Snow Distribution Patterns Revisited: A Physics-Based and Machine Learning Hybrid Approach to Snow Distribution Mapping in the Sub-Arctic

Snowpack distribution in Arctic and alpine landscapes often occurs in repeating, year-to-year patterns due to local topographic, weather, and vegetation characteristics. Previous studies have suggested that with years of observational data, these snow distribution patterns can be statistically integrated into a snow process modeling workflow. Recent advances in snow hydrology and machine learning (ML) have increased our ability to predict snowpack distribution using in-situ observations, remote sensing data sets, and simple landscape characteristics that can be easily obtained for most environments. Here, we propose a hybrid approach to couple a ML snow distribution pattern (MLSDP) map with a physics-based, snow process model. We trained a random forest ML algorithm on tens of thousands of snow survey observations from a subarctic study area on the Seward Peninsula, Alaska, collected during peak snow water equivalent (SWE). We validated hybrid model outputs using in-situ snow depth and SWE observations, as well as a light detection and ranging data set and a distributed temperature profiling sensor data set. When the hybrid results were compared with the physics-based method, the hybrid method more accurately depicted the spatial patterns of the snowpack, areas of drifting snow, and years when no in-situ observations were used in the random forest ML training data set. The hybrid method also showed improvements in root mean squared error at 61% of locations where time-series estimations of snow depth were observed. These results can be applied to any physics-based model to improve the snow distribution patterning to reflect observed conditions in high latitude and high elevation cold region environments. Snowpacks in cold regions often occur in repeated, year-to-year patterns, regardless of the amount of snow that has fallen during the winter. This is due to complex landscape and climate interactions, like the shape and angles of the topography, the direction of the prevailing winter winds, and vegetation cover. Previous studies have used snow pattern maps with models to create more accurate estimations of the snowpack. This study uses machine learning (ML) to create the snow pattern maps, along with previously determined methods for integrating those patterns into estimations of snow depth and snow water content. We spent multiple winter fieldwork campaigns taking measurements of the snowpack, including point-based snow depth and snow water content measurements, and plane-based measurements of snow depth over large areas. These measurements were used to compare with our simulated snowpack characteristics. The hybrid approach, one that used ML plus physics-based models, produced better spatial representations of the snowpack than results when ML was not used. Additionally, the hybrid approach generated improvements to modeled snow drift regions in the study area. These results show that ML and physics-based models can be used together to create better spatial representations of snowpack characteristics in cold regions. Machine learning, coupled with physics-based process models, can predict snow patterns with lower error than the process model alone When machine learning and physics-based process models were used together, drifted snow areas were improved in the model outputs Improvements in root mean squared error were found at 61% of snow depth observation sites throughout the year