Geofluid behavior prior to the 2018 Hokkaido Eastern Iburi earthquake: insights from groundwater geochemistry

Abstract

A notable decrease of groundwater Na/K ratios was observed several months before the 2018 Hokkaido Eastern Iburi earthquake (M6.7) at a site approximately 20 km west of the earthquake’s epicenter. To investigate the cause of the decreased Na/K ratios, we analyzed groundwater samples (commercially bottled drinking water) to assess the contribution of deep-derived fluids using Li and Sr isotopic ratios, which are reliable indicators of deep fluid contributions. No pronounced changes in the ⁷Li/⁶Li and ⁸⁷Sr/⁸⁶Sr ratios were observed; thus, the pre-earthquake decrease of groundwater Na/K ratios did not result from the input of deep-derived fluids. The pre-earthquake decrease in the Na/K ratio observed in this study may instead be related to CO₂ influx into the aquifer. The CO₂ may have leaked from carbon dioxide capture and storage (CCS), because CCS was in operation near the epicenter of the 2018 Hokkaido Eastern Iburi earthquake. Decreases of the Na/K ratios and increases of the CO₂ concentrations in groundwater have been reported before other large earthquakes; thus, CO₂ influx into groundwater may be a common phenomenon preceding earthquakes.

1 Introduction

Geofluids may play an important role in the occurrence of large earthquakes, but the mechanism for this is poorly understood (e.g., Zhao et al. 1996). The term “geofluids,” as elucidated by Czauner et al. (2022), refers to groundwater, geothermal and hydrothermal fluids, and related geo-resources. Geophysical methods using information such as the seismic wave velocity structure and electrical conductivity are extremely useful for understanding the distribution of geofluids over a wide area (e.g., Zhao et al. 1996) and are the primary methods used in geofluid studies. However, small volumes of geofluids cannot be detected by geophysical methods, and thus geochemical methods are powerful tools for obtaining information the origins and behaviors of such geofluids (Nishio 2013). Accordingly, it may be possible to detect the mixing of fluids of different origins before a major earthquake (Liu et al. 2023). Geophysical and geochemical methods are thus complimentary, and their combined application can improve our understanding of the behavior of geofluids prior to major earthquakes.

A variety of elements and elemental ratios are considered when studying geofluids: Two of the useful ones are lithium and strontium. The two stable isotopes of Li are ⁶Li and ⁷Li, and their relative abundances are about 7.5% and 92.5%, respectively. The ⁷Li/⁶Li ratio between the fluid and solid phases changes with reaction temperature (Marschall et al. 2007; Wunder et al. 2006); consequently, the ⁷Li/⁶Li ratio is a useful tool for constraining geofluid reaction temperatures. Additionally, according to You et al. (1996) and James et al. (2003), the concentration of lithium in the fluid reaches its maximum at high temperatures. Thus, geofluid Li concentration that has experienced high temperature is significantly higher than that in surface water that has not experienced high temperature. By the time that a geofluid is emitted at the surface, surface water contamination will have been unavoidable. Li, for which the concentration difference between surface and non-surface waters is extremely large relative to other elements, is less susceptible to surface water contamination. In a case study of the Arima and Kii Peninsula region in southwest Japan, the reported Li concentration in surface water from the Kushida River within the Kii Peninsula area was 0.00034 mg/L (Umam et al. 2022); in contrast, non-surface water sources in the Arima area (Arima hot spring) contained a notably higher lithium concentration of around 40 mg/L (Umam et al. 2022). In addition, the Cl/Li ratio can be used to quantify the extent to which the observed δ⁷Li values are influenced by surface water (Nishio et al. 2010). Data on ⁸⁷Sr/⁸⁶Sr ratios are also useful in elucidating the origins of geofluids. The ⁸⁷Sr/⁸⁶Sr ratio in groundwater provides information about the rocks with which a geofluid has interacted at high temperatures.

