Estimating the Groundwater Recharge Sources to Spring-Fed Lake Ezu, Kumamoto City, Japan from Hydrochemical Characteristics

Abstract

Kumamoto is a city in Japan that relies completely on groundwater for drinking water. Groundwater in the Kumamoto region divided into shallow and deep aquifers. Around Lake Ezu, where one of Kumamoto City’s largest tap-water source wells are located, groundwater from both aquifers mixes, resulting in numerous springs. The aim of this study was to identify and quantify the relative contributions of the groundwater sources that discharge into Ezu Lake. River, lake, spring, and artesian well samples were collected every month between April 2021 and March 2022, and groundwater chemistry data for the shallow and deep aquifers were obtained from previous studies. The NO₃⁻ and SO₄²⁻ concentrations indicated three end-members: (A) high NO₃⁻ from anthropogenic sources, (B) high SO₄²⁻ from Shirakawa River water, and (C) low NO₃⁻ and SO₄²⁻ from denitrification or dilution. Mixing analysis show 60–70% from A, 17–22% from B, and 7–25% from C for the lake waters. Also, the result showed that springs in the Kami-Ezu area were dominated by shallow aquifer water, whereas artesian wells in the Shimo-Ezu area reflected deep aquifer water. This is the first time that the contributions of groundwater sources in this area have been quantified using a three-component mixing approach. Furthermore, it was estimated that Shirakawa River infiltration, including the artificial recharge project from rice paddy, contributed approximately 57% to groundwater discharge into Ezu Lake in 2020. These results provide new insights into the contribution of artificial recharge from agricultural land to groundwater.

1. Introduction

Kumamoto City relies on groundwater from the Kumamoto area, which has an abundance of groundwater, for all its tap water. There are two aquifers in the Kumamoto area: first unconfined aquifer and second confined aquifer. One of the main recharge areas for these aquifers is the midstream area of the Shirakawa River, which has a very high downward infiltration capacity [1]. In this area, there is no impermeable layer between the first and second aquifers, meaning that water from the Shirakawa River infiltrates directly into the first and second aquifers [2].

The abundant groundwater in both aquifers flows mainly in a northwest to southeast direction and then springs up in the Suizenji-Ezu Lake Spring Complex area of Kumamoto City [3], which nearby rivers also fed by springs located around the Ezu Lake area [4]. This reflects the presence of an incomplete impermeable layer around the Ezu Lake area, where groundwater from both aquifers is mixed and wells up [5], resulting in numerous springs, including one of Kumamoto City’s largest tap-water source wells. The groundwater discharge into Lake Ezu had been declining, but as a result of artificial recharge from rice paddies in the midstream area of the Shirakawa River since 2004, it has been suggested to have been increasing since 2006 [6].

Previously, researchers studied the contribution ratios from both the first and second aquifers in the Ezu Lake area using ²²²Rn in lake water and groundwater, and calculated that the monthly contribution ratios of the first and second aquifer were 1–68% and 17–82%, respectively [7]. However, it is doubtful whether there is much variation in the month-to-month mixing ratios of the two aquifers. Also, in another study, the contribution ratio from the second aquifer water was estimated from the nitrate (NO₃⁻) in the groundwater at around 75% [8]. Both studies divided the groundwater source into first and second aquifers but estimated different mixing ratios.

The first and second aquifers have different hydrochemical properties that vary from the Shirakawa River midstream area to Ezu Lake. In particular, the SO₄²⁻ concentration in the river water that permeates the recharge area in the midstream of the Shirakawa River are very high [2,9,10,11], so it has been used as a tracer to estimate the mixing ratio of Shirakawa River water in groundwater in the Kumamoto area [2,9]. Numerous other studies have reported that the first and second aquifers near agricultural areas are contaminated by NO₃⁻ [10,12,13], and have attributed the excessive NO₃⁻ mainly to chemical fertilizers and livestock farms. In contrast, denitrification has been generally confined to the first aquifer on the northern side of Kumamoto City [12,14,15,16], with dilution on the Kumamoto Plain caused by mixing with second aquifer from the outer rim of the Aso Caldera with low NO₃⁻ concentrations [12]. Temporal variations in the major dissolved ion concentrations in this area, including the SO₄²⁻ and NO₃⁻, show minimal variations between the current study dataset, which used samples acquired before the earthquake, and those obtained after the 2016 Kumamoto earthquake [17,18,19]. This indicates the SO₄²⁻ and NO₃⁻ stability along this study area flow paths and demonstrates its ability to be utilized as source tracers. However, it is still unclear whether the water in these two aquifers undergoes mixing processes either within the aquifers or with other sources, from the point where the water infiltrates from the Shirakawa River to when it is discharged in the Ezu Lake area. Hence, groundwater mixing analysis using both SO₄²⁻ and NO₃⁻ as the tracers was a highly suitable method for this study area.

