Assessment of modeled snow cover from General Circulation Models

The Atmospheric Model Intercomparison Project (AMIP-II) is an international effort to determine the systematic errors and differences between atmospheric climate models. This investigation assesses modeled snow cover from a set of AMIP-II models with the aim of identifying the strengths and weaknesses of their snow parameterizations. The motivation behind this model assessment stems from the need for accurate representation of modeled snow cover for hydrology and climatology. Snowmelt runoff provides a large portion of the fresh water in the Northern Hemisphere and is especially important for water resources management in the arid and semi-arid regions of the Western US. Snow cover depresses lower tropospheric temperatures at local and regional scales and impacts atmospheric circulation on regional and hemispherical scales.

Snow cover output from six AMIP-II models are compared with satellite-derived snow covered area for the period 1979-1995. We use a time series of passive microwave snow cover over the Northern Hemisphere as the basis for assessing modeled snow cover. The satellite-derived snow cover data represent weekly maxima and these data are aggregated to monthly values for comparison with the monthly output of the AMIP-II models. Both the remote sensing data and the model output are regridded to a common grid format (Equal Area Lambert Azimuthal). Most of the AMIP-II models are run at T42 resolution (2.8 x 2.8 degrees) whereas the passive microwave snow cover data at a much finer spatial resolution (25 x 25 km). The remote sensing data allow us to assess how well each model computes the fraction of snow cover in a grid cell, the onset and disappearance of snow cover during each snow season, and the total extent of snow cover. This is done for the whole Northern Hemisphere with particular focus on two subregions: the Northern Great Plains and the Western U.S. We then rank the models by the accuracy of their snow cover representations and compare assessment results against snow parameterizations for each model. Differences in these parameterizations include treatment of fractional snow cover in a grid cell, number of model snow layers, treatment of snow albedo, vegetation cover characterization, treatment of surface roughness, and thermal properties of the snowpack and soil. In this way we hope to identify the most successful/unsuccessful GCM parameterizations of snow cover.