Appendix A: Further Details on Landscape Covariates and Spatial Data used to Model Habitat Selection and Fidelity among Central Mountain Woodland Caribou in Western Alberta and Eastern British Columbia, Canada
The list of landscape variables used in the habitat selection (Table A1) and calving range fidelity (Table A2) analyses as well as details on the associated spatial data.
Snow Cover and NDVI
We obtained daily snow cover images from Moderate Resolution Imaging Spectroradiometer (MODIS) data at a 500 m x 500 m spatial resolution (MODIS-Terra Snow Cover 5-Min L2 Swath 500, Hall et al. 2006). The daily images were four-day cloud-free composites, so that pixels identified as cloud on the current day are filled in using values from the previous three days. We calculated snow cover as the percent of days a 500-m pixel was snow covered over a 28 day period beginning at the peak calving date. Vegetation productivity was mapped at a 250-m resolution based on Normalized Difference Vegetation Index (NDVI), an index of vegetation productivity from -1 to 1, obtained from MODIS (MODIS13Q1). The NDVI data are remotely sensed and compiled over consecutive 16 day periods. For each year, we averaged NDVI values between two of these 16-day images, one from 25 May to 9 June and the second from 10 June to 25 June. Snow cover and NDVI spatial data were unavailable for all RSF years (1998-2014) so we averaged values across years, 2006-2013 and 2000-2013 respectively.
Well site data
Our study area encompassed portions of two adjacent provinces of Alberta and British Columbia, and as a result spatial data was often obtained from separate sources in each province. In Alberta, well site blocks were queried from Digital Integrated Dispositions (DIDs) provided by the Government of Alberta but compiled and maintained by Alberta Data Partnerships formerly Spatial Data Warehouse (http://abdatapartnerships.ca/) and Altalis (http://www.altalis.com/products/property/dispositions/dids.html). In British Columbia, locations of well sites were obtained from British Columbia Agency - Oil and Gas Commission. These data were publically available online and last updated on 30 January 2015 (http://catalogue.data.gov.bc.ca/dataset/ogc-well-sites-public-version). The British Columbia well site dataset contained spatial data collected through the Oil and Gas Commission’s electronic Petroleum Application Submission System (ePASS) on or after October 30, 2006. We removed well sites in the “permission to drill” and “cancellation of permission before drilling” because they were not likely to have occurred on the landscape yet. Also because the data are the center coordinates of the well site we applied a square buffer to make polygons of 1.588 ha (15883 m2) representing a well pad footprint. This buffer size matched the average footprint of well pads in the Alberta dataset.
Wild Fires
We obtained spatial representation of burned habitat from perimeters of historical wild fires from the Government of Alberta (http://wildfire.alberta.ca/wildfire-maps/historical-wildfire-information/spatial-wildfire-data.aspx) and the British Columbia Ministry of Forests, Lands and Natural Resource Operations - Wildfire Management (https://apps.gov.bc.ca/pub/dwds/viewOrder.do?orderId=1542399). The Alberta wildfire data were updated April 2013 and includes fires from 1931 to 2012 while the British Columbia data were updated April 2014 and included fires from 1917 to 2013.
Roads
The roads dataset was compiled for Alberta by fRI Research Grizzly Bear Program from several sources including Government of Alberta base features, forest access roads obtained from Weyerhaeuser Grande Prairie, and heads-up digitizing from high-resolution imagery (2.5m or less). Road features were available until and including 2013. On the British Columbia side, road features were obtained from Government of British Columbia – GeoBC’s digital road atlas (http://geobc.gov.bc.ca/base-mapping/atlas/dra/#products). The road atlas represents a compilation of spatial road data from a variety of sources including: British Columbia - Ministry of Environment, British Columbia - Lands and Parks, local governments, 911 service providers and Utilities Companies. The British Columbia road atlas was last updated 25 April 2012.
Cutblocks
In Alberta, cutblock data were provided by Weyerhaeuser Grande Prairie, the only forest management area holder to overlap the Narraway and Redrock-Prairie Creek herd ranges in Alberta. In British Columbia, cutblock polygons were obtained from Government of British Columbia - Forests, Lands and Natural Resource Operations - Forest Tenures Branch (https://apps.gov.bc.ca/pub/geometadata/metadataDetail.do?recordUID=50580&recordSet=ISO19115).
