地理学报 ›› 2011, Vol. 66 ›› Issue (4): 549-561.doi: 10.11821/xb201104011

• 土地利用 • 上一篇    下一篇

基于知识规则的土地利用/土地覆被分类方法——以黑河流域为例

候玉婷1, 王书功2, 南卓铜3   

  1. 1. 兰州大学资源环境学院干旱区水循环与水资源研究中心, 兰州730000;
    2. 匹兹堡大学土木与环境工程系, 美国匹兹堡15261;
    3. 中国科学院寒区旱区环境与工程研究所, 兰州730000
  • 收稿日期:2010-07-22 修回日期:2010-12-20 出版日期:2011-04-20 发布日期:2011-06-01
  • 通讯作者: 南卓铜(1977-), 男, 博士, 研究员, 现从事空间决策支持系统、水文集成建模研究。 E-mail:nztong@lzb.ac.cn
  • 作者简介:候玉婷(1988-), 女, 硕士生, 主要从事黑河流域水文模拟研究。E-mail: yuting.hou88@gmail.com
  • 基金资助:

    科技部863 项目(2008AA12Z205); 中科院重要方向项目(KZCX2-YW-Q10-1)

A Rule-based Land Cover Classification Method for the Heihe River Basin

HOU Yuting1, WANG Shugong2, NAN Zhuotong3   

  1. 1. Center for Hydrologic Cycle and Water Resources Research in Arid Region, College of Earth and Environmental Science, Lanzhou University, Lanzhou 730000, China;
    2. Department of Civil and Environmental Engineering, University of Pittsburgh, Pittsburgh, PA, 15261, USA;
    3. Cold and Arid Regions Environmental and Engineering Research Institute, CAS, Lanzhou 730000, China
  • Received:2010-07-22 Revised:2010-12-20 Online:2011-04-20 Published:2011-06-01
  • Supported by:

    MOST 863 Project No.2008AA12Z205; Knowledge Innovation Project CAS, No.KZCX2-YW-Q10-1

摘要: 本文提出了一种基于知识规则的土地利用/土地覆被分类的新方法。知识规则是基于专家经验建立起来的,反映研究区内不同分类系统下各类别的地理分布特性与地理分布交叉可能性。基于黑河流域90 m 分辨率DEM、2009 年逐月1 km 分辨率NDVI,参考美国地质调查局(USGS) 1 km分辨率土地利用/土地覆被数据在欧亚大陆上各类别的聚类中心,应用在上、中、下游分别建立的知识规则,以知识规则结合最近距离的USGS 类别聚类的方法,制作了一套与USGS全球土地覆被分类标准一致的、可以用于大气模式以及陆面过程模式的黑河流域土地覆被类型分布数据。本方法分类结果与以往研究采用的类别映射方法的分类结果及实际地物影像进行对比,表明知识规则下的分类结果更能准确表达流域地表覆盖特征,对冰雪、冻土类别和沙地荒漠类别的表现更优。

关键词: 黑河流域, 土地利用/覆盖, 知识规则, USGS土地利用/土地覆被数据, 最小距离分类法

Abstract: A novel rule-based land use/land cover classification approach is presented in this study. Rule tables were generated based on geographic characteristics of each class of the China land use classification schema and its possible transferability into other classes of the USGS schema. The USGS land use/land cover (LULC) data product, with a 1-km spatial resolution, was used to locate clustering centers, referred as NDVI fingerprints, of each land use class. A minimum distance approach was then applied to the 1 km NDVI of the year 2009 and 90 m DEM of the Heihe River Basin (HRB), with rule tables considered, to produce a land use/land cover map with schema and attributes consistent with USGS's. The produced map can be used in atmospheric models and land surface models. A comparison to the previous work and satellite images indicates that our rule-based approach is better in distinguishing land cover characteristics, especially for snow-cover, frozen soil and desert types.

Key words: Heihe River Basin, land use/land cover, rule-based classification approach, USGS LULC dataset, the minimum distance classifier