Estimating red deer <i>Cervus elaphus</i> population density using drones in a steep and rugged terrain

Precise and accurate information about population density, crucial for wildlife management, is difficult to obtain for elusive species living in dense forests or steep and inaccessible terrain. Using unmanned aerial vehicles (UAVs), we developed a method for obtaining absolute population estimates of ungulates living in steep, rugged, and partly forested terrain. The UAVs were used to collect data from four different case study areas in two consecutive months in winter/early spring (before leaf growth) in Lærdal, Norway. The detections of red deer Cervus elaphus from UAV images were done using a double observer protocol, and hierarchal state–space models were fitted to the data using a Bayesian framework to obtain posterior distributions of absolute deer density. Results indicate that availability bias was negligible when deer can be searched for in multiple UAV images of the sites (sample plots) taken from different angles by a UAV at 60 m above ground level but detection probability in the images was significantly below one and must hence be accounted for. Although the credible intervals were rather wide, our results demonstrate that absolute red deer density estimates with quantifiable uncertainties can be produced using this method. Lower credible limits for the estimates indicate that densities in two of the study areas were consistently well above management target densities during the two consecutive months. The high spatial variation in deer densities (at various scales) in our study pose a major challenge, as large areas with a high number of sampling sites must be covered to model the spatial variation and obtain reliable and precise density estimates at a scale relevant for management. We nevertheless conclude that the method has potential for reliable spatiotemporal deer density estimation, given sufficient resources and field efforts.