Spatial Characteristics of Groundwater Chemistry in Unzen, Nagasaki, Japan

: Nitrate pollution in groundwater is a serious problem in Shimabara Peninsula, Nagasaki, Japan. A better understanding of the hydrogeochemical evolution of groundwater in vulnerable aquifers is important for health and environment. In this study, groundwater samples were collected at 12 residential and 57 municipal water supply wells and springs in July and August 2018. Nitrate (NO 3 − N) concentration at eight sampling sites (12%) exceeded Japanese drinking water standard for NO 3 + NO 2 − N (10 mg L − 1 ). The highest nitrate concentration was 19.9 mg L − 1 . Polluted groundwater is distributed in northeastern, northwestern, and southwestern areas, where land is used for intensive agriculture. Correlation analysis suggests that nitrate sources are agricultural fertilizers and livestock waste. Dominant groundwater chemistry is (Ca+Mg) − HCO 3 or (Ca+Mg) − (SO 4 +NO 3 ) type. Groundwater with higher nitrate concentration is of (Ca+Mg) − (SO 4 +NO 3 ) type, indicating nitrate pollution affecting water chemistry. Principal component analysis extracted two important factors controlling water chemistry. The ﬁrst principal component explained dissolved ions through water–rock interaction and agricultural activities. The second principal component explained cation exchange and dominant agricultural effects from fertilizers. Hierarchical cluster analysis classiﬁed groundwater into four groups. One of these is related to the dissolution of major ions. The other three represent nitrate pollution.


Introduction
Shimabara Peninsula in Nagasaki Prefecture, Japan is composed of three municipalities: Shimabara City, Unzen City, and Minami-Shimabara City. The peninsula is an important agricultural and livestock region. The cultivated area occupies about 12,100 ha of the total 45,962 ha of the peninsula [1]. Agricultural production areas for potatoes, vegetables, industrial crops, and fruit farms represent about 62.7% of the cultivated area. Crops such as carrots, lettuce and Chinese cabbage are main crops cultivated throughout the year. In terms of livestock, the number of milk cattle and chicken (70% of all livestock) is about 7000 and 1,236,000, respectively [1]. The peninsula includes an important environment. Shimabara Peninsula was recognized as a UNESCO Global Geopark in 2009. The Unzen Volcano, located in the center of the peninsula, has been active during the last half a million years. Thus, the peninsula has been formed by continuous eruptions and mountain collapses. The central part, surrounded by faults, began sinking 300,000 years ago and the subsidence is ongoing. The subsidence zone is called Unzen Volcanic Graben, which has a width of about 20 km in an east-west direction and 10 km in a north-south direction [2]. Its subsidence rate is estimated at 2 mm/year. The graben stores large The geological structure of Shimabara Peninsula was reported by Murakami [3]. The city mainly lies on volcanic rock. An alluvial fan constituted of pyroclastic rocks with a gentle slope spreads from the northern parts of the city (Figures 2b and 3). The rock, which cover Kuchinotsu Layer of basement rock, is called Pre-Unzen volcanic rocks. Above an elevation of 300 m, Unzen volcanic rocks are predominant. The volcanic rocks are andesite with large phenocryst of amphibole. Joints and fissures are well developed. As mentioned above, the central part is a depression zone formed by the Unzen Volcanic Graben. The graben is partitioned by the Chijiwa and Kanahama faults. The old (300-150 ka) and new volcanos are overlapping and form the steep mountainous area. Pre-Unzen volcanic rocks are distributed on a limited area. As a result, most of the central part are covered with Unzen volcanic rocks (Figures 2b and 3). In the south, the Kuchinotsu Formation forms hills and plateaus. Tuff breccia/tuffaceous sand layer and pyroxene andesite lava (Pre-Unzen volcanic rocks) cover the Kuchinotsu Layer over a large area (Figures 2b and 3). The mineralogical data of volcanic rocks were summarized by Sugimoto [23] as follows. The Pre-Unzen volcanic rocks are composed of olivine basalt and two-pyroxene andesite lava flows and pyroclastic. The Unzen volcanic rocks are composed of hornblende andesite to dacite lava domes, lava flows, and pyroclastic. The rocks also include plagioclase, hornblende, and biotite.
