Local-scale wind speed features captured by seasonal forecasts

Thanks to the recent advances in climate modelling, seasonal predictions are becoming more skilful at anticipating the future state of near-surface climate variables over extratropics. Nevertheless, such predictions are delivered on too coarse grids with horizontal resolutions of hundreds of kilometres so that local events happening at much finer scales cannot be reproduced. This is particularly noted for variables with high spatial variability like wind or precipitation: wind speeds can vary substantially over a few kilometres, from the top of a mountain to a valley floor. The differences in magnitude might be relevant for the deriving sectoral indicators, for example, within the wind industry and at a wind farm level.

This work presents and applies a downscaling methodology to generate fine-scale seasonal forecasts ---up to station scale--- for near-surface wind speeds in Europe. The hybrid forecasts are based on a statistical downscaling with a perfect prognosis approach, fitting a multi-linear regression with the four main Euro-Atlantic Teleconnections (EATC) indices as predictors. Seasonal predictions of EATC indices, which are predictable with relatively good skill levels, are later inserted into the multi-linear model. This results in skilful seasonal predictions of surface wind speeds. Indeed, the comparison of the hybrid forecasts against the dynamical forecasts of wind speed shows that the skill of such forecasts is not only maintained but also increased over most of Europe. The hybrid forecasts are generated at 17 locations where tall tower wind speed data are available and at a pan-European scale using the 100-metre wind speeds from the ERA5 reanalysis. Improving the accuracy of seasonal predictions is an essential step to inform weather-and-climate-vulnerable socio-economic sectors of seasonal anomalies a few months ahead.