Predicting Olive-sided Flycatcher (Contopus cooperi) Breeding Habitats in Southwestern Nova Scotia Using LiDAR Metrics Informed by Drone Data.

Abstract

The Olive-sided Flycatcher (OSFL) is a migratory species at risk bird, currently listed as “Threatened” in Nova Scotia, and “Special Concern” federally. The strategy to promote the recovery of OSFL population is expected to revolve around protecting existing breeding habitat as soon as these locations are found. Two studies have previously modeled OSFL habitat in Nova Scotia using tree-stand level input layers which makes it impossible to identify the within-stand characteristics used by the bird when choosing their habitat. These characteristics of fine-scale forest structure are closely tied to their foraging strategies and are the main driver determining OSFL occupancy rates in these habitats. The goal of this study is to use high-resolution drone imagery to inform LiDAR metrics as inputs for a model that predicts OSFL breeding habitat locations in Nova Scotia. The canopy height models (CHM) for the two data types were compared at 19 known OSFL habitat sites across the province by assessing tree spacing, canopy cover, and the vertical heterogeneity of the treetops to determine which LiDAR metrics can show the within-stand characteristics of OSFL habitat. A correlation test identified three metrics in the drone CHM that could be comparatively measured in the LiDAR CHM: canopy cover and the mean and standard deviation of tree heights. These metrics were then used as inputs for a Maximum Entropy (MaxEnt) model alongside other environmental layers important for characterizing OSFL habitat. MaxEnt created a predicted distribution of the species from occurrence data and the environmental input layers to identify where habitat with similar environmental characteristics could occur. After each run of the model, the performance of each input covariate was assessed, and the worst performing covariate was removed before the model ran again. This process was repeated until the best fit model was identified. The final model consisted of four environmental covariates used to predict OSFL habitat locations: proportion of canopy cover, distance to wetlands, mean canopy heights, and distance to spruce stands. The model performed comparatively well to previous predictive habitat models for the OSFL and identified 48.90% of the Kespukwitk area as being suitable habitat for the OSFL. The results showed the importance of capturing the variation within the OSFL’s habitat for predicting habitat locations, evident by the LiDAR-derived covariate measuring the proportion of canopy cover performing better than all other covariate used in any model predicting OSFL habitat in Nova Scotia. Locating where these habitats occur throughout the province is crucial to inform where recovery strategies would be most effectively implemented to protect the OSFL Atlantic population.