The prediction errors of WAsP are analysed by use of five 50 m measuring masts and the differences in the RIX-values in a mountainous site of wind energy development at the Norwegian coast. The RIX-values range from 13% to 18% thus the terrain is quite complex. High correlation (0.96) is found between the prediction error of WAsP and the differences in the RIX-values. The inclusion of a coastal background station contributes strongly to this correlation. Predictions errors of 10%–15% are found based on the background coastal site. One should therefore be careful in applying background meteorological stations to wind resource assessment at the coastal mountains. Adjustments of the wind speed predictions based on the present findings are recommended. For the measuring stations within the wind farm area only a rather high correlation (0.76) between prediction errors and differences in the RIX-values are also encountered. Prediction errors are in the range 0%–6%, which is considered quite low taking the complexity of the terrain into account. Copyright © 2006 John Wiley & Sons, Ltd.