Abstract:
Invasive rats are one of the major threats to native species and ecosystems worldwide. Reliable rat detection is essential to evaluate management, confirm eradication success or detect invaders. Detection relies on interactions between rats and detection devices. Therefore, it is important to understand rat behaviour around detection devices, particularly changes in response to ongoing control. I used camera traps alongside traditional detection devices to answer i) how camera traps can be used to estimate abundance of invasive rats and ii) which factors affect the behaviour, and hence detectability, of invasive rats.
On Goat Island, New Zealand, self-resetting Goodnature A24 rat traps were monitored with camera traps. Counting the number of rat videos was the preferred method when measuring relative abundance and did not result in loss of information compared to the more laboriously counting of individual rats. Estimating density requires identification of individuals. I tested three novel statistical models to estimate density of unmarked animals using camera traps and compared the results to spatially explicit capture-recapture (SECR) estimates from live trapping data in French Polynesia. The spatially explicit Chandler model underestimated density and the estimates were confounded by low precision, presumably due to an insufficient density of detection devices. The random encounter and staying time model had a similar density to SECR in one session but underestimated density in another session. Differences in rat activity and behaviour have influenced density estimates. The random encounter model failed to provide realistic density estimates. The model needs information about rat movement speed and day range, which were not available and difficult to measure. A different random encounter model (REST) obtained information about rat movement speed from the time an individual stays in the field of view, resulting in more precise density estimates. Camera traps were shown to be useful to measure rat relative abundance even when A24s failed to detect the remaining individuals after sustained control.
To understand the behaviour of rats around detection devices, both live traps and Goodnature A24s were monitored with camera traps. Detection probability using live traps for Pacific rats differed between sites and was lower where ship rats were present, and Pacific rat detection was delayed where both species coexisted. Rat density and detection probability varied between times, presumably due to seasonality in food abundance and rat reproduction. Higher rat abundance resulted in more trap encounters and interactions. However, the trapping rate remained low throughout the study. Detectability was influenced by season, abundance and interspecific competition. To improve rat detection from devices that require interaction with a trigger mechanism either the threshold for rats to interact with the trigger must be lowered or attractiveness of the bait must be improved to overcome trigger avoidance.