Estimation of Groundwater and Spring Water Residence Times near the Coast of Fukushima, Japan

The massive Tohoku earthquake, colloquially known as “The 2011 Great East Japan Earthquake,” occurred off the Pacific coast of Japan on March 11, 2011. The coastal area of the Tohoku region was severely affected by the tsunami, and the tsunami also caused severe damage to the Fukushima Daiichi Nuclear Power Plant (FDNPP) releasing a large amount of radioactive material into the atmosphere and environment. Determining the residence time of groundwater is important for evaluating how long radioactive materials are present after nuclear accidents such as FDNPP and multiple methods are needed to account for mixing between old/young water that can occur. The apparent residence times of spring water, groundwater, and artesian well water samples obtained during 2014 to 2018 from the northern coastal area of Fukushima Prefecture were estimated using tritium (3H), chlorofluorocarbons (CFCs), and sulfur hexafluoride (SF6) concentrations. The lowest 3H concentrations were within the background range (1‐5 TU) and were observed in artesian wells in Shinch, Soma and Minamisoma. The highest 3H concentrations (∼8‐15 TU) were observed near the FDNPP, and were probably affected by the accident following the 2011 earthquake. The average apparent residence time of spring water/shallow groundwater was 29 years based on the SF6 concentration and exponential mixing model, whereas that of artesian well water was 62 years based on the CFC‐12 concentration and piston flow model. Considering the relatively short apparent residence time of spring water/shallow groundwater, it is important to conduct continuous observations to understand the influence of the FDNPP accident.


Introduction
The massive Tohoku earthquake, colloquially known as "The 2011 Great East Japan Earthquake," occurred off the Pacific coast of Japan on March 11, 2011. The coastal area of the Tohoku region was severely affected by the tsunami associated with the earthquake (Mori et al. 2012a). Especially within 500 m of the shoreline, the height of the tsunami exceeded 10 m (Sato et al. 2014). The tsunami also caused severe damage to the Fukushima Daiichi Nuclear Power Plant (FDNPP), which is operated by Tokyo Electric Power Company (TEPCO), releasing a large amount of radioactive material (e.g., cesium, strontium, and iodine) into the atmosphere and environment due to the meltdown of the core (Hirose 2012;Katata et al. 2012). In addition, groundwater salinization was subsequently observed in some shallow wells near the coast (Inui et  and studies have reported that groundwater may have been affected by radioactive contamination (Ohta et al. 2012;Xu et al. 2016;Shizuma et al. 2018). After the 2011 earthquake, water quality indicators (e.g., dissolved inorganic ions, trace elements, and stable isotopes of oxygen and hydrogen) were used to estimate the recharge areas of groundwater, spring water, and river water in the northern part of Fukushima Prefecture (Yabusaki 2019(Yabusaki , 2020. The tritium ( 3 H) concentration of groundwater in the southern part of Fukushima Prefecture was measured by Kashiwaya et al. (2017). However, there are no known studies on the residence time of groundwater near the coastal area of Fukushima Prefecture. Determining the residence time and flow of groundwater in and around coastal areas is very important for predicting the impacts of radioactive contamination. And the multiple methods are needed to account for mixing between old/young water. For example, the groundwater longer flow paths are understood by the stable isotopes of oxygen and hydrogen and SiO 2 concentration. The 3 H concentration is useful to recognize whether the radioactive contamination cause from FDNPP accident has influenced, and estimate the residence time if the influence of contamination does not occur. In this study, the concentrations of 3 H, chlorofluorocarbons (CFCs), and sulfur hexafluoride (SF 6 ) in groundwater and spring water were determined and used to estimate the groundwater residence time near the coast of Fukushima Prefecture.

Study Area Geography and Geology
Near the coastal area of the northern part of Fukushima Prefecture, plateaus (river terraces) and lowlands are intricate (Figure 1). Prior to the earthquake disaster, land use near the coastal area (lowland area) was predominantly agricultural, mainly consisting of paddy fields (Minami-soma City Education Committee, Fukushima Prefecture 2019). In this area, the tsunami that occurred after the 2011 Tohoku earthquake not only damaged large areas of paddy fields but also devastated infrastructure, including roads and water channels. Salinization of the groundwater in this region was expected as a result of sea water inundation and the subsequent vertical infiltration of sea water into coastal aquifers. The geologic cross section around the central part of Minamisoma City (cross section a-a in Figure 1A) are shown in Figure 2. And the partial geologic cross section near the downstream of the Ota River (as shown in Figure 1B) was indicated by Chuman (1983). The geology near the coastal area of Fukushima Prefecture is Cenozoic marine sediment (i.e., belonging to last 66 million years). It is separated from the Abukuma granite plateau by the Futuba Fault, which lies approximately 10 km from the coast and runs in an almost south-north direction from Watari Town in Miyagi Prefecture to Minamisoma City. On the Abukuma mountainous terrain, the granitic rocks, and limestone are distributed in some areas to  the northwest of Minamisoma City, in the western part of Soma City, and in Shinch Town. In regards to the vertical geological structure, there is a 20-m-thick sedimentary layer (Quaternary) above the basic layer (Ikude et al. 1989). The upper part of the sedimentary layer is composed of silt and sand layers, whereas the lower part is composed of a peat layer (Shizuma et al. 2018).

Aquifer
There is a multi-layered aquifer in the coastal area of Fukushima Prefecture. Considering the geological column in these areas (Sato and Taketani 2008;Taketani and Osozawa 2013), it has been estimated that shallow (unconfined) groundwater is located at an approximate depth of 0 to 10 m in a sand gravel layer, while deep (confined) groundwater is located at an approximate depth of 30 m or more in the tertiary system (Ministry of the Environment 2020); because there is an impermeable layer (e.g., clay or silt layer with humus) approximately 10 to 30 m deep near the coastal area in locally (Chuman 1983). For the several deep wells, which are located near the coastal area and beneath a confining layer, groundwater is discharging naturally. These wells were defined as "artesian well" in this manuscript.

Meteorological Conditions
The annual precipitation levels averaged from 1991 to 2020 in Soma, Haramachi, and Namie, where the study sites were located, were 1381, 1388, and 1540 mm, respectively. These are slightly lower than the average annual precipitation for Japan (∼1700 mm). Soma and Namie both have an average annual air temperature of 12.7 • C.
The coastal area of Fukushima Prefecture is relatively warm compared with the inland areas (JMA 2021).

Sampling and Chemical Analysis
Samples and in situ data were obtained from 57 monitoring sites (31 spring water, 13 groundwater, and 13 artesian well sites) between April 2014 and August 2018 near the coast of Fukushima Prefecture (Figure 1). Thirteen sites were sampled multiple times. Near the coastal area of Fukushima Prefecture (as this study area), the decontamination from radioactive contamination, replacing soil, and embankment construction, etc. have carried out from 2014 to 2018. However, the agricultural revival (e.g., rice farming) and livestock were only limited areas; therefore, it is assumed the influence of agriculture and livestock (e.g., influence of spraying the fertilizer or livestock wastewater) for the water quality were negligible. The groundwater samples, including those from artesian wells, were mainly collected from private wells. The water samples for analysis of inorganic ions, SiO 2 , δ 18 O, and δ 2 H were collected in 250 mL polypropylene bottles, followed by filtering through a syringe filter immediately. Table 1 provides information on the sampling sites.

