|
Abstract
|
A system has been developed to give probabilistic warnings of severe-weather events for the United Kingdom (UK) on a regional and national basis, based on forecast output from the European Centre for Medium-Range Weather Forecasts (ECMWF) Ensemble Prediction System (EPS). The First-Guess Early Warnings (FGEW) project aims to give guidance to operational forecasters, to help them give earlier warning of severe weather in support of the UK National Severe Weather Warning Service (NSWWS).
Calibration was applied to the EPS model output to optimize the probabilistic early warnings over an initial training period of one winter season, and the resulting warnings were then verified over a 16-month period spanning two winter seasons. The skill of warnings from several versions of FGEW is assessed using a range of probabilistic skill scores, and is also compared with that of warnings issued by forecasters. Results show that the system is capable of providing useful warnings 3-4 days ahead with some probabilistic skill. Most of the skill is attributable to warnings issued at low probabilities, but when higher probabilities do occur, this provides a valuable signal that has been used by forecasters on a number of occasions to issue warnings earlier than was done previously.
Maximum skill of the FGEW warnings is found at a lead time of 4 days, with virtually no skill at shorter lead times of 1 or 2 days. This behavior is found to also occur in equivalent deterministic forecasts and so is not attributable to the ensemble perturbation strategy. Nevertheless it is suggested that while the EPS perturbations work well for the medium range, alternative perturbation strategies may be required for successful short-range ensemble prediction. |