The effects of predators, habitat complexity and water temperature on detection probability and activity patterns of a larval salamander
Analyses of occupancy and/or abundance using count data from field surveys of an organism estimate detection probabilities to correct for imperfect detection. Variation in detection probabilities can occur for many reasons, such as observer bias or sampling efficiency. Biotic processes can also affect detection including indirect interactions with other species (e.g. predator avoidance behavior). Detection and abundance models have the potential to be a unique way to address indirect interactions by using observations of natural processes (e.g. predator-prey interactions) while statistically modeling how detection changes due to these ecological relationships. We applied this modeling techniques to a predation experiment, with the expectation that predator-induced shifts of behavior could be assessed based on detection probabilities. Six combinations of three aquatic predators (newts, mosquitofish, and dragonfly nymphs) was crossed with two habitat complexity treatments within experimental pond mesocosms to estimate detection probabilities and behavioral patterns of larval ringed salamanders (Ambystoma annulatum). Multiple counts of visible salamanders were performed in each tank on 20 days over the first three months of ontogeny. The number of larvae and their position within a mesocosm were recorded, as well as the number of visible predators. Water temperature was also measured during each survey.
Mean daily detection did not vary due to either the cover or predator treatments. Detection was highest in diurnal periods of warmer water, and during nighttime surveys when water temperatures remained high due to heat retention of daily warming patterns. Predator activity did not alter larval activity patterns; larvae were actively foraged within close proximity of all three predator species. Predator detection varied between the three species due to ecological constraints and observer influence. Newts were active during both diurnal and nocturnal periods, but dragonfly larvae were most active at dusk only when water temperatures were high. Fish detection was higher at night due to observer influence during diurnal periods. Overall, it appears that larvae exhibit risky foraging behavior by increasing activity during diurnal periods that would result in higher visibility to some predators. However, this strategy may be adaptive as it would allow them to maximize foraging opportunities on diurnal prey, and grow into a size class invulnerable to gape-limited predators. Overall, this experiment provides evidence that detection models can be a useful tool in analyses of TMIs, but careful consideration of synergistic factors (e.g. temperature in this study) may be critical to elucidate mechanistic processes.