Deep crustal fluid upwelling through structural pathways observed from non-active volcanic regions, western Kumamoto, Japan

Natural springs containing volcanic and magmatic components occur along major volcano-seismotectonic regions over the worlds. However, features of the deep-originated waters were less documented from regions where active volcanic and magmatic activities are not distributed. To characterize the presence of deep fluids of non-volcanic origin 28 groundwater samples (~1,230 m deep) were collected from hot spring sites located at western coast of Kumamoto where the typical subduction related magmatisms are absent. The samples were measured for dissolved ion concentrations and stable isotope ratios (δ 2 H H2O , δ 18 O H2O , δ 13 C DIC and δ 34 S SO4 ) that were compared with data of 33 water samples from vicinity surface systems. The groundwaters were classified into three types based on major hydrochemistry: high Cl - fluid, low concentration fluid, and high HCO 3- fluid. Our dataset suggests that the high Cl - fluid was formed by saline water mixing with aquifer waters of meteoric origin and subsequently evolved by reverse cation exchange. The low concentration fluid is identical to regional aquifer water of meteoric origin that was subjected to cation exchange. The high HCO 3- fluid showed the highest HCO 3-2 concentrations (~3,888 mg/l) with the highest δ 13 C DIC (-1.9‰). Taking recent geophysical mappings under the study area, we suggest that dissolved carbon was of mantle origin and fluids with high HCO 3- generated in lower crust were transported towards surface through structural weakness under open tectonic setting. Observed δ 2 H H2O and δ 18 O H2O shifts support this scenario. The occurrence of deep crustal fluid discharges was sporadic and limited in surface in the study area. Their impacts on surface hydrological systems were minimal except few locations.


Introduction
Artesian springs with high temperature and/or elevated concentrations occur in major active volcanic and seismotectonic regions globally. These highly concentrated dissolved solutes are facilitated with incorporation of volcanic and magmatic gases and fluids (Aiuppa et al. 2009(Aiuppa et al. , 2017, and simultaneous high temperature weathering or alterations that can occur in crust deeper than active hydrological systems in surface (Hedenquist and Lowenstern 1994). Discharge of these fluids is commonly observed in vicinity of active volcano-hydrothermal fields. For example, relevant number of studies tried to depict origins and pathways of deep originated fluids to quantify the impact of volcanic and magmatic activities on lateral material transportations from land to ocean (Chiodini et al. 2000;Dessert et al. 2009;Gaillardet et al. 2011;Rive et al. 2013;Hosono et al. 2018;Perez-Fodich and Derry 2019). Similar high concentration hot spring waters are found in regions where active volcanic and magmatic activities are not existing (Tomaru et al. 2007;Kusuda et al. 2014;Togo et al. 2014;Morikawa et al. 2016;Kusuhara et al. 2020 and references therein), however, origin, distribution, pathway, and thus their impact of hydrochemical features on surface systems are less well known.
The major sources invoked to explain for the origin of deep originated fluids include magmatic fluids, slab fluids and deep crustal fluids (Hedenquist and Lowenstern 1994;Ague 2014;Morikawa et al. 2016). The magmatic fluids present in a crust beneath active volcanoes that are located typically along the volcanic front with some possible exceptions on back arc (Fig. 1a). Dehydration of subducting oceanic plates releases waters and fluids from slab sediments and rocks (here we call slab fluids) into mantle wedges. Some many hot springs in southwest Japan are thought to be originated directly of these fluids on fore arc system (Tomaru et al. 2007;Kusuda et al. 2014;Togo et al. 2014;Yoshida et al. 2015;Morikawa et al. 2016;Kusuhara et al. 2020). Alternatively, fluids generated in lower crust (here we call deep crustal fluids) can also contribute to surface systems along major pathways especially at regions where substantial heat source exists beneath lower crust such as upwelling asthenospheric mantle (e.g., Zhao et al. 2018), however, this phenomenon has less advocated from nature of lived hot spring discharges due to rare opportunity to test this hypothesis. Figure 1a shows the depth contours of subducting slabs under Japanese archipelago illustrating that western Kyushu is located on unique setting where possible subduction plate is in mantle deeper than 200 km and typical subduction related magmatisms are absent (Zhao et al. 2021). Some hot springs occur along the coast of western Kumamoto in western Kyushu where none of active volcanic activities distribute (Fig. 1b). Notably, these historically recognized hot springs and other recently developed hot spring facilities are located along and on extension of the major faults ( Fig. 1) that possibly allow deep fluids in the crust passing toward surface. There have been no reports comprehensively investigating the origin and formation of these hot springs, however, we thought that precise geochemical characterization on these hot springs may be able to reveal a presence, properties, and discharging processes of the deep crustal fluids.
Stable isotopes are useful for identifying the origin of waters and dissolved ions, and for interpreting the biogeochemical processes in aquifers (Kendall 1998;Cook and Herczeg 2000). Stable isotope ratios of oxygen and hydrogen in the water molecule (δ 2 H and δ 18 O) have been used to examine the contribution of fluids that had experienced high temperature water-rock interaction resulting in δ 18 O isotopic shift towards enriched compositions leaving δ 2 H unchanged (Giggenbach 1992). Stable carbon isotope ratio of dissolved inorganic carbon (δ 13 C) is a fundamental tracer for identifying a contribution of magmatic or mantle-derived carbon in fluids (e.g., Chiodini et al. 2000;Yamada et al. 2011;Rive et al. 2013). Sulfur isotope ratio of dissolved sulfate (δ 34 S) is also applicable for evaluating volcanic influence on spring waters and for interpreting redox processes in aquifer environments (e.g., Kusakabe et al. 2000;Hosono et al. 2014aHosono et al. , 2018. This study integrates the results from these isotopic analyses and characterized geochemical features of hot spring waters with comparison of those from other surface waters for trying to find out the signature of deep crustal fluid upwelling and to assess its impact on surface hydrological systems.