Since the groundwater flow is considered interrupted by the faults, the city, and experimental area, was divided into three regions. The unconfined groundwater level contours were drawn using kriging with Surfer 13 (Golden Software) based on the natural water level of municipal waterwork wells (Figure 2b). Following the alluvial fan formation, the groundwater flows from the center to the coastline in the northern region. In the central region, the groundwater flow direction is from east to west. The groundwater level contours are parallel to the coastline and the administrative boundary in the southern region, indicating that the groundwater flows from south-east to north-west. Oxygen and hydrogen stable isotope ratios indicate that groundwater-derived spring and river water are recharged from 300 to 850, 450 to 900, and 300 to 850 m above mean sea level in the northern, central, and southern region, respectively [24]. The isotope studies further show that the groundwater flow systems in the central region, which is inside the rift The geological structure of Shimabara Peninsula was reported by Murakami [3]. The city mainly lies on volcanic rock. An alluvial fan constituted of pyroclastic rocks with a gentle slope spreads from the northern parts of the city (Figures 2b and 3). The rock, which cover Kuchinotsu Layer of basement rock, is called Pre-Unzen volcanic rocks. Above an elevation of 300 m, Unzen volcanic rocks are predominant. The volcanic rocks are andesite with large phenocryst of amphibole. Joints and fissures are well developed. As mentioned above, the central part is a depression zone formed by the Unzen Volcanic Graben. The graben is partitioned by the Chijiwa and Kanahama faults. The old (300-150 ka) and new volcanos are overlapping and form the steep mountainous area. Pre-Unzen volcanic rocks are distributed on a limited area. As a result, most of the central part are covered with Unzen volcanic rocks (Figures 2b and 3). In the south, the Kuchinotsu Formation forms hills and plateaus. Tuff breccia/tuffaceous sand layer and pyroxene andesite lava (Pre-Unzen volcanic rocks) cover the Kuchinotsu Layer over a large area (Figures 2b and 3). The mineralogical data of volcanic rocks were summarized by Sugimoto [23] as follows. The Pre-Unzen volcanic rocks are composed of olivine basalt and two-pyroxene andesite lava flows and pyroclastic. The Unzen volcanic rocks are composed of hornblende andesite to dacite lava domes, lava flows, and pyroclastic. The rocks also include plagioclase, hornblende, and biotite.
Since the groundwater flow is considered interrupted by the faults, the city, and experimental area, was divided into three regions. The unconfined groundwater level contours were drawn using kriging with Surfer 13 (Golden Software) based on the natural water level of municipal waterwork wells (Figure 2b). Following the alluvial fan formation, the groundwater flows from the center to the coastline in the northern region. In the central region, the groundwater flow direction is from east to west. The groundwater level contours are parallel to the coastline and the administrative boundary in the southern region, indicating that the groundwater flows from south-east to north-west. Oxygen and hydrogen stable isotope ratios indicate that groundwater-derived spring and river water are recharged from 300 to 850, 450 to 900, and 300 to 850 m above mean sea level in the northern, central, and southern region, respectively [24]. The isotope studies further show that the groundwater flow systems in the central region, which is inside the rift valley, and the north-south region, which is outside the rift valley, are independent and that the shallow groundwater in the central region is well mixed during flow.

