Near-term efficient predictability of dry and wet years during West African monsoon season

Abstract

The near-term performance efficiency in predicting the dry and wet years in the West African monsoon (WAM) season (May to October) has been studied from 1979 to 2050 using the CRU observational rainfall, NCEP RII atmospheric circulation fields, and CORDEX-Africa outputs in the historical and RCP 8.5 experiments. The dry and wet years from the 6-month SPI at the Western Sudano Sahel (WSS), Eastern Sudano Sahel (ESS), and Guinea Coast (GC) rainfall regions, respectively, have shown consistency in the associated features during such WAM season extremes. The ensemble mean of the historical outputs (1979 to 2005) shows varying simulations of the WAM season; non-significant correlation in rainfall in GC and its overestimation at the three regions, the underestimating (overestimating) of Moisture Flux Convergence (MFC) in the Sahel (GC), and the “non-reasonable” performance by the Kling-Gupta efficiency in simulating the zonal moisture flux in GC and meridional moisture flux in ESS. The bias-correction of the RCP 8.5 outputs has shown improved performance efficiency of the models simulations from 2006 to 2021, however, with the attendant limitations in the technique. The bias-corrected rainfall showed underestimation at all regions although indicating negative significant correlation at the GC (r = -0.33, at 99.9% Confidence level from t-test) whereas the MFC has shown reasonable performance in the GC (KGE = -0.39). However, the ensemble mean of the models presents greater efficiency in projecting the WAM dry and wet years although there are yet huge uncertainties in the projections indicated by the MBE values. The 6-month SPI projections from the improved RCP 8.5 simulation present 2048 to be dry and 2035, 2042 and 2047 to be wet years during the WAM from 2022 to 2050. Noteworthy is the impact of MFC on rainfall being consistent in both the historical and the bias-corrected models’ outputs, having a greater impact by 2050.