SEISMOLOGY AND GEOLOGY ›› 2014, Vol. 36 ›› Issue (4): 1029-1042.DOI: 10.3969/j.issn.0253-4967.2014.04.008

• CONTENTS • Previous Articles     Next Articles

A CLUSTERING APPROACH FOR INCORPORATING SPATIAL DEPENDENCE INTO THE AUTOMATIC FLUVIAL TERRACE EXTRACTION FROM DIGITAL ELEVATION MODEL: A CASE STUDY FROM THE ZHANGLA BASIN ALONG THE UPSTREAM OF THE MINJIANG RIVER

SONG Zhuo-qin1, ZHANG Jun-long1, LI Jian-jun2   

  1. 1. Institute of Earthquake Science, China Earthquake Administration, Beijing 100036, China;
    2. School of Science, North University of China, Taiyuan 030051, China
  • Received:2014-04-01 Online:2014-12-30 Published:2015-01-07

基于模糊聚类和空间自相关性的河流阶地自动提取方法 ——以岷江上游漳腊盆地为例

宋卓沁1, 张军龙1, 李建军2   

  1. 1. 中国地震局地震预测研究所, 北京 100036;
    2. 中北大学理学院, 太原 030051
  • 通讯作者: 张军龙,副研究员,E-mail:zhjulo_2002@163.com
  • 作者简介:宋卓沁,女,1988年生,现为中国地震局地震预测研究所在读硕士研究生,研究方向为构造地貌,电话:010-88015246,E-mail:176578721@qq.com.
  • 基金资助:

    国家自然科学基金(41372215)、中国地震局地震行业科研专项(201408023)和山西省基金资助项目(2012012003)共同资助.

Abstract:

The terraces along the upper reaches of the Minjiang River record rapid uplift around the Tibetan plateau since the Quaternary. However, the common method to obtain the terraces elevation data always relies on single point or line to represent the whole landform. The available results and further analyses are usually not continuous and systematic. All of these, therefore, restrict the development of further study. A large number of qualitative information and the digital terrace models are relied on the improvement of DEM processing technology. Our present paper applies fuzzy C-means algorithm to a bunch of the Minjiang River terrace cross sections in different dimensions to get the integrated distribution characteristics of the terraces in the Zhangla Basin. SPOT5 image is adopted when generating the DEM. Then we consider both the spatial correlation and distribution characteristics of terrace cross sections into analysis and we consequently find a successful way to extract different levels of terraces based on both whole indicator and internal correlation. The precision evaluation suggests that result of terrace extraction is highly consistent with the field survey data. Based on these, we discuss relationship between characteristics of terraces, the incision rates and the regional uplift patterns. We suggest that a clustering approach for incorporating spatial dependence into the automatic fluvial terrace extraction can be used to study and understand the regional tectonic and geomorphic features in the layered landscapes.

Key words: fluvial terrace, fuzzy C-means analysis, DEM, spatial correlation, Zhangla Basin

摘要:

传统研究方法以点、线代表区域地貌面, 数据缺乏连续性和系统性, 限制了对阶地演化的进一步研究.高分辨率DEM处理方法的发展使广泛精确获取地貌面的定量信息、重建阶地模型、进一步研究阶地演化成为可能.以岷江上游漳腊盆地为研究区域, 采用SPOT5高分辨率数据影像构建DEM, 将统一标度后的阶地横剖面作为基本研究单元.通过在不同维度使用模糊C均值聚类, 得到研究区域内阶地总体形态特征.在提取各级阶地面的过程中, 将阶地剖面间的空间相关性与阶地总体形态特征纳入分析, 最终建立了一种基于区域总体特征和区域内部自相关的阶地提取方法.对不同级阶地面提取的精度评价表明, 其结果与野外实测数据有较高的一致性.文中以实验结果为依据, 进一步讨论了漳腊盆地内阶地的形态特征、河流下切速率和区域隆升之间的关系.因此, 将模糊聚类分析和空间自相关性结合的区域地貌面提取方法是简捷易行的, 适用于阶地等层状地貌面发育区域构造-地貌等相关问题的研究.

关键词: 河流阶地剖面, 模糊C均值聚类分析, 数字高程模型, 空间相关性, 漳腊盆地

CLC Number: