Serhan Yeşilköy , Ibrahim Demir

Crop yield and phenological stages are remarkably sensitive to not only environmental factors like atmospheric conditions and physical properties of soils but also agricultural activities. Accurate crop yield prediction plays a crucial role in food security and agricultural sustainability. There are several approaches that a wide range of researchers have tried to predict crop yield at different scales. In this study, we tested AgERA5 reanalysis product and crop phenological stage data to predict winter wheat yields in the agricultural lands of the agroclimatic regions of Turkey. The main objective is to propose a deep learning approach based on the combination of the reanalysis, which was extracted for the agricultural lands of the five most productive agroclimatic zones, and crop phenology data to predict winter wheat yields. Five performance indicators, such as normalized root mean squared error (NRMSE), mean absolute percentage error (MAPE), root mean squared error (RMSE), Nash-Sutcliffe Efficiency (NSE), and coefficient of determination (R2), are chosen to test the model’s accuracy and effectiveness. We have obtained promising findings and suggested that AgERA5 reanalysis data can be used as an input for the crop yield prediction of winter wheat with an error below 10% and a coefficient of determination above 0.9.