ランドサットMSSデータの植生解析への応用 : 福岡県北西部の例

抄録

An examination was made of the validity of Landsat multispectral scanner (MSS) data (May 17,1985) for vegetation classification in the northwestern part of Fukuoka Prefecture, Japan. The digital data (number) of four MSS bands were analyzed by the multi-level slice method, and classified into 8 units : evergreen broad-leaved forest (consisting of Castanopsis cuspidata and Cinnamomum camphora), pine forest (Pinus densiflora), plantation forest (Cryptomeria japonica, Chamaecyparis obtusa), deciduous broad-leaved forest (Fagus crenata, Carpinus laxiflora, C. tschonoskii, Quercus serrata), grassland (semi natural grassland, paddy and upland field), bare land, urban and residential area and open water (sea and river). The ground survey revealed the vegetation and land-use classification obtained by Landsat data to be 83% accurate. However, it was frequently observed that plantation forests located on slopes facing from south to cast were incorrectly classified as evergreen broad-leaved forests. These findings indicate that Landsat MSS data is valid for the vegetation classification at a physiognomical level, but requires some calibration of its digital data in relation to slope aspects.