North Sea wind climatology Part 1: a review of existing wind atlases

The national government needs high quality offshore wind climatology at and around hub height to be able to determine a realistic wind power potential for the North Sea and to be able to assess whether the yields predicted by wind energy companies are reliable. One “reference” wind atlas makes procedures more transparent for all stages in the process of establishing Dutch offshore wind energy: allocation of areas, tendering procedure, allocation of wind energy producer and monitoring of the wind energy yields. It will also save time and money as individual wind energy companies would not have to make their own wind climatologies. It is however essential that the quality of this “reference” wind atlas is scientifically sound and that it has the confidence of the wind energy sector and the banks ( so they would be willing to lend the required funds).

Wind energy developers determine wind climatology at the site where they want to build the wind turbine or wind farm (target site) by transforming long series of near surface wind measurements at a nearby reference site, if available, to hub height at the target site. The transformation requires simultaneous measurements at both sites, but the measurement campaigns at the target site are relatively short. This is called the Measure Correlate Predict (MCP) method. The main problem with using measurements as reference data is that they are predominantly done at levels below hub height. To bridge the height difference between the reference and target measurements, assumptions have to made on the atmospheric stability and the associated vertical wind profile. The temperature profiles required for deriving the actual atmospheric stability are only measured at a few wind masts and for a limited period.

Wind atlases are used for siting purposes. This report gives an overview of the main wind atlases used in the wind energy sector and explains their limitations:

• The existing wind atlases do not include the long-term variability of the wind climate because they are based on short measurement campaigns or short periods of weather forecasting model output. The European Wind Atlas, the OWA-NEEZ wind atlas and the NORSEWInD (based on SAR and ASCAT satellite wind measurements) wind atlases are based on relatively calm periods, the KEMA/SenterNovem Wind Atlas on a relatively windy period. The only wind atlas based on a period in which most of the long-term wind speed variability is captured is the NORSEWInD QuikSCAT map that is based on satellite observations from 1999-2009, but sampling1 and quality control2 errors in the map are not yet accounted for. This is because the NORSEWInD project lacked the financial resources to perform a collocation procedure.

• Most existing wind atlases do not provide information above 100 m (only the European Wind Atlas gives information up to a level of 200m, but the information above 200 m is of limited value).

• As most existing wind atlases are based on measurements at 10 m above the ground or at the sea surface (satellite measurements), these observations have to be extrapolated to higher levels and assumptions have to be made on the influence of atmospheric stability on the wind profile.

o WAsP based atlases (European Wind Atlases and KEMA/SenterNovem Wind Atlas): both the extrapolation formula (Businger-Dyer) and the empirical function (stability function Ψm) that is used to describe the effect of stability on the wind profile have been improved since they were used in the WAsP-based wind atlases. The effect of the height of the planetary boundary layer should e.g. be included in the extrapolation formula (adapted Businger-Dyer). Furthermore, only a climatological average of the non-neutral stability effects is used, which means that hour to hour stability effects on the wind profile can not be reproduced and time series of the wind speed at hub height can not be estimated accurately. Finally, there is only one climatological average wind profile for land and one for sea, whereas in reality it varies from location to location. In contrast, wind data from weather forecasting models do not suffer from either limitation.

o Satellite based wind atlases: SAR3 wind speeds from the Envisat satellite have been lifted to 100 m, using temperature and heat flux (which determine stability) from a numerical weather forecasting model (WRF4 ), but the results underestimated the measured wind speed even more than the WRF model wind speeds did. In the NORSEWInD project, wind mast and LIDAR data have been used to better understand wind profiles. Incorporating this new knowledge has led to better estimates of the 100 m wind speeds but the quality of the results is now limited by the quality of the planetary boundary layer height and the sea surface temperature in WRF.

• The OWA-NEEZ wind atlas is based on different versions of the Hirlam model. Earlier versions are in general less accurate than the later versions which may introduce changes in the wind which are not real. Because the atlas is not based on a homogeneous data set, it cannot be used to derive trends or make time-series.

So all existing wind atlases are based on measurements (except for the OWA-NEEZ wind atlas which is based on an inhomogeneous model data set) and periods that are not long enough to be sure that the long-term variability of the wind has been captured.

Based on the findings presented in this report, we recommend making a high resolution wind atlas for the North Sea that includes information on the long-term variability of the wind climate, provides information up to at least 200 m height and can reproduce more than just a climatological average of stability effects on the wind profile. In part 2 of this report we present a wind climatology which meets all these requirements. It is based on reanalyses model ERA-Interim (to ensure that long-term variability is included) and operational atmospheric weather forecasting model Harmonie (to enhance the resolution). At this stage it is no more (or less) than model climatology and validating against observations is still needed. Therefore validation will be the main recommendation for part 2 of this report.