Use of remote sensing to test and update simulated snow cover in hydrological models

Abstract

Models of daily runoff from seasonal snowpacks and glaciers require knowledge or assumptions about the decline in snow covered area (SCA). Some semi-distributed models rely on satellite data as an input in addition to meteorological data but general purpose hydrological models with a snow component do not normally use earth observation (EO) data. EO data have the potential to verify or update SCA predictions generated by these models, but comparison is hampered by the unrealistic assumption in most models of spatially uniform snow water-equivalent (SWE) within entire zones, SO that SCA decline is stepped. Two possible solutions are either to allow a stepped SWE distribution within a sub-area, or to assume uniform melt over a non-uniform snowpack within a sub-area. In both approaches melt is converted into a reduction in SCA as well as SWE allowing snowpack depletion to be compared directly with EO data. Two examples are given in which EO data is used to verify (and in one case update) SCA. The HBV model is applied to a basin in Arctic Sweden and a recently developed glacier runoff model is applied to a basin in the Swiss Alps. Landsat TM data of both basins revealed considerably less snow than simulated by the models. TM data for the Swedish basin show that only glacier zones were 100% snow covered. Despite over-predicting SCA both models achieved very good discharge fits. It is argued that runoff models should correctly simulate the hydrological system state variables if they are to be transferred to different environments or new climate scenarios with confidence, and that EO data can play a valuable role in this.