IWRA Proceedings

AbstractLake Chad is located in the southern edge of the Sahara Desert in north-central Africa. It is a shallow fresh water endoreic lake whose mean depth is about 4 m while its basin (2.5 106 km2) is the largest endoreic catchment area in the world. Due to its position in the Sahelian belt, Lake Chad provides a vital source of water to surrounding populations for irrigation, fishing, and trade. However, owing to its low depth and to the great variability of semi-arid climatic conditions, Lake Chad experienced several regressions and extensions in the past. The last significant variation was recorded during the last decades, with the contraction of the lacustrine surface from 25,000 to 1,500 km2 during the 1970s and 1980s. Therefore, the extreme sensitivity of Lake Chad hydro-system requires efficient and reliable hydrological models in order to investigate the lake evolution with regard to basin characteristics, climatic change and water resource management. Two distributed models have been applied. They are based on a water routing model THMB (formerly HYDRA) in which runoff and subsurface flows are routed along the drainage network except if water is stored in a depression: in this case, the (P-E) balance is applied. In the first model, IBIS+THMB, runoff is generated off-line with a Land Surface Model, IBIS, whereas in the second one, GR+THMB, runoff is directly computed by the GR2M 2-parameter production internal modules. The first parameter is the Water Holding Capacity as deduced from the FAO soil map, and the second one, C, is equivalent to an interception coefficient whose calibrated value is about 0.27. The two model simulations of Logone-Chari River monthly discharge at N’Djamena (95% of lake tributary input) are compared. Simulated discharge is generally underestimated by 35% for GR+THMB model and 26% for IBIS+THMB model whereas the Nash index is greater for GR+THMB (0.77) than for IBIS+THMB (0.55). Regionalization of C parameter improved significantly the GR+THMB model as shown by Nash index and volume deviation (0.80 and -21%). Concerning lake Chad itself, the monthly water level variability is correctly simulated as the correlation coefficients for the 2 models are about 0.85. However, these agreements hide some local discrepancies: for example, simulated Komadougou river discharge appears to be overestimated and partly compensates for Chari-Logone volume underestimation for the two models. Moreover, additional analysis revealed that lake volume is very sensitive to irrigation withdrawals and lake bottom infiltration parameterization as well as DEM quality. These points are presently under development in order to evaluate the hydrological consequences in response to future climatic change and to water resource scenarios.