April 1999 • NREL/CP-500-26245

Wind Resource Estimation and

Mapping at the National
Renewable Energy Laboratory

M. Schwartz

Presented at the ASES Solar ‘99 Conference

Portland, Maine
June 12-16, 1999

National Renewable Energy Laboratory

1617 Cole Boulevard
Golden, Colorado 80401-3393
NREL is a U.S. Department of Energy Laboratory
Operated by Midwest Research Institute • Battelle • Bechtel
Contract No. DE-AC36-98-GO10337
 NOTICE

This report was prepared as an account of work sponsored by an agency of the United States
government. Neither the United States government nor any agency thereof, nor any of their employees,
makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy,
completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents
that its use would not infringe privately owned rights. Reference herein to any specific commercial
product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily
constitute or imply its endorsement, recommendation, or favoring by the United States government or any
agency thereof. The views and opinions of authors expressed herein do not necessarily state or reflect
those of the United States government or any agency thereof.

Available to DOE and DOE contractors from:

Office of Scientific and Technical Information (OSTI)
P.O. Box 62
Oak Ridge, TN 37831
Prices available by calling 423-576-8401

Available to the public from:

National Technical Information Service (NTIS)
U.S. Department of Commerce
5285 Port Royal Road
Springfield, VA 22161
703-605-6000 or 800-553-6847
or
DOE Information Bridge
http://www.doe.gov/bridge/home.html

Printed on paper containing at least 50% wastepaper, including 20% postconsumer waste
 WIND RESOURCE ESTIMATION AND MAPPING AT THE
NATIONAL RENEWABLE ENERGY LABORATORY

Marc Schwartz
National Renewable Energy Laboratory
1617 Cole Boulevard
Golden, Colorado 80401

ABSTRACT output of the mapping system is a gridded, color-coded wind

power map in units of W/m2 and equivalent mean wind
The National Renewable Energy Laboratory (NREL) has speed for each individual cell. Most of the final wind
developed an automated technique for wind resource resource maps have a 1-km2 grid cell resolution. Despite
mapping to aid in the acceleration of wind energy this high resolution, the maps are not intended to be used for
deployment. The new automated mapping system was micrositing purposes because of terrain induced variability
developed with the following two primary goals: 1) to of the wind resource that can occur within a grid cell.
produce a more consistent and detailed analysis of the wind
resource for a variety of physiographic settings, particularly The computer mapping system uses an empirical and
in areas of complex terrain; and 2) to generate high quality analytical approach to determine the level of the wind
map products on a timely basis. Using computer mapping resource for a particular location. At this time, the wind
techniques reduces the time it takes to produce a wind map mapping system does not use any explicit boundary layer
that reflects a consistent analysis of the distribution of the equations or geostrophic adjustment equations, as some
wind resource throughout the region of interest. NREL's other wind flow models do.
mapping system uses commercially available geographic
information system software packages. Regional wind The major meteorological assumption that underlies NREL's
resource maps using this new system have been produced mapping technique is that there are empirical relationships
for areas of the United States, Mexico, Chile, Indonesia (1), in many parts of the world between the free-air (higher than
and China. Countrywide wind resource assessments are 100 - 200 m above ground level) speed, the wind speed
under way for the Philippines, the Dominican Republic, and over the ocean (where applicable), and the distribution of
Mongolia. Regional assessments in Argentina and Russia are the wind resource over the land areas. Empirical
scheduled to begin soon. relationships have been noticed in previous NREL wind
resource assessment work for well exposed locations with
low surface roughness in diverse areas of the world.
1. APPROACH Accordingly, the wind resource values presented on the
maps are the estimates for a non-sheltered location with low
The mapping system is designed to display regional (greater surface roughness (short grasslands, for example). NREL's
than 50,000 km2) distributions of the wind resource. The mapping system takes a "top-down" method in the
maps are intended to denote areas where wind energy adjustment of much of the available wind data. That is to
projects could be feasible. Figure 1 shows regions of the say, the NREL approach takes the free-air wind profile in
world where the National Renewable Energy Laboratory the lowest few hundred meters above the surface and adjusts
(NREL) has performed recent wind resource assessments. it down to the surface. NREL decided to use the "top-down"
The wind power density, rather than wind speed, is method because of many problems with the available land-
presented on the wind resource maps because it provides a based surface wind data in most of the world. A lack of
truer indication of the wind resource potential. The primary information about observation procedures, and anemometer

1
 RUSSIA (1)
MONGOLIA (1)

POLAND (2)
NREL
CHINA (1)
DOMINICAN
REPUB LIC (1)
MEXICO
(2,3) PHILIPPINES (1)

INDONESIA (1,3)

ARGENTINA (3)
CHILE (1,3)

INDONESIA (1,3)

