Stochastic Simulation of Realistic Continuous Snow Depth Time Series

We propose an approach for stochastic simulation of realistic continuous snow depth time series using a snow depth estimation model and a stochastic weather generation model. The snow depth estimation model consists of three steps: (1) determination of the precipitation type, (2) estimation of  the snow ratio, and (3) estimation of the decreased snow depth. In the first step, air temperature and relative humidity are used as indicators to determine the type of precipitation when precipitation occurs. In the second step, when the type is determined as snow, the snow ratio is estimated, converting the depth of precipitation into depth of fresh snow. Here, the air temperature is used as an indicator to estimate the snow ratio using sigomidal relationship with the snow ratio. In the last step, the amount of decreased snow depth was estimated using a novel temperature index snowmelt equation considering a trend of depth-dependent decreasing snow depth. The snow depth estimation model was applied to the four snowiest meteorological stations of Korea and yielded high Nash Sutcliffe efficiency values which ranged between 0.745 and 0.875 for calibration, and ranged between 0.432 and 0.753 for validation. This calibrated snow depth estimation model was then applied to the simulated weather time series (precipitation, temperature, and relative humidity) from the stochastic weather generation model to simulate continuous snow depth time series. The simulated snow depth data accurately reproduced standard and extreme value statistics of the observed data, the latter of which were consistent with the estimates provided in Korean Building Code. Then, the model was extended to investigate the influence of climate change on the future snow depth. For this, future weather statistics were obtained by applying factor of change to the current weather statistics and then were used to calibrate the weather generation model. Lastly, the future snow depth time series for three future time windows (2021-2040, 2041-2070, and 2071-2100) were simulated using future weather time series and snow depth estimation model.

 

This research was supported by a grant(2022-MOIS61-003) of Development Risk Prediction Technology of Storm and Flood for Climate Change based on Artificial Intelligence funded by Ministry of Interior and Safety(MOIS, Korea).