Sampling and Analysis
Groundwater samples were collected from 12 residential wells and 57 municipal water supply wells and springs ( Figure 1) in July and August 2018. In Japan, wells are generally divided into deep and shallow depending on if the depth is smaller or larger than 30 m. Except for MW39 and MW41, all municipal wells have a depth of 50-150 m. Spring water was sampled with a collection pipe or cut off wall at MW39 and MW41. Of the residential wells, three wells (K14, K24, and K29) are shallow. Other wells are deep. All samples were collected in pre-washed bottles and water from taps connected to the wells. Before sampling, stagnant water in the pipes was removed. Major dissolved ions (Cl − , NO3 − , SO4 2− , Na + , K + , Mg 2+ , and Ca 2+ ) were measured by using ion chromatography (Metrohm 861 Advanced Compact IC and Eco IC). Titration using 0.1N HCl determined HCO3 − . Charge balance errors (CBE) were established for all groundwater samples according to [27]:

Sampling and Analysis
Groundwater samples were collected from 12 residential wells and 57 munic ter supply wells and springs ( Figure 1) in July and August 2018. In Japan, wells erally divided into deep and shallow depending on if the depth is smaller or larg 30 m. Except for MW39 and MW41, all municipal wells have a depth of 50-150 m water was sampled with a collection pipe or cut off wall at MW39 and MW41 residential wells, three wells (K14, K24, and K29) are shallow. Other wells are d samples were collected in pre-washed bottles and water from taps connected to th Before sampling, stagnant water in the pipes was removed. Major dissolved io NO3 − , SO4 2− , Na + , K + , Mg 2+ , and Ca 2+ ) were measured by using ion chromato (Metrohm 861 Advanced Compact IC and Eco IC). Titration using 0.1N HCl dete HCO3 − . Charge balance errors (CBE) were established for all groundwater samples ing to [27]:

Sampling and Analysis
Groundwater samples were collected from 12 residential wells and 57 municipal water supply wells and springs ( Figure 1) in July and August 2018. In Japan, wells are generally divided into deep and shallow depending on if the depth is smaller or larger than 30 m. Except for MW39 and MW41, all municipal wells have a depth of 50-150 m. Spring water was sampled with a collection pipe or cut off wall at MW39 and MW41. Of the residential wells, three wells (K14, K24, and K29) are shallow. Other wells are deep. All samples were collected in pre-washed bottles and water from taps connected to the wells. Before sampling, stagnant water in the pipes was removed. Major dissolved ions (Cl − , NO 3 − , SO 4 2− , Na + , K + , Mg 2+ , and Ca 2+ ) were measured by using ion chromatography (Metrohm 861 Advanced Compact IC and Eco IC). Titration using 0.1N HCl determined HCO 3 − . Charge balance errors (CBE) were established for all groundwater samples according to [27]: where all ion concentrations are expressed in mmol c L −1 . All CBEs were confirmed to be less than about 10% (0.2-8.3%). Hydrochemical parameters such as dissolved oxygen (DO), oxidant redox potential (ORP), electrical conductivity (EC), and pH were measured in-situ by using a handy measure instrument (HORIBA D-51, and D-54).
Multivariate statistical techniques were used to analyze the hydrogeochemical status of the groundwater in the study area [28]. Principal component analysis (PCA) identified factors affecting groundwater chemistry such as anthropogenic and natural processes (e.g., discharge of domestic sewage, application of chemical fertilizer, water-rock interaction, and evaporation). Hierarchical cluster analysis (HCA) can classify water types in addition to traditional analysis methods such as Piper and Stiff diagrams. These methods were applied to the groundwater samples in Unzen City to improve the understanding of nitrate pollution and joint hydrogeochemical processes. The concentrations of eight major dissolved ion concentrations (HCO 3 − , Cl − , NO 3 − , SO 4 2− , Na + , K + , Mg 2+ , and Ca 2+ ) were selected as input data. The principal components were extracted based on the Kaiser criterion to only retain components with eigenvalues greater than 1. The HCA was performed based on Ward's method. The statistical software JMP Pro 13 (SAS Institute Inc., Cary, NC, USA) was used for these multivariate analyses.

Water Chemistry
To improve the understanding how the Chijiwa and Kanahama faults affect the water chemistry in the experimental area, Hydrochemical components are summarized for the three fault separated regions as follows. In general, mean DO of about 8.0 mg L −1 indicates that groundwater is aerobic ( Table 1). The lowest DO is 3.2 mg L −1 ; however, this is still above 2.0 mg L −1 which is the limit for denitrification [29]. Thus, low nitrate levels are not caused by denitrification. In general, there is a small difference regarding pH. Mean pH represents moderate alkalinity (Table 1). A pH from weak acid to weak alkaline is noted in the north and central regions. Unzen City is famous for volcanic hot spring water. One of the hot springs (Obama Hot Spring) located close to MW39 displays weak alkalinity (pH 7.9) [30]. Other hot spring water (Unzen Hot Spring) close to MW41 and MW42 has a pH of less than 2.0 [31]. Two sampling sites (MW41 and MW42) display weak acidity (pH 6.9 and 6.6, respectively). Table 1. Descriptive statistics of hydrochemical components in the three regions. Minimum EC is about the same in all regions (Table 1). On the other hand, the highest EC is significantly lower in the central region than in the other two regions. This is probably a consequence of differences in water-rock interaction time [32]. All ORP values are positive, indicating that groundwater is in oxic conditions ( is often related to redox conditions [33]. In general, young groundwater is likely to be oxic, and older groundwater (e.g., >100-1,000,000 years) is more likely to be anoxic [33]. Thus, groundwater in the study area is relatively young. Minimum ORP of 46 mV was found at MW46. In addition, compared to the two other regions, mean ORP is relatively small in the northern region. According to groundwater levels, the groundwater flow is slower in the northern region. Therefore, groundwater residence time is expected to be longer in this region compared to the other two and as a result, the ORP is smaller.
