Performance evaluation of seasonal precipitation forecasting using NMME over West Africa

Abstract

Abstract In West Africa (WA) in recent years, users in agriculture, hydrology, disaster risk management, health and many other sectors have expressed a growing demand for high-resolution seasonal forecasts with a sufficient lead time for response planning. We present a regional evaluation of precipitation forecasts from 14 North American Multi-Model Ensemble (NMME) seasonal forecast models, using Global precipitation Climatology Centre (GPCC) and Africain Rainfall Climatology Version 2 (ARC2) as a reference over the June-September (JJAS) season. We first assessed the quality of the forecasts in reproducing the climatology, then the quality of each individual model as well as the ensemble model in predicting the quality of forecasts in WA at a 0–5 month lead time. The results show that NMME models capture the seasonal rainfall climatology of the JJAS season over the central, and southeastern parts of WA around 11 mm/day. We found that, in most cases, precipitation skill was highest during the first lead time (i.e. the initialization month's forecasts) and declined rapidly thereafter. The performance of each model in forecasting seasonal precipitation for the JJAS season shows a high ability to predict normal conditions, and appears not to predict below- and above-normal seasons well (Probability of Detection (POD) of less than 40%). Overall, the performance of the NMME ensemble mean (MME) was not consistently better than that of a single individual model, underlining the need for more advanced weight-based averaging schemes. The NMME forecasting system offers a promising skill set for forecasting seasonal precipitation over WA during the JJAS season at first lead time.