FULLTEXT01

ROYAL INSTITUTE
OF TECHNOLOGY
Precise Gravimetric Geoid Model for Iran

Based on GRACE and SRTM Data and the
Least-Squares Modification of Stokes’
Formula with Some Geodynamic
Interpretations
Ramin Kiamehr
Doctoral Dissertation in Geodesy
Royal Institute of Technology (KTH)

Department of Transport and Economics
Division of Geodesy
100 44 Stockholm
October 2006
Ramin Kiamehr, Precise gravimetric geoid model for Iran based on GRACE and SRTM data and the least-
squares modification of Stokes’ formula with some geodynamic interpretations
Supervisor:
Prof. Lars Erik Sjöberg
Faculty Opponent:
Prof. Martin Vermeer,
Helsinki University of Technology, Finland
Evaluation committee,
Prof. Christopher Talbot,
Dept. of Earth Sciences, Uppsala University, Sweden
Prof. Artu Ellmann,
Department of Civil Engineering
Tallinn University of Technology, Tallinn, Estonia
Dr. Margarita Petrovskaya,
Central Astronomical Observatory of Russian Academy of Sciences,
St.Petersburg, Russia
© Ramin Kiamehr 2006
Royal Institute of Technology (KTH)

Architecture and the Built Environment
Department of Transport and Economics
Geodesy Division
SE-100 44 Stockholm, Sweden
Printed by Universitetsservice US-AB.

Stockholm, Sweden, 2006
TRITA 06-003 • ISSN 1653-4468 • ISBN 13: 978-91-85539-06-2 • ISBN 10: 91-85539-06-6
Abstract
Iran is one of the most complicated areas in the world from the view of rough topography,
tectonic activity, large lateral density and geoidal height variations. The computation of a
regional gravimetric geoid model with high accuracy in mountainous regions, especially with
sparse data, is a difficult task that needs a special attention to obtain reliable results which can
meet the needs of the today geodetic community.
In this research different heterogeneous data has been used, which includes gravity
anomalies, the high-resolution SRTM Digital Elevation Model (DEM), recently published
GRACE Global Geopotential Models (GGMs), geological maps and GPS/levelling data. The
above data has been optimally combined through the least-squares modification of Stokes
formula with additive corrections. Regarding the data evaluation and refinement, the cross-
validation technique has been used for detection of outliers. Also, several GGMs and DEMs
are evaluated with GPS/levelling data. The impact of utilizing a high resolution SRTM DEM
to improve the accuracy of the geoid model has been studied. Also, a density variation model
has been established, and its effect on the accuracy of the geoid was investigated. Thereafter a
new height datum for Iran was established based on the corrective surface idea. Finally, it was
found that there is a significant correlation between the lateral geoid slope and the tectonic
activities in Iran.
We show that our hybrid gravimetric geoid model IRG04 agrees considerably better
with GPS/levelling than any of the other recent local geoid models in the area. Its RMS fit
with GPS/levelling is 27 cm and 3.8 ppm in the absolute and relative senses, respectively.
Moreover, the relative accuracy of the IRG04 geoid model is at least 4 times better than any
of the previously published global and regional geoid models in the area. Also, the RMS fit of
the combined surface model (IRG04C) versus independent precise GPS/levelling is almost 4
times better compared to the original gravimetric geoid model (IRG04). These achievements
clearly show the effect of the new gravity database and the SRTM data for the regional geoid
determination in Iran based on the least-squares modification of Stokes’ formula.
Keywords: Gravity database, least-squares modification of Stokes, geoid determination, SRTM,

GRACE, GPS/levelling, density variation model, height datum, geodynamics, Iran
i
Acknowledgements
First and foremost, I would like to express my sincerest gratitude to my supervisor Professor
Lars E. Sjöberg for his guidance and encouragement during the last four years. I am indebted to
him for innumerable discussions, his valuable comments, professional advice and help through the
process of developing the idea for this dissertation and making it a reality. His availability, enthusiasm
and flexibility are wholeheartedly acknowledged. Without his knowledge, experience this thesis
would never have seen the light of day.
Special thanks go to Dr. Jonas Ågren; many impromptu discussions with him contributed greatly
to this work. His contribution in programming and software is greatly appreciated. I would like to
extend my gratitude to all former and present colleagues of Divisions of Geodesy and
Geoinformatics for their support. I would like to thank Drs. Tomas Egeltoft, Huaan Fan, Milan
Horemuz, Mr. Erik Asenjo, Mrs. Solveig Winel and also PhD students Marek Rannala, Johan Vium
Andersson, Yuriy Reshetyuk and Prosper Ulotu for their support, help and warm companionship.
Special thanks to Erik Asenjo and Dr. Hans Hauska for their help with computer and network
problems.
Many thanks to Prof. Will Featherstone as well as Drs. Pavel Novak and Rene Forsberg for their
help, useful suggestions and comments through the determination of geoid model and review of my
articles.
Funding for this research was provided by the Ministry of Science, Research and Technology of
Iran (MSRT). This research could not be initiated and completed without this support. This support is
cordially acknowledged. Also, I wish to express my thanks to the National Cartographic Centre of Iran
(NCC), for releasing part of the GPS/ levelling and gravity data.
Also, I wish to thank my friends in Stockholm, Mahnaz, Ali, Mostafa, Farhanaz, Issa and
Shayesteh for helping me and my family get through the difficult times, and for all the emotional
support, camaraderie, entertainment, and caring they provided.
I would like to dedicate this dissertation to my spouse Noshin and my daughter Pardis. Their
support, patience and understanding have had an enormous contribution to this work. Last but not least,
I would like to thank my parent for their patients and persistent encouragement over my PhD studies.
Stockholm, October 2006
etÅ|Ç ^|tÅx{Ü
ii
Contents
Abstract i
Acknowledgements ii
Table of Contents iii
List of Figures vi
List of Tables vii
PART ONE - THE HYBRID PRECISE GEOID MODEL FOR IRAN 1
1. INTRODUCTION 3
1.1 Preamble 3
1.2 Research objectives and author’s contribution 5
1.3 Review of Part One 6
1.4 Review of Part two 7
1.5 Other Publication and Presentation 9
2. GEOID DETERMINATION BASED ON LEAST- SQUARES MODIFICATION OF

STOKES' FORMULA 11
2.1 Modification of Stokes' formula 11

2.2 Signal and noise degree variances 16
2.3 Determination of the least-squares modification parameters 19
3. ADDITIVE CORRECTIONS IN THE KTH APPROACH 21
3.1 The Topographic Corrections 22

3.2 The Downward Continuation Correction 23
3.3 The Ellipsoidal Correction 25
3.4 The Atmospheric Correction 26
4. EVALUATION AND CONSTRUCTION OF THE DATABASE 29
4.1 Qualification and creation of gravity anomaly database 29

4.2 The GPS/levelling data 31
4.3 The Digital Elevation Models 33
4.3.1 Iranian SRTM based DEM 34
4.4 The Global Geopotential Models 36
4.4.1 GRACE GGM02 37
iii
5. THE NEW GRAVIMETRIC GEOID MODEL (IRG04) 39
5.1 Basic Parameters and Results 39
6. EFFECT OF LATERAL DENSITY VARIATION ON THE GEOID 45
6.1 Introduction 45
6.2 Topographic density models from geological maps 46
6.3 The effect on the geoid 47
7. EVALUATION OF THE GRAVIMETRIC GEOID MODELS 51
7.1 Overview of current Iranian Gravimetric Geoid models 51

7.2 Internal Estimation of accuracy of IRG04 52
7.3 External Estimation of the accuracy of geoid model 54
7.3.1 Estimation of accuracy in an absolute sense 55
7.3.2 Effect of lateral density variation model 60
7.3.3 Estimation of accuracy in a relative sense 61
8. A NEW HEIGHT DATUM FOR IRAN 67
8.1 Creation of the corrective surface 67
9. GEODYNAMICAL RESEARCHES 65
9.1 Impact of the geoid model in studying tectonic activities in Iran 69

9.2 Analysis of surface deformation patterns using 3D finite element method 72
9.2.1 Skåne GPS network 74
10. CONCLUSIONS AND RECOMMENDATIONS 76
BIBLIOGRAPHY 80
PART TWO – PUBLICATIONS 88
PAPER A:
Kiamehr R (2005) Qualification and refinement of the gravity database based on cross-
validation approach, A case study of Iran, Proc. Geomatics 2004 (84) Conferences, National
Cartographic Centre of Iran, Tehran, Iran, Revised version, submitted J. Acta Geodaetica et
Geophysica
PAPER B:
Kiamehr R and Sjöberg LE (2005a) The qualities of Iranian gravimetric geoid models versus
recent gravity field missions. J. Studia Geophysica et Geodaetica, 49:289–304
iv
PAPER C:
Kiamehr R and Sjöberg LE (2005b) Effect of the SRTM global DEM on the determination of
a high-resolution geoid model: a case study in Iran, J. Geodesy, 79(9):540-551
PAPER D:
Kiamehr R (2006a) A strategy for determining the regional geoid in developing countries by
combining limited ground data with satellite-based global geopotential and topographical
models: A case study of Iran, J. Geodesy , 79(10,11): 602–612
PAPER E:
Kiamehr R (2006b) Hybrid precise gravimetric geoid model for Iran based on recent GRACE
and SRTM data and the least squares modification of Stokes formula, J. Physics of Earth and
Space, 32(1) 7-23
PAPER F:
Kiamehr R (2006c) A new height datum for Iran based on combination of the Gravimetric and
GPS/levelling geoid models, Presented and published Dynamic Planet 2005 IAG Symposia
Series Springer Verlag, Revised version in press, Acta Geodaetica et Geophysica, 42(1)
PAPER G:
Kiamehr R (2006d) The Impact of lateral density variation model in the determination of
precise gravimetric geoid in mountainous areas: a case study of Iran, in press Geophysical
Journal International.
PAPER H:
Kiamehr R and Sjöberg LE (2006) Impact of a precise geoid model in studying tectonic
structures- A case study in Iran, J. Geodynamics, 42(2006) 1-11
PAPER I:
Kiamehr R and Sjöberg LE (2005c) Surface Deformation Patterns Analysis using 3D Finite
Elements Method: A case study in Skåne area, Sweden, J. Geodynamics, 39(4) 403-412
v
List of Figures
Fig. 1. Distribution of the gravity anomaly data (BGI, NCC and Ship-borne data presented in
blue, red and green colours, respectively)
Fig. 2. The predicted 80′′ × 90′′ free-air anomaly grid.
Fig. 3. Distribution of the GPS/levelling data in Iran. Red and black points indicate first and
second-order GPS and levelling networks, respectively. Five traverses are chosen to studying
the effect of topography in different areas.
Fig. 4. The 15′′ digital terrain model of Iran (IRD04) based on 3′′ SRTM data.
Fig. 5. Estimated degree variances and degree error variances for GGM02S/C and EGM96 are
shown in geoid height units.
Fig. 6. Additive correction terms for the geoid
a. Combined topographic effect for the geoid in Iran.
b. The DWC effect on the geoid model in Iran.
c. The total topographic effect on the geoid model in Iran.
d. Ellipsoidal correction on the geoid model in Iran.
e. Combined atmospheric correction on the geoid model in Iran.
Fig. 7. The isolines and 3D view of the IRG04 geoid model.
Fig. 8. The lateral density variation model of Iran.
Fig. 9. Effect of the lateral density variation model on geoid.
Fig. 10. Principal of levelling with GPS
Fig. 11. Histogram of difference between the 260 GPS/levelling data and IRG04 geoid model
before (a) and after 7 parameter fitting (b).
Fig. 12. Discrepancy between the GPS/levelling and gravimetric geoid models.
Fig. 13. Comparison of relative accuracies of fitting GPS/levelling data and GGM02 models
versus local gravimetric geoid models in five selected traverses.
Fig. 14. Discrepancies between the IRG04 and TUG geoid models.
Fig. 15. Distribution of earthquakes (black dots) on the lateral geoid slope map.
vi
List of Tables
Table 1. Statistics of the free-air gravity anomaly before and after the blunder removal.
Table 2. Expected global RMS-errors for the least squares modification method with
pessimistic apriori error degree variances. The signal degree variances, M=L=110,
σ = 10 mGal and ψ 0 = 3D .
Table 3. Statistical analysis fitting the 260 GPS/levelling data and gravimetric geoid models
from the absolute accuracy view before and after seven parameter fitting. (The RMS value for
four and five parameter models is given in order to compare results of the fitting between
different models).
Table 4. Evaluation of the IRG04 geoid model based on the cross-validation approach before
and after seven parameter fitting. Unit: m
Table 5. Validation of the IRG04 model versus 35 precise GPS/levelling data.
Table 6. Effect of using the different density variation models (real and isostatic) on the
IRG04 gravimetric geoid model. (Comparison is done before applying the 7 parameter fitting
approach).
Table 7. Statistical analysis fitting the 35 precise GPS/levelling data and gravimetric geoid
models from the relative view of accuracy view (by Eq. 7.13).
Table 8. Statistical analysis fitting the GPS/levelling data and GRACE/GGM02 models from
the absolute accuracy view, before and after applying the 7 fitting procedure.
Table 9. Statistical analysis of fitting between the 35 precise GPS/levelling data and
GRACE/GGM02 models from the view of relative accuracy (Eq. (8)) for GRACE/GGM02
models.
Table 10. Validation of the IRG04 and the new combined geoid model versus 35
GPS/levelling data points.
vii
TÄà{Éâz{ Åç áxÇáxá ãxÜx áxtÜv{|Çz à{x wxáxÜà ytà|zâxÄxááN w|ávÉäxÜ|Çz ÇÉà{|Çz
tÄà{Éâz{ y|Çw|Çz t ÄÉàN Åç áÉâÄ ãtá |ÄÄâÅ|Çtàxw uç t à{ÉâátÇw áâÇáN uâà vÉâÄw
ÇxäxÜ xäxÜ àÉâv{ à{x ÑxÜyxvà|ÉÇ Éy t á|ÇzÄx tàÉÅÊ
Avicenna (Ibn Sina), 980–1037 A.D.
Iranian physician and philosopher
viii
Part ONE
The Hybrid Precise Geoid Model
for Iran
1
(This page left intentionally blank)
2
Chapter 1
________________________________________________________
Introduction
"Geodesy is the discipline that deals with the measurement and representation of the earth,
including its gravity field, in a three-dimensional time varying space."
(Vanícek and Krakiwsky, 1986, p.45)
1.1 Preamble
Geodesy is an interdisciplinary science, which uses space-borne and airborne remotely sensed,
and ground-based measurements to study the shape and size of the Earth, the planets and their
changes; to precisely determine position and velocity of points or objects at the surface or
orbiting the planet, within a realized terrestrial reference system, and to apply these knowledge to
a variety of scientific and engineering applications, using mathematics, physics, astronomy and
computer science. Geodesy is closely related with other Earth Sciences like solid Earth physics,
hydrology, atmospheric sciences, oceanography, glaciology, geophysics and geology, and thus
aids our understanding of the dynamic behaviour within the solid and liquid Earth, the
movements of crustal plates and the behaviour of the oceans and atmosphere. It uses some of the
most advanced satellite measurements, electronic and computer technologies. Radio and visual
astronomy, satellite measurement of location, space-based measurements of atmospheric and
oceanic phenomena, laser and radio measurement of satellite location, use of inertial navigation
and measurement systems, gravity measurement and computer modelling are all part of the work
of geodesists.
One of the major tasks of geodesy is the determination of the geoid, which is defined an
equipotential surface the earth gravity field which coincides on average with mean sea level.
According to C.F. Gauss, the geoid is the "mathematical figure of the Earth", and in fact, of the
gravity field. The geoid surface is more irregular than the ellipsoid of revolution often used to
approximate the shape of the physical Earth, but considerably smoother than the Earth's physical
3
surface. While the latter has excursions of the over of 8 km (Mount Everest) and −11 km
(Mariana Trench), the geoid varies only about ±100 m about the reference ellipsoid of revolution.
The determination and availability of a high-resolution and more-accurate geoid model is