Using commercially available bottled drinking water, Tsunogai and Wakita (1995) reported an increase in chloride ion concentrations in groundwater before the 1995 M7.2 Kobe earthquake. Using a similar approach, Sano and their group focused on the isotopic compositions of the elements constituting volatile elements (i.e., hydrogen, oxygen, carbon, etc.) in groundwater prior to several major earthquakes (Onda et al. 2018; Sano et al. 2020a, 2020b). Sano et al. (2020a) attributed a decrease in δ¹³C values and an increase in total dissolved inorganic carbon (TDIC) observed a few months before the 2018 Hokkaido Eastern Iburi earthquake (M6.7) to CO₂ influx into the aquifer. To understand the nature of the CO₂ influx into the aquifer prior to that earthquake, we investigated whether deep-derived fluids are involved or not in this influx. In this study, we analyzed commercially available bottled drinking water samples that preserve pre-earthquake groundwater compositions to assess temporal changes in the concentrations of non-volatile elements and the Li and Sr isotopic compositions of the groundwater.

2 Methods

The water samples used in this research were obtained from commercially available bottled mineral water that was collected between 2016 and 2019 at two locations: Uenae, 23 km west of the epicenter of the M6.7 Hokkaido Eastern Iburi earthquake; and Eniwa, 40 km northwest of the epicenter (Fig. 1). The hydrogen and oxygen isotopic ratios in the water were previously published by Sano et al. (2020a). At both sites, water is extracted from wells with a depth of approximately 100 m, located in the volcanic–sedimentary aquifer of the Shikotsu pyroclastic flow of Late Pleistocene age (Nakagawa et al. 2018). The Uenae site is closer to the Ishikari–Teichi–Toen active fault zone (ITTFZ) than is the Eniwa site.

Fig. 1

Locations of the Uenae and Eniwa sampling sites. The yellow stars indicate the epicenter locations of the 2018 M6.7 Hokkaido Eastern Iburi earthquake, the M5.1 earthquake on 1 July 2017, and the M5.8 aftershock on 21 February 2019. The red lines on the map represent active faults of the Ishikari–Teichi–Toen fault zone. The inset map shows the location of the study area within Japan

The water from the borehole was placed into plastic bottles after filtration. In this study, bottled water samples were processed using the protocol described by Nishio et al. (2015). After the water samples had been passed through a 0.2 μm PTFE syringe filter, all analytical procedures were performed under a filtered airflow (cleanliness level better than class 1000) and using 18.2 MΩ-grade water prepared by an ultrapure water system (RFU663EA, Advantec). The clean laboratory and all analytical equipment used in this study were at Kochi Core Center, a joint research facility managed by Kochi University and JAMSTEC (Japan Agency for Marine-Earth Science and Technology).

The concentrations of major elements such as K, Ca, Mg, Na, and Cl were determined with an ion chromatograph (ICS-2000, Thermo Fisher Sci.), and those of trace elements (Li, Rb, Cs and Sr) were determined with an inductively coupled plasma mass spectrometer (ICPMS; iCAP Qc, Thermo Fisher Sci.) using diluted sample solutions containing an internal standard of indium (with uncertainty better than ± 3%, as estimated from the reproducibility (2RSD) of standard solutions with salt contents higher than those of the analyzed samples). To determine isotopic ratios, Li and Sr were separated from the sample solutions using a column filled with AG50W cation exchange resin (Bio-Rad Lab., USA). Then, Li and Sr were further purified using columns filled with AG50W cation exchange resin and Sr resin (Eichrom Tech., USA), respectively. After the two-step column separation, Li isotopic ratios were measured with a multi-collector ICPMS (MC-ICPMS; Neptune, Thermo Fisher Sci.) and Sr isotopic ratios with a thermal ionization mass spectrometer (TIMS; TRITON, Thermo Fisher Sci.). A Li standard NIST L-SVEC solution was measured before and after the sample analyses to correct instrumental mass bias. The measured ⁷Li/⁶Li ratios are expressed as permil deviations from the NIST L-SVEC standard, based on the formula δ⁷Li = [[⁷Li/⁶Li]_sample/[⁷Li/⁶Li]_{L-SVEC standard} − 1] × 1000. Our δ⁷Li value of the Institute for Reference Materials and Measurements BCR-403 seawater standard was + 31.3‰, in good agreement with the previously reported value of + 31.0‰ (Millot et al. (2004) and within our uncertainty of less than ± 0.3‰ estimated from the long-term reproducibility. The ⁸⁷Sr/⁸⁶Sr ratios of the samples were acquired using a TRITON TIMS by loading purified Sr fractions onto single tungsten filaments together with a tantalum activator solution. The uncertainty on ⁸⁷Sr/⁸⁶Sr measurements was better than ± 0.000007, as estimated from the long-term reproducibility. The average ⁸⁷Sr/⁸⁶Sr ratio that we obtained for the JB2 rock standards was 0.703679 ± 0.000007 (2 SD, n = 4), as previously reported by Nishio et al. (2010).

3 Results

The geochemical and isotopic data for groundwater collected at the Uenae and Eniwa sites (Fig. 1) are provided in Table 1. The samples analyzed in this study are aliquots of samples previously studied by Sano et al. (2020a), who reported data for elements constituting volatile elements in the samples, such as the δ¹³C values of CO₂. The samples from the Uenae and Eniwa sites analyzed in this study were bottled between June 2015 and May 2019 and between June 2016 and October 2018, respectively. These collections include five and seven samples obtained from the Eniwa and Uenae sites, respectively, before September 6, 2018, the date of the M6.7 Hokkaido Eastern Iburi earthquake. Additionally, our samples include one post-earthquake sample from the Eniwa site and five from the Uenae site.

Table 1 Chemical compositions, elemental ratios, δ⁷Li values, and ⁸⁷Sr/⁸⁶Sr ratios of water samples from the Uenae and Eniwa sites

3.1 Time-series changes in groundwater geochemistry

The most notable result of this study is that the Na/K ratio in the groundwater at the Uenae site decreased markedly before the M6.7 Hokkaido Eastern Iburi earthquake (Fig. 2a). The Na/K ratio at the Uenae site began to decrease around April 2018, about 5 months before the earthquake, reached a minimum value in September 2018, when the earthquake occurred, then increased until around December 2018, approximately 3 months after the earthquake. In contrast, Na/K ratios in groundwater at the Eniwa site were consistent before and after the earthquake (Fig. 2b). With the exception of the Na/K ratios at the Uenae site, no preseismic signals were detected in the analyses in this study, including in the ⁷Li/⁶Li and ⁸⁷Sr/⁸⁶Sr ratios (Figs. 2–4).

Fig. 2

Na/K ratio in groundwater at a the Uenae site and b the Eniwa site before and after the M6.7 Hokkaido Eastern Iburi earthquake of September 2018

The concentrations of Li, Rb, Cs, Sr, Na, K, Mg, Ca, and Cl, and the Cl/Li ratios, increased at the Uenae site about one month after the earthquake (Fig. 3a–j). In contrast, no obvious post-earthquake changes were observed in groundwater at the Eniwa site (Fig. 3k–t).

Fig. 3

Temporal variations in the concentrations of Li, Rb, Cs, Sr, Na, K, Mg, Ca, and Cl and the Cl/Li ratio in groundwater for the Uenae and Eniwa sites before and after the M6.7 Hokkaido Eastern Iburi earthquake of September 2018

3.2 Comparison of groundwater geochemistry at the Uenae and Eniwa sites

The Li concentrations at the Eniwa site (6.9–7.1 µg/L; Table 1) were markedly higher than those at the Uenae site (1.8–2.4 µg/L). At both sites, groundwater Li concentrations were notably higher than the average Li concentration in river water (0.57 µg/L; Nishio et al. (2010)). The concentrations of Na, K, Mg, Ca, Cl, Rb, Cs, and Sr in groundwater were higher at the Uenae site than at the Eniwa site.

Groundwater Cl/Li ratios at the Eniwa site, which ranged from 979 to 1,130, were markedly lower than surface water values (> 20,000; Nishio et al. (2010)), whereas groundwater Cl/Li ratios at the Uenae site (15,900–29,600) were notably higher than at the Eniwa site and comparable to surface water values. The Cl/Sr ratios for the Uenae site (314–368) were notably higher than those of the Eniwa site (145–163).

The δ⁷Li values of groundwater at the Eniwa site (+ 4.7‰ to + 4.9‰) were markedly lower than the δ⁷Li values of groundwater at the Uenae site (+ 12.0‰ to + 12.2‰; Fig. 4a, b). The δ⁷Li values of river waters worldwide range from + 6.0‰ to + 37.5‰ (Huh et al. 1998). At the Uenae site, the δ⁷Li values of groundwater were comparable to those of river water, whereas at the Eniwa site, they were notably lower.

