Thermal unmanned aerial vehicles for the identification of microclimatic refugia in topographically complex areas

Biodiversity loss is one of the most relevant consequences of climate change. Therefore, identifying areas and environmental features that allow certain organisms to be less exposed to the effects of the current global warming is priority for biodiversity conservation. In this study, we describe a novel approach for the identification of microclimatic refugia in rugged mountain areas, specifically for the detection of most thermally stable areas, using an unmanned aerial vehicle (UAV) capable of recording in the visible and thermal infrared spectral bands. We estimated land surface temperatures (LST) at very-high spatial resolution in six topographically complex sectors of the Pyrenees (NE Spain), across seasons with vegetative activity (summer 2020, autumn 2020, spring 2021, and summer 2021), and at two thermally contrasted times of the day (early in the morning: LSTmin, and in the afternoon: LSTmax). LST were validated with a network of miniaturized temperature sensors in the field. LSTmin and LSTmax allowed us to calculate the daily thermal range of each sector across the seasons, and thus the most thermally stable areas over the year. To reveal the importance of different variables on low and narrow thermal ranges we applied Gradient Boosted Models to seven terrain variables derived from ALS-LiDAR (slope, northness, eastness, heat load, wind exposure index, SAGA's topographic wetness index, and vector ruggedness measure) and a proxy of forest density through the three-dimensional point clouds of the UAV data. The northness was the variable that most promoted thermal stability, followed by the slope and forest density, so that microclimatic refugia resulted to be located in northern slopes, small sites under rocky cliffs, and forested areas. Our results demonstrate that thermal UAVs can become promising tools for the identification of microclimatic refugia in topographically complex areas, providing information at unprecedented spatial resolution, and thus of high interest for biodiversity conservation.