Modelling the occupancy of two bird species of conservation concern in a managed Acadian Forest landscape: Applications for forest management

Measuring and modeling bird occupancy in managed forest landscapes can provide useful information for achieving sustainable forestry practices and for informing conservation. Occupancy models are useful tools to describe the habitat use of species and predict how changes in the habitat and landscape may affect future occupancy. Specifically, these models can help support changes in forest management practices to improve habitat quality for species, including species of conservation concern. We used occupancy modelling to quantify the habitat use of two species of conservation concern, Canada warbler (Cardellina canadensis) and olive-sided flycatcher (Contopus cooperi) in the managed Black Brook district forest in northwestern New Brunswick, Canada as case studies to demonstrate an approach for improving sustainability of forest management and improving outcomes for species including those of conservation concern. Bird observations were collected during the breeding season using autonomous recording units over multiple days in plots stratified by forest type and development stage. We extracted environmental variables from forest resource inventory, a digital elevation model, and LiDAR data and classified variables as informing composition, structure, and landscape. The landscape variables were made up of composition and structure variables measured at extents beyond the sampling area. The AIC supported model for Canada warbler included percentage of softwood (+), wetness (+), understory cover(+) and heterogeneity of canopy height (+). However, some of these relationships were weak, with 95 % confidence intervals that included zero. The AIC supported models for olive-sided flycatcher included percentage of softwood (+), understory(+), wetness(+), stand height variation(+) and heterogeneity at 6 m height (-), although none of the variables had 95 % confidence intervals excluding zero. Models for both species were able to effectively classify habitat suitability, and met the criteria for acceptable classification performance. For both species, none of the supported models included landscape level variables. We recommend monitoring areas where the occupancy is predicted to be the highest and uncertainty is lowest to improve identification of occupied habitat. This habitat should then be integrated into forest management planning to support the persistence of these species. Results from the models also suggest habitat features that can be created through forest management to increase the overall amount of good quality habitat for these species.