You are on page 1of 5

MM5-WASP COUPLING FOR EYA IN COMPLEX TOPOGRAPHICAL AND METEOROLOGICAL

CONDITIONS: THE CASE OF LAKE TURKANA (KENYA) 300 MW WIND FARM

L. Claveri, M. Burlando*, F. Durante*


German Wind Energy Institute (DEWI), Wilhelmshaven, Germany
*DEWI Italy, Genoa, Italy

Summary

The Lake Turkana wind farm site is located approximately 130 km west of the town Marsabit and 15 km east of
Lake Turkana, in the region of the Chalbi desert (Kenya). This is a slightly hilly desert area, mainly characterised
by barren soil with scarce vegetation. Despite the rather simple topography of the area where the wind farm is
planned, the surroundings of the wind farm turn out to be of very high complexity both from the topographical and
meteorological point of view. The wind climate which characterises the wind farm area is indeed strongly
influenced by two factors: the synoptic-scale meteorology of the Eastern African continent, and the presence of
the orographical forcing due to the Kenya highlands.
In order to take into account these synoptic meteorological and orographical forcing, as well as the frictional
effects of the local scale topography and roughness on the wind flows within the wind farm area, a methodology
based on the coupling of the mesoscale meteorological model MM5 with the local-scale model WAsP has been
used in the present investigation. The coupling between the two models is performed following a statistical
approach.
This methodology is compared with the application of WAsP as a stand-alone model. It is shown that the MM5-
WAsP coupling is able to capture qualitatively the strong wind gradient that exists from the south-eastern to the
north-western part of the wind farm area, even if only three over ten anemometric stations are used. On the
contrary, this gradient is correctly represented by WAsP in the stand-alone configuration only if all the stations are
used. This also suggests that, in case of large wind farm areas where large-scale meteorological phenomena
could occur and strongly influence the wind climatology, it would be profitable to perform a preliminary numerical
evaluation of the main wind gradients before installing all the anemometric masts, in order to properly locate the
masts along these gradients and reduce as much as possible the number of stations.

1. Introduction zones. The low-level Turkana jet stream [3], which


blows steadily during the whole year in the low
The Lake Turkana wind farm is one of the largest elevation region of the Turkana-Marsabit Corridor,
wind farm projects to be realised in the African occurs in one of the most interesting (for wind
Continent, and the first of its kind in Kenya. After its energy exploitation) of these convergence zones.
fully commissioning in 2012, the wind farm will be The commonly adopted approaches for wind energy
generating 300 MW of clean power almost steadily yield assessment are expected to fail in this
thanks to the very peculiar characteristics of the particular context as they are thought to be applied
wind climate of north-western Kenya. over rather small wind farm areas and forced
The project consists of building 365 wind turbines through the local-scale climatology measured by
Vestas V52 of hub height 45 m and nominal power anemometric masts. On the contrary, the
850 kW. It is also included reinforcing 200 km of methodology needed to manage this kind of
roads and bridges to transport the wind turbines complexity should be able to capture the synoptic
from the Indian Ocean port of Mombasa to the meteorological and orographical forcing, as well as
north-western Kenya, and adding more than 400 km the frictional effects of the local scale topography
of transmission lines and several substations to and roughness on the wind flows within the wind
connect the wind farm and supply power to the farm area. In order to verify whether the large-scale
national electric grid. forcing is really important or not, a comparison
The wind farm area is almost unique over the world between two different methodologies to calculate
from the meteorological and geographical point of wind fields has been performed. Firstly, WAsP has
view, because the winds sweeping this region start been applied as a stand-alone model (WAsP-SA
something like 500 km far east in the Indian Ocean hereafter); then, WAsP has been coupled to the
and are channelled through the “Turkana-Marsabit meteorological model MM5 (MM5-WAsP hereafter).
Corridor” created by Ethiopian and Kenyan
highlands. This behaviour is due to the fact that the
wind climate of Kenya is mainly subject to two 2. Wind climatology of the Lake Turkana
monsoonal regimes, known as East African From a climatological point of view, Kenya lies in
Monsoons [1,2], which are responsible for the one of the most complex regions of the African
seasonal weather conditions over the country. continent. Its synoptic-scale climate is mainly
These synoptic-scale circulations interact with the influenced by the superposition of the Inter Tropical
local orography to generate some convergence Convergence Zone (ITCZ), which moves south and
north of the equator following the seasons, and the km east of the Lake Turkana in the region of the
thermal-driven circulation associated to the East Chalbi desert in Kenya, between Mount Kulal to the
African monsoons, mainly occurring over Kenya, north and Mount Nyiru to the south (both almost
Tanzania, and Uganda. A semi-permanent low 2500 m). The wind farm area extends 19 km in
pressure cell centred over Lake Victoria forces the latitude and 15 km in longitude over a slightly hilly
main air-streams of both the monsoons, which are territory, with elevations ranging between 700 m and
roughly parallel to the coastline over eastern Kenya, 900 m above the sea level.
to flow zonally westward over north-western Kenya, At the site, ten anemometric masts are now
as shown in Figure 1. This thermally-induced available. Three of theme have been installed in
synoptic-scale deflection interacts with the 2008 and have been measuring for more than two
orography of Eastern Africa to generate some years. The other seven masts have been installed
convergence zones over the Ethiopian and Kenyan more recently, less than one year ago. Figure 2
highlands. Just in between the Ethiopian and shows the position of the anemometric masts within
Kenyan highlands, the Turkana-Marsabit Corridor the wind farm area. The height of the top
occurs, which is one of the windiest among these anemometer of the masts range between 38 m and
convergence zones. 53 m, according to the hub height of the planned
wind turbines that should be 45 m.