Outline of the Study Area
The study area is located on the western coast of the Kumamoto Prefecture in Kyushu Islands and consists of four local domains, Uto, Amakusa, Yatsushiro and Ashikita districts (Fig. 1d). Uto and Amakusa are located on western peninsula and small islands, while Yatsushiro and Ashikita are on western coast of the main island Kyushu, respectively. Coastal flat plain in Yatsushiro is the reclaimed land on tidal flat. The Kyushu has humid subtropical climate and average annual temperature and precipitation of Amakusa and Yatsushiro are 16.4 and 16.8°C and 2,076 and 1,979 mm during 1981-2010, respectively (http://www.jma.go.jp/jma/menu/report.html). Around 40% of total precipitation occurs during summer rainy season in June-July.
In the Kyushu Islands the active volcanoes are generally situated in two major graben structures, Beppu-Shimabara and Kagoshima grabens ( Fig. 1b) (Nakada et al. 2016) that are facilitated by unique open tectonic setting partly associated with wet upwelling flow in the mantle wedge (Zhao et al. 2021). The study area is located on just south of the Beppu-Shimabara graben structure, lying between two major active fault systems, Hinagu and Futagawa strike-slip faults (Fig. 1d), for example, which activities generated the 2016 Mw 7.0 Kumamoto destructive earthquakes (e.g., Hosono et al. 2019Hosono et al. , 2020. As stated earlier, neither active volcanic activities nor typical subduction related magmatisms are recognized within and beneath the study area. For instance, some recent tomography images (Zhao et al. 2018(Zhao et al. , 2021 depicted the presence of upwelling asthenospheric mantle in the corner of mantle wedge and there is no major magmas or melts displayed in the crust beneath the study area, which low seismic velocity anomalies are generally observed just beneath active volcanoes like Aso and Kuju (Fig. 1c).
Paleozoic to Mesozoic low-temperature and high-pressure metamorphic rocks and metasedimentary rocks of accretionary prism are widely distributed in the southern part of the Usuki-Yatsushiro Tectonic Line (Fig. 1e). In the northern part of this tectonic line Cretaceous Himenoura Group, alternative sedimentary layers of sandstone, mudstone, shale and conglomerate, comprises of the basement rocks of the study area. Cretaceous metamorphic rocks are also outcropped in Amakusa but their distributions are quite limited. These geological basements are overline by Paleogene sedimentary formations which are further intruded by Miocene dikes and overline by Miocene to Pleistocene volcanic rocks mainly of andesitic to dacitic compositions. Most of these young volcanic rocks are found at Uto peninsula (1.7 to 4.2 Ma; Yokose et al. 2009) and Ashikita district (Fig. 1e). Some limestone formations are interlayered but their distribution is quite limited that are not seen near majority hot spring sites.

