5:15 PM - 6:30 PM
[SSS10-P01] Relationship between fault activity and chemical composition of fault gouge: An attempt by linear discriminant analysis
Keywords:Fault gouge, Active fault, Chemical composition
Introduction: Active faults are identified by the displacement of the modern topography and the deformation of the late Quaternary beds. However, it is difficult to determine the activity of faults in areas where the beds are absent. In this case, the difference between an active fault and an inactive fault can be found by focusing on the physical and chemical properties of the fault. The major difference between them is the time elapsed since the latest event, which is generally considered to be in the order of 100 to 104 years for active faults, while it is 105 years or more for inactive faults. Therefore, even if the phenomena caused by the event are the same, the cumulative changes that occur during the subsequent time is possibly different. If such a cumulative change occurs commonly, it may be applied to the identification of active faults. Based on this idea, the authors have focused on the chemical composition of the faults. Tateishi et al. (2019, 2020) performed linear discriminant analysis using the chemical composition of fault gouge of active and inactive faults in Japan, and showed that the two can be discriminated with high probability and that the difference between the two can be expressed by some elements. In this study, we performed the same method using the data of granite, and discussed the relationship between the results of the analysis and the activity of the fault. The data of active faults are limited to those of strike-slip faults.
Methods: First, chemical composition of fault gouge in Japan is corrected from literatures. Chemical analyses of fault gouge samples owned by JAEA were also carried out, and the data were integrated. Two categories of fault activity were given to these data. Then, the elements were selected so that the number of samples would be as large as possible, and the data were subjected to log-ratio transformation (Aitchison, 1986; Ohta and Arai, 2006). In order to extract elements suitable for discrimination, variables were selected using the Akaike Information Criterion (AIC: Akaike, 1973). Finally, a linear discriminant analysis was conducted using the combinations of elements selected by AIC as explanatory variables.
Results: The chemical composition data of fault rocks were extracted from eight published papers and seven JAEA reports by literature collection. The chemical composition data of the fault gouge was extracted from 327 samples, including the chemical analysis results of 51 samples of fault rocks. The number of fault gouge data was 96 samples in total, 45 samples from active faults and 51 samples from inactive faults. Then, as a result of the AIC, 11 elements were selected as candidates for explanatory variables: TiO2, Al2O3, MnO, CaO, Na2O, K2O, P2O5, Ba, Rb, Sr, and Th. Then, a linear discriminant analysis was conducted using a combination of (1) the 11 elements selected by AIC, (2) the four elements (TiO2, Ba, Rb, and Sr) whose p-values were less than 0.01 by AIC, and (3) the three elements (TiO2, Rb, and Sr) whose p-values were less than 0.001 by AIC. As a result, the discrimination rate between active and inactive faults was 96% for (1), 86% for (2), and 84% for (3).
Discussion on the relationship between chemical composition and fault activity: The discrimination scores and elemental concentrations of the elements with (1) are compared by the latest event stage, where I to IV are active faults and V is inactive faults. The results showed that the discrimination scores tended to increase from Class I to Class V. As for the elemental concentrations, TiO2 and P2O5 tended to decrease from Class I to Class V, while Th and Y tended to increase from Class I to Class V. The transitions of these elements show a breakpoint between Class I-IV and Class V, where a large gap is inferred in the time elapsed after latest event, and the change in V is very small in all cases. These results suggest that TiO2and P2O5 are concentrated during faulting and dissolved during the subsequent period, while Th and Y are dissolved during faulting and concentrated during the quiescence period. Among these elements, Ti is generally considered to be an immobile element. In contrast, Pe-Piper et al. (2011), after discussing the crystallization process of titanium minerals from chalk sandstone, argue that Ti is not an immobile element. If this is correct, even though Ti is hard to move, but may have moved significantly during violent events such as faulting, and then slowly dissolved during the rest period.
This study was carried out under a contract with METI (Ministry of Economy, Trade and Industry) as part of its R&D supporting program for developing geological disposal technology.
Methods: First, chemical composition of fault gouge in Japan is corrected from literatures. Chemical analyses of fault gouge samples owned by JAEA were also carried out, and the data were integrated. Two categories of fault activity were given to these data. Then, the elements were selected so that the number of samples would be as large as possible, and the data were subjected to log-ratio transformation (Aitchison, 1986; Ohta and Arai, 2006). In order to extract elements suitable for discrimination, variables were selected using the Akaike Information Criterion (AIC: Akaike, 1973). Finally, a linear discriminant analysis was conducted using the combinations of elements selected by AIC as explanatory variables.
Results: The chemical composition data of fault rocks were extracted from eight published papers and seven JAEA reports by literature collection. The chemical composition data of the fault gouge was extracted from 327 samples, including the chemical analysis results of 51 samples of fault rocks. The number of fault gouge data was 96 samples in total, 45 samples from active faults and 51 samples from inactive faults. Then, as a result of the AIC, 11 elements were selected as candidates for explanatory variables: TiO2, Al2O3, MnO, CaO, Na2O, K2O, P2O5, Ba, Rb, Sr, and Th. Then, a linear discriminant analysis was conducted using a combination of (1) the 11 elements selected by AIC, (2) the four elements (TiO2, Ba, Rb, and Sr) whose p-values were less than 0.01 by AIC, and (3) the three elements (TiO2, Rb, and Sr) whose p-values were less than 0.001 by AIC. As a result, the discrimination rate between active and inactive faults was 96% for (1), 86% for (2), and 84% for (3).
Discussion on the relationship between chemical composition and fault activity: The discrimination scores and elemental concentrations of the elements with (1) are compared by the latest event stage, where I to IV are active faults and V is inactive faults. The results showed that the discrimination scores tended to increase from Class I to Class V. As for the elemental concentrations, TiO2 and P2O5 tended to decrease from Class I to Class V, while Th and Y tended to increase from Class I to Class V. The transitions of these elements show a breakpoint between Class I-IV and Class V, where a large gap is inferred in the time elapsed after latest event, and the change in V is very small in all cases. These results suggest that TiO2and P2O5 are concentrated during faulting and dissolved during the subsequent period, while Th and Y are dissolved during faulting and concentrated during the quiescence period. Among these elements, Ti is generally considered to be an immobile element. In contrast, Pe-Piper et al. (2011), after discussing the crystallization process of titanium minerals from chalk sandstone, argue that Ti is not an immobile element. If this is correct, even though Ti is hard to move, but may have moved significantly during violent events such as faulting, and then slowly dissolved during the rest period.
This study was carried out under a contract with METI (Ministry of Economy, Trade and Industry) as part of its R&D supporting program for developing geological disposal technology.