Data for: Global change risks a threatened species due to alteration of predator-prey dynamics

Datasets generated and analyzed within the study area located in the
Côte-Nord region of Québec, Canada. To identify species-specific
movement rules that were implemented in the IBM, we used empirical data
collected for caribou, moose, and wolves over the study area.
"DataFinal_SSF_Species_season.csv" (6 files) were used to
develop Step Selection Functions for caribou, moose, and wolves to assess
habitat selection. SSFs were estimated from data for the real animals and
provide the relative probability of selection among a set of options based
on the comparison of observed and random steps (i.e., the linear segment
between successive locations at 8-h interval) using conditional logistic
regression (Fortin et al. 2005). Details on GPS data and SSF models can be
found in the article in Appendix S1: Section S2.  SSFs compare
resource characteristics of observed (scored 1) and random (scored 0)
locations presented in column case. Habitat characteristics (columns
conif_dense, conif_open, mixed, open, other, fire010, fire1020, fire20,
cut010, cut1020, cut20) was extracted from the Canadian National Forest
Inventory (NFI) forest cover maps. Land cover maps were updated every year
by adding roads, recent (<5 years), regenerating (6–20 years) and
old (21–50 years) cutblocks/fires based on information provided annually
by local forestry companies and from the Canadian National Fire Database
(CNFDB).  Columns dist0_0.25, dist0.25_0.50, dist0.5_1.00,
dist1.00_1.5, and dist1.5 are a set of 5 dichotomous covariables
representing the classes of distance to the nearest road (i.e., 1) ≤250 m,
2) 251–500 m, 3) 501–1000 m, 4) 1001–1500 m and 5) >1500 m as the
reference category). "DataFinal_IBM_Caribou_Season.csv" (2
files) corresponded to the IBM outputs with the proportion of caribou
agent killed (Prop.Caribou_killed, number of caribou killed/total number
of caribou), in function of the different scenarios
(CC,LUC,Year,Season,Scenario) and the response (Behavioral-Numerical
responses or Behavioral response). The columns Prop.CutsRoads, Prop.Fire,
Prop.Broadleaf, Homogenization, Isolation correspond to the different
variable we tested to predict the cumulative impact of anthropogenic
disturbance and climate change. To explore how changes in forest structure
and composition impacted the proportion of caribou killed, we used the
proportion of areas disturbed by cuts and roads (Prop.CutsRoads), burned
areas (Prop.Fire), and landscape characteristics, such as the proportion
of deciduous vegetation (Prop.Broadleaf), landscape homogenization
(Homogenization) and isolation (Isolation) of mature conifer stands.