In the last 50 years of the 20th century, West Africa suffered one of the highest observed rainfall variability over the globe. These climate fluctuations have not only modified the characteristics of precipitation but also the entire continental water cycle and have been associated with important land-use and land-cover change. We focus here on the uncertainty linked to simulating recent and future climate changes in this region.

In a first part the future anthropogenic climate change impact on precipitation in West Africa is studied among different coupled climate models. A simple and reproducible measure of the rainy season is proposed using monthly precipitation values. The rainfall band over West Africa is characterized in every model output by its center, width and intensity. It is shown that the most recent climate models (CMIP3) provide control simulations closer to observations than old ones (CMIP2). However, the agreement on the impact of climate change remains still low. Only a weak consensus exists on a narrower but more intense rainfall band and on a slower retreat of the monsoon. Different sources of uncertainty are highlighted through sensitivity experiments. Finally, no link is found between the quality of control simulations and the agreement on the impact of climate change.

Secondly, developments in the Land-Surface Model ORCHIDEE are made to improve the representation of the West African water cycle in offline simulations. They allow for the identification of different sources of uncertainty in the simulation of mean river flows. The precipitation input remains the main uncertainty, owing to the lack of numerous high-quality observations in the region. Vegetation and soil parameters also play a role even if less important. Then key processes are identified for three sub-regions. In Sahel, the main sources of uncertainty are related to the “granularity” of vegetation, which affects the soil evaporation and to the reinfiltration of runoff through ponds in flat areas. In the Guinean region, vegetation has deeper roots and the soil is finer, therefore root-zone infiltration is the key process because it controls a large part of the evapotranspiration. In the equatorial region, evapotranspiration is more limited by the atmospheric demand than by soil moisture and the energy balance at the leaf level is the most important parameterization there. Finally, the role of floodplains is emphasized for large river catchments such as Niger at Niamey or Congo at Kinshasa.

Lastly, the representation of African river discharges by ORCHIDEE is evaluated over the last 50 years. The role of individual rainfall events in the computation of annual evaporation is highlighted for different river basins. In addition to the annual precipitation, the length of the season, the number of dry-spells and the average intensity of rainfall events have a significant impact on the annual evaporation at the basin scale, and thus on annual river discharges for several catchments. ORCHIDEE simulations for the humid (1951-70) and the dry (1971-90) periods are analysed. Comparison with the observations show that ORCHIDEE correctly represents the shift in river discharges around 1971-2 without taking into account any change of land-cover or land-use. This implies that the impact of land-use and land-cover change on river discharges was not significant compared to the impact of precipitation changes in large Guineo-sahelian catchments (more than 50000 km^2) during the years 1950-90.

More information is available at www.lmd.jussieu.fr/~tdolmd