia.org/wiki/File:Crystal_on_graph_paper.jpg. ©
2011 User:APNMJM (CC BY-SA 3.0). Accessed 14th April, 2019. 192
Wikimedia Commons, Corkscrew. URL

https://commons.wikimedia.org/wiki/File:Korkenzieher.jpg. © 2004 Horst
Frank (GFDL). Accessed 14th April, 2019. 267
Wikimedia Commons, Decca Navigator Mk 12. URL

https://commons.wikimedia.org/wiki/File:Decca_Navigator_Mk_12.jpg. ©
2005 User:Stahlkocher (GFDL). Accessed 14th April, 2019. 285
í ¤. û
545
Wikimedia Commons, Exploration theodolite. URL https:
//commons.wikimedia.org/wiki/File:Exploration_theodolite_img_1660.jpg. ©
2005 User:Rama (CeCILL). Accessed 14th April, 2019. 120
Wikimedia Commons, Plane table with alidade. Cut-away view of plane table
with alidade. URL
https://commons.wikimedia.org/wiki/File:Plane_table_cgs00426.jpg. © NOAA.
Wikimedia Commons, The prime meridian (Greenwich). URL

https://commons.wikimedia.org/wiki/File:Prime_meridian.jpg. © 2005
User:Prioryman (GFDL). Accessed 14th April, 2019. 48
Wikimedia Commons, Standard metre in Paris. One of the historical (18th

century), standard-meters (mètre-étalon) by Chalgrin located at 36, rue de
Vaugirard in Paris. URL https://commons.wikimedia.org/wiki/File:
M%C3%A8tre-%C3%A9talon_Paris.JPG. © 2010 User:LPLT (CC BY-SA 3.0).
Wikipedia, 1462 Zamenhof. URL https://en.wikipedia.org/wiki/1462_Zamenhof.

Accessed 26th August, 2019. 457
Wikipedia, 3D rotation group, Topology. URL

https://en.wikipedia.org/wiki/3D_rotation_group#Topology. Accessed 8th
April, 2019. 136
Wikipedia, Alfvén’s theorem. URL

https://en.wikipedia.org/wiki/Alfv%C3%A9n’s_theorem. Accessed 1st
Wikipedia, The AN/UYK-1 (TRW-130) Computer. URL

https://en.wikipedia.org/wiki/Transit_(satellite)#The_AN/UYK-1_(TRW-
130)_Computer. Accessed 1st July, 2019. 287
Wikipedia, Bonferroni correction. URL

https://en.wikipedia.org/wiki/Bonferroni_correction. Accessed 10th April,
2019. 412
Wikipedia, Chemical polarity. URL

https://en.wikipedia.org/wiki/Chemical_polarity. Accessed 9th April, 2019.
203
Wikipedia, Dandelin spheres. URL

https://en.wikipedia.org/wiki/Dandelin_spheres. Accessed 12th April, 2019.
474
Wikipedia, Discrete wavelet transform. URL

https://en.wikipedia.org/wiki/Discrete_wavelet_transform. Accessed 9th April,
2019. 255
í ¤. û
Wikipedia, Earth’s internal heat budget. URL

https://en.wikipedia.org/wiki/Earth’s_internal_heat_budget. Accessed 12th
April, 2019. 487
Wikipedia, Euler spiral. URL https://en.wikipedia.org/wiki/Euler_spiral.

Wikipedia, Gold code. URL https://en.wikipedia.org/wiki/Gold_code. Accessed

9th April, 2019. 293
Wikipedia, Groundwater-related subsidence. URL

https://en.wikipedia.org/wiki/Groundwater-related_subsidence. Accessed 12th
April, 2019. 492
Wikipedia, Kelvin–Stokes theorem. URL

https://en.wikipedia.org/wiki/Stokes%27_theorem. Accessed 12th April, 2019.
522
Wikipedia, Kilobyte. URL https://en.wikipedia.org/wiki/Kilobyte. Accessed 8th

April, 2019. 25
Wikipedia, The LAGEOS time capsule. URL

https://en.wikipedia.org/wiki/LAGEOS#Time_capsule. Accessed 12th April,
2019. 481
Wikipedia, Leap second. URL https://en.wikipedia.org/wiki/Leap_second.

Wikipedia, Magnetosphere. URL https://en.wikipedia.org/wiki/Magnetosphere.

Wikipedia, Non-SI units mentioned in the SI. URL

https://en.wikipedia.org/wiki/Non-SI_units_mentioned_in_the_SI. Accessed
8th April, 2019. 25
Wikipedia, Photon polarization. URL

https://en.wikipedia.org/wiki/Photon_polarization. Accessed 8th April, 2019.
191
Wikipedia, Planimeter. URL https://en.wikipedia.org/wiki/Planimeter. Accessed

9th April, 2019. 259
Wikipedia, Pulsar. URL https://en.wikipedia.org/wiki/Pulsar. Accessed 9th

April, 2019. 523
Wikipedia, Rhumb line. URL https://en.wikipedia.org/wiki/Rhumb_line.

Wikipedia, Schmidt camera. URL

https://en.wikipedia.org/wiki/Schmidt_camera. Accessed 10th April, 2019. 463
í ¤. û
547
Wikipedia, Shoelace formula. URL
https://en.wikipedia.org/wiki/Shoelace_formula. Accessed 9th April, 2019. 258
Wikipedia, Sophie Germain. URL

https://en.wikipedia.org/wiki/Sophie_Germain#Correspondence_with_Gauss.
Wikipedia, Struve Geodetic Arc. URL

https://en.wikipedia.org/wiki/Struve_Geodetic_Arc. Accessed 7th April, 2019.
8
Wikipedia, Synoptic scale in meteorology. URL

https://en.wikipedia.org/wiki/Synoptic_scale_meteorology. Accessed 9th April,
2019. 336
Wikipedia, Theorema Egregium. URL

https://en.wikipedia.org/wiki/Theorema_Egregium. Accessed 8th April, 2019.
54
Wikipedia, Transit 5B-5 — oldest active satellite. URL

https://en.wikipedia.org/wiki/Talk:Transit_(satellite)#Transit_5B-5_-
_oldest_active_satellite. Accessed 30th October, 2019. 283
Wikipedia, Trunnion. URL https://en.wikipedia.org/wiki/Trunnion. Accessed 8th

April, 2019. 144
Wikipedia, Zodiac, Name. URL https://en.wikipedia.org/wiki/Zodiac#Name.

Lydia E. Winkel and Hans de Vries. De ondergrondse pers 1940-1945. Veen, 6th
edition, 1989. URL http://publications.niod.knaw.nl/publications/WinkelDeV
ries_OndergrondsePers_1940-1945.pdf#page=224. entry 748, © 2014 NIOD
(CC BY-SA 3.0). Accessed 5th November, 2019. 425
í ¤. û
Index
ABCDEFGHIJKLMNOPQRSTUVWYZ
A and plane table, 43

Aalto University, 15 alidade level, 127
Aboa (Dronning Maud Land, Antarctica), adjustment, 128
506 alien (LAGEOS), 481
Academy of Sciences of France, 8, 9 Alkmaar (The Netherlands), 3, 5, 42
acceleration, from GPS measurements, almanac, GPS, 291, 293, 328
494 almucantar, 449
accuracy vs. precision, 28 alternative hypothesis, 400, 411
active spare (GPS), 291 ambiguity
adjustment difference, 338
as a projection, 399 distance measurement, 196
average, 366 GPS, 305, 342
correction, 398 spinning encoding circle, 151
linear regression, 367 ambiguity problem, 286, 304
results, 28 distance measurement, 196
weight and correction, 359, 360 ambiguity resolution
weight coefficient, 359 distance measurement, 197
weighting, 359 GPS, 343
adjustment calculus, 28 one-dimensional, 342
gross errors, 400 The American Practical Navigator, 157,
adjustment method 170
by conditions, 396 amplitude modulation, 291
parametric, 363, 396 Amsterdam (The Netherlands), 78, 93
aerial mapping, in detail survey, 228 Amsterdam Ordnance Datum, see NAP
aerial work platform, 260 Andes (mountain range), 9
aerotriangulation, 43 angle measurement, 119
AGA/Minilir, 248 digital, 137
Air Force, USA, 290 angle transformation, 164
air pressure of direction unknown, 165
at GNSS station, 500 angular distance, geocentric, 437, 438
variation, 493, 507 calculation equation, 438
airborne laser scanning (ALS), 252, 254 ANNA (satellite), 464
Akaike information criterion, 395 Antarctica, 506
Åland Sea (Finland, Sweden), 95 antenna (GNSS)
Alexandria (Egypt), 3 calibration, 299
Alfvén, Hannes, 523 choke-ring model, 298, 299
Alfvén’s theorem, 521 electric centre, 298, 300
alidade forced centring, 133
description, 124 phase-centre variation, 298, 299
– 549 –
550 ABCDEFGHIJKLMNOPQRSTUVWYZ I NDEX
antenna reference point (ARP), 298 avskärning, 158

