地理学报 ›› 2018, Vol. 73 ›› Issue (11): 2075-2085.doi: 10.11821/dlxb201811003

• 气候变化与地表过程 • 上一篇    下一篇

滑动平均差检测法的提出及验证

包为民(),沈丹丹,倪鹏,周俊伟,孙逸群   

  1. 河海大学水文水资源学院,南京 210098
  • 收稿日期:2017-12-25 出版日期:2018-11-25 发布日期:2018-11-22
  • 基金资助:
    国家重点研发计划专项(2016YFC0402703);中央级公益性科研院所基本科研业务专项(HKY-JBYW-2017-12);江苏省研究生科研与实践创新计划项目(KYCX17_0423);中央高校基本科研业务费专项资金(2017B 684X14)

Proposition and certification of moving mean difference method for detecting abrupt change points

BAO Weimin(),SHEN Dandan,NI Peng,ZHOU Junwei,SUN Yiqun   

  1. College of Hydrology and Water Resources, Hohai University, Nanjing 210098, China
  • Received:2017-12-25 Online:2018-11-25 Published:2018-11-22
  • Supported by:
    The National Key R&D Program of China, No.2016YFC0402703, The Fundamental Research Funds for Central Public Welfare Research Institutes, No.HKY-JBYW-2017-12, Postgraduate Research & Practice Innovation Program of Jiangsu Province, No.KYCX17_0423; The Fundamental Research Funds for the Central Universities, No.2017B684X14

摘要:

时间序列突变点的检测研究是当前气候变化问题的关键环节,其方法和成果也较多,传统的检测方法因统计量结构不合理、假设条件不符合实际以及受主观因素影响等,结果常常缺乏可信度。本文提出一种新的时间序列突变点的检测方法——滑动平均差检测法,将该方法分别用理想时间序列和实际观测序列进行应用检验,并与现有常用的4种突变检验方法比较,发现滑动平均差法具有:① 结构简单、物理意义明确;② 检测突变点更精确;③ 能同时检测出所有突变点的突变位置和其突变强度3个明显的优势。综合利用滑动平均差检测法和传统的4种方法,对黄河流域几个重点水文控制站实测输沙量资料进行突变分析,结果表明:黄河上游输沙量在1986年发生突变,黄河中游输沙量在1979年和1996年发生突变,华县站和潼关站输沙量在1979年、1996年和2003年发生突变,造成输沙量突变的主要因素为水利工程建设和大规模水土保持措施等人类活动。

关键词: 滑动平均差检测法, 时间序列, 突变检验, 黄河流域, 输沙量

Abstract:

The abrupt change point detection is an important content in the research of current climate change problem. And many methods and achievements have already been proposed. The results of traditional methods are often considered to be unpersuasive due to the irrational structure of statistics, ideal assumptions and subjectivity. In this paper, a new method called moving mean difference method for abrupt change points detection is proposed. It is compared with four widely used methods via both hypothetic series and observed data. The results show that the moving mean difference method has three advantages: (1) The method has a simple structure and definite physical meaning; (2) It is more accurate to detect the mutation point; (3) It can simultaneously detect all the mutation points and calculate the corresponding intensity of the mutations. The sediment discharge abrupt changes in different reaches of the Yellow River are analyzed through the five methods, based on the annual data obtained from the main hydrological stations in the Yellow River Basin. The results indicate that the sediment discharge in the upper reaches the river mutated in 1986, while that in the middle reaches changed suddenly in 1979 and 1996. The mutations of sediment discharge at Huaxian and Tongguan stations occurred in 1979, 1996 and 2003. The main factors for the mutation are human activities, including hydraulic engineering construction and large-scale water and soil conservation measures.

Key words: moving mean difference method, time series, abrupt change point detection, Yellow River, sediment discharge