Groundwater Dating
Residence time is defined as the amount of water in a reservoir divided by either the rate of water addition to the reservoir or the rate of loss from it. However, the calculation of age dates and residence time are obscured due to mixing of old and young water (Weissmann et al. 2002). Various methods can be used to estimate the residence time of groundwater, including groundwater modeling, examination of soil physical conditions, and radioactive isotope dating (Chesnaux and Allen 2007;Post et al. 2013;Chesnaux et al. 2018;Dong et al. 2021). Of these, radioisotope dating using 3 H, 14 C, or 36 Cl is one of the most effective techniques (Clark and Fritz 1997). The halflives of 3 H, 14 C, or 36 Cl are 12.32 years, 5730 years, and 301,000 years, respectively; thus, providing useful dating ranges of several decades for 3 H, 30,000 to 40,000 years for 14 C, and 60,000 to 1,000,000 years for 36 Cl. In general, groundwater that is used by humans in Japan has a relatively short residence time; therefore, the best dating method is that of 3 H. Following the periods of atmospheric testing of nuclear devices that began in 1952, the 3 H concentration of precipitation increased in a series of spikes and peaked during 1963 to 1964 before subsequently decreasing, except for some small increases due to French and Chinese tests in the late 1970s ( Figure 3).
Research in recent years has shown that CFCs and SF 6 are also useful parameters for estimating relatively short residence times of groundwater. Studies have shown that CFCs and SF 6 began to concentrate in the atmosphere from the middle of the last century, offering a convenient way of dating water with an age of up to ∼100 years (Darling et al. 2012;Chambers et al. 2019).
Chlorofluorocarbons are volatile synthetic compounds of carbon, chlorine, and fluorine that were produced commercially at the beginning of the 1930s for use in refrigeration. As nonflammable and noncorrosive compounds with very low toxicities, CFCs are ideal for a variety of industrial and refrigerant applications. For instance, CFC-12 (dichlorodifluoromethane, CF 2 Cl 2 ) and CFC-11 (trichlorofluoromethane, CFCl 3 ), which were produced in 1930 and 1936, respectively, have been widely used in air-conditioning and refrigeration, as blowing agents in foams for insulation and packing materials, as propellants in aerosol cans, and as solvents. Many other CFC compounds were produced later, most notably CFC-113 (trichlorotrifluoroethane, C 2 F 3 Cl 3 ) in 1944, which has been used in the electronics industry for manufacturing semiconductor chips, for vapor degreasing and cold immersion cleaning of microelectronic components, and as a solvent in surface cleaning procedures (IAEA 2006). The concentrations of CFCs increased up until the 1990s (Figure 4), when restrictions were imposed on their usage to protect the ozone layer following the approval of the Montreal Protocol (Chambers et al. 2019). Hence, CFCs have provided useful tools for tracing and dating water with an age corresponding to the post-1945 period (Busenberg and Plummer 1992;Plummer et al. 1998Plummer et al. , 2000. Sulfur hexafluoride is a trace atmospheric gas that is primarily of anthropogenic origin, but also occurs naturally in fluid inclusions within some minerals and igneous rocks, as well as some volcanic and igneous fluids. Unlike CFCs with steady or declining atmospheric mixing ratios (Busenberg and Plummer 1997   atmospheric concentration of SF 6 has increased continuously ( Figure 5); thus, SF 6 has considerable potential as a dating tool for groundwater with an age corresponding to the post-1990 period.

Analysis Method of the 3 H Concentration
The 3 H concentration of water samples was determined using a liquid scintillation detector system based on the β-decay of the 3 H isotope. First, water samples were filtered through a syringe filter, followed by atmospheric distillation at 100 • C. The 3 H concentration was extremely low; therefore, it was necessary to concentrate the samples for measurement. After distillation, the water samples were condensed using a 3 H condensation apparatus (TRIPURE XZ030, De Nora Permelec Ltd., Japan) that could condense 1000 mL of water to 50 mL in approximately 60 h. Subsequently, 10 mL of the concentrated sample was transferred to a clean scintillation counting vial with10 mL of Ultima Gold scintillation cocktail. All vials were then shaken for approximately 15 min before being stored at 4 • C for approximately 24 h. The same procedure was applied to blank and standard samples. The 3 H concentration of each sample was measured over 60 min using a liquid scintillation analyzer (Tri-Carb 3110 TR, Perkin Elmer, USA), and the measurement was repeated 10 times. The concentration factor of the condensation apparatus was considered in the calculations; therefore, the results indicate the 3 H concentration at the time of sample collection. In this study, the concentration unit of 3 H is the tritium unit (TU; 1 TU = 0.118 Bq/L).

