Development of monitoring methods for Hemlock Woolly Adelgid induced tree mortality within a Southern Appalachian landscape with inhibited access

Abstract

Hemlock woolly adelgid (Adelges tsugae Annand, HWA) is an introduced invasive forest pest in eastern North America. Herbivory by this insect results in mortality to eastern hemlock (Tsuga canadensis L. Carr.) and Carolina hemlock (Tsuga caroliniana Engelm.). These species occur in landscapes where extreme topographic variation is common. The vegetation communities within these landscapes feature high diversity of tree species, including several other conifer species. Traditional forest inventory procedures and insect pest detection methods within these limited-access landscapes are impractical. However, further information is needed to evaluate the impacts of HWA-induced hemlock mortality. Accordingly, our goal was to develop a semi-automatic method for mapping patches of coniferous tree species that include the living hemlock component and tree mortality by the HWA using aerial images and LiDAR (light detection and ranging) to increase our understanding of the severity and pattern of hemlock decline. The study was conducted in the Linville River Gorge in the Southern Appalachians of western North Carolina, USA. The mapping task included a two-phase approach: decision-tree and support vector machine classifications. We found that about 2% of the forest canopy surface was covered by dead trees and 43% by coniferous tree species. A large portion of the forest canopy surface (over 55%) was covered by deciduous tree species. The resulting maps provide a means for evaluating the impact of HWA herbivory, since this insect was the only significant coniferous mortality agent present within the study site.

Keywords

• Decision-tree Classification
• Eastern Hemlock
• Hemlock Woolly Adelgid
• Remote Sensing
• Support Vector Machine