Detection of Bursaphelenchus xylophilus infection in Pinus massoniana from hyperspectral data

Hyperspectral remote sensing has great potential for accurate detection of forest pests and diseases. The main objectives of this research were: i ) to determine the best hyperspectral wavelengths or their combinations to discriminate Pinus massoniana trees infected by Bursaphelenchus xylophilus disease from healthy trees over the spectral range of 350-1000 nm; and ii ) to assess the chlorophyll content of infected trees using the hyperspectral algorithm. We also discuss the possibility of an early detection method of B. xylophilus disease dynamics by combining the spectral characteristics and the chlorophyll content of trees. The hyperspectral data were gathered for six stages of healthy to infected trees using a 1-nm-wide handheld spectroradiometer. First derivative (FD) spectra and vegetation indices were used for data dimensionality reduction and to select the most effective wavelengths for detection. The most effective FD spectrum in 759 nm was selected to discriminate the infected and healthy P. massoniana plants. The normalised difference vegetation index (NDVI) (810,450) value in the fully infected stage correlated with the variation of chlorophyll content (). We conclude that the combination of specific spectral characteristics and chlorophyll content is a reliable method for confirming infection about 30 days after B. xylophilus inoculation.