The aims of this study were to identify the sources of first and second aquifer that discharging around Ezu Lake, which will be referred to Shallow and Deep groundwater in this study. Furthermore, this study aims to estimate the mixing ratios of the sources to the Shallow and Deep groundwater, quantify the contribution of the Shirakawa River water to groundwater, and evaluating the impact of artificial recharge from agricultural land around the midstream area of the Shirakawa River.

2. Materials and Methods

2.1. Study Area

The Kumamoto area is located on the western part of Kyushu Island, Japan (Figure 1) and is bounded to the east by Mt. Aso (1592 m a.s.l.), an active volcano with one of the largest calderas in Japan, and to the west by the Ariake Sea. An extension of the Shirakawa River from the Aso caldera flows in a westerly manner and flows into the Ariake Sea. The Kumamoto area is underlain by bedrock, developed from former Aso volcanic rock, that stretches from the west side of Mt. Aso to the Ariake Sea. There are Aso pyroclastic deposits from Mt. Aso’s past eruptions on top of the bedrock. The deep groundwater confined aquifer (second aquifer) is made up of deposits of lava rock, such as the Togawa lava, and Aso-1, Aso-2, and Aso-3 deposits, while the shallow groundwater unconfined aquifer (first aquifer) lies in the Aso-4 deposit [17,20]. The first and second aquifers are divided by low-permeability deposits, such as the Hanabusa and Futa lacustrine deposits. The groundwater in this area generally flows from northwest to southeast [5].

The study area around Ezu Lake, which is in the Suizenji-Ezu Lake Spring Complex, is shown in Figure 2a. This complex consists of the Kami-Ezu and Shimo-Ezu lakes. Kami-Ezu Lake has an area of 0.136 km² and maximum and average depths of 1.8 and 1.2 m, while Shimo-Ezu Lake is larger, with an area of approximately 0.35 km² and maximum and average depths of 5.0 and 2.0 m, respectively [21]. The area around Ezu Lake was a wetland approximately 7000 years ago [7]. The current northwest–southeast long shape of Ezu Lake may be attributed to an embankment that was built in the western area of the current lake as part of a water control project to prevent flooding 400 years ago, and a dam that impeded the flow of the spring water in a southwest direction [22].

2.2. Data and Methods

The sampling sites used in this study are shown in Figure 2a. Water samples, which included five spring-water samples (Kami-Ezu Lake area springs S1–S2 and Shimo-Ezu Lake area artesian wells Ar1–Ar3), four lake-water samples (L1–L4), and two river-water samples (R1–R2) from between the Kase River and Ezu Lake, were collected from 11 sampling points.

Water samples were collected monthly for a 1-year period, from April 2021 to March 2022. The samples were collected two or more days after rain occurred to minimize any effect of rainfall, especially on the lake water samples. The water samples were filtered through 0.2 µm cellulose acetate filters into 100 mL high-density polyethylene bottles without any headspace. The water temperature, electrical conductivity (EC), pH, oxidation-reduction potential (ORP), and dissolved oxygen (DO) were measured in situ using a portable field meter. A range of inorganic dissolved ions, sodium (Na⁺), potassium (K⁺), calcium (Ca²⁺), magnesium (Mg²⁺), chloride (Cl⁻), SO₄²⁻, and NO₃⁻, were determined using ion chromatography (Compact IC 761, Metrohm, Herisau, Switzerland) at the Hydrology Laboratory of Kumamoto University. HCO₃⁻ was measured by alkalinity titration using N/100 H₂SO₄ at a pH of 4.8 and bromocresol green-methyl red mixed indicator within one day of collection. The accuracy of the hydrochemical analyses was checked using a charge balance. Samples with relative errors greater than 5% were reanalyzed. The annual mean Stiff diagram of the samples are shown in Figure 2a.

Data for NO₃⁻ and SO₄²⁻ concentrations in first and second aquifer wells in the area, from the Shirakawa River midstream to Ezu Lake, were obtained from previous studies [12,14,17]. Major ion concentrations and their deviations are also presented in Figure S2. The locations of the wells that tap the first and second aquifers are shown in Figure 1. The contribution ratios of groundwater from different sources were evaluated using end-member mixing analysis (EMMA) [23]. The standard deviation of the sources is also calculated to estimate the standard deviation of the mixing ratio.

