As part of a study to estimate snow parameters based on passive microwave remote sensing and meteorological information, a physically-based, two-layer snow model was used to simulate snow depths for 100 World Meteorological Organization (WMO) sites in the Northern Hemisphere during the period 1992-1995.
At each site, daily observations of precipitation, temperature minima and maxima, humidity minima and maxima, and windspeed were used to estimate hourly time series of temperature, humidity, wind speed, incoming shortwave radiation, downwelling longwave radiation, and precipitation. Observed daily snow depths were compared with modeled snow depths, using a parameterization of density changes due to overburden pressure to convert model-predicted snow water equivalent to depth.
Initially no attempt was made to calibrate the snow model for individual sites, and one set of snow parameters, such as initial density, surface roughness, and albedo was used for all the sites. Although the model performed reasonably well based on this global parameters set, strong regional biases were observed in the residuals of modeled versus observed snow depth. To improve model performance, the WMO stations were grouped according to climate and snow type,
and regional parameter sets were developed for each of the new categories. This regionalization of model parameters led to substantial improvement in model performance