This study examines the assessment of the seasonal forecasts of the North American Multi-model Ensemble (NMME) Project in Central Africa (CA) using deterministic and categorical methods focusing on the rainfall variable. This assessment is made for the June through August, March through May, and December through February seasons at 0-5 months, lead-times which are consistent with many regional climate outlooks. The precipitation observed and predicted by the NMME models has been classified into three categories (rainy, normal and dry). It can be seen that at lead 0 that the average of the multi-model set (MME) favorably represents the average seasonal rainfall in the sub-region. The Taylor diagram shows a promising result at lead0 with significant correlation coefficients greater than 85%. At lead 3, the coefficient values are low compared to lead 0. Note that the probabilities of detection (POD) of the models are more significant for the different seasons (normal). As a result, NMME models appear to be a valuable tool that can provide some key seasonal characteristics up to 5 months in advance in the sub regions, which will allow forecasters to better take into account all the uncertainties linked to natural phenomena and the state of the atmosphere.
Our abstract provides information on the evaluation of a global model for making predictions. The data used are very rich and contribute to the objectives of this SnT2021. conference in the field of atmospheric physics and dynamics.