Estimation of evapotranspiration by spatial remote sensing in West Africa: towards a better knowledge of this key variable for the region.

Summary West Africa is particularly exposed to climatic and anthropogenic changes that exert increasing pressure on water and plant resources, the proper management of which is a major scientific challenge. In particular, it is becoming essential to better understand the energy and matter exchanges within the surface-atmosphere continuum that govern much of the hydrological cycle and vegetation development. In this respect, evapotranspiration is a key variable at the surface-atmosphere interface as it recycles most of the precipitation to the atmosphere and ensures the coupling of the water and energy cycles. In West Africa, current knowledge on this process remains limited because it is mainly based on field measurements representative of small spatial scales, or on complex surface models, whose size of the data sets they require limits their application. In this context, this thesis aims to improve our knowledge of the spatiotemporal variability of evapotranspiration, by analyzing the potential of remote sensing for its estimation in West Africa. The proposed approach is based on a comparison of available remotely sensed evapotranspiration products and the proposal of a new method allowing the generation of new products. The study was conducted on three mesoscale sites (~ 104 km2) providing a sample of the eco-climatic conditions encountered in West Africa, with from North to South: the North-Sahel (in central-eastern Mali), the South-Sahel (in southwestern Niger) and the Sudanian zone (in northern Benin). A method for estimating daily evapotranspiration by remote sensing and its epistemic uncertainty, named EVASPA S-SEBI Sahel (E3S), was developed. E3S was applied to the three study sites using data from the MODIS sensors on board the TERRA and AQUA satellites. Daily evapotranspiration estimates were evaluated against multi-year observations acquired by the AMMA-CATCH Observatory. This study highlights the potential of E3S for estimating daily evapotranspiration over West Africa. However, these estimates are still subject to the vagaries of satellite measurement (image quality, cloud cover, large viewing angle) and are therefore punctuated by gaps. This thesis also proposes new methods for reconstructing daily evapotranspiration series by combining multi-resolution and multi-source estimates. This study shows the relevance of these reconstruction approaches compared to the standard interpolation methods used in the literature. In particular, the proposed approaches allow a better transcription of the response of land surfaces to the drying sequences of the soil between two rainy episodes. The new products generated were introduced in the inter-comparison exercise including eight other products available in West Africa at various spatio-temporal resolutions. These products were evaluated at different spatial and temporal scales against local measurements and spatialized simulations of twenty surface models generated in the ALMIP2 experiment. This study highlights the high inter-product variability, especially in the Sahel. It also highlights the importance of taking into account information related to water stress in the generation of evapotranspiration products. The kilometric resolution of the E3S products provides them with an undeniable advantage regarding the description of the spatial variability of evapotranspiration fluxes compared to other low resolution products. The newly generated products have clear potential for future eco-hydrological and hydrogeological studies in the Sahel.