The utility of a new sea ice model for operational analysis of ice conditions in the marginal ice zone (MIZ) is examined for the Barents and Greenland Seas for select periods during the growth season of 2001-2002. The model has been developed as an operational tool to aid in the production of weekly ice charts at the National Ice Center (NIC) in Washington, DC. Rather than use satellite ice motion data to constrain model dynamics in the MIZ, where such data are less reliable the model uses direct assimilation of passive microwave ice concentration retrievals to drive the model thermodynamics. Results of the model are compared to a suite of synthetic aperture radar (SAR) and visible/infrared satellite data, satellite-derived ice motion vectors, and NIC ice charts. Reliable prediction of the ice thickness distribution depends on accurate detection of opening and closing of the ice cover. Results show that for low to moderate ice concentration, passive microwave data can provide an accurate measure of ice growth. At higher concentrations, ice formation and melt must be controlled by thermodynamic constraints due to errors in ice drift. Despite a simple formulation of model dynamics, it is shown that for the MIZ, predicted ice drift is as accurate, or better, than more sophisticated operational ice models. Comparisons between satellite imagery, model results, and ice charts suggest that in some circumstances the model ice type distribution may be more accurate than those produced through manual analysis of imagery. The results demonstrate that the primary limitation to accurate modeling of the MIZ on short time scales is the accuracy of forcing data. This limits the prediction of ice type distribution as ice drift errors will affect the accuracy of satellite-derived ice growth and melt. Improvements in the accuracy of ice drift predictions through the assimilation of satellite-derived ice motion vectors are discussed. Prospects for using the model to detect ocean heat flux near the ice edge are presented.