On the performances of empirical regressions for the estimation of bulk snow density

Abstract

Snow covers are a seasonal reservoir of water in the solid form. The snowpack accumulates during winter and melts during spring and summer. This process rules streamflow timing and amount in many catchments in temperate areas. The amount of water mass in a snow cover is usually measured as Snow Water Equivalent (SWE, in kg m^-2 or mm w.e.), i.e. the mass of liquid water which would result from the complete melting of that snow cover. To calculate SWE, evaluations of snow depth and bulk snow density are needed. A widely applied solution in conditions of data scarcity implies the measurement of snow depth, and the prediction of bulk snow density using multiple empirical regressions involving, as predictors, a set of proxy variables, such as air temperature, wind velocity, elevation, snow depth, or the age of the snow cover. Here, we reviewed 18 regressions used in the Literature to estimate mean bulk snow density. We compared the estimates of these regressions versus continuous-time measurements of daily bulk snow density collected in western US by the SNOTEL network using snow pillows. This analysis shows that the average percentage difference between predictions and data is around 25% − 45%. In addition, this difference increases with elevation. This shows that particular care is due when using these approaches, especially at high elevations, where snow plays a relevant role in the local hydrologic regime.