anti-aircraft shells, 462 azimuth, 65
anticorrelation, 37 origin of word, 66
apogee, 525 astronomical, 276
apogee height, 476 geodetic, 277
Apollo project, 480 measurement, 385
approach map, 126
approximate co-ordinates, 82 B
approximate value (linearisation), 371 Baarda, Willem, 403
approximate values, vector of, 372 Baeyer, Johann Jacob, 10
APPS (software), 355 Baker, James, 465
aquifer, 492 Baker–Nunn camera, 464
Aquila (Eagle, constellation), 458 Baltic Sea
arctan (function), 70 mareographs, 511
Arctic summer, 505 sea-surface topography, 508
area levelling bar-code staff, 107
explanation, 112, 113 barometric height determination, 95
check, 114 barycentric co-ordinates, 63
earthwork, 260 base extension network, 4
result, 113 base network
argument of latitude, 328 first-order, 137
Aries (Ram, constellation), 459 lower-order, 218
Aries, First Point of, 459 measurement technology, 212
“as-built” mapping, 232, 245 purpose of use, 212
ascending node, right ascension of, 525 base point, 246
Ashtech Z-12 (GPS receiver), 298 base station
asphalt laying, 247 DGPS, 345
asphalt, hot, refraction, 179 virtual, 351
astatisation of spring gravimeter, 427 baseline
asteroid, orbit determination, 362 in Lapland, 42
asthenosphere, 489, 491 in Nummela, 185, 194
astigmatism, 101 of Snellius, 3
astrolabe, 449, 450 base-network measurement, 214
Danjon, 450 methods, 212, 214
Aswan (Egypt), 3 task, 211
atan2 (function), 70 base-station network
atmosphere, 502 DGPS, 350
angular momentum, 466 RTK, 350
propagation delay, 309 battery of GPS satellite, 289
temperature, 500 beehive pattern, 255
atmospheric inversion, 180 BeiDou (positioning system), 283, 344
atomic clock, relativity, 95 bell curve, see normal distribution
atomic time (TAI), 468 benchmark (image), 131
attraction (Newton), 6 benchmark set, 211
attribute data, 233, 234 quality, 214
aurora, 473 requirements, 214
AUSPOS (software), 355 Bergen op Zoom (The Netherlands), 3, 5,
automatic target recognition (ATR), 150, 42
154 Bergensbanen (Norway), 235
automation of measurement, 107 Bergstrand, Erik Ö., 185
auxiliary closing direction, 175 BGI, International Gravimetric Bureau,
auxiliary point, 225, 229 448
auxiliary starting direction, 175 BIFROST project, 491
average, 359 birefringence, 189
adjustment, 366 Birmingham (UK), 202
mean error, 361
í ¤. û
I NDEX ABCDEFGHIJKLMNOPQRSTUVWYZ
551
black body, thermal radiation spectrum, Carloforte (Italy), 467
189 carrier wavelength, 303
black hole, 480 Carrington event, 478
GPS Block I, II, IIA, IIR, IIR-M, IIF, 289 cartography, 46, 53
GPS Block I, 291 Cassini, Jean Dominique, 7
GPS Block IIIA, 289 catenary, 187
block signal, 151 cause-and-effect relationship, 36, 40
blue film, 496, 497 Cayenne (French Guyana), 13
Boller and Chivens, 465 CDMA, see code division multiple access
Bomford, Guy, 279 celestial mechanics, 5
Bonferroni correction, 412 celestial pole
Bonsdorff, Ilmari, 211 elevation angle, 5
Bouguer anomaly, 447 motion, 470
calculation, 447 projection of direction onto horizontal,
Southern Finland, 448 277
Bouguer plate, 447 celestial sphere, 458, 460
Bouguer, Pierre, 9, 448 centigon, 27
Bowditch method, 170, 175, 178 centring of theodolite, 126, 132
Bowditch, Nathaniel, 157, 170 why, 125
“The Bowditch”, 157, 170 crude, 127
Box, George E. P., 55 precise, 128
break line in terrain, 254 certainty, statistical, 33, 34
broadcast ephemeris, 291, 327 CHAMP (satellite), 494, 502
quality, 341 accelerometer, 494
upload, 290 GPS receiver, 494
use, 328 Chandler wobble, 466
Bromarv guest harbour, 51 Chandler, Seth Carlo, 466
Broom Bridge (Dublin), 265 change plate (levelling), 111
Bruns equation, 432, 433 characteristic polynomial, 380
Bruns, Ernst Heinrich, 432 Charjui (Turkestan, Russia), 467
building location, setting out, 245 check using redundancy, 414
building permit, 245 chemical polarity of water molecule, 203
built environment, 17, 251 chip rate, 297
Buys Ballot, Christoph, 508 C/A code, 295
Buys Ballot’s law, 508 P code, 297
chirality of life, 189
C χ2 distribution, 402, 403
C/A code, 292, 293, 295 degrees of freedom, 402
measurement accuracy, 295 MATLAB, 403
modulation state, 295 non-central, 403
CAD, computer-aided design, 53, 255 test, 402–404
cadastral system, 16 χ23 distribution, 407
caesium clock, 26 choke-ring antenna, 298, 299
GPS receiver, 290 Cholesky decomposition, 398
GPS satellite, 287 Cincinnati (Ohio, USA), 467
calcite, 189 Cinemascope (film format), 464
calibration (metrology), 247 circle singularity (GPS), 322
calibration baseline, 200 circular arc, 240
calving of icebergs, 505 setting out, 240
camwheel (modulator), 194 civil day, 457
Canada civil engineering, 19
land uplift, 85, 491 civil time, 461, 467
magnetic pole, 485 Clairaut, Alexis, 13
Cancer (Crab, constellation), 459 Clairaut’s theorem, 13
Capricorn (constellation), 459 Clapeyron, Emile, 500
Caracas (Venezuela), 492 classical mechanics, 5
í ¤. û
Clausius, Rudolf, 500 conservative field, 440

Clausius–Clapeyron equation, 500 construction surveying, 245
climate model, 502 construction, benchmarks needed, 214
climatic global warming, 502, 505, 510 continental ice sheet
climatology, 308 Fennoscandian, 489
clock offset (GPS), 337 is a glacier, 505
clock unknown (GPS), 303 mass change, 505
closing error thickness, 505
check, 398 continuity condition of transition curve,
co-ordinates, 173, 177 244
direction, 176 continuous distribution, 32
triangle, 358, 359 control measure, 246
clothoid, 243, 244 control segment (GPS), 289, 327
principle, 243 convection
equation, 243 in the Earth’s mantle, 486, 487
parameter, 244 in the Earth’s outer core, 485
coaxial solution, 195 in the Sun, 470, 471
code division multiple access (CDMA), 293 Conventional International Origin (CIO),
GLONASS-K, 344 468
codomain, 31 co-ordinate conversion, 49, 84
cofactor, 519 co-ordinate correction (Helmert), 272
coherence length co-ordinate frame
VLBI, 482 geocentric, 269
white light, 191 right-handed, 266, 267
coin flip, 30 topocentric, 269
coin, fair, 32 co-ordinate reference frame, 55, 56, 77
coincidence level, 102 national, 211
coincidence microscope, 124 co-ordinate reference system, 55, 77
collimation axis, see sight axis geocentric, 47, 265
collimation error, 142, 143, 154 inertial or celestial, 266
definition, 122 plane, 47
correction, 144 realisation, 77
determination, 143 terrestrial (ECEF), 266
field check, 144 three-dimensional, 46
from observation notebook, 143, 144 two-dimensional, 46
Colorado Springs (USA), 289 co-ordinate reference systems, guidance,
column vector, 514 213
saving paper, 516 co-ordinate transformation, 55, 56, 213
comet, orbit determination, 362 co-ordinates
“common-mode” error assumption, 335, 2D+1D, 83, 277
336 Cartesian, 268
comparator (length calibration), 186 ellipsoidal, 267
compass direction, 54 geocentric, 266
complete set, of horizontal directions, 163 notation, 265
complete sets, method of, 137, 161, 162 geodetic, 47, 267
error reduction, 161 figure, 49
focusing, 136 construction, 48
complete-set average, 163, 168 term, 49
components of position vector, 309 why use, 50
compound curve, 241 geographical, 47, 267
setting out, 242 in geodesy, 46
comprehensive plan, 235 KKJ, 279
computer (human being), 362 local, 66
computer, submarine, 287 connection to national frame, 67
conic section, 474 map projection, 46, 47
connecting levelling of tide gauge, 510 (N, E, U), 273, 274
í ¤. û
553
polar, 65, 66 cut-off elevation angle, for GNSS
project specific, 67 observations, 291, 298
rectangular, 47, 84, 268 cycle slip, 338, 340
figure, 49 Cygnus (Swan, constellation), 458
why use, 50
spherical, 267 D
temporary, 67 Dandelin spheres, 474
three-dimensional, 46, 47 dangerous circle
rectangular, 48 GPS, 322
topocentric, 47, 50, 268 resection, 160
notation, 269 Danish Sound bridge (Denmark), 247
rectangular, 268 data layer, 234
Coriolis force, 507 data snooping, 405
Coriolis, Gaspard-Gustave de, 507 datum, 77, 79
corkscrew, 266, 267 European, 280
corner-cube prism, 197, 198 geodetic, 77
correction dissemination datum difference (vertical datum), 81
alternatives, 350 datum point
communications satellite, 351 concept, 77
mobile Internet, 351 choice, 81, 82
correlation, 42 proximity, 78
definition, 36 datum transformation, 78
and causation, 40 between two ellipsoids, 272, 275
examples, 37 height, 81
of aerial images, 254 in the plane, 82
of tree rings, 294 small parameter values, 82, 83
of VLBI signals, 481 Decca
perfect, 36 system, 284, 285
correlation method (GPS), 294, 295 carrier wave, 284
correlator, optical (Väisälä), 192 hyperbola, 285, 286
cosine rule on sphere, 86 lane, 286
half-angle version, 87 receiver, 285
cost, of an error, 35, 411 station, 284
covariance, 36 decimilligon, 27
critical inclination, 478 declination, 460
Croll, James, 503 deep-sea trench, 435, 487
cross-dipole antenna, 299 deformation
crosshairs detection, 494
adjustment screw, 105 measurement, 114, 246
better, 137 monitoring, 246
damage, 125 deformation analysis, 395
history, 99 datum point, 418, 419
cross-measure, 224, 246 hypotheses, 417
cross-section measurement, 115, 116 time, 417
crosstalk, 293 deformation monitoring, 493
cruise missile, 256 degree (angle unit), 26
crustal motion degree Celsius, 26
anthropogenic, 492 Delaunay triangulation, 62, 255
monitoring, 493 Delaunay, Boris, 62
culmination of a star, 460 Delaunay, Vadim, 62
cultural data, 233, 234 dendrochronology, 193, 294
curl (operator), 522 Department of Defense, USA, 290
curvature correction dependence, statistical, 40
Earth’s surface, 204 Descartes, René, 268
measurement ray, 204 design matrix
definition, 375
í ¤. û
azimuth, 385 direction vector

