A diagnosis of the preseason Deep Soil Wetness (DSW) and Moist Static Energy (MSE) fields is made to define new types of predictors which could improve rainfall forecasting statistical models in West Africa. It uses the Climatic Research Unit rainfall database, the Global Soil Wetness Project dataset as well as the NCEP/NCAR model outputs, and is mainly based on correlation and composite analyses over the period 1968-1998.

The results show that : a significant positive correlation exists between observed July-September Sahelian rainfall in year 0 and September-November Guinean rainfall in year -1; positive anomalies of DSW from November to March in Guinea precede an abnormally wet rainy season over the Sahel. These DSW positive anomalies increase the MSE gradients over the continent by March-April, gradients which have been shown to be of prime importance for monsoon dynamics and associated rainfall.

These regional signals, used as predictors, allow Multiple Linear Regression and Linear Discriminant Analysis models to efficiently reproduce the West African rainfall variability over the recent period. Moreover, the real-time predictions performed for the last three West African rainy seasons (1999 to 2001) were very close to observations. The statistical models, the real-forecasts and their verifications for the 2001 rainy season are specifically detailed.