Projects
US Dept . of Energy - Nation al
Automated GIS maps since 1996 (1) Re newable E nergy La boratory
Subjective maps from 1993 to 1995 (2)
Automated GIS maps since 1996 and
subjective maps from 1993 and 1995 (3)

Fig. 1: National Renewable Energy Laboratory international wind resource mapping projects.

hardware, height, exposure, and maintenance history are just surface wind information used in our assessments. NREL
a few of the problems. In addition, many areas of the world currently has more than 25 years of DATSAV2 data in its
with the potential to have good-to-excellent wind resource archive, spanning the period from 1973 to mid-1998.
sites have very few or no meteorological stations present. Meteorological parameters such as wind speed, wind
The result is that, overall, the available surface wind data direction, temperature, pressure and altimeter setting are
from meteorological stations are not reliable enough to use extracted from the hourly observations and used to create
directly as input in the wind mapping system. However, statistical summaries of wind characteristics.
these data are critically analyzed to assess the wind
characteristics (e.g., seasonal, diurnal, and directional) in a Marine Climatic Atlas of the World - This data set,
study region. compiled from historical ship observations, presents
summarized wind statistics for a 1º latitude by 1º longitude
grid. Measurements are concentrated along the major
2. DATA SETS AND ANALYSIS METHODS shipping routes. Included are summaries of the monthly
means and standard deviations of wind speed, pressure,
The major data sets used in NREL's wind resource temperature, and wind direction frequency and speed.
assessment methodology are described below.
Special Sensor Microwave Imager (SSMI) - The SSMI,
2.1 Surface Data which is part of the Defense Meteorological Satellite
Program, provides 10-m ocean wind speed measurements.
DATSAV2 - This global climatic database, which was This data set provides much more uniform and detailed
obtained from the U.S. National Climatic Data Center coverage of oceanic wind speeds than the Marine Climatic
(NCDC), contains digital hourly surface weather Atlas of the World. NREL currently has 10 years of SSMI
observations from first-order meteorological stations data covering the period from 1988 to 1997.
throughout the world. This data set is the primary source of

2
2.2 Upper-Air Data upper-air observations. The generated summaries are cross-
referenced against each other to aid in understanding the
Upper-air data can provide an estimate of the wind resource prevalent wind characteristics in the study area. For
at low levels in the atmosphere and contribute to the example, the interannual surface wind speeds are evaluated
understanding of the vertical distribution of the wind to identify obvious trends in the data or periods of
resource. This is useful in estimating the winds on elevated questionable data. Data periods determined to be most
terrain features and for estimating the wind resource at other representative are selected from the entire record for use in
exposed locations in a particular region. assessments. The goal of the data analysis and screening
process is to develop a conceptual model of the physical
Automated Data Processing (ADP) Reports - This data set mechanisms, both regional and local, that influence the wind
contains upper-air observations from rawinsonde flow. The ultimate goal of the analysis is to enable the
instruments and pilot balloons for approximately 1800 meteorological analyst to attain a conceptual model of the
stations worldwide. Observation times include 00, 06, 12, physical mechanism(s), whether produced by large and/or
and 18 Greenwich Mean Time (GMT). Wind information is local scale factors that cause the wind to blow in a particular
available from the surface, the mandatory pressure levels region. The conceptual model guides the development of
(1000 mb, 850 mb, 700 mb, and 500 mb), the significant the empirical relationships that serve as the basis of the
pressure levels and, for some stations, specified geopotential algorithms that calculate wind power. The analyst is
heights above the surface. NREL currently has more than ultimately responsible for preparing the meteorological data
25 years of observations from 1973 through mid-1998. input into the mapping system.
Global Gridded Upper-Air Statistics - This data set contains
monthly means and standard deviations of climatic elements 3. MAPPING SYSTEM
for the mandatory pressure levels on a 2.5 degree global
grid. The data were obtained from the NCDC and cover the The mapping system is divided into three main components:
period from 1980 to 1991. This data set is used to the input data, the wind power calculations, and the mapping
supplement the ADP information in areas where upper-air products. These components are described below.
data are scarce.
3.1 Input Data
Reanalysis Data - This data set, created by the U.S. National
Centers for Environmental Prediction and the National The two primary model inputs are digital terrain data and
Center for Atmospheric Research, contains worldwide formatted meteorological data. The elevation information
information on wind, temperature, and other variables on a consists of Digital Elevation Model (DEM) terrain data that
terrain-following, 208-km resolution grid with more than divide the analysis region into individual grid cells, each
18,000 points. The output is available four times a day at with its own elevation value. The United States Geological
00, 06, 12, and 18 GMT. This data set has better vertical Survey Earth Resource Observing Satellite Data Center
resolution in the boundary layer than the ADP data. NREL recently produced new DEMs for most of the world. The
has obtained 40 years (1958-1997) of the raw data and is new data sets have a resolution of 1 km2.
currently investigating the optimum way of integrating the
reanalysis data into its wind resource assessment The meteorological inputs to the mapping system are
methodology. vertical profiles of free-air wind power density, wind power
roses (which specify the percentage of total power from the
2.3 Data Analysis wind by direction sector), and the open ocean wind power
density, where appropriate. The vertical profiles are broken
An accurate wind resource assessment is highly dependent down into 100-m intervals centered every 100 m above sea
on the quantity and quality of the meteorological data input level (asl), except for the lowest layer, which is at 50 m asl.
into the automated system. NREL reviews the data sets The wind power rose is used to determine the degree of
mentioned above, and previous wind energy assessments, as exposure of a particular grid cell to the power producing
part of its overall evaluation. The use of multiple data sets is winds. The open ocean wind power density is derived from
necessary because the quality of data in any particular the SSMI and ship wind speed observations, converted to
dataset can vary, and because high quality data can be quite wind power density and extrapolated to 30 m for use by the
sparse in many regions of the world. mapping system.

