Validating land cover maps with Degree Confluence Project information

Abstract

We propose the use of Degree Confluence Project (hereby DCP) information as a new method for validating land cover maps. The DCP is a volunteer-based project that aims to collect onsite information from all the degree confluences (intersections of integer level latitude and longitude gridlines) in the world. We assessed the reliability and effectiveness of DCP-derived data in validating land cover maps. As a result, we obtained land cover validation information superior to the validation information obtained by visual interpretation of Landsat images. By using DCP-derived validation information (at 749 confluences), we evaluated existing land cover maps for Eurasia (GLC2000, MOD12, UMD, and GLCC). The agreements between the DCP-derived validation information and the land cover maps were 55% for GLC2000, 58% for MOD12, 54% for UMD, and 50% for GLCC. Although MOD12 and GLC2000 had somewhat better agreements than the other maps, there is no significant difference between the two. Copyright 2006 by the American Geophysical Union.