Detection of spruce beetle-induced tree mortality using high- and medium-resolution remotely sensed imagery

Abstract

Outbreaks of bark beetles cause widespread tree mortality, which have important consequences for wildlife, forest composition and structure, timber production, and water resources. Recent severe, extensive, and synchronous bark beetle outbreaks have motivated research into methods for surveying the location, extent, and attack stage of outbreak using remotely sensed data. In this paper, we adapt methods for classing mountain pine beetle-induced tree mortality to detect spruce beetle-induced tree mortality from single-date high-resolution imagery and multi-date moderate-resolution data. We use freely available aerial imagery from the National Agriculture Imagery Program (NAIP) to produce a tree-scale map of gray-stage spruce beetle kill across a broad spatial extent. This map was then used to train a stand-scale classification of spruce beetle outbreak from Landsat Climate Data Record (CDR) data. Gray-stage spruce beetle outbreak was characterized by higher values of the red–green index (RGI) at the tree-level. At the stand-level, gray stands were associated with high values of a Disturbance Index (DI) and low values of the Normalized Difference Vegetation Index (NDVI). Both tree- and stand-scale classifiers of spruce beetle-attributed tree mortality were highly accurate (overall accuracy and user's accuracy for gray trees/stands greater than 88%). Stand-level classification improved with post-classification correction, which removed high-frequency year-to-year variability in forest condition. Comparison of high-resolution and moderate-resolution classified imagery revealed variable amounts of tree mortality occur in areas of landscape-scale spruce beetle outbreak. Our results highlight the utility of both ecologically informed post-classification correction and coupling fine-scale and moderate-scale resolution imagery for mapping and studying spruce beetle outbreaks.