Effect of model resolution on tropical seasonal rainfall predictions

Seasonal predictions of the African monsoon were performed using three different versions of the NCEP climate forecast system (CFS). The model integrations are the CFS-T62, CFS-T126, and the CFS-T382 for both April and July IC. NCEP reanalysis is used for control of the flow patterns. The CPC PREC/L and the high resolution satellite rainfall estimates (RFE) are used for control of the precipitation. All three models depict the summer time African rainfall band stretching from Ethiopia to the Atlantic coast. The three areas of maximum precipitation located in the Ethiopian Highlands, the Cameroon Mountains, and western Guinea are captured with striking differences. Rainfall is much higher in those regions in the coarser resolution of the CFS-T62. The higher resolution model, the CFS-T382 appears to be much more realistic. To evaluate the degree to which the CFS captures the precipitation interannual and decadal variability of the West African monsoon system, the Sahel rainfall increase of the past 10 years and the physical processes associated with it, are examined. SST anomalies in the Atlantic and the precipitation response are analyzed for both April and July ICs. Both CFS T-382 April IC and CFSv2 April IC captured reasonably well the warming trend in the tropical north Atlantic and the Sahel rainfall recovery. The CFSv1 did not. However, for July IC, all models represented the SST increase in the tropical north Atlantic. While CFS-T382 precipitation anomalies remained close to the observations, CFSv1, Jul IC was clearly an improvement over the April IC run. However, CFS-T126 limited the Sahel rainfall increase to the southern part of the Sahel. The dynamics associated with Sahel rainfall variability for both April and Jul IC are presented. During the past decade, the continental thermal heat low, the LLWJ, the AEJ, and the TEJ were quite prominent. These enhanced features of the West African monsoon system were more reasonably depicted in the CFS T-382, especially for Jul IC, than in any of the other models. Forecast skill scores are examined and are consistently higher in CFS-T382 for both April and Jul IC than in the other models. The CFS-T62 exhibited lower skills for April IC than the T126, however caught up with it for Jul IC. Overall, the CFS-T382 represents clearly an improvement over the CFST-62 and the CFS-T126.