A quantitative approach to conservation planning: using resource selection functions to map the distribution of mountain caribou at multiple spatial scales

Abstract

Summary 1.Visualizing the distribution of rare or threatened species is necessary for effective implementation of conservation initiatives. Generalized linear models and geographical information systems (GIS) are now powerful tools for conservation planning, but issues of data availability, scale and model extrapolation complicate some applications. 2. Mountain caribou are an endangered ecotype of woodland caribou Rangifer tarandus caribou that occurs across central and southern British Columbia, Canada. Currently, conservation professionals use coarse small-scale maps of important habitats to manage forest harvesting and human access across the northern extent of mountain caribou range. These maps were produced before the advent of readily available digital spatial information and are based on expert opinion and limited empirical data. 3. With the purpose of refining existing maps, we used survey results, radio-telemetry locations and GIS data to construct resource selection functions (RSF) that quantified the habitat affinities and predicted the relative probability of occurrence of mountain caribou at two spatial scales. At the scale of the patch, the most parsimonious RSF model consisted of covariates for vegetation and aptly predicted the occurrence of caribou across low-to mid-elevation habitats, but performed poorly across steep alpine terrain. At the landscape scale, a model containing Gaussian terms for elevation and slope was effective at predicting the broader distribution of caribou. 4. We produced a map consisting of the product of the relative probabilities of the patch and landscape RSF. The final map represented the relative probability of occurrence of caribou in vegetative patches weighted by the relative probability of occurrence across the larger study area. We found strong agreement between current definitions of important caribou habitats developed from expert opinion and RSF-based maps generated from empirical data. 5. Synthesis and applications . Both expert opinion and RSF-based approaches offer unique advantages for conservation mapping. Interpretability of results, documentation and repeatability of methods and data, estimates of precision and costs should all be considered when evaluating a technique. We argue that for some species and geographical locations, RSF is a superior technique, but expert opinion should play a role in model development and interpretation.

Modeling influences on winter distribution of caribou in northwestern Alaska through use of satellite telemetry

“…Looking across the northern landscape, caribou ecotype and disturbance (e.g., wildfire and/or industrial development) are also likely to be important factors in determining distribution (Mallory & Hillis, 1998;Johnson et al, 2005;Joly et al, 2007a). The importance of each factor is likely to depend on the scale of the analysis (Wiens, 1989;Rettie & Messier, 2000;Johnson et al, 2004;Gustine et al, 2006;Mayor et al, 2007). Terrain, snow conditions, habitat characteristics and predation pressure are all interrelated to some degree.…”

A synthesis of scale-dependent ecology of the endangered mountain caribou in British Columbia, Canada

“…The most parsimonious landscape-level model included slope and elevation, whereas the patch-scale model contained vegetation cover types extracted from GIS. Unlike Apps et al (2001), Johnson et al (2004) found that stand age was not a predictive variable in their patch-scale analysis. The authors note that the homogenous distribution of old forest in their study area precluded the opportunity for selection, but they expected that if the area was subject to logging or wildfire, age would likely become predictive at this scale (Johnson et al, 2004).…”

“…Unlike Apps et al (2001), Johnson et al (2004) found that stand age was not a predictive variable in their patch-scale analysis. The authors note that the homogenous distribution of old forest in their study area precluded the opportunity for selection, but they expected that if the area was subject to logging or wildfire, age would likely become predictive at this scale (Johnson et al, 2004). Ultimately, the patch and landscape maps were combined to provide a map that was used in conjunction with other information by land-use planners and caribou recovery groups (RIP, 2005).…”

“…However, because not all old stands are protected within caribou habitat in the Columbia Mountains, managers still lacked a tool to identify stands that deserved priority for retention (see 'Fine scale' section). Johnson et al (2004) incorporated the telemetry data of Terry et al (2000) in a GIS environment to evaluate mountain caribou habitat relationships for the North Cariboo Mountains and Hart Ranges east of Prince George. The objectives were similar to the Apps et al (2001) study, whereby the intent was to confirm or modify existing land-use maps that defined caribou management zones.…”

“…The objectives were similar to the Apps et al (2001) study, whereby the intent was to confirm or modify existing land-use maps that defined caribou management zones. They used both census and telemetry information to derive products at the patch and landscape scale (Johnson et al 2004). The most parsimonious landscape-level model included slope and elevation, whereas the patch-scale model contained vegetation cover types extracted from GIS.…”