nowadays a necessity in several geosciences, since it serves as the reference surface of other
measurements and phenomena. The geoid is a basic information in different disciplines related to
Geomatics science from navigation, mapping and surveying to construction, detecting the
variations of the ocean currents, study of the interior properties of the Earth in geophysics,
seismology and plate tectonics. Nowadays geoid determination is getting even more crucial, due
to the development of the Global Navigation Satellite Systems (GNSS). These systems can offer
three-dimensional positioning over the world, but without having the precise geoid model, in the
field of engineering the system can be used only for two-dimensional positioning. This is due to
the fact, that global positioning systems provide ellipsoidal heights, which are geometric heights,
instead of orthometric heights, which have physical meanings. In order to convert the ellipsoidal
height into a more useful orthometric height we need to know the geoid undulation at the station.
The determination of orthometric heights by spirit levelling is known to be a time
consuming and difficult task. This is especially evident in large and topographically rough
countries like Iran, where the establishment of a high resolution levelling network for all parts of
the country would be impractical from the financial point of view. Moreover, levelling over
areas with rough terrain, like the Alborz and Zagros Mountains in north and west or central
deserts of Iran, is very tedious and time consuming. However, in order to get more accurate
results using this method we need a very high resolution and accurate geoid model.
New developments and advances in the gravity field determination from satellite tracking
have taken place in the last few years. Recent satellite gravity missions, such as the
CHAllenging Minisatellite Payload (CHAMP) and Gravity Recovery and Climate Experiment
(GRACE) have provided global scale and very accurate gravity data. However, the up-to-date,
accuracy of the global geoid modelling is a few decimetres, which is not sufficient for many
scientific and engineering applications. So, high resolution regional geoid models are still
necessary for the most practical purposes.
4
1. Introduction
1.2. Research objectives and author’s contribution
Many different methods for regional geoid determination have been proposed during the years,
each preferring its own set of techniques and philosophy, which makes it more difficult to judge
what is the best method in a certain situation. Since 1986, several gravimetric geoid models have
been published in the Iran region based on the Remove-Compute-Restore (RCR) (Forsberg 1990
and Sanso′ 1997), the Helmert’s scheme (Vaníček et al. 1995) and the ellipsoidal Bruns’ formula
Ardalan and Grafarend 2004) approaches.
Based on the primary investigations, which were performed by the currently available
gravimetric geoid models (Kiamehr 1997, 2001, 2003a, 2003b, 2004 and Kiamehr and Sjöberg
2005a), it was found that the standard deviation (SD) of fitting them to GPS/levelling data both in
the relative and absolute senses were almost the same, or in some cases worse than using the
currently available Global Geopotentioal Models (GGMs). This means that for a long period of
time we have not seen much improvement in the accuracy of geoid determination in this region.
However, much of this problem stems from the lacking quality and quantity of the data in the
area. Result of this research clearly indicates that, regardless of the computational method, the
detail assessment on choice and quality of the data plays major roles in the final result. This step
is particularly important in areas with limited data, which is the case of Iran.
Recently, the quantity and quality of terrestrial gravity data slightly increased, and specially
several new GGMs from the recent satellite gravimetric missions (e.g., GRACE) were released.
At the same time, the new high resolution Shuttle Radar Topography Mission (SRTM) global
DEM was released. The main purpose of this research is to test the potential of the Royal Institute
of Technology (KTH) approach based on the Least-Squares Modification of Stokes (LSMS)
formula (Sjöberg, 1984, 1991, 2003c and 2003d) as an optimum way for combination and
determination of a new high-resolution geoid model for Iran. This method was successfully
applied in the determination of several regional geoid models in Zambia (Nsombo 1996),
Ethiopia (Hunegnaw 2001), Sweden (Nahavandchi 1998 and Ågren et al. 2006a, 2006b) and the
Baltic countries (Ellmann 2004), which all, except Ethiopia, mostly have flat to moderate
topography. However, the determination of a geoid model in Iran is the first experience of using
the KTH approach in one of the most crucial areas in the world as viewed from the aspects of
5
topography, geoid, density variation and geodynamics. On the other hand, since 2004, many
theoretical and numerical aspects of this method have been investigated and completed, which
have been tested as a complete package in this research.
Through the research, different theoretical and practical aspects in geodesy and
geodynamics were studied. The most highlighted subjects are:
• The development and use of the cross-validation technique for detecting outliers and the
evaluation of gridding methods of gravity data,
• Evaluation of different DEMs and studying the effect of using high resolution DEMs in
accurate geoid modelling,
• The evaluation of different GGMs,
• The establishment of a density variation model for Iran and studying its effect in geoid
modelling,
• Establishment of a new height datum for Iran based on the corrective surface idea and
• Studying the impact of the geoid model on tectonic activity in Iran.
The results of this research can be used as a procedure for gathering, evaluating and
combining different data for the determination of gravimetric geoid models in developing
countries (specially in mountainous areas), where limited ground gravity data is available. The
dissertation consists of two parts: Part One (review part), which includes 10 chapters, and Part
Two as the publication part, including in 9 papers.
1.2.1. Review of Part One

Chapter 1 starts with the introduction and explaining the general scope of the thesis. In
Chapters 2 and 3 we review the mathematical aspects of the KTH approach to determine the
geoid based on the LSMS formula, including its additive correction terms. The Stokes method
and the mathematical model for the determination of the regional gravimetric geoid are fully
described in this chapter. In Chapter 4 we explain the procedure for the creation and evaluation
of the basic data, which will be used in the geoid determination, including the new generation
and refinement of the gravity database, the choice of the best GGM and DEM models and
GPS/levelling data. Chapter 5 continues with a brief explanation of choosing the best
modification parameters for the construction of the new Iranian geoid model (IRG04). Also, an
6
1. Introduction
approach for evaluating the effect of the lateral topographical mass density variation on the geoid
is proposed in Chapter 6. The procedure for the establishment of a new height datum for Iran by
combining the gravimetric and GPS/levelling geoid models is explained in Chapter 7.
In Chapter 8 we evaluate and compare the accuracy of the newly released geoid models
and the latest geoid models in the study area versus the GPS/levelling data in the absolute and
relative senses. The internal accuracy of the geoid model was estimated by means of the
expected global mean square error, whereas the GPS-levelling data is applied for an external
evaluation of the accuracy of the computed geoid models.
During this research (2003) a large earthquake disaster happened in the south-east of Iran
(Bam city), which for a while focused this author to work on anther major task of geodesy,
namely geodynamics. Chapter 9 includes some of geodynamical experiences of the author
about the application of the geoid in studying the correlation between different geo-spatial data
and earthquakes. The impact of the geoid model was studied for clarifying the tectonic
boundaries of Iran. Also, we did a pilot research project about the application of the 3D
deformation analysis in a GPS network in the south part of Sweden (Skåne area). Results of
this research, including a brief review of the method and final results are presented in the
second part of this chapter.
Finally, the study results are concluded in Chapter 10, which contains a general summary
of all topics, a discussion of the most important results and some recommendations for
future research.
1.2.2. Review of Part two
All of the research contributions described herein have already been accepted/published or
submitted in peer-reviewed international scientific journals. The publications will be referred to
as PAPERS A-I as follows:
7
PAPER A:
Kiamehr R (2005) Qualification and refinement of the gravity database based on cross-validation
approach, A case study of Iran, Proc. Geomatics 2004 (84) Conferences, National Cartographic
Centre of Iran, Tehran, Iran, Revised version, submitted J. Acta Geodaetica et Geophysica
PAPER B:
PAPER C:
Kiamehr R and Sjöberg LE (2005b) Effect of the SRTM global DEM on the determination of a
high-resolution geoid model: a case study in Iran, J. Geodesy, 79(9):540-551
PAPER D:
combining limited ground data with satellite-based global geopotential and topographical models:
A case study of Iran, J. Geodesy , 79(10,11): 602–612
PAPER E:
Kiamehr R (2006b) Hybrid precise gravimetric geoid model for Iran based on recent GRACE and
SRTM data and the least squares modification of Stokes formula, J. Physics of Earth and Space,
32(1) 7-23
PAPER F:
GPS/levelling geoid models, in press Dynamic Planet 2005 IAG Symposia Series Springer
Verlag, Revised version accepted and in press, Acta Geodaetica et Geophysica, 42(1)
PAPER G:
Kiamehr R (2006d) The Impact of lateral density variation model in the determination of precise
gravimetric geoid in mountainous areas: a case study of Iran, Accepted and in press Geophysical
Journal International.
PAPER H:
Kiamehr R and Sjöberg LE (2006) Impact of the precise geoid model in studying tectonic
structures- A case study in Iran, J. Geodynamics, 42(2006) 1-11
8
1. Introduction
PAPER I:
Kiamehr R and Sjöberg LE (2005c) Surface Deformation Patterns Analysis using 3D Finite
Elements Method: A case study in Skåne area, Sweden, J. Geodynamics, 39(4) 403-412
1.2.3 Other Publications and Presentations
The following articles published or presented during the PhD studying period by author which
are not included in this dissertation:
1. Kiamehr R (2003a) Comparison Relative Accuracy of EGM96 and Iranian Gravimetric

Geoid, Proc. 6Th International conference of Civil Eng., Isfahan University of
Technology, Iran, 537-544.
2. Kiamehr R (2003b) The estimation of relative accuracy of GPS/levelling in Iran (In
Persian), Journal of Sepehr, 11(42):13-15.
3. Kiamehr R (2003c) Multi Object Optimization of the Geodetic Network, Proc. Annual
Geomatics 82 Conferences (CD-ROM), Iranian National Cartographic Centre, Tehran,
Iran.
4. Kiamehr R (2003d) Comparison of Ambiguity Resolution Using LAMBDA and KTH
Method, International Symposium on GPS/GNSS, November 2003, Tokyo, Japan.
5. Kiamehr R (2004) The Relative Accuracies of Recent Satellite Gravimetric Models in
Iran, Proc. Geoid and Space Missions Symposium 2004 (GGSM2004) (CD-ROM), August
30th-September 3rd, 2004, Porto, Portugal.
6. Kiamehr R (2004) Success Rate Determination of GPS Ambiguity Resolution Using
Syntactic Data, Position Location and Navigation Symposium (PLANS), California, USA.
7. Ågren J, Kiamehr R and Sjöberg LE (2006) Progress in the Determination of a
Gravimetric Quasigeoid Model over Sweden, Nordic Geodetic Commission General
Assembly, May 29 to June 2, 2006, Copenhagen, Denmark.
8. Ågren J, Kiamehr R and Sjöberg LE (2006) The Swedish geoid as evaluated by the
method of least-squares modification with additive, the first International Symposium of
The International Gravity Field Service (IGFS), August 28 – 1 September 2006, Istanbul,
Turkey.
9
10
Chapter 2
________________________________________________________
Geoid determination based on the least-squares modification of
Stokes’ formula
2.1 Modification of Stokes’ formula
In 1849 one of the most important formulas in physical geodesy was published. This formula
allows the determination of the geoidal height (N) based on gravity data. It is known as Stokes'
formula or Stokes' integral after George Gabriel Stokes. It reads:
R
N=
4πγ ∫∫σ S (ψ ) Δgdσ , (2.1)
where R is the mean Earth’s radius, ψ is the geocentric angle, Δg is the gravity anomaly, dσ is an
infinitesimal surface element of the unit sphere σ, γ is normal gravity on the reference ellipsoid
and S(ψ) is the Stokes function. The orthogonality properties between Legendre polynomials
Pn ( cosψ ) on the sphere allows us to present S(ψ) as:
∞
2n + 1
S (ψ ) = ∑ Pn ( cosψ ). (2.2)
n =2 n − 1
The integration expressed in equation (2.1) must be carried out over the whole earth. In
practice, however, the area of integration is often limited to a spherical cap around the
computation point. Molodensky et al. (1962) showed that the truncation error of the remote zone
11
can be reduced by a modification of Stokes’ formula, which combines the terrestrial gravity
anomalies and the long wavelength (up to degree M) as contribution from a GGM.
With the satellites came the first Global Geopotential Model (GGM) and the possibility to
generate global geoid models. By combining the information from the GGMs with Stokes'
integration over local gravity data, local geoid models may be estimated (e.g. Rapp and Rummel
1975). The GGM gives the long-wavelength information about the geoid and the geopotential,
while the local gravity data gives the short-wavelength information. Due to the use of local
instead of global gravity data coverage, Stokes' kernel will be truncated at the outer range of the
data coverage. This leads to truncation errors due to the loss of gravity information from the
neglected outer zones. These errors may either be ignored or reduced by the modification of
Stokes' kernel.
As mentioned above, the minimization of the truncation errors was first suggested by
Molodensky et al. (1962) by applying a modification to Stokes' kernel. This idea has been further
investigated and the modifications suggested can be divided into two categories: deterministic
and stochastic types of modification. The deterministic approach intends to minimize the
truncation error due to the neglected zones by removing the low degree terms (i.e. the long-
wavelength part) of Stokes' kernel and high-pass filtering the gravity anomalies (Wong and Gore
1969). The major strategies have been suggested by the de Witte (1967); Wong and Gore (1969);
Meissl (1971) in addition to Molodensky et al. (1962). Combinations and variants of these
strategies have been suggested by e.g. Heck and Grüninger (1987) and Featherstone et al. (1998).
The stochastic approach applies additional information about the potential coefficients and the
gravity anomaly errors in combination with least squares modification of Stokes' kernel to
minimize the expected global Mean Square Error (MSE) (e.g. Sjöberg 1984 and 1991).
Assuming a cap σ 0 of spherical radius ψ 0 of integration around the computation point,
Sjöberg (2003d) presented a simple and general modification model for Stokes’ formula by
defining two sets of arbitrary modification parameters ( sn and bn ) as follows (cf. Sjöberg 2003d):
M
c
N = ∫∫ S L (ψ ) Δgdσ + c∑ bn Δgn , (2.3a)
EGM
2π σ0 n=2
12
2. Geoid determination based on the least-squares modification of
where b n = Q nL + s n , c = R / ( 2γ ) and Δg nEGM is the Laplace harmonics of degree n can be
calculated from a GGM (Heiskanen and Moritz 1967, p. 89)
n+2
GM ⎛ a ⎞ n
Δg nEGM = ⎜ ⎟ ( n − 1) ∑ CnmYnm , (2.3b)
a2 ⎝ r ⎠ m =− n
where a is the equatorial radius of the reference ellipsoid, r is the geocentric radius of the
computation point, GM is the adopted geocentric gravitational constant, the coefficients C nm are
the fully normalised spherical harmonic coefficients of the disturbing potential provided by the
GGM, and Y nm are the fully-normalized spherical harmonics (Heiskanen and Moritz 1967, p. 31).
The modified Stokes function is expressed as
L
2n + 1
S L (ψ ) = S (ψ ) − ∑ s k Pk ( cosψ ). (2.4)
n =2 2
The upper limit L is arbitrary and generally not equal to M. The truncation coefficients
Q nL can be calculated by
∞
2k + 1
Q nL = Q n − ∑ s k e nk , (2.5)
n =2 2
where the Molodensky’s truncation coefficients Q n can be expressed
π
Q n = ∫ S (ψ )Pn ( cosψ ) sinψ dψ , (2.6)
ψ0
and e nk are the so-called Paul’s coefficients (Paul 1973), being functions of ψ 0 . By utilizing the
error estimates of the data (terrestrial gravity anomalies Δg and the spherical harmonics Δg nEGM ),
13
and the some approximations (both theoretical and computational), we arrive at an estimate of the
geoid height that we call the approximate geoidal height.
Due to the presence of errors in the estimation of GGM coefficients and terrestrial gravity
data, it is possible to rewrite the estimator of Eq. (2.3a) in the following spectral form (cf. Sjöberg
2003d, Eq. 2.7)
∞
⎛ 2 ⎞
− QnL − sn* ⎟ ( Δg n + ε nT ) + c∑ ( QnL + sn* )( Δg n + ε nS ),
M
N = c ∑ ⎜ (2.7)
n=2 ⎝ n − 1 ⎠ n=2
where ε Tn and ε ns are the spectral errors of the terrestrial and GGM derived gravity anomalies,
respectively. The modification parameters are
sn* =
RSs
n if 2 ≤ n ≤ L
. (2.8)
T0 otherwise
The main objective of the modification procedure is to minimize effects of the errors in
the estimation of the geoid. The modification methods proposed by Sjöberg (1984, 1991, 2003b
and 2003c) allow for minimization of the truncation errors, the influence of erroneous gravity
data, geopotential coefficients and combination of different data sources in the least-squares
sense and at the same time in an optimum way.
The terrestrial gravity observations distributed in Iran are non-homogeneous and often
affected by systematic errors. When the recent GRACE model with very high accuracy in
the low to medium degrees is used, it becomes important to use a kernel modification that
effectively filters out the long-wavelength errors from the gravity anomalies. For this
reason, we need a proper weighting scheme for the gravity data as an a priori or empirical
stochastic model. Of course, the true errors for the gravity data are not well-known. Here we
discuss this problem shortly, as the error degree variances for GGMs and (especially GRACE
only satellite models) will be estimated quite accurately, but the most crucial problem here is that
the gravity anomaly error degree variances are not known. However, the least-squares method is
rather insensitive to the choice of weights, as is well known from other areas of geodesy.
14
According to the numerical investigation performed by Ågren (2004a), the insensitivity to the
weighting is illustrated by using different optimistic and pessimistic prior weighting
values. Furthermore, even though the results are a bit worse, compared to the situation
when the correct a priori errors are utilized, the results are still better than for the other
modification methods explained above (Ågren 2004a). These numerical investigations
clearly show that the least-squares method is a good alternative even though the a priori
error degree variances for the gravity data are not exactly known. Some recent successful
results of applications of these methods can be found in Hunegnaw (2001), Nahavandchi and
Sjöberg (2001), Ellmann (2001) and Ågren (2004a).
Based on the spectral form of the “true” geoidal undulation N (Heiskanen and Moritz
1967, p. 97),
∞
2Δg n
N =c∑ , (2.9)
n =2 n − 1
the expected global MSE of the geoid estimator Ñ can be written:
2
⎧ 1 ⎫ ∞
⎛ 2 ⎞
∫∫σ ( ) dσ ⎬ = c 2 ∑ ( bn2 dcn ) + c 2 ∑ ⎡⎢( bn* − QnL − sn* ) cn + ⎜
M
− QnL − sn* ⎟ σ n2 ⎤⎦ ,
2 2
mN2 = E ⎨ N − N
⎩ 4π ⎭ n=2 n=2 ⎣ ⎝ n − 1 ⎠
(2.10)
where E{} is the statistical expectation operator, c n is the gravity anomaly degree variances, σ n2
is terrestrial gravity anomaly error degree variances, dc n is the GGM derived gravity anomaly
error degree variances and
⎧b if 2 ≤ n ≤ L * ⎧ sn if 2 ≤ n ≤ L
b n* = ⎨ n , sn = ⎨ . (2.11)
⎩0 otherwise ⎩0 otherwise
The first, middle and last terms on the right-side of Eq. (2.10) represent the contributions
due to errors of the geopotential model, the truncation and the influence of erroneous terrestrial
data, respectively. For all the data, the errors are assumed to be random with expectations zero, so
15
the norm of the total error can be obtained by adding their partial contributions. However, in
practice, the GGM and ground gravity data are often correlated. For example, because of using
surface gravity anomalies in the construction of the EGM96 model, we can assume correlations
between these two kinds of data. However, if one utilizes just “satellite-only” harmonics, this
correlation can be avoided. On the other hand, we know that the terrestrial data information is
comprised in the higher degrees of the GGM, but most of the geoid power is in the lower degrees,
so we might assume that the influence of the correlations is insignificant.
2.2. Signal and noise degree variances
The gravity anomaly signal degree variances, the terrestrial gravity anomaly error degree
variances and the GGM derived gravity anomaly error degree variances can be computed as
follows:
1 ⎧ 1 ⎫
∫∫ (ε ) dσ ⎬
2
cn = ∫∫ Δg n d σ , σ n2 = E ⎨
2 T
(2.12a)
4π ⎩ 4π
n
σ σ ⎭
and
⎧ 1 ⎫
∫∫σ (ε ) d σ ⎬⎭ ,
2
dc n = E ⎨ S
(2.12b)
⎩ 4π
n
respectively. In this study we estimate the lower signal degree variance c n by using the spherical
harmonic coefficients Cnm and S nm of the disturbing potential from a GGM models as:
( GM )
2
( n − 1) ∑ ( Cnm2 + Snm2 ) ,
n
2
cn = (2.13)
a4 m=0
16
In practice the infinite sum in Eq. (2.10) must be truncated at some upper limit of
expansion, say n max = 10800 . The higher signal degree variances can be generated synthetically.
Among three well known models for estimation of the signal gravity anomaly degree variances
[e.g. Kaula 1963, Tscherning and Rapp 1974 and Jekeli 1978], the Tscherning and Rapp model
yields the most realistic values for the gravity anomaly truncation RMS errors and gives
reasonable RMS values for geoidal heights (Ågren 2004a). So we decided to use this model to
account for the highest degrees of gravity anomaly degree variance, which are defined by
cn = A
( n − 1) ⎛ R B ⎞
n +2
, (2.14)
( n − 2 )( n + B ) ⎜⎝ R ⎟
⎠
with the parameters chosen to R=6371 km, A=225 mGal 2 , B=4 and R − R B = 3.5 km. However,
this model is valid just for the gravity field uncorrected for any topographic effects. The error
(noise) anomaly degree variance of the erroneous potential coefficients with standard errors
dC and d Snm ( cn ) derived from the GGMs.
nm
( GM )
2
( n − 1) ∑ ( dC2 )
n
2
dcn = + d Snm
2
. (2.15)
a4 m =0
nm
For estimating of the error degree variances for the terrestrial gravity anomalies, two
different error degree variance models will be used to represent gravity anomaly errors, namely
the uncorrelated and the reciprocal distance models. For the latter an isotropic error degree
covariance function C (ψ ) is presented in the closed form (Sjöberg 1986):
⎡ 1− μ ⎤
C (ψ ) = c1 ⎢ − (1 − μ ) − (1 − μ ) μ cosψ ⎥ , (2.16)
⎢⎣ 1 − 2 μ cosψ + μ 2 ⎥⎦
17
and the degree variances σ n2 for the reciprocal distance type function is given by
σ n2 = c1 (1 − μ ) μ n , 0 < μ < 1 . (2.17)
where c1 and μ are constants. From the closed form Eq. (2.16) the parameters c1 and μ can be
computed from a knowledge of the of the variance C ( 0 ) and correlation lengthψ 1/ 2 . In this
research Ridder’s method (Press et al. 1992) was used to find μ from ψ 1/ 2 . Then c1 is computed
as c1 = σ 2 / μ 2 , which follows from Eq. (2.16). Both c1 and μ are depending on the situation of
the study area and the data. The value ψ 1/ 2 = 0.1D gives a reasonable result with the Iran data
which it is matching with the previous studies by (Nahavandchi 1998, Ellmann 2004 and Ågren
2004a). By choosing ψ = 0 in Eq. (2.16) we get:
σ 2 = C ( 0 ) = c1μ 2 ⇒ C (ψ 0 ) = 0.5c1μ 2 , (2.18)
with the solution for μ = 0.99899012912 can be found iteratively (Ellmann 2004).
In the second error model it is assumed that the observation noise is uncorrelated, which
is approximately modelled by band-limited white noise with constant degree-order variances,
σ n2,Δg
σ nm ,Δg = (Rummel 1997 and Jekeli and Rapp 1980). If it is considered that the number of
( 2n + 1)
potential coefficients between degree 2 and the Nyquist degree M N is equal to ( M N + 1) − 4 , the
2
degree variances can be derived as:
σ2
σ nm ,Δg = ( 2n + 1) . (2.19)
(M N + 1) − 4
2
Numerical results of this research show that the second model gives the minimum
expected mean square error and fitting versus the GPS/levelling data. (See Chapter 7 for more
details).
18
2.3. Determination of the least-squares modification parameters
To obtain the Least-Squares Modification (LSM) parameters, Eq. (2.10) is differentiated with
respect to s n , i.e., ∂m N2 / ∂s n . The resulting expression is then equated to zero, and the
modification parameters s n are thus solved in the least-squares sense from the linear system of
equations (Sjöberg 2003d):
∑a
r =2
kr .s r = hk , k = 2,3,..., L , (2.21)
where akr and hk are modification coefficients, which can be expressed via Q n , e nk , c n , dc n and
σ n2 by:
2r + 1 2 2k + 1 2
akr = ( ak2 + dck* ) δ kr − ( σ k + dcn* ) ekr − ( σ r + dcr* ) erk
2 2
(2.22)
2k + 1 2r + 1 ∞
+ ∑ enk enr (σ n2 + dcn* ) ,
2 2 n=2
and
2σ k2 2k + 1 ∞ 2
hk =
k −1
− Qk (σ k2 + dcn* ) + ∑
2 n=2
Qn enk (σ n2 + dcn* ) −
n −1
enkσ n2 ), (2.23)
where
⎧⎪dcn for 2 ≤ n ≤ M
dcn* = ⎨ . (2.24)
⎪⎩cn for n > M
Depending on the quality of local gravity quality, the chosen radius of integration (ψ 0 )
and the characteristics of the GGM, the modification parameters sn vary. The system of equations
in Eq. (2.21) is so ill-conditioned that it often cannot be solved by standard methods like
Gaussian elimination. To overcome this problem, Ellmann (2004) and Ågren (2004a) used the
19
standard Singular Value Decomposition (SVD) procedure provided e.g. by Press et al. (1992).
Ågren (2004a) concluded that the instability of the optimum choice of parameters for the
unbiased LSM method is completely harmless, and the truncated SVD can be used to obtain a
useful solution. After the numerical solution of sn , the corresponding coefficients bn = QnL + sn are
computed. It can be summarized here that the modification methods by Sjöberg (1984, 1991 and
2003d) attempt via minimization of the global MSE to reduce any error in geoid modelling. In
PAPER (E) and Chapter 7, we explain the detail procedure in choosing the final LSM parameters
for our application to Iranian data.
20
Chapter 3
________________________________________________________
Additive Corrections in the KTH Approach
In the determination of the geoid by Stokes’ formula, Eq. (2.1), it is necessary that there are no
masses outside the geoid and the gravity data should be reduced to sea level. However, because
of the presence of topography and atmospheric masses (forbidden masses) above the geoid
surface, we need to add some correction terms to fulfil these necessary conditions.
In the KTH computational scheme for geoid determination (Sjöberg 2003c), we use the
surface gravity anomalies and the GGM for the determination of approximate geoidal heights
( N 0 ). After that, all necessary corrections are added directly to N 0 . In contrast, in the classical
approaches, these corrections are usually applied so that in the first step the surface gravity
anomalies are corrected by removing the effects of topographic and atmospheric external masses
or reducing them inside the geoid as a direct effect, and then, after applying Stokes’ integral, their
effects are restored (indirect effect). Besides, since the gravity anomalies in Stokes’ formula must
refer to the geoid to satisfy the second condition, a reduction of observed gravity from the Earth
surface to the geoid is necessary. This step is called downward continuation (DWC). In the KTH
approach, all these separate effects are replaced by a total topographic effect (total effects of
topographic and downward continuation). The computational procedure for estimation of the
geoidal height N̂ can be summarized by the following formula:
Nˆ = N 0 + δ N comb
Topo
+ δ N DWC + δ N comb
a
+ δ Ne . (3.1)
21
where δ N comb
Topo
is the combined topographic correction, which includes the sum of the direct and
indirect topographical effects on the geoid heights, δ N DWC is the downward continuation effect,
δ N comb
a
is the combined atmospheric correction, which includes the sum of the direct and indirect
atmospherical effects, and δ N e is the ellipsoidal correction for the spherical approximation of the
geoid in Stokes’ formula to ellipsoidal reference surface.
3.1. The Topographic Corrections
As mentioned before, the combined topographic effect is the sum of the direct and indirect
effects, and it can be added directly to the approximate geoidal height values derived in Eq. (2.8)
as:
2π G ρ
δ N comb
Topo
= δ N dir + δ N indir − H2, (3.2)
γ
where ρ is the mean topographic mass density and H is the orthometric height. This method is
independent of the selected type of topographic reduction (Sjöberg 2000 and 2001a). On the other
hand, the direct topographic effect which is usually affected by a significant terrain effect,
including possible lateral density variations, is cancelled in the combined topographic effect on
the geoid.
This formula is very simple and computer efficient. Because of the fact that rough surface
gravity anomalies are Stokes integrated in the KTH approach, some important comments must be
taken into account in using the method. Since the rough surface anomalies are often sampled too
sparsely, the numerical errors of the Stokes’ integration (discretisation error) becomes large.
These errors can be reduced by using a special interpolation technique (Ågren 2004a and
Kiamehr 2005b, p. 548). First, make a smoothing topographic correction that result in reduced
gravity anomalies, which are smoother than the original ones. Then, in the next step, the
observations are interpolated to a denser grid, and the topographic correction is finally restored,
i.e. the masses are restored. The standard planar Bouguer or Residual Terrain Model (RTM)
anomaly can be used in the interpolation stage.
22
3. Additive Corrections in the KTH Approach
By using this interpolation technique on the smooth topographically reduced gravity