The trilinear Piper diagram was applied to classify water types ( Figure 4). With respect to cations, groundwater samples are located close to the center in the cation triangle, indicating that cations have no specific characteristics. For anions, there is a low content of Cl, but HCO 3 and SO 4 +NO 3 contents vary. As a result, water samples are classified into Group I or Group III in the central key diagram, suggesting that the dominant water types are (Ca+Mg)−HCO 3 or (Ca+Mg)−(SO 4 +NO 3 ). Polluted water samples characteristically move from Group I to Group III. In general, groundwater chemistry in Unzen City is affected by nitrate pollution similar to Shimabara City [16].
S.D. 1 4. The trilinear Piper diagram was applied to classify water types (Figure 4). spect to cations, groundwater samples are located close to the center in the cation indicating that cations have no specific characteristics. For anions, there is a low of Cl, but HCO3 and SO4+NO3 contents vary. As a result, water samples are classi Group I or Group III in the central key diagram, suggesting that the dominant wat are (Ca+Mg)−HCO3 or (Ca+Mg)−(SO4+NO3). Polluted water samples character move from Group I to Group III. In general, groundwater chemistry in Unzen C fected by nitrate pollution similar to Shimabara City [16]. Evolution of groundwater chemistry is characterized by Stiff diagrams (Fi Along the groundwater flow from upstream to downstream in the northern region concentrations show an increasing trend, i.e., Na+K and HCO3 tend to increase. A evolution pattern with maximum concentrations of all ions is controlled by ion dis from soil and rock during groundwater flow [16]. Another pattern is with Na+K an when groundwater recharge occurs from paddy fields. Recharge, however, prob fected water chemistry because sampling was performed during the period when fields are flooded. In addition, cation exchange between Ca and Na+K and/or oth of mineral dissolution are considered to adjust water chemistry [16]. In the cen southern regions, the characteristics of water chemistry are similar to that in the n Evolution of groundwater chemistry is characterized by Stiff diagrams (Figure 5). Along the groundwater flow from upstream to downstream in the northern region, all ion concentrations show an increasing trend, i.e., Na+K and HCO 3 tend to increase. A typical evolution pattern with maximum concentrations of all ions is controlled by ion dissolution from soil and rock during groundwater flow [16]. Another pattern is with Na+K and HCO 3 when groundwater recharge occurs from paddy fields. Recharge, however, probably affected water chemistry because sampling was performed during the period when paddy fields are flooded. In addition, cation exchange between Ca and Na+K and/or other kinds of mineral dissolution are considered to adjust water chemistry [16]. In the central and southern regions, the characteristics of water chemistry are similar to that in the northern region, suggesting that the same factors as mentioned above control water chemistry. Near the coast, however, the ion concentrations in the central region tend to be small compared to that of the other regions. Since the EC is relatively low in the central region (Table 1), the water-rock interaction time is short. Although, the nitrate ion sector of the diagrams is noticeable only in the north and south regions, nitrate concentration in the north region tends to be higher than that in the north region. Thus, the characteristics of the three regions can be seen from nitrate and overall ion concentration. region, suggesting that the same factors as mentioned above control water chemistry. Near the coast, however, the ion concentrations in the central region tend to be small compared to that of the other regions. Since the EC is relatively low in the central region (Table  1), the water-rock interaction time is short. Although, the nitrate ion sector of the diagrams is noticeable only in the north and south regions, nitrate concentration in the north region tends to be higher than that in the north region. Thus, the characteristics of the three regions can be seen from nitrate and overall ion concentration.