Analysis Methods of CFCs and SF 6
The water samples for CFC analysis were collected in 125 mL glass bottles with a foil-lined cap, while those for SF 6 analysis were collected in 500 mL glass bottles. The bottles were filled with water in a stainless-steel container and capped with water. The concentrations of CFC-11, CFC-12, CFC-113, and SF 6 were measured with a gas chromatography-electron capture detector following the "purge and trap" cryogenic pre-concentration method (Warner and Weiss 1985;Busenberg and Plummer 2000). The concentrations of CFCs and SF 6 in spring water, groundwater, and artesian well water were converted into their equivalent air concentrations (EAC) corresponding to the time when the water was recharged, based on Henry's Law in the form: where EAC is the equivalent atmospheric concentration (pptv) of CFCs or SF 6 , X gw is the concentration of dissolved CFCs or SF 6 in groundwater (fmol/L), K H is Henry's Law constant for CFCs and SF 6 , P is the total atmospheric pressure, and P H 2 O is the water vapor pressure (Asai et al. 2011). The detection limits for the above-mentioned CFCs and SF 6 in water have been reported to be 0.01 pmol/L and 0.1 fmol/L, respectively (Gooddy et al. 2006), with precisions of 2% to 3% and 1% to 3%, respectively (Wanninkhof et al. 1991;Law et al. 1994). Therefore, even though the atmospheric mixing ratios are relatively small (currently approximately 510 pptv of CFC-12, and 11 pptv of SF 6 , see Figures 4 and 5), it is possible to determine dates from approximately 1960 to the present.
To estimate the residence time using the CFCs and SF 6 concentration, it is necessary to have data on the recharge altitude, salinity, excess air, and water temperature during the water recharge period. In this study, the average annual air temperature at the recharge elevation of groundwater was used as the water temperature during the water recharge period. This calculation was based on the recharge altitude which was estimated by the δ 2 H and δ 18 O, temperature lapse rate (−0.65 • C/100 m), and annual mean air temperature in Soma City (12.7 • C at 6 m a.s.l.). The recharge altitude was estimated from the δ 18 O and δ 2 H values using the isotopic altitude effect (altitude [m] = −370.7 × δ 18 O-2766.5, and altitude [m] = −56.1 × δ 2 H-2689.9: Yabusaki 2020). The salinity was estimated using the EC of groundwater or spring water. The correction by excess air is very important for estimating the residence time (Poulsen et al. 2020;Asai et al. 2021). In this study, value of excess air was set at 2.2 cc STP/kg based on Asai et al. (2021). In addition, for the air concentration of CFCs, the average of observed data in the Northern Hemisphere was used. For the air concentration of SF 6 , the average of observed data from the islands of Japan was used, which was the average of the Northern Hemisphere multiplied by 1.15.

Residence Time Calculation
It is necessary to use a groundwater flow model to estimate residence times based on the concentration of CFCs or SF 6 . Here, we discuss groundwater flow models by comparing the relationships between CFC-12 and CFC-11 concentrations and between CFC-12 and SF 6 concentrations. To quantify the distribution of ages in aquifers, several types of mathematical models have NGWA.org S. Yabusakki and K. Asai Groundwater 61, no. 3: 431-445 been developed during the past decades. The classical age modeling practice consists of using simple analytical models, commonly called lumped-parameter models, to interpret environmental tracer data, for example, lumped-parameter models, simple specific age distributions describing piston flow, exponential mixing, combined piston exponential mixing, or dispersive mixing (Kazemi et al. 2006). The piston flow model, which is an assuming nonmixing of old and new groundwater in the soil, is a simplified way of representing the water movement and available to the case that a groundwater flows through the specific aquifer, that is, this model usually well describes the flow in confined aquifers (Geyh and Mook 2000). However, the piston flow model assumes no mixing and a very simple flow path, it should be taken notice that there is unsuitable case for using the estimation of residence time. As the alternative methods for estimating of the residence time, following methods are regarded as useful.
The Dupuit-Forchheimer model, which is built upon it to develop the simplified forms for both the confined and unconfined aquifers flow, and is usable for studying groundwater flow into wells (Chesnaux et al. 2018;Okuyade et al. 2022). The Vogel method is simplified analytical model introduced by Vogel (1967), and it is available for estimating the groundwater recharge with age tracers (Dong et al. 2021). Post et al. (2013) used the three different approaches to compute the groundwater age in coastal aquifer, and which are referred to as the piston flow model (based on particle tracking), the direct age (based on the concept of age mass), and the tracer-based age (based on a simulation of either 3 H and 3 He, or 14 C decay).
In the case of exponential mixing model, which is based on the assumption of mixing of a theoretically infinite number of different old components, of which the proportions decrease exponentially with increasing age. The exponential mixing model is generally applicable to a groundwater flow near the surface, especially it is useful for the pumped shallow groundwater and particularly for spring water from fissured aquifers (Geyh and Mook 2000). Near the coastal area of this study sites, the upper part of the sedimentary layer is composed of silt and sand layers, which is relatively high permeability layer; therefore, it is considered that the exponential model is suitable for using the shallow groundwater and spring water. And the exponential mixing model is probably the most commonly applied type of the so-called lumped parameter models (Solomon et al. 2010).
In this study, we used the piston flow model and exponential mixing model. These models are widely used for determination of groundwater age based on the CFCs or SF 6 , especially in Japan (Asai and Tsujimura 2010; Sakakibara et al. 2017). As a using these models means that the old and young water mixes; therefore, in this study, the age of water was defined as "apparent residence time". The boundary conditions of these models were set based on the hydrogeological condition and tracer plot (e.g., CFC-12 vs. CFC-11, CFC-12 vs. SF 6 ).