Helmert transformation, 389, 390 to quasar, 481
linear regression, 367 to satellite, 308, 310
pseudo-range, 313 discrete wavelet transform (DWT), 255
slant range, 383, 384 dispersion in ionosphere, 305
station adjustment, 163 distance
design speed, 244 geometric, 302, 309
detail survey intercontinental, 45
definition, 222 distance measurement
carrying out, 228 curvature corrections, 204
data measured, 228 electronic, 44, 195
equipment and software, 229 electro-optical, 44
purpose of use, 229 height reduction, 206
RTK, 222 infrared light, 44
workflow, 229, 230 instrumental correction, 201
detailed shore plan area, 236 laser light, 44, 196
determinant, 518 LED, light-emitting diode, 44, 196
antisymmetry, 519 mechanical, 44, 185
computing, 518 microwaves, 44, 196
interpretation, 374 modulation, 196
vanishing, 518 random error, 201
deuterium, 504 reduction to reference level, 206, 207
dew point, 501 refraction correction, 203
DGPS, see differential GPS second velocity correction, 204, 205
die, 30 steel tape, 185
cast, 30, 31 terrain correction, 205
fair, 31 visible light, 44
Diego Garcia (island, Indian Ocean), 300, white light, 196
302 distance measurement device, see
difference observation, 336, 337, 340 range-finder
double, 336, 339–342 Distomat (range-finder), 195
effect on errors, 338 distribution, see probability density
single, 336, 340 distribution
triple, 336, 339, 340 disturbing potential, 430, 432
difference of ambiguities, 338 div (operator), 522
difference transformation (Helmert), 75 DNA (molecule), 189
differential GPS (DGPS), 345–347 DOP ellipsoid, 316, 317
difference observation, 346 figure, 317
Digges, Leonard, 119 principal axes, 316, 317
digital elevation model (DEM), 256 Doppler positioning system, see Navy
term, 251 Navigation Satellite System
digital height model (DHM), 251 DORIS
digital terrain model (DTM), 256 description, 483
term, 251 beacon, 483
earthwork, 260 double difference (RTK), 347
visualisation, 255 double pentagon prism, 223
volume calculation, 263 downbanding, 293
dilution of precision (DOP), 313, 318, 333 drift
description, 312 gravimeter, 426
azimuth symmetric, 321 quartz oscillator, 338
calculation, 315 range-finder, 199
calculation script, 332 drinking-water extraction, 114, 492
example, 322, 323 Dronning Maud Land (Antarctica), 506
variants, 314 dry air, partial pressure, 499
dimension, 23 dynamo, natural, 471, 485, 521
direction cosine (linearisation), 312
í ¤. û
555
E eccentric anomaly, 526
Earth eccentricity error, 142
average density, 7 eccentricity in height direction, 183
centre of mass, 47, 57, 265 eclipse, 458
core density, 7 ecliptic, 458
core heat, 487 ED50, 11, 277
crustal density, 7 precision, 280
gravitational field, 6 EGM2008, 118, 439
determination, 496, 498 EGNOS, 351
internal structure, 485 Eiffel Tower, 72 names, 194, 330, 362, 453,
magnetic field, 473, 484, 485 500, 508
magnetic poles, 485 eigenvalue problem, 380, 381
mantle density, 7 Einstein, Albert, 189, 426
mantle viscosity, 489 elasticity, 489
mass distribution model, 6 electrical resistivity tomography, 248
mean radius, 86 electromagnetic field, 189
pear shape, 11 electromagnetic field theory, 188
radius, 3 electromagnetic radiation, 188
radius of curvature, 5, 180 in air, 201
transversal, 267, 268 polarisation, 192
rotation rate, 13 quantisation, 189
thermal energy, 487 spectrum, 190
Earth ellipsoid parameters, 7 electromagnetic wave, 188
Earth flattening, 6–8, 13 elevator experiment, Einstein’s, 426
Earth orbit, 457, 458 elevator speech, 282
eccentricity, 505 elimination and back-substitution, 368
major axis, 504 ellipse
Earth orientation parameters (EOP), 85, definition, 474
331, 468 focal point, 474
Earth rotation, 457, 479 ellipsoid of revolution, 7, 49
centrifugal force, 429, 453 as figure of the Earth, 13
Coriolis force, 507 geodesic, 12
monitoring, 461, 482 eLoran (navigation system), 284
period, 291 encoding circle, 150
rate, 467 absolute, 149
variation, 85, 465, 466, 468 incremental, 150
Earth rotation axis energetic correctness, 446
direction, 4, 47, 57, 265 engineering geodesy, 114, 246
momentaneous, 465, 468 engineering surveying, 246
motion, 469, 470 entrance pupil (aperture), 99
tilt, 504 eötvös (unit), 453
VLBI, 482 Eötvös tensor, 452, 453
Earth, figure of the, 15, 476 measurement, 452
mathematical, 10, 15, 435 on Earth’s surface, 453
physical, 6, 15 Eötvös, Loránd, 426, 452
true, 9 ephemeris time (ET), 468
earthquake equinox, vernal, 266, 459
Izmit (Turkey), 495 equipotential surface
plate boundary, 493 description, 90
Sendai (Japan), 493 curvatures, 452
earthquake-prone area, 493 of the geopotential, 432, 433
Earth’s surface forms, 251 of the normal potential, 430, 432, 433
earthwork calculation, 116, 245, 260 radii of curvature, 452
Easter Scheldt storm-surge barrier (The spacing, 442
Netherlands), 247, 248 equi-value curve, 428
easting, 59, 65 era, common or Christian, 2
í ¤. û
Eratosthenes, 1–3 prismatic, 101

Erdmessung, 15
error F
examples, 30 face left/right, 136
gross, 29, 402, 411 error elimination, 161
simulated, 408 Falkland Islands (South Atlantic), 497
of the first kind, 410 false easting, 60
of the second kind, 411 farsightedness, 101
random, 29 fault, geological, 492, 495
systematic, 29 Fennoscandia
elimination, 136 crustal motions, 491
error ellipse, 39, 40, 380 land uplift, 85, 489
and variance matrix, 41 cross-section, 490
axes, 40 pattern, 490
axes directions, 381 FGI, Finnish Geospatial Research
centre, 40 Institute (National Land Survey)
semi-axes, 381 FinnRef, 217
error ellipsoid, 40, 314, 316, 317 GNSS service, 353
azimuth symmetric, 320 field equations
error propagation, 37 Einstein’s, 426
forward geodetic problem, 378 Maxwell’s, 521
eteenpäin leikkaus, 158 field gravimeter, 426
ETOPO1 (terrain model), 252 field sketch, 224
ETOPO2 (terrain model), 253 FIN2000 (geoid model), 94
ETRS89, 57, 279 FIN2005N00 (geoid model), 279
ETRS-GK n, 62 Finnish Transport Infrastructure Agency’s
ETRS-TM35FIN, 61 DGPS service, 352
Euler free nutation, 466 FinnRef, 216, 217
Euler notation, 428, 451 fix solution (RTK), 349
Euler spiral, 243 Fizeau, Hippolyte, 194
Euler, Leonhard, 243, 466 Fizeau’s method, 195
Eurasian plate flat Earth approximation, 14, 89
motion, 281 flight simulator, 256
sea floor, 486 float solution (RTK), 349
EUREF data centre, 86 floating mark (photogrammetry), 43
EUREF, Reference Frame Subcommission flood safety, 246
for Europe, 57 fluid surface as a reference, 94
EUREF89, 57, 280 fluxion, 325
EUREF-FIN, 279, 280 focusing, 99, 100
creation, 57 footprint of satellite altimeter, 506
densification, 217 footscrew, 124
determination, 217 forced centring
map projection, 61 application, 134
Everest, George, 10, 448 principle, 133
Everest, Mount, 10, 220 forced-centring device, 131
excavation work, damage, 249 attachment to tripod, 124
exit pupil, 99 image, 125
expectancy parallel shift, 130
coin flip, 32 forward geodetic problem
continuous quantity, 34 definition, 68
die cast, 32 error propagation, 378
discrete quantity, 31 generally, 68
expectancy (operator), 35, 360 in the plane, 68
eyeglasses foundation of a building, 246
cylindrical, 101 four-parameter transformation, see
of observer, 100, 101 Helmert transformation in the
í ¤. û
557
plane geodynamics, 15, 84, 331, 479
free fall, acceleration in, 425 exotic techniques, 480
free nutation (Euler), 466 local, 492
free-air anomaly, 434 geographic information technology, 233
Southern Finland, 448 geoid
free-stationing survey, 227, 228 definition, 14, 89
figure, 228 determination, 256, 435
setting out, 239 European, 11
with a tacheometer, 227 global, 438
frequency calibration, 201 term, 10, 15
frequency division multiple access geoid height, 433, 439
(FDMA), 344 computation, 437
frequency measurement, 45 near zone, 438
frequency modulation, 291 variation, 436
Fresnel, Augustin-Jean, 189 geoid map
Frisius, Gemma, 3 global, 494, 498
frost heaving, 126, 128 GOCE, 497
“frozen-in” magnetic field, 472, 521, 522 geoid model, 84
Fuglenes (Norway), 10 Bomford, 277
functional model, 397 FIN2000, 94
model error, 403, 404 FIN2005N00, 279
slant range, 385 geoid undulation, 433
fundamental equation of physical geodesy, geological clock, 486
435, 437 geomensuration, 15
geophysical research, 493, 505
G geopotential, 90
Gaithersburg (Maryland, USA), 467 as a co-ordinate, 46, 90
Galilei, Galileo, 194, 425 difference, 45, 443
Galileo (positioning system), 283, 344 energy, 439
Gascoigne, William, 99 fine structure, 494
Gauss, Carl Friedrich, 357, 359, 362, 435 gradient, 429
Gaussian distribution, see normal measurement unit, 441
distribution variation, 436
Gauss–Krüger (map projection), 59, 61 visualisation, 429
conformality, 62 geopotential model, 95
scale distortion, 63 global, 438, 494
zone division, 59 geopotential number, 443, 446
GDGPS, Global Differential GPS, 353 definition, 441
general circulation model (GCM), 256 geopotential table, 429
general plan, 235 geospatial information, 18
generalisation, of a map, 18 shared use, 233
Geneva (Switzerland), 114 geostrophic equations, 508
GeoBasic (Leica), 155 geostrophic flow, 508
geocentricity, 47, 56, 57 Geotrim Oy, 352
GeoCOM (Leica), 154 Germain, Marie-Sophie, 357
geodesic, 12 GIS, geographic information system, 21,
measurement line, 13 233
geodesy, definition, 14, 15 glacial isostatic adjustment (GIA), 85, 488,
geodetic network 490, 509
European, 11 glacial loading, 418
GPS, 342 Global Navigation Satellite Systems, see
orientation, 276 GNSS
reliability, 395 Global Positioning System, see GPS
terrestrial, 219 GLOBE (terrain model), 252
three-dimensional, 218, 219 GLONASS (positioning system), 283, 318
Geodimeter (range-finder), 44, 185 frequencies, 344
í ¤. û
RTK, 350 clock correction, 302