3. Results

3.1. In Situ Measurements

The pH, EC, ORP, DO, and water temperature of the samples are listed in Table S1. The pH, EC, ORP, and DO of the samples ranged from 6.98 to 7.64, from 242.3 to 287.0 μS/cm, from 200.4 to 265.1 mV, and from 7.81 to 9.94 mg/L, respectively. The water temperatures varied spatially and temporally and showed a seasonal pattern. The seasonal variations in the water temperatures of the spring and artesian groundwaters and the lake waters are presented in Figure 2b. The water temperatures in the Kami-Ezu Lake area were relatively stable throughout the year and ranged from 15.8 to 23.9 °C. In contrast, there were pronounced seasonal fluctuations in the water temperatures of Shimo-Ezu Lake, and the summer temperatures were substantially higher (around 28.1 °C) than those in the winter (about 14.1 °C), particularly at sampling point L4. There was a distinct seasonal pattern in the monthly air temperature measured at the Kumamoto Regional Meteorological Office throughout 2020 [24], with high temperatures recorded from June to September (from 24 to 28 °C) and low temperatures recorded from December to February (from 6 to 9 °C). These trends closely align with the seasonal water temperature variations observed in Shimo-Ezu Lake. The water temperature in Kami-Ezu Lake, however, remained stable, mainly because of the high groundwater discharge rate into the lake.

The results of a previous study of Ezu Lake [25] showed that, between 2020 and 2022, the groundwater discharge rate into Kami-Ezu Lake was significantly higher (exceeding 200,000 m³/day) than the discharge rate into Shimo-Ezu Lake (below 26,000 m³/day). Because of this continuous influx of groundwater to the Kami-Ezu Lake which has a smaller lake surface area, the temperature of the water is stable and is comparable to spring water. In contrast, Shimo-Ezu Lake has fewer spring-water sources and a larger lake surface area, resulting in more pronounced seasonal temperature fluctuations that reflect the air temperature variations.

3.2. Chemical Properties

The chemical properties of all the samples are listed in Table S1, and Stiff diagrams are shown in Figure 2a. All the samples belong to the Ca-HCO₃ type and are characteristic of young groundwater (around 16–36 years of residence time, estimated with ⁸⁵Kr [26]) with limited rock interaction. There were no significant differences between the Kami-Ezu spring water (S1–S2) and the Shimo-Ezu artesian groundwater (Ar1–Ar3), or between the Shokebori (R1) and Kengun (R2) river and spring waters (Figure 2a).

The Na, K, Mg, Ca, and HCO₃⁻ concentrations of the samples ranged from 13.23 to 15.31 mg/L, from 4.41 to 5.78 mg/L, from 8.46 to 10.30 mg/L, from 18.85 to 21.79 mg/L, and from 74.11 to 88.78 mg/L, respectively (Table S1). There were no significant differences between the cation and HCO₃⁻ concentrations in the different samples; however, the NO₃⁻ and SO₄²⁻ concentrations in the spring waters from the Kami-Ezu Lake area and artesian groundwater from the Shimo-Ezu Lake were noticeably different. The NO₃⁻ concentrations in the samples from the Kami-Ezu Lake and Shimo-Ezu Lake areas ranged from 17.47 to 21.10 mg/L and from 12.48 to 13.94 mg/L, respectively. However, the NO₃⁻ concentrations of samples L3 and L4 from the Shimo-Ezu Lake, at 16.28 to 17.31 mg/L, respectively, were similar to those from the Kami-Ezu Lake area. In contrast, the SO₄²⁻ concentrations in the Kami-Ezu Lake area ranged from 20.04 to 23.77 mg/L, and were generally lower than those in the Shimo-Ezu Lake area, where the concentrations of most of the samples were between 28.38 and 31.37 mg/L. Nevertheless, the similar pattern also found in the SO₄²⁻ concentrations of samples L3 and L4 from the Shimo-Ezu Lake, which values similar to the spring water from the Kami-Ezu Lake area. It is known that most of the major ions, except for NO₃⁻ and SO₄²⁻, have small variability, making them unsuitable to be used as tracers, as we cannot differentiate them (Figure S3). On the other hand, both NO₃⁻ and SO₄²⁻ show a clearer distinction between sample types, as shown in Figure S3, making it possible to deduce and estimate their sources.

3.3. Seasonal Variations in the SO₄²⁻ and NO₃⁻ Concentrations

The NO₃⁻ and SO₄²⁻ concentrations in the spring waters and water from the Kami-Ezu Lake and Shimo-Ezu Lake areas from April 2021 to March 2022 are shown in Figure 3a,b. The NO₃⁻ concentrations from the Kami-Ezu Lake area averaged about 20 mg/L and were higher than those in the artesian well samples from the Shimo-Ezu Lake area, which averaged about 13 mg/L (Figure 3a). The NO₃⁻ concentrations in the Shimo-Ezu Lake water (L3, L4) were within the range of the concentrations observed in the Kami-Ezu spring water and Shimo-Ezu artesian well water. The NO₃⁻ concentrations of samples L3 (April to June) and L4 (April to September) were similar to those in the artesian well samples from the Shimo-Ezu Lake area and varied widely between the warm and cold periods. The concentration variations may indicate NO₃⁻ reduction during organic processes in spring and summer in Shimo-Ezu Lake (e.g., consumed by plant plankton).

