Candidate model ranking based on QIC for each period of the day and the year.

Models were evaluated using conditional logistic regressions. Only models with ΔQIC≤6 are presented. Number of parameter (K), log-likelihood (LL), difference in QIC values (ΔQIC) and weight (ωi) are given. Model performance was assessed with a Spearman rank correlation (rs±sd). Elevation variables were included in all models tested and models without interactions (i.e. densities of clearcuts edges and roads) are identified with a *.