Downscaling Regional Wind Forecasts for Use in High Resolution, Operational Snow Models

Abstract:

High resolution model forcings are required to adequately simulate snow accumulation, melt, and streamflow in mountain environments. Wind, especially the high winds that induce snow redistribution and drive turbulent heat fluxes during rain-on-snow events, have been shown to play a vital role in these processes. Yet wind observations are sparse and rarely capture the large variability present in alpine regions. High resolution (1-10km) climate data is becoming more readily available but even these data are too coarse to properly represent alpine snow processes. Much attention has been focused on downscaling precipitation and air temperature for fine resolution modeling. However there is very little in the literature that has addressed techniques for deterministically downscaling wind speeds. This work addresses means of downscaling large-scale wind products for high-resolution operational modeling purposes. Though both dynamical and statistical means are available for downscaling purposes, the time constraints imposed by operational modeling restricts this work to the latter. The statistical downscaling is done by means of terrain parameters that determine topographic position related to wind exposure and shelter. First, raw hourly wind data from ~2km and ~7km resolution weather forecasts were compared to observations at well over 100 sites located throughout the Swiss Alps. As might be expected, there was a large range of scatter between model-predicted and observed winds, and predictions at high wind sites were biased low. Terrain parameters derived from a 25m resolution DEM aptly identified high and low wind speed sites and climate model biases related to the higher resolution terrain structure. The statistical downscaling differentiated windward and leeward slopes not resolved in the climate models, reduced modeling errors, and substantially reduced biases at the all-important high wind sites.