災害時の人工衛星資料を用いた地被判読の精度評価と効率化について

抄録

Satellite imagery is often used in environmental research and post-disaster studies. Improved systems of post-disaster information dissemination are now in place, as exemplified by the immediate provision of data on slope failure after the Niigata Chuetsu Earthquake of 2004 and on the tsunami after the Sumatra Earthquake of 2004. The LANDSAT and SPOT (medium-resolution) satellites launched in the 1970s and 1980s, respectively, accumulated great amounts of image databases. Such databases are important in identifying changes in ground cover conditions. One such image can cover an area of over 3,000 km², which is useful in identifying environmental changes in river basins. However, with regard to these medium-resolution satellite images, few discussions have addressed the pixel density required to achieve the desired accuracy of image interpretation.

This study clarifies the accuracy of determining the area of a landform under different resolutions and reports the results of a method that uses GIS data to identify slope failure areas. The study found that an 8:1 ratio of object size to pixel density is required to accurately estimate the area of a landform (accuracy of 80% or more). Also, a method is proposed that enables easy extraction of landslide areas by effectively removing bare ground other than slope failure using commercially available GIS data.