Open Collections

Description

The Garry oak (Quercus garryana) ecosystem is among the most endangered in Canada, yet conservation efforts are hindered by a lack of high-resolution, publicly accessible spatial data. This study addressed this knowledge gap by developing an automated workflow to identify and map individual Garry oak crowns across the Greater Victoria region using a fusion of airborne laser scanning (ALS) and four-band multispectral imagery. This imagery was used to extract structural and spectral metrics to train a Random Forest classifier. Results indicated high model performance with an overall accuracy of 89.7% and a specificity of 96.2%. A sensitivity of 61.0% reflected the challenges of discriminating interlocking crowns within high density forested patches and the spectral diversity of non-native species in an urban environment. However, the model remained robust in open-canopy environments where the unique structural complexity and spectral signature of Garry oak are most distinct. The high specificity of the final output ensures a reliable baseline for land managers and provides a framework for future ground-truthing. This research provides the methodology for the development of the first individual tree level inventory for the region, establishing a non-invasive and scalable approach for monitoring endangered ecosystems, and informing targeted habitat restoration strategies.