Nitrogen is a cornerstone of plant health, directly influencing photosynthesis, growth, and fruit development. However, managing nitrogen in intensive apple orchards is complicated by significant tree-to-tree variability. Traditional laboratory analysis of leaf samples is destructive and time-consuming, offering only a snapshot in time. While hyperspectral remote sensing holds promise for rapid, non-destructive monitoring, its application in orchards has been hindered by the coarse resolution of sensors, which struggles to isolate individual overlapping tree canopies, and the complex three-dimensional (3D) architecture of the trees themselves, which creates spectral mixing from leaves, branches, and shadows. Due to these issues, there is a need to develop more integrated methods that can accurately extract tree-level information from complex, dense canopies.
A team of researchers from the Agricultural Information Institute and the Institute of Fruit Tree Research, both part of the Chinese Academy of Agricultural Sciences, has developed a novel cross-modal framework to tackle these challenges. Their findings, published (DOI 10.1016/j.plaph.2026.100172) in the journal Plant Phenomics on September 7, 2026, detail a method that aligns high-resolution RGB point clouds with hyperspectral orthomosaics. This co-registration allows for precise delineation of individual tree canopies, enabling a stratified analysis of nitrogen distribution and significantly improving estimation accuracy.
The study's innovative workflow is built on several key technical advances. First, it uses RGB-derived 3D point clouds to accurately locate individual tree canopies in dense plantings, overcoming the limitations of low-resolution hyperspectral data. This enabled researchers to quantify a consistent vertical nitrogen gradient, where the lower canopy had a 0.5-9.5% higher nitrogen concentration than the upper layer across different fertilization treatments. To extract the most relevant information from the noisy spectral data, the team applied Continuous Wavelet Transform (CWT) to the canopy reflectance. The CWT-2 scale proved most effective at enhancing subtle, nitrogen-specific spectral features, outperforming traditional original reflectance models. Crucially, K-means clustering was used to partition canopy pixels into two distinct categories: "Cluster-A," representing dense, vegetation-dominated interior pixels, and "Cluster-B," which captured mixed pixels from the canopy's edge and gaps. Models based solely on Cluster-A pixels achieved a significantly higher validation R² of 0.69-0.76, compared to the less reliable 0.48-0.57 from the boundary-affected pixels. This clustering step effectively filters out spectral interference from background elements like soil and shadowing.
"The core of our approach is about looking beyond the superficial signal," said the author. "For years, we've treated the tree canopy as a uniform entity, but it's not. By separating the canopy into layers and, more importantly, purifying the pixels we analyze, we're effectively cutting through the noise created by shadows and structural complexity. The K-means clustering is a game-changer; it allows us to focus on the most reliable, foliage-dominated spectral data, which is where the true nitrogen signal lies. This moves us away from a 'one-size-fits-all' model to a more nuanced, accurate assessment for individual trees."
This research provides a powerful, operational blueprint for precision agriculture. The framework enables orchard managers to create detailed, tree-level nitrogen maps, facilitating variable-rate fertilization that applies nutrients only where they are needed. This not only optimizes fruit yield and quality but also minimizes environmental waste. The study highlights how integrating spatial and spectral data can overcome the inherent complexities of orchard systems, paving the way for more robust biochemical retrieval. This technology can be extended to other high-value tree crops like citrus and olives, offering a scalable solution for sustainable and efficient nutrient management in the future.
###
References
DOI
10.1016/j.plaphe.2026.100172
Original Source URL
https://doi.org/10.1016/j.plaphe.2026.100172
Funding information
This work was supported by the Basic Research Center, Innovation Program of Chinese Academy of Agricultural Sciences (CAAS-BRC-SAE-2025-01 and CAAS-ASTIP-2026-AII).
About Plant Phenomics
Plant Phenomics publishes breakthrough research that integrates genomics, genetics, physiology, molecular biology, bioinformatics, statistics, mathematics and computer science to advance plant phenotyping—from the cellular level to entire populations. The journal tackles key scientific challenges in phenomics by showcasing cutting‑edge technologies for data acquisition, management, interpretation and modelling. Its scope spans practical applications across all plant sciences, aiming to translate novel phenotyping tools into real‑world solutions for breeding, crop management and fundamental plant biology. By fostering interdisciplinary innovation, Plant Phenomics serves as a hub for researchers developing high‑throughput, accurate and scalable methods to decipher complex plant traits, ultimately accelerating progress in agriculture and plant science.