A Multi-model approach for remote sensing-based actual evapotranspiration mapping using Google Earth Engine (ETMapper-GEE)

Accurate estimation of actual evapotranspiration (ETa) through remote sensing (RS) is essential for effective large-scale water management. We developed an EvapoTranspiration Mapper in the Google Earth Engine environment (ETMapper-GEE) to estimate RS-ETa using Landsat satellite data employing four models: Surface Energy Balance Algorithm for Land (SEBAL), Mapping EvapoTranspiration at high Resolution with Internalized Calibration (METRIC), surface temperature-vegetation-based triangle (TriAng), and Operational Simplified Surface Energy Balance (SSEBop). The estimation integrates extrapolation approaches (Evaporative Fraction (EF) and EvapoTranspiration Fraction (ETF)), reference ET types (grass (ETo) and alfalfa (ETr)), and climate forcing datasets (the fifth generation of the European ReAnalysis (ERA5-Land) and the Climate Forecast System version 2 (CFSv2)). The ETMapper was evaluated against observed data from flux towers in Germany for the period 2020 to 2022. The results showed that EF outperformed the ETF approach, with a more than an 8 % higher correlation of determination (R-2) and 35 % lower Root Mean Square Error (RMSE) compared to the other approaches. Among the EF approaches, TriAng (RMSE = 1.38 mm d(-1)) exhibited the best performance, followed by METRIC (1.69 mm d(-1)) and SEBAL (2.07 mm d(-1)). Using ETMapper with ETo resulted in at least 4 % higher R-2 and reduction in RMSE by at least 29 % compared to ETr. Forcing ETMapper with ERA5 yielded better accuracy (R-2 > 4 %, RMSE < 12 %) than when using CFSv2. This study provides an integrated framework for RS-ETa estimation, supporting water-related Sustainable Development Goals, especially in agricultural contexts.