Simulating flash flood hydrographs and behavior metrics across China: Implications for flash flood management.

Flash floods in China are frequent and devastating, yet managing them remains difficult due to the spatial and temporal variability of these events. Prior research has shown that decision-making tools often lack detailed insights into the full range of flash flood behaviors. While existing models can simulate individual events, they rarely capture the broader patterns of flood behavior across diverse regions. This gap motivated the integration of hydrological modeling with clustering techniques to better assess flood reliability. No prior work had resolved how to classify and simulate flash flood types using similarity-based approaches. The challenge lies in handling the sheer number of events and their complex interactions with terrain and weather. Understanding these patterns could improve flood response strategies. This study aims to bridge that gap by analyzing flash flood processes at a broader scale.

The goal of this study was to evaluate the reliability of a hydrological model in simulating flash flood events across China. The researchers aimed to move beyond single-event analysis by classifying flood types based on shared characteristics. They wanted to determine if a modeling approach could accurately reproduce not only flood volumes but also behavior metrics like intensity and timing. The motivation stemmed from the need for better flood management tools that account for variability in event types. Traditional methods often fail to capture the full spectrum of flood behaviors. This study sought to extend the model's application to include behavior metrics and uncertainty measures. By doing so, the researchers aimed to provide more detailed information for decision-makers. The ultimate purpose was to improve flood response strategies through better simulation accuracy.

The study combined a hydrological model called CNFF with cluster analysis to classify flash flood events. A dataset of 207 hourly events from 13 mountainous catchments was used, each with distinct geographical and climatic features. Normalized hydrographs were generated to represent flood patterns. The k-means clustering algorithm grouped events into similar types based on their spatio-temporal characteristics. For each type, the model's ability to reproduce hydrographs and behavior metrics was assessed. Seven metrics were used to measure flood magnitude, intensity, timing, and variability. Uncertainty in model predictions was also quantified. The approach allowed for a similarity-based evaluation of simulation reliability.

Three flash flood types were identified: fast, intermediate, and slow, with an overall silhouette index of 0.45. The CNFF model accurately simulated hourly hydrographs for all types, with runoff errors under 15% and Nash-Sutcliffe Efficiency above 0.55. The slow type had the lowest average relative root-mean-square error of 0.30, followed by intermediate (0.52) and fast (0.58) types. Behavior metrics were best captured for the slow type, with 93.10% of observations within a 95% confidence interval. The slow type also showed the largest uncertainty interval for behavior metrics at 71.96%. Hydrograph uncertainty was 1.24% for the slow type. The model performed well for all types but with varying degrees of accuracy. These findings suggest the model can support flood management decisions.

The study demonstrated that CNFF can reliably simulate flash flood types and behavior metrics across China. The model's performance was highest for slow floods, indicating its usefulness in capturing gradual flood processes. The clustering approach allowed for a similarity-based evaluation of simulation reliability. The results suggest that the model can be used to inform flood management strategies by providing detailed insights into flood behavior. The slow flood type showed the most accurate behavior metrics but also the greatest uncertainty. This finding implies that while the model is useful, uncertainty must be considered in decision-making. The study extended the application of hydrological models to include behavior metrics and uncertainty quantification. The findings support the use of CNFF in flash flood management.