Abstract
Forest insects and pathogens, in addition to fire, contribute to structural diversity by creating snags (dead trees) and dead tops on live trees or “topkill” in conifers throughout western North America. Snags and top-killed trees are important sources of wildlife habitat but quantifying their presence can be challenging at broad scales. Multiple approaches for detecting snags have been developed, but limited methods exist for detecting topkill via remote sensing. There is a need to monitor overlapping disturbances across time and space in addition to the structural diversity in mature and old-growth forests. We used airborne light detection and ranging (lidar) and associated field observations to map individual dead trees and live trees with topkill across a forest dominated by Pinus ponderosa in central Oregon. The lidar point cloud was segmented into individual tree objects (polygons representing tree crown extents as viewed from nadir), for which lidar metrics were computed. Lidar-detected tree objects were paired with 647 field-observed trees with corresponding tree health measurements, and a random forest classifier was developed that separated trees into: (1) live without topkill; (2) live with topkill; and (3) dead classes with 87% accuracy using lidar metrics. Tree mortality was mainly due to fire injury and native bark beetles, such as western pine beetle (Dendroctonus brevicomis), while topkill was primarily due to comandra blister rust (caused by the native fungal pathogen Cronartium comandrae). The classifier was applied to map individual tree health status for 46,444 tree objects from which tree health class density maps were created across the 229-ha study area. Approximately 43% were classified as live without topkill, 22% of trees were classified as live with topkill, and 35% of trees were classified as dead. This approach can be used to improve detection of topkill in conifers, along with tree mortality, caused by disturbances associated with forest insects and pathogens.