Abstract.pdf

This study investigates the skill of hydrological drought forecasts using the high-resolution LISFLOOD model (5km grid) driven by seasonal weather forecasts from ECMWF's IFS S4. Focusing on the 2003 pan-European drought, researchers employed the Variable Threshold Method to identify drought events in forecasted soil moisture, runoff, discharge, and groundwater time series.

Key findings:

Seasonal forecasting skill: The model demonstrates high forecasting accuracy during winter and autumn (up to 6 months lead time), indicating strong potential for long-term drought predictions during these seasons. Skill declines during spring and summer, with a maximum of 2-month lead time achievable.
Variable importance: Drought forecasts utilizing variables with longer memory retention (soil moisture and groundwater) prove more accurate than those using shorter-term indicators (runoff and discharge).
* Regional variations: The model's performance varies across climate regions. Drought forecasts tend to underestimate droughts in Temperate Oceanic and Warm-summer Humid Continental climates, while overestimating them in Subarctic climates. The highest forecasting skill is observed in semi-arid or Mediterranean climates.