Discrimination between active and non-active faults based on the chemical composition of fault gouge; An Attempt using linear discriminant analysis

Tateishi, Ryo*; Shimada, Koji   ; Shimizu, Mayuko   ; Ueki, Tadamasa*; Niwa, Masakazu   ; Sueoka, Shigeru   ; Ishimaru, Tsuneari 

We attempted to discriminate between active and non-active faults by linear discriminant analysis using the chemical composition data of fault gouges in Japan, and then examined the elements that represent the difference between them and better discriminants. As a result, the multiple discriminants obtained could discriminate between them with high probability. In addition, the generalization performance of these discriminants is discussed, and the discriminants that can be expected to have high discriminant performance for unknown samples are presented. Also, from the combination of elements common to these discriminants, we narrowed down the number of elements that represent the difference between active and non-active faults to 6, and showed that the combination of TiO and Sr contributing the most to the discrimination. The method applied in this study is an innovative one that can discriminate the activity by chemical analysis of fault rocks that are universally present in the bedrock.