Improving satellite-based soybean yield mapping across irrigated and rain-fed conditions

Field-scale yield maps of staple crops can provide insights that improve management and scientific understanding. Using satellite imagery in conjunction with crop model simulations can effectively provide field-level estimates at scale. Current approaches, however, often poorly account for variable crop phenology and changes in soil moisture due to irrigation or shallow groundwater. Here we tested methods to improve the reliability of satellite-derived soybean yield maps across a regional environmental gradient, multiple growing seasons, and supplemental water conditions. Using Landsat and Sentinel time series, we calibrated APSIM crop simulations to better match satellite observations in both irrigated and non-irrigated systems. We then applied the calibrated models using the satellite-based Scalable Crop Yield Mapper (SCYM) across the midwestern United States from 2008-2018. Preliminary results show that training models with region-specific weather, management, and phenology parameters improves the ability to account for yield variation. These improvements can transfer to agricultural regions with limited yield data and support assessment of agricultural practices.