Abstract:

[Objectives] This study aims to improve the accuracy and efficiency of flash flood forecasting in the Guanshan River Basin and other similar small mountainous watersheds frequently affected by flood disasters by analyzing the runoff generation mechanisms of flash floods. By comparing the performance of saturation-excess, infiltration-excess, and hybrid runoff generation modes in simulating flash floods of different magnitudes, we also seek to overcome the limitations of single-mode simulation under complex terrain and different rainfall intensities. [Methods] The runoff generation module of the Xin’anjiang model was modified to simulate 38 flood events in the Guanshan River Basin (24 for calibration, 14 for validation) using saturation-excess, infiltration-excess, and hybrid runoff generation modes. Flood magnitudes were classified into small, medium, large, and extra-large according to the Specifications for Hydrological Information and Forecasting. Simulation results were evaluated using Nash-Sutcliffe efficiency coefficient (NSE), peak discharge error, and runoff depth error to compare the applicability and advantages of different runoff generation mechanisms. [Results] The vertical hybrid runoff generation mode demonstrated higher accuracy and stability across different flood magnitudes. It outperformed the other two modes in terms of NSE during both calibration and validation periods, with particularly strong performance in simulating extra-large floods. The saturation-excess mode performed better for small floods but was less stable for large and extra-large events. The infiltration-excess mode achieved the highest accuracy in simulating peak discharges of large floods, but performed relatively poorly in small and extra-large events. Further analysis of the runoff generation mechanisms indicated that runoff generation processes were closely related to rainfall characteristics, soil infiltration rates, and underlying surface conditions. Under intense and short-duration rainfall, infiltration-excess was the dominant mechanism, while under low-intensity and long-duration rainfall, saturation-excess prevailed. The vertical hybrid mode comprehensively integrates both mechanisms, dynamically adjusting the runoff generation approach based on varying rainfall conditions. It enabled effective simulation of flash flood processes under different rainfall scenarios. Additionally, this mode showed higher precision in simulating the recession processes, as it better reflected river basin storage states and the dynamics of interflow and groundwater runoff. [Conclusions] The vertical hybrid runoff generation mode demonstrates significant advantages in simulating flash floods in the Guanshan River Basin, providing robust support for improving the accuracy and efficiency of flash flood forecasting in this area. These findings not only provide a theoretical basis for flood prevention and disaster mitigation in the Guanshan River Basin but also offer innovative approaches for flash flood forecasting in complex mountainous watersheds. The innovation of this study lies in its comprehensive consideration of multiple runoff generation mechanisms and its validation of the hybrid mode’s adaptability under different rainfall conditions through comparative analyses. Future research will further refine the runoff generation module by incorporating more detailed physical processes and parameterization methods, while exploring the coupled applications of hydrological and hydrodynamic models to enhance the model’s capability in simulating complex hydrological processes and provide deeper insights into flood evolution in small mountainous watersheds.

Key words: flash flood disaster, flood forecasting, runoff generation mechanism, Guanshan River Basin