Groundwater, possibly originated from subducted sediments, in Joban and Hamadori areas, southern Tohoku, Japan

Geological settings and water samples

The Joban and Hamadori areas are located in the northeastern part of Ibaraki Prefecture and the southeastern part of Fukushima Prefecture, Japan (Figure 1a). The Joban sedimentary basin is located offshore and consists of a nearly continuous sequence of sedimentary rocks of about 5,000-m maximum thickness that has been deposited on the basement rocks since late the Cretaceous (Figure 1b; Inaba et al.2009). The Maastrichtian-Paleogene coals and coaly mudstones are the most likely sources of natural gas (Iwata et al.2002). Inaba et al. (2009) demonstrated that the Cretaceous argillaceous rocks in the basin can potentially generate petroleum if there is abundant marine organic matter deposited in an anoxic environment. The Abukuma belt is located west of the Joban basin and consists of metamorphic rocks and Cretaceous granite (e.g., Hiroi et al.1998). The Hatagawa Fault is considered to be the boundary between the Abukuma and Southern Kitakami Belts (e.g., Tomita et al.2002). However, we treat basements rocks in the two belts and those beneath the Joban basin as ‘pre-Cretaceous basement rocks’ in this study. Coastal areas, including our sampling sites, are covered with thin Tertiary sedimentary rocks and Quaternary sediments (green and white areas, respectively, in Figure 1a; GeomapNavi2014). Those sedimentary rocks cover the pre-Cretaceous basement rocks and constitute the western margin of the Joban basin. Active faults are shown with red lines in Figure 1a.

We sampled the hot spring water pumped from the depths shown in the third column of Table 1. Because of the limited sample volumes during the first sampling, we collected more samples later. However, the δD, δ¹⁸O, elemental concentration, and temperature for these locations showed small temporal variations even before and after the 2011 Tohoku-oki earthquake, so we treated the data collected at different times in the same way. Table 1 also summarizes the sampling dates, chemical data, and estimated iodine age.

Methods for chemical and isotope analyses

Hydrogen and oxygen isotope compositions (δD and δ¹⁸O) were measured by mass spectrometry or cavity ring-down spectroscopy. The CO₂/H₂O equilibrium method and the H₂ reduction method using Cr metal were used for the analyses of oxygen and hydrogen isotopes, respectively (mass spectrometers, Delta Plus and Delta V advantage; Thermo Fisher Scientific Inc., Waltham, MA, USA). The precision of these measurements was ±0.1‰ for δ¹⁸O and ±1‰ for δD. Oxygen and hydrogen isotope compositions are presented in the δ notation in per mille relative to the Vienna Standard Mean Ocean Water (V-SMOW). The precision for the analysis by cavity ring-down spectroscopy (Picarro cavity ring-down spectrometer L2120-i; Picarro Inc., Santa Clara, CA, USA) was ±0.1‰ for δ¹⁸O and ±0.6‰ for δD. For the analysis of I, Br, and Cl, water samples were filtered using 0.45-μm membrane filters. Iodine concentrations in the groundwater samples were determined by inductively coupled plasma mass spectrometry (Agilent 7700; Agilent Technologies, Santa Clara, CA, USA). Groundwater samples were diluted with 0.5 wt.% tetramethylammonium hydroxide and spiked with Re as an internal standard. Cl and Br concentrations were determined by ion chromatography (Dionex, DX-500; Thermo Fisher Scientific Inc., Waltham, MA, USA).

To determine iodine age, ¹²⁹I/I ratios of deep groundwater were measured by accelerator mass spectrometry (AMS), using the sample preparation scheme developed by Muramatsu et al. (2008). Groundwater samples were purified by solvent extraction and back extraction using CCl₄. Purified iodine was precipitated as AgI by adding AgNO₃, and the AgI precipitate was washed with ultrapure water and NH₄OH. The supernatant was removed after centrifugation, and the AgI was freeze-dried. The ¹²⁹I/I ratio was measured at the Micro Analysis Laboratory Tandem Accelerator (MALT) at the University of Tokyo, following the method detailed in Matsuzaki et al. (2007).

