Predator-prey temporal overlap models
For models of t0 (no lag) with 1-day and 1-week bin-width encounter histories, detections of weasels (Fig. 2a), flying squirrels (Fig. 2b), owls (Fig. 2c), and birds (Fig. 2d) were all positively correlated with tree voles. We observed consistent negative correlations among dependent and independent variables for all models with 1-hour bin-width encounter histories (Fig. 2). Tree vole detections were consistently negatively correlated with weasels in t + 1 andt + 12 models (Fig. 2a). Tree vole detections were weakly positively correlated with flying squirrels and birds digging one time-lag after these predators were detected (t + 1 ; Fig. 2). Interpreting odds ratios, we observed a nearly ten-fold decrease in odds of detecting a tree vole one time lag after a weasel was detected (t + 1 ), from approximately 2 : 1 at t0 (odds ratio = 2.1 (1.2, 3.7; 95% CrI )) to approximately 1 : 5 (odds ratio = 0.19 (0.09, 0.41)) one time-lag after a weasel was detected (t + 1 ; days), and odds of detecting a tree vole remained low 0.16 (0.13, 0.21) when modelling time steps to day 12 (t + 12 , days).
Flying squirrels were weakly but consistently positively correlated with birds digging (except for hour) across bin-widths and time-lags (Fig. 2g). One-hour bin-width showed extreme negative correlations across models except for the t0 model for owls which slightly overlapped 0 (Fig. 2e, f, g).