Characterizing Snow Depth and Ice Surface Elevation on Subarctic Lakes Using a Multi-Sensor Approach

Abstract

Snow on lake ice is highly heterogeneous, which influences ice formation and growth as snow acts as an effective thermal insulator. Despite its significance, spatially resolved snow depth measurements remain challenging, as existing methods are typically limited to point observations. This study presents a multi-sensor approach integrating ground-penetrating radar (GPR) and remotely piloted aircraft system (RPAS) Structure-from-Motion (SfM) photogrammetry to map snow depth at 1-m resolution on two morphologically distinct lakes in the Northwest Territories, Canada: Landing Lake (LL) and Great Bear Lake (GBL). Snow depth was calculated by differencing GPR-derived ice surface elevations and SfM-derived snow surface elevations. Validation against in-situ measurements (n = 1013) yielded relative errors of ~3% (RMSE = 2 cm) for LL and ~12% (RMSE = 3 cm) for GBL. Ice surface variability (9–25 cm) introduced negligible differences when using a mean ice elevation. Application of the method revealed small-scale snow features at comparable spatial scales (19 m LL; 23 m GBL), with greater variability and lower mean snow depth on GBL, likely due to larger wind fetch. Shoreline terrain and pressure ridges produced localized deposition zones ~20 m wide. These results demonstrate the effectiveness of a combined GPR–RPAS approach for mapping lake snow depth.