Skip to main content

Table 1 Comparison between preferred models (∆AICc < 4) and simpler models including only habitats and goose nest density, for arctic fox resource selection functions (RSF). Models are presented for two fox behavioural states and two study periods. Note that no prediction was made regarding goose nest density for foxes resting during goose brooding (Table S1.2). Habitat selection predictors are as follows: distance to territory edges (Edge), distance to the main den (Den), goose nest density (Geese), tundra habitats (Habitat), and reproductive status (Repro)

From: Prey and habitat distribution are not enough to explain predator habitat selection: addressing intraspecific interactions, behavioural state and time

Model

k

-LL

∆AICc

wi

\( \overline{{\boldsymbol{r}}_{\boldsymbol{s}}} \)

(a) Fox active / Goose incubation

 Edge + Den X Repro + Habitat + Geese X Habitat (Complex wetlands)

12

28,529.79

0.00

0.54

0.93 ± 0.02

 Edge + Den X Repro + Habitat + Geese

11

28,530.95

0.32

0.46

0.93 ± 0.03

 Habitat + Geese

7

29,131.90

1194.22

< 0.001

0.73 ± 0.08

(b) Fox resting / Goose incubation

 Edge + Den X Repro + Habitat + Geese

11

22,638.61

0.00

> 0.99

0.86 ± 0.04

 Habitat + Geese

7

25,248.15

5211.07

< 0.001

0.76 ± 0.07

(c) Fox active / Goose brooding

 Edge + Den X Repro + Habitat + Geese

11

22,476.11

0.00

> 0.99

0.80 ± 0.07

 Habitat + Geese

7

22,986.99

1013.75

< 0.001

0.57 ± 0.18

(d) Fox resting / Goose brooding

 Edge + Den X Repro + Habitat

10

16,630.88

0.00

> 0.99

0.97 ± 0.02

 Habitat

6

18,139.79

3009.82

< 0.001

0.76 ± 0.08

  1. k Number of parameters, −LL Negative log-likelihood, ∆AICc Difference in AICc compared to the most parsimonious model, wi AICc Weight of evidence, \( \overline{r_s} \) Mean k-fold cross validation correlation coefficient ± SD