Abstract
Prediction of runoff is of great significance for the sustainable utilization of water resources and flood control and disaster reduction in the basin. In this study, a method for predicting the runoff caused by extreme sub-rainfall events was constructed based on the identification of extreme rainfall events, Mann–Kendall Test, R/S analysis, and regression analysis. The method was applied to the Jingle sub-basin, and the results showed that the extreme precipitation in this basin will maintain a slight rising trend in the future, assuming that the climate and underlying conditions remain the same as they were in the current scenario. There is a more stable correlational relationship between rainfall characteristic factors and runoff in extreme rainfall events. The extreme precipitation of 1–5 consecutive days under the 100a return period designed by the hydrological frequency method is 38.74, 60.01, 66.00, 71.44, and 73.69 mm, respectively, and the possible runoff predicted by the four empirical formulas is 1295−2495, 2108−4935, 2408−5801, and 3051−7062 × 10⁴ m³, respectively. The rainfall designed by the hydrological frequency combination method is 203.64 mm, and the possible runoff predicted by the four empirical formulas is 2.8−5.3 × 10⁸ m³. This study can provide a new reference for predicting the possible incoming runoff under extreme sub-rainfall events.
Introduction
As climate change accelerates, the frequency and intensity of extreme weather events are shifting globally. In China’s Fenhe River Basin—a vital region in the middle reaches of the Yellow River—flood disasters are a recurring threat. While many hydrological models excel at predicting annual or monthly runoff, they often struggle with the "sub-rainfall event" scale—the intense, hours-to-days bursts that actually cause flash floods. The challenge lies in the non-linear and chaotic relationship between rainfall and runoff. Factors like soil saturation, rainfall intensity, and duration create a complex web that standard linear models cannot always capture. This research establishes reliable empirical formulas specifically tuned for extreme events, providing a practical toolkit for engineers and disaster relief planners.
Study Area and Datasets
The researchers focused on the Jingle sub-basin, a critical catchment area within the Fenhe River Basin. The topography of this region, characterized by significant elevation changes, plays a major role in how rainfall is converted into runoff.
The study utilized two primary long-term datasets to build and validate the predictive models:
- Precipitation Data (1960–2019): Daily records from five hydrological stations in the Fenhe River Basin, sourced from the National Meteorological Science Data Center.
- Rainfall-Runoff Process Data (1971–2018): High-resolution (1-hour time step) measured data from the Jingle Hydrological Station, covering 103 specific rainfall events.
Methodology
The core of the study involved a multi-step statistical approach to bridge the gap between simple rainfall measurements and complex runoff outcomes. The researchers identified "extreme" events using a 95th percentile threshold and analyzed historical trends using the Mann–Kendall (MK) test and R/S analysis (Hurst index).
To link these characteristics to runoff, four distinct regression models were developed:
- Formula 1 (Sub-Rainfall Fitting): A simple binomial regression of total rainfall versus runoff.
- Formula 2 (Factor Combination): Incorporates the product of total rainfall and maximum 1-hour intensity ($I_1$).
- Formula 3 (Time-Segment Fitting): Uses multiple stepwise regression to analyze how rainfall distributed across seven sub-periods affects the final yield.
- Formula 4 (Upper Envelop Method): A "worst-case" model that fits a curve to the highest recorded runoff points for any given rainfall amount.
Key Insights and Results
Trend and Sensitivity
Both MK and R/S analyses suggest that extreme rainfall in the basin is increasing slightly and is likely to continue doing so, with Hurst indices between 0.54 and 0.71 indicating "weak persistence" of the upward trend. Crucially, the study found that runoff is significantly more sensitive to rainfall factors during extreme events than during ordinary ones. Once a threshold is crossed, every additional millimeter of rain yields a disproportionately higher amount of runoff.
100-Year Event Predictions
For a 100-year return period (a 1% chance of occurring in any year), the 5-day consecutive rainfall was calculated at 73.69 mm. Under this scenario, the predicted runoff volume ranged from 3,051 to 7,062 × 10⁴ m³, depending on the empirical formula used. For safety-critical infrastructure like dams or levees, the authors argue that the Upper Envelop method (Formula 4) is the most valuable, as it predicts the maximum possible runoff based on historical outliers.
Extreme Combination Scenario
When simulating a "perfect storm"—combining the highest recorded intensities from different historical periods into one event totaling 203.64 mm—the predicted runoff jumped to a staggering 2.8–5.3 × 10⁸ m³.
Conclusion
This study provides a robust framework for predicting runoff from extreme, short-duration rainfall events. By demonstrating that intensity ($I_1$) matters as much as total volume, and by providing a "worst-case scenario" tool through the Upper Envelop method, the research offers vital data for flood management and infrastructure design in the Fenhe River Basin. Future work should aim to incorporate dynamic variables such as real-time soil moisture and changing land-use patterns to further refine these predictions in a changing climate.
Summary updated