Detection of early stage bark beetle infestations in spruce stands using multitemporal low cost hyperspectral imaging

Abstract

The insect outbreaks and resulting forest damages are becoming more intense nowadays. One of preconditions for successful control and fighting the insect attacks is early detection of the outbreaks. Recent advance of low-cost but effective remote sensing solutions offers new opportunities in monitoring forest conditions. The objective of this study was to develop operationalization proposals for using low-cost hyperspectral imaging technology for the identification of early stage bark beetle infestations in spruce stands. The study was carried out in Dubrava forest located in central part of Lithuania. Four imaging flights were conducted in summer seasons of 2017 and 2018, aiming to capture images after the infestation of spruce trees by bark beetle. Imaging system built on simultaneous use of a hyperspectral camera based on a tunable Fabry-Pérot interferometer and colour-near infrared (CIR) camera, installed on ultra-light type aircraft was used for data acquisition. Attacked trees were field identified and mapped, together with sample of not-infested trees. Images were processed to get image mosaics with ground sample distance (GSD) 15-20 cm for CIR images and GSD 40-60 cm for the hyperspectral images. The spectral characteristics of images were used to classify the crowns into two categories – infested and not-infested. The overall accuracy of classification was reached for the best case 77 % (Cohen’s kappa: 0.54). Classification accuracies using hyperspectral data were up to 20% better than using the CIR data. The results suggested that novel remote sensing techniques could offer great potential for assessment of the health condition of spruce stands.