ArticlePDF Available

Mesoscale modeling as a tool for wind resource assessment and mapping

Authors:
  • AWS Truepower,
  • Underwriters Laboratories

Abstract and Figures

The use of mesoscale modeling for the mapping and assessment of wind energy resource is discussed. Mesoscale modeling simulates, with accuracy, complex wind flows in areas where surface measurements are scant or non-existent. The models used in MesoMap system are mesoscale atmospheric simulation model (MASS) and mass-conserving wind flow model (WindMap). The parameterization of surface roughness, atmospheric stability in the lower boundary layer, and the resolution of mesoscale model are the challenges observed during the validation of wind maps created by the MesoMap system.
Content may be subject to copyright.
4.2 MESOSCALE MODELING AS A TOOL FOR WIND RESOURCE ASSESSMENT AND MAPPING
Michael Brower*, J. W. Zack, B. Bailey, M. N. Schwartz, and D. L. Elliott
TrueWind Solutions, Albany, NY, and National Renewable Energy Laboratory, Golden, CO
1. INTRODUCTION
The rapid growth of the wind energy industry in the
past decade, both in the United States and elsewhere,
has led to the development of new techniques for the
systematic identification and evaluation of candidate
wind project sites. One of those techniques is
mesoscale modeling. A familiar tool of weather
forecasting, mesoscale modeling offers a number of
advantages for wind resource assessment, such as the
ability to simulate, with reasonable accuracy, complex
wind flows in areas where surface measurements are
scant or non-existent. TrueWind Solutions has
developed one such mesoscale modeling technique,
which is marketed under the trademark MesoMap.
MesoMap has been used in the past five years to create
wind resource maps of about half of the United States
as well as other parts of the world. Through TrueWind’s
internal efforts and a public/private partnership with the
National Renewable Energy Laboratory (NREL), the
accuracy of the method has been assessed by
comparing estimates with data from nearly 1000 surface
stations. The typical root-mean-square model error,
after accounting for uncertainty in the data, has been
found to be 5-7% of the mean speed at 50 m. This is a
useful level of accuracy, but improvements are still
desirable because of the cubic relationship between the
energy available in the wind and speed. The intensive
validation effort has identified several common sources
of model error. This paper discusses three of the most
important: surface roughness parameterization,
atmospheric stability in the lower boundary layer, and
the mesoscale model resolution. Strategies for
researching and mitigating these issues are discussed.
2. BACKGROUND
The successful development of wind energy
requires a thorough understanding of the wind resource
– in effect, an accurate climatology of the wind at a high
spatial resolution. In the early years of the wind industry
– the 1970s and 1980s – such assessments were done
mainly using field techniques, combined with a practical
understanding of large scale wind patterns and the
effect of topography on wind flow. While often very
effective, these techniques suffered from a lack of
transferability; expertise in one region did not always
translate into expertise in another.
During the 1980s and 1990s, a variety of computer
modeling techniques emerged. Several involved
equilibrium microscale wind flow models, the most
prominent example being the Wind Atlas Statistical
Package, or WAsP, developed by the Risoe National
Laboratory of Denmark based on the theory of Jackson
and Hunt (1975). This model creates a wind map and
climatology of a region using data from a single
reference mast. It and its cousins (MS-Micro, WindMap,
and others) are best suited to estimating the wind
resource in areas of simple to moderate terrain slopes
at distances of up to tens of kilometers from the
reference mast (Bowen and Mortensen, 1996;
Walmsley, Troen, Lalas and Mason, 1990).
In the 1990s, the National Renewable Energy
Laboratory (NREL) developed a “computer mapping
system” that uses upper-air wind data from balloon
soundings and various mathematical relationships
between the wind and topography to estimate the wind
resource over large regions at a grid scale of 1 km
(Schwartz, 1999; Schwartz and Elliott, 2001). This
method produced some of the first detailed wind
resource maps of states in the United States (Vermont,
North and South Dakota, and Illinois) as well as other
countries (the Philippines and Mongolia, among others).
By the late 1990s, mesoscale modeling techniques
were beginning to emerge as a major focus of research.
One of the first was the KAMM-WAsP method
developed by Risoe. This method uses the KAMM
mesoscale model to simulate a representative number
of static “cases” sampled from a distribution of upper-air
wind statistics (Frank, Rathmann, Mortensen, and
Landberg, 2001). The output of the model, at a typical
grid scale of 2-5 km, is used to drive WAsP, which
produces wind resource estimates at a much higher
resolution.
TrueWind Solutions developed its own mesoscale
modeling approach, MesoMap, in the late 1990s, with
funding from the New York State Energy Research and
Development Authority (NYSERDA), the US
Department of Energy (DOE), and private sources
(Brower, Bailey, and Zack, 2001). Aside from the
different models used (described below), a key
distinction between MesoMap and KAMM-WASP is that
MesoMap’s mesoscale model is run in a dynamic mode
with the energy equations. This allows the development
of non-equilibrium mesoscale flows (sea breezes being
an obvious example) within the model domain.
3. THE MESOMAP SYSTEM
The MesoMap system has several major
components. First, there are the models: a mesoscale
atmospheric simulation model (MASS) and a mass-
__________________________________________
* Corresponding author address: Michael Brower,
TrueWind Solutions, 255 Fuller Road, Suite 274,
Albany, NY 12203; e-mail: mbrower@truewind.com.
conserving wind flow model (WindMap). MASS is similar
in many respects to the MM5 family of mesoscale
models, but is a commercial program developed by
MESO, Inc., a co-owner of TrueWind (Manobianco,
Zack and Taylor, 1996). In mapping projects, MASS
normally operates at a scale of 1-3 km. WindMap is
based on the NOABL program, which was developed in
the 1970s and 1980s for wind resource studies and
subsequently updated by Brower & Company
(Sherman, 1978; Brower, 1999). Starting with an initial
mesoscale wind field (provided, in this case, by MASS),
it finds a solution that conserves mass at the
microscale. It normally operates at a grid scale of 100 to
200 m, which is roughly comparable to the spacing
between turbines in wind projects.
