Geochemical monitoring of deionized seawater injected underground during construction of an LPG rock cavern in Namikata, Japan, for the safety water curtain system

An underground liquified petroleum gas (LPG) storage facility was constructed between 2003 and 2013 in Namikata, Imabari City, Ehime Prefecture, Japan, to increase domestic LPG stockpiles. The most important issue during construction and operation of this facility is gas leakage prevention. To thwart water leakage, the water curtain system was constructed according to design standards, and a large amount of deionized seawater (seal water) was continuously injected into the rock mass around the cavern to keep the water level constant during both construction and operation. It is possible to distinguish three end member waters (existing groundwater, seawater or fossil seawater, and seal water) using the salinity and isotope (δ18O) difference because seal water injected underground has almost the same δ18O value as seawater. In this study, continuous observation is carried out using the geochemical techniques for flow analysis with a mixing proportion of three end-members in the initial construction period (April 2005 to March 2006) of the LPG underground storage facility. It is determined that existing groundwater and fossil seawater originally distributed in this region are partly replaced by seal water in the cavern.


Introduction
Liquified petroleum gas (LPG) is a gaseous fuel that is mainly composed of butane and propane extracted from byproduct gases in oil fields, natural gas fields, and oil refineries. It can be easily liquefied at room temperature by simple compressors and cooling. It is transported, stored, and delivered in liquid form. LPG is among the most important energy resources for social and economic prosperity that is widely used as a power source for homes and automobiles (Agency for Natural Resources and Energy 2017). In addition to storage in aboveground tanks, many methods of storing LPG in underground facilities have been implemented in countries other than Japan. To store LPG safely, the groundwater level and the chemical and isotopic composition must be monitored (Nenoen and Blindheim 1989;Hamberger 1991;Kim et al. 2000;Lee and Cho 2008;Yamamoto and Pruess 2004;Eric et al. 2005;Park et al. 2005;Raghavan et al. 2007;Li et al. 2009;Lim et al. 2013;Saikat and Kannan 2015;Lin et al. 2016). Lee et al. (2007) calculated the mixing proportion of seawater and groundwater based on the chlorine (Cl) concentration and oxygen isotopic composition (δ 18 O) of tunnel spring water. Using tritium and helium isotopes, they analyzed groundwater around an LPG stockpiling location in a coastal area of Korea. Lim et al. (2013) analyzed the flow of groundwater in oil storage depots for which seawater is injected as seal water into the ground. The main components and hydrogen and oxygen isotopic compositions were analyzed to determine the mixing ratio of injected seawater to existing groundwater.
To increase domestic LPG stockpiles, the Japanese government started construction of two underground LPG stockpiling locations in Namikata, Imabari City, Ehime Prefecture, and Kurashiki City, Okayama Prefecture, in 2003 (Fig. 1a). Namikata base construction was started by the National LPG Stockpiling Company, which was established in December 1998 to promote the construction. In 2004, the Japan Oil, Gas and Metals National Corporation (JOGMEC) took over the construction operations through a transition of the national stockpiling system. JOGMEC continued construction work on behalf of the national government and completed the project in March 2013. The Namikata base, like the Kurashiki base, is Japan's first underground LPG storage facility that uses a water-sealed rock mass tank. Approximately 450,000 tons of LPG are stored at the base, making it the world's largest LPG storage facility (Otake 2000;Okazaki et al. 2014;Kurose et al. 2014;JOGMEC 2000JOGMEC , 2020. Among the important safety measures in the construction of underground LPG storage facilities is preventing LPG leakage into the ground. Because the annual precipitation (1219 mm; Japan Meteorological Agency 2020) in Ehime Prefecture is less than that in other areas in Japan, a large amount of seawater (approx. 1500 L/min) is deionized (Cl concentration < 220 mg/L) and continuously injected as seal water into the rock mass around the cavern from the water gallery; this is intended to keep the groundwater level constant during the period of tunnel construction for the storage facility (Fig. 1b). It is important to maintain a safe water-sealing system with injected seal water, and methods such as water-level observation using observation pits that continuously monitor the groundwater table are implemented before and during construction. Monitoring of the injected seal water's underground flow and changes in water quality because of mixing with existing groundwater and seawater is necessary to understand the flow of water around the cavern. The chemical composition of groundwater originally distributed in the Namikata area is similar to that of seal water, with a low salt concentration, and existing groundwater and seal water cannot be distinguished from each other by chemical composition. On the tank's north side, seawater gushes into the tunnel.
The purposes of this study were to identify the contribution proportion of seal water, existing groundwater, and seawater in the tunnel and observation wells and to examine the changes in the groundwater environment as the construction progresses from the distribution of these three end-member waters. This is based on the fact that the hydrogen and oxygen isotopic compositions of seal water show deuterium (δD) and δ 18 O values that are almost the same as those of seawater. In this study, the hydrogen and oxygen isotopic compositions and Cl concentration values of water samples were measured monthly during the initial construction period (April 2005to March 2006, and the main chemical components were analyzed to examine the groundwater behavior geochemically.

