A remote sensing data fusion method for continuous daily evapotranspiration mapping at kilometric scale in Sahelian areas

International audience Thermal infrared (TIR) remote sensing-based ET estimates are very appealing for a wide range of scientific applications in hydrology. However, they are prone to infrequency due to satellite revisit interval and cloud cover. Temporal interpolation techniques or multi-resolution and multi-frequency data fusion approach have thus recently been studied to provide continuous ET estimates. It has been already shown that the interest of temporal interpolation techniques is limited to short return interval that is troublesome in Sahelian regions where mesoscale convective systems linked to the West African Monsoon (WAM) renders unusable most of TIR images during the rainy season. Here we developed a data fusion approach to provide remote sensing-based continuous daily ET estimates at kilometric resolution in Sahelian areas. The proposed algorithm fill gaps in MODIS-based ET estimates from EVASPA S-SEBI Sahel (E3S) using the Global Land Evaporation: the Amsterdam Methodology (GLEAM) product and/or ET estimates from a simple parametric model for typical Sahelian ecosystems as a normalized basis. The fusion algorithm is evaluated at the pixel scale against eddy-covariance measurements and simulations of a locally calibrated and validated land surface model (LSM) on a millet crop and a fallow of shrubby savannah in the southwest of the Republic of Niger. Consistency of the fusion approach is also evaluated at mesoscale by comparing it with a set of 20 regional LSMs. Globally both level of comparison highlight the very good agreement of ET estimates based on the fusion approach with both in situ measurements and LSMs simulations. We also show the benefit of such a fusion approach compared to linear temporal interpolation techniques of the ratio between ET and either incoming solar radiation or reference evapotranspiration. The main benefits are observed during the first months of the rainy season in the depiction of the surface response to rainfall events and consecutive drying up of the surface soil layer. We suggest that such a fusion approach could be later used in disaggregation chain to enhance field scaled ET estimates by the combination of coarse, moderate and high resolution remote sensing-based ET estimates.