Hydrological extremes within the UK have increased in intensity, frequency and persistence over recent years and are predicted to increase in variability throughout the 21st century. Past and future changes in hydrological extremes relative to climate change were investigated within Severn Uplands, a climate sensitive catchment. Using the Mann- Kendall trend detection test, time-series analysis over a 30-year period revealed a significant increase in winter and autumn precipitation and a decrease in summer precipitation. The analysis of flow time-series indicated an increase in winter and July flows and a decrease in spring flows. Changes in climate variability over the same period showed increases in air temperature and SST, and a reduction in snow cover. Climate variables were found to largely correlate with hydrological extremes which were characteristic of certain weather types and largely influenced by the NAO. To model future flows within the Severn Uplands a hydrological model (HEC-HMS) was used to simulate hydrological processes. The extreme hydrological event of November- December 2006 was used to calibrate the model. The difference between using radar and gauge precipitation data to drive the model was quantified. Radar data resulted in the smallest prediction accuracy followed by gauge-corrected radar data (corrected using the mean-field bias where gauge rainfall was interpolated using cokriging) and then gauge precipitation which had the largest prediction accuracy. Model accuracy was sufficient using the gauge corrected radar and gauge precipitation data as inputs, so both were altered for future predictions to investigate the propagation of uncertainty. Predicted changes in temperature and precipitation by the UKCIP02 scenarios were used to alter the baseline extreme event to predict changes in peak flow and outflow volume. Both radar- and gaugedriven hydrological modelling predicted large flow increases for the 21st century with increases up to 8% by the 2020s, 18% by the 2050s and 30% by the 2080s. Discrepancies between predictions were observed when using the different data inputs.