anomalies, the combined estimator can be expected to yield more or less similar accuracy as when
the remove-compute-restore estimator is used. Ågren (2004a) shows, using synthetic models, that
this interpolation step reduces the discretisation errors significantly in the combined approach of
Sjöberg (2003c), and that it improves the geoid determination result. Notice that a good DEM
should be available with at least the same resolution as the new interpolated grid. Preferably it
should be denser.
3.2. The Downward Continuation Correction
The analytical continuation of the surface gravity anomaly to the geoid is a necessary correction
in the application of Stokes’ formula for geoid estimation. It means that, after reduction of the
topographic effect, the observed surface gravity anomalies must also be downward continued to
the geoid. This correction is applied to the surface gravity anomalies in the classical approaches.
Traditionally, different methods are used for DWC, but using the inversion of Poisson’s
integral (Heiskanen and Moritz 1967; Martinec 1998; Press et al. 1992) is the most common one.
The inversion of Poisson’s integral is converging well for a low resolution gravity grid
(e.g. 5′ × 5′ ), but the convergence is doubtful for a denser grid (Martinec 1998). On the other
hand, this method has large discretisation errors, especially in rough areas, and it is also an
extremely time consuming procedure (Martinec 1998). Sjöberg (2003a) designed a new method
for the DWC of the full field gravity anomalies. In this method, the DWC effect is computed
directly for the geoid height rather than for the gravity anomaly. (He treats all corrections in this
manner as an adaptive scheme.) The downward continuation effect in this case is given by:
(ψ ) ( Δg * − Δg ) dσ ,
c
δ N DWC =
2π ∫∫ S
σ0
L (3.3)
where Δg is the gravity anomaly at the surface computation point P and Δg * is the corresponding
quantity downward continued to the geoid. The use of the smoothed data in the Stokes operator
( Δg * − Δg ) makes this formula particularly advantageous, and we can therefore obtain more
23
accurate results. To sum up, the final formulas for Sjöberg’s DWC method for any point of
interest P based on LSM parameters can be given by (for more details, see Ågren 2004a):
δ N DWC ( P) = δ N DWC
(1)
( P) + δ N DWC
L1, Far
+ δ N DWC
L2
( P) , (3.4)
where
⎛ Δg ( P) N P0 1 ∂Δg ⎞
δN (1)
= HP ⎜ +3 − HP ⎟, (3.5)
⎝ γ rP 2γ ∂r P
DWC
⎠
and
⎡⎛ R ⎞ n + 2 ⎤
= c ∑ ( s + Q ) ⎢⎜ ⎟ − 1⎥ Δg n ( P ) ,
M
δN L1, Far
DWC
*
n
L
n (3.6)
n=2 ⎢⎣⎝ rP ⎠ ⎥⎦
and
⎛ ∂Δg ⎞
⎟ ( H P − H Q ) dσ Q ,
c
δ N DWC = ∫∫ S (ψ ) ⎜
L2
(3.7)
2π ⎝ ∂r P ⎠
L
σ 0
where rP = R + H P , σ 0 is a spherical cap with radius ψ centred around P and it should be the
same as in modified Stokes formula, H is the orthometric height of point P and gravity gradient
∂Δg
in point P can be computed based on Heiskanen and Moritz (1967, p.115):
∂r
∂Δg R2 ΔgQ − Δg P 2
∂r
P=
2π ∫∫σ A 3
0
dσ Q −
R
Δg ( P) , (3.8)
0
ψ PQ
where A 0 = 2 R sin .
2
24
3.3. The Ellipsoidal Correction
Stokes’ formula determines the geoidal height from the gravity anomaly on the sphere with a
spherical approximation of 0.3%. However, in theory, the boundary surface for the gravity
anomaly is the geoid, which can be better approximated by an ellipsoid. The spherical
approximation can cause a relative error of 0.3% in the estimation of a geoid model. In precise
geoid determination, this approximation should be taken into account by some correction term
(ellipsoidal correction). Different authors have studied the ellipsoidal correction for the original
Stokes formula, see e.g. Molodensky et al. (1962), Moritz (1980), Martinec and Grafarend
(1997), Fei and Sideris (2000), Heck and Seitz (2003) and the references therein. A new integral
solution for the ellipsoidal correction was published by Sjöberg (2003b). The ellipsoidal
correction for the original and modified Stokes formulas is derived by Sjöberg (2003e) and
Ellmann and Sjöberg (2004) in a series of spherical harmonics to the order of e 2 , where e is the
first eccentricity of the reference ellipsoid.
Nowadays most local geoid models are determined using a modified version of Stokes’
formula, which means that the major parts of the long-wave features of the geoid are estimated
from a GGM. As the GGMs can correctly be applied at sea level (approximately the ellipsoid),
the remaining ellipsoidal correction for Stokes’ formula is limited to the integration cap. The
ellipsoidal correction to the modified Stokes’ formula ( δ N e ) to order e 2 is (Sjöberg 2004):
∞
⎛ 2 ⎞⎛ a − R EGM ⎞
Δg n ( P ) + (δ ge )n ⎟ , (3.9)
R a
δ Ne ( P ) =
2γ
∑ ⎜⎝ n − 1 − s
n=2
*
n − QnL ⎟⎜
⎠⎝ R R ⎠
where sn* = sn if 2 ≤ n ≤ M and sn* = 0 otherwise. Furthermore,
(δ ge )n = 2a ∑ {⎡⎣3 − ( n + 2 ) Fnm ⎤⎦ Tnm − ( n + 1) GnmTn−2,m − ( n + 7 ) EnmTn+2,m }Ynm ( P ) ,

e2 n
(3.10)
m =− n
where Tnm are spherical harmonic coefficients for the disturbing potential. See Sjöberg (2004c)
for the ellipsoidal coefficients Enm , Fnm and Gnm . Ellmann and Sjöberg (2004) concluded that the
absolute range of the ellipsoidal correction in the least-squares modification of Stokes’ formula
does not exceed the cm level with a cap size within a few degrees. It proves the common
25
assumption that the ellipsoidal correction could be neglected in the modified Stokes formula and
in many practical applications. However, in regions with large gravity anomalies or large
integration cap, the ellipsoidal correction cannot be ignored in precise geoid modelling.
3.4. The Atmospheric Correction
As mentioned before, the presence of forbidden atmospheric masses outside the geoid surface
(same as topographic masses), means that it is necessary to add an additional correction term to
satisfy the boundary condition in Stokes’ formula. In the computation of the atmospheric
correction in the classical IAG approach, we suppose the Earth is a sphere with a spherical
atmospheric ring and the topography of the Earth is completely neglected (Moritz 1992). Based
on this assumption, some tabulated correction terms are usually added as a direct effect to gravity
anomaly before using Stokes’ formula, and the indirect effect is so small that it is usually
neglected.
It was emphasised by Sjöberg (1998, 1999b, 2001a and 2006) that the application of the
IAG approach using a limited cap size, and especially in the modified Stokes’ formula, can cause
a very significant error in the zero-order term (more than 3 m). In the KTH scheme, the combined
atmospheric effect δ N comb
a
can be approximated to order H by (Sjöberg and Nahavandchi 2000)
2π R ρ 0 Μ
⎛ 2 ⎞ 2π R ρ 0 ∞
⎛ 2 n+2 L⎞
δ N comb
a
( P) = −
γ
∑ ⎜
n=2 ⎝ n − 1
− sn − QnL ⎟ H n ( P ) −
⎠ γ
∑ ⎜
n = M +1 ⎝ n − 1
− Qn ⎟ H n ( P ) ,
2n + 1 ⎠
(3.11)
where ρ 0 is the atmospheric density at sea level, and H n is the Laplace harmonic of degree n for
the topographic height. The elevation H of the arbitrary power v can be presented to any surface
point with latitude and longitude (ϕ , λ ) as
∞ n
H v (ϕ , λ ) = ∑ ∑H v
Y
nm nm (ϕ , λ ) , (3.13)
m = 0 m =− n
26
where H nm
v
is the normalized spherical harmonic coefficient of degree n and order m that can be
determined by the spherical harmonic analysis
1
= ∫∫σ H (ϕ , λ )Y (ϕ , λ ) dσ . (3.14)
v v
H nm
4π
nm
For more details about the properties of the DEM model which used to generate the normalized
spherical harmonic coefficient H nm
v
, see Section 4.3.
27
28
Chapter 4
________________________________________________________
Evaluation and construction of the base data
4.1. Qualification and creation of gravity anomaly database
The computation of gravimetric geoid models for the Iranian region suffers mainly from the few
gravity observations available. The total area of Iran is 1,648,195 km 2 , so it is simple to show
that we have about one gravity point per 65 km 2 . The largest gap areas are mostly located in the
Zagros and Alborz mountain areas, Lout and Kavir central desert areas, Sistan & Balochestan
state (in south-east of the country) and the marine areas of the Persian Gulf, Oman and Caspian
seas. Fig. 4.1 shows the distribution of the gravity data in Iran. The quality of the data is
questionable because they have been gathered from different organisations for different
purposes and with various accuracies during long time. In order to obtain reasonable
results, we started this research by collecting and investigating the gravity data available
from different sources. Thus, all available gravity data has been collected for both land and
marine regions and edited by a blunder-removal processing scheme to generate an optimal
gravity dataset for use in geoid determination.
We propose a technique for precise cleaning of the gravity anomaly database based on the
cross-validation approach. In this technique, the terrestrial gravity anomaly is compared to a
GGM and the effect of topography is taken into account in this comparison. The efficiency of the
cross-validation technique is illustrated both in outlier detection and in the choice of the proper
gridding technique as a case study in the construction of the Iranian new gravity database. The
points removed represent 4.3% of the total database, while those remaining were 25105 gravity
29
observations. The mean value of the statistics before and after removing the outliers were 992
and 234 mGal, respectively, which shows a significant refinement in the gravity database (Table
1). (For more details, see Kiamehr 2005; PAPER A).
Table 1. Statistics of the free-air gravity anomaly before and after the blunder removal. Unit: mGal
Gravity Data No. of data Min. Max. Mean RMS

