5:15 PM - 6:30 PM
[SVC29-P01] Groundwater composition around the Aso caldera: multivariate statistical analysis and its implications for independent variables and geographical variations
Keywords:groundwater, composition, volcano, deep-seated fluid, statistical analysis
The compositional variability of groundwater (including hot/mineral spring waters) reflects the origin and circulation processes of groundwater, such as fluid sources, fluid-rock reactions, flow rates and residence times; e.g., identification of fluid sources by hydrogen-oxygen isotope ratio (Matsubaya et al., 1973), reaction with bedrock by Sr isotope ratio (Notsu et al., 1991), and evaluation of residence time of deep-seated fluid based on helium isotope ratio (Morikawa et al., 2005). By integrating these analyses, it is possible to forward our understanding of the origins of fluid and the circulation processes more comprehensively. Kusuda et al. (2014) integrated the geochemical data of the Arima spring waters with the dehydration simulation of the subducting plate to infer that the Arima-type brine is originated from the subducted Philippine Sea plate. Nakamura et al. (2016) and Iwamori et al. (2020) discussed the origin of Arima-type brine in the Arima-Kii Peninsula to be rooted from the subducting plate and geochemical processes during ascent to the surface, based on multivariate statistical analysis of the spring water compositions, including rare earth elements.
In this study, the existing composition data set of groundwater (Takahashi et al., 2018) is used to identify the geochemical sources and processes recorded in the groundwater composition over the area of about 120 km east-west and 80 km north-south surrounding the Aso caldera in central Kyushu. In particular, we aim to distinguish the volcanic and deep-seated fluid components from those related to meteoric water circulation, and ultimately to understand (1) fluid processes associated with magmatic activity under the Aso caldera, and (2) active tectonics and the large-scale fluid circulation from the subducting plate to the surface, including magmatism in central Kyushu.
As a pilot study for this purpose, in this presentation, we examine whether information on independent sources and processes can be extracted from the high-dimensional composition data set (590 samples containing 12 major dissolved elements of Na, K, Mg, Ca, Li, Cl, SO4, HCO3, F, NO3, Br, and total C), based on multivariate statistical analyses. The statistical methods used were "whitened-data based k-means cluster analysis (KCA)" and "independent component analysis (ICA)" (Iwamori et al., 2017). Principal component analysis (PCA), a standard multivariate analysis, can extract information for independent sources and processes only if the data show a (joint) normal distribution. For data showing a non-normal distribution, the above KCA and ICA should be used. The data set used in this study shows clear non-normality, for which KCA and ICA are suitable. Based on the eigenvalue analysis of the data, we set the number of clusters and independent components to eight and attempted to detect statistically independent features in their compositional and geographical spaces.
As a result, a concentric spatial structure (distinguished with the 8 clusters) was found both in the Aso caldera and in the wide area surrounding the Aso volcano, with a broad decrease in the concentrations of Na, Li, Cl, HCO3, Br, total C toward the outside. Even when focusing only on the inside of caldera, four clusters were mainly distributed concentrically, and two of them were distributed so as to surround basalt and felsic rock bodies and lava. Except for the Aso caldera, the two clusters occur characteristically in the Kuju volcano centered around Mt. Kuju, which suggests that the two clusters are of magmatic origins.
The distribution and compositional characteristics of these clusters suggest that KCA and ICA are useful for the detection and identification of independent sources and processes involved in the origin of groundwater, especially the magmatic components. In the future, we plan to characterize the individual clusters with more elements and isotopic compositions, and to apply this statistical approach to other volcanic areas. This research was partly supported by JSPS KAKENHI Grant Numbers JP18H03747, JPJSBP120204804, JP17KK0018, JPJSBP120194830, and the Nuclear Regulation Authority "Research on Knowledge Development of Giant Eruption Processes, FY 2020".
In this study, the existing composition data set of groundwater (Takahashi et al., 2018) is used to identify the geochemical sources and processes recorded in the groundwater composition over the area of about 120 km east-west and 80 km north-south surrounding the Aso caldera in central Kyushu. In particular, we aim to distinguish the volcanic and deep-seated fluid components from those related to meteoric water circulation, and ultimately to understand (1) fluid processes associated with magmatic activity under the Aso caldera, and (2) active tectonics and the large-scale fluid circulation from the subducting plate to the surface, including magmatism in central Kyushu.
As a pilot study for this purpose, in this presentation, we examine whether information on independent sources and processes can be extracted from the high-dimensional composition data set (590 samples containing 12 major dissolved elements of Na, K, Mg, Ca, Li, Cl, SO4, HCO3, F, NO3, Br, and total C), based on multivariate statistical analyses. The statistical methods used were "whitened-data based k-means cluster analysis (KCA)" and "independent component analysis (ICA)" (Iwamori et al., 2017). Principal component analysis (PCA), a standard multivariate analysis, can extract information for independent sources and processes only if the data show a (joint) normal distribution. For data showing a non-normal distribution, the above KCA and ICA should be used. The data set used in this study shows clear non-normality, for which KCA and ICA are suitable. Based on the eigenvalue analysis of the data, we set the number of clusters and independent components to eight and attempted to detect statistically independent features in their compositional and geographical spaces.
As a result, a concentric spatial structure (distinguished with the 8 clusters) was found both in the Aso caldera and in the wide area surrounding the Aso volcano, with a broad decrease in the concentrations of Na, Li, Cl, HCO3, Br, total C toward the outside. Even when focusing only on the inside of caldera, four clusters were mainly distributed concentrically, and two of them were distributed so as to surround basalt and felsic rock bodies and lava. Except for the Aso caldera, the two clusters occur characteristically in the Kuju volcano centered around Mt. Kuju, which suggests that the two clusters are of magmatic origins.
The distribution and compositional characteristics of these clusters suggest that KCA and ICA are useful for the detection and identification of independent sources and processes involved in the origin of groundwater, especially the magmatic components. In the future, we plan to characterize the individual clusters with more elements and isotopic compositions, and to apply this statistical approach to other volcanic areas. This research was partly supported by JSPS KAKENHI Grant Numbers JP18H03747, JPJSBP120204804, JP17KK0018, JPJSBP120194830, and the Nuclear Regulation Authority "Research on Knowledge Development of Giant Eruption Processes, FY 2020".