SP3-c (precise ephemeris), 330 clock offset, 302, 303, 309, 338
GLONASS-K, 344 dual-frequency, 297, 303
GNSS, 45, 198 geodetic, 300
definition, 284 received signal, 293
base-network measurement, 221 GPS satellite, 287
co-location, 509 atomic clock, 287
Earth rotation monitoring, 468 clock correction, 289, 290, 327
geocentricity, 47 clock offset, 303, 309, 339
geodynamics, 479 communication channel, 289
land-ice research, 506 components, 287
tidal loading, 493 Doppler shift, 328
tropospheric sounding, 501 fingerprint, 293, 294
GNSS buoy, 510 frequencies, 288
GNSS limbsounding, 502, 503 health status, 289, 327
GNSS monitoring, 493 lifetime, planned, 289
land uplift, 492 orbit, 324, 326
GNSS network orbit data, 289
continuously operating, 216, 217 orbit perturbation, 289
global, 503 orbital elements, 325
GNSS radio occultation, 502, 503 orbital inclination, 291, 324
GNSS receiver orbital period, 291
on satellite, 502, 507, 509 orbital plane, 291, 325
software-defined, 300, 344 radio transmitter, 288
GOCE (satellite), 95, 497 solar panels, 289
name, 498 three-axes stabilisation, 289
accelerometer, 497 thruster, 289
climate research, 508 transmission power, 288
gravitational gradiometer, 497, 498 GPS signal, 291, 292
sea-surface topography, 499 bandwidth, 293
gon (angle unit), 26, 27 carrier frequencies, 304
Google Earth, 56, 255 carrier phase, 297, 303, 305, 340
Google Maps, 117, 118 carrier phase, raw, 304
Google Street View, 223 correlation, 295
GPS, 45, 283 lock-on, 293, 327
development, 289 modulation, 292
modernisation, 291 replica, 294, 295
original system, 284 wavelength, effective, 296
GPS constellation, original, 291, 292 GPS time, 469
GPS measurement GPS/MET (satellite), 502
geometry, 308, 311 GPU, geopotential unit, 441
goodness, 309, 318 GRACE (satellite pair), 496, 497
repeat, 291 accelerometer, 496
observable, 295, 300, 340 land-ice research, 505
pseudo-range, 295, 309 microwave link, 496
relative, 319, 342 GRACE Follow-On (GRACE-FO), 496
static, 319 grad (angle unit), see gon
vector, 340, 342 grad (operator), 450
water-vapour influence, 203 grade measurement
GPS positioning, 284 arc length, 2
geodetic, 303, 354 in Lapland, 8, 9, 42, 185
geometry, 308, 318 in Peru, 8
real-timeness, 344 of Eratosthenes, 2
task, 308 of Snellius, 3–5, 42
GPS post-processing, 344, 353 of Struve, 8
GPS receiver
í ¤. û
559
plumb-line directions, difference, 2, 4, Great Belt bridge (Denmark), 247
5, 8 Greenland, 484
gradient, 428 Greenwich Apparent Sidereal Time
gradient vector, 428 (GAST), 459
graduation circle Greenwich Mean Time (GMT), 48
eccentricity error, 122 Greenwich meridian
manufacture, 137 image, 48
reading, 140, 141 direction, 266
graduation error, 142 meridian treaty, 48
graduation machine, 137 plane, 48
graduation of levelling staff, 109 Greenwich Observatory, Royal, 47
bar code, 109 meridian circle, 467
chessboard, 108 Greenwich paradox, 282
“E”, 108, 110 Greenwich sidereal time, 459, 461
line, 108 Groningen (The Netherlands), 492
gravimeter, 426, 427 ground control point (GCP), 254
drift, 426 ground-penetrating radar, 248
gravimetric point height, 434 group index of refraction
gravimetry, 426 dry air, 202
monitoring, 493 information content, 202
gravitation, 425 ionosphere, 306
universal law of, 6 group velocity, 305, 306
gravitational constant, universal, 6 GRS80 reference ellipsoid, 49
gravitational gradient, 451 GSI (Leica), 154, 155
gravitational gradient tensor, 451 GTOPO30 (terrain model), 252
gravitational gradiometer, 497, 498 guidance of geodetic works, 213
gravity gyroscope, 469
definition, 426 gyroscope phenomenon, 4
along plumb line, 443, 444
centrifugal force, 429 H
change, land uplift, 492 half-angle formula (arc tangent), 70
monitoring, 492 half-angle formula (cosine rule on sphere),
on Earth’s surface, 426 87
plumb-line mean, 443, 444 Hamilton, Sir William Rowan, 265
strength (scalar), 430 hand-held GNSS, 126
variation, 436 Hannover (Germany), 362
gravity anomaly harbour logistics, 247
definition, 434 hard-drive pack (VLBI), 481
and disturbing potential, 435 Härmälä, Seppo, 16
and height, 434 harmonisation of significance levels, 413
calculation from measurements, 434 harp, Phoenician, 428
variation, 435 hat notation, 362
gravity disturbance, 433 Hatanaka compression, 354
gravity field, 430 Hayford ellipsoid, 59
fine structure, 494 Heaviside, Oliver, 521
is conservative, 440 height
temporal changes, 496, 497 definition, 89
gravity flattening, 13 above sea level, 91, 92, 95
gravity vector, 429, 434, 439 concepts, 93
definition, 430, 440 dynamic, 445
as gradient of geopotential, 450 from levelling, 14
direction, 448 from the reference ellipsoid, 14, 49,
properties, 440 91, 267
gravity-gradient tensor, 451 metric, 90, 441, 446
Gray code, 149 normal, 445
Gray, Frank, 149 definition, 445
í ¤. û
orthometric, 89, 92, 93, 442–444 Hipparchus, 469

definition, 444 histogram, 32
approximate value, 444 Holmes, Arthur, 486
height anomaly, 432 horizon plot, 318
height base point, 246 horizon, plane of the, 119
height contour, 428 horizontal angle, 120
calculation, 256 definition, 120
earthwork, 260 figure, 121
on topographic map, 96 measurement, 157, 158, 386
height deformation horizontal axis (theodolite), 121
analysis, 418 horizontal circle
displacement, 419 electronic readout, 151
monitoring, 419 graduation-error minimisation, 161
shifting variate (E ), 419 locking screw, 161
height determination, 90 horizontal datum, 78, 81, 82
barometric, 95 horizontal deformation
height location, of a building, 246 analysis, 419
height system, 14, 96, 97 co-ordinate displacement, 422
orthometric, 92 displacement vector, 420
height type, 91, 446 mean error of unit weight, 423
properties, 446 monitoring, 421
Heiskanen, Veikko A., 16 shifting variate (E ), 420, 423
helium-neon laser, 26, 202 weight-coefficient matrix, 423
Hellenes, 1 horizontal direction, 65, 227
Helmert transformation, 71 hot spot, 488
centre-of-mass co-ordinates, 391 hour angle, 460
equations, 75 humidity, relative, 500, 501
free stationing, 228 Huygens, Christiaan, 189
in the plane, 72–74, 76, 389 Earth flattening, 7, 13
open traverse, 173 nature of light, 188
residuals, 391, 393 polarisation, 189
rotation angle, 73, 74 hydrogen-maser clock, 287
scale distortion, 76, 271 hygrometer, 501
scale ratio, 73 hyperbola (Decca), 285, 286
scaling, 74 hyperbolic positioning system, 45, 284
“small”, 272 hyperboloid of revolution, 347
steps, 73 hypothesis-freeness, 444, 446
three-dimensional, 270
translation parameters, 73, 75 I
translation vector, 74, 75 IAG
two-dimensional, 270 EUREF subcommission, 281
variances of parameters, 391 history, 11
Helmert, Friedrich Robert, 72 International Earth Rotation and
Helmholtz brightness illusion, 108 Reference Systems Service, 466
Helmholtz, Hermann von, 108 International GNSS Service, 331
Helsinki (Finland), 79, 336 International Gravimetric Bureau
Helsinki astronomical observatory, 78, 467 (BGI), 448
Helsinki mean sea level, 78, 93 iceberg, calving, 505
Hermite, Charles, 420 Iceland spar, 189, 191, 192
Hermitian conjugate, 420 IJsselmeer (The Netherlands), 95
Hermitian variance matrix, 420 illumination system (theodolite), 141
heterodyning, 293 image plate, photographic, 463
heterophoria, 101 image processing, 255
Hexagon Oy, 353 impossibility, statistical, 33
highway levelling, 112 incremental measurement, 286
Himalayas, 10 independence, statistical, 41
í ¤. û
561
independent, identically distributed invar staff, 109
(i.i.d.), 364 invar tape, 110
index error, 146, 147 invariant of point variance matrix, 382
definition, 145 inverse geodetic problem, 69
field check, 148 inverse mapping, 373
removal, 147 inverse matrix, 516–518
index of refraction, 201 inverse problem, 160
air, 203 inversion, mathematical, 498
air, microwaves, 203, 307, 499 inverted barometer (IB), 507
air, visible light, 202 ion, 521
dry air, 202 ionic engine (GOCE), 497
glass, 139 ionosphere
representativeness, 205 dispersion, 305
industrial measurement, 246 electron density, 306, 503
inertia, 425 propagation delay, 302–304, 306
inertial measurement unit, 223 research, 308
infrastructure, 236 sounding, 503
construction, 16, 236, 247 total electron content (TEC), 503
planning, 16, 17, 116, 236 Ionosphere Map Exchange Format
in-phase (modulation), 297 (IONEX), 503
InSAR, 494, 495 ionosphere model, 503
inskärning, 158 ionosphere-free linear combination, 307
insolation iron-ore particle, 485
on edge of ice sheet, 503 irrigation-water extraction, 114, 418, 492
summer, 504 isostatic adjustment, 489
INSPIRE (EU directive), 252 isostatic compensation, 449
installation measurement, 114, 245, 246 isothermal surface, 206
instrument co-ordinates, 50, 119, 268, 269 isotope
instrument height, 183 deuterium, 504
instrument of appeal (zoning), 17 radioactive, 486
instrument, terrestrial measurement, 45 ITRS, International Terrestrial Reference
integrity monitoring (GPS), 351 System, 272
interference fringe (SAR), 494, 495 Izmit (Turkey), 495
interference measurement (Väisälä), 194
interference of light waves, 188 J
International Association of Geodesy, see dynamic flattening ( J2 ), 476
IAG Jacobi, Carl Gustav Jacob, 373
international collaboration, 438 Jacobi, matrix of, 373
International Earth Rotation and Jason (satellite), 509
Reference Systems Service Java (Dutch Indies, Indonesia), 435
(IERS), 466, 468 Jeffreys, Harold, 395
International Ellipsoid of 1924, 59 jet fighter, 256
International GNSS Service (IGS) Jet Propulsion Laboratory (JPL)
description, 331 APPS, 355
Central Bureau, 331 GDGPS, 353
ionosphere modelling, 503 International GNSS Service, 331
precise ephemeris, 330 JHS recommendation
tracking stations, 330 benchmark measurement, 213
International Gravimetric Bureau (BGI), EUREF-FIN, 213
448 municipalityGML, 213
International Latitude Service, 466 N2000, 213
International Polar Motion Service, 467 zoning base map, 213
International Space Station, 478 journalist, 282
intersection, 158, 159 JPEG 2000, 255
intrinsic angular momentum (spin), 191 JUHTA, 213
invar (alloy), 44, 185 Jupiter’s flattening, 13
í ¤. û
Jyväskylä (Finland), 78 Laplace azimuth phenomenon, 278