Nitrate Pollution
Nitrate levels (NO3−N) in Unzen City range from 0.6 to 19.9 mg L −1 , with a mean of 5.2 mg L −1 . The highest concentration appeared in a residential well (Az659). Compared to Japanese drinking standard for NO3+NO2−N (10 mg L −1 ) 4 of 57 (7%) for municipal wells and 4 of 12 (33%) for residential wells, respectively, do not meet this standard. A threshold of 3 mg L −1 is generally considered to be polluted by human activities [34], and 25 of 57 (44%) for municipal wells and 11 of 12 (92%) for residential wells exceeded this criterion. This means that most of the groundwater in the study area can be considered as polluted by nitrate. Nitrate levels in groundwater decrease with increasing sampling depth [35]. Since the depth of the most residential wells are shallower than that of municipal wells, the residential wells showed higher concentrations. The nitrate polluted wells are concentrated to the northern and southern regions ( Figure 6). These locations coincide with the upland fields of Unzen City (Figure 2). This indicates that nitrate pollution in this area is associated with agricultural activities, especially crop production.
The correlation matrix of eight dissolved ion concentrations are summarized as follows (Table 2). NO3 − shows relatively high correlation with Cl − , SO4 2− , Mg 2+ , and Ca 2+ . Similar correlation pattern was reported for Shimabara City [16]. Positive correlation between NO3 − and Cl − (r = 0.62) indicates use of manure and livestock waste. Positive correlation (r = 0.51) between NO3 − and SO4 2− reflects application of chemical fertilizers such as (NH4)2SO4. Chemical fertilizers MgCO3 and CaCO3 are applied with (NH4)2SO4 [16]. Thus, NO3 − is positively correlated with Mg 2+ (r = 0.80) and Ca 2+ (r = 0.75), respectively. Applied calcareous material to modify soil acidification probably results in correlation between NO3 − and Ca 2+ . Nitrogen load from livestock waste in Unzen City was estimated to be 4452 kg day −1 in 2014 [1]. Total fertilizer application of nitrogen was estimated to 3740 kg day −1 Figure 5. Major ion variation using stiff diagrams.

Nitrate Pollution
Nitrate levels (NO 3 −N) in Unzen City range from 0.6 to 19.9 mg L −1 , with a mean of 5.2 mg L −1 . The highest concentration appeared in a residential well (Az659). Compared to Japanese drinking standard for NO 3 +NO 2 −N (10 mg L −1 ) 4 of 57 (7%) for municipal wells and 4 of 12 (33%) for residential wells, respectively, do not meet this standard. A threshold of 3 mg L −1 is generally considered to be polluted by human activities [34], and 25 of 57 (44%) for municipal wells and 11 of 12 (92%) for residential wells exceeded this criterion. This means that most of the groundwater in the study area can be considered as polluted by nitrate. Nitrate levels in groundwater decrease with increasing sampling depth [35]. Since the depth of the most residential wells are shallower than that of municipal wells, the residential wells showed higher concentrations. The nitrate polluted wells are concentrated to the northern and southern regions ( Figure 6). These locations coincide with the upland fields of Unzen City (Figure 2). This indicates that nitrate pollution in this area is associated with agricultural activities, especially crop production.
The correlation matrix of eight dissolved ion concentrations are summarized as follows ( Table 2). NO 3 − shows relatively high correlation with Cl − , SO 4 2− , Mg 2+ , and Ca 2+ . Similar correlation pattern was reported for Shimabara City [16]. Positive correlation between NO 3 − and Cl − (r = 0.62) indicates use of manure and livestock waste. Positive correlation (r = 0.51) between NO 3 − and SO 4 2− reflects application of chemical fertilizers such as (NH 4 ) 2 SO 4 . Chemical fertilizers MgCO 3 and CaCO 3 are applied with (NH 4 ) 2 SO 4 [16]. Thus, NO 3 − is positively correlated with Mg 2+ (r = 0.80) and Ca 2+ (r = 0.75), respectively. Applied calcareous material to modify soil acidification probably results in correlation between NO 3 − and Ca 2+ . Nitrogen load from livestock waste in Unzen City was estimated to be 4452 kg day −1 in 2014 [1]. Total fertilizer application of nitrogen was estimated to 3740 kg day −1 in 2013 [1]. Nitrogen load from domestic wastewater was estimated to 258 kg day −1 in 2014 [1]. Compared to livestock waste and fertilizers, domestic wastewater load contains much less nitrogen. The relative difference in pollution load from livestock waste and fertilizer can be seen in the observed SO 4 2− concentration. For water samples with NO 3  in 2013 [1]. Nitrogen load from domestic wastewater was estimated to 258 kg day −1 in 2014 [1]. Compared to livestock waste and fertilizers, domestic wastewater load contains much less nitrogen. The relative difference in pollution load from livestock waste and fertilizer can be seen in the observed SO4 2− concentration. For water samples with NO3−N concentrations exceeding the Japanese drinking water standard, SO4 2− concentrations are relatively small at MW9, MW32, and M12. SO4 2− concentrations ranged from 3.2 to 5.4 mg L −1 at these locations. On the other hand, SO4 2− concentrations at other locations (MW10, MW12, K35, Az659, and Ai29) are in the range between 18.6 and 30.6 mg L −1 . At locations with high NO3−N and SO4 2− concentrations, chemical fertilizers such as (NH4)2SO4 are supposed to be a dominant pollution source.  -------1.00

Multivariate Analysis
According to the PCA results (Figure 7), eight chemical variables were reduced to the significant two principal components based on the Kaiser criterion. Principal component 1 (PC1) with the eigenvalue of 5.23 explained 65.4% of the total variance. PC1 is clearly related to the effect of all ions. The dissolution of minerals contained in Pre-Unzen and Unzen volcanic rocks increase cations and HCO3 − concentrations [17,36]. Thus, the positive loadings for cations and HCO3 − indicate mineral dissolution through water-rock interaction. Influence from fertilizer, manure and livestock waste loading as noticed in the correlation analysis, is displayed in Cl − , NO3 − , SO4 2− , Mg 2+ , and Ca 2+ positive loadings. PC 2 has an eigenvalue of 0.96, accounting for 12.0% of the total variance. PC2 has a positive

Multivariate Analysis
According to the PCA results (Figure 7), eight chemical variables were reduced to the significant two principal components based on the Kaiser criterion. Principal component 1 (PC1) with the eigenvalue of 5.23 explained 65.4% of the total variance. PC1 is clearly related to the effect of all ions. The dissolution of minerals contained in Pre-Unzen and Unzen volcanic rocks increase cations and HCO 3 − concentrations [17,36]. Thus, the positive loadings for cations and HCO 3 − indicate mineral dissolution through water-rock interaction. Influence from fertilizer, manure and livestock waste loading as noticed in the correlation analysis, is displayed in Cl − , NO 3 − , SO 4 2− , Mg 2+ , and Ca 2+ positive loadings. PC 2 has an eigenvalue of 0.96, accounting for 12.0% of the total variance. PC2 has a positive loading for HCO 3 − and negative loading for NO 3 − implying denitrification. However, according to the DO content in groundwater, denitrification does not occur. Since HCO 3 − concentration decreases due to nitrification of NH 4 + [37], HCO 3 − shows higher concentration without nitrate pollution. PC2 has a negative loading for SO 4 2− too. Thus, PC2 is related to the effects of chemical fertilizer (NH 4 ) 2 SO 4 . In addition, positive loading for Ca 2+ and negative loading for Na + represent cation exchange [38].
loading for HCO3 − and negative loading for NO3 − implying denitrification. However, according to the DO content in groundwater, denitrification does not occur. Since HCO3 − concentration decreases due to nitrification of NH4 + [37], HCO3 − shows higher concentration without nitrate pollution. PC2 has a negative loading for SO4 2− too. Thus, PC2 is related to the effects of chemical fertilizer (NH4)2SO4. In addition, positive loading for Ca 2+ and negative loading for Na + represent cation exchange [38]. Applying HCA, 69 water chemistry samples were classified into four different groups (Table 3). Depending on nitrate content, the four groups can be condensed into two groups: non-polluted group (Group 1) and nitrate polluted group (Group 2-4). Furthermore, based on the Japanese drinking water standard and the threshold of 3 mg L −1 indicating effects of anthropogenic activities, nitrate affected samples can be classified into two main groups: severely polluted by nitrate (Group 3) and moderately polluted by nitrate (Group 2 and 4). Most samples (59%) belong to Group 1. Samples from this group are found at higher altitudes in the central region (Figure 8), corresponding to the forested area ( Figure 2). Group 2-4 mainly cover the upland agricultural fields. A scatter plot of PC1 against PC2 depends on pollution group (Figure 9). According to Figures 8 and 9, there are two chemical evolution paths. One path is explained by evolution from Group 1 to Group 2, and then to Group 4 with increase of PC1. The second path is from Group 1 to Group 2, and then to Group 3. The difference between Group 3 and Group 4 is characterized by PC2. PC2 for Group 3 tends to be negative, indicating that chemical fertilizers have a more pronounced effect on nitrate pollution as compared to livestock waste. According to the positive