Dynamic Analysis of Early Stage Pine Wilt Disease in Pinus massoniana Using Ground-level Hyperspectral Imaging

Abstract

Pine wilt disease (PWD) is caused by the pine wilt nematode and is a tremendous threat to coniferous forests. Remote sensing, particularly hyperspectral remote sensing, has been utilized to identify PWD. However, most studies have focused on distinguishing the spectra between infected and healthy pine trees and ignored further visualization of spectral symptoms, which could greatly improve the pre-visual diagnosis of PWD. This research used the false color feature maps (FCFMs) synthesized using the normalized difference vegetation index (NDVI) and the ratio vegetation index (RVI) calculated from selected feature bands to analyze the changes in the spectral and image dimensions of the hyperspectral data. Our main findings were (1) the confirmed feature bands were 440, 550, 672, 752, 810, and 958 nm; and (2) NDVI (810, 440), NDVI (810, 672), NDVI (550, 672), RVI (810, 550), RVI (810, 672), and RVI (550, 672) were suitable to synthesize the FCFMs. As PWD developed, the color of the infected needles changed from blue and white to red on the NDVI-based feature maps and from blue to red on the RVI-based feature maps. Importantly, the color changes were captured by the FCFMs when the symptoms were not visible on the true color images, indicating the ability to identify PWD during the early infection stage. Study Implications: Many studies on PWD detection using remote sensing only focus on spectral information but ignore image information. In this article, a method was proposed to comprehensively utilize the spectral and image information of hyperspectral data. In addition, the ground-level imaging spectrometer was used to collected hyperspectral data of lateral branches of infected pine trees, which has rarely been the focus of other remote sensing platforms. This research helps to identify PWD as early as possible and thereby reduces the damage of PWD to pine forest resources.