Detection and occupancy modelling of invasive deer in Namadgi National Park

Abstract

Invasive deer in Australia have been associated with threats to native vegetation structure and composition, water quality, and economic impacts on agriculture. As deer numbers continue to grow and their range expands, management is becoming increasingly necessary, which requires a robust system of monitoring. Multiple methods have been recommended for deer monitoring, but these have not been broadly applied to forested, subalpine areas of south eastern Australia. Namadgi National Park has an established deer management program, but no landscape-scale analysis of deer occupancy has been undertaken. In this thesis I aimed to (1) compare deer monitoring via scat surveys and camera trapping to establish the most effective system for use in forested, mountainous landscapes; (2) determine factors influencing camera detections; and (3) understand deer occupancy at a landscape scale in Namadgi National Park. This research is intended to assist park managers in ongoing deer management and contribute to the current understanding of deer in less-studied forested sites and at lower population densities. To address my first aim, I compared camera trap records from November 2024 to April 2025 with 90 scat surveys at 45 locations in three different vegetation formations. Only 15 scats were found at 5 sites, while cameras detected deer at 23 sites, demonstrating that field based surveying was much less effective than camera trap monitoring in this landscape. To address the second aim, I developed detection-occupancy models on both my 45 field sites and the full 73-camera array using the �unmarked� package in R, focusing on detection factors which had been identified in the literature. Current low deer density in the park meant that fallow and red deer detections were low, so modelling complexity for those species was limited. Midstory density, measured along the scat transects, was significant to sambar detection in the 45-site model, and remote-sensed canopy cover was significant to sambar and fallow deer in the 73-site model, though sambar detectability decreased with increased canopy cover, and fallow deer detectability increased with increased cover. To address my third aim, both the 45-site and 73-site models were evaluated for occupancy predictors. Distance from water, or elevation, was the main covariate in all three species models, and was particularly strongly associated with fallow deer occupancy. As this modelling was single-season only, however, focusing on the summer, it could be expected that the effect of water proximity would be lower for all three species in winter and spring modelling when water is more available in the landscape. Deer populations in Namadgi National Park are currently low density, but will expand without management. My research can inform both monitoring practices within the park and priority areas for ongoing deer management, and can be used more broadly by land managers engaged in the monitoring and management of deer at low densities in similar forested locations across south-eastern Australia.