Sea ice thickness is an important parameter in the surface energy exchange and mass balance of the polar oceans. Remote sensing offers the opportunity to monitor thin ice routinely. We combine measurements of surface temperature from an advanced very high resolution radiometer (AVHRR) with a thermodynamic ice growth model to estimate thin ice and its thickness distribution during the cold season. In a Beaufort Sea area of perennial ice we find that thin ice, up to a meter thick, had a concentration of 18% in December 1990 and 3% in April 1991; thin ice concentrations near the ice edge in the Greenland Sea range from 20% to 50%. These AVHRR-derived estimates generally compare well with thickness observations from upward looking sonar. Given the uncertainties in all the model inputs estimated from AVHRR, such as radiative and turbulent heat flux components and surface albedo, we find that ice thickness is estimated with an uncertainty of roughly 50%. The uncertainty in the cumulative thickness distribution is about 3% for ice less than 20 cm thick and about 9% for ice around 1 m thick; these magnitudes are significantly smaller than the variations in thin ice observed by satellite in different months and regions. We conclude that AVHRR is capable of resolving the regional and seasonal variations in thin ice to a level of accuracy useful both for establishing a climatology and for testing ice models. ¿ American Geophysical Union 1996