Laplace field equation, 437, 453
K Laplace, Pierre-Simon, 453
K XVIII (submarine), 425 Lapland, 8
Kakkuri, Juhani, 205 working from the large to the small, 214
Karttakomitea, 211 Large Hadron Collider (LHC), 114
kelvin (temperature scale), 26 Larmor precession, 469
Kelvin, Lord (Sir William Thomson), 522 laser level
Kelvin–Stokes theorem, 522 principle of operation, 115, 116
Kepler orbital elements, 326 earthwork, 260
description, 525 vertical plane, 116
figure, 475 laser plummet, 124, 134, 154
GPS, 327 laser scanning
Kepler, Johannes, 473 airborne, 252, 254
Kepler’s law of areas, 474 earthwork, 260
Kepler’s laws, 473, 474 terrestrial, 223
Kepler’s third law, 476 laser-station network, global, 480
Kern (instrument manufacturer), 44, 130 latency, 344
Kevo (Finland), 78 lateral refraction, 29
kilocalorie, 25 latitude
kinematic measurement, 286 astronomical, 5, 449
Kitab (Uzbekistan, Soviet Union), 467 geocentric, 267, 268
KKJ, 56, 57 geodetic, 49, 267, 268
basic co-ordinate system, 59 variation, 466, 467
YKJ (Uniform Co-ordinate System), Lattamiehentie
61 Joensuu, Finland, 89
zone geometry, 60 Vantaa, Finland, 89
KM2 (terrain model), 252 Laurentide land uplift area, 491
KM10 (terrain model), 252 law of large numbers, 31, 35
Kozai, Yoshihide, 328 leap second, 468
Kukkamäki method, 103 least-squares method, 395
Kukkamäki, T. J., 103 invention, 362
Kvarken (Baltic Sea), 490 name, 360
least-squares solution, 364
L precision, 365
L 1 , L 2 , GPS frequencies, 288, 291, 293, weighted, 396
304, 307 legend of map, 231
L 5 , GPS frequency, 291 Legendre, Adrien-Marie, 362
LAGEOS (satellite), 481 Leica (instrument manufacturer), 152, 155
Lagrange interpolation, 330 Leiden (The Netherlands), 3
Lagrange, Joseph-Louis, 330 length (quantity), 26, 185
land subsidence length of day (LoD), 85, 466
anthropogenic, 114 monitoring, 466, 483
mineral extraction, 418, 492 levelling, 45, 95
water extraction, 492 geometric, 97
land uplift, post-glacial, 114, 466, 479 geometry, 96
horizontal motions, 491 hydrostatic, 94
measurement, 491 safety, 112
physical character, 489 staff distance, 107, 112
Land Use and Building Act 1999/132, 235 trigonometric, 182
land, purpose of use, 16 levelling instrument, 97, 98
landscape visualisation, 256 accuracy, 97
land-use planning, 17 automatic, 105
lane (Decca), 286 classification, 98
Langmuir, Irving, 523 construction, 97
Laplace azimuth equation, 277 digital, 107, 110
í ¤. û
563
field check, 102, 103 ocean tidal, 493
horizon adjustment, 103, 104 Local Apparent Sidereal Time (LAST), 459
purpose of use, 97 local detailed plan, 235
self-levelling, 105, 106 local master plan, 235
traditional, 97 location data, 233, 234
levelling network, 45, 96 location review, 246
loop, 45 logarithmic scale, 26
planning, 112 longitude
levelling of theodolite, 126, 132 astronomical, 461
crude, 127 geocentric, 267
precise, 127, 128 geodetic, 49, 267
levelling rod, see levelling staff longitude determination, 461
levelling spike, 110, 111 Loran-C (navigation system), 284
levelling staff lorry, articulated, 243
description, 108 lower mantle, 489
bull’s-eye level, 110 loxodrome (rhumb line), 54
choice, 98 lunar eclipse, 2
classification, 110 lunar laser-ranging (LLR), 480
for base network, 110 Lunokhod (lunar rover), 480
purpose of use, 110 Lyra (Lyre, constellation), 458
reading through foliage, 108
self-calculating, 113 M
levitation, magnetic, 523 “Maan toinen kuu” (article), 462
life, chemistry of, 189 machine guidance, 247
lifting screw, 97, 103, 105 magma, 485, 487
light, nature of, 188 magnetic bottle (radiation belts), 473
line microscope, 138, 139 magnetic flux, 523
line of force magnetisation stripe (sea floor), 484
of gravity, 429, 432 magnetism of pendulum compensator, 106
of normal gravity, 430, 432 magnetohydrodynamics (MHD), 472, 523
of solar magnetic field, 473 magnetosphere, 473
linear regression, 200, 361 magnification of telescope, 99
adjustment, 367 malfunction of device, 29
definitions, 362 map
design matrix, 367 legend, 20
example, 369 purpose of use, 18
in terms of averages, 362 map algebra, 234
observation equations, 367 map co-ordinates, 55
shifting variate (E ), 407, 409 map north, 179
simulated error, 409, 410 map projection, 53
testing, 406, 408 application, 221
linearisation, 370 choice, 221
one-dimensional, 372 conformal, angle-preserving, 54, 221
pseudo-range, 312, 313 distortion, 55
vector mapping, 373 equidistant, 54
linearised quantity, 371, 375 equivalent, equal-area, 54
link between significance levels, 412 gnomonic, 55
Lippershey, Hans, 119 purpose of use, 54, 55
liquid compensator (height index), 148 scale reduction, 207
Listing, Johann Benedict, 10 map projection co-ordinates, 46, 47
lithosphere, 489 map projection plane as computation
living space, human, 219 surface, 221
loading map projection science, 54
atmospheric, 493 map projection zone, 58, 62
glacial, 418 KKJ, 58
ocean, 493 municipality, 62
í ¤. û
map scale, 18, 61 measurement unit, 23, 24

nominal, 62 measuring pole (GPS), 298
on central meridian, 62 measuring tape, 126
map scale number, 61 calibration, 186
map sheet ( x, y), 46 of steel, 44
map-sheet division, 61, 213 sag, 186, 187
marble, glass, 429 temperature correction, 186
mareograph, 14, 492 measuring telescope, 98, 100
description, 509 levelling instrument, 97
co-location, 509 adjustment screw, 105
spatial sampling, 509 tasks, 99
mareograph pillar, 510 theodolite, 122
Mariana Trench, 220 measuring wire, precision, 44
Mars, Olympus Mons, 219 mechanics, classical, 5
Martikainen, Martti, 18 Mekometer (range-finder), 44, 196
Masala, Kirkkonummi (Finland), 216 Mercator (map projection), 54, 221
mass meridian circle, 449, 461, 467
heavy, 425, 426 graduation circle, 461
inertial, 425, 426 refraction correction, 461
mass flow, under Earth crust, 492 meridian transit, 449, 461, 467
Master Control Station (GPS), 289 metadata, 125
MATLAB (software), 36, 517 definition, 20
statistical testing, 403 concept, 231
matrix of theodolite measurement, 149
definition, 513 recording, 231
addition and subtraction, 513 metal detector, 248
column, 515, 516 meteor observation, 55
inversion, 517 meteorology, 308
main diagonal, 516 metre, 185
multiplication, 513, 515 public, Paris, 24
with a constant, 514 SI definition, 26
orthogonal, 274 metric correctness, 446
row, 515, 516 metrology, 23, 247
singularity, 518 Metsähovi radio telescope, 483
transpose, 516 Metsähovi research station, 78, 79, 86, 354
Maupertuis, Pierre L. M. de, 1, 8, 185 aerial image, 117
Maxwell, James Clerk, 188, 189 benchmark, 78
Maxwell–Faraday equation, 522 co-ordinates, 49, 118
mean anomaly, 526 EUREF-FIN co-ordinates, 48
mean error, 28 VLBI, 483
mean error of unit weight, 317, 376 microgal (µGal), 426
pseudo-range, 315 micrometer, optical, 124, 139
station adjustment, 167 levelling, 98, 109
mean height of satellite, 525 microwaves and water vapour, 203
mean sea level Mid-Atlantic Ridge, 485, 487
geopotential, 441 Mid-Indian Ocean Ridge, 487
global, 510 Mid-Pacific Ridge, 487
mean sea surface, global, 438 Milanković cycle, 504
measure (mathematics), 34 Milanković hypothesis, 503
measured quantity, 23, 24 Milanković, Milutin, 503
measured value, 28 milligal (mGal), 426
measurement base of project, 245 mine surveying, 170, 212, 247
measurement class (zoning), 235, 236 minute (angle unit), 26
measurement error, 28 mistake, human, 29
measurement plan, 18, 21 Misuzawa (Japan), 467
measurement planning, 401 Mitteleuropäische Gradmessung, 10
í ¤. û
565
mobile mapping system, 223 description, 284
Moby (musician), 235 Doppler shift, 286
model error, 398 submarine computer, 287
moisture, effect on levelling staff, 109nearsightedness, 101
monitoring measurement, 246 network adjustment, 359, 362
Mont Valérien (Suresnes, France), 195 in map co-ordinates, 222
Montmartre (Paris), 195 in map plane, 221, 222
monument-type choice, 125, 126 joint European, 11
Moon three-dimensional, 218
attraction, 469 network densification, 111, 214
phases, 188 network hierarchy
mortgage, 16 concept, 214, 215
motorway base-network measurement, 211
asphalt laying, 247 GNSS, 212
planning, 244 levelling, 96, 111
MRI, magnetic resonance imaging, 469 significance, 215
MSAS, 351 network RTK archive service, 353
multipath, 298, 319 new degree, new minute, new second, 27
municipal technology, 237 Newton, Sir Isaac, 5
municipality dot notation, 281
local co-ordinates, 66 Earth flattening, 13
planning, 17 gravitation, 425
utility lines, 247 nature of light, 188
Munk, Walter, 479 Principia, 5
Mylar (PAGEOS), 464 Nivavaara, 1, 9
NLS, National Land Survey of Finland
N co-ordinate transformation service,
N60, 93 86, 87
datum point, 78 levelling, 96
N2000, 279 terrain models, 251, 252
description, 93 non-centrality parameter of χ2
datum point, 78 distribution, 404
fundamental benchmark, 79 non-Euclidean geometry, 358
nabla, 428 nonius, 140
naïve world model, 1 Nordic gravity profiles, 492
nanogal (nGal), 426 normal distribution, 34
NAP (height system), 78, 93 expectancy, 34
NASA (USA), 465 inflection points, 35
National Geodetic Survey (NGS, USA), mean error, 34, 35
328 multivariate, 396
forward geodetic problem, 68 probabilities, 35
inverse geodetic problem, 86 properties, 34
National Geospatial-Intelligence Agency statistical testing, 406, 410, 411
(NGA, USA), 439 two-dimensional, 39
National Map Grid Co-ordinate System, two-sided test, 405
see KKJ normal equation, 364
National Oceanic and Atmospheric as system of equations, 365
Administration (NOAA, USA), name, 364
252 solution, 364
natural-gas extraction, 114, 418, 492 normal gravity, 430
nature, data on, 233, 234 calculation, 430
navigation, 54, 344 GRS80, 431
navigation message, GPS, 291–293, 327, plumb-line mean, 445
469 strength (scalar), 430
Navy Navigation Satellite System (NNSS, vertical gradient, 451
“Transit”), 45, 286, 287 GRS80, 431
í ¤. û
normal gravity field, 430, 431 octave (software), 393, 517

