Using drones to improve wildlife monitoring in a changing climate

Abstract

This thesis advances knowledge of wildlife monitoring techniques and demonstrates the potential of high-resolution, remotely sensed data to inform species conservation, improve ecosystem management and assess mitigation strategies for biodiversity loss. Drones can easily collect systematic, high spatial and temporal resolution data to detect fluctuations in key parameters such as abundance, range and condition of some species. Advances in drone-facilitated wildlife monitoring of sentinel species will provide rapid, efficient insights into ecosystem-level changes. This thesis focused on resolving knowledge gaps within three key areas of wildlife drone-ecology: disturbance, population monitoring and body condition. From the outset, we recognised drones might have undesirable or unforeseen behavioural and physiological effects on wildlife. To address this, I led a time-critical publication that advocated researchers adopt a precautionary approach given the limited understanding of the impacts. It also provided recommendations for conducting drone-facilitated research around wildlife as the basis for a code of best practice. Then, using colonial birds as a study group, we tested the utility of drone-derived data for population monitoring. First, life-sized, replica seabird colonies containing a known number of fake birds were used to robustly assess the accuracy of our intended approach compared to the traditional ground-based counting method. Drone-derived abundance data were, on average, between 43% and 96% more accurate, as well as more precise, than estimates from the traditional approach. Our open-source, semi-automated detection algorithm estimated abundance 94% similar to manual counts from the remotely sensed imagery. To apply this in the field, we collected drone-derived abundance data by repeatedly surveying representative, wild colonial birds (a tern, cormorant and pelican species). We used these data to develop a transferable technique requiring minimal user-input for adaptable and high spatiotemporal population monitoring. Finally, to investigate the use of drone-facilitated photogrammetry, we used a representative pinniped species to test if non-invasively acquired, morphometric data could infer body condition. Drone-derived measurements of endangered Australian sea lions (Neophoca cinerea) of known size and mass were precise and without bias. These two- and three-dimensional measurements from orthomosaics and digital elevation models were highly correlated with animal mass and body condition indices and not significantly different to those generated from ground-collected data. This work addresses and informs a range of issues arising from human activity in the Anthropocene, including rapid habitat loss, species extinctions and an altered climate. We have shown that using technology for wildlife monitoring enables timely, proactive environmental and conservation management.