Feasibility study of UAV based ecological monitoring and habitat assessment of cervids in the floating meadows of Keibul Lamjao National Park in Manipur, India

[Display omitted] •Lightweight optical-thermal quadcopter can detect deer in inaccessible wetland terrains.•Suitable height for deer detection was 25 m and habitat mapping was 90 m.•Dawn survey gave the best results for deer count with optical and TIR camera.•Animals were disturbed by noise of more than 45 Hz (Hz) at a flying height below 20 m.•UAV survey can assist wildfire estimation and detection. Technological interventions such as remote sensing can play a crucial role in conserving inaccessible wetlands and their associated species. The floating meadows of Keibul Lamjao National Park (KLNP) in Manipur, India, are the only home to the last remaining wild population of the endangered and endemic Eld’s Deer, locally known as Sangai (Rucervus eldii eldii). Besides the rampant change in land use and altering water regimes induced by anthropogenic pressure, monitoring the population and habitat quality remains challenging. Moreover, conventional in-situ surveys that involve infiltrating animal habitats could adversely affect the habits and behavior of such shy species as a response to human presence. This study assesses the scope of Uncrewed Aerial Vehicles (UAVs) to non-invasively access such a challenging habitat and detect the ungulate individuals of interest. A quadcopter, mounted with an optical (RGB) and a thermal camera, was tested in this landscape conditions for its durability, failsafe range, flexibility, camera capabilities for visualization and detection, and resolution. The results of the drone survey based on three different time frames suggest that it is feasible to soar through the floating meadows to detect deer, best during dawn and dusk. Detection yielded the best results without creating a disturbance in animal behavior, with thermal infrared (TIR) imaging at night and both TIR and optical during the early morning. A flying height of nearly 90 m with RGB for vegetation and 20–25 m with TIR for animal presence yielded the best results. We have also established the least distracting sampling technique by analyzing the behavioral response of the deer to the presence of a drone. This study has paved a safe and practical way for acquiring accurate ground information that can assist in population estimation, habitat assessment, and burnt area estimation due to wildfire, to aid the conservation of an invaluable natural asset.