Multivariate statistical analyses of rare earth element compositions of spring waters from the Arima and Kii areas, Southwest Japan

Abstract

Multivariate statistical analyses have been applied to a compositional data set for spring waters from the Arima and Kii areas in southwest Japan, for understanding sources and processes that produced the deep-seated Arima-type brine and the related spring waters in the areas. The data set consists of 14 rare earth elements (REEs), six major solute species (HCO 3 , Cl, Na, Ca, K, Mg) and isotopic ratios of oxygen, hydrogen and helium (δD, δ 18 O, 3 He/ 4 He) from 19 sites. Independent Component Analysis (ICA) and k-means cluster analysis on the whitened data (KCA) are useful to extract independent features hidden in such a high-dimensional geochemical data set. As a result of ICA and KCA, four ICs and eight clusters were found to sufficiently and effectively describe the observed variability. Comparison of the results with δD, δ 18 O and 3 He/ 4 He provides geochemical interpretations on the extracted four ICs as follows: IC1 represents the overall REE concentration level; IC2 measures the amplitude of Eu anomaly inherited from local basement rocks/minerals; IC3 reflects the amplitude of (reversed) W-shaped pattern, which developed on mineral precipitation event(s) when the deep brine mixes with meteoric water in a relatively shallow aquifer under oxidative conditions, and IC4 measures the amplitude of Ce anomaly possibly by precipitation or mechanical filtering of CeO 2 . Combination of the four ICs may create a complex and various REE pattern to explain the actual data. Although these geochemical features with Eu, Ce and Wshaped patterns have been broadly suggested to be important for understanding origin of the spring waters in the areas, this is the first time to demonstrate that the above features represent the independent processes and to compare them systematically with the regional compilation of the basement rock compositions. In addition, several clusters were found only in either the Arima area or the Kii area, which reflects broad geographical provenance of IC3 and IC4. These spatial variations can be attributed to compositional variations of the original deep brine, which possibly varies with slab depth and temperature to control the solute concentrations and subsequent geochemical evolution to produce different types of spring waters.

Upwelling of Deep-seated Fluid in the Sikhote-Alin Region, Far East of the Eurasian Plate

“…This W-shaped pattern is thought to be caused by the precipitation of Fe oxide (Kawabe 2015). When the redox conditions are changed drastically, such as meteoric water flowin (Iwamori et al 2020), the mid-REEs tend to be co-precipitated with iron hydro-oxide, which results in the depletion of mid-REEs compared to LREEs and HREEs in the residual 1 from Kharitonova et al (2016) fluid phase. In addition, the Eu anomaly from feldspar-like plagioclase abundant in crustal rocks is added to the mid-REE depletion, which leads to a W-shaped pattern in the residual fluid.…”

Stratigraphic Variations of Fe–Mn Micronodules and Implications for the Formation of Extremely REY-Rich Mud in the Western North Pacific Ocean

“…In this study, we employed K-means cluster analysis (KCA) to classify the Fe-Mn micronodules based on their geochemical characteristics. The KCA is a widely used classification method based on which samples are partitioned into a set of K-clusters using multivariate data [38][39][40][41][42]. In this study, we used the Fe, Mn, Co, Ni, Cu, and Mo contents as a set of variables because these elements are abundant in the Fe-Mn (oxyhydr)oxide phase of the micronodules and thus reflect the geochemical features of the micronodules [37].…”

“…Subsequently, they were uncorrelated using a principal component analysis (PCA) algorithm and scaled based on the standard deviations (i.e., square roots of the eigenvalues) along the principal components. This preprocessing, called "whitening," highlights the essential structure of the multivariate dataset [14,39,40]. In addition, a "dimensionality reduction" considering the eigenvalues was performed to exclude noise and stabilize the numerical process.…”

Combined tracers in hot spring waters across the Kii Peninsula, Japan: Implications for the origins of metamorphic fluids of the SW Japan forearc

“…Thus the evolution pattern may be useful for estimating the parent rock types of spring waters. We also investigated patterns of rare earth elements (REEs), which can be used to estimate the mixing components, and identify the types of host rocks and minerals (Nakamura et al, 2015(Nakamura et al, , 2016Iwamori et al, 2020). Previous studies applied statistical processing to group spring water geochemical features, which revealed several mixing components along the MTL.…”

“…According to observations from across the Kii Peninsula, non-meteoric δDδ 18 O signatures are specifically distributed on the western side of the median tectonic line (MTL), despite the wide distribution of groundwater with high 3 He/ 4 He ratios (Umeda et al, 2006;Morikawa et al, 2016). Spatial variations of spring waters may be attributed to compositional variations of the original deep fluid (Iwamori et al, 2020). Moreover, non-meteoric δD-δ 18 O signatures can also be produced by clay mineral dehydration (Dählman and Lange, 2003).…”