Results
Altogether, the motion sensor triggered cameras captured 24,596 and 45,387 images and the time lapse cameras 134,979 and 142,506 images in 2020 and 2021, respectively.
The best fitting count model had a negative Binomial distribution (NB1 parameterization) and it accounted for zero-inflation (zero-inflation p <0.001, Table 1). It included the fixed factors ‘trigger type’ and ‘year’ without their interaction and the random intercept of ‘site’ (random effect variance (σ2) of 0.9062 and standard deviation (SD) of 0.952). Taiga bean goose count was significantly higher in 2021 compared to 2020 (p <0.001, Fig. 3), however, the camera trigger type had no significant effect on the number of geese (p = 0.129).
As the AIC values were within <2 of each other, the best fitting logistic model on goose capture probability was selected based on parsimony and it included the fixed terms ‘trigger type’, ‘year’ and ‘effort’ without interactions and the random intercept of ‘site’ (random effect σ2 = 0.6215 and SD = 0.7883, Table 1). The capture probability increased significantly with increasing camera effort (p <0.001) and was significantly higher in the year 2021 (p <0.001) and marginally significantly higher (p = 0.049) with game cameras set to time lapse (Fig. 4).