Determination of critical nitrogen loss zones and high-resolution nitrate leaching modeling in drinking water protection areas with satellite-based algorithms

Abstract Purpose This study developed and validated an integrated framework to identify critical nitrate loss zones at sub-field scale in a drinking water protection area. The key question was whether satellite-based estimation of grain N uptake, combined with site-specific N balancing and soil–water–nitrate modeling, can reliably capture spatial patterns of nitrate leaching under practical farming conditions. Methods A satellite-based algorithm was developed to estimate winter wheat N uptake using data from georeferenced plots in southern Germany. Multiple linear regression and Random Forest models were applied based on Sentinel-2 vegetation indices, as well as soil, topographic, and climatic variables. Results and Discussion The Random Forest model performed slightly better (R 2 = 0.74, RMSE = 14.8 kg ha⁻ 1 , MAE = 11.2 kg ha⁻ 1 ) than the linear regression model (R 2 = 0.70, RMSE = 16.0 kg ha⁻ 1 , MAE = 12.5 kg ha⁻ 1 ). Predicted N uptake was combined with field-specific fertilization data to calculate site-specific N balances, which were integrated into a soil–water–nitrate leaching model. Low-yield zones showed lower N uptake but higher N surplus, modeled nitrate leaching, and measured nitrate–N stocks in deeper soil layers compared to high-yield zones. During validation (Field A), low-yield zones showed N surpluses of 98 kg ha⁻ 1 and modeled nitrate concentrations of 120 mg L⁻ 1 in seepage water, compared to 33 kg ha⁻ 1 and 49 mg L⁻ 1 in high-yield zones. Modeled nitrate concentrations correlated with measured nitrate–N stocks (R 2 = 0.68). The approach shows strong potential for identifying nitrate-prone areas and enabling targeted measures for drinking water protection.