Fig. 4

Variations in δ⁷Li values and ⁸⁷Sr/⁸⁶Sr ratios before and after the M6.7 Hokkaido Eastern Iburi earthquake of September 2018 for a, c the Uenae site and b, d the Eniwa site

The ⁸⁷Sr/⁸⁶Sr ratios in the Uenae and Eniwa groundwater during the observation period were 0.70474–0.70477 and 0.70551–0.70554, respectively (Table 1). The basement rocks at both sites are mainly Quaternary volcanic rocks. The ⁸⁷Sr/⁸⁶Sr ratios of groundwater collected at the study sites are similar to the ⁸⁷Sr/⁸⁶Sr ratio of 0.70485 for Quaternary volcanic rocks reported by Notsu et al. (1991). Marked temporal changes in the ⁸⁷Sr/⁸⁶Sr ratios of samples were not observed during the study period at either site (Fig. 4c, d).

4 Discussion

4.1 Origin of groundwater collected from the Uenae and Eniwa sites

The Cl/Li ratio is a useful indicator for quantitatively understanding the extent to which δ⁷Li values have shifted as a result of surface-fluid contamination (e.g., Nishio et al. 2015, 2010). Surface fluids have never experienced high temperatures, whereas non-surface fluids such as volcanic and deep-derived fluids may have experienced such temperatures. The observed δ⁷Li values of groundwater are the result of binary mixing between surface and non-surface fluids. Previous analyses of water samples from southwest Japan and Ontake volcano spring waters yielded Cl/Li ratios of volcanic fluids and/or deep-derived fluids of less than 1100 (Nishio et al. (2010) and 1000 (Kazahaya et al. 2014), whereas measured Cl/Li ratios of surface water were markedly higher, e.g., 20,000 for a water sample collected from the Shirakawa River in the Ontake area (Nishio et al. (2010)).

Therefore, water samples with Cl/Li values between 1100 and 20,000 are considered to be mixtures of surface water and deep-derived fluids. For the Uenae groundwater samples, the Cl/Li ratios varied widely from 15,900 to 29,600 but the δ⁷Li values were uniform, ranging from + 12.0 to + 12.2‰ (Fig. 5; Table 1). As shown in the δ⁷Li–Cl/Li diagram (Fig. 5), the Uenae site samples have almost constant δ⁷Li values while the Cl/Li ratio varies significantly. If this is the result of binary components mixing of surface water (high δ⁷Li values and Cl/Li ratios) and deep-derived fluids (low δ⁷Li values and Cl/Li ratios), the δ⁷Li values should increase with Cl/Li ratio. In the Uenae region, the δ⁷Li values of both the surface water component (high Cl/Li ratio) and the non-surface water component (low Cl/Li ratio) are around + 12‰, indicating that the groundwater samples at the Uenae site had been affected by mixing with surface water during the sample collection period. However, the samples from the Eniwa site, with an average Cl/Li ratio of ~ 1000 and δ⁷Li values of ca. + 4.7‰, did not vary during the sample collection period.

Fig. 5

Cross-plot of Cl/Li weight ratio and δ⁷Li for the Eniwa site (orange circles) and the Uenae site (blue circles)

In most minerals, Li is found in sixfold coordination, whereas in fluids it occurs in fourfold coordination. Thus, because of the preference of the lighter isotope for the higher coordination site, the δ⁷Li value of the fluid phase is higher than that of the solid phase (Oi et al. 1989; Wunder et al. 2006). This Li isotopic fractionation between liquid and solid phases (Δ⁷Li_fluid-solid) is temperature-dependent, decreasing with increasing temperature (James et al. 2003). This relationship can be expressed as (Wunder et al. 2006):

$${\Delta}^{7} {\text{Li}}_{{\text{fluid - solid}}} = 4.61 \times \left( {1000/T} \right) - 2.48$$

(1)

where T represents the reaction temperature associated with the water–rock interaction. The δ⁷Li values of the volcanic rocks in the study area are unknown; however, fresh volcanic rocks erupted in nearby areas of the subduction zone are expected to have the same δ⁷Li value, so we calculated the reaction temperature by using the average value for volcanic rocks in Northeast Japan (+ 2.1‰ to + 4.4‰) reported by Moriguti et al. (2004), which are subducted with the Pacific Plate as were the volcanics in our study. From this calculation, we infer that Li in the non-surface water component of the Uenae groundwater experienced temperatures above 120 °C.