The SO₄²⁻ concentrations in the artesian wells in the Shimo-Ezu Lake area averaged about 30 mg/L and were higher than those in the springs in the Kami-Ezu Lake area, which averaged about 22 mg/L (Figure 3b). The SO₄²⁻ concentrations in the lake and river water samples in both lake areas were similar. The SO₄²⁻ concentrations between the springs at Kami-Ezu and the artesian wells at Shimo-Ezu were clearly different, but those in the lake water from Kami-Ezu and Shimo-Ezu were similar. This can be explained by the difference in the groundwater recharge in Kami-Ezu Lake and Shimo-Ezu Lake [25] that has already been discussed in the previous section about water temperature.

The NO₃⁻ and SO₄²⁻ concentrations were lower in July and August than in the other months and may reflect dilution by precipitation, as the 30-year mean monthly precipitation amounts for June and July (395.2 mm) were higher than the mean monthly amounts in other months (164.4 mm) [27]. There were seasonal variations in lake water temperature, NO₃⁻ and SO₄²⁻ were also noted in [7], and since the Lake Shimo-Ezu was not sampled, the differences in seasonal variation are small.

4. Discussion

4.1. Sources of Groundwater Discharge Around Ezu Lake

The NO₃⁻ concentrations were high, and the SO₄²⁻ concentrations were relatively low, in the Kami-Ezu Lake area (Figure 3a,b); the opposite was observed in the Shimo-Ezu Lake area, where the NO₃⁻ concentrations were low, but the SO₄²⁻ concentrations were high. Therefore, we used the NO₃⁻ and SO₄²⁻ concentrations to identify the sources of groundwater discharge at Ezu Lake.

In this study, first and second aquifer wells from previous study [14] were utilized. As shown in Figure 4, the red and blue with white-colored fills are the first and second aquifer wells, as described in [14]. The variability of major ion concentrations in aquifer wells (Figure S3) shows distinct differences between the first aquifer, second aquifer, and river water for its NO₃⁻ and SO₄²⁻ concentrations, making it suitable to act as a tracer. However, not all wells are selected as tracer wells. This are due to various reasons such as its spatial location deemed as far from the study area thus have little to no impact to the groundwater flows to Lake Ezu. The red, blue and purple shaded areas in Figure 4 represent the Shallow, Deep and mixed groundwater, respectively. These area classifications, especially for the Shallow and Deep groundwater, are explained further in later paragraphs with parameters explained in details in the Supplementary Materials.

The annual mean NO₃⁻ and SO₄²⁻ concentrations of all the samples from Ezu Lake are shown in Figure 4, and those of the first and second aquifer wells are shown in Table S2. We adopted three hypothetical end-members, namely high NO₃⁻ sources (A), high SO₄²⁻ sources (B), and dilution by low NO₃⁻ and SO₄²⁻ sources (C), which are the source wells that assumed to be the sources of the groundwater that flows and discharging at Lake Ezu. Further explanation can be found in the Supplementary Materials. All of the samples were within the triangle ABC. The springs in the Kami-Ezu Lake area and lake waters were relatively close to source A, while the artesian wells in the Shimo-Ezu Lake area were relatively close to source B. Similarly, the Shallow groundwater was at almost the same place as the springs in the Kami-Ezu Lake area. The Deep groundwater was close to the artesian wells in the Shimo-Ezu Lake area, but was closer to source B. To further confirm the sources (A, B, C), we examined the relationships between the NO₃⁻ and SO₄²⁻ concentrations and both the well depth (Figure 5) and the land use in the Kumamoto area (Figure 6). The mixed groundwater falls in between the Shallow and Deep groundwaters with wells located between them (Figure 6). Many wells were also relatively close to source C, which indicates dilution process by the groundwater with low NO₃⁻ and SO₄²⁻ concentrations.

The first aquifer well of k84 and second aquifer well of k109 were classified as source A. These wells have high NO₃⁻ concentrations (30–40 mg/L) and high well-bottom elevations (6 and 7 m a.s.l.) (Figure 5a) and are in agricultural and residential areas south of the Shirakawa River (Figure 6). The high NO₃⁻ in this area may be strongly influenced by anthropogenic activities, mainly agricultural fertilizer applications and livestock waste [10,12,13]. However, the second aquifer well of k3 was not included in source A. The exceptionally high NO₃⁻ concentrations in the k3 well may reflect localized anthropogenic contamination even though the well depth is deep (−45 m a.s.l.). Source B was characterized by extremely high SO₄²⁻ levels (>60 mg/L) and contained the Shirakawa River sites (R7, R9, and R10) and the second aquifer well of k21. This well had a bottom elevation of −51 m a.s.l. (Figure 5). Source C, with NO₃⁻ levels close to 0 mg/L and SO₄²⁻ concentrations between 0 and 4 mg/L (Figure 5a,b), was composed of the first aquifer well of k9 and second aquifer wells of k9 & k13, which was located north of Ezu Lake (Figure 6). This area was highlighted as a denitrification hotspot [19]. There was evidence of dilution in the second aquifer well of k17 and first aquifer well of k117 wells that were already mixed with low NO₃⁻ waters [12].

The average mixing ratios of the spring water at the Kami-Ezu Lake, artesian wells at the Shimo-Ezu Lake, and lake waters from the EMMA and its standard deviation are shown in

Table 1. (a) Annual mean mixing ratio results of the spring samples with their standard deviation (A, B, and C represent high NO₃⁻ source, high SO₄²⁻ source, and dilution by low NO₃⁻ and SO₄²⁻ source, respectively). (b) Annual mean mixing ratio results of the spring samples with their standard deviation, excluding July and August. (c) Annual mean mixing ratio results of the spring samples with their standard deviation, only July and August.

(a)
ID	Annual M.
		A			B			C
Springs and artesian wells
S1	57.4%	±	4.7%	17.2%	±	0.9%	25.4%	±	4.9%
S2	52.4%	±	4.5%	20.2%	±	1.1%	27.4%	±	5.4%
Ar1	37.3%	±	3.5%	33.5%	±	1.9%	29.2%	±	4.6%
Ar2	32.6%	±	4.1%	39.6%	±	1.7%	27.8%	±	4.0%
Ar3	33.7%	±	2.7%	39.0%	±	0.9%	27.3%	±	3.1%
Lake water
L1	59.8%	±	6.0%	20.7%	±	1.4%	19.5%	±	7.2%
L2	70.8%	±	3.1%	21.6%	±	1.6%	7.7%	±	3.8%
L3	65.5%	±	5.4%	21.9%	±	2.5%	12.6%	±	7.5%
L4	57.5%	±	3.8%	17.2%	±	2.8%	25.3%	±	6.2%
(b)
ID	Excluding July and August
		A			B			C
Springs and artesian wells
S1	59.1%	±	2.5%	17.4%	±	0.8%	23.5%	±	2.1%
S2	54.2%	±	1.8%	20.6%	±	0.9%	25.3%	±	2.3%
Ar1	38.6%	±	2.0%	34.0%	±	1.4%	27.4%	±	1.7%
Ar2	33.0%	±	4.3%	40.0%	±	1.4%	27.0%	±	3.6%
Ar3	34.5%	±	2.1%	39.2%	±	0.7%	26.3%	±	2.1%
Lake water
L1	61.7%	±	4.4%	21.2%	±	1.2%	19.0%	±	5.2%
L2	71.2%	±	3.3%	21.8%	±	1.6%	7.1%	±	3.9%
L3	64.1%	±	4.6%	21.7%	±	2.7%	14.2%	±	7.1%
L4	63.4%	±	3.8%	21.5%	±	2.8%	15.1%	±	6.1%
(c)
ID	Only July and August
		A			B			C
Springs and artesian wells
S1	48.5%	±	1.1%	16.5%	±	0.9%	35.1%	±	1.9%
S2	43.4%	±	0.8%	18.6%	±	0.4%	38.0%	±	1.2%
Ar1	30.8%	±	0.8%	30.8%	±	1.4%	38.4%	±	2.3%
Ar2	30.6%	±	2.2%	37.6%	±	2.2%	31.9%	±	4.3%
Ar3	29.9%	±	1.9%	37.9%	±	1.2%	32.2%	±	3.1%
Lake water
L1	50.2%	±	0.8%	18.7%	±	0.2%	31.1%	±	0.6%
L2	68.8%	±	1.6%	20.5%	±	0.0%	10.8%	±	1.6%
L3	72.1%	±	4.1%	23.3%	±	0.5%	4.5%	±	3.5%
L4	60.6%	±	2.2%	19.8%	±	1.4%	19.6%	±	3.7%

. The spring water and artesian well samples were divided into two datasets based on two different time periods, SJ (September to June) (Table 1b) and JA (July to August) (Table 1c). The mixing ratio based on the annual mean (Table 1a) shows that the spring water in the Kami-Ezu Lake area, represented by S1 and S2, was mainly from the high NO₃⁻ source (A), which accounted for between 52 and 57% of the contributions, followed by source C (25–27%) and source B (17–20%). Water from the Shimo-Ezu Lake area artesian wells (Ar1 to Ar3) was mostly from source B (high SO₄²⁻) (33–40%), followed by source A (32–37%) and source C (27–29%).

