Estimating relative density of an invasive ungulate in a biodiversity hotspot using drone-based thermal video surveys

New Caledonia’s unique terrestrial habitats – primarily mountain rainforest and sclerophyll forest – face significant threats from invasive ungulates, particularly Javan deer (Rusa timorensis), introduced in the 19th century and now widespread. Since deer are an important game and food source for local communities, the management strategy aims to mitigate their detrimental impact on ecosystems while maintaining populations at levels that continue to support hunting. To achieve these objectives, effective and reliable population monitoring methods are essential. Unmanned aerial vehicles (drones) equipped with thermal sensors represent a potentially superior alternative to conventional ground-based methods for ungulate inventories, particularly in remote and difficult-to-access areas. In this study, we explored the feasibility of using a rotor-wing drone and a thermal camera to estimate the relative density of Javan deer in two protected areas: Domaine de Déva (7,319 ha) and Parc provincial des Grandes Fougères (8,098 ha). Visual line-of-sight flights were conducted after sunset at altitudes of 80 and 100 m above ground level, with the camera fixed at 40° or 0° angles. In Déva, we surveyed 10 sampling blocks, and in Grandes Fougères, 4 blocks (each ranging from 56 to 92 ha). In the predominantly open savanna area of Déva, the estimated relative population density was 116 deer/km2 (SE = 26.8), with some blocks exceeding 400 deer/km2. In contrast, in the dense canopy rainforest of Grandes Fougères, the relative density was 7 deer/km2 (SE = 2.8), with a local maximum of 18 deer/km2. Differences in deer counts between consecutive flights over the same blocks (with time gaps of less than 100 minutes) were minor, demonstrating that drone surveys are highly repeatable – an essential quality for a reliable population monitoring program. To detect a 25% population reduction with a statistical power of 0.8, surveying 10 sampling blocks was sufficient in the high deer density savanna area, whereas approximately 28 blocks would be required in the rainforest. The logistics of drone operations were relatively straightforward in the savanna; however, in the rainforest, we encountered practical difficulties, including a limited number of suitable take-off sites and restricted visibility due to the dense canopy cover. Despite these obstacles, the method proved to be an effective and efficient approach for monitoring deer populations in the challenging landscapes of New Caledonia.