A new method for satellite-based derivation of site-specific yield potentials of winter wheat for precision farming applications

Accurate site-specific yield potential determination is essential for sustainable agricultural practices, enabling efficient resource use and reducing environmental impacts. This study presents a satellite- based method to predict winter wheat yield using a six-year dataset from conventional  nd organic farming systems. Two models were evaluated: a non-linear support vector machine (SVM) model and a multilinear model with stepwise backward regression. The SVM model demonstrated higher predictive accuracy, with an R² of 0.80, a MAE of 0.58 t/ha, and an RMSE of 0.79  /ha, compared to the multilinear model (R²=0.75, MAE=0.69 t/ha). Expanding the dataset to include diverse soils, climates, and growth stages could further enhance model robustness and applicability.