In order to reduce the impact of flooding and improve the management of flood disasters, it is crucial to assess the impact of extreme land use changes and rainfall types on flood characteristics. The distributed hydrological soil vegetation model was used to simulate 20 floods in the upper watershed of Tangpu Reservoir, Zhejiang Province. And then the fuzzy decision tree was adopted to categorize flood types into flash flood, short duration rainfall flood and long duration rainfall flood, so as to investigate the effects of extreme land use and rainfall types on the characteristic indices of flood. The results showed that the model simulation was effective, with Nash efficiency coefficients in both calibration period and validation period greater than 0.7. Due to the limited area of cropland in the region, the reduction of flood peaks and flood volumes by both returning farmland to forest and restoring farmland land to grassland was less than 2%. The flood peak of the flash flood was the most sensitive to land use change, its growth rate reached 12.87% and 60.88% under the extreme scenarios of deforestation and urban expansion respectively. Changes in subsurface had the greatest impact on the flood volume of long duration rainfall floods, with flood volume growth rates of 10.48% and 20.70% under the deforestation and urban expansion scenarios, respectively. The urban expansion scenario advanced the flood peak onset time by 2.10 h on average; changes in rainfall type advanced the onset of flash-short duration rainfall floods by an average of 3.63 h, and reduced the flood peak by 34.73%; while the onset of short duration rainfall-flash floods were delayed by an average of 7.11 h, and increased the flood peak by 22.87%.