Simulating Snow Redistribution and its Effect on Ground Surface Temperature at a High-Arctic Site on Svalbard

In high-latitude and mountain regions, local processes such as redistribution by wind, snow metamorphism, and percolation of water produce a complex spatial distribution of snow depths and snow densities. With its strong control on the ground thermal regime, this snow distribution has pronounced effects on ground temperatures at small spatial scales which are typically not resolved by land surface models (LSMs). This limits our ability to simulate the local impacts of climate change on, for example, vegetation and permafrost. Here, we present a tiling approach combining the CryoGrid permafrost model with snow microphysics parametrizations from the CROCUS snow scheme to account for subgrid lateral exchange of snow and water in a process-based way. We demonstrate that a simple setup with three coupled tiles, each representing a different snow accumulation class with a specific topographic setting, can reproduce the observed spread of winter-time ground surface temperatures (GST) and end-of-season snow distribution for a high-Arctic site on Svalbard. For the 3-year study period, the three-tile simulations showed substantial improvement compared to traditional single-tile simulations which naturally cannot account for subgrid variability. Among others, the representation of the warmest and coldest 5% of the observed GST distribution was improved by 1-2 degrees C, while still capturing the average of the distribution. The simulations also reveal positive mean annual GSTs at the locations receiving the greatest snow cover. This could be an indication for the onset of localized permafrost degradation which would be obscured in single-tile simulations.

Automated summary

In high-latitude and mountain regions, complex spatial distribution of snow depths and densities significantly affects ground temperatures, which are typically not well captured by traditional simulation methods. Utilizing a tiling approach can lead to better simulation results for observed temperature and snow distribution, especially in polar regions. This approach demonstrates the potential for improving simulation accuracy in capturing small-scale variability.