地球信息科学学报 ›› 2014, Vol. 16 ›› Issue (3): 460-469.doi: 10.3724/SP.J.1047.2014.00460

• 遥感科学与应用技术 • 上一篇    下一篇

基于决策树方法的蒙古高原土地覆盖遥感分类——以蒙古国中央省为例

田静1, 王卷乐2, 李一凡2,3, 周玉洁1, 郭海会1, 祝俊祥2,3   

  1. 1. 中国矿业大学北京地球科学与测绘工程学院, 北京100083;
    2. 中国科学院地理科学与资源研究所资源与环境信息系统国家重点实验室, 北京100101;
    3. 中国科学院大学, 北京100049
  • 收稿日期:2013-07-29 修回日期:2013-10-29 出版日期:2014-05-10 发布日期:2014-05-10
  • 通讯作者: 王卷乐(1976- ),男,河南洛阳人,博士,副研究员,研究方向为地理信息共享与遥感应用。E-mail:wangjl@igsnrr.ac.cn E-mail:wangjl@igsnrr.ac.cn
  • 作者简介:田静(1987- ),女,宁夏银川人,硕士生,研究方向为GIS与遥感应用。E-mail:tjalwaysuse@hotmail.com
  • 基金资助:

    中国科学院重点部署项目(KZZD-EW-08);中国科学院2013、2014年度俄乌白人才专项项目;中国科学院信息化专项项目(XXH12504-1-01)。

Land Cover Classification in Mongolian Plateau Based on Decision Tree Method: A Case Study in Tov Province, Mongolia

TIAN Jing1, WANG Juanle2, LI Yifan2,3, ZHOU Yujie1, GUO Haihui1, ZHU Junxiang2,3   

  1. 1. College of Geoscience and Engineering, China University of Mining & Technology Beijing, Beijing 100083, China;
    2. State Key Laboratory of Resources and Environmental Information System, Institute of Geographic Sciences and Natural Resources Research, CAS, Beijing 100101, China;
    3. University of Chinese Academy of Sciences, Beijing 100049, China
  • Received:2013-07-29 Revised:2013-10-29 Online:2014-05-10 Published:2014-05-10

摘要:

蒙古高原包括蒙古全部、俄罗斯南部和中国北部部分地区。蒙古高原的土地利用/覆盖格局与变化,对揭示该区域乃至整个东北亚地区的资源、环境和生态特征,促进该区域可持续发展具有重要的现实和科学意义。本文以在蒙古国中央省及其所含首都乌兰巴托市为研究区,利用空间分辨率为30m的TM影像,采取QUEST(Quick Unbiased and Efficient Statistical Tree)决策树方法,通过图像目视解译,获取了研究区2010年土地覆盖分类数据。结果显示,草地占据研究区总面积的70.88%,其次是森林占14.83%、裸地占10.73%、农田占2.98%、水体占0.31%、建筑用地占0.27%、湿地占0.02%。通过野外实地采集的139个GPS验证点进行精度评价发现,一级土地覆盖类型的总体精度可达72.66%。针对草地的二级分类的总体精度有较明显下降,其主要是由于中蒙科学家对于草地类型分类体系的差异所造成的典型草地和荒漠草地的混分。

关键词: 土地覆盖, 蒙古高原, 遥感, QUEST

Abstract:

Global and regional land use/cover data processed by remote sensing images is a supportive part for the researches related to environment and so on. Mongolia Plateau includes all of Mongolia, parts of southern Russia and northern China. Study on land use/cover changes in Mongolia Plateau has important practical and scientific significance to discover the resources, environmental and ecological characteristics of this area and even in northeast Asia area, and also it is beneficial to enhance international cooperation in regional resources utilization, ecological and environmental protection and sustainable development. According to the land use/cover mapping requirement in Mongolia Plateau, a land cover classification approach was designed based on QUEST (Quick Unbiased and Efficient Statistical Tree) decision tree method in the representative area of Mongolia Plateau, where includes Tov Province and Ulaanbaatar City in Mongolia. Land cover classification dataset in this area was captured using Landsat TM images, through QUEST automatic classification and visual interpretation. Results show that the total area of steppe occupies 70.88%, followed by forest accounting for 14.83%, barren 10.73%, cropland 2.98%, water 0.31%, built area 0.27%, and wetland 0.02%. 139 GPS verification points for accuracy assessment were collected by fieldwork, which was held in August 2013. Accuracy assessment found that land cover overall accuracy is up to 72.66% in class I, while the overall accuracy in class II is obviously descent, mainly due to the confusion of real steppe and desert steppe. This was caused by the different understanding of these two steppe types between scientists from China and Mongolia. In general, the QUEST decision tree method used in this paper is proved to be feasible for higher accuracy land cover classification mapping in this area.

Key words: remote sensing, Mongolian Plateau, QUEST, land cover