The Canadian Land Surface Scheme (CLASS) was modified to correct an underestimation of the winter albedo in evergreen needleleaf forests. Default values for the visible and near-infrared albedo of a canopy with intercepted snow, αVIS,cs and αNIR,cs, respectively, were too small, and the fraction of the canopy covered with snow, fsnow, increased too slowly with interception, producing a damped albedo response. A new model for fsnow is based on zI*, the effective depth of newly intercepted snow required to increase the canopy albedo to its maximum, which corresponds in the model with fsnow = 1. Snow unloading rates were extracted from visual assessments of photographs and modelled based on relationships with meteorological variables, replacing the time-based method employed in CLASS. These parameterizations were tested in CLASS version 3.6 at boreal black spruce and jack pine forests in Saskatchewan, Canada, a subalpine Norway spruce and silver fir forest at Alptal, Switzerland, and a boreal maritime forest at Hitsujigaoka, Japan. Model configurations were assessed based on the index of agreement, d, relating simulated and observed daily albedo. The new model employs αVIS,cs = 0.27, αNIR,cs = 0.38 and zI* = 3 cm. The best single-variable snow unloading algorithm, determined by the average cross-site d, was based on wind speed. Two model configurations employing ensemble averages of the unloading rate as a function of total incoming radiation and wind speed, and air temperature and wind speed, respectively, produced larger minimum cross-site d values but a smaller average. The default configuration of CLASS 3.6 produced a cross-site average d from October to April of 0.58. The best model employing a single parameter (wind speed at the canopy top) for modelling the unloading rate produced an average d of 0.86, while the two-parameter ensemble-average unloading models produced a minimum d of 0.81 and an average d of 0.84. © 2015 Her Majesty the Queen in Right of Canada. Hydrological Processes published by John Wiley & Sons, Ltd.