A post-processing method applied to simulated wind speeds in Romania

Abstract

The quality of a numerical weather forecast is influenced by uncertainties that may arise from the initial conditions or the design of the model itself. Specifically, our evaluation of a limited area model wind speed forecasts over Romania shows that forecast errors are significant in areas with complex topography, or other local characteristics. The purpose of this paper is to reduce these errors obtained from a deterministic model forecast. A bias correction method that post-processes the model output was implemented in order to improve the accuracy of the wind speed prediction. The new forecast was evaluated using standard statistical scores, for winter and summer seasons of the year 2016, at six stations in Romania located in areas with different topography. The corrected forecast was also compared to that obtained from a simple linear regression model using both simulated and observed wind speed data. The results show that the application of such methods generally leads to an improvement of the wind speed forecast. Furthermore, the reduction in the forecast error depends both on the particularities of the area and the period under review.