The ³⁶Cl/Cl ratios of the samples were measured by AMS at the Australian National University (Fifield et al.2010) or at the Purdue Rare Isotope Measurement Laboratory (PRIME Lab), Purdue University, USA (Sharma et al.2000) to investigate the age of Cl. The samples were acidified with concentrated HNO₃ to precipitate AgCl by adding AgNO₃. The AgCl precipitate was separated by centrifugation and dissolved in NH₄OH. The SO₄^2- was precipitated as BaSO₄ by adding a saturated Ba(NO₃)₂ solution to remove isobaric interference from ³⁶S, and the BaSO₄ precipitate was eliminated by filtration. The AgCl was precipitated again and separated by centrifugation before drying. Sample numbers 11 and 12 were prepared following the procedures described in Tosaki et al. (2011), and their ³⁶Cl/Cl were measured with the AMS system at the Tandem Accelerator Complex, University of Tsukuba (Sasa et al.2010). Isotope ³H is one of the most commonly employed radioisotopes used to identify the presence of a modern water component. The ³H concentrations in groundwater were measured after electrolytic enrichment, using a liquid scintillation analyzer (Packard 3100TR, PerkinElmer 1220 Quantulus (PerkinElmer Inc., Waltham, MA, USA), or LSC-LB5 (Hitachi Aloka Medical, Ltd., Mitaka, Tokyo, Japan).

Isotope compositions and the estimation of hypothetical end-members

Figure 2 exhibits isotope and chemical compositions of groundwater from areas A to D. The d parameter (d = δD - 8δ¹⁸O) after Dansgaard (1964) is mostly in the range of 15 to 20 for the Abukuma area (Takahashi et al.2004) and 10 to 12 for the Kanto Plain (Inamura and Yasuhara2003). The δ¹⁸O values of several samples collected from the A and D areas are slightly heavier than the meteoric water lines using d values of 10 and 20 (MWL in Figure 2a). The positive shift of δ¹⁸O can be explained by isotopic exchange between water and rock (Figure 2a), and this characteristic is often found in connate water such as oil field brine (Clayton et al.1966; Mahara et al.2012). The relationship between the Cl and δ¹⁸O of the samples from areas A and B cannot be explained by simple mixing between meteoric water and seawater (Figure 2b). The Cl concentration of the hypothetical end-member was determined to be 361 mM by extrapolating the best-fit line for the data from area D to 0‰ of δ¹⁸O. Higher Cl in some data from areas A and B may be due to the mixing of seawater through the sedimentary rocks.

The concentration of Br correlates well with that of Cl (correlation coefficient, R = 0.99), suggesting simple mixing between the meteoric water and seawater (Figure 2c). On the other hand, the iodine concentration in deep groundwater is much higher than in meteoric water and seawater (Figure 2d). The highest iodine concentration was approximately 150 times higher than the seawater value. The I and Br concentrations of the hypothetical end-members were inferred from the mixing lines between the meteoric water and data from area D (asterisks in Figure 2c,d), using the Cl concentration at the end-member in Figure 2b. The Br and Cl concentrations of the hypothetical end-members are diluted (Figure 2c). The I and Cl concentrations for some samples from areas A and B can be explained by the mixing of the hypothetical end-member (asterisk), meteoric water (filled black circle), and seawater (filled black square), except for two samples with low I and Cl concentrations (Figure 2d). In contrast, those for samples from areas C and D in granitic basement can be explained by the mixing of the hypothetical end-member with meteoric water.Figure 3 exhibits an interesting correlation among the four areas. The I/Cl increases with increasing temperature of deep groundwater, and the ratio and temperature for areas C and D are much higher than for areas A and B.

Age of I estimated from ¹²⁹I/I

We estimated the ages of I in the deep groundwater from the measured ¹²⁹I/I of the water samples. Three gray curves in Figure 4 indicate mixing lines, on logarithmic scales, of the water sample with the highest iodine content (number 2 in Table 1), with the anthropogenic meteoric water, pre-anthropogenic seawater, and pre-anthropogenic meteoric water. Snyder and Fehn (2004) report five ¹²⁹I/I values of anthropogenic meteoric water in Japan that vary widely from 2.0 × 10^-10 to 7.9 × 10^-9 (cross marks in Figure 4a). We did not measure the ¹²⁹I/I of meteoric water, using their lowest value instead, because it falls on the mixing line. To draw the other two mixing lines, we used a reported value of ¹²⁹I/I (1.5 × 10^-12) for the pre-anthropogenic water (Moran et al.1998; Fehn et al.2007a) and iodine concentrations of Abukuma River near the sampling site after Tagami and Uchida (2006) and typical seawater (Table 1). The results for the samples from areas A, B, and D can be explained by the mixing of the water with the highest iodine content with the pre-anthropogenic meteoric and seawater (Figure 4). This is consistent with the relationship between iodine and Cl in Figure 2d. The two samples from area C have higher ¹²⁹I/I values (green filled circles in Figure 4) than the pre-anthropogenic water. They also contained detectable ³H (half-life: 12.32 years) of 1.28 ± 0.04 and 0.6 ± 0.1 TU, where 1 TU is defined as the ratio of 1 tritium atom to 10¹⁸ hydrogen atoms. Thus, the water from area C is likely to have mixed with shallow groundwater containing anthropogenic ¹²⁹I, and we did not use them for the age determination. Sample number 6 from area B contained detectable ³H, but no significant contamination of anthropogenic ¹²⁹I was recognized.

