Cloud optical thickness and droplet effective radius retrievals from solar reflectance measurements are traditionally implemented using a combination of spectral channels that are absorbing and nonabsorbing for water particles. Reflectances in nonabsorbing channels (e.g., 0.67, 0.86, 1.2 μm spectral window bands) are largely dependent on cloud optical thickness, while longer-wavelength absorbing channels (1.6, 2.1, and 3.7 μm window bands) provide cloud particle size information. Cloud retrievals over ice and snow surfaces present serious difficulties. At the shorter wavelengths, ice is bright and highly variable, both characteristics acting to significantly increase cloud retrieval uncertainty. In contrast, reflectances at the longer wavelengths are relatively small and may be comparable to that of dark open water. A modification to the traditional cloud retrieval technique is presented. The new algorithm uses only a combination of absorbing spectral channels for which the snow/ice albedo is relatively small. Using this approach, retrievals have been made with the MODIS airborne simulator (MAS) imager flown aboard the high-altitude NASA ER-2 from May to June 1998 during the Arctic FIRE-ACE field deployment. Data from several coordinated ER-2 and in situ University of Washington Convair-580 aircraft observations of liquid water stratus clouds are examined. MAS retrievals of optical thickness, droplet effective radius, and liquid water path are shown to be in good agreement with in situ measurements. The initial success of the technique has implications for future operational satellite cloud retrieval algorithms in polar and wintertime regions. ¿ 2001 American Geophysical Union