We investigated the effect of changing the horizontal resolution of a regional climate model (RCM) on the simulation of hydrological extremes. We employed the results of three experiments of the RCM HIRHAM using a grid size of approximately 12, 25 and 50 km. These simulations were used to drive the hydrological model LISFLOOD, developed for flood forecasting at European scale. The discharge simulations of LISFLOOD were compared with statistics of observed river runoff at 209 gauging stations across Europe. The largest discrepancies in peak flow occurred in climates with a seasonal snow cover, which may be explained by inaccuracies in the simulated precipitation that accumulate over winter. Although previous studies have found that high resolution climate simulations result in more realistic patterns of extreme precipitation, especially in mountainous regions, we did not find conclusive evidence that the 12-km HIRHAM run generally yields a better simulation of peak discharges. At some gauging stations the model performance is increasing with increasing horizontal resolution of the RCM, while at other stations it is decreasing. However, the differences between the three experiments become less important in larger river basins. Above about 30 000 km² and 120 000 km², respectively, the 25- and 50-km runs generally provided a good approximation of the simulations based on the 12-km climatology. Under the A2 scenario of climate change, the changes in extreme discharge levels were similar between the three experiments at continental scale. At the scale of individual river basins, however, there were occasionally important differences. If we assume the 12-km HIRHAM simulation to be more realistic, the use of lower-resolution climate simulations may lead to an underestimation of future flood hazard. This means that results obtained with lower-resolution RCM simulations should be interpreted with care, as the grid scale of the climate model adds to the uncertainty.