normal gravity vector, 430, 434 Ölander, Victor Rafael, 182
normal matrix, 365 “The Old Survey”, song, 119
azimuth symmetric, 319 Ollikainen, Matti, 280
calculation of average, 366 Olympus Mons (Mars), 219
Helmert transformation, 390 Omega (navigation system), 284
linear regression, 368 one-minute theodolite, 138
example, 370 one-second theodolite, 138
pseudo-range, 314 on-the-fly (RTK), 348, 349
normal potential, 430 optical activity, 189
GRS80, 431 optical lattice clock, 95
normal section, 12 orbit determination, 362
Norsk Rikskringkasting, 235 orbital ellipse
North Star, 470 eccentricity, 525
northing, 59, 65 focal point, 474
no-shadow orbit, 477 semi-major axis, 474, 476, 525
nuisance parameter, 163 semi-minor axis, 525
null hypothesis, 400, 410 orbital inclination, 525
numerical weather prediction (NWP), 256, critical, 478
466 mapping area, 475, 476
Nummela (Finland), 194 orbital period of satellite, 476
Nummela baseline, 185, 194 orientation unknown, 163
Núñez, Pedro, 140 Orion (constellation), 458
Nunn, Joseph, 465 Ørsted (satellite), 502
nutation, 469, 470 orthogonality of pseudo-random codes, 294
NUVEL (plate-tectonics model), 487 orthonormal basis, 519
orthophoto map, 256
O ortho-rectification, 256
observation equation, 363, 398 oscillation period of Cs atom, 26
azimuth, 386 outlier detection, 395, 405, 406
calculation of average, 366 overall validation, 402
carrier phase, 304, 338 over-determination by observations, 398
formation, 363 ozone layer, 502
GNSS, 500
Helmert transformation, 389–391 P
horizontal direction, 386 P code, 292, 293, 295
linearisation, 374 encryption, 297
linearised, 375, 382 measurement accuracy, 297
non-linear, 375 wavelength, effective, 297
pseudo-range, 302, 309, 313 PAGEOS (satellite), 464, 465
redundancy, 363 palaeomagnetism, 486
slant range, 384 Pantometria (book), 119
standard form, 366 paper machine, 246
station adjustment, 163, 164 paper, printed maps, 47
tacheometer measurement, 388 parallax of measuring telescope, 100, 101
VLBI, 482 parallel shift of forced-centring device, 130
zenith angle, 388 parallelepiped, 374
observation errors, vector of, 396, 398 parallelogram, 374
observation function, 375 parasol, 125
observation-station catalogue, 461 parcel boundary, 245
Observatory Hill, Helsinki, 78 parcel division, binding, 236
ocean current partial derivative (linearisation), 312
map, 498 Pasteur, Louis, 189
thermal energy transport, 508 pathological function, 263
transversal tilt, 508 pendulum clock, 13
velocity vector, 508, 509 pendulum compensator, 105
í ¤. û
567
principle of operation, 105, 107 plasma, 521
magnetism, 106 name, 523
pentagon prism, 116 plasma frequency, 306
perigee, 525 plasticity, 489
perigee height, 476 plate boundary, strike-slip, 492
perigee, argument of, 525 plate co-ordinates, 463
periglacial bulge, 488, 509 plate tectonics, 85, 483, 487
perturbance in receiver location, 310 mechanisms, 488
Peru, 8 Plesetsk (Russia), 494, 497
petroleum extraction, 114, 418, 492 plumb line
phase error of signal, 136, 137 definition, 120
phase index of refraction curvature, 121
dry air, 202 plumb-line deviation
ionosphere, 306 definition, 276
phase measurement, electronic, 196, 303 effect on azimuth, 278
phase modulation, 291, 293 equations, 275
phase quadrature, 297 figure, 276
phase velocity, 305, 306 plumb-line direction
photodiode, 149 astronomical determination, 10, 276,
photogrammetry 448
description, 43 variation, 448
earthwork, 260 plummet
terrain model, 254 optical, 124, 130, 131, 134, 135
photon, 189 adjustment, 134
photosphere, 470, 471 checking, 134, 135
physical geodesy, 15, 46 index, 130
definition, 90 separate, 133
pillar, for instrument, 128 precision, 128
Pisa, tower of, 114 rod, 129, 130
Pisces (Fishes, constellation), 459 string, 124, 128, 129
pitch, viscosity, 489 Poder, Knud, 57
Pittacus of Mytilene, 23 Poincaré-Prey reduction, 444
pixel graphics, 255 point cloud, 254
plan (map), 53, 83 point data base, 21
Planck, Max, 189 point description, 126
Planck’s constant, 189 point mean error, 316, 382
plane angle, 24 point number, 126
plane co-ordinates, 47, 66 point precision, 382
as complex numbers, 419 point variance, 382
in geodesy, 65 point-grid presentation, 254, 255
rectangular, 65 polar amplification, 505
plane surveying, 15 polar drift, 466
plane table, 42, 43 polar motion, 465–467
planimeter equation, 259 annual, 465
planimeter, polar Chandler wobble, 466
principle of operation, 259 monitoring, 466, 483
image, 259 polarimeter, 189
planning Polaris (star), 5, 470
of GPS measurements, 312, 314, 333 polarisation, 189, 191
of infrastructure, 17 figure, 192
of land use, 17 and life, 189
of measurement networks, 414 circular, 299
of roads and railways, 243 elliptical, 191
of the built environment, 17 Porvoo (Finland), 392
technical, 17 position determination, 462
planning and measurement process, 401 astronomical, 4, 448, 449
í ¤. û
position vector projection measurement (Väisälä), 193,

of observation station, 308, 309 194
of receiver, 310 propagation
of satellite, 308–310, 326, 328 general, of variances, 377
positioning error, 340, 341 of errors, 37, 38
potassium–argon, 486 of expectancies, 376
potential energy, 90, 439 of variances, 38, 41, 376
Pothenot, Laurent, 160 propagation delay modelling, atmospheric,
power of a test, 411 351
PP2000 (benchmark), 78 property
image, 79 monetary value, 16, 17
PPP, precise point positioning, 353 owner rights, 17
pre-amplifier of GPS antenna, 293, 297 property formation, 236
precession, 469 proxy, 294
apsidal, 505 pseudo-observation, 170
astronomical, 504 pseudo-random code, 291, 293, 294
climatological, 505 pseudo-range, 295, 302, 309, 340
Larmor, 469 equivalent, 304, 305
of the equinoxes, 459, 469, 470 linearisation, 312, 313
satellite nodal, 477 observation, 295, 303
satellite orbital, 11 pseudo-range offset, 345
precise ephemeris, 328, 329, 331 public debate, 256
quality, 341 puff pastry (metaphor), 521
precise levelling pulsar, 523
definition, 98 pulse shape, 507
Finnish national, 78
of Finland, first, 111 Q
of Finland, second, 111 Q ℓℓ metric, 398, 402, 405
repeated, 491 quadrant of the plane, 65, 66, 70
precise-levelling network, Finnish quadrature
national, 96, 111, 510 alternatives, 262
precise-levelling staff Simpson’s rule, 260
invar staff, 109 quadrature (modulation), 297
spring, 109 quality
staff unit, 109 of benchmark set, 214
precision behaviour of a network, 81 of broadcast ephemeris, 341
precision farming, 247 of detail survey, 228
precision theodolite, 138 of measurement results, 28, 148
precision vs. accuracy, 28 of precise ephemeris, 341
primordial heat, 487 quantum theory, 189
Principia, 5 quartz gauge, 192
Earth flattening, 6 quartz oscillator drift, 338
probability quasar, 480, 482
of a die cast, 31 quasi-geoid, 445
of a value interval, 33 quasi-two-dimensional network geometry,
probability density distribution, 32 218
as limit of histograms, 33 quaternion, 265
three-dimensional, 40
two-dimensional, 39 R
problem situation, field-work, 132 R (software), 36
Procyon (star), 458 radial survey, 225–227
profile measurement, 114, 115 figure, 226
projected co-ordinate reference frame, 58 as a numerical mapping method, 226
projection horizontal-angle measurement, 158
onto map plane, 221 metadata, 227
onto reference ellipsoid, 221 setting out, 237, 238
í ¤. û
569
slope reduction, 226 choice, 276
radian, 26, 27 goodness of approximation, 220
radiation belt, 289, 473 surface normal, 49
radio navigation, 284 reference sequence (tree rings), 294
radioactivity, 486, 487 reference sphere, goodness of
radiometric dating, 486 approximation, 219
radio-traffic disruption, 473 reference surface, 93, 218, 220
rail shoe, 111 computational, 183
railway reflective panel on Moon, 480
centrifugal force, 244 reflector, 197
design speed, 244 zero-point error, 200
railway levelling, 112 refraction coefficient k, 179, 204
random-point method, 253 refraction modelling, atmospheric, 182
range-finder, 44, 194 regional plan, 235
calibration, 199 rejection bound, 410
constant correction, 201 choice, 409
constant error, 199, 200 normal distribution, 411
construction, 195 two-sided, 406
damage, 125 rejection of a hypothesis, 410
drift, 199 relativity
frequency calibration, 199 general, 95, 426
frequency correction, 201 special, 307
frequency error, 199, 200 reliability, 395, 398, 412
determination by frequency example, 413, 414
standard, 200 exterior, 404, 416
hand-held, 197 interior, 404, 416
instrumental errors, 198 reliability and precision, 414, 416
scale error, 199 resection, 158–160
zero-point error, 199 three-dimensional, 322
ranging rod, 223 reservoir dam (seismics), 492
rapid orbits, 330 residual ̸= error, 360
rapid prototyping language, 333 residuals, 397
reading device, graduation circle, 123, 138 final, 168
reading microscope, 138, 141 first, 168
eyepiece, 123 in quality control, 398
real estate, monetary value, 16, 17 properties, 398
real number 1, 516 size, 398
real numbers, in computers, 32, 302 sub-space of, 364
real-time kinematic measurement, see weighted, 405
RTK reticule, 99
real-time positioning, 344 reversed curve, 242
real-timeness, 344 reversion staff, 110
reconnaissance, 212 rhumb line (loxodrome), 54
reconnection (magnetism), 473 Richer, Jean, 13
rectangle rule (quadrature), 262 ridge, mid-ocean, 484, 487, 493
red lead, 126 right angle, 26, 27
reduction of observations, 29 right ascension, 460
reduction to reference surface, 13, 219 determination, 461
redundancy, 357, 398 right-angle survey, 223, 224
meaning, 417 check, 224
redundancy matrix, 415 setting out, 239
redundancy, degree of, 417 tools, 223
reference ellipsoid, 48, 84 right-hand side vector
as computation surface, 219 calculation of average, 367
as equipotential surface, 430 linear regression, 368
centre shift, 273, 275
í ¤. û
RINEX, Receiver-Independent Exchange widespread use, 57

