Grid map analysis and its application for detecting vegetation changes in Japan

. Grid maps are used as a basic vegetation data base in Japan; they are simplified from vector‐based vegetation maps. We estimated the frequency error or lack of information corresponding to reduced resolution and examined the reliable limits of this data base. We produced 10 grid maps on five different scales from 50 m to 1000 m using two different methods using both the whole cell ( W ‐method) and only the central circle ( C ‐method) from a vegetation map at scale 1: 25 000. We found that patches larger than the area of a cell on a vector‐based map could be kept almost certainly on any map, but many patches of less than the cell size were lost. The number of missing patches with the C ‐method is fewer at every scale than those with the W ‐method. Though the value of Morisita's C λ ( p ) index showed that the similarity with the original map was high ‐ from the 50‐m to the 200‐m resolution ‐ it was increasingly lower on the 400‐m and 1000‐m grid maps. The values of the Shannon index on the original map, 50‐m and 100‐m grid maps were not different, but they decreased from the 200‐m to 1000‐m grid maps. Because the vegetation data base of the Japanese Environment Agency used a 1000‐m C ‐method grid map, we found that much information on patches less than 100 ha had disappeared. Information about dominant vegetation or large patches is almost accurate in this data base.