A temperature-based snow module has been coupled with a grid-based distributed hydrological model, to improve simulations of river flows in upland areas of Britain subject to snowfall and snowmelt. The coupled model has been driven with data from an 11-member perturbed-parameter climate model ensemble, for two time-slices (1960–1990 and 2069–2099), to investigate the potential impacts of climate change. The analysis indicates large reductions in the ensemble mean of the number of lying snow days across the country. This in turn affects the seasonality of peak river flows in some parts of the country; for northerly regions, annual maxima tend to occur earlier in the water year in future. For more southerly regions the changes are less straightforward, and likely driven by changes in rainfall patterns rather than snow. The modelled percentage changes in peak flows illustrate high spatial variability in hydrological response to projected climate change, and large differences between ensemble members. When changes in projected future peak flows are compared to an estimate of current natural variability, more changes fall outside the range of natural variability in southern Britain than in the north.