From data on spring water from sites around Ontake volcano, Nishio et al. (2010) pointed out that the δ⁷Li values of samples with Cl/Li < 1100 were not affected by surface water or seawater mixing. On this basis, the δ⁷Li values (+ 4.7 to + 4.9‰) of groundwater at the Eniwa site (Cl/Li = 980–1100) can be regarded as unaffected by surface water components. From Eq. (1), it can be inferred that the Li in the groundwater at Eniwa experienced a temperature of 780 °C. Thus, the Li in the non-surface water component of the groundwater at the Eniwa site experienced higher reaction temperatures than the groundwater at the Uenae site.

As mentioned in Sect. 3, no marked temporal changes were observed in the ⁸⁷Sr/⁸⁶Sr ratios of the samples collected and analyzed during the study period at either site (Fig. 4c, d). This is consistent with the fact that the ⁸⁷Sr/⁸⁶Sr ratios of waters in Quaternary volcanic regions tend to correlate with those of the surrounding volcanic rocks (Notsu et al. 1991).

4.2 Pre-earthquake decrease in Na/K ratio at the Uenae site

4.2.1 Mixing model

The water samples from the Uenae and Eniwa sites can be regarded as mixtures of a deep-derived fluid component and surface water, as discussed in Sect. 4.1. The Na/K ratios and δ⁷Li values of possible fluid components in the region are illustrated in Fig. 6. One of the components is surface water in which the meteoric water has not experienced high temperature. The Na/K ratios and δ⁷Li values for the surface water component are 5 and + 23‰, respectively. The Na/K ratio of the surface water is the average value of river waters of western Hokkaido with low total dissolved solids as reported by Shigino (2011) (Appendix 1). The δ⁷Li value for surface water is the global average value reported by Huh et al. (1998). Due to the the presence of Tarumae volcano in the area, volcanic fluid resulting from the reaction of hot magma and meteoric water is also plotted in Fig. 6. As there are no available data on hot spring waters around Tarumae, the δ⁷Li values of volcanic fluids obtained from hot spring waters around Ontake volcano (Nishio et al. (2010) are shown. Because the average δ⁷Li value of Ontake volcanic rocks (+ 2.7‰; Nishio et al. (2010)) is very close to the average δ⁷Li value of volcanic rocks in Northeast Japan (+ 3.0‰; Moriguti et al. 2004), in this study, the δ⁷Li value of the volcanic fluids from Ontake volcano can be regarded as representative of those of volcanic fluids from Tarumae volcano. The Na/K ratio and δ⁷Li values for the volcanic fluid component are 7 and + 3‰, respectively. As Na/K ratios and δ⁷Li values for volcanic fluids, we used the average values of hot spring waters from two sites within 5 km of the crater of Ontake volcano reported by Nishio et al. (2010).

Fig. 6

Mixing lines between surface water and deep-derived fluids. Surface water data are for river waters in western Hokkaido (Shigino 2011) and global surface waters (Huh et al. 1998); data for deep-derived fluids are for Arima deep fluid (Kusuda et al. 2014; Umam et al. 2022) and Ontake volcanic fluid (Nishio et al. 2010). Also shown are data for groundwater from the Eniwa site (orange circles) and the Uenae site (blue circles). The Uenae groundwater samples exhibit movement along the vertical axis (red dashed arrow) and not along the gray mixing lines; thus, the mixing of upwelling deep-derived fluids with surface water cannot explain the observed pre-earthquake decrease in the Na/K ratio. The red dashed arrow in the figure represents the movement of water samples along this direction, indicating the potential reaction with CO₂

In this study, we used data from Arima hot spring water to represent deep-derived fluids in the forearc region. The hydrogen and oxygen isotopic ratios of hot spring waters occurring in several forearc regions, including Arima, are different from those of meteoric waters, suggesting that Arima-type deep fluids may be derived from subducted slabs (Matsubaya et al. 1973). The Na/K ratio of 8 and δ⁷Li value of + 1‰ used for the deep-sourced fluid are the average values of previously reported data for Arima hot spring waters (Kusuda et al. 2014; Umam et al. 2022). For groundwater relatively enriched in surface water components, such as at the Uenae site, the δ⁷Li value would be expected to decrease with a constant Na/K ratio when the contribution of fluids that have experienced high temperatures, such as volcanic fluids and deep source fluids, increases (Fig. 6). Therefore, the decreasing Na/K ratio with constant δ⁷Li values observed in the groundwater at the Uenae site before the Hokkaido earthquake did not result from a change in the contribution of fluids that had experienced high temperatures, such as volcanic fluids or deep fluids.

