Climatic Research Unit

Regional case studies

Statistical modelling of North Sea winds

High-resolution information about changes in marine winds is needed for offshore wind farms and other marine constructions ­ but fitting of statistical models is difficult because of the lack of marine observations. Thus a new approach has been developed where "pseudo observations" from the ENSEMBLES ERA-40 forced runs are used in a two step downscaling approach.

V-wind speed component for January 1961-1990 used in a two-step statistical downscaling approach applied to the ECHO-G GCM.

Climate variability and its possible changes are important for marine regions, especially if one thinks about offshore wind farms and other marine constructions. To investigate climate variability on highly spatial and temporal scales under different global climate conditions two approaches are commonly used. One is to run regional climate models, which need much computational resources, or the other is to use statistical downscaling methods. But for marine regions only few or no observations with long enough time series are available to fit such statistical models.

In this study the output of regional climate models within the ENSEMBLES project are used as "pseudo observations" in a two step downscaling approach to investigate marine surface winds. The model approach consist of a multi linear regression (MLR) model for spatial downscaling and a multi variate autoregression (mvAr) model to generate highly temporal time series of wind components.

Validation of this method show good correlation of wind speed and direction (see wind rose figures below). However in the presented method the regional scale climate variability under different climate conditions depends strongly on today's variability. In future studies the regional scale variability should be related to the large scale variability under different climate conditions.

10m wind components (u & v) from the two step statistical downscaling approach applied to a last millennium transient simulation (ERIK 1) from the general circulation model ECHO-G. Column A shows mean January wind speed for the period 1961-1990. Column B show the difference between period 1751-1780 and 1961-1990 (A). Column C show the difference between 1911-1940 and 1961-1990(A). Red (blue) colours in column B and C indicate higher (lower) wind speed compared to the period 1961-1990.

Groll, Nikolaus:"A two step statistical downscaling approach for marine surface winds", in preparation.

 

return to top