Seismic Lines
In the British Columbia portion of the study area seismic line data were obtained from a combination of British Columbia Oil and Gas Commission (well site point data; www.bcogc.ca/public-zone/gis-data) and a digitization of orthoimagery by DeCesare et al. (2012). The British Columbia data were verified and augmented with SPOT 5 imagery circa 2005 at a 5 m resolution (www.blackbridge.com/geomatics). In Alberta, seismic lines were initially identified from a base layer provided by AltaLIS (http://www.altalis.com/products/base/) then verified and augmented with LiDAR surveys between 2003 and 2008.
Vegetation Cover
Caribou habitat modelling required an accurate spatial representation of vegetation cover and forest structure. To meet this need, we utilized a vegetation cover classification based on Landsat-8 OLI spectral data and slope and elevation as ancillary datasets. The vegetation cover classification was based on the scheme used for grizzly bear in Alberta, which extensively uses LiDAR data to identify forest structure (Nijland et al. 2015).
The vegetation cover classification was based on a maximum likelihood classifier using the Visual, NIR and SWIR bands of a 2008 (12 September) Landsat-8 OLI image (Bands 2 to 7, US Geological Survey 2013), combined with slope and elevation derived from the ASTER Global Digital Elevation Model (GDEM) dataset (ASTER GDEM Validation Team 2011). The classification scheme has 15 classes as shown in Table A3. Training data were collected in the Alberta part of the study area where LiDAR based measurements of canopy height and cover were also available. The regeneration class is based on disturbances in the last ten years before 2013, and includes all recently disturbed areas originally classified as forest.
Conifer Cover was modelled based on reflectance corrected Landsat TM data from 2010 based on Tasselled Cap Brightness and Wetness (Huang et al. 2002). The layer is only provided for the forested cells in the study area, as the model produces invalid results over other vegetation cover classes.
Table A1. Variables used to model caribou (Rangifer tarandus caribou) calving site selection and post-parturition habitat selection among central mountain caribou (Narraway, Redrock-Prairie Creek) during the spring calving season in western Alberta and eastern British Columbia between 1998 and 2014. The covariates included in the analysis at the herd-range and home-range scales are marked by an x in their respective columns.
Scale of SelectionVariable / Description / Range / Herd-Range / Home-Range
Disturbance
Seismic5k / Seismic line density (km/km2) in 5-km radius / 0 – 2.4 / x
Seismic70 / Seismic line density (km/km2) in 70-m radius / 0 – 49 / x
DSeismic / Distance (km) to seismic line as decay function [1-exp (-2 x distance)] / 0 – 1 / x
Road5k / Road density (km/km2) in 5-km radius / 0 – 3.2 / x
Road70 / Road density (km/km2) in 70-m radius / 0 – 43 / x
DRoad / Distance (km) to road as decay function [1-exp (-2 x distance)] / 0 – 1 / x
Cut5k / Proportion cutblock in 5-km radius / 0 – 0.8 / x
Cut70 / Proportion cutblock in 70-m radius / 0 – 1 / x
DCut / Distance (km) to cutblock as decay function [1-exp (-2 x distance)] / 0 – 1 / x
Well5k / Proportion well site in 5-km radius / 0 – 0.02 / x
Well 70 / Proportion well site in 70-m radius / 0 – 1 / x
DWell / Distance (km) to well site as decay function [1-exp (-2 x distance)] / 0 – 1 / x
Burn5k / Proportion burn in 5-km radius / 0 – 0.8 / x
Burn 70 / Proportion burn in 70-m radius / 0 – 1 / x
DBurn / Distance (km) to burn as decay function [1-exp (-2 x distance)] / 0 – 1 / x
Topography
Elevation / Elevation (km) / 555 – 3279 / x / x
TPI1k / Difference in elevation (m) between a central cell and the mean elevation within a 1-km radius / -420 – 627 / x / x
Slope / Terrain slope in degrees / 0 – 90 / x / x
Aspect
Northness / cosine of aspect in radians / -1 – 1 / x / x
Eastness / sine of aspect in radians / -1 – 1 / x / x
Cover / x / x
Snow / seasonal % daily coverage / 0 – 100 / x / x
NDVI / Index of vegetation productivity / -1 – 1 / x / x
Vegetation Cover / See Table A3 for details / Factor / x / x
Table A2. Disturbance variables used to model caribou (Rangifer tarandus caribou) calving range fidelity among central mountain caribou (Narraway, Redrock-Prairie Creek) during the spring calving season in western Alberta and eastern British Columbia between 1998 and 2014.
