Impacts of precipitation events and land-use changes on West African river discharges during the years 1951–2000

The aim of this article is to determine the main sources of the variability of large West African river discharges during the years 1951–2000. The analysis is based on river discharge simulations of ORCHIDEE Land-Surface Model forced by NCC (NCEP Corrected by CRU) over this period. In a first part, an analysis of the partitionning of precipitation in the different basins is given. It is first shown that total runoff is more variable and better correlated to precipitation than evapotranspiration over every basin on annual means. Then the more complex link between evapotranspiration and precipitation is investigated through correlation and regression analyses. Over the “dry” (soudano-sahelian) catchments, evapotranspiration is the most variable and mainly correlated to the annual precipitation. Over the “humid” (equatorial) catchment, it is mainly correlated to the season length and does not depend on other characteristics of the precipitation. Over the “intermediate” (guinean) catchments, annual precipitation and season length both play a role, along with the average intensity of rainfall events, and other characteristics such as the number of long breaks in the rainy season. A second part focuses on the way ORCHIDEE reproduces the variability of river discharges for the years 1951–2000. It is shown that relative anomalies are correctly simulated by ORCHIDEE forced by NCC over every catchment without taking into account any land-use change. Moreover, no significant difference is found in the accuracy with which ORCHIDEE simulates the humid (1953–1970) and the dry (1973–1990) periods over two selected catchments. This implies that the impact of land-use changes was much less important than the impact of precipitation changes over the years 1951–2000 in large West African catchments.