Stochastic Generation of Snow Depth in Canada

Abstract

Snow depth is a significant component in the hydrological cycle and global energy and water balances, contributing to climate change impacts. Weather stations with gauges for snow depth are scarce, especially in complex terrain regions, and require high accuracy for measurements. Advances in observational systems offer unconventional solutions yet are expensive. To bridge these gaps, stochastic generation methods offer a cost-effective solution to reproduce time series of hydrological variables, preserving their stochastic properties. Stochastic generation methods are well-established for total precipitation but lack snow depth generation. Here, we introduce a stochastic method to exclusively generate snow depth time series that preserve their distinct statistical properties on different time scales. We use 450 observed snow depth time series and 470 CMIP6 simulations to detect Canada's observed and physical statistical properties. The results indicate that snow depth has a light tail, and the distribution might change daily. The probability of zero snow depth shows a clear seasonal pattern. The synthetic snow depth time series can be an alternative to climate models’ outputs, offering a computationally effective solution to investigate the snow depth variability. This method advances the generation of stochastic time series of snow depth and can be applied to investigate catastrophes from snowmelt processes and avalanches that lead to severe damage and fatalities.