Evaluation of snow albedo in land models for weather and climate studies

Snow albedo plays an important role in land models for weather, climate, and hydrometeorological studies, but its treatment in various land models still contains significant deficiencies. Complementary to previous studies that evaluated the snow albedo as part of an overall land model study, the snow albedo formulations as used in four major weather forecasting and climate models [European Centre for Medium-Range Weather Forecasts (ECMWF), National Centers for Environmental Prediction (NCEP) "Noah" land model, National Center for Atmospheric Research (NCAR) Community Land Model (CLM3), and NCEP global model] were directly evaluated here using multiyear Boreal Ecosystem-Atmosphere Study (BOREAS) in situ data over grass and forest sites. First, four idealized cases over grass and forest sites were designed to understand better the different albedo formulations in these models. Then the BOREAS data were used to evaluate snow albedo and relevant formulations and to identify deficiencies of each model. Based on these analyses, suggestions that involve only minor changes in parameters or formulations were made to significantly reduce these deficiencies of each model. For theECMWFland model, using the square root of snow water equivalent (SWE), rather than SWE itself, in the computation of snow fraction would significantly reduce the under-estimation of albedo over grass. For the NCEP Noah land model, reducing (increasing) the critical SWE for full snow cover over short (tall) vegetation would reduce the underestimate (overestimate) of snow albedo over the grass (forest) site. For the NCAR CLM3, revising the coefficient used in the ground snow-fraction computation would substantially reduce the albedo underestimation over grass. For the albedo formulations in the NCEP global model, replacing the globally constant fresh snow albedo by the vegetation-type-dependent Moderate-Resolution Imaging Spectroradiometer (MODIS) maximum snow albedo would significantly improve the overestimation of model albedo over forest.