Samples and Methods
In total 28 groundwater samples were collected from borehole of the hot spring facilities except one sample that was collected from artesian spring during June 2019 and May 2020 (Fig. 1d, Table S1). The depth of borehole is ranging from 70 to 1,230 m and are generally deeper than 800 m and temperature of the waters were ranging from 19.8 to 45.2°C (Table S1). Shallow boreholes (70-100 m) are of Yunoko and Shimoda hot springs, which hot springs with Oyano hot spring were first discovered as artesian flowing from the ground (Fig. 1b) and then later explored subsurface by drilling. Other facilities explored groundwater with elevated temperature though originally there were no hot spring discharge symptoms on a surface land. In addition, 19 stream waters, 10 mountain spring waters, and 4 shallow well (~ 45 m in depth) waters were collected within the same area during July 2019 and April 2020 for isotopic and geochemical comparisons ( Fig. 1d), which samples are representative for waters in meteologically-driven surface hydrological systems (we termed 'surface water' for all these waters). No precipitation occurred during 3 days before the sampling.
Groundwaters from hot spring sites were collected directly from the pipes of the pumping or artesian flows. Stream and artesian mountain spring waters were collected on sites, while shallow well waters were obtained using electric pumping or Baylor tube sampler. Field parameters such as temperature, electrical conductivity, pH, oxidation reduction potential, and dissolved oxygen were monitored using field meters (HORIBA D-54) and water samples were collected after these values became constant. The HCO3concentration was determined by in situ alkalinity measurement following titrating with sulfuric acid. Water samples were stored in 100, 500 and 2,000 ml polyethylene bottles for major ion concentrations (Na + , K + , Ca 2+ , Mg 2+ , Cl -, SO4 2-, and NO3 -), δ 13 C and δ 34 S analyses, while 20 ml glass vials were used for collecting water samples for δ 2 H and δ 18 O analyses, respectively. The samples were filtered through ADVANTEC ○ R 0.2 µm celluloseacetate filters onsite before storing for 100 ml bottles and immediately after taking back to laboratory for 500 and 2,000 ml bottles, respectively. The filtered water samples in latter two bottles were adjusted to pH 12 to 12.5 with NaOH and to pH 2.5 to 3 with 6 M HCl, respectively, then dissolved inorganic carbon and SO4 2were collected as BaCO3 and BaSO4 by adding 10% BaCl2 for δ 34 S and δ 13 C analyses.
Major ion concentrations were measured by ion chromatography (Compact IC 761, Metrohm, Switzerland). Both δ 13 C of BaCO3 and δ 34 S of BaSO4 were determined using a continuous-flow gas-ratio mass spectrometer (Delta V Advantage, Thermo Fisher Scientific, USA) interfaced with an elemental analyzer (Thermo Fisher Flash 2000, USA) (Hosono et al. 2014a(Hosono et al. , 2015. International standards (NBS123 and OGS-1) and several carbonate and sulfate materials were used for quality check of the measurement data. The precision of both δ 13 C and δ 34 S analyses were generally better than ±0.2‰. The δ 2 H and δ 18 O of water molecular were determined by a gas-ratio mass spectrometer (Delta V Advantage, Thermo Fisher Scientific, USA) coupled with an automatic water-gas equilibrization devise (Nakano Denshi Co., Ltd., Japan). Based on replicate measurements of standards and samples, the analytical precisions for δ 2 H and δ 18 O were better than ±0.5‰ and ±0.05‰, respectively.