EGM96 Model -151.94 168.47 5.68 45.97
BGI (International Gravimetric Bureau) 9566 -163.20 234.60 9.36 50.23
NCC (National Cartographic Centre) 8949 -292.99 1220.11 94.88 280.06
Ship-borne 7610 -128.46 40.04 -53.85 32.58
All Data (Before outlier) 26125 -343.03 992.27 -83.61 136.65
All Data (After outlier) 25105 -163.2 234.6 -7.08 54.83
Database (Filled Gaps) 27172 -163.2 234.6 -6.34 54.04
Fig. 1. Distribution of the gravity anomaly data (BGI, NCC and Ship-borne data presented in blue, red and
green colours, respectively)
30
4. Evaluation and construction of the base data
The geoid determination methods can be classified mainly in two categories: one uses
reduced gravity data (e.g., RCR which also use topography and EGM reduced data) and other
uses unreduced data (e.g., combined approach of Sjöberg 2003c). The main problem of using the
unreduced gravity data in the combined model is that the free-air anomaly is known to be more
sensitive to the topography. If there is rough topography in the computation area (e.g., Iran), the
free-air anomalies will be rough, and due to loss of short-wavelength gravity information (so
called discretisation error), the interpolation cannot always be successful. In order to overcome
this problem, a special interpolation techniques was used in the gridding scheme of the gravity
data (For more details, see Section 3.1 of PAPER A). The numerical investigation shows the
effect of this interpolation step in reducing the discretisation errors in the KTH approach of
Sjöberg (2003c). Based on the 1 km GLOBE (2003) DEM, we found a significant improvement
in the order of 0.34m in the approximate geoid height between the unreduced and reduced gravity
grid. (See Kiamehr and Sjöberg 2005b; PAPER C, Table 4).
The predicted 80′′ × 90′′ grid of free-air gravity anomalies is presented in Figure 2. Within
the whole target area free-air anomalies vary from -182 to +352 mGal.
4.2. The GPS/levelling data
In this research GPS/levelling data were used as an external and independent tool for the
estimation of the absolute and relative accuracy of the different global and local gravimetric
geoid (or DEM) models. (For more details, see Ch. 7.)
The establishment and measurement of the Iranian GPS network started in August 1988
using single frequency GPS receivers. The observations and adjustment of this network was
renewed and completed in recent years using dual frequency GPS receivers and data processed
with Bernese software (Beutler et al., 1996). From the 260 available GPS/levelling points, 35
points belong to the first-order and the rest belong to second-order national GPS and levelling
networks (Fig. 3). The mean standard deviation of the geodetic heights was estimated to 0.2 m.
(Nankali, 2005)
31
Fig. 2. The predicted 80′′ × 90′′ free-air anomaly grid. Unit: mGal
Fig. 3. Distribution of the GPS/levelling data in Iran. Red and black points indicate first and second-order GPS and
levelling networks, respectively. Five traverses are chosen to studying the effect of topography in different areas.
Unit: m
32
The estimated absolute accuracy for the first-order spirit levelled heights (35 points) is
about 0.7 m. Different sources of errors e.g., neglecting the effect of the Sea Surface Topography
in the 1989 adjustment of levelling networks (Hamesh 1991 and Abbolgasem 1994), the presence
of systematic errors (e.g., refraction, staff settlement, neglecting the correction of gravity on
orthometric heights) and some uncertainty in definition and establishment of the height datum in
the adjustment of the network are reduced the absolute accuracy of the orthometric heights.
However, the accuracy of the second-order levelling network should be lower than the first-order
network, because it was computed based on the fixed first-order network stations. From the
view of the relative accuracy, fortunately the field observation of GPS and levelling networks
were performed with good observation and instrumentation standards under strict rules.
According to Hamesh (1991), the average accuracy of the relative orthometric heights of the first
order national levelling network is about 3 ppm.
4.3. The Digital Elevation Models
In this section, a brief overview of the current DEM of the region and its accuracy is given. Two
different local DEM models have been constructed for Iran since 1997. The computation of the
first Iranian DEM was conducted jointly by the Institut für Angewandte Geodäsie (IfAG),
Germany, the Institute of Geophysics, Tehran University (IGTU) and the National Cartographic
Centre of Iran (NCC). The height of the DEM, extracted from the 1/250000 base maps (paper
version) of the country has 1 km resolution. It is clear that the accuracy of heights, in the best
case estimated as 1/3 of contour interval, could be near 80 m in best case. The horizontal
accuracy of this model is estimated to 125 m. This DEM is used in the computation of the current
official gravimetric geoid model of Iran by the IfAG group. In 2001, NCC decided to produce a
new precise national DEM based on the 1/25000 photogrammetric maps with 10 m resolution. At
present, 80% of the country is covered by this DEM, but still it is under construction and
evaluation, so it was not available for our research.
Because of presence of outliers and also low accuracy of the IfAG DEM, the old version
of the global GTOPO model was used in the computation of some recent geoid models (e.g.,
Ardalan and Grafarend 2004 and Najafi 2004).
33
Based on the investigations by Kiamehr and Sjöberg, 2005b (PAPER C) investigations,
the absolute accuracy of different DEMs in the study region were evaluated versus GPS/levelling,
and also the procedure for creating a new precise DEM model for Iran was explained. We
summarized that the effect of using high-resolution Shuttle Radar Topography Mission (SRTM)
in gridding of gravity anomalies and terrain correction for the geoid model compared with
previous DEMs is significant, and that it is important to use the new SRTM DEM for
computation of the new gravimetric geoid. We found very large differences between the GLOBE
and SRTM models, in the range of -750 to 550 metres. These differences cause an error in the
range of -160 to 140 mGal in free-air correction and -60 to 60 mGal in the simple Bouguer plate
correction. Using the adaptive formula of KTH approach for terrain correction of geoid, we got a
maximum difference 3 cm between the two DEMs.
4.3.1. Iranian SRTM based DEM
The Shuttle Radar Topography Mission (SRTM) data products result from a collaborative
mission by the National Aeronautics and Space Administration (NASA), the National Imagery
and Mapping Agency (NIMA), the German Space Agency (DLR) and Italian Space Agency
(ASI), to generate a near-global DEM of the Earth using radar interferometry.
The SRTM data flight occurred on February 11-22, 2000, and it successfully fulfilled all
mission objectives. The mission collected 3-D measurements of the Earth's land surface using
radar interferometry, which compares two radar images taken at slightly different locations to
obtain elevation or surface-change information. The collected radar images are converted to
DEMs spanning the globe between 60 D North and 58 D South. The "virtual Earth" will be
reconstructed as a mesh of 30 m spacing, and it is accompanied for each point by a measure of
the reflected energy of the radar signal, the intensity image. The SRTM is a valuable asset for
many applications ranging from geodesy, geology, tectonics, hydrology, cartography, to
navigation and communications.
The memorandum of understanding between NASA and NIMA for SRTM specifies that
data processed at 3′′ (~100 m) for anywhere on the globe will be unrestricted, as will 1′′ (~30 m)
data for the United States and its territories. Detailed documentation with technical specification
of the SRTM data can be found at http://www2.jpl.nasa.gov/srtm/index.html.
34
The first version of SRTM data used in this research is unedited, and it contains numerous
voids (areas without data), water bodies may not appear flat, and coastlines may be ill-defined.
There are many holes or gaps in the old version of the SRTM data, especially in mountainous
areas, also other data quality limitations are too complicated to be described here. To summarize,
there is a vertical error of 16m in 90% confidence interval. This model does not include any
bathymetric data.
There are several software tools for patching the gaps in the SRTM files, but no tools are
available for using shoreline vectors to fix the noise on SRTM water areas. For the Iranian DEM
model, the gaps in land and also marine areas were filled with ComputaMaps 500 m SRTM data.
Heights of all points in the gravity data grid were extracted directly from the original SRTM (3)
model with 100 m resolution. The newly release of SRTM data with significant improvements in
filling mentioned gaps are now available from: http://srtm.csi.cgiar.org/index.asp.
The new Iranian DEM was computed by 15 second resolution (approximately 480 m) by
use of recent 3 second (100 m) SRTM DEM data (Kiamehr and Sjöberg, 2005b). The output grid
size is limited to 23D < ϕ < 42D , 42D < λ < 67D . Small gaps in land areas was patched and filled by
an interpolation procedure, for the large land and marine areas information from the new version
of GTOPO DEM was used. This DEM is intended to be used in interpolation of free-air
anomalies and to compute topographic correction in the new geoid model of Iran. The minimum,
maximum, mean and standard deviation of heights in IRD04 DEM are -84.5, 5033.1, 758.1 and
760.6 m, respectively (Fig.11). PAPAER C presents detail investigations about the procedure of
the evaluation of different DEMs and their impacts on the accuracy of the geoid.
35
Fig. 4. The 15′′ digital terrain model of Iran (IRD04) based on 3′′ SRTM data.
In applying the atmospheric correction term (Eq. 3.11) in this research, a 30′ × 30′ DEM is
generated by averaging the Geophysical Exploration Technology (GETECH) 5′ × 5′ DEM
(GETECH 1995), using area weighting. The harmonics coefficients ( H nm
v
) used here (Eq. 3.13)
were provided by Fan (1998), and they were computed to degree and order 360.
4.4. The Global Geopotential Model
The new satellite gravity missions CHAMP and GRACE lead to significant improvements of our
knowledge of the long wavelength part of the Earth’s gravity field, and thereby of the long-
wavelengths of the geoid. They provide a homogeneous and near-complete global coverage of
gravity field information. Since 1990 different gravimetric local geoid models have been released
for the region of Iran. During the same time several new Global Gravity Models from the recent
satellite gravimetric missions CHAMP and GRACE were released. For the computation of a new
36
gravimetric geoid model for Iran we need a new investigation on the choice of the best GGM
model for the combined solution with local gravimetric data. For the validation of different
global models in the absolute and relative senses, Kiamehr and Sjöberg (2005a) have used 260
GPS/ levelling data as an external tool (See PAPER B). This comparison was utilized with the
new GRACE satellite-only and combined models GGM02S, GGM02C (Tapley et al. 2005) and
EIGEN-02S, combined CHAMP and GRACE model EIGEN-CG01 (for more information, see
http://op.gfz-potsdam.de/index_GRAM.html), GPM98C (Wenzel 1998) and EGM96 (Lemoine et
al. 1998).
Among different satellite-only, combined and tailored GGMs, the study shows that the
combination of the newly released GRACE model (GGM02C) with EGM96 geoid model fits the
GPS/levelling data in Iran with the best absolute and relative accuracies among the GGMs.
However, in practice, because of interaction between the terrestrial data and the GGM in
computing the gravimetric geoid model using the least-squares modification of Stokes’ formula
(for more information, see PAPER B), we find that the GRACE GGM02S model gives the same
results as the GGM02C and EGM96 models. Thus, we use the satellite-only GGM02S model that
is more suitable for the LSMS approach. Here we give a brief presentation of the newly released
GGM02 model.
4.4.1. GRACE GGM02
The GGM02S gravity model (Tapley et al. 2005) was estimated from 363 days (spanning April
2002 through December 2003) of GRACE K-band range-rate, attitude and accelerometer data.
GGM02S is completely pure GRACE-only model because in its construction they do not use any
Kaula constraint, other satellite information, surface gravity information and other a priori
conditioning. The GGM02S field was estimated to degree and order 160, and the solution appears
to retain the correct signal power spectrum up to about degree 120 (see Figure 6). However, it is
recommended that GGM02S should not be used as is beyond approximately degree 110.
(http://www.csr.utexas.edu/grace/gravity/). The GGM02C was constructed by combination of the
GGM02S with terrestrial gravity information (surface gravity and mean sea surface) using the
TEG4 covariance (complete to 200 × 200) to constrain the higher degrees to the harmonic
coefficients of the EGM96. The GGM02C solution was created to degree and order 200, and
37
retains correct signal power spectrum to this resolution, see Figure 4.4. This solution can also be
smoothly extended to 360 × 360 by using the EGM96 coefficients to fill in above degree and
order 200. Paper (B) explains detail information about the procedure of the evaluation of different
GGMs in Iran.
Figure 5. Estimated degree variances and degree error variances for GGM02S/C and EGM96 are shown
in geoid height units. (Tapley et al. 2005).
38
Chapter 5
________________________________________________________
The New Gravimetric Geoid Model (IRG04)
5.1. Basic Parameters and Results
In Chapters 2 and 3 we explained the mathematical procedures for determining the geoid based
on the LSMS method. The basic data including the gravity database, the GGM and DEM, were
evaluated and created. These data were evaluated also through the computation of the
modification parameters in the LSM approach. We tried different DEMs and GGMs with
different degree and order, different integration cap sizes and also different accuracy for the
gravity data to find the best parameters. These comparisons give us very good information about
the properties of different sources of data and their interactions and effects in geoid models.
As we mentioned in Section 4.4, the combination of the newly released GRACE model
(GGM02C) with EGM96 geoid model fits the GPS/levelling data in Iran with the best absolute
and relative accuracy among the GGMs. However, as mentioned before, in practice, GRACE
GGM02S model gives the same results compared combined model of GGM02C and EGM96
models. The full potential of the GGM02S model was used with the maximum degree and order
of 110 in determination of the least-squares modification parameters.
So, the final geoid model was computed based on the free-air gravity anomalies in the
80′′ × 90′′ grid size, GGM02S GRACE-pure GGM and 500 m SRTM global DEM. Because of the
presence of different systematic errors in gravity data and observation of data for special
engineering purposes from different organisations with different accuracy, we found that there
are large local correlations between data. The presence of local properties of data calls for
choosing an optimum cap size for integration. We try different cap sizes (1 to 5 degrees), and
found that results of the computation of geoid with a 3 degree cap works very good versus
39
GPS/levelling derived geoid models (See Kiamehr 2006a and 2006c, PAPER D). On the other
hand, we got the best results with our pre-estimated accuracy for gravity data through the cross-
validation step ( σ Δg =10 mGal) (See PAPERS A, D and E). All additive correction terms were
applied to the approximate geoidal height based on the methods, which explained in Chapter 3.
Figures 6 (a-e) show the 3D view of combined topographic, DWC, total topographic, ellipsoidal
correction and combined atmospheric effects on approximate geoidal heights, respectively.
Figures 7 show the 3D and contour maps of the IRG04 geoid model. Notice that all corrections
depend on the modification, cap size and maximum degree for the GGM being used. As
mentioned before, we can see that the atmospheric and ellipsoidal corrections are small for the
LSMS.
Fig. 6.a. Combined topographic effect for the geoid in Iran. Unit: m
40
5. The New Gravimetric Geoid Model (IRG04)
Fig. 6.b. The DWC effect on the geoid model in Iran. Unit: m
Fig. 6.c. The total topographic effect on the geoid model in Iran. Unit: m
41
Fig. 6.d. Ellipsoidal correction on the geoid model in Iran. Unit: mm
Fig. 6.e. Combined atmospheric correction on the geoid model in Iran. Unit: mm
42
5. The New Gravimetric Geoid Model (IRG04)
Fig. 7. The isolines of the IRG04 geoid model. Unit: m
43
44
Chapter 6
________________________________________________________
Effect of Lateral Density Variation on the Geoid
6.1. Introduction
The existence of topography above the geoid violates the basic assumption of Stokes’ formula for
the determination of the geoid. Usually a constant density 2.67g/ cm3 is used in the determination
of the geoid. This research presents the results of some preliminary experiments conducted using
topographic density data in gravimetric geoid computations in Iran. There are many different
reasons for starting the current research in Iran. Iran is one of the most complicated areas in the
world from the view of rough topography, tectonic activity, density and geoidal height variation.
The minimum, maximum, mean and standard deviation of the topographic heights in Iran are -26,
5678, 1059 and ±734 m, respectively. Also, the geoidal height varies in the range of 50 m. After
the determination of the gravimetric geoid model (IRG04) (Kiamehr 2006a), we attempt to
improve the accuracy of this model, specially in mountainous areas by considering the effect of
lateral density variations.
In this research we utilise the very simple and practical formula for determining the effect
of density variation in geoid models given by Sjöberg (2004a), that can be applied easily as an
additive correction for geoid heights for any gravimetric geoid model. This method includes all
density variation of the topographic and downward continuation effects. In contrast to other
works, here we applied the effect of density variation in a very large area with the large density
variation and also studied the density effect on the geoid based on Pratt-Hayford’s isostatic
model.
45
6.2. Topographic density models from geological maps
The Iranian Geological Survey (GSI) published the Geological Map of Iran in the scale of
1:1000000, displaying bedrock formations at or near the Earth surface (Haghipour et al. 1985).
The bedrock units are grouped according to composition and geological age (see Fig. 3).
Unfortunately, the digital version of this map is still not accessible for public use. The U.S.
Geological Survey also published a similar geological map over Iran in scale 1:2500000 in a
digital version. This map has been digitally compiled and abstracted from the original geological
map of Iran, which greatly facilitates its use, by allowing a direct import into GIS software
(Pollastro et al. 1999). About 541476 geometrical polygons in 59 categories are used to delimit
the bedrock units over Iran. These polygons form the fundamental density units.
For generating the two-dimensional topographical mass density map using the digital
geological maps in a GIS, we used legend information of the GSI geologic map, which each
geological unit is assigned a range of rocks. Then we assigned a mean value of the density range
as a representative density value to each geological unit by using the table of density values
(Carmichael 1989 and http://www-geo.phys.ualberta.ca/~vkrav/Geoph223/Gravity-Density.htm).
The resolution of a simulated density value is taken at a grid size of 80′′ × 90′′ corresponding to the
gravimetric geoid model grid size. The results are displayed in Figure 8, which shows large
contrasts in density variation of rocks with maximum and minimum values of 1.6 and 3.4
gm/ cm3 , respectively.
46
6. Effect of Lateral Density Variation on the Geoid
3
Fig. 8. The lateral density variation model of Iran. Unit: gm/ cm
6.3. The effect on the geoid
Instead of computing separate effects of lateral density variation for direct, DWC and indirect
effects, Sjöberg (2004a) showed in two independent ways that the total geoid effect due to the
lateral density anomaly can be represented as a simple correction proportional to the lateral
density anomaly and the elevation squared of the computation point. If the density of the
topography at the computation point is
ρ = ρ0 + Δρ , (6.1)
47
where ρ 0 is the standard density (2.67 g/cm 3 ) and Δρ = Δρ (θ , λ ) is the lateral density anomaly
with respect to the standard density, the total effect of Δρ on the geoid height becomes
2π GΔρ
δ N Δρ ≈ − H2. (6.2)
γ0
The topographic model used in this study, the Shuttle Radar Topography Mission (SRTM), has a
resolution of 15′′ (Kiamehr & Sjöberg 2005b, PAPER C). Figure 9 shows the effect of using the
density variation model on the geoid. The minimum, maximum, mean and standard deviations of
the lateral variation of density effect on geoid model are -0.143, 0.224, 0.004 and 0.015 m,
respectively.
Fig. 9. Effect of the lateral density variation model on geoid. Unit. m
48
6. Effect of Lateral Density Variation on the Geoid
Also, the lateral isostatic density anomaly effect on the geoid based on the Pratt-Hayford’s model
(Heiskanen & Moritz 1967, p. 138) can be computed by
2π G ρ0 H3
δ N Δρ = , (6.3)
γ0 D+H
where D is the depth of the crust below sea level and ρ 0 is the density for H. The minimum,
maximum, mean and standard deviations of using this isostatic model in Iran are 0, 0.290, 0.008
and 0.012 m, respectively. If we consider D set to 100 km (standard value), then the minimum,
maximum, mean and standard deviation of change are 0, 0.138, 0.004 and 0.005 m, respectively.
Figure 10 presents the effect of using the isostatic model on geoid heights. The result indicates a
significant effect on geoid based on the Pratt-Hayford’s isostatic model. Our results suggest that
the effect of topographical density lateral variations is significant and ought to be taken into
account, specially in mountainous regions, in the determination of a precise geoid model. For
more information and details, see PAPER G.
49
50
Chapter 7
________________________________________________________
Evaluation of the Gravimetric Geoid Model
In order to show the potential of the IRG04 geoid model, we will now compare different
available gravimetric geoid models in the study versus GPS/levelling data. Before this
comparison, let us have a brief and quick overview of the available geoid models. Most of these
models were published recently with different well-know approaches.
7.1. Overview of current Iranian Gravimetric Geoid models
Since 1986 different local gravimetric geoid models have been computed for Iran using various
methods. The computation of the first (and currently only official) Iranian geoid model was
conducted jointly by the Institut für Angewandte Geodäsie (IfAG) Germany, the Institute of
Geophysics of Tehran University (IGTU) and the National Cartographic Centre of Iran (NCC) by
using a regional geopotential model improvement approach (Weber and Zomorrodian 1988). The
method was based on the GPM2 GGM (Wenzel 1985).
In a second stage, the 1 km resolution DEM (Hamesh 1992), which was extracted from
scanned 1/250000 scale maps, including the OSU89B geopotential model (Rapp and Pavlis 1990)
and International Gravimetric Bureau (BGI) gravity database, were utilized for determination of
the geoid. The method of remove-compute-restore (RCR) had been used for the transformation
of gravity anomalies ( Δ g) to geoidal undulations (N). The Fast Fourier Transform (FFT)
approach was used to compute the integral formula (Hamesh 1992). Further results in this model
showed that the fitting of the geoid model with GPS/levelling points improved when eliminating
51
the terrestrial gravity data ( Δ g) from the geoid model, implying that the final IfAG geoid model
was computed only based on the OSU89B and 1 km DTM data.
Another research project was conducted by the NCC and Department of Geodesy of
Tehran University, resulting in another gravimetric geoid model (TUG) computed in a new
manner based on the ellipsoidal Bruns formula without applying Stokes’ formula (Ardalan and
Grafarend 2004). They used the BGI database gravity data together with the recent accurate
gravity database, which was observed by the NCC. For the computation of the terrain correction,
a 1 km resolution the GLOBE global DEM was used. Recently, this group (Safari et al. 2005)
published also another geoid model, which was computed based on idea of the gravimetric,
satellite altimetry, astronomical ellipsoidal boundary value problem.
In yet another effort, a gravimetric geoid model was computed by the Department of
Geodesy of K.N. Toosi University (KNTUG), using the Stokes-Helmert scheme (Vaníček et al.
1995). The result of this model was presented just for the central part of Iran
( 30D ≤ ϕ ≤ 34D and 50D ≤ λ ≤ 54D ). The long-wavelength part of the model was determined by the
CHAMP satellite-only model EIGEN-01S, and the short-wavelength contributions determined by
the total BGI and NCC terrestrial gravity data and National Imagery and Mapping Agency
(NIMA) GLOBE (2003) 1 km global DTM model (Najafi 2004). In Sect. 7.3 we will evaluate
and compare the accuracies of these geoid models in the absolute and relative senses.
7.2. Internal Estimation of accuracy of IRG04
The internal accuracy of the geoidal height estimations can be investigated in form of the global
mean square error of the estimators. The expected Global Mean Square Error (GMSE) of the
unbiased least-squares model is derived as (Sjöberg 1991):
⎡⎛ 2
nmax
L⎞
2
⎤
δ N = c ∑ ⎢⎜ − sn − Qn ⎟ σ n2 + ( QnL + sn* ) dcn* ⎥ .
2 2 * 2
(7.1)
⎢⎝ n − 1
n=2 ⎣ ⎠ ⎦⎥
52
7. Evaluation of the Gravimetric Geoid Model
For the IRG04 geoid model, with nmax =1080, maximum degree of modification and
expansion L=M=110 (for GGM02S) and truncation radius of ψ 0 = 3D , the global root mean square
error is estimated to about 5.2 cm. However, we should add here the MSE is only a theoretical
estimate for the accuracy of the geoid, which is usually too optimistic and does not match exactly
with practical results. We estimated the GMSE based on the band-limited white noise and
reciprocal distance degree variance models (Table 2). The table shows that the total expected
RMS error based on the band-limited white noise model is more than 3 times better than the
reciprocal distance degree variance model. So, we choose white-noise degree variance model to
estimate the LSM parameters in this research.
Nowadays, the most reliable way to estimate the real potential of the gravimetric geoid is
the comparison of its result with the externally derived geoidal height from GPS/levelling. This
will be studied next.
Table 2. Expected global RMS-errors for the least squares modification method with pessimistic apriori error degree
variances. The signal degree variances, M = L = 110, σ = 10 mGal and ψ 0 = 3 .Unit: [mm]. (a). Band-limited
D
white noise degree variance model (b). Reciprocal distance degree variance model.
(a)
Degrees All 2-M (L+1) - ∞ (10800)
Δg part 50.8 24.6 44.4
EGM part (GGM02S) 11 11 -

