Spectral analysis of Scotch pine infested by Sirex noctilio

Abstract

Remote sensing methods for monitoring forest health have advanced with the development of sophisticated new tools and techniques in recent years. Prior research had explored needle-level analysis of Sirex infestation and this research explores a new method of investigating the impact of Sirex woodwasp (Sirex noctilio) infestation in Scotch pines using 8-band multispectral WorldView-2 imagery. The goal of the project was to assess if the broadband spectral regions associated with this sensor can characterize subtle changes in spectral reflectance caused by Sirex infestation. Eight different spectral indices were derived from the images and statistical analysis was performed. While the needle-level analysis previously reported showed statistical differences, none of the eight spectral indices showed statistically significant differences between the healthy and infested trees at a 0.05 significance level, though some showed differences with a weaker significance level. An automated calibration model based on exhaustive search techniques was used to consider the optimum threshold to determine the capability of the vegetation index (ices) to detect changes in spectral reflectance from the infested trees. The accuracy assessment results were poor to moderate. The best overall accuracy was 68% with a Kappa coefficient of 0.33. While there is an opportunity to explore multispectral spectral bands of WorldView-2 data in mapping forest health using spectral indices, in our study, the broadband spectral indices were not capable of accurately differentiating the subtle changes in spectral properties of healthy and infested tree at the individual tree level.