The melt of seasonal snowpack in mountain regions provides downstream river basins with a critical supply of freshwater. Snowdrift-permitting models (1 m – 250 m resolution) have been proposed to simulate snowpack heterogeneity that stems from differences in energy inputs, over-winter redistribution, sublimation, melt, and variations in precipitation. However, these spatial scales can be computationally intractable for large extents. In this work, the multiscale Canadian Hydrological Model (CHM) was applied to simulate snowpacks at snowdrift-permitting scales (≈ 50 m) across the Canadian Cordillera and adjacent regions (1.37 million km^2) forced by downscaled atmospheric data. Model outputs were compared to a set of multiscale observations including snow-covered area (SCA) from Sentinel and Landsat imagery, snow depth from uncrewed aerial system lidar, and point surface observations of depth and density. The multiscale approach reduced computational elements by 98%. Including snow redistribution processes improved the summer SCA r^2 from 0.7 to 0.9. At larger scales, inclusion of snow redistribution processes delayed full snowpack ablation by an average of 33 days. These simulations show how multiscale modelling can improve snowpack predictions to support prediction of water supply.