Diagnosis of nitrogen nutrition status in spring maize based on multi-source data fusion analysis from unmanned aerial vehicles

Maize ( Zea mays L.) is a vital food crop and plays a critical role in ensuring global food security. Efficient diagnosis of nitrogen (N) deficiency and excess is essential for optimizing spring maize cultivation. The nitrogen nutrition index (NNI) serves as a key indicator for assessing N status, but its application in large-scale agricultural production remains limited due to cumbersome and costly measurement procedures. In this study, a two-year field experiment was conducted using multiple maize varieties under varying N application rates. Multimodal data-including multispectral imagery and LiDAR point clouds-were collected at key growth stages using a quad-sensor unmanned aerial vehicle (UAV) system. A customized dynamic model based on cumulative growing degree days was developed to predict the optimal normalized difference vegetation index (NDVI) trajectory. Variety-specific parameterization of the model achieved high predictive accuracy (R²: 0.955–0.988; RMSE: 0.142–0.216). The NNI was calculated by dividing the UAV-derived NDVI values by the predicted optimal NDVI. By integrating multispectral and LiDAR data through machine learning techniques, a robust NNI prediction model was established. Fusion of multi-source data significantly enhanced prediction accuracy, with the k-nearest neighbors algorithm yielding the best performance (R² = 0.76, RMSE = 0.007). Spatiotemporal maps of N status were generated to facilitate precision nutrient management. This study presents a high-throughput, non-destructive method for NNI diagnosis based on multi-source data fusion, providing a practical framework for efficient and precise N management in maize and potentially other field crops.