Geological overview of the Namikata area
The Namikata area is located on a peninsula in Imabari City, Ehime Prefecture, facing the Seto Inland Sea (Fig. 1a). The Namikata LPG terminal had already been operating as a ground facility. For underground storage of LPG, three caverns (maximum width of 26 m, height of 30 m, and length of 500 m) at elevations from − 180 to − 150 m above sea level (mASL), water galleries (− 120 mASL), and working tunnels were constructed. The area was largely separated into a Propane cavern area and a Butane/Propane cavern (Fig. 1b). A geological map of the study area is shown in Fig. 2 (modified from Otake 2000). The study area belongs to the Inner Zone of Southwest Japan, and granite formed in the Mesozoic's Cretaceous period is widely distributed. The main rock is Namikata granite, and Takanawa granodiorite is distributed near the northern coast. Takanawa granodiorite is inferred to be widely distributed under sea surface in northern areas, although its distribution over land areas is small. In the northwestern part of the land area, quartz porphyry dikes penetrate the granite, and it is estimated that they are continuously distributed in the northern sea area. Quaternary sediments cover these plutons. Namikata and Takanawa granodiorite exhibit a course, semi-granular structure. The constituent minerals are mainly composed of plagioclase, quartz, and biotite; trace amounts of K-feldspar and amphibole are included; and zircon was observed in Takanawa granodiorite (this study). Plagioclase, which constitutes Namikata granite and Takanawa granodiorite are relatively homogeneous feldspar rich in andesine (Na; this study). Namikata granite is classified as a weathered part, a weathered-fresh transition part, and a fresh part, according to the degree of weathering. Fresh top elevations range from about + 20 mASL (ridge part) to about − 30 to − 40 mASL (coastal area) and tend to be deeper toward the coast. The fresh rock mass below this depth is composed of an almosthomogeneous rock mass with low permeability, except for some sections. Takanawa granodiorite is distributed from the cavern's northern part to the adjacent site, and it is seen as bedrock similar to Namikata granite. According to outcrop observation, the quartz porphyry has more cracks than both granitic rocks and is estimated to be highly permeable. Crack systems in the Namikata area tend to be relatively steep and run predominately in north-northwest to north-northeast and northwest to east-northeast directions (Otake 2000).

