地球信息科学学报 ›› 2012, Vol. 14 ›› Issue (2): 240-249.doi: 10.3724/SP.J.1047.2012.00240

• 遥感技术与应用 • 上一篇    下一篇

MODIS和GLOBCOVER全球土地覆盖数据集对比分析——以黑龙江流域为例

宁佳1 2, 张树文*1, 蔡红艳3, 卜坤1   

  1. 1. 中国科学院东北地理与农业生态研究所, 长春 130012;
    2. 中国科学院研究生院, 北京 100049;
    3. 中国科学院地理科学与资源研究所, 北京 100101
  • 收稿日期:2012-04-24 修回日期:2012-04-24 出版日期:2012-04-25 发布日期:2012-04-24
  • 作者简介:宁佳(1987),女,硕士研究生,主要从事地理信息系统应用和土地利用变化研究。Email:ningjia09@mails.gucas.ac.cn
  • 基金资助:

    国家科技基础性工作专项重点项目专题"中国北方及其毗邻地区LUCC遥感制图"(2007FY110300-1-2)和国家"973"项目专题"俄罗斯西伯利亚LUCC数据生成、验证及LUCC驱动机制分析"(2010CB950901-2-1)。

A Comparative Analysis of the MODIS Land Cover Data Sets and Globcover Land Cover Data Sets in Heilongjiang Basin

NING Jia1 2, ZHANG Shuwen1, CAI Hongyan3, BU Kun1   

  1. 1. Northeast Institute of Geography and Agroecology, CAS, Changchun 130012, China;
    2. Graduate University of Chinese Academy of Sciences, Beijing 100049, China;
    3. Institute of Geographic Sciences and Natural Resources Research, CAS, Beijing 100101, China
  • Received:2012-04-24 Revised:2012-04-24 Online:2012-04-25 Published:2012-04-24
  • Contact: 张树文(1955-),男,研究员,主要从事环境与资源遥感、地理信息系统应用及土地变化科学研究。 E-mail: zhangshuwen@neigae.ac.cn E-mail:zhangshuwen@neigae.ac.cn

摘要: 随着全球气候变化的日益加剧,全球变化研究对全球土地覆盖数据的需要也越来越迫切。目前全球土地覆盖数据产品主要包括由欧洲和美国生产的5类数据产品,其中,美国波士顿大学生产的全球土地覆盖数据产品(即MODIS数据集)和欧洲空间局通过全球合作生产的全球土地覆盖数据产品(即Globcover数据集)具有较好的实效性,应用越来越广泛。由于数据来源、分类系统和分类方法不同,两个数据集在土地覆盖类型的数量和空间分布上有明显的差异。本研究从数据使用者的角度,对比了MODIS和Globcover数据集在黑龙江流域上数量和空间分布的差异,并采用LANDSAT TM/ETM+影像随机采样和野外照片验证两种方式对两个数据集的分类精度进行了验证。结果表明,在黑龙江流域,两个数据集数量和空间分布差异较大。在数量上,两个数据集一级土地覆盖类型均以森林和农田为主,草地次之,但二级土地覆盖类型差异较大。在空间上,二级类空间一致性区域和一级类空间一致性的区域分别仅占流域的22.5%和53.6%。两个数据集精度均不高,一级土地覆盖类型精度约为60%,Globcover数据较MODIS数据破碎化明显,整体精度略低于MODIS数据集,不同的二级土地覆盖类型精度不同。考虑到黑龙江流域的代表性,我们认为Globcover数据集和MODIS数据集可满足较低要求的土地覆盖分析需求。本研究为全球气候变化研究选择合适的数据集提供了基础。

关键词: Globcover数据集, MODIS数据集, 一致和不一致, 精度验证, 黑龙江流域

Abstract: As global change is becoming serious, the global land cover data sets which the global change research needs have become more and more important. The current global land cover data products mainly include five products which are produced by European and U.S. The global land cover data sets produced by Boston University (i.e. MODIS data sets) and by European Space Agency with global cooperation (i.e. Globcover data sets), which are somehow better and more effective than the others, are used more and more widely. Because of the difference of data sources, classification systems and classification methods, there are significant differences between two land cover data sets both in quantity and spatial distribution. In this study, as a products' user, we compared the MODIS and Globcover data sets in Heilongjiang Basin on the quantity and spatial distribution, using LANDSAT TM / ETM + images and field photos to verify the accuracy of the classification of two data sets. The results show that, in Heilongjiang Basin, there are great differences in the quantity and spatial distribution between the two data sets. As to the quantity, the main land cover types (forest and cropland, followed by grassland) in both data sets are similar. When it comes to the spatial distribution, they are of great difference. The two data sets are both of low precision, with a precision of only 60%. The Globcover data sets is more clustering than MODIS data sets, and the overall accuracy is slightly lower than MODIS data sets, while the accuracy of different land cover types is different. Taking into account the representation of Heilongjiang Basin, we believe that Globcover data sets and MODIS data sets are able to meet lower demand for land cover information. This study provides a basis for global change research in selecting the appropriate data sets.

Key words: Globcover global data sets, MODIS global data sets, agreement and disagreement, accuracy validation, Heilongjiang Basin