Abstract
Understanding the spatiotemporal variability of actual evapotranspiration (ET) is a critical issue for better water resources management from plot to basin scale. It is particularly true in the Sahelian region which is very vulnerable in terms of water and agricultural resources, but also very scarcely monitored. This study aims to improve understanding of ET in Sahelian agrosystems by comparing twenty ET products and collections on two areas in Central Senegal and South-Western Niger: Breathing Earth System Simulation (BESS), Derive Optimal Linear Combination Evapotranspiration (DOLCE) v2.1 and v3.0, ERA5 and ERA5-Land, FLUXCOM-RS, Global Land Data Assimilation System (GLDAS) on Noah and Catchment Land Surface Model (CLSM), Global Land Evaporation: the Amsterdam Methodology (GLEAM) v3.5a, v3.5b, v3.6a and v3.6b, Modern-Era Retrospective analysis for Research and Applications v2 (MERRA-2), MODIS Global Evapotranspiration products (MOD16/MYD16), Penman-Monteith Leuning version 2 (PML_V2) v0.1 and v0.1.7, Reliability Ensemble Averaging (REA), Simplified Surface Energy Balance for operational applications (SSEBop) and Water Productivity through Open access of Remotely sensed derived data (WaPOR). In order to assess the abilities and drawbacks of each product, a multi-scale analysis is first performed at local scale with different temporal aggregation levels (daily, decadal, monthly). The best performing products, like GLEAM, MERRA-2 or GLDAS-Noah, have low spatial resolution (≥0.25°), but shows really good agreement with ET measurements (RMSE ∼ 0.5 mm.d-1, R2 ≥ 0.8). A multi-scale spatial analysis with different spatial aggregation levels (1, 10 and 25 km) is then performed on interannual composite maps of annual and seasonal ET. Results highlight the different spatiotemporal behaviour of all the products. Products such as ERA5-Land and PML_V2 0.1.7 stand out in this comparison as their higher resolution allows them to describe the spatial patterns of ET realistically and more precisely while having relatively good performances at local scale (RMSE ∼ 0.6 mm.d-1, R2 ≥ 0.7). Overall, the comparison highlights the need to consider soil moisture to accurately estimate ET in the Sahelian region, as the products considering it generally show better performances. It also points out the lack of a high quality and resolution ET product over the Sahel. Future satellite missions, as well as data fusion techniques, could help to fill up this gap and propose a reference product on Sahelian ecosystems.