Hydrochemical classification
According to trilinear diagram hydrochemical features of groundwater samples from hot spring sites clearly differ to those for surface waters (Fig. 2a). The surface water samples are broadly plotted on compositional field between Ca-HCO3type with pristine signature and modern sea water, suggesting presence of sea water mixing into fresh surface waters with lower dissolved solutes (electric conductivity of around 50 μS/cm increased near to 1,000 μS/cm, Table S1). On the other hand, the groundwater samples from hot spring sites are exclusively plotted along two different trends with increased concentrations (electric conductivity of ~42,500 μS/cm), high Clwith variable Ca 2+ ratio and high Na + with variable HCO3ratio (Fig. 2b). Thus, these groundwaters changed their hydrochemical features owing to several factors including mixing of waters with other origins and geochemical evolutions such as cation exchange and redox processes.
The groundwater samples categorized by the former trend are relatively more enriched in Cland Na + and thus we classified this type of waters as High-Clfluid ( Fig. 2b and c, Table S1). The groundwater samples categorized by the latter trend are subdivided into two groups, one characterized with greater HCO3concentrations (~3,888 mg/l) and the other with least dissolved concentrations but with relatively higher SO4 2ratio (Fig. 2d). We classified these two types of groundwaters and named Low-concentration fluid and High-HCO3fluid, respectively ( Fig. 2b and d, Table S1). Geographical distribution of three types of fluids, High-Cl -, Low-concentration and High-HCO3fluids, are shown in Figure  1d. This hydrochemical classification and terminology are used in the following discussion. Figure 3 shows δ 2 H versus δ 18 O diagrams plotting all collected samples and data of some hypothesized source waters for comparison. This comparison allows us to identify the possible origins of water samples of our concern. All surface water samples were closely plotted on local meteoric water line except one stream water sample that is slightly deviated towards composition of modern sea water (Fig. 3a). This suggests that all these waters are mainly of meteoric origin except one sample that was clearly affected by a mixing of sea water. We admit that some of these waters are affected by sea water component in terms of major ion chemistry. However, that effect was small and not obviously reflected on δ 2 H and δ 18 O shift and thus their compositions were regard as representative feature of waters in surface hydrological systems over the study area.

Origin of water
On Figure 3b, samples of High-Clfluid are broadly plotted along an array between compositional field of surface waters and modern sea water. In combination with Clconcentration (Fig. S1), our simple mixing model suggests that this fluid was originated of the surface fresh water admixed with modern sea water with variable mixing proportions ranging from few to nearly 100% due to sea water intrusion and/or contribution of saline pore waters in sediments. This interpretation agrees well with the fact that all High-Clfluids are located just on the coasts or on the reclaimed land in Yatsushiro (Fig. 1d). In turn, for Low-concentration fluid all sample plots were fallen on the compositional field of the surface water (Fig. 3c), indicating their intrinsic meteoric origin analogous to surface circulated waters.
In contrast to High-Cland Low-concentration fluids, some samples of High-HCO3fluid are plotted on right to the local meteoric water line towards hypothesized magmatic water proposed by Giggenbach (1992) (Fig. 3d), although this isotopic shift seems not to be much drastic compared to ones reported from some volcanic fields and other hot spring waters of subduction slab origin (e.g., Kita et al. 2009;Kusuda et al. 2014;Morikawa et al. 2016). Nevertheless, observed isotopic shift possibly indicates incorporation of deep shield fluids that have experienced high temperature water-rock interaction in lower crust particularly in case of our study setting as explained more later.

Cation exchange
High-Clfluid showing variable Ca 2+ ratio (Fig. 2b) possibly indicates the occurrence of other processes than source inputs, i.e., geochemical evolution. The chloro-alkaline indices (CAI) (Schoeller 1967) expressed by the flowing equations in milliequivalent concentrations can be used to test the occurrence of cation exchange reactions that control cation concentrations in aquifers. Here, the cation exchange and reverse cation exchange are expected to be occurred when both CAI-1 and CAI-2 show negative and positive values, respectively. These equations are applicable when pH of solutions is lower than 10.3. The pH measured for our samples were all below 10.3 and thus we can test these indices. Figure 4 displayed both CAI-1 and CAI-2 yielded positive values and the values are increased with increase in Ca 2+ ratio for the High-Clfluid. We can thus conclude that hydrochemical features of High-Clfluid was formed by the sea component mixing and subsequent progressive reverse cation exchange reaction in aquifer. Exchange of calcium ion in exchangeable site in aquifer's solid phase was released by Na + in water, assuming to easily be enhanced by additives of saline water with high Na + concentration. Moderately, but similar variation in Ca 2+ ratio was observed for Low-concentration and High-HCO3fluids (Fig. 2). However, both fluids had negative CAI-1 and CAI-2 values which tended to decrease with decrease in Ca 2+ ratios. Thus, these fluids had evolved along normal cation exchange reaction in contrast to High-Clfluid.

