基于K-means聚类的降雨时空特征提取对水文模型预测结果的影响

梁峰铭; 谭超; 黄锋华; 赵璧奎; 刘志敏; 程涛; 李泽君; 左斌斌

doi:10.13928/j.cnki.wrahe.2026.01.015

水利水电技术（中英文） ›› 2026, Vol. 57 ›› Issue (1) : 192 -204. DOI: 10.13928/j.cnki.wrahe.2026.01.015

水文水资源

基于K-means聚类的降雨时空特征提取对水文模型预测结果的影响

梁峰铭 ¹^,²^,³^,⁴ ,
谭超 ¹^,²^,³^,⁴ ,
黄锋华 ¹^,²^,³^,⁴ ,
赵璧奎 ¹^,²^,³^,⁴ ,
刘志敏 ¹^,²^,³^,⁴ ,
程涛 ¹^,²^,³^,⁴ ,
李泽君 ¹^,²^,³^,⁴ ,
左斌斌 ¹^,²^,³^,⁴

作者信息 +

The impact of precipitation spatiotemporal feature extraction based on K-means clustering on hydrological model predictions

Fengming LIANG ¹^,²^,³^,⁴ ,
Chao TAN ¹^,²^,³^,⁴ ,
Fenghua HUANG ¹^,²^,³^,⁴ ,
Bikui ZHAO ¹^,²^,³^,⁴ ,
Zhimin LIU ¹^,²^,³^,⁴ ,
Tao CHENG ¹^,²^,³^,⁴ ,
Zejun LI ¹^,²^,³^,⁴ ,
Binbin ZUO ¹^,²^,³^,⁴

Author information +

文章历史 +

PDF (12976K)

摘要

【目的】降雨作为水文模型的重要输入变量,其时空特性对模型的预测能力具有重要影响。研究降雨在不同时空尺度下的变化特征,对提高水文模型稳定性和预测精度具有重要意义。【方法】选择河南省下孤山流域为研究对象,利用SCS-CN水文模型和新安江模型模拟历史较大洪水场次,统计分析洪水场次的降雨时空特征和模型模拟精度,探讨降雨的时空变化对水文响应和水文模型模拟能力的影响。【结果】新安江模型和SCS-CN水文模型平均确定性系数分别为0.83和0.85,SCS-CN水文模型略优于新安江模型。两种水文模型对历时短、暴雨中心集中的洪水模拟效果最佳,而对暴雨中心接近流域出口的洪水模拟精度较低。【结论】低洪水条件下,两种水文模型精度受平均降雨强度影响较大。中洪水条件下,新安江模型受平均降雨强度和降雨空间变异系数影响较大,而SCS-CN水文模型主要受降雨位置指数影响。高洪水条件下,降雨位置指数是影响两种模型模拟能力的主要因素,贡献率分别为30.23%和33.97%。因此,可根据不同降雨的时空特征优化模型参数,以提升模型的模拟精度和可靠性。

Abstract

[Objective] Rainfall, as an important input variable in hydrological models, significantly affects the prediction capabilities of these models. Studying the temporal and spatial variation characteristics of rainfall at different scales is crucial for enhancing the stability and prediction accuracy of hydrological models. [Methods] The Xiagushan watershed in Henan Province was selected as the study area. Historical flood events were simulated using the SCS-CN and Xin'anjiang models. The temporal and spatial characteristics of rainfall and the simulation accuracy of the models were statistically analyzed to explore the impact of rainfall variability on the modeling capabilities of these hydrological models and hydrological response. [Results] The average coefficient of determination for the Xin'anjiang model and the SCS-CN model were 0. 83 and 0. 85, respectively, with the SCSCN model slightly outperforming the Xin'anjiang model. Both models showed the best performance in simulating floods with short durations and concentrated rainfall centers, whereas the simulation accuracy was lower when the rainfall center was closer to the watershed outlet. [Conclusion] Under low flood conditions, the accuracy of both hydrological models is significantly influenced by the average rainfall intensity. Under moderate flood conditions, the Xin'anjiang model is most affected by the average rainfall intensity and the rainfall spatial variability coefficient, while the SCS-CN model is mainly influenced by the rainfall location index. Under high flood conditions, the rainfall location index becomes the main factor affecting the simulation performance of both models, with contribution rates of 30. 23% and 33. 97%, respectively. Therefore, model parameters can be optimized based on the temporal and spatial characteristics of rainfall to improve the simulation accuracy and reliability of the models.

关键词

SCS-CN水文模型 / 新安江模型 / 降雨雨型 / 降雨时空分布特征 / 时空变化 / 气候变化 / 洪水预报 / 全球水文循环

Key words

SCS-CN hydrological model / Xin'anjiang model / rainfall pattern / temporal and spatial distribution characteristics of rainfall / spatiotemporal changes / climate change / flood forecasting / global hydrological cycle