•An Improved Karst-Liuxihe (IKL) model for karst basins was developed successfully.•19 karst flood events under 4 extreme land use scenarios were analyzed in detail.•The annual runoffs of 2030, 2040, and 2050 were forecasted effectively.•The critical proportion of urbanization area throughout the watershed area was 45%. Urbanization, especially land use pattern changes, has a great impact on natural flood events in a karst basin. Quantitative simulations of the effects of urbanization on karst flood events and forecasts of future evolution trends are difficult based on the current hydrological models. These models often require a large amount of data when used in karst areas due to the complex model structure and model parameters. To overcome the limitations of model applications, a physically based and fully distributed karst hydrological model, i.e., the improved Karst-Liuxihe (IKL) model, was proposed in this study to simulate and forecast karst flood events under the influence of urbanization. This IKL model was developed through the overall improvement of the Karst-Liuxihe (KL) model. The main additions were the improved runoff generation algorithm and the underground river confluence module. The karst flood simulation results of the IKL model were much better than those of the KL model: the average values of the Nash–Sutcliffe coefficient, correlation coefficient, and coefficient of the water balance increased by 23%, 23% and 26%, respectively, while the process relative error, flood peak flow relative error, and flood peak flow time error decreased by 21%, 22%, and 3 h, respectively, which confirmed that the improvements to the model were effective and feasible. Therefore, this paper used the IKL model to simulate karst flood events and annual runoff under the influence of urbanization based on 4 extreme land use scenarios. In addition, the model effectively forecasted the future runoff in 2030, 2040, and 2050. The results indicated that the critical proportion of urbanized area throughout the watershed area was 45%. When the urban proportion exceeded 45%, waterlogging could occur in the study area.