When samples from July and August were excluded (Table 1b), the contributions shifted somewhat, and the Kami-Ezu zone had a higher contribution from source A (54–59%) for the 10 months than for the whole year, followed by source C (23–25%) and source B (17–20%), while the sources for the Shimo-Ezu zone showed little change, with the highest contribution from source A (33–38%), followed by source B (34–40%) and source C (26–27%). For the July and August samples only (Table 1c), the highest contributor to the Kami-Ezu zone water was source A (43–48%), followed by source C (35–38%) and source B (16–18%). The source C contribution to the Shimo-Ezu zone, at 31–38%, was higher for July and August than for the whole year, followed by source B (30–37%) and source A (29–31%). The high contribution of source C (>10%) in July and August may reflect dilution during the high precipitation season, as previously discussed in Section 3.3.

The high standard deviation of the source C (±5.2%) further confirms that it is susceptible to seasonal changes, i.e., precipitation rate. Source A, which shows the second-highest standard deviation (±4.2%), confirms that it is mostly flowing near the surface, which also makes it easily affected by changes that may occur seasonally. Source B, on the other hand, demonstrates that it indeed mainly comes from the second aquifer and is mostly insusceptible to changes in the surface and seasonal variation as well, shown by a small standard deviation value (±1.6%) as shown in Table 1.

As with the spring and artesian well samples, the lake water’s annual average mixing ratios and its mixing ratios for two different periods were examined. The mixing ratios based on the annual mean (Table 1a) show that the lake water in the Kami-Ezu zone, represented by L1 and L2, was mainly from source A (60–70%), followed by source B (20–21%) and source C (between 7 and 19%). The lake water in the Shimo-Ezu zone, represented by L3 and L4, was mainly from source A (57–65%), with similar contributions from sources B (17–21%) and C (17–21%). When data from July and August were excluded (Table 1b), the results for both zones were similar to the annual results, with the Kami-Ezu zones still mainly fed by source A (61–71%), followed by source B (around 21%) and then source C (7–17%), while the Shimo-Ezu zone was mainly from source A (57–64%), with similar contributions from the B (16–21%) and C (14–26%) sources. For July and August only (Table 1c), the Kami-Ezu zone was still fed from source A (50–68%), but at a slightly lower percentage than for the whole year, followed by source C (10–31%) and source B (18–20%). The pattern for the Shimo-Ezu zone was different from that for the Kami-Ezu zone, with contributions of 60–72% from source A, 19–23% from source B, and 4–19% from source C. Based on the mixing ratio calculation, the mixing ratio of the lake water samples (L1–L4) shows similarity with those from Kami-Ezu Lake spring water samples (S1–S2). This shows that main contributor of the groundwater discharge in the Lake Ezu is the spring from Kami-Ezu Lake area, which shows good agreement with previous study [29].

4.2. Relationships with the Shallow and Deep Groundwaters

Study of mixing ratios of the groundwater discharge in Lake Ezu that were calculated in other studies [7,8] were mostly using a simple approach of two-component mixture, which is the first aquifer and second aquifer. In current study, it is believed that this approach might not be the most accurate one, since the first and second aquifer waters that discharging at Lake Ezu might already be mixed among itself before discharging into Lake Ezu. Hence, we coined the term Shallow and Deep groundwater to refer to the first and second groundwater that is specifically discharging into Lake Ezu. Therefore, study also attempted to estimates the mixing ratio from the Shallow and Deep groundwaters, both that discharging to the Lake Ezu and also its own mixing ratio with the three sources A, B and C. First and second aquifer wells data (Table S2) were utilized to observe the which of them are the representative tracer wells for Shallow and Deep groundwater.

