Noah land surface model modifications for improved snowpack and water resource prediction in the Colorado Rockies

The Noah land surface model typically under-predicts the snowpack in mountainous regions. Sensitivity simulations are conducted to test the effects of several parameter and physics changes on the modeled snow water equivalent (SWE) in the Rocky Mountains. The control simulation is the standard Noah model forced with WRF output of temperature, humidity, wind, pressure, long wave radiation and precipitation along with short wave radiation forcing from GOES satellite estimates. Five subsequent simulations are completed using 1. time-varying albedo formulation; 2. adjusted downward solar radiation by the grid point slope and aspect; 3. surface exchange coefficient (Ch) adjustment during stable boundary layers; 4. setting maximum albedo to 0.85; and 5. adjusting surface roughness length to account for snow. These simulations are all completed over a 2km domain for the time period 1 Nov 2007 to 1 Aug 2008. Once per day, modeled SWE, snow depth, snow cover, snowmelt, snow evaporation and precipitation are archived.

Three metrics are chosen to assess the effects on simulated SWE: last day with greater than 5mm SWE, maximum SWE during simulation, and date when maximum SWE occurred. Results are further divided into groups of grid points with similar terrain, and similar slope and aspect.

Results of all model changes are shown for the entire grid and are also compared to SNOTEL observations. The last date with snow is much improved but still too early in the season for low elevation sites. Total effect of all model changes increases the day of maximum SWE by about 20 days for grids in the 2500-3500m range, but this is still about 20-30 days too early in the season compared to SNOTEL. Maximum SWE using all changes increases by 100-120 mm for 2500-3500m grids, but is still under-predicted by 200mm compared to SNOTEL.