We evaluate different methodological choices for seasonal drought prediction over the Mediterranean region with the multi-dimensional Standardised Evapotranspiration Precipitation Index accumulated over a 3-month time-scale (SPEI-3), based on the ECMWF SEAS5.1 operational prediction system. We analyse two strategies for constructing the index backfilling data prior to model initialization, using real-time quasi-observations from the ERA5 reanalysis (SPEI-3-R), or model data from previous initializations of the same prediction system (SPEI3-M), and show that model skill is sensitive to these methodological choices. The long 42-year hindcast/prediction record available for this model (1981–2022) allows for a robust skill assessment. A window of significant skill, extending from May to October, is detected over the Iberian Peninsula. This window arises from the cumulative and multivariate nature of the index and cannot entirely be explained by the individual skill of the components, nor by the warming trend during the validation period. Based on these results, seasonal drought predictions relying on the SPEI are currently being enabled in the framework of a new generation of climate services developed in Spain. These go beyond alternative applications available to-date, which usually rely on simpler indices and/or shorter model verification periods.

This research work was funded by the Spanish Ministry for Ecological Transition and Demographic Challenge (MITECO) and the European Commission NextGenerationEU (Regulation EU 2020/2094), through CSIC’s Interdisciplinary Thematic Platform Clima (PTI-Clima). We also acknowledge the support of ECMWF’s Copernicus Climate Change Service national collaboration programme, under contract C3S2_461-1_ES_CSIC.