Large-scale watershed modelling presents a unique challenge in terms of physiographic and climatological heterogeneity, and spatially varied hydrologic responses. In particular, the spatial variability in hydrologic processes may introduce a high degree of uncertainty in the modelling of a large watershed. This study assessed the uncertainties in annual/seasonal streamflow and annual peak flow simulations with respect to selection of climate data and model parameter sets for the variable infiltration capacity (VIC) model of the Athabasca River basin (ARB) in Alberta, Canada. Two high-resolution gridded climate data sets over the 1979 to 2010 period, and six different model parameter sets calibrated corresponding to different time periods and various hydrologic patterns, were employed to quantify the uncertainty in VIC simulations. Moreover, the possibility of an ensemble approach to predict hydrologic responses in the ARB has been investigated. The results indicated that streamflow simulations near the headwater and along the Athabasca River mainstream have high uncertainty corresponding to selection of climate data mainly because of greater difference of precipitation between the two climate data sets, whereas sub-basin stations at low elevations were more sensitive to the selection of parameter set for interflow-dominated runoff cycle. All stations showed higher uncertainty corresponding to the selection of parameter set for annual peak flows. In addition, this study confirmed that the ensemble means can provide more accurate and consistent hydrologic information for the low-elevation area where higher internal variability exists.