Factors influencing detection probability
Model selection suggested bait-type (fish versus honey) without temporal
trend using a half-normal detection function was the best fitting and
outperformed alternate temporal and detector-level models based on AICc
(Table 1). Detection probabilities at g0 =0 based on this model
were far higher for fish-bait (0.33 [95% C.I. 0.25-0.43]) compared
to honey-bait (0.01 [95% C.I. 0.01-0.03]) (Fig. 2). Alternate
models that included temporal effects of weather and rebaiting periods
were not found to be as influential as bait-type (Table 1). Likewise,
detector-level covariates (i.e., tree species, tree height and trap
height) were also found to not be as important as bait-type.