Redox process (sulfur cycle)
The redox process is another important geochemical evolutional process that must be considered. This is true because some groundwaters from hot spring sites showed low dissolved oxygen (< 2 mg/l) and negative oxidation reduction potential (mV) while surface waters generally showed aerobic regime (Table S1). In general, Clconcentrations are not affected by redox changes and NO3was not highly concentrated (mostly < 2.0 mg/l except three samples with > 10 mg/l due to local contamination) in the water samples that we investigated (Table S1). Thus additional source characteristics and redox processes will be discussed hereafter for SO4 2with conjunction of δ 34 S (Fig. 5a). For this, compositional variations for surface waters were considered as juvenile indices that are defined as -7‰ < δ 34 S < 15‰ and 2 mg/l < SO4 2-< 30 mg/l, respectively.
On Figure 5b, some High-Clfluids showed elevated SO4 2-(~1,700 mg/l) towards compositions of modern sea water (δ 34 S: ca. 20.3‰, SO4 2-: ca. 2649 mg/l; Rees et al. 1978), suggesting significant saline water contribution as explained previously. However, other samples showed lower concentrations with elevated δ 34 S up to 28‰, indicating an occurrence of microbial reduction of concentrated sulfate of marine origin (Strebel et al. 1990;Tuttle et al. 2009). Some samples were not measured for δ 34 S because of their low SO4 2concentrations probably due to progressive sulfate reduction reaction. Such phenomenon is commonly seen for coastal aquifers either at near to study fields in Kumamoto (Hosono et al. 2014a) or some other countries (Hosono et al. 2010(Hosono et al. , 2011(Hosono et al. , 2014b. The Low-concentration fluid displayed compositions almost identical to those for surface water except two samples with moderately elevated SO4 2-(~280 mg/l) and δ 34 S (~17.4‰) and one sample with slightly elevated SO4 2-(~57 mg/l) and depleted δ 34 S (~13.7‰) (Fig. 5b). The observed signature for former two samples can be explained by mixing of sulfate of marine origin similar to High-Clfluids but that intensity was too small to clearly be detected on stable isotope ratios of water molecular (Fig. 3c). The signature of later sample can be explained by the oxidation of sulfide with lower δ 34 S (Taylor et al. 1984;van Everdingen and Krouse 1985). In summary, overall hydrochemical formation process of Low-concentration fluid resembles to that for surface waters except cation exchange reaction.
The High-HCO3fluid has the least SO4 2concentrations, which are lower than 16 mg/l, and for majority samples, ranging between detection limit and 5 mg/l (Table S1). Two samples had similar δ 34 S compositions to those for surface waters but other four samples showed most enriched δ 34 S signature (~45‰) among all samples with depleted SO4 2concentrations (Fig. 5b). No striking signatures of contribution from marine sulfate are observed, corresponding to results of δ 2 H and δ 18 O analyses. Overall, hydrochemistry of the High-HCO3fluid was changed their features with similar manner as found for Lowconcentration fluid, except less contribution from marine sulfate and tremendous HCO3enrichment (Fig. 2d).

