Forest Road Status Assessment Using Airborne Laser Scanning

Forest roads allow access for silvicultural operations, harvesting, recreational activities, wildlife management, and fire suppression. In British Columbia, Canada, roads that are no longer required must be deactivated (temporarily, semipermanently, or permanently) in order to minimize the impact on the overall forested ecosystem. However, the remoteness and size of the road network present challenges for monitoring. Our aim was to examine the utility of airborne laser scanning data to assess the status and quality of forest roads across 52,000 hectares of coastal forest in British Columbia. Within the forest estate, roads can be active or deactivated, or have an unknown status. We classified road segments based on the vegetation growth on the road surface, and edges, by classifying the height distribution of airborne laser scanning returns within each road segment into four groups: no vegetation, minor vegetation, dense understory vegetation, and dense overstory vegetation. Validation indicated that 73 percent of roads were classified correctly when compared to independent field observations. The majority were classified as active roads with no vegetation or deactivated with dense vegetation. The approach presented herein can aid forest managers in verifying the status of the roads in their management area, especially in remote areas where field assessments are costly and time-consuming. Study Implications: Large-area assessments of road status are critically important for operational forest management. We classified road segments based on the vegetation growth on the road surface, and edges, by classifying the height distribution of airborne laser scanning returns within each road segment into four classes: no vegetation, minor vegetation, dense understory vegetation, and dense overstory vegetation. Validation indicated that 73 percent of roads were classified correctly when compared to independent field observations. The majority of road segments were classified as either active roads with no vegetation or deactivated roads with dense vegetation. This information is potentially valuable for forest planners, as it allows for the identification of road segments that require maintenance or, alternatively, informs on successful road deactivation. Additional information on precise road centerlines is a required first step for our approach. We would propose that if this approach were to be made operational, future work should integrate automated road detection using the airborne laser scanning data into the proposed current workflow. Road extraction could also assess the exact width of each road and would further improve the classification results and greatly enhance the automation process.