Sampling and analytical procedures
The sampling points for water samples were spring water points in the working tunnel, the water gallery, the cavern, and the tunnel connecting them (Fig. 3). Water samples from the water-gallery facilities on the ground, the surrounding streams, and observation wells were also collected (Table 1,  supplement data table). In this study period, analysis was advanced for the sample water for 1 year, from April 2005 to March 2006. In this period, excavation was completed for the working tunnel and the water gallery. In the cavern, the 30-m height was excavated in five stages, and excavation of the arch section (7.5 m from the top) was in progress.
Spring water in the tunnel was collected monthly in 100-mL polyethylene bottles for isotope analysis (Fig. 3). For major chemical component analysis, samples were collected monthly from April to October 2005, for 7 months, and once every 2 months from November to March 2006 in polyethylene bottles (indicated in red in Fig. 3). Seawater was sampled on May 25 and August 10, 2005. Groundwater from well 15 near the study area was sampled five times between May 2005 and January 2006. At these sampling times, water was pumped upward using an electric pump and sampled from a faucet. Stream water in the study area was sampled on July 12, 2005. Measurements at the time of sampling were water temperature, flow rate, pH, electric conductivity (EC), and oxidation-reduction potential (ORP). The pH, EC, ORP, and water temperature were measured using a portable pH/EC meter (WM-32EP; DKK-TOA) and an ORP meter (RM-12P; DKK-TOA). The flow rate was measured using a measuring cylinder or beaker. For areas from which water could not be sampled by measuring cylinders or beakers because of flow down the wall, the approximate flow rate was determined by measuring part of the flow rate.
Hydrogen and oxygen isotopic compositions and main chemical components were analyzed at the Analysis Center of Mitsubishi Materials Techno Co. The oxygen isotopic composition (δ 18 O) of the water samples was measured by the H 2 O-CO 2 isotope exchange reaction method (Epstein and Mayeda 1953). The Optima Micromass mass spectrometer was used. The analytical precision was ± 0.1 ‰. The hydrogen isotopic composition (δD) was determined by reacting a water sample of about 3 µL with about 1 g of metallic zinc in a vacuum at 410 °C for 2-5 h to generate H 2 gas (Coleman et al. 1982) and then introducing the hydrogen gas into the mass spectrometer. The analytical precision was ± 1 ‰ Na and K were determined by flame atomic absorption spectrometry (Hitachi Z-8200). Calcium (Ca), magnesium (Mg), aluminum, silicon, boron, iron, and manganese were determined by inductively coupled plasma (ICP) emission spectrometry (7700cx ICP mass spectrometry; Agilent Technologies). Cl and sulfate (SO 4 ) were determined

Calculation method of the mixing proportion of three end-member waters
The δ 18 O value and Cl concentration values of the water samples in the tunnel and observation wells are shown in Fig. 4a. Most of these samples were plotted in the area surrounded by three points of seawater, seal water, and existing groundwater. Because there were significant differences in isotopic composition between seal water and groundwater in the study area, and because seawater and seal water have a similar isotopic composition but different salt concentrations (Cl concentration and EC value), these three endmembers can be distinguished (Fig. 4a). Assuming that the water samples in the study area were mixed with seal water, existing groundwater, and seawater (water of high salinity north or south) at an arbitrary proportion, the mixing proportion of these three end-members was determined from the δ 18 O value and Cl concentration value using the following formula.
where D, S, and G denote seal water, seawater, and existing groundwater and X denotes the proportion of each endmember. The settings for the three end-members are shown in Table 2. Only δ 18 O was used for the mixing proportion calculation, because as shown in Fig. 5, there was a 1:1 correlation between δ 18 O and δD values; thus, δ 18 O values, which are relatively easy to apply to isotopic composition analysis, were used. The seal water sample showing same Cl concentration but varies greatly in δ 18 O in Fig. 4a was collected in April 2005 during the early stages of tunnel construction. Since the desalination facility did not have sufficient processing capacity at that time, tap water was added after desalination of seawater and supplied as seal water. However, since it was only for a short period of time, it will not affect the analysis of this study. Considering the analytical error of δ 18 O and Cl concentration value, the error of the obtained mixing proportion is about ± 10%. (1)

Results and discussion
The major chemical components and the hydrogen and oxygen isotopic compositions of water samples in the Namikata area are shown in Table 1. The samples in which only the oxygen isotopic compositions and the Cl concentration values of spring water in the tunnel were analyzed are shown in the supplement data.

