Can UAV-Derived Maize Canopy Traits Inform Soil Mineral Nitrogen Estimation? Linking Above- and Below-Ground Nitrogen Dynamics in Black Soils

Accurately characterizing crop–soil nitrogen dynamics is essential for precision nitrogen management in maize production, yet soil mineral nitrogen (soil Nmin) remains difficult to monitor under field conditions. This study evaluated whether UAV-derived maize canopy traits could be used to indirectly estimate soil Nmin in black soils through a cascaded modeling framework. Multi-stage UAV multispectral observations, agronomic variables, and machine learning were integrated into a cascaded framework in which aboveground nitrogen uptake (ANU) was first predicted and then used as an intermediate variable for soil Nmin and yield estimation. UAV-derived vegetation indices showed stronger relationships with ANU than with soil Nmin across growth stages, indicating that canopy spectral signals more directly reflected plant-level nitrogen accumulation. XGBoost achieved the best performance for ANU and soil Nmin prediction, with R2 values of 0.94 and 0.82, RMSE values of 15.99 kg ha−1 and 6.42 mg kg−1, and rRMSE values of 13.4% and 17.6%, respectively. Predicted ANU was the most influential variable for soil Nmin estimation, and the framework also captured the nitrogen response pattern of yield. These results indicate that UAV-based canopy sensing can support the indirect estimation of soil Nmin through crop nitrogen status, as well as nitrogen response diagnosis and data-informed nitrogen management in maize production.