Parameter sensitivity in simulations of snowmelt

A simple snow model with only three parameters (fresh snow albedo, albedo decay rate for melting snow and surface roughness) is used to simulate snow accumulation and melt at four sites in Europe and North America, and the extent to which the model's parameters can be calibrated against observations is investigated. Results from the model are compared with observations of snow water equivalent (SWE) and the range of results from models that participated in an intercomparison project for the same sites. Good simulations of SWE are obtained by parameter calibration, but sensitivity analyses show that the SWE observations do not contain enough information to uniquely determine parameter values even for this very simple model. Comparisons of simulated snow albedo with observations for two of the sites give stronger constraints on the model parameters, but the model is unable to give good simulations of SWE and albedo simultaneously with a single parameter set, revealing a weakness due to the model's neglect of internal snowpack processes; an enhanced version of the model representing heat storage in the snow performs better in simultaneous simulations of SWE and albedo. In comparison with observations of snow surface temperature, it is found that sensible heat fluxes in low windspeed conditions have to be enhanced to prevent the model from simulating unrealistically low night-time temperatures at a sheltered site.