The observed samples from the Ezu Lake area fell into three distinct groups (Figure 4). The first group, marked with red shading, comprised the first (k72, k73, k81, and k89) and second (k82, k95, and k103) aquifer wells, and had high NO₃⁻ and low SO₄²⁻ values, and were similar to the spring water and lake water from Kami-Ezu Lake. The well-bottom elevations were shallow (between −42 and 0.4 m a.s.l.) (Figure 5a), so the high NO₃⁻ water was likely from the land surface. The first (k72, k73, k81, and k89) aquifer wells were in residential areas near Kami-Ezu Lake, and the second (k82, k95, k103) aquifer wells were in agricultural areas far from Ezu Lake (Figure 6). These wells may represent the Shallow groundwater. The second group, marked with blue shading, had low NO₃⁻ and high SO₄²⁻ values. This group comprised the first (k80) and second (k11, k33, k46, k46, k48, and k94) aquifer wells that were near the Shimo-Ezu Lake artesian wells (Figure 4). The well-bottom elevations were both shallow and deep (between −205 and 4 m a.s.l.) (Figure 5b), and the high SO₄²⁻ values were from the Shirakawa River water, which infiltrates both the first and second aquifers in the midstream area of the Shirakawa River [14]. Most of these wells were in residential areas near Shimo-Ezu Lake, except the k33 well (Figure 6). These wells may represent the Deep groundwater. The third group, marked with purple shading, included the 1st (k45, k75) and 2nd (k30, k31, k32, k47, and k97) aquifer wells. This group had NO₃⁻ and SO₄²⁻ values between the spring and lake water at the Kami-Ezu Lake and the Shimo-Ezu Lake artesian wells (Figure 4). Like the Shallow groundwater, the well-bottom elevations were shallow (between −51.9 and −1.8 m a.s.l.) (Figure 6). Most of the wells were between the Shallow and Deep groundwater near Ezu Lake in residential areas, and the 2nd (k30, k31, and k97) aquifer wells were in agricultural areas far from Ezu Lake (Figure 6). These wells may represent a mix of the Shallow and Deep groundwaters. This mixed groundwater wells was not observed in the previous studies, which strengthen the assumption of the mixing process of both aquifers during the flow process before discharging into Lake Ezu.

The average mixing ratios of the Shallow and Deep groundwaters from the EMMA are shown in

Table 2. The three sources of the Shallow and Deep groundwater.

Groundwater Type	Sample Counts	Well-Bottom Elevation (masl)	SO42− (mg/L)	NO3− (mg/L)	Calculated Mixing Ratio
					A	B	C
Shallow	5	−28–0	21.0–23.0	17.0–20.0	50.4–53.3%	10.9–12.3%	35.8–38.7%
Deep	6	−205–4	29.0–36.0	9.0–12.0	39.8–44.2%	18.5–23.6%	34.4–39.9%

. Most of the 1st groundwater, ranging from 50.4% to 53.3%, was from the high NO₃⁻ source (A), while the contribution was around 10.9% to 12.3% from the high SO₄²⁻ source (B) and around 35.8% to 38.7% from the low NO₃⁻ and SO₄²⁻ source (C). The 2nd groundwater had a contribution of around 39.8% to 44.2% from source A, a contribution of around 18.5% to 23.6% from the high SO₄²⁻ source (B), and a contribution of about 34.4% to 39.9% from source C.

The NO₃⁻ and SO₄²⁻ concentrations of the Kami-Ezu lake area springs (S1 and S2) and the Shimo-Ezu lake area artesian wells (Ar1–Ar3) were similar to those of the Shallow groundwater wells and close to those of the Deep groundwater wells, respectively (Figure 4). This means that they were mainly from the first and second aquifers. Calculations that assumed a simple two-component mixture of the Shallow and Deep groundwaters components showed that the Shallow groundwater made up almost 100% of the supply to the springs in the Kami-Ezu lake area, the Shallow and Deep groundwaters made up about 20% and 80% of the supply to the artesian wells in the Shimo-Ezu Lake area, and the Shallow and Deep groundwaters accounted for around 85% and 15% of the lake waters, respectively. Previous studies mentioned that the amount of spring water from the second aquifer was estimated at almost 75% from the NO₃⁻ concentrations [8] and 58% from ²²²R [7]. Hence, the current study offers a novel interpretation and originality by considering not only the spring water of Lake Kami-Ezu and the artesian well spring of Lake Shimo-Ezu were mixtures of Shallow and Deep groundwater based on EMMA, but the Shallow and Deep groundwater itself also already mixed. However, it should be noted that the calculated mixing ratios can vary significantly depending on the tracer elements and the wells chosen to contribute to the Shallow and Deep groundwater.

4.3. Contribution of the Shirakawa River Water to Ezu Lake

In this study, we estimated the contribution of the Shirakawa River water from the EMMA. This allows us to estimate the volumetric contribution of the Shirakawa River water from the input in the midstream area to the output in the Ezu Lake area, which the first quantitative estimation attempt done in this area and replicable to other regions with similar settings. The groundwater recharge rate from the Shirakawa River midstream area in 2020 was estimated at about 70,000,000 m³/year and reflects agricultural activity, including artificial recharge [6]. Another study showed that the total groundwater discharge at Ezu Lake for 2020–2022 averaged 549,000 m³/day [25], which gave a rough average of around 200,000,000 m³/year. In this study, the contribution ratio of the Shirakawa River water (B) to the lake water (L) averaged about 20% (Table 1). So, the groundwater discharge to Ezu Lake that was sourced from the Shirakawa River was estimated at around 40,000,000 m³/year. This implies that 57% (40,000,000/70,000,000) of the Shirakawa River water that recharges in the Shirakawa River midstream area discharges to the Ezu Lake area.