Bias (truncation error) 3.8 0.0 3.9
Total 52.1 26.9 44.6
(b)
Degrees All 2-M (L+1) - ∞ (10800)

Δg part 144.5 121.3 109.5
EGM part (GGM02S) 72 72 -

Bias (truncation error) 25 0.0 25
Total 163.4 121.3 109.5
53
7.3. External Estimation of the accuracy of geoid model
One of the most interesting and practical applications of a geoid model is levelling with GPS.
Nowadays, for many applications, conventional spirit levelling is being replaced by height
determination using GPS. It has been used for levelling projects, e.g., to monitor local subsidence
due to water or natural gas removal, earth crustal movements and to control heights across water
bodies in connection with a bridge construction.
The heights directly derived from GPS measurements are geodetic heights referred to the
ellipsoid defined by WGS84. Figure 11 shows the basic relationship between geodetic height,
(h), geoid height and orthometric height (H). In the first approximation, the heights are related
by:
h H + N. (7.2)
Fig. 10. Principal of levelling with GPS
The primary problem of transforming GPS-derived geodetic heights into orthometric

heights thus reduces to the determination of reliable and precise geoidal undulations, which
involves a combination of satellite and terrestrial gravity measurements or other data.
54
However, as mentioned in Chapter 4, in practice because of the presence of different

random and systematic errors in the values of h, H and N [e.g., datum inconsistencies and long-
wavelength systematic errors in N, distortions in the orthometric height datum due to an over-
constrained adjustment of the levelling network, effect of various geodynamic effects (e.g.,
specially plate tectonic deformation in Iran), improper or non-existent terrain/ density modelling
in the geoid modelling and orthometric heights and negligence of the sea surface
topography (SST) at the tide gauges], the simple model in Eq. (7.2) does not work properly.
7.3.1 Estimation of accuracy in an absolute sense

In order to reduce effect of the mentioned random and systematic errors in evaluation of the
geoid models, several systematic parameter models can be used in order to fit the quasi-geoid to a
set of GPS levelling points through an integrated least-squares adjustment. Such a comparison is
traditionally based on the following model (Kotsakis and Sideris 1999):
ΔN = N iGPS − N i = hi − H i − N i = aiT x + ε i , (7.3)
where x is a n ×1 vector of unknown parameters, ai is a n × 1 vector of known coefficients, and
ε i denotes a residual random noise term. The parametric model aiT x is supposed to describe the
mentioned systematic errors and datum inconsistencies inherent in the different height data sets.
Its type varies in form and complexity depending on a number of factors. The parametric models
tested in this research are:
4-parameter model:
ai = ( cos ϕi cos λi cos ϕi sin λi sin ϕi 1) and x = ( x1 x4 ) .

T T
x2 x3 (7.4)
5-parameter model:
ai = ( cos ϕi cos λi sin ϕi 1 sin 2 ϕi ) and x = ( x1 x5 ) . (7.5)

T
cos ϕi sin λi
T
x2 x3 x4
55
7-parameter model:
⎛ cos ϕi cos λi ⎞ ⎛ x1 ⎞
⎜ ⎟ ⎜ ⎟
⎜ cos ϕi sin λi ⎟ ⎜ x2 ⎟
⎜ sin ϕi ⎟ ⎜ x3 ⎟
⎜ ⎟ ⎜ ⎟
ai = ⎜ cos ϕi sin ϕi cos λi / Wi ⎟ and x = ⎜ x4 ⎟ . (7.6)
⎜ cos ϕi sin ϕi sin λi / Wi ⎟ ⎜ x5 ⎟
⎜ ⎟ ⎜ ⎟
⎜ sin 2 ϕi / Wi ⎟ ⎜ x6 ⎟
⎜ 1 ⎟ ⎜x ⎟
⎝ ⎠ ⎝ 7⎠
where ϕ and λ are the horizontal geodetic coordinates of the network or baseline points, and
Wi = (1 − e2 sin 2 ϕi ) ,
1/ 2
(7.7)
where e is the first eccentricity of the reference ellipsoid. Then we obtain the following matrix
system of observation equations
Ax = ΔN − ε , (7.8)
where A is the design matrix composed of one row aiT for each observation ΔN i . The least
squares adjustment to this equation utilizing the mean of squares of the residuals ε i , becomes
xˆ = ( AT A ) AT ΔN .
−1
(7.9)
yielding the residuals
εˆ = ΔN − Axˆ = ⎡⎢ I − A ( AT A ) AT ⎤⎥ ΔN .
−1
(7.10)
⎣ ⎦
56
The standard deviation of the adjusted values for the residuals εˆ are traditionally taken as
the external measure of the absolute accuracy of the geoid model. It is important to mention here
that the final residual values are not the exact errors of the gravimetric or GGM geoid models,
because they include also some part of errors from GPS and levelling observations. According to
our numerical results for these three models, the seven-parameter model gives the best fit with
minimum standard deviation in all selected GGMs and gravimetric geoid models. For the
evaluation of the geoid models in the absolute sense, we used all the available 260 and 35 precise
GPS/levelling data separately. Table 3 (a and b) shows the results of the estimation of the
absolute accuracy of IRG04 and current local gravimetric geoid models in Iran versus 260
GPS/levelling points using Eq. (7.10).
The cross-validation approach used both for detection of outliers in GPS/levelling data
and evaluation of the IRG04 geoid model. Based on this investigation, from 260 GPS/levelling
data, 13 points were detected as outliers. The IRG04 geoid model was evaluated again, with 247
points using the cross-validation method. Table 4 shows the results of this evaluation. By using
the cross-validation approach, the overall estimation for absolute accuracy of the IRG04 model
improved about 12 cm.
Table 3. Statistical analysis fitting the 260 GPS/levelling data and gravimetric geoid models from the absolute
accuracy view before and after 7-parameter fitting. (The RMS value for four and five-parameter models is given in
order to compare results of the fitting between different models). Unit: m
(a)
N GPS − Lev - N KNTUG NGPS − Lev - N IfAG

Gravimetric
(Helmert Scheme) (R-C-R Method)
Geoid Models
Najafi 2004 Hamesh et al. 1992
22 points 260 Points
Before After Before After
Min. -11.244 -1.631 -2.46 -1.934
Max. -4.525 1.809 2.792 3.259
Mean -9.559 0.000 -0.559 0.000
RMS 1.324 7P: 0.844 0.801 7P: 0.763
________ ________
5P: 0.864 5P: 0.796
4P: 0.911 4P: 0.775
57
(b)
N GPS − Lev - NTUG NGPS − Lev - N IRG 04 N GPS − Lev - N IRG 04

Gravimetric
(Ellipsoidal Bruns (LSMS Method) (Ellipsoidal boundary
Geoid Models
Formula) Kiamehr 2006a value problem)
Ardalan et al. 2004 Safari et al. 2005
258 points 260 points 51 points
Before After Before After Before After
Min. -5.14 -4.261 -2.257 -1.678 -2.148 N/A
Max. 3.458 3.496 2.063 2.570 2.066 N/A
Mean -0.517 0.000 -0.74 0.000 0.161 N/A
RMS 1.262 7P: 1.074 0.577 7P: 0.551 1.068 N/A
_________ _________
5P: 1.089 5P: 0.568
4P: 1.123 4P: 0.571
Table 4. Result of evaluation of the IRG04 geoid model based on the cross-validation approach, before and after
seven- parameter fitting. Unit: m
IRG04 Model (N = 247) Min Max Mean RMS

Before 7-parameter fitting -1.891 0.489 -0.659 0.446
After 7-parameter fitting -1.219 1.223 0.000 0.427
However, validation of the IRG04 geoid model based on the precise 35 GPS/levelling
points, gives much better accuracy (almost 0.29 m) comparing all available GPS/levelling data
(see, Table 5). As mentioned before, 225 GPS/levelling points belongs to the second-order GPS
and levelling networks which have a lower accuracy comparing the 35 precise points.
Table 5. Validation of the IRG04 model versus 35 precise GPS/levelling data. Unit: m.
Model Min Max Mean RMS
IRG04 gravimetric geoid model -1.284 0.223 -0.652 0.362
IRG04 after 7-parameter fitting approach -0.518 0.924 0.000 0.288
Figure 11 shows the histogram of difference between the 260 GPS/levelling points and
IRG04 geoid model. Figure 12 shows the discrepancy between GPS/levelling data and
gravimetric geoid models in the study area.
58
(a) (b)
Fig. 11. Histogram of difference between the 260 GPS/levelling data and IRG04 geoid model before (a) and after 7
parameter fitting (b). Unit: m.
(a). TUG (Ellipsoidal Bruns formula) (b). IfAG (R-C-R Method)
Fig. 12. Discrepancy between the GPS/levelling and three

gravimetric geoid models. (Contour interval 0.5 m). Unit: m
(c). IRG04 (LSMS formula)
59
7.3.2. Effect of lateral density variation model
In order to study the effect of using the density variation models on the IRG04 geoid model, 260
GPS/levelling points were used. The minimum, maximum, mean and standard deviations of these
levelled heights are -21.45, 2551.55, 1083.68 and 651.99 m, respectively. We added either the
effect of actual or isostatic models on the gravimetric geoid model as correction term (Fig. 9).
Table 6 shows a summary of a statistical analysis for the evaluation of the two corrected geoid
models versus GPS/levelling data.
Table 6. Effect of using the different density variation models (real and isostatic) on the IRG04 gravimetric geoid
model. (Comparison is done before applying the 7 parameter fitting approach).
Geoid models With actual With Isostatic With constant

density variation density variation density value
Minimum -2.257 -2.257 -2.257
Maximum 2.059 2.073 2.064
Mean -0.751 -0.674 -0.741
RMS 0.576 0.589 0.578
Although there is a slight indication that the correction for lateral density variation works
in the right direction, from Table 1 we cannot see a significant improvement by using the actual
density variation model, because most of the GPS/levelling data are located in the moderate
topographic areas with a maximum height of 2551 m. Also, we think that the evaluation of the
effect of density models with GPS/levelling data seems not very reasonable, because basically
our levelling Helmert orthometric height is not a true orthometric height. This difference is due to
in the error in the estimation of the mean value of gravity along the plumb line between the
surface and the geoid.
The Helmert height is based on a model of an infinite Bouguer plate with a uniform
density of 2.67 g/ cm3 . Variations in density and topographic relief will cause departures of
Helmert heights from true orthometric heights. As a gauge on the influence of rock density
variation, Heiskanen and Moritz (1967, pp.169) show a 4 mm error in Helmert height for a point
at 1000 m elevation and with a constant 0.1 g/ cm3 surface density departure from 2.67 g/ cm3 .
Such error is proportional, so that if one assumes an average density of 2.87 g/ cm3 (e.g.,
60
diorite/gabbro combination or an alkaline basalt as found in the Rocky Mountains) distributed as

a Bouguer plate with an elevation of 3000 m, then one would obtain a Helmert height error of
0.024 m. For any realistic evaluation of the effect of actual density on the gravimetric geoid
model, we need some GPS/levelling points in rough areas as well as correct orthometric heights
for them.
7.3.3. Estimation of accuracy in a relative sense
In order to give more precise and acceptable results, the geoid models were evaluated also in the
relative sense by using the 35 precise GPS/levelling data. For this purpose, we computed the
orthometric height differences using the relative geoidal undulations Δ N for selected baselines
by using Eq. (7.2) as follows:
ΔH = Δh − ΔN , (7.11)
GPS-GGM GPS GGM
where Δ symbol means difference. We can easily derive differences between two different
orthometric height differences, i.e. from levelling ( ΔH Level ) and from GPS minus GGM
( ΔH ) using Eq. (7.11):