Water Quality Characteristics
The water quality characteristics varied between the monitoring sites. The ion concentrations at each sampling site are shown in Table S1 and Stiff diagram are shown in Figure S1. The main endmembers were Ca-HCO 3 and Na-HCO 3 ; however, Na-Cl and Ca-SO 4 type waters were also observed. The sites where Na-Cl type water and high dissolved concentration was detected (sites 15, 16, 17, 18, and 41) were affected by the tsunami in March 2011; The main endmember of shallow groundwater and spring water was Ca-HCO 3 , whereas that of deep groundwater and artesian well water was Na-HCO 3 .
The stable isotopic ratios of oxygen (δ 18 O) and hydrogen (δ 2 H) at each site are shown in Table S1. In general, the δ 18 O and δ 2 H values of meteoric water decrease with increasing altitude, which is called the isotopic altitude effect (Clark and Fritz 1997). At the observation sites, the δ 18 O values of groundwater and spring water, which are located near the Abukuma Mountains, were relatively low (below −8.0‰). In contrast, the δ 18 O values were relatively high near the coastal area, and the isotopic ratios of spring water in this area were especially high across many sites; therefore, these sites may have been affected by evaporation (e.g., water affected by evaporation was added). Although there was no obvious trend between δ 18 O or δ 2 H and well depth, nearly all the sites where the δ 18 O values were below −8.0‰ had a well depth exceeding 20 m. In a previous study, the altitude effect in the northern part of the coastal area of Fukushima Prefecture was reported to be −0.26‰/100 m (Yabusaki 2020). The difference in the δ 18 O value between shallow and deep groundwater was approximately 0.4 to 1.0‰; therefore, the difference in recharge elevation was estimated to be approximately 150 to 400 m due to the altitude effect.

H Data
As it is easier to measure the 3 H concentration than the concentrations of CFCs or SF 6 , we investigated the characteristics of the 3 H data for all sites. As shown in Figure 3, the 3 H concentration data of spring water and groundwater were divided into groups corresponding to three areas (Shinch; Soma and Minamisoma; and Namie and Okuma). The 3 H concentration of spring water and groundwater ranged from 1.9 TU to 14.8 TU, and the 3 H concentration at all sites was above the detection limit (Table S2). For spring water and groundwater recharged before 1953 (i.e., before the nuclear weapons test in the atmosphere), 3 H would not be detected owing to radioactive decay. In general, the detection of 3 H in water samples indicates that the water was recharged after 1953. However, considering the influence of the FDNPP accident is need for interpreting of the 3 H data in this study site. The 3 H concentration of precipitation in Japan has been 1-5 TU since the late 1990s (Yabusaki et al. 2003;Saito et al. 2013; IAEA/WMO 2021); however, the 3 H concentration ranged from 8 TU to 15 TU at about half of the observation sites in this study (i.e., approximately three times the background value). In particular, the 3 H concentration of spring water and groundwater samples from Namie and Okuma, which are located near the FDNPP, was higher (8.6 to 14.8 TU) than that of samples from other areas. Tritium was included with the material released into the atmosphere and environment following the FDNPP accident in March 2011, and the 3 H concentration of precipitation in Tokyo increased after this accident (Matsumoto et al. 2013). However, the 3 H concentration of precipitation in Fukushima Prefecture from January to December in 2013 was approximately 0.8 to 0.9 Bq/L (6.8-7.6 TU), which is similar to the background level before the FDNPP accident (Yabusaki et al. 2015). Consequently, the cause of the 3 H concentration exceeding the background level (approximately 5 TU) can be considered to be the FDNPP accident, and not the nuclear test in the 1950s. Accordingly, it was not possible to estimate the residence time using the 3 H concentration of spring water and groundwater at this study site because of the influence of the FDNPP accident. Although it was not possible to measure the 3 H concentration of this study area when the FDNPP accident occurred, it is expected that the 3 H concentration of precipitation (i.e., source of spring water and groundwater) was very high at that time. The 3 H concentration of precipitation on March in 2011 was observed at Chiba Prefecture (in Chiba City), where approximately 240 km far from Fukushima Prefecture (in Minamisoma City), and the 3 H concentration was approximately 12 TU, it had been increased cause from FDNPP accident (Kano et al. 2021). In addition, the 3 H concentration of spring water and shallow groundwater in a lot of sites in Shinch, Soma, and Minamisoma also exceeded the background level; however, the 3 H concentration of artesian well water, and some sites of spring water and groundwater was within the range of the background level ( Figure 3). Therefore, the influence of the FDNPP accident differed between the shallow and deep groundwater systems, and 3 H data of artesian well water, and some sites of spring water and groundwater described above are can be used for estimating the residence time. Focusing on the sites where the 3 H concentration was within the range of background level, 3 H concentration ranged from 1.8 TU to 3.7 TU, it seems to reflect the difference of residence time.

