The dominant component of the polar surface energy budget during half the year is the downwelling flux of longwave radiation (DLF), yet little is known about its spatial and temporal variability except on monthly timescales. As surface measurements will always be sparse, the most promising opportunity for diagnosing the DLF is provided by satellite data. Estimating this flux from space, however, presents challenges over all surface types and particularly in polar environments where cloud detection and cloud fraction estimation are less certain. A new method is presented to estimate DLF from measurements by the TIROS-N operational vertical sounder (TOVS). Temperature profiles, humidity estimates, and cloud cover are retrieved from TOVS radiances using the improved initialization inversion algorithm, which has been modified to produce more accurate results over snow and sea ice. This information is combined with brightness temperature differences from pairs of infrared and near-infrared TOVS channels. These differences are used to infer cloud phase and geometric thickness. Longwave fluxes are then calculated using a forward radiative transfer model. Results during winter 1988 and spring 1992 are compared with hourly radiation measurements from the Coordinated Eastern Arctic Experiment in the eastern Arctic basin and from the Lead Experiment in the Beaufort Sea. Error analyses yield a bias of approximately 3 W m-2, a standard deviation of 23 W m-2, and a correlation coefficient of about 0.75. These errors are comparable to results from similar studies over midlatitude land and ocean areas where clouds are more easily identified.¿ 1997 American Geophysical Union