Relative probability of elk occurrence in predator scats
To predict the relative probability of elk occurrence in a predator-specific scat (Pij ), we contrasted locations of scats containing elk (n = 157: 24 bear, 75 wolf, 42 coyote, and 16 cougar) to locations of a larger set of scats (n = 870: 257 bear, 363 wolf, 223 coyote, and 27 cougar) from the same predator species collected but not analysed for prey contents (Spilker, 2019), similar to a use/available design (Manly et al., 2002). By using scat locations as our available locations (rather than random locations), we controlled for the influence of landscape features on where predator scats were located. We avoided a presence-absence of elk in the scat design because the methods we used to detect elk in scats could not reliably confirm absence of elk. We used a model selection approach based on a ΔAICc > 4 to determine the model with the most support. We determined model covariates as the density or proportion within a buffer around a scat whose radius was derived from the mean gut passage time and movement per day (i.e., 3-km for wolf, 1.5-km for bear, and 2-km for cougar and coyote, Appendix S1, Supporting Information).
Covariates included green herbaceous biomass (g m-2) at the peak of the growing season (1 August), derived from a general linear model based on field sampling of 983 sites across the summer extent of the Ya Ha Tinda elk herd (Hebblewhite et al., 2008), and updated for changes in forage availability caused by climate, timber harvest, and fires (Smolko, 2014; MacAulay, 2019). Because elk must use the area to be found in the scat and encounters with elk are likely related to their intensity of use of an area, we also quantified relative intensity of elk use using a population-level resource utilization function (RUF; Marzluff et al., 2004) based on utilization distributions from 6-hour GPS relocations of 66 adult female elk during 2013 – 2016. Variables in the RUF included herbaceous forage biomass, herbaceous land cover (Hebblewhite et al., 2008), distance to nearest forest edge, burned vegetation (< 15 years), and wolf (Hebblewhite & Merrill, 2007) and grizzly bear predation risk (Nielsen et al., 2002). For more detail, see MacAulay (2019).