Multi-scales analyses of the West African Monsoon water cycle with GPS observations.

Abstract

The West African Monsoon (WAM) regional climatic system is characterized by a strong seasonal cycle in humidity and precipitation and land-surface atmosphere interactions.The water cycle is a primordial element of the WAM, the study of which is a ma jor ob jective of the AMMA campaign (African Monsoon Multidisciplinary Analysis). Within this context, six GPS stations have been installed along the meridian climatic gradient in West Africa, beginning in 2005. This dissertation focuses on the analysis of atmospheric humidity through GPS data, while also proposing a more in-depth study of atmospheric water budgets based on different data sets resulting from the AMMA campaign. The analysis of temporal GPS series shows a spatial and temporal variability of water vapor (seasonal cycle, inter-seasonal variability with 10-20 day modes, synoptic scales and diurnal cycle) that is shown to be linked with the WAM atmospheric processes. The precision of this data also enables the identification of humidity biases in radiosounding data and through assimilation processes in meteorological prevision systems. A method is then developed to compute water budgets vertically integrated at regional scale based on a "hybrid" data set. This method combines satellite precipitation estimations, simulations from a grouping of land surface models, all of which are forced by the same precipitation and other elaborate products. The analysis of the functioning of the hydrological cycle at the surface-atmosphere interface allows us to confirm or not certain hypotheses that have been developed in the past, but especially to shed new light on the seasonal cycle and the interannual variability of water budgets and surface energy. We focus in particular on the coupling between the terms of these budgets (moisture convergence, precipitation, evaporation, runoff, soil moisture, and surface net radiation), which differ according to the scale in question. This dataset is then used as a r