Assimilation of Earth Observation Data for Crop Yield Estimation in Smallholder Agricultural Systems

Crop yield estimates are an important data output of agricultural monitoring systems. In sub-Saharan Africa, large input requirements of crop growth models, fragmented agricultural systems, and small field sizes are substantial challenges to accurately estimate crop yield. Multisensor data fusion can be a valuable source of high spatial and temporal resolution data to meet the requirements of crop growth models in Africa. In this study, we estimated crop yield in smallholder agricultural systems of Ethiopia by assimilating Landsat and MODIS fused data in the simple algorithm for crop yield estimation (SAFY) model. Enhanced spatial and temporal adaptive reflectance fusion model (ESTARFM) adapted for fragmented agricultural landscapes was used for data fusion. We assimilated LAI and phenology information derived from Landsat-MODIS fusion, MODIS and field data for comparison. The model was validated with in situ LAI and yield measured in rice and maize fields during the 2019 growing season. Data fusion minimized the yield estimation error (rRMSE = 16% for maize and rRMSE = 23% in rice) more than MODIS (rRMSE = 20% for maize and rRMSE = 35% in rice) because of its higher LAI and phenology estimation accuracy. Data fusion improved the calibration accuracy of the field and crop-specific model parameters and better captured the spatial variability of yield, which is vital for crop production monitoring and food security in smallholder agricultural systems in Africa. Considering the promising results, further investigation into the transferability of the approach to other smallholder agricultural landscapes and hybridization with machine learning is needed for large-area applications.