Hydrogen and oxygen isotopic compositions
The hydrogen and oxygen isotopic compositions of water samples in this study are shown in Fig. 4. The solid line in the figure shows a global meteoric water line (δD = 8δ 18 O + 10; Craig 1961;Dansgaard 1964). The water samples showed the composition near the straight-line connecting groundwater in the observation well and seawater. The δD and δ 18 O values of the collected seawater were − 7 ‰ and − 0.5 ‰, respectively, which were similar to those of Vienna Standard Mean Ocean Water (δD = δ 18 O = 0 ‰). Because this seawater came from the Seto Inland Sea, seawater in the bay was slightly mixed by surrounding river water or precipitation. Seal water was obtained by desalting the surrounding seawater using the osmotic membrane method, and the EC value was as low as 58 to 777 μS/cm. The δD and δ 18 O values were − 7 ‰ to − 8 ‰ and − 0.8 ‰, respectively, which were similar to those values for seawater ( Table 1). The seal water sample collected in April 2005 is plotted in the middle of the mixing line. Since the desalination facility did not have sufficient processing capacity during the early stages of tunnel construction as mentioned above, tap water was temporally added after desalination of seawater and supplied as seal water.
All groundwater and well water samples except for observation well No. 39 showed EC values ranging from 200 to 580 μS/cm, δD values ranging from − 54 to − 47‰, and δ 18 O values ranging from − 8.4 to − 7.4‰. These values were lower than those of seal water and seawater (Table 1). River water and groundwater in Ehime Prefecture had δD values of − 60 ‰ and − 42 ‰ and δ 18 O values of − 8.9 ‰ and − 7.1 ‰, respectively (Mizota and Kusakabe 1994). Groundwater in the study area showed similar values, so it was judged to reflect the average precipitation in this area. The δD and δ 18 O values of precipitation are known to vary widely from month to month, but groundwater in this area showed almost constant values, suggesting that recharge water is the results of well mixed system coinciding with mean weighted average composition of yearly precipitation and recharged directly to aquifer without passing an evaporation. The δD value of groundwater in well No. 39 was − 30 ‰ and the δ 18 O value was − 4.4 ‰; these are intermediate values between those of seal water/seawater and existing groundwater (Fig. 5). However, because the EC value was 1033 μS/cm and the pH value was extremely high (pH 11.6), it is highly likely that half of the seal water was mixed into existing groundwater and reacted with the cement injected around the tunnel to increase both the Ca concentration and the pH value. Seal water was used in borehole drilling for this well, and it is possible that it was not replaced by groundwater and instead remained in the mix; however, seal water may have infiltrated to this point. The proportion of seal water was estimated to be around 40%, with the remainder as existing groundwater. The δD value of groundwater in well 15 was − 50.6 ‰, which was the δD value of existing groundwater in this area.

Time variation of the mixing proportion
As an example of the time variation of the mixing proportion, the results at sample point B-6 of water gallery No. 1 in the Butane/Propane are shown in Fig. 4b. The horizontal axis shows the monthly values for the survey period from March 2005 to February 2006, and the vertical axis shows the percentage of each source water contained in the water sample. At this point, seawater was about 90%, and the remaining water was existing groundwater. However, from July 2005, as construction progressed, the impact of seal water began to appear, and it can be seen that the proportion of three end-members fluctuated by about 10% monthly. It can also be seen that although the contribution of seawater decreased relatively, the groundwater portion increased slightly. In the Namikata area, oxygen isotope and Cl concentration analysis for working tunnel, water gallery, carven, and surrounding groundwater and seal water was determined for about 400 samples at 103 sites (Table 1 and supplement data). Working tunnel (0 to − 125 mASL) At the working-tunnel level (Fig. 6a), several springs were found. Water at sample points T-6 (TD (tunnel distance) 62 m), T-7 (TD 270 m) in water supply tunnel and T-2 (TD 400 m), T-3 (TD 600 m), and T-9 (TD 507 m) in working tunnel were considered fresh seawater based on their chemical composition (Table 1). These sample points are located in the northeastern part of the water supply tunnel, where fresh seawater gushes out. At sample point T-4 (TD 717 m) and T-8 (TD 710 m), the contribution of groundwater was large, but seawater is mixed in, accounting for around 10%. These springs are transformed into Ca-rich seawater (Table 1), which distinguishes them from the northeastern regions where fresh seawater gushes out. At sample point T-5 (TD 948 m), which sample point up on the west side of the working tunnel, seal water replaced 90% in March and 100% in April. At point T-1 (TD 250 m), seawater accounts for 20%, groundwater for 70%, and seal water seal for 10%. From the chemical composition, fresh seawater contributes to this seawater component, and it is judged that it is spring water near the seawater/freshwater interface.