The second major component is a distributed
computer processing system consisting of 94 Pentium
III and IV processors connected in a network. It is the
parallelization of the mapping process that makes it
possible to produce high-resolution maps using this
technique in a reasonable amount of time. A typical
MesoMap project requires two CPU-years of
processing, but can be completed on this system in
about a week.
Global meteorological data bases (reanalysis,
surface, and rawinsonde) and geophysical data bases
(topography, land cover, vegetation greenness, sea
temperatures, snow cover, soil moisture) make up the
third component. The reanalysis data, which are
produced by the National Centers for Environmental
Prediction (NCEP), provide a three-dimensional
snapshot of global weather conditions every 6 hours
over the past several decades on a 2.5 degree grid.
Along with rawinsonde and surface data, they provide
the initial conditions for the MASS simulations, and
they provide updated lateral boundary conditions. The
topographic and land cover data are essential, of
course, to properly simulating interactions between the
atmosphere and land or ocean surface.
The mapping process begins by defining several
grids around the area to be mapped. The largest is
typically more than 2000 km wide, with a mesoscale
grid spacing of 30 km. Within that large grid there are
usually two or three levels of nested grids, each
covering a smaller area at higher resolution, with the
last extending perhaps 200-400 km at a grid scale of
1-3 km. The mesoscale model then simulates weather
and wind conditions throughout the area at all levels of
the atmosphere for 366 days randomly sampled from a
15 year period. The three-dimensional output of the
model (including wind, temperature, pressure, and
other parameters) is stored every hour of simulated
time, resulting in a total of 8784 samples at each grid
point.
The results of the mesoscale simulations are then
summarized in data files containing gridded wind rose
and Weibull statistics at 11 levels above the surface.
These files are input into the Windmap model.
Windmap, in effect, perturbs the MASS wind field to
account for differences in the topography and land cover
as seen by Windmap and MASS.
4. VALIDATION PROGRAM
MesoMap has been applied in the past several
years to mapping wind resources in nearly 30 states of
the United States as well as in almost 30 other
countries. The successful application of MesoMap is in
part the result of an unusual public/private collaboration
between TrueWind and NREL, with support from the US
Department of Energy’s Wind Powering America
y = 0.9852x
R2 = 0.8963
2.0
3.0
4.0
5.0
6.0
7.0
8.0
9.0
10.0
11.0
2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 12.0
Predicted Speed (m/s)
Measured/Extrapolated S peed (m/s
)
Figure 1. Annual average wind speed map of New Mexico
created using MesoMap.
Figure 2. Estimated and measured long-term mean wind
speeds at 59 stations in New Mexico. The data have been
extrapolated to a reference height of 50 m.
initiative. Recognizing the critical importance of verifying
the accuracy of the maps and making corrections, if
necessary, NREL has provided both data and
meteorological expertise, which have greatly bolstered
TrueWInd’s own validation efforts.
Figure 1 shows a typical wind speed map produced
by MesoMap, this one of New Mexico. Figure 2 shows
the corresponding validation results, here represented
as a scatterplot of estimated and measured mean wind
speed. Typically, the root-mean-square discrepancy
between model and data is 7-10%. After accounting for
uncertainties in the data (resulting from limited periods
of record, low mast height, and other factors), the rms
error ascribed to the model alone usually falls in the
range of 5-7%. To place this in perspective, the error
margin in a high-quality measurement program is
usually 3-4%. (It should be noted that only the mean
annual wind speed and power estimates have been
validated, not seasonal or diurnal patterns or other
aspects of the wind climate.)
Although the typical error in mean wind speed is
moderate, it has a considerable impact on wind project
feasibility because the energy available in the wind
varies as the cube of the mean speed (assuming a fixed
speed frequency distribution). A 5-7% rms error in
speed implies a 16-22% rms error in available energy.
Because wind turbines do not convert all of the available
energy to electricity, the rms error in wind turbine output
is actually somewhat smaller – about 10-15%.
Nevertheless this represents a substantial uncertainty
for the financial evaluation of wind projects.
Consequently, it is important to try to determine and
mitigate the largest sources of model errors. While in
any particular region a variety of factors may be at work,
in our experience, three have often been significant: the
parameterization of surface roughness, the simulation of
the stable boundary layer, and the mesoscale model
grid scale.
5. SURFACE ROUGHNESS PARAMETERIZATION
In similarity theory, surface roughness is a property
that helps determine the vertical profiles of wind,
temperature, turbulence, and other atmospheric
parameters. It is represented in the familiar logarithmic-
linear formulas by a surface roughness length, in
meters, which is related in principle to the height, width,
porosity, and average spacing between “rough
elements” (such as vegetation and man-made
structures) on the land surface. For given conditions of
thermal stability, a high roughness induces a greater
wind shear and lower wind speed at the surface,
whereas a low roughness has the opposite effect. In
mesoscale models, the roughness length does not
directly determine the vertical profile, but rather
influences the flux of energy and momentum between
the land and atmosphere at the lowest model level.
A key challenge in producing accurate wind
resource maps of large areas is to estimate the surface
roughness using remotely sensed data. To do this we
make use of a table of common equivalences, drawn
from the literature (e.g., Garrat, 1992), between various
land cover types (forest, cropland, etc.) and roughness
length. In making such estimates, at least three
problems can occur:
5.1. Errors in land cover description. Remotely
sensed data must be properly interpreted to determine
the land cover type. Rather than do such interpretation
ourselves, we rely on publicly available land cover data
sets. For the mesoscale model, the source is the global
1 km Advanced Very High Resolution Radiometer
(AVHRR) data set (Brown, Loveland, Ohlen, and Zhu,
1999). For the microscale model, a 30 m Landsat data
set is used in the United States (Vogelmann, Sohl,
Campbell, and Shaw, 1998), while a 250 m CORINE
data set is available for most of Europe (Fuller and
Brown, 1994); for other regions, we usually use the
global 1 km Moderate Resolution Imaging
Spectroradiometer (MODIS) data set (Hodges, Friedl,
and Strahler, 2001).