A critical analysis of the climatic data is performed to ensure

that the meteorological input is as precise as possible. Wind
characteristic summaries are generated for surface and

3
 97
97° 96
96° 95°
95
94°
94

4 5°
4 5°

45° Ê Montevideo 45°

Ñ Montevideo Hutchinson Ê Wind Power Classification
Brownton Ñ
a
Wind Wind Power Wind Speed
Power Density at 50 m at 50 m
Class W/m2 m/s
Marshall
Ê
Redwood Falls 2 200 - 300 5.6 - 6.4
Marshall Ñ Ê
Brookings 3 300 - 400 6.4 - 7.0
Ñ Lynd
MN
9 7°

Ê% New UlmÊ% 4 400 - 500 7.0 - 7.5

Ê

9 4°
Ruthton Mankato
% 5 500 - 600 7.5 - 8.0
Ñ
Currie Ñ Mountain Lake 6 600 - 800 8.0 - 8.8
44° Pipestone Ê Ñ 44°

4 4°
4
4 4°
4°

4°
a
4

Ñ Ihlen Ñ Chandler Wind speeds are based on a Weibull k value

Brewster Winnebago
Ñ
of 2.0 typical for the Buffalo Ridge area.
Ñ Luverne
SD Worthington %Ê
Ñ
Ê Fairmont %
Ê Ê Meteorological Station with Wind Data
Ê% Sioux Falls Jackson
Ñ Wind Monitoring Site used for Validation

9 4°
Ñ Sibley Ê Estherville % City or Town
9 7°

Ñ Ñ
Inwood Estherville
Spencer Ê%
Algona
Ñ N
43° Ñ 43°
4 3°
4
4 3°
3°

Sutherland 3°
4

20 0 20 40 60 80 100 Kilometers
Alta
Ñ
Cherokee %
IA
% Sioux City Fort Dodge Ê% US Dept. of Energy - National
Ê Sioux City Renewable Energy Laboratory
NE
Ñ
Winnebago
42° 42°
42
4
2°
2°

97° 96° 95° 94°

Fig. 2: Wind resource map for the Buffalo Ridge area.

3.2 Wind Power Calculations wind flow near the surface and the amount of fetch the
prevailing winds have across the water body. The final wind
The mapping system is designed to calculate the wind power power estimates for these coastal or lakeshore cells are
for those grid cells that meet certain exposure and slope based on the maximum value calculated from either of these
requirements. Therefore, the mapping system identifies the methods.
most favorable wind resource areas. The vertical profiles of
free-air wind power density serve as the input for a base 3.3 Mapping Products
wind power density value for a particular grid cell. The
wind power calculations in the wind mapping system are The primary product of the mapping system is a wind power
adjustments to the base free-air wind power density values density map of favorable wind resource areas. The wind
of each grid cell. The factors that either decrease or increase resource values presented are estimates for non-sheltered
the base wind power value for a particular grid cell are locations with low surface roughness (e.g., short grasslands).
terrain considerations, relative and absolute elevation, the The DEM data can be used to create a color-coded
aspect of the terrain, distance from ocean or lake shoreline, elevation map, a hill-shaded relief map, and a map of the
and the influence of thermal or other types of small-scale elevation contours. When combined with the wind power
wind flow patterns. When a coastal area or an area with maps, these products enable the user to obtain an
large water bodies (e.g., lakes, estuaries, and fjords) is being understanding for the three-dimensional distribution of the
mapped, the distance between a grid cell and the body of wind power for the study region.
water also becomes an important component of the wind
power calculation. For areas not located near a large water Two examples of NREL's wind resource map are shown.
body, the wind power is derived from the vertical profile of The first, (Figure 2) is a wind resource map of the Buffalo
wind power density. However, for areas within a few Ridge area of the northern Great Plains. Several wind
kilometers of an ocean, lake, or other large water body, the energy plants are either being installed, or have recently
mapping system calculates the power based on the ambient been installed, in this region. The map, which covers a 3º