Water gallery (− 125 mASL)
At the water-gallery level (Fig. 6b), the source water mixing proportion greatly changed with time, depending on the gushing point. A major feature was the large contribution of seawater to the tunnel's north side. In the southern part, the water from sample points P-16 (TD 917 m) and P-15 (TD 937 m) from the investigation hole was judged to be a mixture of groundwater and Ca-rich fossil seawater (Table 1). Existing groundwater with 100% of the water sample in the tunnel could be identified only at the following four locations: sample point B-1 (TD 63 m) in the Butane/ Propane water gallery No. 1, sample point B-45 (TD 467 m) in Butane/Propane water gallery No. 2, sample point P-97 (TD 68 m) in Propane water gallery No. 2, and sample points P-66 (TD 17 m), P-105 (TD 93 m) and P-20 (TD 97 m) in Propane water gallery No. 3 (Fig. 6b). In addition, a range of mixed water was found in the horizontal tunnel (sample point P-58). Such water was distributed to the south and was considered to correspond to areas with high groundwater potential. As for seal water, it made up 100% of the contribution in the central zone in the water gallery, but in other places, it was mixed with water of another origin. The mixing proportion of each end-member in water gallery as of February 2006 is shown in Figs. 7a-c. The black circles in the figure represent measurement points, and the other sections are interpolated data. This figure also shows that seawater (or fossil seawater) was distributed to the north and south, as mentioned previously, and that more water gushed out on the Butane/Propane side. A considerable amount of groundwater flowed out not only to the south but also to the north.

Cavern (− 150 mASL)
At the cavern level, it was difficult to track long-term changes because many sites were unable to collect spring water due to the progress of excavation work, and also because the source site was changed. The inflow of water into Carven, compared to the water gallery tunnel, indicates that the seal water is flowing out predominantly, replacing existing groundwater and seawater (Fig. 6c). The mixing proportion of each end-member in carven as of February 2006 is shown in Figs. 7d-f. Seawater springs in the northeastern region account for less than 20% of spring water, but they still spring out. Similarly, the proportion of groundwater has not fluctuated significantly during the study period but has emerged in the southeast and northwest regions. shows the difference between April and December in the observation well's water level. If this value is negative, it means that the water level was slightly lower than in April. The results show that seawater increased in the south and north during the 8 months. This was considered the effect of increasing the pressure of seal water at the south side during this period. The amount of spring water originating  from seawater decreased in the north, which is in harmony with the proportion of seal water. Therefore, the amount of seawater decreased because of the increase of seal water.