MM367
284000

282000
980

MM264
940
280000 MM365
Latitude (m) - UTM WGS84 Zone 37

Height above sea level (m)


900

278000
MM369
860
MM368
276000
MM370 820

780
MM265
274000
740
MM263
700
272000

Fig. 1 Schematic model of the flowat the entrance of MM371 660

the Turkana-Marsabit Corridor during (a) northern 270000


MM364 620
summer and (b) northern winter (after Kinuthia and
Asnani, 1982). 268000

Firstly, [3] and [4] observed the presence of the so- 250000 252000 254000 256000 258000 260000 262000 264000
Longitude (m) - UTM WGS84 Zone 37
called Turkana low-level jet stream in this channel.
They associated the presence of this jet, which flows
steadily during both summer and winter monsoons Fig. 2 Orography of the wind farm area and position
and partially disappears only during intermediate of the ten anemometric masts: stations installed in
seasons, to the branch of the monsoons that enters 2008 (red) and in 2010 (green).
the Turkana channel and intensifies into the Turkana
easterly low-level jet. All measurements have been extended to a 3-year
More recently, Indeje et al. [5] simulated the period by means of an MCP procedure with a
dynamics of the Turkana low-level jet and medium-term anemometric station located
investigated the role of different forcing factors in approximately 25 km northwest of the wind farm
order to understand the kinematics of the jet. They area. Then, the extended data sets have been
found that the orographic forcing is the most corrected to the long-term average with
important mechanism responsible for sustaining the NCEP/NCAR Reanalysis data, in order to decrease
jet, the large-scale monsoon background flow is the influence of year-to-year wind variations on the
important in determining the wind speed and jet short-term statistics.
cores, and the depth of the channel influences the
vertical structure and location of the jet cores.
Finally, they state that the frictional forcing at the 4. Simulation setup and model coupling
surface plays a role equivalent to that of the large-
scale monsoonal background winds in the formation In the present paper two different models have been
and maintenance of the jet. used to claculate the energy yield assessment of the
Lake Turkana wind farm, namely the mesoscale
meteorological model MM5 and the diagnostic
3. Wind farm area and measurements model WAsP. Meteorological models are general-
purpose models and, in principle, they can be used
The wind farm site Lake Turkana is located for whatever problem concerning atmospheric
approximately 130 km west of the town Marsabit, 15 physics. They are not specific, however, for wind
energy applications and the maximum resolution According to [10], WAsP is mainly a wind climate
they can manage is of the order of one kilometre in model, and it is certainly the most widely used tool
their non-hydrostatic versions. For wind energy for wind energy resource assessment. In a standard
applications specific post-processing tools are project, a time series of wind measurements is firstly
needed and higher resolution are usually required. analysed to provide statistical summary of the
For this reason, MM5 has been used coupled with observed, site specific wind climate. Then, the
the local-scale model WAsP in order to achieve the analysed wind data can be converted into a wind
required resolution and post-processed output. atlas data set or regional wind climate.