Mistakes frequently arise in the global data sets
because of their relatively coarse resolution. In a
systematic ground-truthing of the AVHRR land cover
data, it was found that the surface roughness
designation based on land cover class was correct
about 80% of the time (Defries and Los, 1999). In our
experience, the MODIS data are somewhat more
accurate than AVHRR, while the Landsat and CORINE
data are probably more accurate than either. As far as
we know, however, no systematic validation of
roughness derived from these other data sets has yet
been performed.
5.2. Errors in roughness assignment. Even when the
land cover type is correct, it is not always easy to
determine the appropriate roughness length. Surface
roughness depends on vegetation height and density,
among other things, which land cover descriptions say
nothing about. Forests can be sparse and deciduous or
dense and evergreen, with the roughness length
ranging, as a result, from as low as 0.3 m to as high as
2 m. The cropland designation can encompass
everything from vast open spaces of wheat to small
fields separated by wind breaks, farm buildings, fences,
and other structures; in the first instance, the surface
roughness length could be as low as 0.02 or 0.03 m; in
the latter, it could be as high as 0.1 to 0.15 m. Seasonal
variations can be important as well. In winter, snow
cover can create a very smooth surface, whereas in the
summer and fall when grasses and crops are at their
tallest, the roughness is comparatively high.
5.3. Errors in displacement height. Where vegetation
is particularly tall and dense, the wind flow can be
displaced upwards off the ground. The displacement
height is defined as the height at which a logarithmic
wind profile reaches zero. A typical estimate of the
displacement height is two-thirds of the average
vegetation height (Garrat, 1992). For densely spaced
trees 20 meters tall, according to this rule of thumb, the
displacement height would be about 17 m. Such a large
displacement can substantially reduce the observed
wind speed at typical wind turbine hub heights of 50-80
m. Like roughness, however, it cannot be determined
exactly from the land cover designation.
5.4. Roughness corrections at the microscale. The
surface roughness at the microscale is often quite
different from what it is at the mesoscale – in part
because of the higher resolution, and in part because
we generally use a different, region-specific land-cover
data set for the microscale simulations. Our microscale
model, WindMap, must adjust the vertical profile and
surface winds to account for the different roughness.
Since WindMap lacks the turbulent mixing equations of
a mesoscale model, this necessitates some
approximations – for example, to describe the rate of
growth of the internal boundary layer downwind of a
roughness change, or to describe the maximum height
to which a different equilibrium roughness will affect the
wind.
The wind profiles in Figure 3 illustrate the potential
effect of errors in roughness. In this example, changing
the roughness from 0.03 m to 0.4 m would cause the
estimated mean speed at 50 m to decrease by nearly
1.5 m/s, or 12%. The impact could be even greater –
20% or more – under stable atmospheric conditions.
While such extreme errors in roughness –
equivalent to confusing cropland with forest – are
relatively rare and confined to small regions, moderate
errors are quite common and can be widespread. A
case in point is New Mexico. Initially areas denoted as
shrubland in the Landsat data set – including much of
southeast New Mexico and valleys west of the Rockies
– were assigned a roughness of 0.02 m. After a round of
validation, it was concluded that this value was too low,
and it was raised to 0.07 m. This led to a reduction of
about 5% in the estimated 50 m wind speed.
Our main strategies for addressing such errors are,
first, to verify to the extent possible the roughness
designations for different land cover types by examining
photographs and other information; and second, to
acquire more accurate land cover data. The Landsat
and CORINE data sets for the United States and
western Europe have proven invaluable. Elsewhere, we
have adopted the MODIS data for the final stage of
mapping, and will eventually replace AVHRR with
MODIS for the mesoscale modeling as well.
6. THE STABLE BOUNDARY LAYER
The thermal stability of the atmosphere has an
equally important effect on the vertical wind profile and
on estimates of the wind resource at a particular height.
In a stable boundary layer, the air has negative
buoyancy, so that a parcel of air displaced upwards or
downwards tends to return to its previous level. As a
result, mixing and friction are confined to a shallow layer
near the surface, above which the wind increases
rapidly. In an unstable atmsosphere, in contrast,
momentum is spread by convective mixing more evenly
throughout a much deeper boundary layer. This typically
raises the wind speed very near the surface and
reduces it above.
Mesoscale models like MASS approximate these
effects by changing the mixing parameterization
depending on the stability class in which the boundary
layer falls. The stability class is determined by the
predicted potential temperature profile. Although the
equations work well most of the time, highly stable
conditions pose a particular challenge. Sometimes the
model allows more momentum to penetrate the stable
layer to the surface than occurs in reality. This results in
an overestimation of the wind resource.
Figure 4 illustrates a typical range of overestimation
of nocturnal winds using data from Kansas City
International Airport, Missouri, and Douglas Bisbee
International Airport, Arizona. The problem is mild in the
first but much more significant in the second. As these
examples suggest, the overestimation tends to be
greater where winds are light because highly stable
conditions can occur more frequently and persist longer.
The dynamic behavior of the stable atmosphere
also presents modeling challenges. Nocturnal jets, for
example, are a significant feature of wind climates in
parts of the United States. They are caused by the
sudden decoupling of friction in the atmosphere that
occurs as night falls. That decoupling alters the balance
of frictional and Coriolis forces, causing an oscillation in
the wind vector. The rebound effect can be pronounced
at heights of 50 to 200 m – precisely the zone of
importance to large wind turbines. Because of the
nocturnal jet, the wind speed is often at a maximum at
night in this height range in many locations.
3.0
3.5
4.0
4.5
5.0
5.5
6.0
6.5
7.0
7.5
0 50 100 150 200 25
0
Hei ght (m)
Speed (m/s)
Figure 3. Theoretical vertical wind profiles for three
values of roughness length: 0.03 m (pink), 0.1 m
(yellow), and 0.4 m (blue), assuming a thermally
neutral atmosphere. Similarity theory was used, with
the assumption that the wind is unaffected by
rou
g
hness differences at hei
g
hts above 500 m.
Another class of stability-related phenomens is
mountain-valley circulations, which include katabatic
winds. Such flows up and down mountain slopes and
through valleys are driven by differential solar heating
and nocturnal cooling of the earth’s surface. In the
absence of strong synoptic forcing, the wind speed is
highly sensitive to a variety of factors, such as surface
properties that affect the rates of heating and cooling,
and the precise depth of the nocturnal boundary layer.