4
 118° 118°30' 119° 119°30'

Wind Power Classification

a
Wind Power Wind Speed
ÊÑ Ñ
Resource Density at 30 m at 30 m
2
Potential W/m m/s

ÊÑU% Fair 100 - 200 4.4 - 5.6

25°30' Ñ Putian 25°30'
Ñ
% Moderate 200 - 300 5.6 - 6.4
Ñ Pingtan Good 300 - 500 6.4 - 7.5
Ñ
Excellent 500 - 1600 7.5 - 10+
a
Wind speeds are based on a Weibull k value of 2.0
Honglai estimated f or many coastal areas of SE China.
%

25° 25°
Ñ Ê GTS Surface Meteorological St ation
Quanzhou% Ñ U GTS Surface Meteorological St ation
QingyangÊ
Ñ
% Ñ Addit ional Surf ace Meteorological Station
%
% City
Ñ
Anhai
N

Ê 24°30'
24°30'
UÑ% ÊU 20 0 20 40 60 80 Kilometers
Xiamen

US Dept. of Energy - National

Renewable Energy Laboratory

24° 24°

118° 118°30' 119° 119°30'

Fig. 3: Wind resource map for the central coast of Fujian Province.

latitude by 3º longitude area centered on the western section hope to obtain future wind energy measurements in this
of the Minnesota-Iowa border, shows significant variation of region in order undertake a validation study similar to the
the wind resource in what is generally perceived as a flat one done for the Great Plains.
area (slopes less than 3%). Nevertheless, based on the
upper-air meteorological input, elevation changes in the area 3.4 Limitations of the Mapping Technique
have a notable impact on the estimated resource level. This
area has numerous stations designed specifically for wind There are several limitations to the mapping technique, the
energy measurement, enabling a validation study (2) to be first of which is the resolution of the DEM data. Significant
performed on the mapping results. Nineteen stations were terrain variations can occur within the DEM's 1-km2 area;
used as validation sites. The wind power estimate from the thus, the wind resource estimate for a particular grid cell
mapping system was within 20% of the measured power at may not apply to all areas within the cell. A second
15 of the 19 sites. potential problem is the development of the conceptual
model of the wind flow and its extrapolation to the region.
Figure 3 is a wind resource map for the central coast of There are many complex factors in the wind flow that make
Fujian Province in southeastern China. The map was this an inexact process, including the structure of low-level
produced for a project jointly funded by the U.S. jets and their interaction with the boundary layer, and
Department of Energy and the U.S. Environmental localized circulations, such as land-sea breezes, mountain-
Protection Agency. The region's geographical features valley flows, and channeling effects in mountainous areas.
include an indented shoreline with hills along the Pacific These factors affect the amount of momentum that is
Ocean (South China Sea) coast, a narrow coastal plain in transported to the surface from free-air levels; such
some areas, and mountainous terrain within 50 km of the processes can only be approximated in the mapping system.
coast. The Pacific coast of Fujian has been an area of Finally, the power estimates are valid for areas with low
interest to U.S. wind energy developers; this map is surface roughness. Estimates for areas with a higher surface
designed to help concentrate their efforts on the areas with roughness need to be adjusted down by as much as 25% to
the greatest resource. The best resource is found on the 60%, depending on the amount of obstructions to the wind
South China Sea coast and the higher inland ridges. We flow.

5
4. CONCLUSIONS

During the course of NREL's wind mapping studies, we

encountered many different types of wind climates and
boundary layer interactions for a wide variety of regions.
This factor causes the general structure of the individual
algorithms used in the mapping system to be continually
evolving in order to continue to produce the most useful
wind maps. We have started to validate some of the system
results, recognizing that this is not currently feasible in some
regions because of the lack of quality wind measurements.
Validation results will be used to help improve the mapping
system.

5. ACKNOWLEDGEMENTS

I would like to thank members of NREL's wind resource

assessment group Dennis Elliott, George Scott, Steve
Haymes, Ray George, and Donna Heimiller for their
contributions to software development, data processing, and
GIS mapping applications. This paper was written at the
National Renewable Energy Laboratory in support of the
U.S. Department of Energy under contract number
DE-AC36-GO10337.

6. REFERENCES

(1) Elliott, D.L., Schwartz, M.N., Recent Wind Resource

Characterization Activities at the National Renewable
Energy Laboratory, Windpower '97 Proceedings, American
Wind Energy Association, 1997
(2) Elliott, D.L., Schwartz, M.N., Validation of Regional
Wind Resource Predictions in the Northern Great Plains,
Windpower '98 Proceedings, American Wind Energy
Association, 1998