Chemical composition
The Piper diagram (Piper 1944) in Fig. 9 shows water samples in the study area indicating that the water samples are mixture of seawater, seal water and groundwater. The relationship between each chemical component and the Cl concentration are shown in Fig. 10a. The broken line in Fig. 10a shows a simple mixing of seawater and seal water or groundwater. There was no significant chemical difference in water sample from the Butane/Propane and Propane areas. Na, Mg and SO 4 ions are plotted near this simple mixing line or show slightly lower values (Fig. 10a). K ion for most samples except for those in working tunnel were clearly lower than the line. In contrast, Ca and HCO 3 ions were quite higher than the mixing line. These results suggested that the saline spring waters in the tunnel are of current sea water in some places and the fossil sea water in other places. This fossil seawater was enriched in Ca components and other elements due to the reaction with the rocks before the start of construction.
To examine the presence of rock-water reaction after mixing, Fig. 10b shows the relationship between Ca/Cl and HCO 3 /Cl equivalent ratios and classifies the Cl concentration into two categories by 10 meq/L. The mixed water lower than 10 meq/L is a mixture of existing groundwater and seal water, while the higher one indicates a mixture of fossil seawater/seawater and groundwater/seal water. Two dashed lines in Fig. 10b are due to the reaction Eqs. (4) and (5) and indicates dissolution of carbonate minerals by water ± CO 2 .
As a result, no increase in Ca was observed in the mixed water with high Cl concentration, but most of the mixed water with low Cl concentration showed intermediate values between the two dashed lines, where the spring waters in tunnel are plotted along the line in Eq. (1) and the observation well waters are between (4) and (5). In particular, the Ca/Cl ratios of the spring waters with 100% existing groundwater (B-1, P-4, P-20, and P-105) are high, while those of the other spring waters are low due to mixing with seal water. This is interpreted to mean that the Ca concentration in the tunnel spring waters have increased due to the dissolution of carbonate minerals prior to mixing, which is then diluted by the seal water. In the spring water in the tunnel, which is 100% groundwater, Ca and HCO 3 concentrations increase due to the reaction between carbonate minerals and water. On the other hand, the groundwater in the observation wells is shallow groundwater and is always in contact with the atmosphere, so it is interpreted that CO 2 in air contributes to the reaction (Fig. 10b).
In saline spring waters (Cl > 10 meq/L) in tunnels, the HCO 3 /Cl ratio is almost constant even though the Ca/Cl ratio increases, and the Cl concentration changes due to mixing with existing groundwater or seal water (Fig. 10b). This suggests that these Ca are not derived solely from the dissolution of carbonate minerals. From Fig. 10a, Ca concentration in the fossil seawater before mixing is estimated to be about 400 meq/L. The Ca concentration in fossil seawater in other regions is for example about 160 meq/L in the seawater-groundwater mixture that reacted with the Green Tuff Formation of the Seikan Tunnel, Japan (Mizukami et al. 1977). In the Namikata area, the host rock is granite, which is expected to have reacted with fossil seawater for a long time. The reason for such a high Ca concentration is not only the dissolution of carbonate, but also the ion exchange with clay materials and plagioclase (albitization) as suggested before (e.g., Russell 1970;Mizukami et al. 1977;Davisson and Criss 1996;Labotka et al. 2015).
The Stiff diagram (Stiff 1951) of the water samples is shown in Fig. 11. In the northern part of a working tunnel, the water was characterized by its high salt concentration, and two water samples (T-7 and T-9) in working tunnel had a Cl concentration value almost equal that of current seawater (Fig. 10a). At sample point T-1, the spring water consisted of 10% seal water, nearly 70% groundwater, and the remainder was sea water (Fig. 6a), which was judged to be current sea water (Fig. 11). On the other hand, T-4 and T-7 are spring waters in which more than 90% of the ground water is mixed with about 10% of the sea water, but because of their Ca-rich composition, this sea water is highly likely to be fossil sea water. That is, sea water was altered to the Na·Ca − Cl type. Lee et al. (2007) reported that a major chemical composition analysis and hydrogen and oxygen isotopic compositions of groundwater at an LPG stockpiling base constructed in the coastal area of South Korea resulted in mixing of seawater and existing groundwater. Most spring waters in the tunnels were plotted far away from the 1:1 line of sea water and groundwater mixing line except bromide. They discussed using a principal component analysis and showed that cations in the ground water, such as Ca, Na, Mg, and K, are irregularly enriched or depleted by various hydrogeochemical reactions between host rock and the ground water along the flow path. In general, seawater that changes its water quality by reacting with rocks is called fossil seawater and has been reported worldwide (e.g., Meyers 1968;Capuano 1990Capuano , 1992. Fossil seawater has also been reported in many places in Japan (e.g., Sakai and Matsubaya, 1974;Mizukami et al. 1977;Ueda et al. 2010;Okano et al. 2020).
In water gallery, there are many types of spring waters were gushing out. As same as the working tunnel, there are saline spring waters in the northeastern part (Fig. 11). Water sample at B-15, 100% current sea water gushes out from the chemical composition, and at P-59, sea water was diluted with seal water (Fig. 6b). Other spring waters with high salt concentration are considered to be formed by the dilution of fossil seawater into groundwater or sea water at various rates. In the water gallery, 100% of the groundwater is gushed out at B-1 and P-20 and its water quality is Ca-HCO 3 type. Other spring waters (< 10 meq/L) are mixtures of the seal water and groundwater and show Na-Cl type (Fig. 11). At sites P-15 and P-16 of the south investigation tunnel in the water gallery, there is a mixture of groundwater and fossil sea water.
The chemical compositions of spring waters in carven were analyzed for 1 and 5 samples in the Propane No. 1 and No. 2 carven, respectively, and other two spring waters (P-69, B-43) from connecting tunnels nearby the Propane No. 2 and Buran/Propane carven were also analyzed ( Fig. 11). Spring waters (P-42, P-53, P-81) in the central part of the carven are dominant in the seal water with small amount of groundwater, and in the northern part (B-43, P-114, P-49) and point P-95 of the No. 1 Propane carven are judged that fossil sea water is diluted with seal water and groundwater. In the northwest part of the No. 2 Propane carven, water samples (P-114, P-115) are adjacent springs, and the former is a dilution of fossil seawater, while the latter is mostly seal water. Thus, it can be seen that the quality of spring water is different even in areas that are only a few meters away.
Groundwater in 13 observation wells was sampled in May 2005 (blue) and November 2005 (green) (Fig. 11). In wells 7, 10, 25, and 40, groundwater was analyzed at different depths. Most chemical components for groundwaters in the wells remained almost constant at different depth. Groundwater in the observation wells was weakly acidic to weakly alkaline with a pH of 6.0 to 8.8, and the EC value was 600 μS/cm or less (most samples contained approx. 300 μS/ cm). This is similar to the EC value of the water sample from sample point B-1 in the Butane/Propane water gallery (TD 63 m), which was considered to contain 100% existing groundwater (300 to 428 μS/cm) (Figs. 6b and 11). From the δD and δ 18 O values of water samples in the observation wells, groundwater in this area showed almost constant values, suggesting that recharge water is the results of well mixed system coinciding with mean weighted average composition of yearly precipitation (Fig. 5). The chemical composition of the water samples shows the Ca-HCO 3 type and the EC value is high in a part, but the relation with the geology has not been found. This is because the observation wells are distributed in the Namikata Granite, which occupies most of the study area, and it is assumed that the composition is rich in Ca and HCO 3 due to the reaction with carbonate minerals and clays in the weathering zone. These results indicated that seal water did not infiltrate into these wells. In well 39, water sample showed the intermediate values between seal water and groundwater and the proportion of seal water was estimated to be around 40% (Fig. 5). The pH value was extremely high (pH 11.6), but for this well, it was probably high because seal water was used during borehole drilling and because some of the depth was filled with cement.

