This effort to improve prediction of forest snowpack evaporation was aimed at understanding differences in winter snowpack among forested aspects. Theoretical arguments combined air temperature, humidity, wind speed, and radiation into a sublimation index. Monitoring weather factors and the mass of a 65-cm-diameter pan of snow set in a forest snowpack provided calibration of the index. Changes in snow surface structure caused a decrease with time in the empirical ratio of snowpack-to-index sublimation. This decrease was proportional to total sublimation since snowfall. We applied the model to compute daily snowpack sublimation on two adjacent forested slopes in central Colorado, U.S.A., during a 40-d accumulation period in March and April 1996. The estimate for both slopes combined was 0.52 mm/d, with sublimation from a south slope snowpack (0.61 mm/d) averaging 1.2 times that from an adjacent north-facing slope (0.43 mm/d). Extending the average difference of 0.18 mm/d through a winter season (150 d) would account for a 27-mm difference in snowpack water between the two aspects. Sublimation of 78 mm in a snow season, estimated from the combined average daily rate, represents about 20% of the normal peak water equivalent of these snowpacks. Effects of surface structure on snowpack sublimation should also apply to surfaces of snow intercepted by canopies.