Ecological Archives A022-060-A2

Nicholas J. DeCesare, Mark Hebblewhite, Fiona Schmiegelow, David Hervieux, Gregory J. McDermid, Lalenia Neufeld, Mark Bradley, Jesse Whittington, Kirby G. Smith, Luigi E. Morgantini, Matthew Wheatley, and Marco Musiani. 2012. Transcending scale dependence in identifying habitat with resource selection functions. Ecological Applications 22:1068–1083.

Appendix B. Description of GIS-based spatial resource data.

We used a suite of raster data sets to characterize spatial variation in anthropogenic, climatic, vegetative, and topographic resources found to be important in previous caribou ecology research (Table B1). We compiled spatial layers of forestry cut-blocks using a combination of provincial stand inventory data for British Columbia and a layer developed for Alberta which combined provincial and industry inventory data with remote sensing-based detection of landscape change (Foothills Research Institute Grizzly Bear Program, Hinton, Alberta, unpublished data). We compiled linear feature vector data using a combination of seismic line data digitized from orthoimagery by both ourselves and Alberta Sustainable Resource Development (ASRD) and trail data developed by Parks Canada and ASRD. Attributes of cut-block data included the year of creation, which allowed us to backcast annual landscapes of cut-block density for each year of study (1998–2009). For S₂ and S₃ analyses we used cut-block density layers specific to the year of capture for each animal, and for S₁ analyses we used the cut-block density layer specific to 2004, which approximated the conditions present concurrent with the mean start date of monitoring across radio-collared caribou (6 Jan 2004). We also used the cut-block layer specific to 2004 for preliminary analyses concerning the most predictive radii of feature density and for model validation. Linear feature data did not include reliable time attribute data and were thus treated as constant throughout the study period using 2007 vector data provided by ASRD and data digitized from 2005 SPOT imagery in British Columbia.

We used remotely sensed Moderate Resolution Imaging Spectroradiometer (MODIS) data to characterize seasonal snow cover and vegetation biomass at 500 and 250 m resolution, respectively. We used MODIS10A2 data to estimate seasonal (winter and summer) mean percentage of days each pixel was covered by snow. We used MODIS13Q1 data to estimate the normalized difference vegetation index (NDVI), an index of vegetation productivity from -1 to 1, that has been predictive of ungulate ecological responses including reindeer body mass (Pettorelli et al. 2005a,b). Snow coverage was estimated as the seasonal average per pixel across a 10 year period from 2000-2009, and NDVI as the seasonal average across a 5 year period from 2005-2009.

To categorize vegetation across our study area, we began with continuous and categorical mapping products derived from Landsat 5 Thematic Mapper (TM) or Landsat 7 TM sensors as described by McDermid (2006), and including categorical rasters of land cover types, forestry cut-blocks, and burned areas and continuous rasters of percent conifer species and percent canopy closure within forested types. Caribou response to non-forest vegetation may differ in alpine versus subalpine regions (Gustine et al. 2006), so we estimated the elevation of the alpine treeline across the study area with polynomial regression of the maximum elevation of forested pixels and latitude (Paulsen and Körner 2001). We then assigned an "alpine vegetated" land cover type to those pixels with elevations within or above 200 m of treeline and designated as shrub or herbaceous vegetation according to pre-existing layers (McDermid 2006, McDermid et al. 2009). We also used province-maintained vector layers of burns ≤ 60 years old and our cut-block layer described above to distinctly identify recently burned or logged forests. For S₂ and S₃ analyses we back-casted burn and cut-block layers specific to the year of capture for each animal, by re-assigning burns or cut-blocks that occurred after the animal was captured to the closed conifer forest land cover type. For S₁ analyses we used the burn and cut-block layers specific to 2004, which approximated the conditions present concurrent with the mean start date of monitoring across radio-collared caribou (6 Jan 2004). All land cover layers were consolidated into a single categorical land cover raster including 11 classes (Table B1).

We used digital elevation data to estimate elevation at a 30 m resolution. We used the Spatial Analyst extension for ArcGIS 9.3 to estimate slope (°), and a topographic position index (TPI; Jenness 2006). We also incorporated the elevation model to estimate the 3-dimensional distances to water bodies and to the alpine treeline, and we transformed distances in meters to indices from 0 to 1 using a decay function (1-exp^{-α × distance}) described by Nielsen et al. (2009). The TPI and decay functions for distances to water and treeline were scale-dependent. We estimated TPI and distance decay functions within 1 km radii for S₃ selection models (Nielsen et al. 2009; Whittington et al. In press), and we incorporated the radii found to be most predictive when estimating selection of anthropogenic feature density (as described above) for S₁ (12 km) and S₂ (5 km) models. We estimated TPI within circular neighborhoods of each radius, and we adjusted the distance decay terms (α) for each order of selection such that 95% of the possible distance index values occurred within a distance equal to the specified radii (α_S1 = 0.000250, α_S2 = 0.000599, α_S3 = 0.00300).

Table B1. Resource variables and their respective units and resolutions as measured for modeling woodland caribou resource selection across three scales of selection.

Literature Cited

Gustine, D. D., K. L. Parker, R. J. Lay, M. P. Gillingham, and D. C. Heard. 2006. Interpreting resource selection at different scales for woodland caribou in winter. Journal of Wildlife Management 70:1601–1614.

Jenness, J. 2006. Documentation for the Topographic Position Index (tpi_jen.avx) extension for ArcView 3.x, v. 1.3a. Jenness Enterprises, Flagstaff, Arizona.

McDermid, G. J., 2006. Remote sensing for large-area, multi-jurisdictional habitat mapping. Dissertation. University of Waterloo, Waterloo, Ontario, Canada.

McDermid, G. J., R. J. Hall, G. A. Sanchez-Azofeifa, S. E. Franklin, G. B. Stenhouse, T. Kobliuk, and E. F. LeDrew. 2009. Remote sensing and forest inventory for wildlife habitat assessment. Forest Ecology and Management 257:2262–2269.

Nielsen, S. E., J. Cranston, and G. B. Stenhouse. 2009. Identification of priority areas for grizzly bear conservation and recovery in Alberta, Canada. Journal of Conservation Planning 5:38–60.

Paulsen, J., and C. Körner. 2001. GIS-analysis of tree-line elevation in the Swiss Alps suggests no exposure effect. Journal of Vegetation Science 12:817–824.

Pettorelli, N., J. O. Vik, A. Mysterud, J.-M. Gaillard, C. J. Tucker, and N. C. Stenseth. 2005a. Using the satellite-derived NDVI to assess ecological responses to environmental change. Trends in Ecology and Evolution 20:503–510.

Pettorelli, N., R. B. Weladji, Ø. Holand, A. Mysterud, H. Breie, and N. C. Stenseth. 2005b. The relative role of winter and spring conditions: linking climate and landscape-scale plant phenology to alpine reindeer body mass. Biology Letters 1:24–26.

Whittington, J., M. Hebblewhite, N. J. DeCesare, L. Neufeld, M. Bradley, J. Wilmshurst, and M. Musiani. In press. Caribou encounters with wolves increase near roads and trails: a time-to-event approach. Journal of Applied Ecology.

[Back to A022-060]