Diagnosis of nitrogen nutrition in winter wheat across years based on multi-source remote sensing data from unmanned aerial vehicles

Unmanned aerial vehicle (UAV) remote sensing has been widely employed for crop nitrogen status diagnosis and plays a crucial role in optimizing fertilization strategies. The nitrogen nutrition index (NNI) is a key parameter for assessing crop nitrogen status. However, studies focusing on cross-year NNI prediction for the same growth stages remain limited. In this study, field experiments on winter wheat with different nitrogen treatments were conducted in 2023 and 2024. Plant nitrogen concentration (PNC), aboveground biomass (AGB), and spectral reflectance data from multispectral and hyperspectral sensors were collected. Year-specific critical nitrogen dilution curves were constructed based on AGB to calculate NNI. The top 5 and top 10 multispectral vegetation indices (VIs) most correlated with NNI in 2023, along with hyperspectral reflectance bands, were selected as input variables. Using multiple linear regression (MLR) models and four machine learning algorithms—Random Forest (RF), Extreme Gradient Boosting (XGBoost), K-Nearest Neighbors (KNN), and Extremely Randomized Trees (ExtraTree)—were used for cross-year NNI prediction and spatial distribution mapping. Results showed that the VIs at three critical growth stages of winter wheat across both years exhibited significant correlations with NNI (P<0.01). When using the top 10 multispectral VIs as input variables, model prediction accuracy was generally higher, with the ExtraTree model achieving the best performance, reaching a coefficient of determination (R²) of 0.60, a root mean square error (RMSE) of 0.14, and a mean absolute error (MAE) of 0.12. The spatial prediction maps generated by the ExtraTree model clearly depicted the spatial variability of winter wheat nitrogen status, providing strong support for precision nitrogen management.