Estimated Apparent Age Based on CFC and SF 6 Concentrations
In the study area, the concentrations of CFC-12, CFC-11, and CFC-113 exhibited almost the same trends between different types of water (Table S2); therefore, CFC-12 was used as a representative of CFCs. The concentrations of CFC-12 and SF 6 in spring water, groundwater, and artesian well water are shown in Figures. 4 and 5, respectively. And, the temporal variations in the concentrations of CFC-12 and SF 6 in air are also shown in these figures. The apparent age was obtained by comparing between the EAC and historical variation curve of CFCs or SF 6 in ambient air.
The EAC of CFCs and SF 6 are shown in Table S2. We could not estimate the residence time using CFC-12 at four spring water sites (sites 10, 12, 13, and 14) and one groundwater site (site 9) in Minamisoma as a result of CFC contamination from local industries. At the other 15 sites, the CFC-12 concentration tended to be relatively high in spring water but low in artesian well water. In particular, the CFC-12 concentration of artesian well water at all sampling sites in Shinch, Soma, and Minamisoma was <50 pptv, and the apparent recharge time was estimated to be between the 1950s and the 1960s (i.e., the apparent age was approximately 70 to 50 years, see Figure 4). Whereas, the CFC-12 concentration of spring water varied with the location. The CFC-12 concentration at two sites in Minamisoma (sites 30 and 52) was <13 pptv, and the apparent recharge time was estimated to 1950s (i.e., the apparent age was approximately 70 years); however, at two sites in Shinch (sites 3 and 4), and one site in Minamisoma (site 8), CFC-12 concentration was 340 to 423 pptv, the apparent recharge time was estimated late 1980s (i.e., the apparent age was approximately 35 years).
The SF 6 concentration at all sites was within the range of the air concentration; therefore, the apparent recharge time could be estimated at all sites ( Figure 5). The SF 6 concentration exhibited a clear trend of being relatively high in spring water and groundwater (2.32 to 6.80 pptv), and relatively low in artesian well water (0.47 to 3.08 pptv) (Table S2), which similar to the trend observe for the CFC-12 concentration. The apparent recharge time of artesian well water (except at one site) was estimated to be between the 1960s and 1970s (i.e., the apparent age was approximately 60 to 50 years), while that of spring water and groundwater was estimated to be between the late 1990s and late 2000s (i.e., the apparent age was approximately 30 to 20 years).

Comparison Between Water Apparent Residence Time Based on CFC-12 or SF 6 and Mixing Models
To consider whether this hypothesis is adequate to estimate the apparent residence time, EAC values in water samples were shown in the tracer plot of CFC-12 vs. CFC-11 ( Figure 6) and CFC-12 vs. SF 6 (Figure 7). These figures also showed the distribution of the recharge time obtained by the piston flow model or exponential mixing model.
Owing to the relatively limited number of spring water and groundwater samples, it was difficult to decipher which model best fitted the measured CFC-12 concentrations; however, they tended to plot closer to the curve of the exponential mixing model curve than the curve of the piston flow model. The SF 6 concentrations of spring water and groundwater also plotted near the curve of the exponential mixing model; therefore, it was suggested that spring water and groundwater in this study were formed by mixing of a number of different old NGWA.org S. Yabusakki and K. Asai Groundwater 61, no. 3: 431-445  component water, which recharged at different times. Almost all the spring water samples were obtained from the end of a small mountainous drainage basin or at the edge of a hilly area, while the groundwater samples (site 9) were from a shallow aquifer (well depth of 0.7 m). Therefore, it was assumed that spring water and groundwater were a mixture of various water types (i.e., a water which had a different apparent residence time) in the region of the soil surface, with recharge occurring in many places within the drainage basin. This result that the exponential mixing model is appropriate for estimating the apparent residence time of spring water and groundwater is not contradictory to the hypothesis mentioned above. Based on the SF 6 concentration and exponential mixing model, the estimated apparent residence time of spring water and shallow groundwater ranged from 14 to 47 years (average of 29 years). The concentrations of CFC-12 and SF 6 were relatively low in the artesian well water samples, which can be explained by the fact that the artesian well water in the study area originates from a confined aquifer with a clay layer. Although no clear trend was observed in the distribution of the SF 6 concentration of artesian well water, the CFC-12 concentrations plotted near the piston flow model curve; therefore, it is assumed the artesian well water was formed by the groundwater which recharged a certain year with no mixing of many different old groundwater. This result is consistent with the estimation considering the hydrological conditions as mentioned above. Based on the CFC-12 concentration and piston flow model, the estimated apparent residence time of artesian well water ranged from 32 to 71 years (average of 62 years).

