Assessing the skill of European seasonal forecast models for extreme climate events over Africa

Abstract:

Extreme climate events such as heavy rains, floods, heatwaves and droughts have a large societal impact if not appropriately monitored and if adaptation measures are not developed and modified based on early warning of these events. Currently, there are only a few global seasonal climate forecast models available with a high temporal (e.g., sub-daily) but coarse spatial resolution (~1°) that can provide early warning operationally. In this study, we assessed the skill of the five European seasonal forecast models available as part of the Copernicus Climate Change Service over the African continent: European Centre for Medium-Range Weather Forecasts (ECMWF), UK Met Office, Météo-France, Deutscher Wetterdienst (DWD), and Centro Euro-Mediterraneo sui Cambiamenti Climatici (CMCC). The Multi-Source Weighted-Ensemble Precipitation (MSWEP) and Princeton Global Forcings (PGF) gridded products were used as reference observational datasets for precipitation, and maximum/minimum temperature, respectively. Multiple statistical measures such as anomaly correlation and root mean square error were used to evaluate the skill of individual models and their unweighted and skill-weighted multi-model ensembles on daily, monthly, seasonal, and climatological time scales. Forecasts of drought were evaluated based on the Standardised Precipitation Index (SPI) and forecasts of other climate extremes were based on the ETCCDI extreme indices (e.g., heavy rainy days and hot days) using probabilistic skill measures. The results generally show useful skill only at lead-1 month, with regional and seasonal differences across models, suggesting the use of a skill-weighted multi-model ensemble. The weighted ensemble forecast shows significant improvement in skill relative to individual models and the unweighted ensemble for some regions, suggesting potential for operational use to provide early warning for a range of climate extremes.