Quantitatively Simulating Flood Processes Under Different Karst Landforms and Water‐Bearing Media

Abstract Landforms and water‐bearing media in karst regions greatly influence the patterns of hydrological processes (runoff generation and confluence) and water circulation, resulting in notable differences in the formation and propagation of flood processes between karst and nonkarst areas. Moreover, it is difficult to simulate floods accurately and quantitatively under the influence of complex landforms and karst water‐bearing media. In this study, an improved Karst‒Liuxihe (IKL) model was adopted to simulate karst flood processes in three typical karst basins in Southwest China, and the influences of karst landforms and water‐bearing media on the flood simulation results were evaluated quantitatively. The model improvements included optimizing the rainfall‒runoff algorithms and adding a mode of excess overland runoff infiltration to capture runoff generation on exposed karst landforms. The flood simulation results revealed that different karst landforms and water‐bearing media greatly influenced the flood simulation performance. For example, the relative error of the simulated peak flow decreased by 33%, and the relative error of the simulated flood volume decreased by 50% from the Qingmuguan Basin (conduit dominated) to the Mulian Basin (fissure dominated). The flood process simulation results obtained with the IKL model were much better than those obtained with the Karst‒Liuxihe (KL) model. The evaluation indices for the flood simulation performance included the Akaike information criterion (AIC), for which the average value decreased by 27%; the Nash–Sutcliffe coefficient, for which the average value increased by 6%; and the correlation coefficient, for which the average value increased by 8%. The values of the indices reflecting simulation errors decreased. For example, the relative error of the average process decreased by 28%, and the flood peak error decreased by 38%. The better flood simulation results indicate that the improvement in the KL model is effective, and this model could provide key technical support for hydrological forecasting and flood risk management in karst regions in Southwest China.