Towns and Trails Drive Carnivore Connectivity using a Step Selection Approach

Abstract

Global increases in human activity threaten connectivity of animal populations. Protection and restoration of animal movement corridors requires robust models to forecast the effects of human activity on connectivity. Recent advances in the field of animal movement ecology and step selection functions offer new approaches for estimating connectivity. We show how a combination of hidden Markov movement models and step selection functions can be used to simulate realistic movement paths with multiple behavioral states. Simulated paths can be used to generate utilization distributions and estimate changes in connectivity for multiple land use scenarios. We applied movement models to 20 years of grizzly bear (Ursus arctos) and gray wolf (Canis lupus) data collected in and around Banff National Park, Canada. These carnivores avoided areas near towns in all seasons, avoided areas of high trail density in most seasons, and campgrounds during summer and fall. We simulated movement paths for three landscape scenarios: reference conditions with no anthropogenic development, current conditions, and future conditions with expanded town footprints and trail networks. We counted the number of paths that crossed valley-wide, digital transects through mountain tourist towns of Banff and Canmore, Alberta. We divided current and future crossing rates by the reference crossing rates to estimate connectivity. Current connectivity rates ranged between 7 and 45% of reference values with an average of 21% for grizzly bears and 25% for wolves. Potential town expansion and increased development of trails further decreased connectivity an average of 6% in future scenarios. Anthropogenic developments reduced the amount of available high quality large carnivore habitat in the Bow Valley by an average of 14% under current conditions and 16% under future conditions. Our approach for estimating connectivity provides a robust and flexible method for combining movement models with step selection analyses to estimate connectivity for a variety of species.

Predicting areas important for ecological connectivity throughout Canada

“…For the buffer area outside of Canada, we used six layers from the Global Human Footprint [38] (GHF): built environments, nighttime lights, croplands, pasturelands, railways, and roads. In addition to human modified land cover features, we included natural features considered to affect the movement of terrestrial, non-volant fauna, namely elevation and slope, glaciers, permanent sea ice, as well as large lakes and rivers [6,[39][40][41][42] for within both Canada and the buffer areas in the United States of America (hereafter, U.S.). Vector layers (e.g., roads, railways, rivers) were rasterized and all input layers were resampled to a resolution of 300 x 300 m, where necessary, to match the resolution of layers from the CHF.…”

“…High movement costs were also assigned to natural features known to inhibit the movement of terrestrial, non-volant fauna based on data from published studies and consultations with experts (S1 Table ). These included areas with high elevations (> 2300 m [40,41,45]), steep slopes (> 30 degrees [39,42,[45][46][47]), large lakes (> = 10 ha) and rivers with large flow (> 28 m 3 /sec [6,34,48]). Conversely, because some mammal species are known to travel across permanent sea ice among islands in the Canadian Arctic Archipelago, we assigned a medium-low cost to that layer [49].…”

Areas Important for Ecological Connectivity Throughout Canada

“…We used a recently developed national road layer 43 that included resource-access roads, along with a national railway layer. In addition to human modified land cover features, we included natural features considered to inhibit the movement of animals, namely elevation and slope, glaciers, lakes, and rivers 10,[44][45][46][47] . We also included a layer for permanent sea ice because it is known to facilitate connectivity among islands in the Canadian Arctic archipelago for some mammal species 48 .…”