Lidar-based estimation of snow depth and SWE in north boreal and sub-arctic sites

Snow is an important part of the hydrological cycle in high-latitude and mountainous regions, influencing global climate, ecosystems, water resource management, and human societies. Accurate, high-resolution snow cover data are increasingly needed for model inputs, predictions, and societal risk management. Snow distribution is influenced by weather and topography, often exhibiting consistent patterns across locations, such as areas prone to faster melting or wind-blown accumulation. Thus, there is major local variation, making modelling and predictions challenging.

This study tests a novel measurement of snow water equivalent in the boreal landscape through the combination of UAV lidar technology, machine learning and ground measurements. We focus on three different study sites in Finnish Lapland, Pallas, Sodankylä and Oulanka, each representing different vegetational and topographical conditions typical of the boreal and sub-arctic landscapes. The field data were collected in four campaigns during the winter of 2023–24 from UAV-based lidar, manual snow course measurements, and snow depth sensor network. Based on measurements, we defined clusters for variable snow accumulation sections in study sites using a k-means machine learning algorithm, and daily snow height estimates were created for each cluster from reference snow depth measurements. The created clusters and their daily snow heights were then used as input for the Δsnow model (Winkler et al., 2021) to estimate catchment-scale daily snow water equivalent (SWE) and its distribution.

Three different clusters were defined in all sites by the lidar-based snow depth maps, typically corresponding to open areas, transition zones and forested areas. Each established cluster represents three different snow development patterns during the winter, from early winter to melt. The clustering approach allowed the upscaling of snow course measurements with reasonable accuracy, producing daily SWE and snow depth estimates that aligned with observed measurements.

The results show a promising contribution of UAV lidar mapping to catchment-scale snow monitoring, providing improved spatial and temporal accuracy for daily snow depth and SWE mapping in different areas. The work is important for estimating snow cover and melting for flood prediction, hydropower operation and water availability estimation.