| Sumario: | This paper aims to find best predictors for rainfall statistics over Western Africa. We investigated skill level for rainfall total and number of rainy days over a 0.8-degree gridded resolution over five season windows from March to September: March-April-May (MAM), April-May-June (AMJ), May-June-July (MJJ), June-July-August (JJA) and July-August-September (JAS). The NMME data set, which is available in the International Research Institute for Climate and Society (IRI) Data Library was used as the predictors. Results presented here aim to find the best downscaled forecast using both the recent climate model produced by climate community and also high resolution gridded data produced by the Enhancing National Climate Services (ENACTS) platform. Among North American Multi-Model Ensemble (NMME) parameters, soil moisture from CMC1-CanCM4 offers good predictive skill for early season, and the GFDL-CM2p5-FLOR-B01 rainfall captures rainfall during the June to September period very well. The transition season when the Inter Tropical Convergence Zone (ITCZ) is moving from South (Gulf of Guinea) to the North (Sahel) is the most challenging period to predict (April, May, June, and July). For each of the three-month seasons from March to September, a statistical model is proposed with the best skill associated. The skill found in these exercises are the highest skill found among previous efforts for downscaled rainfall over Sahel. We present all skills and the model approach used. Finally, an example of real time forecast is presented for the March-April-May (MAM) season and April-Mary-June (AMJ) season in 2018. Further suggestions for improvement and practical usage is also discussed.
The overall objective is to improve the outlook seasonal forecasts (Prévisions Climatiques Saisonnières en Afrique Soudano-Sahélien [PRESASS] and Prévisions Climatiques Saisonnières pour les pays du Golfe de Guinée [PRESAGG]) in spatial resolution (downscaling) as well as in skill.
This work was done in collaboration with AGRHYMET.
|
|---|