Multivariate Satellite Data Assimilation to Improve Yield Estimates in Deep South, USA

Crop growth simulation and yield prediction are integral to developing food security policies and agricultural strategies, particularly amidst the ongoing environmental challenges. However, crop models often fall short in regional yield estimation due to the spatial variation and uncertainties in soil properties, canopy state variables, and meteorological data. This study proposes a framework that integrates remote sensing observations with the DSSAT crop model using data assimilation aiming to enhance the accuracy of yield prediction at regional scales. Specifically, we assimilated remotely sensed Leaf Area Index (LAI) and Soil Moisture (SM) into the DSSAT crop model to estimate maize yield in the Mobile River Basin (MRB), the sixth-largest river basin in the United States. Sentinel-2 optical and Sentinel-1 Synthetic Aperture Radar satellite images were employed to retrieve LAI and SM. Three data assimilation schemes were implemented for maize yield simulation: assimilating LAI alone, SM alone, and LAI and SM jointly. The assimilation results demonstrated an improved performance in yield prediction compared to the open-loop simulation. Our study substantiates the feasibility and effectiveness of coupling remote sensing observations and crop models to better estimate regional crop yields, paving the way for food security policies and sustainable agricultural practices.