where R _obs is the measured ¹²⁹I/I, R _i is the initial ¹²⁹I/I (assumed to be 1.5 × 10^-12), and λ₁₂₉ is a decay constant of 4.41 × 10^-8 year^-1 (Fehn2012). Most samples have a ¹²⁹I/I value of (0.2 to 0.3) × 10^-12 irrespective of the iodine concentration and I/Cl ratio (Figure 4b), which gives iodine ages of 36 to 46 Ma. We neglect fissiogenic ¹²⁹I produced by spontaneous fission of ²³⁸U because its influence is small (Fehn et al.2000; Tomaru et al.2007a,[c],2009a,[b]). The homogeneity of iodine age (39 ± 4 Ma) suggests that the source of iodine is limited.

Secular equilibrium of ³⁶Cl of the deep groundwater

We calculated the ³⁶Cl/Cl ratios at secular equilibrium between the production and decay of ³⁶Cl using the chemical compositions of the host rocks (Additional file1: Table S1). Figure 5 shows the relationships between ³⁶Cl/Cl and 1/Cl for the samples from the four areas. The secular equilibrium ³⁶Cl/Cl (R_e) can be calculated by the following equation (Andrews et al.1986; Snyder and Fabryka-Martin2007; Morikawa and Tosaki2013):

$R_e=\frac{\varnothing {\mathrm{\sigma }}_{35\mathrm{C}\mathrm{l}}}{{\mathrm{\lambda }}_{36}}\mathrm{N}$

(2)

The thermal neutron absorption cross section of ³⁵Cl (σ_35Cl) is 43.6 × 10^-24 cm², the decay constant (λ₃₆) of ³⁶Cl is 2.3 × 10^-6 year^-1, and the isotopic abundance (N) of ³⁵Cl is 0.7577. The thermal neutron flux (∅), in neutron per square centimeter per year, was estimated from the whole-rock major and trace element compositions of various sedimentary rocks and granitic rocks collected in this area (Additional file1: Table S1; Morikawa and Tosaki2013). The secular equilibrium ³⁶Cl/Cl (R _e ) are calculated and plotted as horizontal dashed lines for areas A, C, and D with granitic basement (R _e  = (1.77 ± 0.82) × 10^-14), and as horizontal solid lines for area B with sedimentary rocks (R _e  = (9.87 ± 3.10) × 10^-15) in Figure 5. The equilibrium between ³⁶Cl production and decay is established in approximately 1.5 Ma (Andrews et al.1986). The ³⁶Cl/Cl values of most of our samples are plotted at or near the horizontal lines, and they have nearly reached secular equilibrium. Note that a state of true equilibrium may not be attained with a single, specific type of host rock in a strict sense, possibly leading to some uncertainty in the above interpretation. Overall, the deep groundwater in the Joban and Hamadori area has not been affected by modern seawater, even though the areas are located near the coast.

The origin of iodine

We discuss the possible sources of iodine in the deep groundwater collected at ten hot springs in the Joban and Hamadori areas and implications for water circulation in the Tohoku subduction zone. The measured concentrations of iodine in the groundwater were much higher than in seawater (Figure 2d), whereas Cl and Br concentrations were lower than in seawater (Figure 2b,c). Iodine is strongly bound to organic matter under marine conditions (Elderfield and Truesdale1980) and is released during the diagenesis of organic material to form natural water enriched in iodine, such as oil field brines and pore waters in marine sediments (e.g., Tomaru et al.2009a,[b]; Fehn2012). Previous papers report that Br is also released during decomposition of organic materials (Tomaru et al.2007c,2009b), but the enrichment of Br did not occur in the deep groundwater we studied. A possible reason for this is that the release rate of iodine from organic matter is higher than that of Br (Tomaru et al.2009b). The iodine of about 40 Ma in age in the deep groundwater must have derived from sedimentary rocks, and there are two possible sources, the Joban sedimentary basin, east of the hot springs where we did the sampling (Figure 1), and the subducted marine sediments from the Japan Trench beneath Tohoku, as discussed below.