4.1 Meteorological model MM5 4.3 MM5-WAsP coupling


MM5 is a numerical weather prediction model The coupling between the meteorological model and
developed by the Pennsylvania State University and WAsP is performed following a statistical approach
National Center for Atmospheric Research based on two steps. Firstly, a number of joint
(PSU/NCAR) with the ability to simulate atmospheric probability distributions of the wind speed and
conditions with resolutions ranging from 100 to 1 km. direction evenly spaced over the whole wind farm
MM5 Version 3 [6] is a non-hydrostatic, prognostic area are obtained from the mesoscale model, and
model with explicit description of pressure, calibrated with the wind measurements at the
momentum and temperature. available anemometric stations. Secondly, the local
MM5 requires a number of input data, in terms of model WAsP is applied with the aforementioned
initial and boundary conditions, to simulate the probability distributions as information for the wind
evolution of the atmosphere. These conditions climate over the wind farm area. This kind of
concern both the geometry of the computational approach allows evaluating the wind gradients
domains and the state of the atmosphere, described induced by the large-scale meteorological and
trough the independent variables of the system. topographical forcing, which could exist within the
Elevation and vegetation cover derive from different wind farm area.
datasets, appropriately re-gridded to meet the
resolution of the computational domains. The
meteorological initial and boundary conditions used
for the simulations were the National Center for
WAsP-SA (three masts)

264 264

Environmental Prediction (NCEP) Final Analyses


(FNL). This data is provided by NCEP in GRIB

MM5-WAsP
format with a resolution of 1 × 1 degrees for the
period from 2000 to present. FNL analyses are 265 265

produced trough the combination of the Global Data 263 263

Assimilation System (GDAS) and the general


circulation model AVN/GFS.
MM5 simulations have been performed over four
two-way nested domains with increasing resolution, 367

from 27 km to 1 km. MM5 setup is similar to that


used by [7]. A total number of 122 simulations have
been carried on for the period from 2009-01-01 to 264
WAsP-SA (ten masts)

365

2009-12-31. Each run consists of a hindcast of 78 369

hours, with 6 hours of spin-up period in order to 370


368

allow the model to develop high-resolution features 265

over the inner domains. Also, the model’s solution is 263

nudged towards the analysis in the outer coarser 371

domain at each time step. 364

4.2 Meteorological model


Fig. 3 Mean wind field at 45 m a.g.l. calculated
For the local-scale simulations, the computational
through three different methods: WAsP-SA (top left)
model Wind Atlas Analysis and Application Program
and MM5-WAsP (top right) forced by 3 masts, and
(WAsP), developed by the Wind Energy Division at
WAsP-SA forced by 10 masts. The color scale is
Risø DTU, has been used. This is a program for the
the same for all the maps.
vertical and horizontal extrapolation of wind climate
statistics [8]. It contains several models to describe
wind flow over different terrains and close to
5. Comparison between the mean wind fields
sheltering obstacles. WAsP utilises the linear BZ-
obtained with WAsP-SA and MM5-WAsP
model (Bessel expansion on a Zooming grid) of [9]
to calculate the wind velocity perturbations induced
Two approaches have been used in order to verify
by orographic features such as single hills or more
how important is the large-scale forcing in evaluating
complex terrain. The basic theory behind the model
the wind fields over the Lake Turkana wind farm
assumes the upstream wind speed to vary
area. The first approach is based on using WAsP
logarithmically with height, and flow perturbations
forced through the wind climate of only three over
introduced by the complex terrain are treated as
ten anemometric masts, namely the three masts
small perturbations on the basic state wind profile.
MM263, MM264, and MM265, which have been
installed in 2008 (method 1). The second approach very low and reversed with method 1. The
is based on coupling MM5 with WAsP by means of polynomial regressions of order 3, which retain more
the approach described in Section 4.3 (method 2). information from the original maps, depict also a
Also in this case, only the aforementioned three wind gradient towards south-west and north with
masts have been used. The results of these two method 2. These secondary gradients are not
methods have been compared to the results present with method 3, whereas method 1 shows
obtained using WAsP forced through the wind again a reversed gradient. It is worth noting,
climate of all the ten available anemometric masts, however, that in this case it is not possible to state
as this is considered our best estimate of the real exactly whether method 3 produces a more reliable
wind fields, at least for the areas where the results gradient than method 2 as these secondary
are not only a WAsP extrapolation (method 3). gradients are outside the area where the masts are
located. In the MM5 output, there is also some
evidence that an hydraulic jump occurs in the lee of
mount Kulal, so that the wind gradient to the north
WAsP-SA (three masts)

