面向延长洪水预见期的动态分位数雨型洪水预报方法——以尖山、河子口流域为例

张达文; 高玉丹; 江平

doi:null

欢迎来到能源电力期刊网！

English Version 期刊登录

您当前的位置：

首页 >

文章列表页 >

面向延长洪水预见期的动态分位数雨型洪水预报方法——以尖山、河子口流域为例

由编辑部分配 | 更新时间：2026-10-08

- 面向延长洪水预见期的动态分位数雨型洪水预报方法——以尖山、河子口流域为例
- A Dynamic Quantile Hyetograph-Based Flood Forecasting Method for Extending Flood Lead Time: A Case Study of the Jianshan and Hezikou Basins
- 人民珠江 2026年页码：1-16
- 作者机构：
  
  广东省水文局梅州水文分局，广东梅州 514000
- 作者简介：
  
  张达文（1985—），男，高级工程师，主要从事水文监测预报、洪涝预警及水文建模研究等工作。
  江平（1982—），男，硕士研究生，正高级工程师，主要从事水文监测预报、中小河流洪水预警及水文模型研发等工作。
- 基金信息：
- DOI：
  中图分类号：
- 收稿：2026-07-18，
  
  修回：2026-09-19，
  
  录用：2026-09-20，
  
  网络首发：2026-10-08，
- 稿件说明：
移动端阅览
张达文，高玉丹，江平.面向延长洪水预见期的动态分位数雨型洪水预报方法［J］.人民珠江.

ZHANG Dawen,GAO Yudan,JIANG Ping.A Dynamic Quantile Hyetograph-Based Flood Forecasting Method for Extending Flood Lead Time: A Case Study of the Jianshan and Hezikou Basins[J].PEARL RIVER.
张达文，高玉丹，江平.面向延长洪水预见期的动态分位数雨型洪水预报方法［J］.人民珠江. DOI：

ZHANG Dawen,GAO Yudan,JIANG Ping.A Dynamic Quantile Hyetograph-Based Flood Forecasting Method for Extending Flood Lead Time: A Case Study of the Jianshan and Hezikou Basins[J].PEARL RIVER. DOI：

摘要

针对中小流域洪水预报存在气象雨量总量与水文逐时输入时空尺度错配、传统固定雨型易低估洪峰的问题，构建基于动态分位数雨型的洪水预报方法。以雨型不对称特征截然相反的尖山（不对称系数

=3.32，峰后拖尾）、河子口（

=0.43，峰前集中）流域36场洪水为试验样本，依托1～24 h历时最大累积降雨分位数与峰前－峰后贪心分配策略搭建连续动态雨型库，结合安全系数与蒙特卡洛模拟量化预报低估风险。结

果表明：动态雨型可适配两类差异化流域汇流特征，尖山流域静态雨型合格率均为66.7%，动态P50提升至83.3%；河子口流域不匹配的静态雨型（固定P80、Huff）合格率仅57.1%，动态雨型达85.7%。气象雨量总量偏差是预报不确定性主导来源（贡献率约65%），研究流域洪水平均低估概率为8.5%。方法实现降雨时程集中度与总量安全裕度的解耦调控，可为中小流域山洪预警、水库调度及滚动洪水预报提供技术支撑。

Abstract

In the context of increasingly frequent extreme rainstorms

extending flood forecast lead time while controlling the risk of missed flood-peak warnings has become a core requirement for flash-flood evacuation and reservoir pre-release in small and medium-sized basins. Although radar-based nowcasting and numerical weather prediction can provide basin-total rainfall in advance

a spatiotemporal scale mismatch persists between the areal rainfall totals supplied by meteorological models and the hourly series required by hydrological models. Conventional fixed hyetographs

which rely on a single discrete typical curve

tend to underestimate flood peaks and cannot represent the continuous risk spectrum from normal to extreme rainstorms. This study therefore proposes a dynamic quantile hyetograph method for flood forecasting. Two basins with sharply contrasting rainfall temporal structures were selected: the Jianshan Basin (1

578 km²; asymmetry coefficient A = 3.32; post-peak trailing) and the Hezikou Basin (1

