A GEE TSEB workflow for daily high-resolution fully remote sensing evapotranspiration: Validation over four crops in semi-arid conditions and comparison with the SSEBop experimental product

Abstract

Accurate and synoptic estimation of Evapotranspiration (ET) is crucial for water management. A Google Earth Engine workflow is implemented to estimate daily ET at 30m. The algorithm uses Landsat and ERA5-Land datasets and includes the Two Source Energy Balance (TSEB) model, an Artificial Neural Network for Leaf Area Index, and a gap-filling approach based on crop coefficient. The outputs were evaluated against four local flux towers in a semi-arid site in Morocco (wheat, maize, watermelon, olive), and compared to another high-resolution ET (SSEBop product). The results demonstrated good performances (RMSE between 0.67 mm/day and 2 mm/day, low MBE), while SSEBop product generally underestimated ET. Better performance of the TSEB-GEE workflow was found when aggregating ET to weekly and monthly timescales. The workflow offers ease of model implementation to deliver reliable daily plot-scale ET estimates, offering the potential for broader-scale applications in semi-arid Mediterranean regions, encompassing various crops and facilitating historical analysis.