Seasonal predictions of precipitation over Africa using coupled ocean-atmosphere general circulation models: skill of the ENSEMBLES project multimodel ensemble forecasts

ENSEMBLES stream 2 single-model and multimodel ensemble seasonal precipitation forecasts are evaluated over the African continent with respect to Global Precipitation Climatology Centre (GPCC) precipitation data for the 1960-2005 time period using deterministic and probabilistic skill scores. Focus is set on three regions of Africa during main precipitation seasons: West Africa during boreal summer, southern Africa during austral summer and the Greater Horn of Africa during the 'long rains' and 'short rains' transition seasons. The 45-member multimodel improves the ensemble spread-skill ratio over all regions, which translates into enhanced skill in terms of anomaly correlation and ranked probability skill scores for climatological precipitation deciles over West Africa and southern Africa. Results are contrasted depending on the region and probabilistic formulations of the ensemble predictions after a quantile-quantile calibration give valuable information essentially over areas where deterministic skill is found. Probabilistic skill scores illustrate the range of possibilities for more user-related applications of ensemble seasonal forecasts. A simple illustration using a cost-loss model shows that model potential economic values can reach over 10% depending on the regions and occurrences studied.