Part of the problem with the simulation of stable
conditions is having enough vertical resolution. Often
there is a very shallow, highly stable layer (e.g., an
inversion) which acts as an effective barrier to mixing. If
the model layers are too widely spaced, the simulated
barrier may be too weak. The coefficients of the mixing
parameterization equations in stable conditions are also
rather uncertain, as the experiments on which they are
based show a good deal of scatter.
One step we have taken to improve the simulation
of the stable atmosphere is to adopt a new method of
determining the depth of the stable boundary layer, one
based on turbulent kinetic energy (TKE). This has
generally reduced the depth of the layer, which has, in
turn, increased the shear and reduced the speed at the
surface. Based on available wind shear data, the model-
estimated nocturnal shear from 10 to 100 m is now quite
realistic in most cases. At the same time, the change
has increased the intensity of nocturnal jets and
katabatic winds, with more mixed results. In some
regions, such as the coastal mountain passes of
California, the change has been for the better, whereas
in others, it has accentuated the tendency towards
overestimation.
As a further step, TrueWind Solutions is now
engaged in a research effort funded by the California
Energy Commission to improve the simulation of winds
at a height range of 50-200 m in a variety of climates,
and in particular under the stable conditions that occur
frequently from night through morning in the California
interior. Model simulations will be coordinated with
measurements using 100 m masts, sodar profilers, and
other instruments.
7. MESOSCALE GRID SCALE
The ability of the mesoscale model to resolve major
features of the topography and surface properties (such
as coastal boundaries) is of obvious importance to
developing accurate wind climatologies. It is the reason
TrueWind has invested in a large distributed computer
processing system. While the microscale model,
WindMap, is able to adjust for simple acceleration over
small hills and ridges, it cannot account for channeling,
blocking, circulations created by thermal gradients, and
other phenomena. Unfortunately, it is difficult to
determine, a priori, the mesoscale resolution needed to
achieve a desired level of accuracy. Certainly, the more
complex the terrain, the higher the resolution should be.
Figure 5 demonstrates the effect of mesoscale grid
scale on the ability of the model to resolve channeling
through a mountain pass – in this case, San Gorgonio
Pass, a major center of wind projects that lies between
Los Angeles and Palm Springs, California. At 8 km, the
pass is barely visible, and the predicted mean speed
through it is about 7.5 m/s; at 2 km, the pass is nicely
resolved, and the predicted speed is, at maximum,
about 9 m/s. In reality, annual mean wind speeds
through the pass exceed 9.5 m/s. It is likely that an even
smaller grid spacing, such as 1 km, would be needed for
the model to develop the channeled winds to their full
intensity.
Increasing the mesoscale grid resolution poses
practical problems, however. The first is the computer
processing required. In a simulation covering a fixed
area, the number of grid cells increases inversely with
the square of the grid cell size. In addition, to control
gravity and sound waves, the model time step must be
reduced roughly in proportion to the grid cell size. Thus,
there is a cubic relationship between total run time and
grid cell size; reducing the size from 2 to 1 km requires
a factor of 8 increase in run time.
Second, as the resolution increases, non-
hydrostatic effects become more significant in the
simulations. Normally, the atmosphere is very close to
hydrostatic equilibrium. However, where rapid vertical
Figure 4. Model-generated durnal wind speed profiles
(red) at 10 m height compared with actual profiles
(green) at 6 or 10 m height, for Kansas City, Missouri
(top), and Douglas Bisbee, Arizona (bottom). The dips
in model speed at 0100 and 1300 UTC are caused by
data assimilation ever
y
12 hours.
accelerations occur – such as over a steep mountain
ridge or in a thunderstorm – the non-hydrostatic
pressure response can be significant. Running MASS in
a non-hydrostatic mode entails roughly a 50% increase
in run time. One strategy to cope with this is to allow the
model to go into non-hydrostatic mode only when the
vertical acceleration demands it. This capability has not
yet been implemented, however.
Lastly, as the model resolution increases, numerical
rounding errors between the finite elements in the
terrain-conforming grid used by MASS (and similar
models) begin to accumulate and create noise in the
simulations. We are presently exploring the limits of the
terrain-conforming grid, but expect they will become
important in complex terrain at grid scales below about
500 m. Wherever the limit occurs, a change in the
coordinate system will be required to achieve further
gains in resolution.
Ultimately, as computer power increases, TrueWind
plans on adopting a new mesoscale model with a non-
conforming grid. One option is the OMEGA model,
which was developed jointly by MESO, Inc., and
Science Applications International Corp. OMEGA does
not use a terrain-conforming grid; instead, the terrain is
represented as a lower boundary condition. The model
also uses a unique, unstructured, adaptive grid in the
horizontal dimension, which allows high resolution to be
concentrated on features of interest (such as a
mountain pass). This feature has the advantages of
improving computational efficiency and eliminating the
need for grid nesting.
9. CONCLUSIONS
Mesoscale modeling has been widely applied to
wind energy resource assessment. Although it cannot
substitute for on-site measurements, it can provide
useful information for the identification and preliminary
evaluation of wind project sites. In the course of
validating wind maps created by the MesoMap system,
a number of interesting modeling challenges have been
encountered. The first of these, surface roughness
parameterization, is being addressed through field
verification and the acquisition of better sources of land
cover data. The second, the stable boundary layer, has
already led to modifications in the formulation of the
boundary layer depth in the mesoscale model and is the
subject of a coordinated modeling and measurement
program in California. The third, mesoscale grid
resolution, is being addressed through continuing
expansion of TrueWind’s distributed computer capability
and by employing the mesoscale model in a non-
hydrostatic mode. Ultimately, however, the desire to
reduce mesoscale grid spacings below 500-1000 m will
necessitate switching to a new mesoscale model that
does not use a terrain-conforming grid.
