ランドサットデータMSSデータを用いた早川流域の斜面崩壊の研究

新規崩壊地の探索

抄録

Japanese land has been attacked frequently by natural disasters caused mainly by earthquakes, volcanic activities, typhoon and heavy rain. Landslides due to the heavy rain occur very often, bringing about sometimes rather serious damage.
Landslides in mountainous districts can be studied by using the aerial photograph. But this method has an economic disadvantage for studying landslides over a wide area. Therefore the authors have used Landsat MSS data because of their wide coverage and the acquistion of repeated data.
This research is aimed at pursuing the possibility of identifing landslide by Landsat MSS data. The Hayakawa watershed on the branch of Fuji River in Yamanashi prefecture was chosen as the test area. The MSS data of Oct. 1981 and Oct. 1982 are used in this study. After the comparison between these two MSS data, changes of the CCT counts were observed in 951 pixels, which indicate the change of vegetation coverage into a naked area after landslides. But when checked with other data from investigations by air survey, it was recognized that those pixels show not only the landslides but also the transmutations in the surface conditions.
By projecting those 951 pixels on the stereoscopy image of Landsat MSS data, the authors interpreted that 211 pixels were caused by landslides, 202 pixels show the changes of the riverbed conditions, 144 pixels identified as the chenged conditions of the stream, 44 pixels as those of the rocks on the mountaintop and 13 pixels are determained some chenges of vegetation on a gentle slop. But 367 pixels can not be classified.
For statistical method, the authors employed the analysis by the models of quantification type 2, in which not only the CCT counts of band 7 and band 5 of MSS data but also the ground level, directions and grades of the slopes and the elements of geology are incorporated as factor items. By this analysis, the following useful results were obtained : cross relations in frequency between the categories of each factor items ; the weight of each categories ; and the range of each factor items. From the combined results of the stereoscopy image identification and statistical analysis, the identifications of the landslide pixels and non -landslide pixes were performed . The 284 pixels in 367 pixels which had not been classified by the stereoscopic image method were identified as the non - landslide pixels by statistical analysis.