One of the key aspects of mesoscale quantitative precipitation forecasts for hydrological purposes, particularly in regions of complex terrain, is adequate characterization of snowfall and snowpack heterogeneity within catchments. This study utilizes a mesoscale atmospheric model (the Regional Atmospheric Modeling System, RAMS) configured with full liquid- and ice-phase microphysics to evaluate the potential for seasonal mesoscale forecasting of snowfall and snowpack accumulation. As a first step, a severe case of snowfall,
primarily east of the Continental Divide, was chosen for model verification at small grid spacing. In this case, a powerful winter storm affected the Front Range of the Rocky Mountains during the period 23-26 October 1997, with some snowfall amounts exceeding 140 cm. Due to terrain relief and storm dynamics, snowfall amounts varied by over 100 cm over a 15 km horizontal distance in some locations. As such, this case presents a challenging test of mesoscale model capability to reproduce snowfall heterogeneity. RAMS was configured for operational use by using initialization and boundary conditions from existing national weather forecasts from the National Center for Environmental Prediction's 48 km grid spacing Eta Model.
Multiple nested grids were employed to focus on the region where snowfall heterogeneity and terrain relief were greatest, at the mountain/plains interface. The outermost grid (Dx = Dy = 15 km) encompasses the western U.S. to allow for the synoptic evolution of the storm system over a 60 hour simulation period. The second grid (Dx = Dy = 5 km) telescopes in on the Front Range of Northern Colorado. The innermost nest (Dx = Dy = 1.67 km) captures the local region
in Colorado in the Denver-Boulder corridor, where the storm snowfall maximum occurred (to the accuracy of observations). The results were excellent. Improved resolution on each successive grid caused sequential improvement in model comparison with observation. The average snow water equivalent error on the innermost grid was less than 10% and snowfall heterogeneity was properly captured. Half-hourly liquid equivalent measurements from the Marshall, CO test site during the storm compare remarkably well to the model output. These results suggest that application of catchment scale atmospheric modeling to
seasonal snowpack accumulation could be successful.