Recognition of Spectral Pattern Characteristic of Land Cover for Assisting Visual Interpretation of Landsat ETM^+ :A Forest Degradation Mapping in Tropical Rain Forest of Sabah, Malaysia

概要

論文の詳細を見る
The study was originally designed to assess vegetation diversity in Borneo's tropical rain forest with ecological consequence that can be retrieved from Landsat ETM^+. We were particularly interested in recognition of spectral pattern of land cover for developing land cover and forest degradation mapping by visual analysis of satellite imagery in Sabah, Malaysia. In addition unsupervised, supervised classification, band ratio of TM4, TM4/TM5, and TM2/TM7, and pattern decomposition analysis were tested to evaluate the interpretation. We proposed a reliable method of forest degradation mapping in Sabah from land cover map incorporated with their spectral pattern. Dense forest was found clearly discriminated from water and soil by using visual interpretation and pattern decomposition analysis, particularly in channel 3, 4, and 5. However, it was difficult to discriminate spectral pattern of non-forest through decomposition analysis. Logged-over forest was found dominant in the study area (36%), followed by plantation (21%), and only about 17% covered with natural dense forest. It is estimated that approximately 76% of the forest in Sabah, have been threatened with forest degradation.
森林計画学会の論文