Behavior of seal water injected into the LPG carven
In the Namikata LPG carven, a large amount (1500 L/min) of seal water (de-ionized sea water) had been continuously injected into ground. From the isotopic composition and main chemical composition of water samples gushing in the tunnels, three types of source water gushed out in the water gallery and cavern. The source water was sea water or Ca-rich seawater (fossil seawater), existing groundwater, and seal water. In the center part of the tunnel, only seal water gushed but in other areas, spring waters in the tunnels are mixture of three end-members. Saline water spouted in the north and south portions of the water gallery. Groundwater mainly distributes in the southwestern part in the tunnels because water potential is high due to high altitude on the surface (Fig. 6c). Seal water is infiltrating into the main part of the water galleys (Fig. 6b). The groundwater and sea water initially distributed in this area cannot be completely replaced by seal water. Also, sea water-groundwater interface has not been largely changed. The groundwater and fossil seawater are assumed to have already been in the ground for a long time and simply mixed with the seal water, and the rock-water reaction during this period is assumed to be negligible, if any.

Conclusion
During the construction of the underground LPG storage terminal at Namikata, a large amount of seal water, which is deionized seawater, was continuously injected through the water gallery to maintain a constant water level. During the initial stage of this construction, the δ 18 O value and Cl concentration of the tunnel spring water had been analyzed for 1 year from April 2005 to see how the injected seal water would displace the existing groundwater and seawater/fossil sea water distributed in the northern area, and studied the proportion of seal water, seawater/fossil seawater, and existing groundwater, as well as the main chemical components. As a result, it was found that the existing groundwater and seawater/fossil seawater before construction were not completely replaced by seal water, and the location of the boundary between seawater and seal water or existing groundwater did not change much. From this study, it was possible to find out the route by which the seal water flows underground as the construction progresses, and this information can be used as a study material for efficient water level maintenance.