GPS-GGM
δΔHGGM-level = ΔH GPS −GGM − ΔH Level . (7.12)
The relative differences between a GGM/gravimetric geoid model and levelling becomes
in part per million (ppm):
(δΔH GGM − Level )mm

ppm = mean , (7.13)
Dij( km )
61
where Dij is the length of the baseline. The average distance between the 35 GPS/levelling
points is 80.65 km. Table 7 shows the results of comparison between the IfAG, TUG and IRG04
geoid models from the view of relative accuracies on five selected traverses (see Fig. 3). As the
KNTUG model is available only in a limited area, we can just find two precise GPS/levelling
points there, so it is not possible to make any comparison for this model.
Table 7. Statistical analysis fitting the 35 precise GPS/levelling data and gravimetric geoid models from the relative
view of accuracy (by Eq. 7.13). Unit: m
TUG IfAG IRG04

Traverse
(Ellipsoidal Bruns (R-C-R method) (LSMS formula)
formula)
RMS RMS RMS

1 (West) 1.942 1.17 0.48
2 (North) 1.183 1.20 0.58
3 (Centre) No DATA 1.33 0.48
4 (East) 1.93 0.49 0.18
5 (South) 0.407 2.00 0.14
Min. -2.686 -3.419 -0.82
Max. 2.551 2.750 0.79
Mean 0.009 0.132 0.02
RMS 1.239 1.310 0.40
ALL Baselines 15.4 16 3.8
(ppm)
With simple comparison between results of Tables 3 and 7, we found that the comparison
of geoid models from the relative accuracy sense gives very interesting and realistic results about
the real potential of gravimetric geoid models.
In order to show a better view of the improvement in accuracy of IRG04 model, we also
make a comparison with the GRACE satellite-only (GGM02S) and the combined (GGM02C)
models in the absolute and relative senses (Tables 8 and 9). The statistical analyse of absolute and
62
relative accuracy for recent GRACE models before and after 7-parameter fitting procedure are
shown in Tables 8 and 9.
Table 8. Statistical analysis fitting the GPS/levelling data and GRACE-GGM02 models from the absolute accuracy
view, before and after applying the 7-parameter fitting procedure. Unit: m
GGM02S GGM02C
GGM GRACE GRACE
degree and order degree and order
160 200
Before After Before After
Min. -4.076 -3.671 -2.96 -2.558
Max. 2.714 3.013 1.87 2.018
Mean -0.239 0.000 -0.243 0.000
RMS 1.230 1.218 0.854 0.837
Table 9. Statistical analysis of fitting between the 35 precise GPS/levelling data and GRACE GGM02 models from
the view of relative accuracy (Eq. (8)) for GRACE GGM02 models. Unit: m
GGM-02S GGM02C
Traverse degree and order degree and order
160 200
RMS RMS
1 0.451 0.500
2 2.461 1.530
3 1.860 1.477
4 0.778 0.733
5 1.348 1.223
Min. -3.221 -2.598
Max. 3.569 1.943
Mean -0.01 0.03
RMS 1.540 1.11
ALL (ppm) 19 13.8
63
3
2.5
West
2
North
1.5 Center
East
1
South
0.5
0
GGM02S GGM02C IRG04 IfAG TUG
Fig. 13. Comparison of relative accuracies of fitting GPS/levelling data and GGM02 models versus local gravimetric
geoid models in five selected traverses. Unit: m
It is evident that most of the progress in IRG04 is because of using the most recent GGM
and DEM. However, except using some extra satellite altimetry data and EGM96 (original data)
in marine and gap areas inside and outside of boarders, the quantity of the used ground data in
IRG04 is almost the same as in KNTUG and TUG. We think that there are three different reasons
that come into play:
a) Our method for eliminating outliers from the gravity database.
b) The interpolated denser gravity observations using the high-resolution SRTM DEM before Stokes'
integration. This operation improved the RMS of the fit between the geoid model and GPS/levelling
data up to 0.37 m (Kiamehr and Sjöberg 2005)
c) From the theoretical point of view, the LSM kernel matches the errors of the terrestrial gravity
data, the GGM and the truncation errors in an optimum way.
Figure 14 shows the discrepancies between IRG04 and TUG. The largest differences are
mostly located in the rough topographic areas in the Albourz and Zagrous mountains (north and
west). We think that a large part of the differences can be attributed to the high-resolution SRTM
DEM used in IRG04, which is considerably more accurate than the GLOBE DEM used in TUG.
The IRG04 model fits better with GPS/levelling data in these areas, which supports this
conclusion. Another factor, that might explain the differences between IRG04 and TUG, is that
64
Fig. 14. Discrepancies between the IRG04 and TUG geoid models. Contour maximum and minimum are +4.5 m
(brightest region) and -4 m (darkest region), respectively, and contour interval is 0.5 m.
the most recently released GRACE-only GGM is used in the former. It is also interesting to note
that the GGM02C has better accuracy compared to both the TUG and RCR models (see Tables 7
and 9). In Papers D and E we present a general comparison between several geoid models using
different interpolation methods, cap sizes and standard deviations of σ Δg and GGMs.
65
66
Chapter 8
________________________________________________________
A New Height Datum for Iran
The statistical behaviour and modelling of the misclosure of Eq. (7.2), computed in a network of
GPS/levelling benchmarks, have been the subjects of many recent studies (e.g. Kotsakis et al.
2001a and 2001b, Kotsakis and Sideris 1999 and Fotopoulos 2003). Using the corrective surface
idea is one of the most interesting and practical subjects in this area. The development of
corrective surfaces can be used for the optimal height transformation between ellipsoid and
levelling datum surfaces for reducing the long-wavelength gravimetric geoid errors. Some studies
from recent years emphasising this fact can be mentioned, like Kotsakis and Sideris (1999) and
Lee and Mezera (2000).
8.1. Creation of the corrective surface
In order to define a new height datum for Iran, we attempt to combine the IRG04 high resolution
gravimetric geoid model with GPS/levelling data by using a corrective surface idea. The
corrective surface is constructed based on 224 GPS/levelling points and then evaluated with the
35 independent points. By applying the 7-parameter fitting approach (Eq. 7.6), the adjusted
values for the residuals ( εˆ ) are estimated based on the least-square model (Eq. 7.10). Then, the
continuous surface is generated from the discrete GPS/levelling data by using the prediction
techniques. Different interpolation techniques were tested for the creation of the corrective
surface; among them the Kriging method gives the minimum RMS and minimum spread of
residuals versus the GPS/levelling data. The use of a corrective surface to the IRG04 gravimetric
67
geoid model has been shown to significantly improve the determination of heights from GPS in
Iran. The RMS of fitting the new combined model IRG04C versus GPS/levelling data is 0.09 m,
which gives a more than 4 times better fit compared with the IRG04 model (see Table 10).
Hence, this corrective surface should be convenient and useful in the definition of a new height
datum, specifically in engineering and GPS/levelling projects. (For more information and details
about the definition of the new height datum, see Kiamehr 2006c, PAPER E)
Table 10. Validation of the IRG04 and the new combined geoid model versus 35 GPS/levelling data points. Unit: m.
Model Min Max Mean RMS
IRG04 model -1.284 0.223 -0.652 0.362
IRG04 after 7-parameter fitting -0.518 0.924 0.000 0.288
The combined model (IRG04C) -0.474 0.433 0.000 0.088
68
Chapter 9
________________________________________________________
Geodynamical Researches
As mentioned in the introduction of the thesis (Chapter 1), during this research we also make
some investigations about geodynamical applications of geodesy. There are two published papers
in this area in Part 2 of the thesis. In the fist paper (Kiamehr and Sjöberg 2006, PAPER G) we
studied the impact of different geospatial information (specially a high resolution geoid model) in
studying of tectonic activity of Iran. The second Paper (Kiamehr and Sjöberg 2005c, PAPER H)
explains the analysis of some surface deformation patterns based on the 3D finite element
method. In this chapter we have a brief review of the most important parts of these two articles.
9.1. Impact of the geoid model in studying tectonic activities in Iran
Geoid signals can give important information about the underlying density structure of the earth,
and it can be used to locate the source depth of mass anomalies. The importance of the geoid in
geophysics and geodynamics has been recognised by the studies of the correlation between the
geoid and deep-Earth mass density anomalies (Bowin 1983); the constraints provided by the
geoid on mantle rheology and flow (Hager, 1984); correlation between the geoid and westward
drift of the geomagnetic field (Khan, 1971), geoid bulge and its correlation with land uplift and
Moho depth (Sjöberg 1983 and Sjöberg et al. 1991) and correlation between the geoid and plate
tectonic features and seismic tomography (Silver et al., 1988)
The Iran region is known as an extraordinary natural laboratory for the study of seismo-
tectonic processes. This area is geologically and geophysically as well as geodetically one of the
most studied regions on Earth. It is one of regions that are located in the line of the most recent
69
foldings along the length of the Alp strip in Europe to the Himalayas in the north of India. Here
Iran is in the place of the junction of the Saudi-Arabia, India and Eurasia plates. This junction has
led to the seismicity of Iran, and, as a result, most regions and cities of Iran are exposed to
frequent earthquakes.
An integrated approach, combining gravity, geoid, and topography and seismology data,
allow us to have a detail study of the crustal and lithospheric structure of the region. In the
current study we used the most recent information, which come from the high resolution Shuttle
Radar Topography Mission (SRTM), free air gravity grid database (Kiamehr, 2005a) and the new
precise combined geoid model (IRG04C), the complete earthquakes database of Iran, which was
gathered by the International Institute of Earthquake Engineering and Seismology (IIEES). The
seismological database has 5900 records, and it includes the general information of the
earthquakes since 1902. (See Kiamehr and Sjöberg 2006d, PAPER G - Table 2).
The statistical analysis of the extracted heights of the earthquake locations from the
SRTM model shows that the mean height of the seismological database points is 1091 m, but the
classification of earthquakes with the magnitude over 6 says that their mean height is 1275 m. We
can conclude that the flat and moderate topographic areas (like the central of Iran) have a lower
risk of earthquakes compared to the mountainous areas in the north and south-west.
The gravity anomaly is a differential quantity, exhibitively small scale structure, whereas
the geoid is an integrated quantity, being optimal for studying long-wavelength effects and major
tectonic features. A detail study of the thematic map of the free-air anomaly and the statistical
analysis of the extracted values from the grid for the seismological data (see PAPER H, Fig. 2
and Table 2) clearly shows very local and complicated structures (like topography), which are
mostly related with local mass variations and shallow density contrasts below sea level, and they
could therefore be useful in studying local tectonic patterns. The large positive anomalies in the
Alborz and Zagros in Figure 2 are interpreted to be a result of higher-density mantle lithosphere.
Combining these two observations, we conclude that the higher-density material in the uppermost
mantle is the likely candidate for the cause of this anomaly.
Similar to gravity, the geoid reflects mass variations in the Earth’s interior of various
wavelengths, but, in contrast to gravity, the geoid primarily reflects the long and medium
wavelengths. Hence, it is more useful in studying global and regional mass anomalies in the
70
9. Geodynamical Researches
Earth, such as deep lithosphere structures. We found interesting results from the gradient map of
the geoid (geoid deflection map). It clearly shows the presence of significant correlation between
the lateral geoid gradient (NS) and the distribution of earthquakes in Iran (see Fig. 15). Our
statistical analysis shows that almost 72% of the earthquakes in Iran occurred in areas with geoid
slopes over 5%. The minimum, maximum, mean and standard deviations of earthquake
magnitudes are 1.7, 7.7, 4.15 and 0.84 Richter, respectively. It is also interesting that all of the
78 earthquakes with magnitude over 6 Richter are in areas with geoid slope 7.5% or more. The
statistical analysis of correlation values between different parameters gives us very interesting
results which can be concluded as follows (For more information see Kiamehr and Sjöberg 2006d
and PAPER G):
• All correlations are significant except those between NS (geoid slope) and N (geoid
height) as well as between NS and R (earthquake magnitude).
• There is a strong correlation between topographic height and each of free-air anomaly and
simple Bouguer anomaly.
• There is a positive and significant correlation between the geoid slope and earthquakes
depth parameters in the order of +0.46 in the Zagros area.
• There is a significant and remarkable correlation between the depth and magnitude of
earthquakes in the study area (r = 0.4).
• The geoid slope parameter (NS) and free-air anomaly have the highest correlation value
among different parameters with the depth of the earthquake (D). It could be an important
sign from the earthquake study view. We think using the simple and clear map of the
geoidal slope and the complicated map of the free-air anomaly are complementary for the
local and regional interpretations of earthquakes and plate tectonics.
71
Fig. 15. Distribution of earthquakes (black dots) on the lateral geoid slope map. (Slope in percentage)
9.2. Analysis of surface deformation patterns using 3D finite element method
Typical geodetic observables are discrete functions in space and time, and, consequently, the
height and displacement vectors, deduced from the geodetic data are of discrete nature. The
finite element method (FEM) (e.g., Burnett 1987) has found a manifold of applications in
geodesy and particularly in Earth deformation studies, because a geodetic network can be
viewed as a typical example for a set of irregularly shaped finite elements in two or three
dimensions. The domain of the problem is divided into smaller regions or sub domains, called
elements. Adjacent elements touch each other without overlapping, and there are no gaps
between the elements. The shapes of the elements are intentionally made as simple as possible,
72
such as triangles and quadrilaterals in 2-dimensional domains, and tetrahedral and pentahedra
in 3-dimensions. It yields what is known as a Delaunay triangulation, one in which the
triangles formed are as near equilateral as possible for the given positions of irregularly spaced
nodes. The analysis of geodetic observations is typically performed with models involving coordinates,
which refer to more-or-less arbitrary reference frames. In this way the determination of the
displacement vector can not show the real physical behaviour of the body, because it is completely
dependent to the chosen reference frame.
Since deformation refers to the change of a shape, independent of position (Grafarend and
Schaffrin 1976), it is only reasonable to study it by properly defined parameters, which are
independent of frame definitions. Here, we first have a quick view to the principal 3D FEM.
Then, we will use FEM for determination of deformation parameters in the area of Skåne in the
south of the Sweden. The detail investigation and information about the principal and the
theoretical aspects of this method can be found in Kiamehr and Sjöberg 2005c, PAPER I.
Here we summarize the practical procedure for the determination of the deformation
parameters in a 3D view (Dermanis and Grafarend 1983). For modelling the three dimensions
FEM we use 4 vertices pentahedra element, say Pα , Pβ , Pγ , Pδ . The linear field of coordinate
changes from X to X ′ is given by the transformation
X ′ = JX + G = ⎡⎣ xα′ x ′β xγ′ xδ′ ⎤⎦ = J 3×3 ⎡⎣ xα x β xγ xδ ⎤⎦ + ⎡g g g g⎤ , (9.1)