264 264

could really exist.

265 265 6. Conclusions


263 263

The present investigation mainly suggests that,


when a project extends over a wide area and large-
scale meteorological or topographical forcing can
occur, it could be profitable to perform a certain
367 367
number of numerical simulations and analyses
before installing the complete set of anemometric
264 264 masts. A meteorological model, indeed, could be of
WAsP-SA (ten masts)

365 365

369 369
help to identify the main wind gradients from a
370
368
370
368
qualitatively point of view, so that the positioning and
265 265
setup of the anemometric masts could be done a
263 263
posteriori in a more reliable and efficient way.
371 371

364 364

7. References

[1] Findlater J. (1977) Observational Aspects of the


Low-Level Cross-equatorial Jet Stream of the
Western Indian Ocean. Pure Appl. Geophys. 115,
MM5-WAsP (three masts)

264 264

1251–1262.
[2] Hart J.E. (1977) On the theory of the East African
Low Level Jet Stream. Pure Appl. Geophys. 115,
265 265
1263–1282.
263 263
[3] Kinuthia J.H., and G.C. Asnani (1982) A Newly
Found Jet in North Kenya (Turkana Channel). Mon.
Wea. Rev. 110, 1722–1728.
[4] Kinuthia J.H. (1992) Horizontal and Vertical
Fig. 4 Polynomial regressions of order 1 (left) and Structure of the Lake Turkana Jet. J. Appl. Meteor.
3 (right) of the mean wind field maps in Figure 4. 31, 1248–1274.
[5] Indeje M., F.H.M. Semazzi, and L. Xie (2001)
Figure 3 shows the mean wind field calculated at the Mechanistic Model Simulations of the East African
hub height of 45 m a.g.l. with the three independent Climate using NCAR Regional Climate Model:
methods. The greatest difference between methods Influence of Large-Scale Orography on the Turkana
1 and 2 is the presence of quite a strong wind Low-Level Jet. J. Climate 14, 2710–2724.
gradient from south-east to north-west in the MM5- [6] Grell G., J. Dudhia, and D. Stauffer (1994) A
WasP map. This gradient seems to be confirmed by description of the fifth-generation Penn State/NCAR
method 3, even if in this case the gradient seems mesoscale model (MM5). National Center for
lower. Atmospheric Research, Boulder, CO, tn-398+IA.
In order to have a deeper understanding of the [7] Jimenez B., F. Durante, B. Lange, T. Kreutzer,
structure of the maps shown in Figure 3, the main and J. Tambke (2007) Offshore wind resource
trends of these maps have been analysed by fitting assessment with WAsP and MM5: Comparative
the maps with a polynomial regression of order 1 study for the German Bight. Wind Energy 10, 121–
(planar surface) and of order 3 (cubic surface). The 134.
results of these regressions are shown in Figure 4. [8] Troen I., and E.L. Petersen (1989) The European
The polynomial regressions of order 1 (left) clearly Wind Atlas. Risø National Laboratories, Roskilde,
show that the wind gradient is enhanced with Denmark.
method 2 with respect to method 3, but also that it is
[9] Troen I. (1990) A high resolution spectral model
for flow in complex terrain. Proc. Workshop on
Turbulence and Diffusion, May 1990, Risø,
Denmark, American Meteorological Society, Boston.
[11] Petersen E.L., N.G. Mortensen, L. Landberg, J.
Højstrup, and H.P. Frank (1998) Wind Power
Meteorology. Part II: Siting and Models. Wind
Energy 1, 55–72.

You might also like