Using imagery from unmanned aerial vehicles to investigate variation in snag frequency among forest stands

Abstract

Advances in unmanned aerial vehicle (UAV) technology have opened new opportunities for measuring canopy mortality processes through high-resolution aerial imagery. Unlike traditional plot-based sampling, UAVs can easily survey areas several hectares in size that encompass hundreds of canopy trees. The resulting imagery provides extensive data on the number and locations of dead trees (snags), which can be related to neighbourhood conditions and environmental factors that vary among stands. Here, we used a UAV to measure the relative proportions of live and dead canopy trees in stands varying in soil moisture across a mixed forest landscape. Crowns of live trees and snags were first segmented and classified to species from high-resolution canopy imagery. We then modeled whether crowns of each species were alive or dead as a function of tree height, neighbourhood competition, soil moisture, and the relative abundance of conspecifics. Short lodgepole pine (Pinus contorta var. latifolia) trees were more likely to be dead when surrounded by a tall canopy in sites with greater soil moisture, reflecting effects of competition on this shade-intolerant species. Trembling aspen (Populus tremuloides) was more likely to be alive when surrounded by tall neighbours, perhaps because this clonal species can benefit from facilitation through root connections with neighbouring aspen stems. There was a higher frequency of white spruce (Picea glauca) snags in sites where it had a high relative abundance, suggesting strong effects of intraspecific competition for this species. Soil moisture did not appear to have a direct effect on the snag frequency of any of these species, despite pronounced niche partitioning along an elevation-driven moisture gradient. Our models explained 39% and 33% of the variation in white spruce and lodgepole pine snag frequency, respectively, but did not have much predictive power for trembling aspen or total snag frequency. Our results reflect the important role of competition in determining tree mortality, but also indicate that stochastic or unexplained processes account for considerable variation in snag frequency among stands. As UAV technology becomes more widely used by ecologists, it may enable a better understanding of how biotic and abiotic processes produce local variation in canopy mortality.