It is believed that the remaining water of the Shirakawa River (43%) is transported through the groundwater system beneath Ezu Lake because there are many artesian wells on the western side of the Ezu Lake area. Furthermore, numerous wells from the Shirakawa River midstream area to the Ezu Lake also supply groundwater to Kumamoto City and local industries and agriculture. This estimation, however, has potential uncertainty as the dataset for the groundwater recharge estimation with the mixing ratio calculation results were not in the same period, which may differ. Future study with the same dataset between the two might be needed to confirm these findings.

5. Conclusions

The seasonal variations in the water temperature, NO₃⁻ and SO₄²⁻ concentration in lake water confirmed that Kami-Ezu Lake was the main contributor to the groundwater discharge in Ezu Lake. This result shows a good agreement with the estimated groundwater discharge results obtained through flow analysis [29]. By analyzing the NO₃⁻ and SO₄²⁻ concentrations in wells located between the midstream of the Shirakawa River and Ezu Lake, we divided the groundwater sources in this area into three end-members, namely a high NO₃⁻ source (A), a high SO₄²⁻ source (B), and a low NO₃⁻ and SO₄²⁻ source (C). These sources were characterized by their respective unique signatures, based on their chemical properties and spatial location. The A source was characterized by high NO₃⁻ concentrations (30–40 mg/L) and shallow well-bottom depths (6–7 m a.s.l.) and was situated in agricultural and residential areas south of the Shirakawa River. The high SO₄²⁻ source (B) consisted of Shirakawa River water and its intrusion to the second aquifer (well-bottom elevation of −51 m a.s.l.) and had SO₄²⁻ concentrations of more than 60 mg/L. The low NO₃⁻ and SO₄²⁻ source (C) had NO₃⁻ concentrations close to 0 mg/L and SO₄²⁻ concentrations from 0 to 4 mg/L, which probably reflect denitrification and dilution processes that take place along the flow zones.

Based on the EMMA of annual data, the springs in the Kami-Ezu Lake area were mainly from the high NO₃⁻ source (A) (52–57%) and the high SO₄²⁻ source (B) (17–20%), with dilution percentage from the low-concentration water (C) (25–27%). The artesian wells in the Shimo-Ezu Lake area were supplied by the B source (33–40%), followed by the A source (32–37%) and the C source (27–29%). The lake water in both the Kami-Ezu Lake and Shimo-Ezu Lake areas was mainly from the A source (60–70%), followed by the B source (17–22%) and the C source (7–25%). These results show that the water in both lakes was mainly from spring water from around the Kami-Ezu Lake area. In addition, the springs in Kami-Ezu Lake were almost the same as the groundwater in the first aquifer, and the artesian wells in Shimo-Ezu Lake were close to the groundwater in the second aquifer. Furthermore, analysis of the well-bottomed elevations and land use distribution showed that the well bottoms of the second aquifer well were limited in shallow wells (between −50 and 0 m a.s.l.), while the well bottoms of the second aquifer well were widespread (from −200 to 0 m a.s.l.). A comparison of the amount of water infiltrating into the midstream area of the Shirakawa River and the amount of water originating from the Shirakawa River that flows out of Ezu Lake showed that 57% of the discharge to Ezu Lake was from groundwater that originated in the Shirakawa River.

These results of the three-component mixing analysis represent the first quantitative findings for the area based on this approach and provide new insights into the groundwater sources, which are replicable for lakes that have similar characteristics of water sources and are spring-fed. Furthermore, the estimated contribution rate of the Shirakawa River can be used to quantitatively assess the impact of artificial recharge from midstream paddy fields and their effect on groundwater in this area in general.

The current study has several limitations that might be addressed and improved in future study. The first is that this study relies heavily on two ion tracers, NO₃⁻ and SO₄²⁻, which may not be able to be conservative in certain situations or conditions. This makes this study harder to replicate in other places, even if they have similar characteristics on the lake, which is spring-fed. Usually isotopic validation can be utilized to further confirm findings in this study and minimize its potential error in mixing ratio analysis, however the variability of the stable isotope of oxygen and hydrogen that usually utilized were notably small in this area, making it unsuitable to use in this study. Lastly, the mixing ratio calculation method used in this study, EMMA, might be improved by utilizing other methods that further employ not only hydrochemical and isotopic data, but also quantitative parameters such as groundwater level and discharge volume.