Detection of predator-free New Zealand 2050 mammalian pest species with thermal AI cameras using a range of audio lures : A dissertation submitted in partial fulfilment of the requirements for the Degree of Bachelor of Science with Honours at Lincoln University

Abstract

Limitations identified in current best-practice pest monitoring tools have directed research to seek alternatives capable of overcoming key management concerns. Some promising options identified include – thermal cameras, a more sensitive alternative to trail cameras; audio lures to increase the conspicuousness of monitoring devices and invoke interactions; and automatic classification AI to identify target species from in-field camera footage, reducing time and costs associated with manually analysing camera data. This research aimed to further assess these three novel approaches through trialling the Project Cacophony AI thermal camera and sound lure device on free-ranging animals in regenerating native forest. Three categories of animal sound lure, containing three different noises were trialled – including possum (Trichosurus vulpecula), rat (Rattus spp.), and calls from three species of common avian prey present in the area (chicken chicks (Gallus gallus domesticus), bellbird (Anthornis melanura), and fantail (Rhipidura fuliginosa). The results showed a highly significant interaction between the pest species detected and the sound category of the audio lure played (P<0.0001). Possums were the most detected pest in the area (n = 104, 59, 76 for each monitoring trial). They responded predominantly to possum and prey noises – within those respectively, the shrill possum call and fantail call were the most favoured. Possums generally showed inquisitive and exploratory behaviour, with some investigating the equipment closely and staying by the monitoring stations for multiple playbacks of the audio lure. Rodents were the second most detected pest (n = 102, 20, 80 for each monitoring trial). However, individuals passed by the monitoring equipment at a distance and did not interact with it. This is consistent with behaviour observed in wild and captive rats and mice, which tend to be very sensitive and even show neophobia to novel stimuli, such as unnatural noises. Rodents favoured possum and prey noises overall; however, they preferred the slow and quiet possum noise and the bellbird call. Rabbits and hedgehogs were detected in far fewer numbers in all three trials (n = 5, 6, 12 and n = 3, 12, 14, respectively) and were observed to survey the area around the monitoring station but not interact directly with the equipment. Cats were only detected in the possum and rat audio trials (n = 1, 12) from a distance and showed no interest in the monitoring stations. These species responded sparingly across the trials and showed no significant difference in response levels between the sound categories, making it difficult to estimate their audio lure preferences. Although limitations with time and sample size were faced in this study, the results provide a basis for improving the methodology of assessing free-ranging pest behaviour using thermal AI cameras and animal audio lures. There is still much to explore using these technologies, as limited research has been conducted in a pest management context globally. However, preliminary research in New Zealand shows that these novel approaches are worthwhile avenues to investigate for developing an extensive range of control options to achieve a predator-free status.