A project undertaken at the University of New South Wales, and supervised by Dr Neil Jordan.
Dingoes are an iconic and controversial part of the Australian landscape. Interactions with humans and livestock, or the potential impacts of these, have led to substantial broad-scale management efforts. Dingo management includes widespread culling and the construction and maintenance of the world’s longest exclusion fence. Dingoes also play a key role as apex predators, potentially suppressing populations of herbivores and smaller introduced predators, and are a culturally significant animal in some contexts. Balancing the benefits of apex predators while minimizing the costs of living alongside them requires a more nuanced management approach.
Motivated by the intent to develop a tool to reduce negative impacts at a local scale, while maintaining the ecological and cultural benefits of retaining dingoes in the broader landscape, we tested whether dingo signals could deter dingo from specific areas.
In common with some other canids, dingoes emit a long-range “howl” to communicate their presence in an area (see image 1). To test whether these howls are effective territorial signals, and specifically whether howls could be used to deter dingoes from target areas, we established 12 remote playback stations in the coastal forest of the Myall Lakes region of New South Wales (see image 2). In successive months we played back three treatments (during three 90-minute windows each night). Treatment order was balanced across the experiment, before being randomly assigned to each site. For one month, we played back dingo howls, one month we played ambient recordings from the area, and another month we played back nothing, but left the equipment in place.
The playback station was positioned on a crossroads in the forest, and camera traps were placed there and 600m and 1.2km away on each of the four trails going away from it (see image 4). During each treatment, we recorded any dingoes in the area using these camera-traps, and predicted that we would detect less dingoes during the howl playbacks. Logger files also recorded the exact time of the playback, and songmeters validated the occurrence of the playback and recorded any response vocalisations from dingoes in the environment.
Camera-traps take pictures when they detect an animal moving by, but can also be triggered by moving grass and changes in temperature. Our approach generated >1.2million images, within which we identified approximately 60,000 images containing animals using an Artificial Intelligence (AI) approach – in this case the Microsoft Megadetector algorithm. These animal images were uploaded to a custom project (“Dingo? Bingo!”) we created on the Zooniverse citizen science platform, and >5000 volunteers worked through these images classifying to species level any animals that they detected (see image 5). To ensure accuracy of results, each image was classified by 10 independent volunteers, and the research team reviewed any images where there was less than 70% agreement between volunteers as to the species captured. Analysis is in progress.
The coastal forest contains a rich diversity of animals, and many might be affected by dingo signals. In addition to testing whether the howls deterred dingoes from the area, this dataset will also allow an assessment of the impact on the detection of other species in the area, consideration of which is important from a management perspective and also in understanding how the signals of an apex predator can affect their prey and competitors. Understanding these impacts on dingoes and other species is a key part of this project. Ultimately this project will determine whether howls can be used to deter dingoes from a given area, and will contribute to our understanding of animal communication and its potential role in animal management and conservation.