Integration of vegetation indices derived from Sentinel-2 and Google Earth Engine to estimate actual evapotranspiration in wheat

Evapotranspiration is a major component of the water cycle and agricultural water balance. However, although different methods exist to estimate evapotranspiration, achieving accurate results remains a challenging task. Therefore, these methods require calibration and validation under local conditions. Agricultural systems must integrate innovative tools that not only optimize irrigation water management but also enable the spatial assessment of water use under different conditions. We present a methodology to calibrate agricultural variables using multiple vegetation indices (VIs) through the VICAL tool implemented in the Google Earth Engine environment. Wheat actual evapotranspiration (ETa) and actual crop coefficient (Kca) were estimated from Sentinel-2 satellite imagery. The model’s calibration was carried out using data from three experimental sites and validation in five additional sites located in southern Sonora, Mexico. Wheat phenology and accumulated growing degree-days (GDD) were considered for model development; therefore, the analysis was divided into two periods: (1) early to mid-season growth stage and (2) late-season growth stage. The error parameters indicated better performance during the first analysis period. For estimating Kca and ETa, it is recommended that (1) during the early- and mid-season growth stage (0 to 1000 accumulated GDD), models based on the ratio vegetation index (RVI) and normalized difference vegetation index (NDVI) indices can be used interchangeably, and (2) during the late-season growth stage (>1000 accumulated GDD), a model based on wide dynamic range vegetation index (WDRVI) should be used. This approach, which evaluates more than 15 VIs to efficiently calibrate agricultural variables, offers a practical tool that may be extended to a wide variety of crops with proper local calibration and consideration of site-specific conditions.