031 km²; A = 0.43; pre-peak concentrated). They provided 15 and 21 flood events

respectively

yielding 36 events in total at a 1-h time step. Areal mean rainfall was derived by Thiessen polygon weighting; a 24-h rainfall sequence centred on the 6-h storm core was extracted

and each event was normalized into a dimensionless hyetograph. Empirical quantiles from 0.3 to 0.9 of the maximum cumulative rainfall over continuous 1–24 h durations were estimated to construct a duration-nested hyetograph library

and the incremental rainfall of each time step was allocated using a greedy pre-peak/post-peak strategy constrained by the basin-specific asymmetry structure. Runoff was simulated with a coupled saturation-excess and infiltration-excess module and routed by a time-invariant unit hydrograph

with all parameters held fixed so that differences could be attributed solely to the temporal rainfall distribution. A one-sided safety factor Kq

derived from the 90th percentile of the underestimation ratio

together with a 200-run Monte Carlo simulation

was used to quantify the underestimation risk induced by total-rainfall bias and quantile-selection error. The results show that the dynamic hyetographs adapt well to the two contrasting runoff concentration regimes. In the Jianshan Basin

all four static hyetographs gave a pass rate of 66.7%

whereas the dynamic P50 hyetograph raised it to 83.3% and reduced error dispersion by 42%. In the Hezikou Basin

the mismatched static hyetographs (fixed P80 and Huff) gave a pass rate of only 57.1%

with a mean underestimation of 22.3% and a maximum of 37.8%

whereas the dynamic P50 hyetograph reached 85.7% and the optimal q reduced the maximum underestimation to 7.1%. Variance decomposition indicates that total-rainfall error is the dominant uncertainty source (about 65%)

while q-selection error accounts for about 35%; the mean probability of peak underestimation exceeding 20% is about 8.5%. The quantile q therefore acts as a risk-control variable that shifts the error distribution rather than as a simple accuracy on/off switch

revealing a clear trade-off between overestimation and missed warnings. By decoupling the regulation of rainfall temporal concentration from the total-rainfall safety margin

the proposed method provides a standardized interface from meteorological rainfall totals to hydrological model inputs and supports a three-tier rolling forecast product system comprising baseline

attention

and safety levels. It offers a low-cost and operationally feasible technical basis for flash-flood warning

reservoir pre-release

and rolling flood forecasting in small and medium-sized basins.

关键词

Keywords

references

RAVURI S ， LENC K ， WILLSON M ， et al . Skilful precipitation nowcasting using deep generative models of radar ［J］. Nature ， 2021 ， 597 ： 672 - 677 .

李涛，夏润亮，李冰，等 . 洪水预报研究与展望［J］. 中国农村水利水电， 2022 （ 12 ）： 107 - 114 .

Li Tao ， Xia Runliang ， Li Bing ， et al . Research and prospects of flood forecasting ［J］. China Rural Water and Hydropower ， 2022 （ 12 ）： 107 - 114 . （in Chinese）

谢小燕，高熠飞，徐凯莉，等 . 基于气象水文耦合的水库实时洪水预报方法［J］. 中国农村水利水电， 2024 （ 8 ）： 81 - 87 .

Xie Xiaoyan ， Gao Yifei ， Xu Kaili ， et al . Real-time reservoir flood forecasting method based on meteorological-hydrological coupling ［J］. China Rural Water and Hydropower ， 2024 （ 8 ）： 81 - 87 . （in Chinese）

司琪，金保明，卢旺铭，等 . 崇阳溪流域PRBP神经网络洪水预报模型研究［J］. 人民珠江， 2025 ， 46 （ 6 ）： 68 - 74 .

Si Qi ， Jin Baoming ， Lu Wangming ， et al . Research on PRBP neural network flood forecasting model in Chongyang River basin ［J］. Pearl River ， 2025 ， 46 （ 6 ）： 68 - 74 . （in Chinese）

马瑞雪，刘建卫，逄晓腾 . 挠力河流域极端土地利用变化与降雨类型对洪水过程的影响［J］. 人民珠江， 2025 ， 46 （ 5 ）： 30 - 40 .

Ma Ruixue ， Liu Jianwei ， Pang Xiaoteng . Effects of extreme land use changes and rainfall types on flood process in Naoli River basin ［J］. Pearl River ， 2025 ， 46 （ 5 ）： 30 - 40 . （in Chinese）

HAPUARACHCHI H A P ， BARI M A ， KABIR A ， et al . Development of a national 7-day ensemble streamflow forecasting service for Australia ［J］. Hydrology and Earth System Sciences ， 2022 ， 26 ： 4801 - 4821 .