Deep fluid upwelling
Significant contribution of sulfur from some heat sources was not detected from any water samples collected from our study area (Fig. 5), which δ 34 S values resembling to arc volcanic magmas typically in a range between 0 and 10‰ (Ueda and Sakai 1984). On the other hand, elevated HCO3in High-HCO3fluid may be originated of CO2 gas released from some heat sources that was dissolved in waters and preferentially transported towards surface leaving SO2 and other gasses (e.g., Ohsawa et al. 2002;Gultekin et al. 2011). The use of δ 13 C tracer in dissolved inorganic carbon can provide clue to test this hypothesis (Chiodini et al. 2000;Yamada et al. 2011;Rive et al. 2013).
On the δ 13 C versus HCO3concentration diagram (Fig. 6a), all surface water samples fall on δ 13 C compositions between -31.3 and -11.8‰ with the lowest HCO3concentrations generally less than ca. 100 mg/l (Table S1), except one sample with relatively elevated δ 13 C (-8.9‰) and HCO3concentrations (295 mg/l). These geochemical features are of the typical ecosystem-mediated carbon globally (Clark and Fritz 1997) and aquifer systems near the study areas (Ohsawa et al. 2002;Hosono et al. 2014) and thus are reasonably assigned as a representative composition of surface water systems before alteration by additional source inputs. On the same diagram, almost all groundwater samples from hot spring sites are plotted along the exponential curved trend based on two end-component mixing between hypothesized soil CO2 and deep originated CO2, suggesting progressive contribution of CO2 from deep crust into shallower hydrological systems.
A contribution of deep originated CO2 dominates in High-HCO3fluid compared to other fluids (High-Cland Low-concentration fluids) which contains only ~5% of carbon from that source (Fig. 6a). Therefore, significant contribution of CO2 from heat source is responsible for the formation of major hydrochemical features of High-HCO3fluid. Figure  6b displayed geographical distribution of relative contribution of deep originated CO2. It is notable from this map that groundwaters with the greater contribution are sporadically distributed along the fault systems, and interestingly, these anomalies are characteristically located near the historical hot springs sites, i.e., Yunoko, Shimoda and Oyano, regardless of their host geological formations (Fig. 1). It seems plausible therefore that the presence of relatively young volcanic rocks (Yokose et al. 2009), and presumably still lived magma beneath surface, is not necessarily associated for the occasion of observed deep originated CO2, instead, another heat source must be more critically important.
Considering geotectonic setting of the study area, upwelling asthenospheric mantle (Zhao et al. 2018(Zhao et al. , 2021 is the most plausible origin of deep sourced CO2 that was released into metasomatized lower crust triggered by crustal assimilation with this huge heat source. Similar phenomenon was reported from the Hitoyoshi Basin in non-volcanic regions at western Kumamoto (Fig. 1b) based on the same isotopic systematics  with some additional noble gas isotopes, e.g., 3 He/ 4 He ). In addition, groundwaters with greater δ 2 H and δ 18 O compositional shifts tend to be seen on locations where deep originated CO2 occurs (Fig. 6c and d), supporting above scenario. Hata et al. (2017Hata et al. ( , 2020 provided key information for putting more constraint on the realization of the deep crustal fluids. Their recent electrical resistivity image finely identified significant high conductivity blocks exist in western Kyushu under shallow depth (~50 km) in the upper mantle and lower crust, and argued the presence of deep hydrothermal fluids particularly under Hishikari (hydrothermal gold-silver ore deposits) (Fig. 1b), Hitoyoshi, Ashikita and Amakusa, where deep crustal fluid upwelling was anticipated. Thus, we propose fluids generated in a lower crust are the major origin of hypothesized deep fluids, although a possibility of partial involvement of recycled subduction slab fluids can't be neglected as their ultimate origin (Nakamura et al. 2019). However, observed degree of isotopic shift on δ 2 H and δ 18 O compositions (Fig. 6c) was not much dramatic as seen for deep originated CO2 contributions (Fig. 6a) probably due to both dispersion of CO2 gas and mitigation by the mixing with water of meteoric origin while the deep crustal fluids ascent through long distance in the crust.
Discharging slab originated fluids ubiquitous in southwest Japan is thought to be transported through major tectonic structures such as Median Tectonic Line, forearc thrust faults in the accretionary prisms and branching faults, driven by geo-and heat-pressures under compressional tectonic setting (Kusuda et al. 2014;Togo et al. 2014;Yoshida et al. 2015;Morikawa et al. 2016). Deep crustal fluids proposed in this study could ascend also through structural breaks in the Hinagu and Futagawa strike-slip faulting systems (Fig. 1) and possibly in relation to pressure release under open tectonic setting, with a similar law demonstrated from previous works in Kyushu (Hosono et al. 2004 and from other open tectonic regions (Cerón et al. 1998). It's interesting to note that stronger signature of the deep fluid crustal upwelling ( Fig. 6b and d) characteristically occurs near the locations where more than two different faults are crossing (Fig. 1d). This may be due to generation of lower pressure around cross point of the different stresses in the crust where pressurized deep crustal fluids can more easily release towards air pressure near surface (Cappa et al. 2009;Takashima and Kano 2005).
The results of this study showed δ 13 C was the most sensible tracers among all measured stable isotope ratios for fingerprinting deep originated components in groundwaters. Giving that the mantle sourced carbon mediated groundwaters (High-HCO3fluid) shared nearly half of total analyzed groundwaters (46%), this type of waters must to be extended wide in space bellow few hundred meters deeper in crust. Despite their prosperity in deep aquifers their discharges to near surface systems look somehow restrictive only where major fluid pathways exist, and thus, their overall impacts on surface hydrological systems seem to be small on this particular study setting, except only minor contamination with mantle originated carbon, i.e., ca. 5%, observed locally in shallow well water (20 m in depth) near Yunoko hot spring (Fig. 6a). These aspects can be taken when we consider carbon cycle on earth surface systems in other fields.

