Bayesian Rainfall Variability Analysis in West Africa along Cross Sections in Space–Time Grid Boxes

Abstract

This paper proposes an approach for analyzing rainfall variability over West Africa during the 1950?90 period. Three grid boxes, corresponding to three selected areas over West Africa, have been constructed. For each candidate area the set of annual grid maps are stored in 3D matrices, reflecting time and geographical position, called here space?time grid boxes. Each space?time grid box contains grid points corresponding to a given gauging year. The Bayesian procedure, based on a single-shifting model, is applied to grid points extracted from mean meridional and latitudinal cross sections of each space?time grid box. Two different problems are considered: the first is the detection of a change, while the second is the estimation of the changepoint and its amplitude under the assumption that a change has occurred. The Bayesian single-shift model is applied on grid points extracted from each cross section. A latitude?latitude and longitude?longitude analysis of the rainfall climatology changes is, thus, carried out. It is pointed out that the most significant rainfall climatological changes in the Sahel most probably occurred between 1965 and 1970 with the decrease of the mean level of annual rainfall. This deficit is very high over the coastal region of Senegal (25%) and over the central region of the Sahel (15%?20%). Under approximately 9°?10°N, over the humid West Africa region, a zone without any significant change extending from 6° to 10°N was highlighted. A similar zone with nonsignificant rainfall change was identified along the cross section at 1.5°E, which follow the border of Togo and Benin. However, over the zones in edge of the coast of Ivory Coast, a deficit about 17% is observed.