We do not exclude the Joban basin as a possible iodine source, but we consider it unlikely for the following three reasons: (1) The lower Miocene formation in the Joban basin contains a thick layer of fine to coarse sandstone and conglomerate, which comprise a gas reservoir, and the iodine-containing water may have migrated up along the formation towards the hot spring wells where we sampled the deep groundwater (Figure 1b). However, the geothermal gradient at the Joban gas field (Figure 1a) is 18°C km^-1 (Tanaka et al.2004), and the expected temperature at maximum depth (approximately 3 km) for the formation is about 70°C. This is lower than the temperature of the groundwater in area D (close to 80°C maximum, Figure 3), and the high temperature of the groundwater cannot be explained by the water in the Joban basin as its source (there is no heat source such as a volcano in the sampling area). (2) Deep groundwater samples in areas A, C, and D were collected in basement granitic rocks, but the temperature of the water is higher than that of water from the Joban sedimentary rocks in area B (Figure 3). Presumably, hot groundwater from depths comes up close to the surface in areas A, C, and D through the basement rocks. (3) Several active faults are developed in the sampled and neighboring areas, and surface ruptures formed along the Itozawa and Yunodake Faults during the 2011 Iwaki earthquake (red lines in Figure 1a; The Research Group for Active Faults of Japan1991; Toda and Tsutsumi2013). This earthquake discharged thermal water in Iwaki City for at least 2 years (Sato et al.2011; Kazahaya et al.2013). The earthquake was of a normal-fault type with complex ruptures and aftershocks extending to a depth of almost 15 km (Fukushima et al.2013). The active tectonics of the sampled areas must have promoted discharge of deep groundwater from depths through the basement rocks, rather than from the Joban basin. The geothermal gradients in the Joban-Hamadori areas are 24°C km^-1 to 33°C km^-1 (Tanaka et al.2004), but the temperature of the hot spring water is higher than the geothermal gradient in about half of the hot springs where we sampled groundwater (Figure 3).

The deep groundwater we sampled came through the pre-Cretaceous basement rocks. Thus, the geological situation in our study area is different from those studied by Fehn et al. (2003) and Tomaru et al. (2007b,2009a), who sampled pore waters in sediments on the continental slopes in the Nankai Trough and off the Shimokita Peninsula, northern Japan. In our study area, it is hard to expect sedimentary rocks younger than 40 Ma beneath the pre-Cretaceous basement. The subduction zone in the Japan Trench is regarded as a classical erosion boundary and sediments on the continental slope deposited on Cretaceous rocks (von Huene and Culotta1989). Moreover, von Huene and Lallemand (1990) proposed that the subduction of seamounts would play an important role in the erosion and estimated that the rate of tectonic erosion at a subducting plate boundary is comparable to the rate of sediment accretion. Thus, rocks constituting the upper plate in the Japan Trench are unlikely to be accreted large-scale sedimentary rocks, younger than the Cretaceous. However, a small-scale accretionary prism is developed throughout the Japan Trench, and a thin interplate layer, possibly composed of sedimentary rocks, is recognized along the plate interface down to a depth of almost 20 km (Figure 6; Tsuru et al.2002; Miura et al.2003). The types of subducting sediments are unclear in the southern part of the Japan Trench, but DSDP drilling at leg 56 site 436 revealed that incoming sediments are mostly pelagic sediments of a few hundred-meter thickness and of Miocene to Pliocene in age (about 20 to 2 Ma) (The Shipboard Scientific Party1980; Kimura et al.2012, Figures two panel b and four). Chester et al. (2013) report very similar sediments beneath the coseismic fault during the Tohoku-oki earthquake in the J-FAST drill cores in the Japan Trench. Moreover, sediments in the western Pacific Ocean near the Japanese Islands contain organic carbon of 0.5 to 1.0 wt.%, much higher than in sediments in the Pacific Ocean away from land (Premuzic et al.1982, Figure one; Klauda and Sandler2005, Figure three). We thus consider that subducted sediments in the Japan Trench are possible sources of iodine in the deep groundwater in the Joban-Hamadori areas.

