Key points are not available for this paper at this time.
The predictability of drought events in the Spree region is analyzed, aiming at developing hydrological extreme events forecast and warning systems and long-term solutions regarding sustainable, interdisciplinary and integrated water resources management in the project SpreeWasser:N. Predictors acting as potential indicators of imminent drought risk are inferred from statistical analyses, modeling and literature. Connections between certain states of the atmosphere (large-scale weather patterns) and local drought events are drawn, focussing mainly on agriculture as a user group. Special attention is paid to the succession of certain weather patterns and their impact on precipitation. A drought forecast based on k-nearest neighbor regression is being developed using an algorithm which automatically selects the meteorological variables and regions yielding the largest forecast skill as input predictor variables during a hindcast period. This machine learning approach supports the discovery of underlying physical links in atmospheric phenomena. The analysis and software development is based on ECMWF ERA5 reanalysis data and the objective weather type classification by the German Weather Service (DWD), spanning the years 1980 to 2021.
Hauke et al. (Fri,) studied this question.