The usefulness of Extended-Range Probabilistic Forecasts for Heat wave forecasts in Europe

Abstract. Severe heat waves lasting for weeks and expanding over hundreds of kilometres in horizontal scale have many harmful impacts on health, ecosystems, societies, and economy. Under the ongoing climate change heat waves are becoming even longer and hotter, and as proactive adaptation, the development of early warning services is essential. Weather forecasts in extended range (2 weeks to 1 month) tend to indicate a higher skill in predicting warm extremes than average temperature events in Europe. We verified hindcasts of the European Centre for Medium-Range Weather Forecasts (ECMWF) in forecasting heat wave days, i.e., periods with the 5-day mean temperature being above its 90th percentile. The verification was done in 5° × 2° resolution over Europe, based on the forecast week (1 to 4 weeks). In the first forecast week, it is evident that across Europe, the accuracy of ECMWF heat wave forecasts surpasses that of a mere climatological forecast. Even into the second week, in many places in Europe, the ECMWF forecasts prove to be more reliable than their statistical counterparts. However, if we extend the forecast lead time to 3–4 weeks, predictability begins to lower to such a level that it can no longer be said, with the exception of Southeastern Europe, that the forecasts in general were statistically significantly better than the statistical forecast. Nonetheless, intense and prolonged heat waves during the third forecast weeks appear to have a higher-than-average level of predictability.