Combining multiple data sources with multi-physics simulation frameworks offers new potential to extend snow model inter-comparison efforts to the Himalaya. As such, this study evaluates the sensitivity of simulated regional snow cover and runoff dynamics to different snowpack process representations. The evaluation is based on a spatially distributed version of the Factorial Snowpack Model (FSM) set up for the Astore catchment in the upper Indus basin. The FSM multi-physics model was driven by climate fields from the High Asia Refined Analysis (HAR) dynamical downscaling product. Ensemble performance was evaluated primarily using MODIS remote sensing of snow-covered area, albedo and land surface temperature. In line with previous snow model inter-comparisons, no single FSM configuration performs best in all of the years simulated. However, the results demonstrate that performance variation in this case is at least partly related to inaccuracies in the sequencing of inter-annual variation in HAR climate inputs, not just FSM model limitations. Ensemble spread is dominated by interactions between parameterisations of albedo, snowpack hydrology and atmospheric stability effects on turbulent heat fluxes. The resulting ensemble structure is similar in different years, which leads to systematic divergence in ablation and mass balance at high elevations. While ensemble spread and errors are notably lower when viewed as anomalies, FSM configurations show important differences in their absolute sensitivity to climate variation. Comparison with observations suggests that a subset of the ensemble should be retained for climate change projections, namely those members including prognostic albedo and liquid water retention, refreezing and drainage processes.