宝鸡市2024年“7·16”暴雨清姜河水位演变特征  

Characteristics of the water level evolution in Qingjiang River during the rainstorm on July 16,2024 in Baoji City

作　　者：张翀[1,2] 谢娅 张俊辉 黄云博[1] 韩子逸 张嘉珊 朱永乐 上官书睿 ZHANG Chong;XIE Ya;ZHANG Jun-hui;HUANG Yun-bo;HAN Zi-yi;ZHANG Jia-shan;ZHU Yong-le;SHANGGUAN Shu-rui(College of Geography and Environment,Baoji University of Arts and Sciences,Baoji 721013,Shaanxi,China;Shaanxi Key Laboratory of Disaster Monitoring and Mechanism Simulation,Baoji 721013,Shaanxi,China)

机构地区：[1]宝鸡文理学院地理与环境学院,陕西宝鸡721013 [2]陕西省灾害监测与机理模拟重点实验室,陕西宝鸡721013

出　　处：《宝鸡文理学院学报(自然科学版)》2025年第1期92-101,共10页Journal of Baoji University of Arts and Sciences(Natural Science Edition)

基　　金：陕西省自然科学基础研究计划项目(2021JM-513)。

摘　　要：目的探究清姜河流域极端暴雨事件中降水、土壤湿度与水位变化的耦合关系,验证长短期记忆网络(LSTM)在水位模拟中的适用性,为洪水预警提供科学支撑。方法基于2023年6月-2024年7月流域气象水文数据(降雨量、土壤湿度、水位),采用多元线性回归、通径分析及时滞分析揭示水文要素相互作用机制,构建LSTM模型模拟水位并与传统方法对比。结果(1)极端暴雨中,0~1 h累积降雨和40 cm深度土壤湿度剧增是水位激增主因;(2)土壤湿度对水位影响随深度异质:10 cm和30 cm呈正相关,20 cm和40 cm呈负相关;(3)LSTM模型在水位稳定期精度高,但对极端降水响应存在1 h时滞,且未考虑人类活动导致预测值偏高。结论LSTM模型适用于常规水文模拟,但极端事件中需融合土壤湿度时空异质性和人类活动因子。时滞效应是水位预测的关键参数,本研究为流域洪水预警系统优化提供理论依据。Purposes—To investigate the coupling relationships among precipitation,soil moisture and the changes in water level during extreme rainstorm events in the Qingjiang River basin,validate the applicability of the long short-term memory(LSTM)network in water level simulation,and provide scientific support for flood prewarning.Methods—Based on meteorological and hydrological data(such as rainfall,soil moisture and water level)from June 2023 to July 2024 in the basin,multivariate linear regression,path analysis,and time-lag analysis were employed to reveal the interaction mechanisms among hydrological elements.An LSTM model was developed to simulate water levels and compared with traditional methods.Results—(1)During extreme rainstorms,the cumulative rainfall within 0~1 h and a sharp increase in soil moisture at 40 cm depth were the primary drivers of the surge in water level;(2)The impact of soil moisture on water levels varied with depth:Positive correlations were observed at 10 cm and 30 cm depths,while negative correlations occurred at 20 cm and 40 cm depths;(3)The LSTM model achieved high accuracy during stable water level periods but exhibited a 1-hour time lag in responding to extreme rainfall events,with overestimated predictions due to unaccounted human activities.Conclusions—The LSTM model is suitable for routine hydrological simulations,but its performance in extreme events requires integration of spatiotemporal heterogeneity in soil moisture and the factors of human activities.The time-lag effect is a critical parameter in water level prediction.This study provides a theoretical foundation for optimizing flood prewarning systems in the basin.

关 键 词：水位模拟 时滞分析 LSTM 清姜河流域 

分 类 号：TV147.1[水利工程—水力学及河流动力学]