Mapping postglacial succession of coniferous tree species in high elevation ecosystems of British Columbia

Abstract In western North America, most glaciers are predicted to disappear by 2100 under a 2-3 degrees Celsius warming scenario. Despite this, little research has been done on postglacial succession in the region. To address this, we leveraged multispectral drone imagery and machine learning techniques to investigate patterns of post-glacial conifer succession in two glacial bays in Garibaldi (Nch'kaSI) Provincial Park: Sphinx Bay and Sentinel Bay. We classified rasters at a 10 cm resolution using multispectral band values from ground-based training data. The random forest classification predicted subalpine fir (Abies lasiocarpa) dominated in Sentinel Bay and mountain hemlock (Tsuga mertensiana) as the dominant tree in Sphinx Bay. Higher elevations, less sun exposure, and more recent deglaciation in Sentinel Bay indicate that subalpine fir and mountain hemlock follow distinct post-glacial successional patterns, with subalpine fir more tolerant of poorly developed paraglacial soils and mountain hemlock requiring more well developed, mesic soils. These patterns confirm previously documented patterns of forest succession among these species in other areas. We found that machine learning tools can effectively be combined with drone imagery and ground surveys to map vegetation responses to deglaciation, but results can be limited by subjective photointerpretation and sparse data.