Bridging the gap between leaf and canopy nitrogen content in almond orchard with UAV-based sensor fusion

Efficient nitrogen management is essential for optimizing yield and minimizing environmental impacts in commercial orchards. However, conventional nitrogen assessment methods, based on destructive leaf or soil sampling, are labor-intensive and fail to capture the spatial and temporal variability of nitrogen uptake at the tree level. A key challenge in orchard-scale nitrogen modeling is scaling leaf-level measurements to canopy-level estimates. This study presents a novel, Unmanned Aerial Vehicle (UAV)-based multi-sensor fusion framework for non-invasive, high-resolution monitoring of nitrogen seasonal dynamics at the leaf and canopy levels in almond orchards. This UAV-based multi-sensor fusion framework was used to estimate canopy nitrogen content (CNC) and leaf nitrogen content (LNC) at tree levels. CNC represents the total nitrogen content per unit ground area by integrating leaf-level nitrogen with canopy structure, while LNC reflects the nitrogen concentration within individual leaves. By combining LiDAR-derived canopy structural data, multispectral imagery, and environmental variables in ML models, we estimated nitrogen dynamics at the tree scale. We further linked these dynamics to yield in almond orchards across two cultivars (UEF and 53) over two growing seasons (April–November 2022–2023). Monthly field measurements were synchronized with UAV campaigns and the acquisition of meteorological data. Random Forest models trained demonstrated high predictive performance (CNC: R²₍_Val₎ = 0.862, RMSE = 0.672 kg/tree; LNC: R²₍_Val₎ = 0.894, RMSE = 0.165 % dry weight). Growing degree days, cultivar identity, phenological stage, and chlorophyll vegetation indices were key explanatory features. CNC showed a strong correlation with yield (R² = 0.87), outperforming LNC, and reflecting its integrative value. Spatial-temporal analysis revealed cultivar-specific nitrogen responses and increasing divergence during the dry season. These results highlight the potential of the proposed framework, which provides a scalable, data-driven solution for nitrogen monitoring and precision fertilization, offering new insights into nitrogen dynamics for sustainable orchard management.