Discrimination of adelgid-damage on single balsam fir trees with aerial remote sensing data

Abstract

Compact Airborne Spectrographic Imager (CASI) multi-spectral and panchromatic images were acquired in July 1992 over two forest plots infested by the balsam woolly adelgid (Adelges piceae) in western Newfoundland. A panchromatic image (pixel size approximately 25 cm) was used as a georeference for the coarser resolution multi-spectral data which were resampled to 1 m and 0.5 m, then carefully tied to a detailed plot plan showing the locations of 159 trees. Field observations for each of these trees included a ranking for adelgid damage class according to a standard Forest Insect and Disease Survey (FIDS) scheme. A discriminant analysis of the multi-spectral CASI imagery (together with semi-variance parameters and texture derivations) indicated that damage caused by the balsam woolly adelgid on balsam fir (Abies balsamea) trees could be detected and separated into severity classes with a range of accuracy between 40-76 per cent depending on the classification scheme and the type of remote sensing variables available to perform discrimination. The 'best' discriminant results obtained were based on a single pixel sample extracted from a 0.5 m multi-spectral image comprised of six bands centred on 454, 550, 590, 662, 836, and 895 nm. A central theme of this continuing effort is that the combination of multi-spectral, spatial and high spectral resolution image processing may provide further insight into optimal damage detection-and subsquent hazard ratings-using digital remote sensing imagery.