额尔齐斯河库威站日尺度的降雨融雪径流模拟

Simulation of rainfall and snowmelt runoff on the daily scale of the Kuwei Station in the Irtysh River

额尔齐斯河流域受地理条件的影响,流域内水文气象站点较少,基础资料匮乏,而融雪洪水在该流域的汛期及水资源管理上有着较大影响。本研究通过应用降水和气温的再分析产品及AVHRR积雪数据,利用K-means聚类法进行不同径流时期特点的划分,并在不同时期构建相应SRM+LSTM模型,并使用2009年数据及2023年实地观测的径流数据进行验证。结果表明:再分析产品CMFD能够较好地应用于额尔齐斯河流域,并能根据降水、温度、积雪及径流间的关系得到不同径流划分时期,即12月11日—次年4月10日为积雪退水期、4月11日—8月10日为融雪降水产流期、8月11日为降水产流期。SRM模型模拟效果较差,大部分径流纳什效率系数(NSE)<0;而SRM+LSTM模型能够较好地模拟该流域的不同时期的径流,决定系数R2均能达到0.5以上,纳什效率系数也能达到0.5以上,证明SRM+LSTM模型能够较好地应用于该地区,精度较高。

Due to geographical conditions,there are limited hydrometeorological stations and a lack of basic data in the Irtysh River Basin,and the snowmelt flood exerts a considerable effect on the flood season and water resources management in the basin.In this study,precipitation and temperature reanalysis products and AVHRR snow cover data were applied,the K-means clustering method was used to divide the characteristics of different runoff periods,the corresponding SRM+LSTM model in different periods was constructed,and the runoff data observed in the field in 2023 were used.Results showed that the reanalysis product CMFD can be well applied to the Irtysh River Basin according to precipitation and temperature.The relationship between snow cover and runoff was divided into different runoff periods,as follows:December 11th to April 10th of the following year was the snow retreat period,April 11th to August 10th was the snowmelt precipitation runoff period,and August 11th was the precipitation runoff period.The simulation effect of the SRM model was poor,and the Nash efficiency coefficient of most runoff was<0.The SRM+LSTM model could better simulate the runoff in different periods of the basin,the deterministic coefficient could reach>0.5,and the Nash efficiency coefficient NSE could also reach>0.5,which confirms that the SRM+LSTM model can be better applied to the area with high accuracy.