Predicting Habitat Value for Elk in the Central East Slopes of Alberta
Shevenell M. Webb and Robert B. Anderson
In collaboration with the University of Alberta and Alberta Sustainable Resource Development (ASRD), the Alberta Conservation Association (ACA) facilitated the development of a geographic information system (GIS) habitat‐disturbance planning tool that incorporated information from a five year wolf (Canis lupis) and elk (Cervus elaphus) radio telemetry study in the Clearwater forest of west‐central Alberta. The specific objectives of this project were to: (1) develop a user‐friendly GIS‐based Elk Tool that can be used to evaluate the influence of proposed landscape treatments on elk
occupancy and survival; (2) test the Elk Tool by predicting the effect of prescribed burn treatments on elk habitat in the R11 Forest Management Unit (FMU); (3) evaluate whether the models generated by the Elk Tool could be extrapolated to geographic areas outside of the original test area, but still within the Foothills Natural Region; and (4) evaluate the efficacy of using remote trail cameras for detecting elk occupancy and therefore validating Elk Tool predictions in the expanded study area.
Ultimately, ACA was interested in producing a practical application from the long‐term elk study in the East Slopes that would be useful to managers and project partners. ACA worked with the University of Alberta and Foothills Research Institute (FRI) to incorporate rigorous resource selection function (RSF) models into a GIS tool that predicts current and proposed elk habitat suitability as a result of alternative treatment or industrial disturbance scenarios. The tool allows the user to conduct scenario
evaluations by defining the study area extent within the Clearwater forest, then adding new roads, seismic lines, well sites, cutblocks, and/or burns to the landscape. Eight total maps are created each time the tool runs; elk occurrence, wolf occurrence, elk survival, and elk habitat states are predicted for both summer and winter seasons. These habitat states integrate the predicted relative occurrence and survival of elk, thereby delineating primary and secondary sink and source habitats across the landscape. As with all models, these scenarios are forecasts of possible outcomes and
not a guarantee of any particular outcome.
A suite of proposed prescribed burn units in the R11 Forest Management Area provided a useful case study to verify that the tool was working and to evaluate the potential effects of different scenarios on elk habitat. Overall, the tool predicted an increase in source habitat for elk by 3 ‐ 4% or an additional 17 ‐ 24 km2 (winter) and 28 ‐ 47 km2 (summer) of source habitat in the R11 FMU. In the Cline River watershed subbasin (05DA), which includes the large proposed Upper North Saskatchewan Prescribed Burn, elk source habitat was predicted to increase by 15 ‐ 22%. However, we found that there was little variation in the amount of source habitat that each hypothetical burn treatment created. In general, placement of burns in secondary source habitats away from roads will most likely increase habitat for elk.
We found that landcover layers used to build the RSF models in the Clearwater forest were classified differently from the available layers in the expanded Foothills Natural Region. After attempting to translate FRI Grizzly Bear Program landcover classes into those required by the elk models, we compared tool predictions at different spatial scales using the original and expanded landcover layers. We concluded that the FRI and Central East Slopes Wolf and Elk Study (CESWES) landcover layers and the predictions generated from them were statistically and biologically different.
Therefore, we do not recommend expanding the Elk Tool beyond the original study area extent at this time.
Finally, although we had a small sample size, we determined that remote trail cameras did not show promise as an efficient method for detecting elk occurrence in our study area. On average, each camera was activated for 29.94 ± 0.96 nights, producing a combined total of 1,501 photos in 479 camera nights. Although variation was high, cameras located in primary habitats tended to take more wildlife photos (n = 10 cameras, mean ± SD = 117.8 ± 67.67 photos) than those in non‐primary habitat locations (n = 6 cameras, mean ± SD = 55.67 ± 53.82 photos); however, all photos were of nontarget species. Trail cameras were useful in passively detecting wildlife species such as
deer, coyotes, moose, and fox and operated successfully in a variety of winter temperatures and weather conditions. We recommend the use of bait and/or lures, increasing the number of cameras and stations, and setting cameras where animal movement may be restricted to specific travel corridors to increase the likelihood of capturing wildlife photos.