A Short-Range to Early-Medium-Range Ensemble Prediction System for the European Area

Abstract

An ensemble prediction system, especially designed for the short to early-medium range for the European domain, is presented. The initial perturbations of each ensemble are based on singular vectors that maximize the 3-day total energy error growth above the European area and the northern Atlantic. In total, a set of 51 ensembles, each consisting of 51 members, has been integrated, comprising 28 winter cases in 1996 and 1997, 2 spring, and 21 autumn cases in 1997. The impact on performance, relative to the operational ensemble prediction system at ECMWF, appears to be modest for the winter set. However, for the combined spring and autumn set, the impact is significantly positive, especially for rare (extreme) events of large-scale precipitation, 2-m temperature, 10-m wind speed, and the pressure at mean sea level. For the targeted ensembles, the situation that a complete ensemble misses an extreme event occurs much less than for the operational ensemble system. As a result, the range of cost?loss ratios for which a user has benefit is larger. The benefit of the targeted ensembles is maximal between days 2 and 3. For this range in spring?autumn, an ensemble system consisting of a subset of 21 targeted members performs comparable to the 51-member operational ensemble prediction system (EPS), when evaluated for the same area of interest. The spread within a targeted ensemble prediction system (TEPS) is somewhat larger than within the EPS ensembles. However, both systems appear to be underdispersive for large-scale precipitation, 2-m temperature, and 10-m wind speed. Only for mean sea level pressure is the spread of the TEPS ensemble, on average, comparable to the magnitude of the forecast error of the ensemble mean between days 2 and 3.5.