# Spatiotemporal pattern analysis of juglans leaf necrosis disease occurrence and development in southern Xinjiang, China, based on UAV

## TL;DR

This study uses drones with advanced imaging to track and map walnut leaf disease in southern Xinjiang, China, offering a faster and more accurate monitoring method.

## Contribution

The study introduces a UAV-based hyperspectral monitoring system for tracking Juglans leaf necrosis with high accuracy and identifies key vegetation indices for severity assessment.

## Key findings

- MRESRI, CRI1, and PRI were the most effective vegetation indices for mapping disease severity.
- Random Forest achieved 86% accuracy and a Cohen’s kappa of 0.825 in classifying JLN severity.
- Disease hotspots were found in low-lying areas, near roads, and in dense walnut stands.

## Abstract

Juglans leaf necrosis (JLN) is a physiological disease primarily associated with abiotic stressors such as high temperatures, drought, and soil salinity, though biotic factors may also exacerbate its severity. It is a global concern affecting walnut production in multiple regions, including Xinjiang, China. In recent years, climate change, shifting agricultural practices, and disease transmission have increased its incidence, severely affecting tree growth, yield, and quality. Traditional field-based monitoring is labor-intensive and often inaccurate, underscoring the need for advanced remote sensing. To provide fast and objective monitoring, we used hyperspectral and high-resolution RGB imagery acquired by an unmanned aerial vehicle (UAV) to track JLN from June to September 2024 in southern Xinjiang. Five survey rounds captured the progression of disease severity. Among 17 vegetation indices, the modified red edge simple ratio (MRESRI), carotenoid reflectance index 1 (CRI1), and photochemical reflectance index (PRI) were the most informative for severity mapping. A Random Forest classifier achieved 86% overall accuracy and a Cohen’s kappa of 0.825. Spatial patterns showed persistent hotspots in low-lying areas, near roads, and in dense stands. These findings provide an effective, scalable approach for early detection and severity assessment, enabling timely, targeted interventions. Adoption of UAV-based hyperspectral monitoring can improve field surveillance, optimize resource allocation, and support sustainable walnut production.