correlation, ion concentration (Cl − , SO4 2− , Mg 2+ , and Ca 2+ ) increases as NO3 − content increases. In spite that NO3 − concentration does not increase from Group 2 to Group 4, the increase in cation concentration in Group 4 is higher than that of Group 3. This indicates that water-rock interaction (mineral dissolution and cation exchange) is the predominant factor controlling water chemistry in Group 4. Applying HCA, 69 water chemistry samples were classified into four different groups (Table 3). Depending on nitrate content, the four groups can be condensed into two groups: non-polluted group (Group 1) and nitrate polluted group (Group 2-4). Furthermore, based on the Japanese drinking water standard and the threshold of 3 mg L −1 indicating effects of anthropogenic activities, nitrate affected samples can be classified into two main groups: severely polluted by nitrate (Group 3) and moderately polluted by nitrate (Group 2 and 4). Most samples (59%) belong to Group 1. Samples from this group are found at higher altitudes in the central region (Figure 8), corresponding to the forested area ( Figure 2). Group 2-4 mainly cover the upland agricultural fields. A scatter plot of PC1 against PC2 depends on pollution group (Figure 9). According to Figures 8 and 9, there are two chemical evolution paths. One path is explained by evolution from Group 1 to Group 2, and then to Group 4 with increase of PC1. The second path is from Group 1 to Group 2, and then to Group 3. The difference between Group 3 and Group 4 is characterized by PC2. PC2 for Group 3 tends to be negative, indicating that chemical fertilizers have a more pronounced effect on nitrate pollution as compared to livestock waste. According to the positive correlation, ion concentration (Cl − , SO 4 2− , Mg 2+ , and Ca 2+ ) increases as NO 3 − content increases. In spite that NO 3 − concentration does not increase from Group 2 to Group 4, the increase in cation concentration in Group 4 is higher than that of Group 3. This indicates that water-rock interaction (mineral dissolution and cation exchange) is the predominant factor controlling water chemistry in Group 4.

Conclusions
To improve the understanding of hydrogeochemical evolution of groundwater in vulnerable aquifers and spatial characteristics of nitrate pollution and water chemistry in Unzen City, groundwater samples were collected from 12 residential and 57 municipal water supply wells and springs, and major dissolved ions and hydrogeochemical variables were analyzed. The general groundwater types are (Ca+Mg)−HCO 3 or (Ca+Mg)−(SO 4 +NO 3 ). Dominant anions are transformed due to nitrate pollution. In total eight wells (four municipal water supply wells and four residential wells), NO 3 −N levels were exceeding Japanese drinking water standards. Land use and correlation analysis indicate that excessive NO 3 −N is related to application of chemical fertilizers and discharge of livestock waste. Among water samples displaying high NO 3 −N concentrations exceeding Japanese drinking water standard, relatively high SO 4 2− concentrations of 18.6-30.6 mg L −1 were observed at 5 out of 8 locations (MW10, MW12, K35, Az659, and Ai29) compared to the others (MW9, MW32, and M12). Since NO 3 −N and SO 4 2− concentrations increase simultaneously, the groundwater is significantly polluted by chemical fertilizers such as (NH 4 ) 2 SO 4 . PCA results show that factors controlling water chemistry in Unzen City can be explained by two main PCs. PC1 accounts for mineral dissolution (water-rock interaction) and agricultural effects. PC2 explains the effects of chemical fertilizers and cation-exchange. HCA classified water samples into four groups. Further, these groups can be separated into a non-polluted group (Group 1), a moderately nitrate affected group (Group 2 and 4), and severely polluted group (Group 3). Combination of PCA and HCA reveals that Group 1 is evolving into Group 2 and eventually into Group 4 with the dominant effect of water-rock interaction. Group 2 is evolving into Group 3 by the dominant impacts of chemical fertilizers.
We confirmed that only specific wells in Unzen City are unsuitable for drinking purposes. A similar situation has been found in Shimabara City. However, since sampling was done during the rainy season in July and August, sampled groundwater may have been affected by recharge and thus dilution. In any case, since most of the samples exceed 3 mg/L, the groundwater in this area is clearly affected by anthropogenic activities. To certify that no negative health effects occur in the area, it is important to continuously monitor the nitrate concentrations in the groundwater.

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.