9. ACKNOWLEDGMENTS
We would like to thank the Department of Energy's
Wind Powering America Initiative and other
organizations for their support of the wind resource
mapping and validation. We also appreciate the support
of the private wind energy meteorological consultants
who have participated in the validation effort. The
validation of the maps would not have been as
successful without their expertise and data. We would
like to thank NYSERDA and DOE for their support for
the development of the MesoMap system.
9. REFERENCES
Bowen, A.J. and N.G. Mortensen, 1996: Exploring
the limits of WAsP: the Wind Atlas Analysis and
Application Program. Proceedings of the 1996
European Union Wind Energy Conference and
Exhibition, Göteborg, Sweden, May 20-24, 584-587.
Brower, M.C., 1999: Validation of the WindMap
Model and Development of MesoMap, Proc. of
Windpower 1999, American Wind Energy Association,
Washington, DC.
Brower, M.C., B. Bailey, and J. Zack, 2001:
Figure 5. The mean wind speed through San Gorgonio
Pass as simulated at 8 km grid scale (top) and 2 km
grid scale (bottom).
Applications and Validations of the MesoMap System in
Different Climatic Regimes, Proc. of Windpower 2001,
American Wind Energy Association, Washington, DC.
Brown, J.F., Loveland, T.R., Ohlen, D.O., Zhu, Z.,
1999: The Global Land-Cover Characteristics Database:
The Users' Perspective. Photogrammetric Engineering
and Remote Sensing, v. 65, no. 9, p. 1,069-1,074.
Defries, R.S. and Los, S.O., 1999: Implications of
Land-Cover Misclassification for Parameter Estimates in
Global Land-Surface Models. Photogrammetric
Engineering and Remote Sensing, v. 65, no. 9, pp.
1083-1088.
Frank, H. P., O. Rathmann, N. G. Mortensen, and
L. Landberg, 2001: The Numerical Wind Atlas - the
KAMM/WAsP Method. Risoe National Laboratory,
Denmark.
Fuller R M and N J. Brown, 1994: A CORINE map
of Great Britain by automated means: a feasibility study.
Institute of Terrestrial Ecology, Report to the
Department of the Environment, Department of the
Environment, UK, London.
Garrat, J. R., 1992: The Atmospheric Boundary
Layer. London: Cambridge University Press, p. 290.
Hodges, J. C. F., M. A. Friedl, and A. H. Strahler,
2001: The MODIS Global Land Cover Product: New
Data Sets for Global Land Surface Parameterization.
Proc. of the Global Change Open Science Conference,
Amsterdam, 2001.
Jackson, P.S. and J.C.R. Hunt, 1975: Turbulent
wind flow over a low hill. Quart. J. Roy. Met. Soc. 101,
929-955.
Manobianco, J., J. W. Zack and G.E. Taylor, 1996:
Workstation-based real-time mesoscale modeling
designed for weather support to operations at the
Kennedy Space Center and Cape Canaveral Air Station.
Bull. Amer. Meteor. Soc., 77, 653-672. Embedded
equations are described in Zack, J., et al., 1995: MASS
Version 5.6 Reference Manual. MESO, Inc., Troy, NY.
Schwartz, M.N., 1999: Wind resource estimation
and mapping at the National Renewable Energy
Laboratory. NREL/CP-500-26245. Golden, CO: NREL.
Schwartz, M.N., and D. Elliott, 2001: Remapping of
the wind energy resource in the Midwestern United
States. NREL/AB-500-3103. Golden, CO: NREL.
Sherman, C.A., 1978: A Mass-Consistent Model for
Wind Fields over Complex Terrain, J. Applied
Meteorology, 98, 875-883.
Vogelmann, J.E., T. Sohl, P.V. Campbell, and D.M.
Shaw, 1998: Regional land cover characterization using
Landsat Thematic Mapper data and ancillary data
sources. Environmental Monitoring and Assessment 51:
415-428.
Walmsley, J.L., I. Troen, D.P. Lalas and P.J.
Mason, 1990: Surface-layer flow in complex terrain:
Comparison of models and full-scale observations.
Boundary Layer Meteorology, 52, 259-281.
... Existem dois métodos computacionais principais para o cálculo dos recursos eólicos: a implementação numérica de modelos físicos simplificados e modelos de dinâmica computacional de fluidos (CFD). Os principais programas que utilizam o primeiro dos dois métodos apresentados são o Wind Atlas Analysis and Application Program (WAsP) (modelo linear [5]) e Openwind (modelo de conservaação de massa [6,7] As incertezas dos modelos numéricos são utilizadas na estimação da incerteza padrão na estimativa de longo prazo do Produção Energética Anual (AEP). Esta incerteza padrão é utilizada no cálculo da probabilidade de superação P90, que é exigido para participação nos leilões de energia no Brasil feitos pela EPE. ...
... O Openwind utiliza um modelo de conservação de massa [6,7] [11,12]. Nestes artigos, simulações foram realizadas na mesma área e com as mesmas torres meteorológicas deste artigo. ...
Conference Paper
Full-text available
In the current work, a procedure and preliminary results for the evaluation of the numerical model uncertainty, named cross checking, will be presented. The numerical models uncertainties are used in the estimation of the standard uncertainty in the long-term Annual Energy Production (AEP) estimate. This standard uncertainty is used in the exceedance probability P90 calculation, which is a requirement of the Energy Research Company (EPE) for participation in the auctions of energy in Brazil.
... El primero fue realizado en 1984 [9]. Entre 2008 y 2009 el ICE y la Universidad de Costa Rica emplearon modelos de meso-escala, similar al estudio de [23], para elaborar mapas y archivos con datos del recurso eólico en Costa Rica y hasta 20 km dentro del mar [5], un estudio de 2011 presenta el mapa de velocidad del viento media anual [9], siendo el más reciente de 2017, cuando el Banco Interamericano de Desarrollo (BID) realiza un estudio para manejar e incorporar una mayor variedad de energías renovables en Costa Rica [10]. ...