Variable / Description / RangeSeismicyear1 / Seismic line density (km/km2) in 5-km radius, averaged across all GPS-collar locations in first year of calving / 0.17 – 0.37
Seismicyear2 / Seismic line density (km/km2) in 5-km radius, averaged across all GPS-collar locations in second year of calving / 0.17 – 0.23
Roadyear1 / Road density (km/km2) in 5-km radius, averaged across all GPS-collar locations in first year of calving / 0.15 – 0.33
Roadyear2 / Road density (km/km2) in 5-km radius, averaged across all GPS-collar locations in second year of calving / 0.16 – 0.35
Cutyear1 / Proportion cutblock in 5-km radius, averaged across all GPS-collar locations in first year of calving / 0.02 – 0.06
Cutyear2 / Proportion cutblock in 70-m radius, averaged across all GPS-collar locations in second year of calving / 0.03 – 0.07
Wellyear1 / Proportion well site in 5-km radius, averaged across all GPS-collar locations in first year of calving / 0.00 – 0.001
Wellyear2 / Proportion well site in 70-m radius, averaged across all GPS-collar locations in second year of calving / 0.00 – 0.001
Burnyear1 / Proportion burn in 5-km radius, averaged across all GPS-collar locations in first year of calving / 0.01 – 0.09
Burnyear2 / Proportion burn in 70-m radius, averaged across all GPS-collar locations in second year of calving / 0.01 – 0.10
Table A3. Descriptions of the original 15 vegetation cover classes derived from 8 Operational Land Imager (OLI) spectral data and the new classes they were grouped into for subsequent caribou (Rangifer tarandus caribou) calving site selection (CSS) and post-parturition habitat selection (PPHS) models.
Original Class / Description / CSS Class / PPHS ClassConifer Dense / >75% crown closure; >80% conifer / Reference / Reference
Conifer Moderate / 40-75% crown closure; >80% conifer / Conifer Open / Conifer Open
Conifer Open / 6-40% crown closure; >80% conifer / Conifer Open / Conifer Open
Deciduous Closed / >50% crown closure; >80% broadleaf / Deciduous Forest / Deciduous Forest
Deciduous Open / 6–49% crown closure; >80% broadleaf / Deciduous Forest / Deciduous Forest
Mixed Closed / >50% crown closure; 26–79% broadleaf / Mixed Forest / Mixed Forest
Mixed Open / 6–49% crown closure; 26–79% broadleaf / Mixed Forest / Mixed Forest
Herb / <25% shrub cover; <6% tree cover / Herb / Herb
Alpine Herb / <6% vegetation cover, Alpine Elevations / Herb / Herb
Alpine Bare / <25% shrub cover; <6% tree cover; Alpine Elevations / Barren / Barren
Barren / <6% vegetation cover / Barren / Barren
Water / >6% standing or flowing water / Reference / Wetland
Agriculture / Landuse: Agriculture, crops, pasture etc. / Reference / Reference
Wetland / >10% Vegetation Cover; ‘wet’ or ‘aquatic’ moisture regime / Reference / Wetland
Regeneration / >25% shrub or tree cover; Canopy < 5m height / Regeneration / Regeneration
Literature Cited
ASTER GDEM Validation Team. 2011. ASTER Global Digital Elevation Model Version 2 – Summary of Validation Results [online]. Available from https://lpdaac.usgs.gov/dataset_discovery/aster/aster_products_table/astgtm [accessed 29 September 2016]
DeCesare, N. J., Hebblewhite, M., Schmiegelow, F., Hervieux, D., McDermid, G. J., Neufeld, L., Bradely, M., Whittington, J., Smith, K.G., Morgantini, L.E., Wheatley, M., and Musiani, M. 2012. Transcending scale dependence in identifying habitat with resource selection functions. Ecol. Appl. 22: 1068–1083.
Hall, D. K., Riggs, G. A., and Salomonson, V. V. 2006. MODIS/Terra Snow Cover 5-Min L2 Swath 500. Version 5. Boulder, Colorado, USA.
Huang, C., Wylie, B., Yang, L., Homer, C., and Zylstra, G. 2002. Derivation of a tasselled cap transformation based on Landsat 7 at-satellite reflectance. Int. J. Remote Sens. 23: 1741–1748.
Nijland, W., Coops, N. C., Nielsen, S. E., and Stenhouse, G. B. 2015. Integrating optical satellite data and airborne laser scanning in habitat classification for wildlife management. International Journal of Applied Earth Observation and Geoinformation, 38: 242–250.
US Geological Survey. 2013. Landsat 8: U.S. Geological Survey Fact Sheet [online]. Available from http://pubs.usgs.gov/fs/2013/3060/ [accessed 29 September 2016]