Skip to main content
Research article - Peer-reviewed, 2022

Classification of pine wilt disease at different infection stages by diagnostic hyperspectral bands

Li, Niwen; Huo, Langning; Zhang, Xiaoli

Abstract

Pine wilt disease (PWD) is a very destructive forest disease that causes the mortality of pine. The infected trees usually die within three months, and the disease spreads fast with the long-horned beetle as the medium if the infected trees are not removed from the forest in time. Therefore, detecting the infected trees at different infection stage, especially the early infection, is crucial for preventing PWD spread. This study aims to exhibit the spectral differences of the pine needles between healthy pines and infected pines at different infection stages and reveal the diagnostic spectral bands for classifying the different infected stage trees. We collected needle samples from healthy, early-, middle-, late-stage infected trees in a Japanese pine (Pinus densiflora) forest and a Korean pine (Pinus koraiensis) forest in northern China to explore the spectral and biochemical properties differences of these four classes, and selected the sensitive bands combining competitive adaptive reweighted sampling (CARS) and successive projections algorithm (SPA). The selected bands were used for the four infection stages classification by linear discriminant analysis (LDA) algorithm. The results show that Chlorophyll a, chlorophyll b, carotenoids, and moisture content decreases with the aggravation of infection. The green (510–530 nm), red-edge (680–760 nm), and short-wave infrared (1400–1420 nm and 1925–1965 nm) bands are the sensitive bands, and the overall accuracy is 77 % and 78 % for the Japanese pine and Korean pine respectively when using these bands for classifying healthy, early-, middle-, late-stage infected trees. The results demonstrate that physiological parameters including Chlorophyll a, chlorophyll b, carotenoids, and moisture content can be used as the diagnostic parameters of PWD, and the selected sensitive spectral bands are feasible for detecting the stress symptoms of the Japanese pine and Korean pine.

Keywords

Hyper spectral; Pine wilt disease; Infection stages; Diagnostic spectral bands screening; Forest health monitoring

Published in

Ecological Indicators
2022, volume: 142, article number: 109198

Authors' information

Li, Niwen
Swedish University of Agricultural Sciences, Department of Forest Resource Management
Li, Niwen
Beijing Forestry University
Swedish University of Agricultural Sciences, Department of Forest Resource Management
Zhang, Xiaoli
Beijing Forestry University

UKÄ Subject classification

Forest Science

Publication Identifiers

DOI: https://doi.org/10.1016/j.ecolind.2022.109198

URI (permanent link to this page)

https://res.slu.se/id/publ/118881