Technical Report
Full-text available
Proyecto de investigación financiado por la Vicerrectoría de Investigación https://repositoriotec.tec.ac.cr/handle/2238/14624
... Moreover, the estimation of the roughness length can vary significantly depending on the land cover product use for the calculation [37], although it not necessarily depends on the resolution of the product [38]. Moreover, roughness assignment is not a straightforward task due to it depends on vegetation height and density, among other things, which is not depicted in the land cover resource and might affect to the vertical distribution of the wind speed [39]. Many previous studies have analysed the impact of the uncertainty on the wind resource estimation [36,38]. ...
Article
Full-text available
Renewable energies play a significant role to mitigate the impacts of climate change. In countries like Spain, there is a significant potential of wind energy production which might be a key resource. In this research, we obtain wind power at 80 m height and wind turbine energy (assuming a specific turbine). To achieve this objective we produce an optimal mapping of the hourly “instantaneous surface wind speed” (height 10 m), based on the available data. An extensive region (Granada Province, south Spain) is studied with a spatial resolution of 300 m, during a long period (1996–2016). It allows us to assess the intra- and inter-daily variability of wind energy resources. Several interpolation approaches are tested and a cross validation experiment is applied to identify the optimal approach. The obtained maps were compared with the results obtained in the stations with two common frequency distributions (Rayleigh and Weibull). This is the first time that this sensitivity integrated analysis is performed over an extensive region (12600 km²) for a long time period (20 years) at fine spatiotemporal resolution (300 m, hourly scale). The results can be very valuable for a preliminary analysis of potential optimal location of wind energies facilities.
... El primero fue realizado en 1984[9]. Entre 2008 y 2009 el ICE y la Universidad de Costa Rica emplearon modelos de meso-escala, similar al estudio de[23], para elaborar mapas y archivos con datos del recurso eólico en Costa Rica y hasta 20 km dentro del mar[5], un estudio de 2011 presenta el mapa de velocidad del viento media anual[9], siendo el más reciente de 2017, cuando el Banco Interamericano de Desarrollo (BID) realiza un estudio para manejar e incorporar una mayor variedad de energías renovables en Costa Rica[10].Considerando este panorama, se puede afirmar que Costa Rica no cuenta con estudios sobre el recurso eólico en regiones específicas del país, a partir de datos de estaciones meteorológicas. Al profundizar en este recurso, se podría revelar un potencial en alguna localidad particular, partiendo de un análisis más detallado, como el desarrollado por[11], que basado en variables como la velocidad y, dirección del viento, evidencia un gran potencial en una región colombiana.Estudios similares en Reino Unido han permitido identificar las mejores zonas para el uso de turbinas de pequeña escala[12]. ...
Article
Full-text available
La evaluación del recurso eólico es vital para el desarrollo e implementación de sistemas de aprovechamiento eólico para la generación de energía. Costa Rica no posee estudios por zonas de manera específica como se plantea, que permita a los profesionales en el área tomar decisiones en función del potencial y características del viento. Con el presente estudio se brinda por primera vez la caracterización del recurso eólico en la provincia de Cartago. La caracterización se realiza para la capa límite superficial, con datos medidos a 10 m de altura sobre el nivel del suelo tomando la magnitud de la velocidad y dirección del viento obtenidas de ocho estaciones meteorológicas. Durante la caracterización del recurso eólico se analizaron los datos de forma estadística y usando códigos computacionales que permitieron obtener resultados de la información colectada. Al procesar la información brindada por las estaciones meteorológicas se determina que los promedios de velocidades del viento se encuentran entre los 3 m/s y 5 m/s a una altura de 10 m sobre la superficie. Además, la intensidad de turbulencia se registra entre 15% y 30% y se determinaron dos periodos bien marcados en cuanto a la magnitud del viento, de noviembre a febrero para intensidades fuertes y septiembre a octubre para intensidades más bajas. Para la provincia de Cartago las direcciones que predominan son las componentes Norte o Este, siendo las dos componentes más recurrentes en los datos analizados.
... El primero fue realizado en 1984 [9]. Entre 2008 y 2009 el ICE y la Universidad de Costa Rica emplearon modelos de meso-escala, similar al estudio de [23], para elaborar mapas y archivos con datos del recurso eólico en Costa Rica y hasta 20 km dentro del mar [5], un estudio de 2011 presenta el mapa de velocidad del viento media anual [9], siendo el más reciente de 2017, cuando el Banco Interamericano de Desarrollo (BID) realiza un estudio para manejar e incorporar una mayor variedad de energías renovables en Costa Rica [10]. ...
Preprint
Full-text available
Resumen La evaluación del recurso eólico es vital para el desarrollo e implementación de sistemas de aprovechamiento eólico para la generación de energía. Costa Rica no posee estudios por zonas de manera específica como se plantea, que permita a los profesionales en el área tomar decisiones en función del potencial y características del viento. Con el presente estudio se brinda por primera vez la caracterización del recurso eólico en la provincia de Cartago. La caracterización se realiza para la capa límite superficial, con datos medidos a 10 m de altura sobre el nivel del suelo tomando la magnitud de la velocidad y dirección del viento obtenidas de ocho estaciones meteorológicas. Durante la caracterización del recurso eólico se analizaron los datos de forma estadística y usando códigos computacionales que permitieron obtener resultados de la información colectada. Al procesar la información brindada por las estaciones meteorológicas se determina que los promedios de velocidades del viento se encuentran entre los 3 m/s y 5 m/s a una altura de 10 m sobre la superficie. Además, la intensidad de turbulencia se registra entre 15% y 30% y se determinaron dos periodos bien marcados en cuanto a la magnitud del viento, de noviembre a febrero para intensidades fuertes y septiembre a octubre para intensidades más bajas. Para la provincia de Cartago las direcciones que predominan son las componentes Norte o Este, siendo las dos componentes más recurrentes en los datos analizados.
... Offshore wind energy has recently become a rapidly growing renewable energy resource worldwide triggering many projects in Europe either in developmental or in grid-connected operational stages [1]. Mesoscale modeling is one broadly used technique to evaluate a prospective wind project and lead wind resource assessment (WRA) which has made it an undoubtedly indispensable part of wind energy research [2]. The main aim of mesoscale modeling is to explicitly resolve the atmospheric equations and produce precise simulations applying the prognostic factors at the model grid points. ...
