LOCAL CONTROLS ON SNOW DYNAMICS USING MULTI-SENSOR UAV IMAGERY AND MACHINE LEARNING

Seasonal snow serves as a critical input for water resources in many regions, making accurate monitoring and prediction of snow dynamics crucial for water management strategies. Understanding local scale variability in snow accumulation and melt patterns remains a critical challenge for hydrological modeling and satellite-based snow cover monitoring. Changes in snow depth can be mapped at an unparalleled scale by leveraging high-resolution, repeat imagery from unmanned aerial vehicles (UAV). This data can provide insight into fine-scale controls on snow dynamics at a local level. We present data from repeat UAV surveys throughout the 2025 snow season across a small alpine field site, including multispectral imagery, thermal data, and snow depth products derived from Structure-from-Motion. We assess potential controls on snow depth distribution from terrain variables, vegetation characteristics, and snow surface properties derived from thermal and multispectral imagery. We run a random forest model to predict snow accumulation and melt patterns and evaluate the importance of terrain, vegetation, and surface controls across the field site. UAV results are contrasted with satellite observations, snow model output, and automated weather station data to discuss method- and scale-dependent variability across the study site. Our findings provide a framework for understanding fine-scale snow dynamics and give context for broader-scale snow cover monitoring, with implications for modeled and remotely sensed approaches for water resource forecasting.