4.2.2 CO₂ influx model

A decrease in δ¹³C (CO₂) and an increase in total dissolved inorganic carbon (TDIC) were reported in groundwater at the Uenae site before the Hokkaido Eastern Iburi earthquake in 2018 (Sano et al. 2020a). In addition to changes in carbon isotopic ratios, Sano et al. (2020a) reported variations in ¹⁴C activity, δ¹⁸O, and δD values prior to the earthquake. The beginning of the pre-earthquake decrease in the Na/K ratio at the Uenae site documented in this study coincides well with the beginnings of the changes in δ¹³C, ¹⁴C, and TDIC values reported by Sano et al. (2020a) (Fig. 7). Based on the δ¹³C, ¹⁴C, and TDIC results in groundwater, Sano et al. (2020a) suggested that CO₂ injected into the Tomakomai subsurface for carbon dioxide capture and storage (CCS) may have entered the aquifer at the Uenae site a few months before the 2018 earthquake.

Fig. 7

Na/K ratios (blue circles) compared with a δ¹³C, b ¹⁴C activity, and c total dissolved inorganic carbon (TDIC) (Sano et al. 2020a) at the Uenae site before and after the M6.7 Hokkaido Eastern Iburi earthquake of September 2018. The light blue and green arrows from February 2018 to the time of the earthquake represent changes in the Na/K ratio and TDIC of Uenae groundwater samples, respectively; notably, these changes are negatively correlated during this period

Using a bottled water method similar to that applied herein, Onda et al. (2018) proposed that significant changes in groundwater oxygen isotopic ratios observed prior to the 2016 M6.6 Tottori earthquake were directly related to changes in volumetric strain. However, similar to our results, the changes in oxygen isotopic ratios prior to the 2016 M6.6 Tottori earthquake reported by Onda et al. (2018) may be explained by oxygen isotopic exchange with carbon dioxide entering the groundwater aquifer.

The concentration of each element in solution would be expected to change as the pH decreases as a result of the influx of CO₂ into the aquifer. The contrasting behavior of potassium (K) compared to other elements, possibly related to the precipitation of K-bearing minerals, which would be affected by water–rock interaction to a greater degree than other minerals in the groundwater source region, has been previously demonstrated by Skelton et al. (2014) in groundwater in northern Iceland before an earthquake there. However, the mechanism by which only the Na/K ratio decreased markedly, while the other element ratios did not, remains unresolved.

The δ⁷Li data suggest that the pre-earthquake decrease in the Na/K ratio at the Uenae site was not caused by the contribution of volcanic fluids and/or deep-derived fluids, as discussed in Sect. 4.2.1. The coincidence with the decrease in δ¹³C values indicates that it is very likely that the decrease in the Na/K ratio was also due to the CO₂ injection into the aquifer proposed by Sano et al. (2020a).

4.2.3 Low Na/K ratio in groundwater before a large earthquake

A decrease in the Na/K ratio associated with an increase in the dissolved CO₂ concentration of groundwater before a major earthquake has been reported previously at Slanic Moldova, Romania, before a M6.0 earthquake in 2004 (Fig. 8; Mitrofan et al. 2008). The first assessment of geological storage capacity in Romania was conducted between 2006 and 2008 by EU GeoCapacity (2009). Therefore, since no CCS had been carried out in Romania prior to 2004, the CO₂ influx into the aquifer responsible for the negative correlation between dissolved CO₂ concentration and Na/K ratio in groundwater observed before that earthquake was of natural origin.