Spatial Distributions of Estimated Apparent Residence
Times Based on CFC-12 and SF 6 Figure 8 displays a distribution map of the apparent residence times of spring water, groundwater, and artesian well water, which could be divided into three stages: 0 to 20 years, 20 to 40 years, and 40 to 75 years. As mentioned, the apparent residence time of artesian well water was determined using the piston flow model and CFC-12 concentration, while that of spring water and groundwater was determined using the exponential mixing model and SF 6 concentration owing to the industrial contamination of CFC-12. Spatially, the longest apparent residence time of artesian well water was observed near the coastal area in the southern part of Minamisoma and in Shinch.
Regarding the relationship between water types and apparent residence time, samples with a Ca-HCO 3 endmember (i.e., shallow groundwater and spring water; sites 3, 8, 12, and 13) had a relatively short apparent residence time, whereas samples with a Na-HCO 3 endmember (i.e., artesian well water; sites 6, 20, 26, 31, 36, and 37) had a relatively long apparent residence time ( Figure S1). These results suggest that the endmember types were generally consistent with the estimated apparent residence times.
As an important point, when apparent residence time is estimated it needs to use multiple residence time estimation methods and check to see if they are consistent. In this study, average elevation estimated by the δ 18 O and δ 2 H, SiO 2 concentration, and 3 H concentration were used in order to confirm the consistent of apparent residence time. Figure 9 shows the relationship between apparent residence time and the difference in elevation, which was taken as the average elevation of the recharge area (estimated from the δ 18 O and δ 2 H values using the isotopic altitude effect) minus the elevation of the sampling site. The difference in elevation can indicate the scale of the groundwater flow system. In Figure 9, the apparent residence time of spring water/groundwater, which had a relatively small difference in elevation, was short; therefore, the scale of the flow system of spring water and groundwater was probably relatively small. However, the apparent residence time of two site in spring water (sites 30 and 52) was relatively longer than that of other spring water; therefore, the groundwater flow system of these spring water was probably large. In contrast, the apparent residence time of artesian well water, which had a relatively large difference in elevation, was long; therefore, the scale of the groundwater flow system of artesian wells was also probably large.
The relationship between the SiO 2 concentration and apparent residence time of artesian well water, spring water, and groundwater is shown in Figure 10. Silicon (Si) is released to groundwater during rock weathering, particularly the chemical weathering of feldspar. Therefore, Si can be used as an indicator to evaluate the length of apparent residence time. We found that the apparent residence time of artesian well water with a high SiO 2 concentration tended to be relatively long, whereas the apparent residence time of spring water/groundwater except two sites (sites 30 and 52) with a low SiO 2 concentration tended to be short. Accordingly, it was considered that artesian well water was more evolved than spring water due to chemical weathering. For spring water with long apparent residence time (sites 30 and 52) was also probably evolved due to chemical weathering. Thus, it agreeing with the finding that the endmember type of most artesian well water was Na-HCO 3 (Yabusaki 2020). Figure 11 presents the relationship between the 3 H concentration and estimated apparent residence times based on the CFC-12 concentration of artesian well water and the SF 6 concentration of spring water/groundwater. As mentioned in Section "Results-3 H data," the 3 H concentration of some groundwater samples, especially shallow groundwater samples, obtained near the coastal area exceeded the upper background level of 5 TU. This was also the case for most spring water samples; therefore, these spring water and shallow groundwater sites were probably affected by the FDNPP accident. This indicates that the 3 H concentration of spring water/shallow groundwater in these sites was not applicable to estimate the apparent residence time. In contrast, the 3 H concentration of precipitation in Tokyo only increased in March 2011 (Matsumoto et al. 2013), and then decreased to almost natural levels after April 2011. Therefore, it is expected that the 3 H concentration of spring water and groundwater in the study area will decrease gradually in the near future through radioactive decay or by mixing recent precipitation having low 3 H concentration because the influence of FDNPP accident for groundwater recharge was limited. Assuming the 3 H concentration of water decreases only owing to radioactive decay, it will be necessary approximately 20 years for the highest concentration of 3 H in water samples (=14.8 TU) to decrease below the background level (<5 TU). The 3 H concentration of artesian well water was <3.5 TU; therefore, these wells were probably not affected by the FDNPP accident. This understanding is consistent with the estimated apparent residence time of artesian well water, which ranged from 50 to 70 years (Figure 11), indicating that recent precipitation had not mixed with artesian groundwater under piston flow. Although the water recharged in March 2011 with a relatively high 3 H concentration due to the FDNPP accident may enter deeper groundwater in the future, the 3 H concentration of artesian well water might not increase. This can be explained by the long apparent residence time and large scale of the groundwater flow system of artesian wells, whereby the 3 H concentration would reduce due to radioactive decay. Whereas, considering the relatively short apparent residence time of spring water and shallow groundwater estimated by the CFCs and SF 6 , even if the site whose 3 H concentration was low at the time of the sampling, the effect of the FDNPP accident (i.e., effect of the water which was recharged immediately after the FDNPP accident with relatively high radioactive materials concentration) may be evident in the future. Therefore, it is especially important to continuously monitor the water quality in the study area to understand the effect of the FDNPP accident. Additionally, the observation using the 3 H, CFCs, and SF 6 concentration of water such as described above could be useful to recognize the water apparent residence time or influence of radioactive contamination for other area of Japan where the topographical and geological conditions are similar.

