Mapping dead wood distribution in a temperate hardwood forest using high resolution airborne imagery

Dead wood, in the form of coarse woody debris and standing dead wood, or snags, is an essential structural component of forest ecosystems. It plays a key role in nutrient cycling, ecosystem functions and provision of habitat for a wide variety of species. In order to manage dead wood in a temperate hardwood forest, an understanding of its availability and spatial distribution is important. This research evaluates airborne digital camera remote sensing for mapping temperate forest dead wood across an area within Gatineau Park, Canada. Two approaches were evaluated: (1) direct detection and mapping of canopy dead wood (dead branches and tall snags) through the combination of three techniques in a hybrid classification: ISODATA clustering, object-based classification, and spectral unmixing, and (2) indirect modelling of coarse woody debris and snags using spectral and spatial predictor variables extracted from the imagery. Indirect modelling did not provide useful results while direct detection was successful with field validation showing 94% accuracy for detected canopy level dead wood objects (i.e. 94% of validation sites with canopy dead wood were detected correctly) and 90% accuracy for control sites (i.e. 90% of validation sites with no canopy level dead wood were identified correctly). The procedures presented in this paper are repeatable and could be used to monitor dead wood over time, potentially contributing to applications in forest carbon budget estimation, biodiversity management, and forest inventory.