Conclusions
The hydrochemical features of hot spring waters were characterized to understand the origin, distribution and pathway of deep fluids in western Kumamoto using multiple stable isotope ratios (δ 2 H, δ 18 O, δ 13 C and δ 34 S) for an area where subduction related active volcanic and magmatic activities are not anticipated. Our results successfully implied the occasion of deep fluid transportation to shallower aquifers along active faults under open tectonic setting, which dissolved carbon and waters are assumed to be originated from upper mantle and lower crust, respectively. This study demonstrates the genesis of some carbon rich hot springs appearing on 'back-arc-side' non-volcanic regions. Additional analysis on noble gas isotopes may provide firmer evidence. The aspects from our results are useful for interpreting data from other fields and for more global generalization, and beneficial to users for hot spring resources.

Abbreviations
MTL: Median Tectonic Line; TL: tectonic line; DIC: dissolved inorganic carbon. Fig. 1 Map showing the study area and its geological background. a Tectonic setting of the Japanese archipelago with locations of active volcanoes (triangle). Red and blue lines show the depth contours of Pacific and Philippine Sea subducting plates, respectively (Nakajima and Hasegawa 2007;Hirose et al. 2008). Blue dotted line indicates extension of deep subduction of Philippine Sea slab (Zhao et al. 2021). b Tectonic structure of the Kyushu Islands with some important localities explained in the main text. The Beppu-Shimabara and Kagoshima grabens are after Matsumoto (1979) and Hayasaka (1987). Yellow hatch indicates an area where subsurface high conductivity blocks distribute in and beneath lower crust which were depicted from recent electrical resistivity survey (Hata et al. 2017). c The A-A' cross section in panel b showing low seismic velocity zones in the mantle wedge and crust (Zhao et al. 2018). d Sampling locations shown on topographical map with locality name and major fault systems. e Simplified geological map modified after Moreno et al. (2016). Geographical distribution of three types of fluids, High-Cl -, Low-concentration and High-HCO3fluids, are also shown.  for High-Clfluid, Lowconcentration, and High-HCO3fluids, respectively, with comparison of those for surface waters. Isotopic compositional field for hypothesized magmatic water is after Giggenbach (1992). Local meteoric water line in Kumamoto (ca. 20 km northeast from Uto district) are after Okumura et al. (2018).

Fig. 4
Biplot diagram between Ca 2+ ratio and chloro-alkaline indices (CAI-1 and CAI-2) for three types of groundwaters from hot spring sites, High-Clfluid, Low-concentration, and High-HCO3fluids. Ca 2+ ratio denotes proportion of calcium ion among four major cations (Na + , K + , Ca 2+ and Mg 2+ ) in milliequivalent concentrations.  Rees et al. (1978). Fig. 6 Identification for deep crustal fluids. a HCO3concentration versus δ 13 C diagram for all collected water samples. Simple two components mixing curved lines with relative contribution rates from mantle sourced carbon in 10% interval are shown in the graph. Mixing of two sets of end-components, soil carbon (HCO3and δ 13 C: 20 mg/l and -30‰, 160 mg/l and -30‰, respectively) and mantle sourced carbon (HCO3and δ 13 C: 4,000 mg/l and -4.5‰, 4,000 mg/l and -1.5‰, respectively), were hypothesized taking compositional variations observed from this study, with a same method applied elsewhere (e.g., Chiodini et al. 2000;Yamada et al. 2011;Rive et al. 2013 Table S1 Location, property, and measurement and analytical results for all water samples collected in this study from western coast of Kumamoto in western Kyushu, Japan.

Fig. S1
Clconcentration versus δ 2 H diagram for High-Clfluids. Simple two components (surface water and modern sea water) mixing lines with relative contribution rates from modern sea water in 10% interval are shown. End-component for surface water with variable compositions (Cland δ 2 H: 13.8 mg/l and -52.6‰ ~ 17.2 mg/l and -37.3‰) were considered.