Fig. 8

Dissolved CO₂ concentrations (orange circles) and Na/K ratios (blue circles) before the M6.0 Romania earthquake of 2004 (data from Mitrofan et al. (2008))

In contrast, no change in the Na/K ratio was observed in groundwater at the Eniwa site (Fig. 2b), despite the observed decrease in δ¹³C (CO₂) before the M6.7 Hokkaido Eastern Iburi earthquake reported by Sano et al. (2020a). A possible explanation for the absence of a Na/K change at the Eniwa site before the earthquake is a time lag, resulting from the different distances of the Eniwa and Uenae sites from the CO₂ source that produced the decrease in the Na/K ratio.

The Uenae site, where the pre-earthquake Na/K change was observed, is located near the ITTFZ (Fig. 1). Because the fault zone provides a pathway for fluids rising from depth, it is possible that the groundwater near the surface at Uenae may have been affected quickly by subsurface events such as the effects of CO₂ inflow.

4.3 Post-earthquake changes in chemical compositions and elemental ratios

Changes in chemical composition after an earthquake may occur in response to fracturing of a hydrological barrier between aquifers or as a result of a change in the relative pressures of connected aquifers (Thomas 1988). Because the Uenae site is close to the ITTFZ fault zone and the earthquake epicenter, observed changes in the chemical composition and elemental ratios of groundwater caused by the earthquake can be explained as a result of changes in the stress–strain state of the fracture zone. Such changes can alter aquifer properties and/or cause local hydrostatic releases, mixing of water from different aquifers, or switching between different groundwater sources along the faults (Rosen et al. 2018).

In the Unae site samples, an increase was observed in the Cl/Li ratios (Fig. 3j) and in the concentrations of Li, Rb, Cs, Sr, Na, K, Mg, Ca, and Cl one month after the earthquake (Fig. 3a–i). The observed increase in the Cl/Li ratios in these samples could be related to post-earthquake mixing of Uenae groundwater with another water source, such as seawater with a higher Cl concentration. The increases in the concentrations of Li, Rb, Cs, Sr, Na, K, Mg, Ca, and Cl can be explained by mixing with another water source at the Uenae site.

The presence of water or fluid after the M6.7 earthquake has also been proposed by Hua et al. (2019), who employed P- and S-wave attenuation tomography; a clear low-Q (high attenuation) belt at depths of 10 and 40 km exist under central Hokkaido, including the area of the earthquake epicenter. Groundwater mixing after the earthquake through the ITTFZ fault system near the Uenae site, where the pore pressure was high and permeability along the fault plane was enhanced, has been discussed by Sano et al. (2020a). After the earthquake, the Uenae site groundwater aquifer mixed with another new fluid; however, meaningful changes in hydrochemical compositions and the Cl/Li ratio were not observed in Eniwa site samples before or after the earthquake (Fig. 3k–t). This difference can be attributed to the larger distance of the Eniwa site from the epicenter and the active faults in the ITTFZ, or to the small number of measurements.

5 Conclusions

We documented a marked decrease in Na/K ratio in groundwater in the Uenae area, approximately 20 km west of the epicenter, a few months before the September 2018 M6.7 earthquake in Hokkaido, Japan.

The pre-earthquake decrease in the Na/K ratio at the Uenae site did not result from an input of deep-derived fluid because no variations in the ⁷Li/⁶Li and ⁸⁷Sr/⁸⁶Sr ratios of the groundwater were observed before the earthquake. The timing of the pre-earthquake decrease in the Na/K ratio observed in this study coincided well with the beginnings of the changes in δ¹³C, δ¹⁸O, and δD values reported by Sano et al. (2020a); these changes were related to the injection of CO₂ into the aquifer.

Therefore, the pre-earthquake Na/K decrease observed in the Uenae groundwater may be attributed to injection of CO₂ into the groundwater layer, without the involvement of deep-derived fluids, because the concentration of each element in the solution was altered as a result of the CO₂ influx into the aquifer.

Appendix 1

Na/K ratios of the waters of several rivers in western Hokkaido (Shigino 2011) that were used to calculate the average Na/K ratio of surface waters in the study area.

Name	Longitude (deg. E)	Latitude (deg. N)	Na/K (w/w)
Changliuchan	141.118	42.642	3.6
Tobi River	141.153	42.539	6.8
Shiraoi River	141.186	42.652	5.1
Tobi River	141.266	42.513	4.8
Tobi River	141.266	42.513	5.1
Average			5.1