Performances of climatic indicators from seasonal forecasts for ecosystem management: The case of Central Europe and the Mediterranean.

• Seasonal forecasts performed well over some locations in the Euro-Mediterranean. • Multi-model forecast are encouraged to increase the extent of skillful predictions. • Quantile mapping is the most preferable bias correction technique. Climate change is affecting the interannual variability and the seasonal cycle for key meteorological variables, as well as the frequency, intensity, duration and timing of out-of-the-norm and extreme events. In such a context, seasonal forecasts (i.e., climate predictions from a few weeks to several months ahead) are compelling instruments able to anticipate upcoming climate risks and guide tactical decisions. In this study, we assess the performance of seasonal forecasts for the summer period across Central Europe and the Mediterranean region in predicting some key indicators serving agriculture and forestry sectors, i.e., Potential EvapoTranspiration (PET), Potential Soil Moisture Deficit (PSMD), and Fire Weather Index (FWI). We exploit authoritative climate data from Copernicus Climate Data Store, i.e., ERA5 reanalyses and hindcasts from CMCC SPSv3 and ECMWF SEAS5 seasonal prediction systems (SPSs), to evaluate, using both deterministic and probabilistic evaluation scores, differences in performance across the two SPSs, two start dates (March and May) and four correction techniques applied to overcome modelling bias, namely bias correction (BC), calibration (CAL), quantile mapping (QM) and detrending (DET). Results show that seasonal predictions of PET perform better in Western and Eastern Europe and some areas of North Africa. PSMD predictions follow a similar spatial pattern as PET, except that for some areas in Central (Eastern) Europe in which the performance increases (decreases). FWI predictions reveal better results in some areas of the Iberian Peninsula, North-Western Africa, Balkan Peninsula, and Ukraine. Results also suggest that QM might be the most suitable technique for bias correction. Furthermore, the start date of the forecast might imply varying correlation significance, with the start date closest to the forecasted period not always being the best. Overall, our study suggests the potential usefulness of seasonal forecasts for decision making under different geographical and environmental contexts, considering sensitivity inherent to different processing chains. [ABSTRACT FROM AUTHOR]