Estimating uncertainties in the newly developed multi-source land snow data assimilation system

The snow simulations from the recently developed multivariate land snow data assimilation system (SNODAS) for the Northern Hemisphere are assessed with regard to uncertainties in atmospheric forcing, model structure, data assimilation technique, and satellite remote sensing product. The SNODAS consists of the Data Assimilation Research Testbed (DART) and the Community Land Model version 4 (CLM4). A series of experiments are conducted to estimate each of the above uncertainty sources. The experiments include several open-loop model cases and data assimilation cases that assimilate the snow cover fraction (SCF) data from the Moderate Resolution Imaging Spectroradiometer (MODIS) and the terrestrial water storage (TWS) from the Gravity Recovery and Climate Experiment (GRACE). The atmospheric forcing uncertainty in terms of precipitation and radiation is found to be the largest among the various uncertainty sources examined, especially over the Tibetan Plateau (TP) and most of the mid- and high-latitudes. Model structure and choice of data assimilation technique are also two big sources of uncertainty in SNODAS. The uncertainty of model structure is represented by two different parameterizations of SCF. The density-based SCF scheme (as used in CLM4) generally results in better snow simulations than does the stochastic SCF scheme (as in CLM4.5) within the data assimilation framework. The choice of TWS products retrieved from GRACE has relatively the least impact on the snow data assimilation.