Article
Towards the improvement of the mesoscale modeling for offshore wind application, the real time observational nudging capability of the Weather Research and Forecasting (WRF) model has been implemented aiming for enhanced model performance. Utilizing three different horizontal levels of the offshore meteorological mast, FINO3, in the North Sea, wind speed observations were integrated into the model core. The performance of this modified model was then assessed for three different atmospheric stability conditions. Results from this study, illustrate that for all three stratification cases, there is a significant improvement in model performance when using observational nudging showing a reduction in Root Mean Square Error of up to 27% when compared to the observations from FINO1 platform. This study suggests that observational nudging takes a step towards more accurate simulations in wind resource assessment (WRA).
... Mesoscale numerical wind models have also been used frequently in the creation of wind resource atlases: in Bolivia (3TIER, 2009), in the United States, the Philippines and Mongolia (Brower et al., 2004), Montenegro (Burlando et al., 2009), Ireland , the Iberian peninsula (Bravo et al., 2008), offshore in New ...
Thesis
Full-text available
El viento es una fuente de energía renovable que ha tomado mucho auge en las últimas décadas. En particular en Costa Rica, alrededor del 10 % de la matriz eléctrica es energía eólica. Tomando en cuenta que este país posee una abundante vegetación, este trabajo se enfoca en determinar las características aerodinámicas óptimas de un rotor de turbina eólica de eje horizontal, de menos de 3 metros de diámetro, mediante simulación numérica, para aplicaciones en zonas boscosas de bajo potencial eólico. En particular, orientando los esfuerzos a brindar una fuente de energía a viviendas ubicadas en zonas indígenas, que están lejos de las redes de distribución eléctrica. En la primera parte de esta investigación se describe el recurso eólico en Costa Rica, aprovechando los datos de 37 estaciones meteorológicas distribuidas a lo largo y ancho del territorio; las cuales ofrecen registros entre 2007 y 2017. Se encuentra que, a 10 metros sobre el suelo, el viento sopla principalmente entre 3 y 5 m/s con una intensidad de turbulencia de hasta 30 %, estos resultados validan el foco de esta investigación, al demostrarse que el potencial eólico es bajo debido a las bajas velocidades y a la elevada intensidad de turbulencia. La segunda parte del documento contiene un modelado del viento en las zonas de interés, con la finalidad de conocer de forma detalla las condiciones en las cuales operará el rotor que se pretende diseñar. Para estos efectos se instrumentan varias torres meteorológicas con anemómetros a diferentes alturas y se determina que la temperatura, junto con la altura, son factores relevantes para describir la velocidad del viento en zonas boscosas. Se encuentra un valor de intensidad de turbulencia de alrededor del 30 % como el más frecuente entre los datos tomados. También se determina que una zona localmente sin obstáculo se asocia a una menor intensidad de turbulencia, lo que deberá ser tomado en cuenta para una eventual instalación del rotor que se diseña en esta tesis doctoral. El capítulo 3 contiene el diseño propiamente del rotor, una vez conocido el recurso eólico de manera general en el país, y de forma específica en zonas boscosas. Se emplea simulación numérica para conocer el desempeño del perfil SG6043 en condiciones de alta turbulencia y posteriormente con los resultados se diseña el rotor mediante el programa SWRDC, siglas de Small Wind turbine Rotor Design Code. El cual fue diseñado en Matlab por terceros y en este caso implementado con los datos de viento propios de la zona de interés. Se obtiene un rotor que puede generar hasta 1070 kWh de energía anual, frente a los 114 kWh que generaría un rotor de turbina comercial en las mismas condiciones. Finalmente en la cuarta parte de esta investigación se presentan los resultados de la pasantía realizada en la Universidad de Kyushu en Japón, donde se investiga el efecto un difusor en flujo turbulento. Para ello se emplea un túnel de viento con una sección de pruebas de 2 metros de alto por 3.6 metros de ancho. En dicho túnel se coloca una rejilla generadora de turbulencia y el difusor. Se encuentra que el efecto positivo del difusor en el flujo se ve potenciado por la turbulencia. Por lo que se concluye que este tipo de dispositivos son adecuados para turbinas que operan en condiciones de flujo turbulento.
Article
Muito do conhecimento que possuímos atualmente sobre o comportamento do sistema terrestre é proveniente de simulação numérica. Isso se deve ao fato de soluções analíticas para as equações de governo do sistema serem inexistentes, para casos de interesse prático, e dos dados observacionais disponíveis serem esparsos e irregularmente distribuídos no planeta. Em conjunto, essas restrições praticamente impedem investigações por meios analíticos e observacionais nos tempos atuais, obrigando os pesquisadores a recorrer à modelagem computacional. Contudo, simulações numéricas são apenas tão boas quanto o modelo matemático e os dados de entrada utilizados. Por isso, é importante compreender o processo através do qual a dinâmica atmosférica é representada computacionalmente, quais seus limites e quais as suas restrições. Este é um texto introdutório destinado a revisar conceitos importantes de simulação numérica da atmosfera. Ele começa descrevendo resumidamente os programas de computador utilizados para este fim, depois prossegue apresentando o conhecimento atual sobre a influência de grandes corpos de água sobre a baixa atmosfera, parte dele baseados em simulação. Termina realizando um estudo de caso para ilustrar os limites atuais da simulação de alta resolução aplicada ao reservatório de uma grande usina hidroelétrica brasileira.
Article
Full-text available
A unique global land-cover characteristics database developed by the U.S. Geological Survey has been available to users since mid-1997. Access to the data is through the Internet under the EROS (Earth Resources Observation Systems) Data Center's home page (http://edcwww.cr.usgs.gov/landdaac/glcc/glcc.html). Since the release of the database, the data have been incorporated into various environmental research and modeling applications, including mapping global biodiversity, mesoscale climate simulations, carbon cycle modeling, and estimating habitat destruction. Since the early stages of the project, user feedback has provided a means to understand data utility in applications, garner suggestions for data improvements, and gain insights into the technical challenges faced by users. Synthesis of user feedback provided a means to generate a user profile and derive a list of applications-critical criteria for land-cover data. User suggestions have lead to revisions in the database, including label changes, alternative classification schemes, and additional projections for the data.