3×4 3×4 ⎣ ⎦ 3×4
where J and g are the Jacobian matrix and the translation vector, respectively. Based on this
model, the three-dimensional Jacobian matrix J can be computed directly from the coordinates
using the solution given by
−1
⎛ xβ′1 − xα′1 xγ′1 − xα′1 xδ′1 − xα′1 ⎞ ⎛ x1β − xα1 xγ1 − xα1 xδ1 − xα1 ⎞
⎜ ⎟⎜ ⎟
J = ⎜ xβ′2 − xα′2 xγ′2 − xα′2 xδ′2 − xα′2 ⎟ ⎜ xβ2 − xα2 xγ2 − xα2 xδ2 − xα2 ⎟ . (9.2)
⎜ 3 xδ′3 − xα′3 ⎠⎟ ⎝⎜ xβ3 − xα3 xδ3 − xα3 ⎠⎟
⎝ xβ′ − xα′ xγ′3 − xα′3 xγ3 − xα3
3
The strain matrix S can easily be computed by
73
1 t
S=
2
(J J − I ) . (9.3)
If vi are the unit eigenvectors of S corresponding to the eigenvalues ei , the strain matrix can be
diagonalised as:
S = V ΛV t ⇔ V t SV = Λ = Diag ( e1 , e2 , e3 ) , (9.4)
where V is the orthogonal matrix with columns being the eigenvectors vi , and Λ is the diagonal
matrix with diagonal elements being the ordered eigenvalues ei of S. Also, the planar dilatation ( Δ )
and maximum shear strain ( γ ) components can be defined for each of the three principal planes
perpendicular to the principal directions v1 , v2 , v3 , respectively:
Δ12 = e1 + e2 , γ 12 = e1 − e2 ,
Δ 23 = e2 + e3 , γ 23 = e2 − e3 ,
Δ13 = e1 + e3 , γ 13 = e1 − e3 . (9.5)
The total dilatation is given by:
1
Δ= ( Δ12 + Δ 23 + Δ13 ) , (9.6)
3
and the direction of each principal axis is defined by the corresponding vector vi = ( vi1 vi2 )
vi3 , or
by two appropriate direction angles, e.g. "longitude" and "latitude":
⎛ ⎞
⎛ vi2 ⎞ ⎜ vi1 ⎟
Λ i = arctan ⎜ 1 ⎟ and , Φ i = arctan ⎜ ⎟. (9.7)
⎝ vi ⎠ ⎜
⎝ (v ) + (v )
1 2
i
2 2
i
⎟
⎠
9.2.1. Skåne GPS network
Since 1993 a large and complete GPS network with more than 40 stations with continuous daily
observations established in the Swedish part of the Baltic shield for studying postglacial rebound
74
and tectonic activities under the BIFROST project (Scherneck et al. 1998 2000 and 2001). This
network is useful specially for studying and analysis of the general activities of major faults and
geodynamics phenomena in the large scale. Nowadays, using local geodetic networks with
Electronic Distance Measurement (EDM) or GPS observations is so common for studying local
geodynamic activities (e.g. landslide monitoring) and analysis of deformation for large man made
structures (e.g. dam, bridges and nuclear buildings). Instead of using national geodynamic
networks (e.g. BIFROST) which usually have continuous observations, for local projects, based
on the field conditions and limitation and specially because of economic reasons, we frequently
prefer to have repeated observations in separate campaigns.
Pan et al. (2001) processed the GPS data of the Skåne campaigns in 1996 and 1998. The
Skåne network is a local network, which was basically established for monitoring possible plate
tectonic activity in this area. The major guidelines of how they dealt with the data are as follows.
The data were processed session by session, and normal equations for all the sessions were
stored. The results were then combined to obtain final solutions for each campaign of 1992, 1996
and 1999. In processing the GPS carrier phase data, the coordinates of one station usually have to
be fixed (as known) in the WGS84 or the International Terrestrial Reference Frame (ITRF) coor-
dinate system. The fiducial station Stavershult was selected as the fixed station in the ITRF96
system. (For more detail see, PAPER I)
The results of the research shows that the areas with maximum shear strain and dilation
(regions G, B and A) are the most active areas and are located exactly in the active fault zones
and their intersections (see Figure 5, PAPER I). However, further observations in a denser
network as well as integration with geological and geophysical data are needed to fully explore
the recent crustal motions over the Tornquist zone.
75
Chapter 10
________________________________________________________
Conclusions and Recommendations
The computation of a regional gravimetric geoid model with proper accuracy in a developing
country with sparse and limited data is a difficult task, which needs a special notice to produce
good results. In this research we try to investigate the procedure for gathering, evaluating and
combining different data for the determination of a gravimetric geoid model for Iran. The least-
squares modification of Stokes’ method by the KTH approach was used for combining different
heterogeneous data in an optimum way.
The first and most important step for geoid determination is to evaluate and choose the
best available data in the region. The basic data which we used in this research are gravity
anomaly data (from different sources), a DEM, a GGM and GPS/levelling data. The available
data are mostly unofficial and sparse (e.g., gravity, GPS and levelling data) or mostly based on
global data sources (e.g., DEM and GGM).
During this research, a new gravity anomaly database (with contributions partly from
ship-borne and satellite altimetry data) has been created by gathering all available data in the
Iranian region from different local and international organizations, including data from the NCC
and BGI. Different procedures were used for the refinement of the data (outlier detection) and
also for taking into account the effect of topography in the gridding scheme.
Due to the lack of any high resolution photogrammetric based DEM in the region, a new
DEM model (IRD04) was constructed with the resolution 15′′ based on the 90 m high resolution
SRTM DEM. A detail investigation was performed for studying the impact of using the high
76
10. Conclusions and Recommendations
resolution SRTM data versus 1 km resolution models in the geoid modelling. The IRD04 model
fitted the heights of the 260 GPS/levelling data at the level of 6 m.
Also, we make a detailed investigation of all available GGMs for choosing the best fitting
model in the study area. From the results of this investigation we estimated that the GRACE
model GGM02C extended with EGM96 fits the GPS/levelling data in Iran with 0.69 m and 7
ppm in the absolute and relative senses, respectively. This is the best model among all GGMs. It
is also better compared to some local gravimetric geoid models. However, because of the
interaction (correlation) between data in the least-squares scheme, in practice we got the same
results by using the GRACE-GGM02S (satellite-only) model, which is theoretically preferred in
the LSMS model.
The IRG04 gravimetric geoid model was evaluated in the absolute and relative senses,
both internally by the estimation of the global MSE and externally by GPS/levelling data. The
standard deviation of fitting between the IRG04 and the 35 precise GPS/levelling data is
estimated to 0.29 m. However, because of the presence of different systematic errors in the
observation and adjustment of the GPS/levelling and gravity data, the estimation of the accuracy
in the absolute view usually can not show the real potential of a geoid model. However, it is well-
known that GPS and levelling observations have very good accuracies in the relative sense,
because most of the systematic errors are cancelled or eliminated through the differencing of
observations. This lead us to evaluating the relative accuracy of the geoid models based on
ΔH (GPS/geoid) versus levelling data (see Sect. 7.3.3). The results of this investigation shows
that the IRG04 model fits the GPS/levelling with 3.8 ppm, which means that it is the best
available geoid model in the study region.
Also, the influence on the geoid height stemming from a laterally density variation model
was studied. We used a geological map to construct a rock density map of Iran. The numerical
results show that the effect of using a lateral density model reaches up to 0.22 m, which is not
negligible in a precise geoid determination with expected centimetre accuracy. Our results
suggest that the effect of lateral topographical density variations is significant and ought to be
taken into account specially in mountainous regions.
According to the various systematic errors that have been encountered in the terrestrial
gravity and GPS/levelling data, we tried to find an optimal combination by combining the
77
gravimetric and geometric geoid models based on the corrective surface idea. The RMS
difference between of fitting the independent GPS/levelling data and the IRG04C combined
model is 9 cm, which is almost 4 times better than the IRG04 gravimetric model. However, it
should be mentioned here that the combined surface model is not an equipotential surface but it
can be used in eliminating systematic errors and datum inconsistencies between the GPS and
levelling height systems.
Based on this progress, the IRG04 geoid model combined with GPS observations can
replace spirit levelling in most of engineering projects. This holds definitely in the relative form
and preferably for short baselines.
Our main recommendations for further work are as follows:
1- According to the results of this research we cannot recommend further use of the current
IfAG official gravimetric geoid model. We strongly recommend using the IRG04 geoid model in
future projects.
2- One of the most important factors, that limit the accuracy of the gravimetric geoid in Iran, is
the poor quality and quantity of the gravity data. It is evident that with the available distribution
of gravity data (1 data per 65 km 2 ) it is almost impossible to obtain cm accuracy for the geoid.
There are large mountains areas in Alborz and Zagros that do not have any gravity data at all.
Most of these areas will not be surveyed for terrestrial gravity data in the near in future because
of funding and logistics reasons. In mountainous and large countries like Iran it is strongly
recommended to use airborne gravimetry to increase the quality and quantity of data.
3- For any future test it is recommended to use much denser high quality GPS/levelling points.
Well distributed GPS/levelling data (especially in mountainous areas) is important in evaluating
and refining the gravimetric geoid model.
6- Variance component estimation can be applied to the common adjustment of the
heterogeneous heights (N, h and H). This leads to an in-depth analysis of the effects of correlation
among heights of the same type and the intrinsic connection between the proper modelling of
systematic errors and datum inconsistencies with the estimated variance components. It can be
used for the optimum combination of these three parameters through the combined adjustment
model. Unfortunately due to lack of information of covariance matrixes for h and H components
we could not apply this approach in this research. The numerical studies including the scaling the
78
10. Conclusions and Recommendations
GPS/derived ellipsoidal height covariance matrix and evaluation of the accuracy of orthometric
heights obtained from the national adjustments of levelling data is recommended for future
works.
79
________________________________________________________
Bibliography
Abbolgasem A (1994) Iranian sea surface topography, M.Sc. Thesis, K.N. Toosi University, Tehran.
Ardalan R and Grafarend E (2004) High Resolution geoid computation without applying Stokes’ formula
case study: High resolution geoid of Iran, J. Geodesy, 78, 138 - 156.
Beutler G, Brockmann E, Fankhauser S, Gurtner W, Johnson J, Mervart L, Rothacher M,Schaer S,
Springer T, Weber R (1996) Bernese GPS Software Version 4.0 Documentation, Astronomical Institute
University of Berne, September 1996 Burnett DS (1987) Finite element analysis, Addison-Wesley
publishing Co., U.S.A
Carmichael S (1989) Practical Handbook of Physical Properties of Rock and Minerals, CRC Press, Boca
Raton, FL United-States.
de Witte L (1967) Truncation errors in the Stokes and Vening Meinesz formulae for different order
spherical harmonic gravity terms. Geophys J R Astron Soc 12: 449-464
Ellmann A (2001) Least squares modification of Stokes' formula with applications to the Estonian
geoid. Royal Institute of Technology, Division of Geodesy Report No. 1056 (Licentiate Thesis),
Stockholm
Ellmann A (2004) The geoid for the Baltic countries determined by the least squares modification of
Stokes' formula. Doctoral Dissertation in Geodesy No. 1061, Royal Institute of Technology
(KTH), Stockholm
Ellmann A and Sjöberg LE (2004) Ellipsoidal correction for the modified Stokes formula. Boll.
Geod. Sci. Aff., 63 (No. 3).
Fan H (1998) On an Earth ellipsoid best-fitted to the Earth surface. Journal of Geodesy, 72(9): 511-515
Featherstone WE, Evans JD and Olliver JG (1998) A Meissl-modified Vanicek and Kleusberg kernel to
reduce the truncation error in gravimetric geoid computations, Journal of Geodesy, 72(3): 154-160.
Fei ZL and Sideris MG (2000) A new method for computing the ellipsoidal correction for Stokes' formula.
J. Geod., vol. 74, pp. 223-231 and 671.
Forsberg R (1984) A Study of Terrain Reductions, Density Anomalies and Geophysical Inversion
Methods in Gravity Field Modelling. Rep 355, Dept Geod Sci, Ohio State Univ, Columbus.
80
Bibliography
Forsberg R (1997) Terrain effects in geoid computation. Lecture Notes, International School of the
Determination and Use of the geoid. Int Geoid Service, Politecnico di Milano.
Forsberg R (2001) Development of a Nordic cm-geoid – with basics of geoid determination. In: Harsson
BG (ed.) Lecture notes for autumn school organised by the Nordic Geodetic Commission, held in Fevik,
Norway, 28 Aug. – 2 Sept. 2000, Statens kartverk, Hønefoss, Norway
Fotopoulos G (2003) An analysis on the optimal combination of geoid, orthometric and ellipsoidal height
data, Ph.D. dissertation, University of Calgary, Canada.
GETECH (1995) Global Digital Elevation Model, Geophsical Exploration Technology (GETECH),
University of Leeds
GLOBE (2003) The Global Land One-km Base Elevation (GLOBE) Project. A 30-arc-second (1-km)
gridded, quality-controlled global Digital Elevation Model (DEM). National Geophysical Data Centre
(NGDC), National Oceanic and Atmospheric Administration (NOAA).
http://www.ngdc.noaa.gov/seg/topo/globe.shtml.
Haghipour, A. & Aghanabati, A., 1985. Geologic Map of Iran: Ministry of Mines and Metals, Geological
Survey of Iran.
Hamesh M (1991) Technical report about adjustment of Iranian first order levelling network, NCC J.
Surveying, 1, 9-20.
Hamesh M (1992) Iranian gravimetric geoid determination Second step. NCC J. Surveying, 6, 17-24, 52-
63.
Heck B and Grüninger W (1987) Modification of Stokes’ integral formula by combining two classical
approaches. Proc. XIX General Assembly of the International Union of Geodesy and Geophysic, Vol. 2
Vancouver, Canada, 309-337
Heck B and Seitz K (2003) Solutions of the linearized geodetic boundary value problem for an ellipsoidal
boundary to order e2 J Geod, 77: 182-192.
Heiskanen WA, Moritz H (1967) Physical Geodesy. W H Freeman and Co., New York, London and San
Francisco
Hunegnaw A (2001) Geoid determination over Ethiopia with emphasis on downward continuation of
gravity anomalies. Royal Institute of Technology, Division of Geodesy report No. 1057 (Doctoral
Dissertation), Stockholm.
Jekeli C (1978) An investigation of two models for the degree variances of global covariance
functions. Dept Geod Sci Rep No. 275, The Ohio State University.
81
Jekeli C, Rapp R (1980) Accuracy of the determination of mean anomalies and mean geoid
undulations from a satellite gravity field mapping mission. Dept Geod Sci Rep No. 307, The Ohio
State University.
Kaula WM (1963) The Investigation of the Gravitational Fields of the Moon and Planets with
Artificial Satellites. Adv. in Space Sci. and Tech., vol. 5, pp. 210-226
Kiamehr R (1997) Iranian geoid problem and recommendation, 1997, 4th International Conferences of
civil engineering, Sharif University of Technology, Tehran, Iran.
Kiamehr R (2001) Potential of Iranian geoid for GPS/Levelling, National Cartographic Centre of Iran
Geomatics 80 conferences, Tehran, Iran.
Kiamehr R (2003a) Comparison relative accuracy of EGM96 and Iranian gravimetric geoid. In:
Proceedings sixth international conference of civil engineering. Isfahan University of Technology, Iran,
pp. 537–544
Kiamehr R (2003b) Discussion about relative accuracy of IFAG geoid, Journal of Sepehr, National
Geography Organization, (3) 23-29.
Kiamehr R (2004) The Relative Accuracies of Recent Satellite Gravimetric Models in Iran, IAG
International Symposium Gravity, Geoid and Space Missions - GGSM2004, Porto, Portugal.
approach, A case study of Iran, Proc. Annual Geomatics 84 Conferences, Iranian National Cartographic
Centre, (CD-ROM), Tehran, Iran. Submitted J. Studia Geophysica et Geodaetica.
Kiamehr R (2006a) A strategy for determining the regional geoid in developing countries by combining
limited ground data with satellite-based global geopotential and topographical models: A case study of
Iran, J. Geodesy , 79(10,11): 602–612.
Kiamehr R (2006b) Hybrid precise gravimetric geoid model for Iran based on recent GRACE and SRTM
data and the least squares modification of Stokes formula, J. Physics of Earth and Space, 32(1) 7-23
GPS/levelling geoid models. Dynamic Planet 2005, IAG Symposia Series by Springer Verlag
Kiamehr R (2006d) The Impact of the lateral density variation model in determination of the precise
gravimetric geoid in mountainous areas: A case study of Iran, Submitted to Geophysical Journal
International.
Kiamehr R and Sjöberg LE (2005a) The qualities of Iranian gravimetric geoid models versus recent
gravity field missions. J SGG 49:289–304
Kiamehr R and Sjöberg LE (2005b) Effect of the SRTM global DEM in the determination of a high-
resolution geoid model of Iran. J Geodesy 79(9):540–551
82
Bibliography
Kiamehr R and Sjöberg LE (2005c) Surface Deformation Patterns Analysis Using 3D Finite Elements
Method: A case study in Skåne area, Sweden, J. Geodynamics, 39( 4) 403-412.
Kiamehr R and Sjöberg LE (2006) Impact of the precise geoid model in studying tectonic structures- A
case study in Iran, J. Geodynamics.
Kotsakis C, Fotopoulos G and Sideris MG (2001a) A Study on the Effects of Data Accuracy and Datum
Inconsistencies on Relative GPS Levelling. In Proceedings of the International Association of Geodesy
Symposium on Vertical Reference Systems, Cartagena, Colombia, February 20-23, IAG Symposia, vol.
124, pp. 113-118.
Kotsakis C, Fotopoulos G and Sideris MG (2001b) Optimal fitting of gravimetric geoid undulations to
GPS/levelling data using an extended similarity transformation model. Presented at the 27th Annual
Meeting joint with the 58th Eastern Snow Conference of the Canadian Geophysical Union, Ottawa,
Canada, May 14-17.
Kotsakis C, Sideris MG (1999) On the adjustment of combined GPS/levelling/geoid networks. Journal of
Geodesy, Vol 73(8), 412-421.
Lee J.T., D.F. Mezera (2000) Concerns Related to GPS-derived Geoid Determination. Survey Review 35:
pp 379-397
Lemoine FG, Kenyon SC, Factor JK, Trimmer RG, Pavlis NK, Chinn DS, Cox CM, Klosko SB,
Luthcke SB, Torrence MH, Wang YM, Williamson RG, Pavlis EC, Rapp RH and Olson TR (1998)
The Development of the Joint NASA GSFC and NIMA Geopotential Model EGM96. NASA
Goddard Space Flight Center, Greenbelt, Maryland, 20771 USA
Martinec Z and Grafarend EW (1997) Solution to the Stokes' boundary value problem on an ellipsoid of
revolution. Studia geophysica et geodaetica, 41, 103-129.
Martinec Z (1998) Boundary-value problems for gravimetric determination of a precise geoid. Lecture
Notes in Earth Sciences No. 73, Springer.
Meissl PA (1971) A Study of Covariance Functions from Discrete Mean Values. Rep. 151, Dept Geod
Sci, Ohio State University, Columbus
Molodensky MS, Eremeev VF, Yurkina MI (1962) Methods for study of the external gravitational field
and figure of the earth, Israeli Programme for the Translation of Scientific Publication, Jerusalem.
Moritz H (1980) Advanced Physical Geodesy. Wichmann, Karlsruhe Omang OCD, Forsberg R (2003)
How to handle the topography in geoid determination: three examples. J Geod 74: 458-466
Moritz, H (1992) Geodetic Reference System 1980, Bull Geod, 66: 187-192
Najafi MA (2004) The KNT University geoid model, Department of research, Technical Report, National
Cartographic Centre (NCC), TOTAK project, Tehran, Iran
83
Nahavandchi H (1999) Precise gravimetric-GPS geoid determination with improved topographic
corrections applied over Sweden. PhD Thesis, Division of Geodesy Rep 1050, Royal Institute of
Technology, Stockholm
Nahavandchi H, Sjöberg LE (2001) Two different views of topographical and downward-continuation
corrections in the Stokes-Helmert approach to geoid computation. J Geod 74: 816-822
Nankali H, (2005) National Cartographic Centre of Iran, Personal Communication.
Nsombo P (1996) Geoid Determination Over Zambia, PhD Thesis, Royal Institute of Technology (KTH)
Pan M, Sjöberg LE and Talbot CJ (2001) Crustal movements in Skåne, Sweden, between 1992 and
1998 as observed by GPS, Journal of Geodynamics, 31, 311-322.
Paul MK (1973) A method for evaluating the truncation error coefficients for geoidal heights. Bull Geod
110: 413-425
Pollastro RM, Persits and Steinshouer FD (1999) Map showing geology, oil and gas fields, and geological
provinces of Iran: U.S. Geological Survey Open-File Report 97-470G, U.S. Geological Survey, Denver,
CO, USA.
Press WH, Teukolsky SA, Vetterling WT, Flannery BP (1992) Numerical recipes in FORTRAN,
2nd edition. Cambridge University Press, Cambridge
Rapp RH and Pavlis NK (1990) The development and analysis of geopotential coefficients models to
spherical harmonic degree 360, J. Geophys. res., 95(B13) 21885-21911.
Rapp RH and Rummel R (1975) Methods for the computation of detailed geoids and their accuracy.
Report 233, Department of Geodetic Science and Surveying, Ohio State University.
Sandwell DT and Smith WHF (1997) Marine gravity anomaly from Geosat and ERS 1 satellite altimetry,
Journal of Geophysical Research, v. 102, No. B5, p. 10039-10054.
Safari A, Ardalan AA and Grafarend EW (2005) A new ellipsoidal gravimetric, satellite altimetry and
astronomic boundary value problem, a case study: the geoid of Iran. J. Geodyn. 39, 545–568.
Sansó F (1997) Lecture Notes, International School of the Determination and Use of the geoid. Int Geoid
Service, Politecnico di Milano
Scherneck HG, Johansson JM, Mitrovica JX and Davis JL, (1998) The BIFROST project: GPS
determined 3-D displacement rates in Fennoscandia from 800 days of continuous observations in the
SWEPOS network. Tectonophysics, 294, 305–321.
Scherneck HG, Johansson JM, Elgered G, Davis JL, Jonsson B, Hedling G, Koivula H, Ollikainen M,
Poutanen M, Vermeer M, Mitrovica JX and Milne GA (2000) BIFROST: Observing the Three-
Dimensional Deformation of Fennoscandia, in Glacial Isostatic Adjustment and the Earth System,
84
Bibliography
edited by J.X. Mitrovica and B.L.A. Vermeersen, Geodynamics Series, Volume 29, American
Geophysical Union, Washington, D.C., in press.
Scherneck HG, Johansson JM, Vermeer M, Davis JL, Milne GA and Mitrovica JX (2001) BIFROST
project: 3-D crustal deformation rates derived from GPS confirm postglacial rebound in Fennoscandia, J
Earth Planets Space, 53 (703-708).
Silver PG, Carlson RW and Olson P (1988) Deep slabs, geochemical heterogeneity and the large scale
structure of mantle convection: investigation of an enduring paradox: Earth and Planetary Science
Letters 16, 477-541.
Sjöberg LE (1975) On the discrete boundary value problem of physical geodesy with harmonic
reductions to an internal sphere. Doctoral Thesis, the Royal Institute of Technology, Stockholm
Sjöberg LE (1977) On the errors of spherical harmonic developments of gravity at the surface of
the Earth. Dept Geod Sci Rep No. 257, The Ohio State University
Sjöberg LE (1980a) A recurrence relation for the βn-function. Bull Geod 54: 69-72
Sjöberg LE (1980b) Least squares combination of satellite harmonics and integral formulas in
physical geodesy. Gel Beitr Geophys 89(5): 371-377
Sjöberg LE (1981) Least squares modification of satellite and terrestrial data in physical geodesy.
Ann Geophys 37(1): 25-30
Sjöberg LE (1983) The Fennoscandian land uplift spectrum and its correlation with gravity. Proc. XVIII
IUGG general assembly, Ohio state university, Ohio, pp. 166-179
Sjöberg LE (1984) Least squares modification of Stokes' and Vening Meinesz' formulas by accounting for
truncation and potential coefficient errors. Manuscr Geod: 9, 209-229
Sjöberg LE (1986) Comparison of some methods of modifying Stokes' formula. Proc. 10th General
Meeting of the NKG, pp. 268-278. Also in Boll Geod Sci Aff, 46(2): 229-248
Sjöberg LE (1991) Refined least squares modification of Stokes' formula. Manuscr Geod: 16: 367-375
Sjöberg LE (1995) On the quasigeoid to geoid separation. Manuscr Geod 20:182-192
Sjöberg LE (1996) The terrain effect in geoid computation from satellite derived geopotential models.
Boll Geod Sci Aff 55: 385-392
Sjöberg LE (1998) The atmospheric geoid and gravity corrections. Boll Geod Sci Aff, 57(4): 421-435
Sjöberg LE (1999a) On the downward continuation error at the Earth's surface and the geoid of
satellite derived geopotential models. Boll Geod Sci Affini 58(3): 215-229
Sjöberg LE (1999b) The IAG approach to the atmospheric geoid correction in Stokes formula and a
new strategy. J Geod, 73(7): 362-366
Sjöberg LE (2000) On the topographic effects by the Stokes-Helmert method of geoid and quasi-geoid
determinations. J Geod, 74(2): 255-268
85
Sjöberg LE (2001a) Topographic and atmospheric corrections of gravimetric geoid determination with
special emphasis on the effects of degree zero and one. J Geod 75: 283-290
Sjöberg LE (2001b) The effect of downward continuation of gravity anomaly to sea-level in Stokes'
formula. J Geod 74: 796-804
Sjöberg LE (2003a) A solution to the downward continuation effect on the geoid determined by
Stokes' formula. J Geod 77: 94-100
Sjöberg LE (2003b) Improving modified Stokes’ formula by GOCE data. Boll Geod Sci Aff 61 (3): 215-
225
Sjöberg LE (2003c) A computational scheme to model the geoid by the modified Stokes formula without
gravity reductions. Journal of Geodesy, 74, 255-268.
Sjöberg LE (2003d) A general model of modifying Stokes’ formula and its least squares solution. J Geod
77: 459-464
Sjöberg LE (2003e) The ellipsoidal corrections to topographic geoid effects. J Geod 77: 804-808
Sjöberg LE (2004a) The effect on the geoid of lateral density variations. J Geod 78: 34-39
Sjöberg LE (2004b) A local least-squares modification of Stokes' formula. Accepted for publication
by Studia Geodaetica et Geophysica
Sjöberg LE (2004c) A spherical harmonic representation of the ellipsoidal correction to the modified
Stokes formula, Journal of Geodesy, 78(3) 1432-1394
Sjöberg L.E. (2006) The Effects of Stokes’ Formula for an Ellipsoidal Layering of the Earth’s
Atmosphere, J. Geodesy, 79(12): 675 - 681
Sjöberg LE and Featherstone WE (2004) Two-step procedures for hybrid geoid modelling. J Geod 78:
66-75
Sjöberg LE and Hunegnaw A (2000) Some aspects of the modification of Stokes formula. J Geod
74: 232-238
Sjöberg LE, Nord T and Fan H (1991) The Fennoscandian geoid bulge and its correlation with land uplift
and Moho depth, Presented in the XX IUGG General Assembly, Vienna, Austria
Sjöberg LE and Nahavandchi H (2000) The atmospheric geoid effects in Stokes' formula. Geophys. J.
Int. 140: 95-100
Tapley B, Ries J, Bettadpur S, Chambers D, Cheng M, Condi F., Gunter B, Kang Z, Nagel P, Pastor R,
Pekker T, Poole S and Wang F (2005)GGM02 - An improved Earth gravity field model from GRACE,
Journal of Geodesy , DOI 10.1007/s00190-005-0480-z
86
Bibliography
Tscherning CC and Rapp (1974) Closed Co variance Expressions for Gravity Anomalies, Geoid Undulations,
and Deflections of the Vertical Implied by Anomaly Degree-Variance Models. Rep of the Dept of Geodetic
Sci and Surv No 208 The Ohio State Univ, Columbus, Ohio.
Rummel R (1997) Spherical spectral properties of the Earth’s gravitational potential and its first and
second derivatives, in: Geodetic Boundary Value Problems in View of the One Centimeter Geoid, F
Sansò and R Rummel (eds.), Lecture Notes in Earth Sciences 65, pp. 359–404.
Vaníček P, Najafi M, Martinec Z, Harrie L and Sjöberg LE (1995) Higher-degree Reference Field in the
Generalized Stokes-Helmert Scheme for Geoid Computation. Journal of Geodesy 70,176-182.
Weber G and Zommorrodian H (1988) Regional geopotential model improvement for the regional Iranian
geoid determination. Bulletin Géodésique, 62, 125-141.
Wenzel HG (1985) Hochauflösende Kuglefunktionsmodelle für das Gravitationspotential der Erde.
Wenzel HG (1998) Ultra-high degree geopotential models GPM98A and GPM98B to degree 1800, Report
98:4, Finnish Geodetic Institute, Masala, 71-80.
Wissenschaftliche Arbeiten der Fachrichtung Vermessungswesen der Universität Hannover Nr. 135.
Wong L, Gore R (1969) Accuracy of geoid heights from modified Stokes kernels. Geophys J R Astron
Soc 18: 81-91
Vermeer M (1995) Mass point geopotential modelling using fast spectral techniques; historical
overview, toolbox description, numerical experiment, manuscripta geodaetica, 20: 362-378
Ågren J (2004a) Regional geoid determination methods for the era of satellite gravimetry, Ph.D. thesis,
Royal Institute of Technology, Stockholm, Sweden.
Ågren J (2004b) The analytical continuation bias in geoid determination using potential coefficients
and terrestrial gravity data. J. Geodesy, 78, 314-332.
Ågren J, Kiamehr R and Sjöberg LE (2006a) Progress in the Determination of a Gravimetric Quasi-geoid
Model over Sweden, Nordic Geodetic Commission General Assembly, May 29 to June 2, 2006,
Copenhagen, Denmark.
Ågren J, Kiamehr R and Sjöberg LE (2006b) The Swedish geoid as evaluated by the method of least-
squares modification with additive, the first International Symposium of the International Gravity Field
Service (IGFS), August 28 – 1 September 2006, Istanbul, Turkey.
Ågren J and Sjöberg LE (2004) Comparison of some methods for modifying Stokes' formula in the
GOCE era. Poster at the 2nd International GOCE user workshop, held at ESA-ESRIN, Frascati,
Italy, March 8-10, 2004
87
Part Two
Publications
PAPER A:
approach, A case study of Iran, Proc. Geomatics 2004 (84) Conferences, National Cartographic
Centre of Iran, Tehran, Iran, Revised version, submitted J. Acta Geodaetica et Geophysica
Link: http://www.ncc.org.ir/_DouranPortal/documents/articles/kiamehr.pdf
PAPER B:
Link: http://dx.doi.org/10.1007/s11200-005-0011-7
PAPER C:
Kiamehr R and Sjöberg LE (2005b) Effect of the SRTM global DEM on the determination of a
high-resolution geoid model: a case study in Iran, J. Geodesy, 79(9):540-551.
PAPER D:
combining limited ground data with satellite-based global geopotential and topographical models:
A case study of Iran, J. Geodesy , 79(10,11): 602–612.
PAPER E:
Kiamehr R (2006b) Hybrid precise gravimetric geoid model for Iran based on recent GRACE and
SRTM data and the least squares modification of Stokes formula, J. Physics of Earth and Space,
32(1) 7-23.
Link: http://geophysics.ut.ac.ir/En/Journal/Archive.asp
88
PAPER F:
GPS/levelling geoid models, Published Dynamic Planet 2005 IAG Symposia Series Springer
Verlag, Revised version in press, Acta Geodaetica et Geophysica, 42(1)
Link: http://www.dynamicplanet2005.com/downloads/DP%20Abstracts%20poster.pdf
PAPER G:
Kiamehr R (2006d) The Impact of lateral density variation model in the determination of precise
gravimetric geoid in mountainous areas: A case study of Iran, in press, Geophysical Journal
International.
http://www.eproof.techbooks.com/cgi-bin/al?aid=30651gF8883qPeO2898XW
PAPER H:
Kiamehr R and Sjöberg LE (2006) Impact of the precise geoid model in studying tectonic
structures- a case study in Iran, J. Geodynamics, 42(2006) 1-11
Link: http://dx.doi.org/10.1016/j.jog.2006.04.001
PAPER I:
Kiamehr R and Sjöberg LE (2005c) Surface Deformation Patterns Analysis Using 3D Finite
Elements Method: A case study in Skåne area, Sweden, J. Geodynamics, 39(4) 403-412.
Link: http://dx.doi.org/10.1016/j.jog.2005.03.001
89