Implications for water circulation in subduction zones

The subducted sediments undergo a series of dehydration reactions (Figure 6). Kimura et al. (2012) showed using a new temperature calculation that the opal A-CT-quartz transition and smectite-illite transition occur at the plate interface between 40 to 80 and 80 to 120 km, respectively, landward of the Japan Trench (Figure 6). The hypothetical end-member determined in this study shows higher iodine and lower Cl concentrations than seawater. The dehydration of hydrous minerals releases water and dilutes the Cl concentration in pore fluid; hence, the source of deep groundwater is probably deeper than those dehydration depths. Moreover, the isotope and trace element composition revealed the involvement of sediments in the generation of island-arc magma in northeast Japan (Shibata and Nakamura1997), strongly suggesting that the subduction component is carried to depths exceeding 150 km. The mantle wedge may be serpentinized, in view of a result of seismic wave tomography (Figure 6; e.g., Miura et al.2003), also suggesting the release of a large amount of water beneath the mantle wedge.

Microseismicity has been recognized in the mantle wedge to form the aseismic front (AF in Figure 6) roughly below the coastline (Yoshii1975,1979). Thus, the overriding plate off the coast of the study area is characterized by the mantle wedge and basement crust, both seismogenic, and slope sediments covering the basement rocks. Shale is very impermeable, particularly in the direction normal to the bedding plane (Kwon et al.2004a,[b] and references therein), and even the clay-bearing fault gouge has a permeability of as low as 10^-17 to 10^-22 m² (Faulkner and Rutter2000). Therefore, fluids probably cannot easily penetrate through the slope sediments containing low-permeability shale, although we are not aware of any permeability data on the slope sediments in the Tohoku subduction zone, whereas fractured basement rocks have much higher permeability (e.g., 10^-15 to 10^-18 m² in the case of cataclasite along the Median Tectonic Line; Uehara and Shimamoto2004). Hence, the overall permeability structures might have promoted fluid flow through fractured basement rocks from the subducting plate towards the coastline.

In addition, Imanishi et al. (2012) recognized that normal-fault type earthquakes had occurred locally in the Joban and Hamadori areas even before the 11 March 2011 Tohoku-oki earthquake, despite the stress field in Tohoku being dominated by E to W compression. The 11 April 2011 Iwaki earthquake and associated surface ruptures were of normal-fault type (Fukushima et al.2013; Toda and Tsutsumi2013). Moreover, Imanishi et al. (2012) found a clear normal-fault type earthquake sequence extending from the subducting plate to the aftershock area and called it a branching fault (shown by a thick red line in Figure 6; small gray circles are aftershocks traced along profile EE’ in Figure four of their paper). The study area is the only place in the focal area of the Tohoku-oki earthquake where aftershocks are distributed continuously from the subducting plate to the coastal area of Tohoku. The branching fault is a broad zone of microseismicity and probably acted as a fluid conduit from the subducting plate to the study area. The discharge of a large amount of thermal water after the Iwaki earthquake (Sato et al.2011; Kazahaya et al.2013) strongly suggests that the earthquakes promote upward movement of deep groundwater in the areas. If the deep groundwater in the study area came out through the broad branching fault zone, the path of the water along the fault would be about 80 to 90 km. The ages of the incoming sediments range from 2 to 20 Ma, as reviewed above, and it takes about 2 Ma for the sediments to reach the intersection of the subducting plate and the branching fault. Thus, the iodine age of about 40 Ma gives 2 to 5 mm year^-1 as the average speed of H₂O migration in the overlying plate. The low speed may imply that fractures are sealed rapidly, and fault valve behavior (Sibson1992,2013) controls the fluid flow. Whether an average velocity of a few to several millimeters per year is appropriate or not needs a test from additional study in the future.

According to Kazahaya et al. (2014, Figure two panel B), the end-member of Arima-type deep-seated fluids in southwest Japan, which has been considered to be derived from the subducting Philippine Sea Plate, is within the range of magmatic gas in the δD to δ¹⁸O diagram. The values of δD and δ¹⁸O for our deep groundwater samples (Figure 2a) fall on the trends for altered seawater in their diagrams. In addition, our samples have lower Cl concentration than seawater, whereas the Cl concentration of Arima-type deep-seated fluids is more than 4 wt.% (Kazahaya et al.2014). It should be emphasized that the δD, δ¹⁸O, and Cl data of our samples do not have Arima-type deep-seated fluid characteristics, although they might have migrated through the mantle wedge. One possible cause of the difference is the difference in temperature between the Nankai Trough and Japan Trench; that is, the calculated temperature along the slab/mantle interface at a depth of 50 km is only 200°C in northeast Japan (Peacock and Wang1999). More detailed geochemical investigations need to be undertaken to determine the origin of iodine and fluid-flow paths in the Tohoku subduction zone. The Joban and Hamadori area, where we sampled deep groundwater from basement rocks, may be a good window to look into the fluid circulation in a subduction zone related to subducted sediments.