Format, 300, 302, 353 satellite radar altimetry, 498, 506
road technique, 506
centre line, 243 ice sheet, 505
centrifugal force, 244 measurement geometry, 506
curvature, 243 on-board GNSS, 509
radius of curvature, 244 satellite-based augmentation systems
structures, 245 (SBAS), 351, 352
transversal tilt, 243, 244 satellite-to-satellite tracking (SST), 497
road plan, 244 saturation (water vapour), 500
roadbuilding, 244 saturation partial pressure, 500, 501
rod plummet, 129, 130 SBAS differential correction, 351
roede (length unit), 3, 4 scalar product, 519
root, mountain, 449 scale height, 502
rounding of corners, 241 scale microscope, 124, 138
circular arc, 240 scale model, of network, 3
compound curve, 241 scale reduction
rounding residue, 143 Gauss–Krüger (map projection), 207
rover location (RTK), 348 UTM (map projection), 208
row vector, 514 scattered settlement area, 236
RTK, 347 Schmidt camera, 464
principle of operation, 348 Schmidt, Bernhard, 463
base-network measurement, 218 Schmidt–Väisälä telescope, 463
detail survey, 228 scilab (software), 517
essence, 349 screw thread, geodetic standard, 124, 298
rubidium clock, 287 S-curve, 242
rugby ball (figure of the Earth), 7 sea currents, geophysical modelling, 95
rust-protection paint, 126 sea level
monitoring, 491
S variation, 493
S waves, 484 sea-floor mapping, 14
saddle point, 428 sea-floor pressure, 496, 497
Safety of Life (SoL), 291, 351 sea-floor spreading, 486
sag correction, measuring tape, 186, 187 sea-level rise, 509
Sagan, Carl, 481 Seasat (satellite), 511
Sagittarius (Archer, constellation), 459 sea-surface topography, 498, 508
Sagittarius A*, 480 definition, 14
sample, 35 causes, 507
sample mean, 35 GOCE, 499
sample mean error, 360 magnitude, 508
sample variance, 35 observed, 509
San Andreas fault (California, USA), 492 theory, 507
satellite geodesy, 265 second (angle unit), 26
co-ordinate reference system, 265 second (time unit), 26
geopotential model, 438 SI definition, 26
images (NASA), 465 selective availability (SA), 345
orbit perturbations, 475 self-calculating staff, 110
positioning, 45 Sendai (Japan), 493
satellite laser-ranging (SLR), 45, 480 sensitivity analysis, 310
satellite orbit choice, 475 sensitivity of observations, 309, 311
satellite orbital motion, 473 sensitivity of spirit level, 102
satellite orbital period, 474 setting out, 236
satellite positioning building location, 245
base-network measurement, 212, 214 process description, 237
geodynamics, 479 radial survey, 237, 238
of sea surface, 510 site reference point, 246
í ¤. û
571
zoning plan, 236 Kepler, 251, 260
setting-out measure proof, 260
pre-calculated, 236 volume calculation, 261
radial survey, 238 singularity
straight line, 239 of a mapping, 374
surface-area calculation, 256 of resection, 160
sewer, functioning, 114, 246 Sirius (star), 458
SFS 3161, utility-line mapping standard, site plan, 246
247 Sjökulla (Kirkkonummi, Finland), 483
shifting variate (E ) ski piste, 256
term, 403 sky plot, 313
expectancy sky survey, 465
alternative hypothesis, 404 slant-range measurement, 382–385
null hypothesis, 403 slope percentage, 187
shimmer (levelling), 112 slope reduction, 187, 188, 206
shipyard, 246 radial survey, 226
shoebox model, 83, 89 SmartNet (Hexagon Oy), 353
shoelace formulas, 258 Snell van Rooyen, Willebrord (Snellius,
shore area, 236 Snell), 3, 160
Shuttle Radar Topography Mission Sodankylä (Finland), 336
(SRTM), 252 Sodankylä Geophysical Observatory, 86,
SI system, 24 354
accepted unit, 25 sodium, viscosity, 489
base unit, 24 solar eclipse, total, 471
derived unit, 24 solar flare, 473
prefix, measurement-unit, 24, 25 solar maximum, 473
unit symbols, 23 solar wind, 471
SI unit solid Earth
of distance, 185 elasticity, 493
of geopotential, 441 mass distribution, 475
sidereal clock, 458 motions, 331
sidereal day, 457 tide, 85
sidereal time, 457, 460 solid-angle element, 437
Greenwich, 266 solstice, 459
sight axis solution space, 364
adjustment, 142 SP3 (ephemeris format), 328, 329
levelling instrument, 97, 101 space geodesy, 85, 265
realisation, 142 space plasma, 521
theodolite, 121 space segment (GPS), 289
sighting azimuth, 276 space weather, 473
sighting direction projected onto the space-time
horizontal, 277 curvature tensor, 426
signal geometry, 426
good, 136 spatial correlation, 336
problem, 136 spatial planning, 17
with prism, 198, 199 definition, 17
signal height, 183 figure, 19
significance level, 411 topographic surveying, 17
link, harmonisation, 412, 413 speed of light, 189, 201
normal distribution, 411 definition, 195
two-sided, 406 determination, 194
silicone oil in compensator, 148 spherical excess, 358
similarity transformation, see Helmert spherical-harmonic coefficient, 476
transformation spider silk, 99
Simpson, Thomas, 260 spike (measurement), 112
Simpson’s rule, 260, 262 spin of a particle, 191
í ¤. û
spinning encoding circle, 151, 152 Stokes kernel, 437

spirit level in levelling instrument, 97 Stokes loop integral theorem, 522
spring balance, 426 Stokes, George Gabriel, 436
Sputnik 1 (satellite), 11 straight angle, 26
stadium (length unit), 3 straight line
staff (levelling), see levelling staff fit, 361, 367
staff calibration, 108 setting out, 239
staff distance, 107, 112 straight setting-out method, 239
staff support, 110, 111 stratification, atmospheric, 318
staff unit, 110 stratosphere, 502
standard strength of network, 412
of construction surveying, 245 stress-energy tensor, 426
of geodetic works, 213 string plummet, 124, 128, 129
standard deviation, 28 Struve chain, 8, 10
standard ellipse, see error ellipse Struve, Georg Wilhelm von, 8
standard uncertainty, 28 subduction, 435, 487
standing axis (theodolite), 121 sum of squares
star catalogue, 461 of plumb-line deviations, 276
star programme, 449 of residuals, 398
static base-network measurement, 218 Sun
station adjustment, 162 convection, 470
calculation, 164 convection layer, 470, 471
template, 169 core thermal energy, 470
defect, 169 corona, 471
degrees of freedom, 163, 170 heating, 471
design matrix, 163 magnetic field, 471, 472
error model, 167 lines of force, 473
in observation notebook, 162 topology, 473
observation equations, 163, 165 nuclear power, 470
redundancy, 163 plasma, 471
residuals, 166 X-rays, 470
shifting variate (E ), 167 Sunda deep-sea trench, 435
summary, 167 sunspot, 472
steel grid mast, 118 Sun-synchronous orbit, 477
stellar triangulation, 43, 462 superconductivity, 471, 521
equipment, 462 superconductor, 522
flashtube, 462, 464 surface area
high signal, 463 calculation, 256
measurement geometry, 463, 464 with co-ordinates, 257, 258
network adjustment, 464 with setting-out measures, 256,
plate reduction, 464 257
using satellites, 464 measurement by planimeter, 259
weather balloon, 462 Survey of India, 10
steradian, 26 surveying science, 15
stereo model restitution, 43, 44 Syene (Egypt), 3
sticker, reflective, 198 symmetric GPS measurement geometry,
stochastic quantity, 30 316
linear combination, 37 synoptic scale, 336
multi-dimensional, 39, 40 synthetic-aperture radar (SAR), 494, 505
realisation, 30, 31 terrain model, 252
real-valued, 32 system calibration, 107, 110
two-dimensional, 33
Stokes equation, 436, 437 T
computation point, 438 taaksepäin leikkaus, 158
integration point, 438 tablet (computer) in detail survey, 229
Stokes function, 437 tabloid press, 282
í ¤. û
573
tacheometer instrumental errors
term, 45 horizontal-angle measurement, 142
electronic, 45 zenith-angle measurement, 148
observation equations, 388 invention, 119
tape correction, 186 measurement, 119
tartaric acid, organic, 189 sighting and targeting, 135
Tasmania (Australia), magnetic pole, 485 thermal expansion coefficient
Taurus (Bull, constellation), 459 invar, 44, 109, 185
Taylor series expansion (linearisation), steel, 186
312, 371, 375 thermal radiation spectrum of black body,
TCA2003 (theodolite), 152, 153 189
automatic target recognition (ATR), thermodynamics, arrow of time, 307
154 three-axes stabilisation, 289
data exchange, 155 three-sigma rule, 29, 36, 411
distance reduction, 154 tidal force field, 451, 453
laser plummet, 154 tidal loading, 493
level, electronic, 154 tide gauge, see mareograph
monitoring measurement, 154 tie measurement of a profile, 116
self-calibration, 154 tie-in survey, 224
tectonics, 418 check, 225
intra-plate, 493 tiling of images, 255
telescope tilt meter, digital, 148
description, 99 time (co-ordinate), 46, 84
invention, 119 time (quantity), 26, 468
Tellurometer (range-finder), 44 time determination, 461
terminology work, 213 time difference measurement, 285
terrain attraction, 447 time, precise, dissemination of, 286
terrain correction (distance), 205 time-zone system, 48, 468
terrain correction (gravity), 256 Tobler, Waldo R., 335
terrain model toise (length unit), 8
applications, 255 Tom and Jerry, see GRACE
measurement geometry, 253 top (spinning), 469
measurement technology, 254 TOPEX/Poseidon (satellite), 479, 509, 510
presentation form, 255 orbital inclination, 509
terrain presentation, 255 topographic attraction, 447
testing strategy, 411 topographic data
testing, statistical, 28, 29, 400 catalogue, 231
deformation, 419, 494 classification, 233
gross errors, 400 encoding, 231–233
hypotheses, 400 topographic information system, 230
normal distribution, 410 national, 231
outlier detection, 395 topographic surveying, 16, 18, 19, 21
theodolite computation, 229
image, 120 data, 231
axes, 121, 122, 127 end products, 20
axis errors, 142 map, 18
base, 124 in spatial planning, 17
calibration, 142 metadata, 231
carrying case, 125 skills, 18
circles, 121, 122 software, 229
classification, 137 tasks, 17
construction, 123 terrain model, 254
electronic, 148–152 work volume, 17
gyro-, 269, 386 Topographic Surveying and Mapping
handling, 124 (booklet), 16
in mapping, 42 Torge, Wolfgang, 15
í ¤. û
Torne river (Sweden, Finland), 42, 185 principle, 3

