Evaluation of ECMWF subseasonal-to-seasonal forecast skill over Ireland for flood and drought events

Floods and droughts can occur in sequence, with one intensifying the other. For example, a drought may alter soil moisture and infiltration capacity, resulting in more surface runoff for a subsequent rainfall event. Anticipating these sequences weeks to months ahead would support water resource management, agriculture, and disaster preparedness. Ireland is particularly exposed to both Atlantic storm-driven flooding and periodic drought. However, subseasonal-to-seasonal (S2S) prediction skill is limited and depends on the variable, region, and time of year, and no study has assessed S2S forecast skill for these hazards over Ireland. S2S forecasts are usually assessed for individual variables and impacts in isolation, leaving a gap between what is verified and what actually happens when hazards occur in sequence.

This research evaluates ECMWF's sub-seasonal and seasonal reforecasts over Ireland. The two systems differ in ensemble size, model physics, resolution, and initialisation frequency. We first assess skill for individual meteorological variables against a climatological baseline. We then use case studies of flood and drought events over Ireland to assess impact skill. By comparing individual variable skill with impact skill, we determine whether useful forecast skill persists for multi-hazard events across lead times.

 

Forecasts are verified against Met Éireann station observations and ERA5-Land reanalysis using deterministic and probabilistic metrics including RMSE, ACC, BSS, and CRPS. Skill is evaluated as weekly means of daily data at lead times of 3 to 10 weeks for temperature, precipitation, mean sea level pressure, and wind. This research identifies the forecast windows where useful skill exists for downstream multi-hazard analysis.