FULLTEXT01

Uploaded by

FULLTEXT01

Uploaded by

ROYAL INSTITUTE

Precise Gravimetric Geoid Model for Iran

Doctoral Dissertation in Geodesy

Royal Institute of Technology (KTH)

© Ramin Kiamehr 2006

Royal Institute of Technology (KTH)

Printed by Universitetsservice US-AB.

Keywords: Gravity database, least-squares modification of Stokes, geoid determination, SRTM,

Stockholm, October 2006

PART ONE - THE HYBRID PRECISE GEOID MODEL FOR IRAN 1

2. GEOID DETERMINATION BASED ON LEAST- SQUARES MODIFICATION OF

2.1 Modification of Stokes' formula 11

3. ADDITIVE CORRECTIONS IN THE KTH APPROACH 21

3.1 The Topographic Corrections 22

4. EVALUATION AND CONSTRUCTION OF THE DATABASE 29

4.1 Qualification and creation of gravity anomaly database 29

5.1 Basic Parameters and Results 39

6. EFFECT OF LATERAL DENSITY VARIATION ON THE GEOID 45

7. EVALUATION OF THE GRAVIMETRIC GEOID MODELS 51

7.1 Overview of current Iranian Gravimetric Geoid models 51

8. A NEW HEIGHT DATUM FOR IRAN 67

8.1 Creation of the corrective surface 67

9.1 Impact of the geoid model in studying tectonic activities in Iran 69

10. CONCLUSIONS AND RECOMMENDATIONS 76

PART TWO – PUBLICATIONS 88

Table 5. Validation of the IRG04 model versus 35 precise GPS/levelling data.

The determination and availability of a high-resolution and more-accurate geoid model is

1.2. Research objectives and author’s contribution

1.2.1. Review of Part One

1.2.2. Review of Part two

1.2.3 Other Publications and Presentations

1. Kiamehr R (2003a) Comparison Relative Accuracy of EGM96 and Iranian Gravimetric

2.1 Modification of Stokes’ formula

where b n = Q nL + s n , c = R / ( 2γ ) and Δg nEGM is the Laplace harmonics of degree n can be

calculated from a GGM (Heiskanen and Moritz 1967, p. 89)

The modified Stokes function is expressed as

where the Molodensky’s truncation coefficients Q n can be expressed

respectively. The modification parameters are

the expected global MSE of the geoid estimator Ñ can be written:

error degree variances and

2.2. Signal and noise degree variances

σ n2 = c1 (1 − μ ) μ n , 0 < μ < 1 . (2.17)

σ 2 = C ( 0 ) = c1μ 2 ⇒ C (ψ 0 ) = 0.5c1μ 2 , (2.18)

degree variances can be derived as:

2.3. Determination of the least-squares modification parameters

equations (Sjöberg 2003d):

Additive Corrections in the KTH Approach

geoid in Stokes’ formula to ellipsoidal reference surface.

3.1. The Topographic Corrections

By using this interpolation technique on the smooth topographically reduced gravity

3.2. The Downward Continuation Correction

3.3. The Ellipsoidal Correction

where sn* = sn if 2 ≤ n ≤ M and sn* = 0 otherwise. Furthermore,

(δ ge )n = 2a ∑ {⎡⎣3 − ( n + 2 ) Fnm ⎤⎦ Tnm − ( n + 1) GnmTn−2,m − ( n + 7 ) EnmTn+2,m }Ynm ( P ) ,

3.4. The Atmospheric Correction

Evaluation and construction of the base data

4.1. Qualification and creation of gravity anomaly database

Gravity Data No. of data Min. Max. Mean RMS

4.2. The GPS/levelling data

4.3. The Digital Elevation Models

4.3.1. Iranian SRTM based DEM

4.4. The Global Geopotential Model

4.4.1. GRACE GGM02

The New Gravimetric Geoid Model (IRG04)

5.1. Basic Parameters and Results

Fig. 7. The isolines of the IRG04 geoid model. Unit: m

Effect of Lateral Density Variation on the Geoid

6.3. The effect on the geoid

Fig. 9. Effect of the lateral density variation model on geoid. Unit. m

Evaluation of the Gravimetric Geoid Model

7.1. Overview of current Iranian Gravimetric Geoid models

7.2. Internal Estimation of accuracy of IRG04

EGM part (GGM02S) 11 11 -

Degrees All 2-M (L+1) - ∞ (10800)

EGM part (GGM02S) 72 72 -

Fig. 10. Principal of levelling with GPS