Conclusions
In this study, the concentrations of 3 H, CFCs, and SF 6 were determined in spring water, groundwater, and artesian well water near the coastal area of northern Fukushima Prefecture. The apparent residence time of spring water and shallow groundwater was estimated using the SF 6 concentration and exponential mixing model, while that of artesian well water was estimated using the CFC-12 concentration and piston flow model. The apparent residence time of spring water/shallow groundwater ranged from 14 to 47 years (average of 29 years), whereas that of artesian well water ranged from 32 to 71 years (average of 62 years). Spatially, the apparent residence time of artesian well water was relatively long near the coastal area in Shinchi and in the southern part of Minamisoma. The endmember type of these artesian wells was dominantly Na-HCO 3 , and the SiO 2 concentration was higher than that of spring water and shallow groundwater, implying that artesian Figure 11. Relation between apparent residence time and tritium concentration. The apparent residence time of artesian well was estimated by CFC-12, and that of spring water was estimated by SF 6 . well water was more evolved than spring water/shallow groundwater due to chemical weathering.
The 3 H concentration of spring water in Okuma and Namie, which are near the FDNPP, was relatively high; therefore, these sites were likely influenced by the FDNPP accident in 2011. In these areas, 3 H concentration is useful for recognizing of presence/no presence the influence of the FDNPP accident. In contrast, the 3 H concentration of artesian well water was <3.5 TU, indicating that these wells were not affected by the FDNPP accident. In the future, it is expected that water that was recharged in March 2011 with a high 3 H concentration due to the FDNPP accident will be discharged into artesian well water; however, the 3 H concentration of artesian well water might not increase. This can be explained by the long apparent residence time and large scale of the groundwater flow system of artesian wells, whereby the 3 H concentration should reduce due to radioactive decay. However, the relatively short apparent residence time of spring water and shallow groundwater, even if the site whose 3 H concentration was low at the time of the sampling, means that the effects of the FDNPP accident may be evident in the future. Therefore, it is especially important to continuously monitor the water quality in the study area to understand the effect of the FDNPP accident. The results obtained from this study will contribute to the groundwater use for future in and around the coastal areas of Fukushima Prefecture. Table S1. Concentration of dissolved inorganic materials and stable isotope ratios of oxygen and hydrogen in water samples. Table S2. Concentration of CFCs and SF 6 in water samples. Figure S1. Stiff diagrams and residence time of water samples. The diagram was expressed in terms of milliequivalents per liter (meq/L). Scale of diagram of No.41 is one-tenth. Residence time are estimated by the CFCs and SF 6 .