valley, 8, 78 Finnish national, 57
torsion balance, 452 first order, 215
total station, 45 horizontal-angle measurement, 157
automatic, 493 lower order, 215
traceability (metrology), 185, 186, 247 method, 42
traffic route planning, 90, 256 of Snellius, 3
trajectory (road) triangulation network, 359
calculation, 245 adjustment, 42
setting out, 245 continental, 11
transformation formula, local (RTK), 350 intercontinental, 465
transformation parameters principal side, 185
EUREF89 to ED50, 280 triangle, 358
Helmert transformation, 73, 74, 389 trigonometric heighting, 180
ITRF2005 to ETRF2005, 281 fundamental equation, 181
Transit 5B-5, 283 zenith angle, 180, 181
Transit positioning system, see Navy Trimble AB, 44
Navigation Satellite System Trimnet VRS, 352
transit through meridian, 449, 461, 467 tripod head, central hole, 128
transition curve, 243 tropic, 459
transpose, 516 tropopause, 502
of a vector, 516 tropospheric propagation delay, 302, 303,
saving paper, 42, 516 305, 307
transversal wave motion, 484 tropospheric research, 308
trapezoid rule (quadrature), 262 true anomaly, 526
travel time (GPS signal), 294 true value, 28, 35
traverse of an observed quantity, 34, 396
definition, 170 truncated co-ordinates, 61, 67
auxiliary points, 171 trunnion axis (theodolite), 121
calculation in the plane, 179 trunnion-axis tilt, 144, 145
closed, 170, 171, 174, 175 definition, 121
closing point, 171, 178 determination, 144
computation, 170, 172 truth, formal (datum), 81, 82
computation template, 178 tubular level
co-ordinate adjustment, 176 figure, 102
directions, adjustment of, 175 adjustment screw, 101, 103
horizontal-angle measurement, 157 construction, 101
incompletely closed, 171 horizon, 97, 101
least-squares adjustment, 178 task, 101
open, 170–172 tunnel surveying, 158, 170, 212, 247
starting direction, 172 tunnels, network of (metaphor), 92
starting point, 171, 178 Tuorla, Wizard of, see Yrjö Väisälä
straight, 178 Turku (Finland), 78, 79, 462
weighting, 177 Turku Cathedral, 79
traverse levelling, 95, 111, 112 Tuusula (Finland), 57
traversing, 170, 215
tree line, 294 U
tree ring, 293, 294 UDP/IP protocol (VLBI), 481
correlation, 294 Ukiah (California, USA), 467
trend in time, 40 ultra-rapid orbits, 330
triangle condition, 29, 358 ultraviolet radiation (PAGEOS), 464
triangulated affine transformation, 62, 64 UNAVCO
triangulated irregular network (TIN), 255 geoid model, 118
triangulated-network presentation, 254, plate motion calculator, 488
255 uncertainty, 30
triangulation, 3 type A, 30
í ¤. û
575
type B, 30 automatic, 148
uncertainty ellipse, see error ellipse video game, 256
unit element of a set of numbers, 514 video tape (VLBI), 481
unit matrix, 514, 516 virtual reference station (RTK), 352
Universal Time Co-ordinated (UTC), 468 viscosity, 489
University of Helsinki, 15 pitch, 489
uranium, depleted, 480 sodium, 489
uranium–lead, 486 water, 489
urban canyon, 170, 223 visibility
urban landscape, 158 free stationing, 227
US Geological Survey (USGS, USA), 252 in the terrain, 256
“use point”, 245 of GPS satellites, 284
user segment (GPS), 290 visibility conditions of benchmarks, 212
utility line, 247 visible light, nature of, 188
map, 247 visual ellipse of point variance matrix,
mapping methods, 248 380, 381
utility-lines marking service, 249 visual ellipsoid of gravitational gradient
UTM (map projection), 61 tensor, 451
description, 62 visualisation of spatial data, 234
conformality, 62 VLBI
scale distortion, 62, 63 description, 480
Earth rotation monitoring, 468
V Mark II, 481
Vaakitsijantie, Oulu, Finland, 89 measurement geometry, 482
Väisälä interferometry, 191, 193 observable, 481
Väisälä, Yrjö, 43, 191, 457 observation campaign, 480
value set, 31 principle of operation, 482
Vanguard 1 (satellite), 11 vector solution, 482
Vantaa (Finland), 352 Vocabulary of Geoinformatics, 214
variance, 35, 376 volcanism, 488
variance matrix, 40, 376 volume calculation
definition, 40 Simpson’s rule, 260
of co-ordinate displacements, 423 terrain model, 263
of co-ordinates, 315, 316 VRS-RTK, 352
of GPS solution, 315 VVJ (Helsinki system), 57
of observations, 365, 397 map projection, 61
of solution, 397
variance of unit weight W
a posteriori, 402 W code, 292, 297
pseudo-range, 315 WAAS, 351
Vasa, the good ship, 114 wall measure, 245, 246
vector graphics, 255 Washington DC (USA), meridian treaty, 48
vector mapping, 373, 374 water gauge in interior waters, 95
vectorial product, 519 water molecule, chemical polarity, 203
Vega (star), 470 water vapour
velocity correction, second, 204, 205 content of air, 307
velocity vector of satellite, 326, 328 integrated, 502
Venezuela, 492 total precipitable, 502
Venice (Italy), 114 in stratosphere, 502
Vening Meinesz, Felix A., 425, 435 partial pressure, 308, 500, 501
vernier, 140 wave function, complex, 188
Vernier, Pierre, 140 wave motion
vertical angle, see zenith angle frequency, 189
vertical axis (theodolite), 121 on open sea, 509
vertical datum, 80 phase, 188, 189
vertical-circle index, 145 wavelength, 189
í ¤. û
wave packet, 306 topographic data classification, 233

wave-motion theory of light, 188 zoning-plan types, 235
weather observations, representativeness,
204, 206
Wegener, Alfred, 483
Wegener’s continental-drift theory, 484
weight matrix of unknowns, see normal
matrix
weight-coefficient matrix, 376
of observations, 396
of pseudo-ranges, 315
of residuals, 400
of unknowns, 365
West Siberian land uplift area, 491
WGS84, 58
what if. . . (reliability), 414
white noise (VLBI), 482
wide-area differential GPS (WADGPS),
351
wide-laning, 343, 350
wind drag, 507
wine-barrel formula, 251
wood dating (tree rings), 193
work, 45
path integral, 440, 441
world model, naïve, 1
writing error in observation notebook, 29
Y
Y code, 292
year ring, see tree ring
Young, Thomas, 189
Z
1462 Zamenhof, 457
Zamenhof, L. L., 457
Zeiss (Trimble) DiNi12 (levelling
instrument), 107
zenith angle, 120
definition, 120
figure, 121
measurement, 145, 146, 387
refraction, 179, 180
zenith propagation delay, 500
zenith tube, 449, 467
zero element of set of numbers, 514
zero error of level, 127, 128
zodiac, 458
zoning, 16
Zoning Base Map Guide, 213
zoning plan, 16
boundaries, 237
calculation, 237, 238
Zoning Survey Guide, 213
new (2014), 235
setting out, 236
í ¤. û

Geode Sy

Uploaded by

Geode Sy

Uploaded by

Martin Vermeer

The science underneath

© 2019 Martin Vermeer

ISBN (pdf) 978-952-60-8872-3

Graphic design: Cover: Tarja Paalanen

Keywords geodesy, land surveying, mapping, geodetic measurements, networks, co-ordinates,

A LTHOUGH F INLAND IS, and has been since independence, a superpower

and volume calculations.

Helsinki, 9th December, 2020,

A second, extensively corrected and improved edition was published on

In drafting the various manuscript versions, lecturing transparencies and

Õ ! A. Properties of matrices . . . . . . . . . . . . . . . . . . . . . . . . . 513

List of Tables xii

List of Figures xiii

1. The history and societal status of geodesy 1

2. Geodetic measurements and co-ordinates 23

3. Map projections, datums and transformations 53

5. The theodolite 119

6. Angle measurement 157

7. Distance measurement 185

8. Base-network and detail-survey measurement 211

9. Construction surveying 235

10. Digital terrain models and volume calculation 251

11. The third dimension 265

12. The Global Positioning System (GPS) 283

13. Processing GPS observations 335

14. Adjustment calculus in geodesy 357

14.5 Examples of the least-squares method . . . . . . . . . . . . 366

15. Statistical methods in geodesy 395

16. Gravity in geodesy 425

17. Space geodesy 457

18. Geodesy and geophysics 479

A. Properties of matrices 513

B. A short introduction to magnetohydrodynamics 521

C. The Kepler orbital elements for satellites 525

C.2 Elements describing the orbit’s size and shape . . . . . . . 525

1.1 Topographic surveying . . . . . . . . . . . . . . . . . . . . . . . 21

2.1 Measured quantities, units and their symbols . . . . . . . . . 24

3.1 Alternative vertical datums A and B . . . . . . . . . . . . . . 80

4.1 Classification of levelling instruments . . . . . . . . . . . . . 98

6.1 Calculation template for station adjustment . . . . . . . . . . 169

7.1 Calculating the constant and frequency error of a range-

8.1 Methods for base-network measurement . . . . . . . . . . . . 214

11.1 Transformation parameters between EUREF89 and ED50 . 280

12.1 Codes included in the GPS signal . . . . . . . . . . . . . . . . 292

13.1 Effect of forming difference observations on the magnitude

14.1 Measurement results for linear regression . . . . . . . . . . . 369

15.1 The planning and measurement process . . . . . . . . . . . . 401

16.1 Normal potential and normal gravity according to GRS80 . 431

17.1 Kepler’s third law for Earth satellites . . . . . . . . . . . . . . 476

1.7 The Lapland grade measurement . . . . . . . . . . . . . . . . 9

2.1 A public standard metre in Paris . . . . . . . . . . . . . . . . . 24

4.1 Different height types map geopotential numbers in different

5.1 An old-fashioned theodolite . . . . . . . . . . . . . . . . . . . . 120

5.11 A benchmark seen though an optical plummet . . . . . . . . 131

6.1 Use situations for horizontal angle measurement . . . . . . . 158

7.1 Sag correction of measuring tape . . . . . . . . . . . . . . . . 186

8.1 Significance of network hierarchy, mistakes often made . . . 215

9.1 Setting out onto the terrain, process description. . . . . . . . 237

10.1 Terrain model ETOPO2 version 2 on the Finnish territory . 253

11.1 Inertial and terrestrial co-ordinate reference systems . . . . 266

12.1 The Decca system . . . . . . . . . . . . . . . . . . . . . . . . . . 285

13.1 The “common-mode” error assumption . . . . . . . . . . . . . 335

14.1 Triangulation network . . . . . . . . . . . . . . . . . . . . . . . 358

15.1 Least-squares adjustment as an orthogonal projection . . . . 399

15.7 Another example of reliability . . . . . . . . . . . . . . . . . . 414

16.1 Absolute or ballistic gravimeter . . . . . . . . . . . . . . . . . 427

18.1 A LAGEOS satellite . . . . . . . . . . . . . . . . . . . . . . . . 481

DHM digital height model 251

InSAR interferometric synthetic-aperture radar 494, 495

SoL Safety of Life 291, 351

Maupertuis (1738), PDF page 44, page 16

^ 1.1 The figure of the Earth, early conceptions

^ F IGURE 1.2. The grade measurement of Eratosthenes.

1 “Before the common (or Christian) era”.

See figure 1.2.