Modeling Actual Evapotranspiration with MSI-Sentinel Images and Machine Learning Algorithms

The modernization of computational resources and application artificial intelligence algorithms have led to advancements in studies regarding the evapotranspiration crops by remote sensing. Therefore, this research proposed machine learning estimate ETrF (Evapotranspiration Fraction) sugar can crop using METRIC (Mapping Evapotranspiration at High Resolution with Internalized Calibration) model data from Sentinel-2 satellites constellation. In order achieve goal, images MSI sensor (MultiSpectral Instrument) OLI (Operational Land Imager) TIRS (Thermal Infrared Sensor) sensors Landsat-8 were acquired nearly same time between years 2018 2020 for cane crops. Images TIR intended calculate through (target variable), while images, explanatory variables extracted two approaches, 10 m (approach 1) 20 2) spatial resolution. results showed that able identify patterns predict model. For approach 1, best predictions XgbLinear (R2 = 0.80; RMSE 0.15) XgbTree 0.15). 2, algorithm demonstrated superiority was 0.91; 0.10), respectively. Thus, it became evident algorithms, when applied sensor, a simpler way than one involves energy balance thermal band used