The increased frequency of flood events has motivated interest in natural flood management (NFM), in particular the potential for woodlands to reduce flooding. Woodlands can reduce the risk of rainfall-generated flooding through increased interception, soil infiltration, and available storage. Despite growing evidence, there is still low confidence in woodlands as a flood mitigation method due to limited empirical data available, particularly for semi-natural woodlands. We established a correlation catchment study in Haweswater, Cumbria, UK. Nine small upland catchments, each less than 0.2 km2 in area, were established on semi-natural broadleaf woodland sites where no stock grazing occurs or pasture with varied grazing intensity. At each site soil characteristics were investigated, namely soil moisture, permeability and bulk density. In addition, a v-notch weir was installed within in each catchment to calculate flow. The specific peak discharge (SPD), peak runoff coefficient, volume runoff coefficient and time taken to flow response was determined at each site for 28 storm events, of up to 205 mm, identified over a 13-month period. We found that semi-natural woodland reduced SPD by 33-52 % compared with pasture, reducing SPD by 36 % during larger storms (> 1 mm/hr peak discharge). Woodland reduced the peak runoff coefficient by 31-52 % and the volume runoff coefficient by 13-22 % compared to pasture. Additionally, response to storm events took 1-4 hours longer in woodland. These differences in flood response can be somewhat explained by the more permeable woodland soils, 4.6 times greater than pasture soil. Our analysis strengthens the argument that woodlands can reduce rainfall-generated flooding as a land use management method of NFM. Data collected here should be used to inform the parameters in flood prediction models and contribute to the evidence base for NFM.