A rigorous methodology is described for quantifying come of the uncertainties of climate change impact studies, excluding those due to downscaling techniques, and applied on a sit of five catchments in Great Britain. Uncertainties in climate change are calculated from a set of 25,000 climate scenarios randomly generated by a Monte Carlo simulation, using several Global Climate Models, SRES-98 emission scenarios and climate sensitivities. Flow series representative of current and future conditions were simulated using a conceptual hydrological model. Generalised Pareto Distributions were fitted to Peak-Over-Threshold series for each scenario, and future flood scenarios were compared to current conditions for four typical flood events. Most scenarios show an increase in both the magnitude and the frequency of flood events, generally not greater than the 95% confidence limits. The largest uncertainty can be attributed to the type of GCM used, with the magnitude of changes varying by up to a factor 9 in Northern England and Scotland. It is therefore essential that climate change impact studies consider a range of climate scenarios derived from different GCMs, and that adaptation policies do not rely on results from only very few scenarios. (C) 2003 Elsevier Science B.V. All rights reserved.