关键词: 地表覆盖, 自发地理信息, GlobeLand 30, OpenStreetMap, 一致性

Abstract:

Land cover data, which plays a significant role in national geographical condition monitoring, ecological environmental protection and some other areas, is an important resource of the information on land use. At present, land cover data is produced mainly through the interpretation of remote sensing imagery and field measurement, and some limitations still exist to a certain extent. With the rapid development and wide spread of Web2.0, internet technology and various kinds of GPS equipment, the general public have the opportunity to participate in crowd sourced mapping. Volunteer users can identify the spatial distribution and attributive characters of the land cover effectively. Therefore, the classification accuracy of land cover map can be improved in the meantime. In this paper, OpenStreetMap, the most successful item of volunteered geographic information, was taken as an example on the comparison analysis with GlobeLand 30, the newly developed land cover data produced in China with 30m resolution. Firstly, the data was preprocessed and topologically checked, and then the feature relationship was established. Finally, a confusion matrix was built to analyze the consistency between the two kinds of data. The experimental results show that the missing part of the OpenStreetMap is mainly cultivated land, and the grassland and water elements are more abundant than those of GlobeLand 30. The consistency of OpenStreetMap and GlobeLand 30 is high with a value of 75%. Forest and artificial surface have the highest accuracy, and cultivated land and water body take the second place, while grassland possesses the worst consistency. The key point is to verify and determine the land cover type within the inconsistent areas, and try to find classification errors of GlobeLand 30 so as to provide basis for further modification and optimization. Volunteered geographic information contains abundant land cover information, so it can provide great potential for the development and evaluation of land cover maps. The research methods and conclusions of this paper can provide basis for exploring the application of OpenStreetMap data for land cover mapping, and provide support for assessment and improvement of the classification accuracy of GlobeLand 30.

Key words: land cover, volunteered geographic information, GlobeLand 30, OpenStreetMap, consistency