HUFF F A . Time distribution of rainfall in heavy storms ［J］. Water Resources Research ， 1967 ， 3 （ 4 ）： 1007 - 1019 .

CHIN D A . Application of Huff rainfall distributions in stormwater management ［J］. Journal of Sustainable Water in the Built Environment ， 2023 ， 9 （ 1 ）： 04022024 .

PILGRIM D H ， CORDERY I . Rainfall temporal patterns for design floods ［J］. Journal of the Hydraulics Division ， 1975 ， 101 （ 1 ）： 81 - 95 .

YUAN W L ， TU X Y ， SU C G ， et al . Research on the critical rainfall of flash floods in small watersheds based on the design of characteristic rainfall patterns ［J］. Water Resources Management ， 2021 ， 35 （ 10 ）： 3297 - 3319 .

OH B ， KIM J ， HWANG S . Influence of rainfall patterns on rainfall-runoff processes： indices for the quantification of temporal distribution of rainfall ［J］. Water ， 2024 ， 16 （ 20 ）： 2904 .

PELEG N ， WRIGHT D B ， FOWLER H J ， et al . A simple and robust approach for adapting design storms to assess climate-induced changes in flash flood hazard ［J］. Advances in Water Resources ， 2024 ， 193 ： 104823 .

YUAN W L ， JING B H ， XU H S ， et al . A dynamic early warning model for flash floods based on rainfall pattern identification ［J］. International Journal of Disaster Risk Science ， 2024 ， 15 （ 5 ）： 769 - 788 .

YANG P ， XU Z X ， YAN X F ， et al . Establishing a rainfall dual-threshold for flash flood early warning considering rainfall patterns in mountainous catchment， China ［J］. Natural Hazards ， 2024 ， 120 （ 7 ）： 6657 - 6684 .

瞿思敏，陆威村，蒋思军，等 . 洪水预报的比例系数系统响应误差修正方法［J］. 水资源保护， 2025 ， 41 （ 2 ）： 88 - 94， 105 .

Qu Simin ， Lu Weicun ， Jiang Sijun ， et al . Proportional-coefficient system-response error correction method for flood forecasting ［J］. Water Resources Protection ， 2025 ， 41 （ 2 ）： 88 - 94， 105 . （in Chinese）

BENNETT J C ， ROBERTSON D E ， WANG Q J ， et al . Propagating reliable estimates of hydrological forecast uncertainty to many lead times ［J］. Journal of Hydrology ， 2021 ， 603 ： 126798 .

张春堂，张珂，陈国鼎，等 . 基于智能网格降水和水文滑坡耦合模型的全国洪水和滑坡灾害滚动预报［J］. 水资源保护， 2025 ， 41 （ 3 ）： 83 - 92 .

Zhang Chuntang ， Zhang Ke ， Chen Guoding ， et al . Rolling forecast of nationwide flood and landslide disasters based on intelligent grid precipitation and a coupled hydrological-landslide model ［J］. Water Resources Protection ， 2025 ， 41 （ 3 ）： 83 - 92 . （in Chinese）

GB/T 22482-2008 水文情报预报规范［S］. doi: 10.1002/hyp.7014 http://dx.doi.org/10.1002/hyp.7014

GB/T 22482-2008 Standard for hydrological information and hydrological forecasting ［S］. （in Chinese） .

刘源，纪昌明，马皓宇，等 . 基于集合Kalman滤波的中长期径流预报［J］. 水资源保护， 2024 ， 40 （ 1 ）： 93 - 99 .

Liu Yuan ， Ji Changming ， Ma Haoyu ， et al . Medium- and long-term runoff forecast based on ensemble Kalman filter ［J］. Water Resources Protection ， 2024 ， 40 （ 1 ）： 93 - 99 . （in Chinese）

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

基于CMIP6气候情景的西江流域未来径流预测及其不确定性分析

崇阳溪流域PRBP神经网络洪水预报模型研究

堤防工程险情风险评估与管控研究进展

基于洪水分类与产汇流机制增强的改进API-SCS-CN模型

融合Transformer深度学习与卫星遥感数据的桓仁水库洪水预报方法