Article
AN ANALYTICAL SOLUTION IS PRESENTED FOR THE FLOW OF AN ADIABATIC TURBULENT BOUNDARY LAYER ON A UNIFORMLY ROUGH SURFACE OVER A TWO-DIMENSIONAL HUMP WITH SMALL CURVATURE E.G.A LOW HILL.THE THEORY IS VALID IN THE LIMIT L/YO TO OO WHEN H/L % 1/8(YO/L)/SUP 0/.1 AND GS/L=2K/SUP 2//IN%GS/YO< WHERE L AND H ARE THE CHARACTERISTIC LENGTH AND HEIGHT OF THE HUMP, YO THE ROUGHNESS LENGTH OF THE SURFACE AND GS THE THICKNESS OF THE BOUNDARY LAYER.FOR RURAL TERRAIN, TAKING GS # 600M THESE CONDITIONS IMPLY THAT 10/SUP 2/ %L%10/SUP 4/M AND H/L%0.05.CONSIDERATIONS OF THE TURBULENT ENERGY BALANCE SUGGEST THAT THE EDDY VISCOSITY DISTRIBUTION FOR EQUILIBRIUM FLOW NEAR A WALL MAY STILL BE USED TO A GOOD APPROXIMATION TO DETERMINE THE CHANGES IN REYNOLDS STRESS.THIS RESULT IS ONLY REQUIRED IN A THIN LAYER ADJACENT TO THE SURFACE - IN THE MAIN PART OF THE BOUNDARY LAYER THE PERTURBATION STRESSES ARE SHOWN TO BE NEGLIGIBLE AND THE DISTURBANCE TO BE ALMOST IRROTATIONAL.THE THEORY SHOWS THAT FOR A LOG-PROFILE UPWIND THE INCREASE IN WIND SPEED NEAR THE SURFACE OF THE HILL IS O%H/L<UO%L<< WHERE UO%L< IS THE VELOCITY OF THE INCIDENT WIND AT A HEIGHT L.THUS THE INCREASE IN SURFACE WINDS CAN BE CONSIDERABLY GREATER THAN IS PREDICTED BY POTENTIAL FLOW THEORY BASED ON AN UPWIND VELOCITY UO%H<.IT IS ALSO FOUND THAT, AT THE POINT ABOVE THE TOP OF A LOW HILL AT WHICH THE INCREASE IN VELOCITY IS A MAXIMUM, THE VEL OCITY IS APPROXIMATELY EQUAL TO THE VELOCITY AT THE SAME ELEVATION ABOVE LEVEL GROUND UPWIND OF THE HILL.THE SURFACE STRESS IS HIGHLY SENSITIVE TO CHANGES IN THE SURFACE ELEVATION, BEING DOUBLED BY A SLOPE AS SMALL AS ONE IN FIVE.THE TURNING OF THE WIND IN THE ELKMAN LAYER MAY INDUCE A CHANGE IN DIRECTION OF THE WIND ABOVE THE HILL.THE MAIN OBJECT OF THIS ANALYSIS IS TO SHOW HOW THE CHANGES IN WIND SPEED AND SHEAR ARE RELATED TO THE SIZE AND SHAPE OF THE HILL AND TO THE ROUGHNESS OF THE SURFACE.SOME COMPARISONS ARE MADE WITH MEASUREMENTS OF THE NATURAL WIND AND WIND TUNNEL FLOWS.%FROM AUTHOR'S SUMMARY
Article
A recent increase in interest and development of wind energy in the Midwestern United States has focused the need for updating wind resource maps of this area. The wind resource assessment group at the National Renewable Energy Lab., a U.S. Department of Energy (DOE) laboratory, has produced updated high-resolution (1-km) wind resource maps for several states in this region. This abstract describes the computerized tools and methodology used by NREL to create the higher resolution maps.
Article
Land cover is a key parameter for global modeling activities related to climate and biogeochemistry. Until recently, AVHRR observations were the only viable source of data for global land cover mapping. While many useful insights have been gained from analyses based on AVHRR data, the availability of MODIS data, with dramatically improved spectral, spatial, geometric, and radiometric attributes relative to AVHRR data, provides significant new opportunities and challenges for remote sensing-based land cover mapping research. On April 1, 2001 the first official land cover product based on MODIS will be released. We describe the algorithms and databases being used to produce the MODIS global land cover parameter. This parameter provides maps of global land cover at 1-km spatial resolution using the IGBP classification system. To generate these maps, a supervised classification methodology is used that exploits a global data base of training sites interpreted from high resolution imagery in association with ancillary data. While this product has not been fully validated, and is therefore officially a "beta" release, initial assessment and validation is very encouraging. At global scales, the distribution of vegetation and land cover types is qualitatively very realistic. At regional scales, comparisons among heritage AVHRR products, TM-based classifications, and results from MODIS show that the algorithm is performing well. As a longer time series of data is added to the processing stream and the global representativeness of the site database is improved, the quality of the MODIS land cover product will improve accordingly. The quality and timeliness of land cover maps produced from MODIS data should be of high utility to the global climate and weather forecast modeling communities.
Article
The influence of rugged terrain on the accuracy of predictions by the Wind Atlas Analysis and Application Program (WAsP) is investigated using a case study of field measurements taken over 3Ω years in rugged terrain. The parameters that could cause substantial errors in a prediction are identified and discussed. In particular, the effects from extreme orog raphy are investigated. A suitable performance indicator is developed which predicts the sign and approximate magnitude of such prediction errors. This procedure allows the user to assess the consequences of using WAsP outside its operating envelope and could provide a means of correcting for rugged terrain effects.
Article
An adjustment model was developed to provide a pollutant transport model with input wind fields that are mass-consistent, three-dimensional, and also representative of the available meteorological measurements. Interpolated three-dimensional mean winds were adjusted in a weighted least-squares sense to satisfy the continuity equation within the volume specified. The upper and lateral boundaries above topography were assumed to be open air and thus allowed mass flow through the boundaries. The bottom boundary was determined by the topographic elevations of the area being studied. The topographic boundary was assumed to be solid.