| description abstract | his study presents a method that improves extreme windstorm early warning in regards to past events that hit France during the last 30 years. From a 21-member ensemble forecast, the extreme forecast index (EFI) and the shift of tails (SOT) are used to produce calibrated forecasts for a selection of 59 windstorm cases. The EFI and SOT forecasts are evaluated for windstorms of different levels of severity and for various forecast index thresholds using the Heidke skill score (HSS), hit rate (HR), and false alarm rate (FA). The HR and FA show that a ?zero misses? level always goes conjointly with a high level of false alarms. The HSS shows maxima that are associated with EFI (or SOT) thresholds that could be used as a rationale for decision-makers to issue warnings. For most extreme events, it is found that a higher level of HR can be achieved using the SOT rather than the EFI. Overall, most of the windstorms are well anticipated 3?4 days ahead. To facilitate the use of EFI or SOT forecasts, it is suggested that extra information in the form of conditional probabilities be added, hence linking the EFI (or SOT) values to a risk of occurrence of a severe event. Finally, this anticipation of extreme events is illustrated by maps of EFI and SOT for four historical windstorms. | |
