Generating time-series land cover/land use dataset of SE Asia

点击翻译按钮获取中文摘要

1) Time series dataset of land use/land cover in 1970's, 1980's and 1990's for Indochina Peninsula are developed, though here are several areas with no data due to cloud cover of satellite imagery.2) Land use/land cover classes consisting of forest (evergreen, deciduous), paddy field, crop field, grassland/barren land, urbanized area/artificial structures and water body are discriminated. Size of grid cells are different by land cover/land use classes. 1km ^* 1km for forest and 500m ^* 500m for the others.3) Validation reveals that the overall classification accuracy is more than 80%. Forest have relatively higher accuracy, while crop field and grassland/barren land are relatively poorly discriminated. For the validation, 10 check sites with the size of approximately 180km by 180km are generated where detailed and reliable land cover/land useclassifications are conducted mainly by visual interpretation with aid of many external knowledge and information such as existing topo-maps and land use maps. Time-series land use/land cover dataset generated for these check sites.4) For partial check, ground-based photos are archived in a database by software developed by Chiba University.5) Methodologies and software are developed in a systematic manner for generating time-series land use/land cover dataset using Landsat MSS and TM,NOAA AVHRR data. Typical example are automated mosaic system and land use/land cover classification system.6) All data are now being edited and related documents are being generated for publishing as a CD-ROM and over the net.