Gaps and ghosts: How empty space and dead trees complement each other for a fuller picture of forest structure

Link To Share This Poster: https://cdmcd.co/zY9Jzq

Jessica M. Stitt, Fish & Wildlife Sciences, University of Idaho, Moscow, ID, Andrew Hudak, Rocky Mountain Research Station, USDA Forest Service, Moscow, ID, Lee A. Vierling, Natural Resources and Society, University of Idaho, Moscow, ID and Kerri T. Vierling, Department of Fish and Wildlife Sciences, University of Idaho, Moscow, ID

Background/Question/Methods
Many forest canopy gaps are ghosts of trees past, where the creation of gaps via the loss of foliage might be concurrently associated with snag development. Snags are keystone structures providing important substrates to wildlife species for breeding, roosting, and foraging. However, remotely sensing snags in closed canopy forests is challenging because snags are cryptic, rare, and randomly distributed. In previous work, we found that snags have distinct canopy gap profiles compared to live trees. Snags had a greater gap area within 36m2 of tree center than did live trees, and this difference was significant at heights 10–20m above ground level. The objective of this current study was to determine whether canopy gap information distinguishing individual snags from live trees was evident at the 25m-radius plot level. We evaluated dense conifer stands of the Idaho Panhandle National Forest where we paired airborne lidar with ground reference data collected for fixed-radius survey plots (0.2ha area) to evaluate gap structures. The R package ForestGapR was used to locate canopy gaps at 10m above ground and generate lidar-derived canopy gap metrics within each plot, such as total gap area, number of gap fragments, and multiple statistics about the canopy heights within (below) each gap fragment. We modeled snag density as a function of these canopy gap metrics for a subset of plots and applied the best model to remaining plots in order to evaluate the relative importance of canopy gap characteristics in predictive snag modeling.
Results/Conclusions
We divided our survey plots into training (n = 20) and validation (n = 12) sets, and fit a generalized linear model to the training set to predict snag densities per plot. There were a total of 180 snags across these 32 plots, with snag densities ranging from 0–23 snags/plot. Though our preliminary sample sizes were small, true snag densities were significantly correlated with predicted snag densities (p < 0.05). Expansion of this approach into the rest of our study sites and survey plots will yield additional information on these relationships. Because snags are important forest components for wildlife, fuel modeling, and carbon assessments, studies that can connect remotely sensed data to snag distributions and densities will impact multiple fields of ecology.