北京大学学报自然科学版 ›› 2022, Vol. 58 ›› Issue (4): 753-762.DOI: 10.13209/j.0479-8023.2022.052

上一篇    下一篇

流域水文模型的参数全局敏感性分析

蔡开奎1, 李金城1, 胡梦辰1, 马文静2, 叶瑞2, 刘永1,†    

  1. 1. 北京大学环境科学与工程学院, 国家环境保护河流全物质通量重点实验室, 北京 100871 2. 南京智水环境科技有限公司, 南京 210012
  • 收稿日期:2021-08-16 修回日期:2021-09-15 出版日期:2022-07-20 发布日期:2022-07-20
  • 通讯作者: 刘永, E-mail: yongliu(at)pku.edu.cn
  • 基金资助:
    国家自然科学基金(51721006)资助 

Global Sensitivity Analysis of Hydrological Parameters of the Watershed Simualtion Model

CAI Kaikui1, LI Jincheng1, HU Mengchen1, MA Wenjing2, YE Rui2, LIU Yong1,†    

  1. 1. College of Environmental Sciences and Engineering, State Environmental Protection Key Laboratory of All Material Flux in Rivers, Beijing 100871

    2. Nanjing Innowater Co. Ltd., Nanjing 210012

  • Received:2021-08-16 Revised:2021-09-15 Online:2022-07-20 Published:2022-07-20
  • Contact: LIU Yong, E-mail: yongliu(at)pku.edu.cn

摘要:

基于LSPC构建内蒙大佘太流域水文模拟模型, 利用Morris和Sobol两种全局敏感性分析方法, 识别水文敏感参数及敏感下垫面类型, 评估不同敏感性分析方法和模型输出度量方法对敏感性分析结果的影响。主要结论如下: 1) 模型对日、月两尺度的径流模拟效果好, 模拟值与观测值的决定系数R2>0.6, 纳什系数NSE>0.5, 说明LSPC模型适用于对内陆干旱半干旱地区的流域水文模拟; 2) 两种敏感性分析方法对参数敏感指数排序和敏感参数识别均有影响, 而MAE和MSE两种度量方法的影响主要体现在敏感参数识别方面; 3) 敏感水文参数为下层土壤含水量(LZSN)和地下水蒸发系数(AGWETP), 敏感下垫面类型为草地、耕地、林地和水域, 与大佘太流域的降水和土地利用类型有关。

关键词: 全局敏感性分析, 水文参数, LSPC模型, 大佘太流域

Abstract:

A hydrological simulation model was developed for the Dashetai watershed in the Inner Mongolia Autonomous Region with LSPC (Loading Simulation Program in C++). Two global sensitivity analysis (GSA) methods, Morris and Sobol, were applied to identify hydrological sensitive parameters and sensitive surface landuse types. The impacts of two GSA methods and model output measures on sensitivity analysis results were also evaluated. The main conclusions are as follows. 1) The model has a good fit for daily and monthly runoff simulation. The R2 of the simulated value and the observed value is greater than 0.6, and NSE between them is greater than 0.5, which indicates that LSPC model is suitable for hydrological simulation of inland arid and semiarid areas. 2) The two GSA methods have impacts on the identification of sensitive parameters and the ranking of sensitive indexes; while the impact of the two measurement methods of MAE and MSE is mainly reflected in the identification of sensitive parameters. 3) The sensitive hydrological parameters are lower zone nominal storage (LZSN) and active groundwater evapotranspiration (AGWETP), and the sensitive underlying landuse types are grassland, farmland, forest and water. All have close relationships with precipitation and landuse types in the study area.

Key words: global sensitivity analysis, hydrological parameters, LSPC model, Dashetai watershed