Journal of Hydrology 549 (2017) 703–714

Contents lists available at ScienceDirect

Journal of Hydrology
journal homepage: www.elsevier.com/locate/jhydrol

Research papers

Hydrochemical assessment of freshening saline groundwater using

multiple end-members mixing modeling: A study of Red River delta
aquifer, Vietnam
Ji-Hyun Kim a, Kyoung-Ho Kim a, Nguyen Thi Thao a,b, Bayartungalag Batsaikhan a, Seong-Taek Yun a,⇑
a
Department of Earth and Environmental Sciences and KU-KIST Green School, Korea University, Seoul 136-701, South Korea
b
Institute of Geological Sciences, Vietnam Academy of Science and Technology, 84 Chua Lang, Hanoi, Viet Nam

a r t i c l e i n f o a b s t r a c t

Article history: In this study, we evaluated the water quality status (especially, salinity problems) and hydrogeochemical
Received 22 June 2016 processes of an alluvial aquifer in a floodplain of the Red River delta, Vietnam, based on the hydrochem-
Received in revised form 5 February 2017 ical and isotopic data of groundwater samples (n = 23) from the Kien Xuong district of the Thai Binh pro-
Accepted 17 April 2017
vince. Following the historical inundation by paleo-seawater during coastal progradation, the aquifer has
Available online 21 April 2017
This manuscript was handled by L. Charlet,
been undergone progressive freshening and land reclamation to enable settlements and farming. The
Editor-in-Chief hydrochemical data of water samples showed a broad hydrochemical change, from Na-Cl through Na-
HCO3 to Ca-HCO3 types, suggesting that groundwater was overall evolved through the freshening process
Keywords:
accompanying cation exchange. The principal component analysis (PCA) of the hydrochemical data indi-
Freshening of salinized aquifer cates the occurrence of three major hydrogeochemical processes occurring in an aquifer, namely: 1) pro-
Multiple end-members mixing modeling gressive freshening of remaining paleo-seawater, 2) water-rock interaction (i.e., dissolution of silicates),
Cation exchange and 3) redox process including sulfate reduction, as indicated by heavy sulfur and oxygen isotope com-
Bacterial sulfate reduction positions of sulfate. To quantitatively assess the hydrogeochemical processes, the end-member mixing
Hydrogeochemistry analysis (EMMA) and the forward mixing modeling using PHREEQC code were conducted. The EMMA
results show that the hydrochemical model with the two-dimensional mixing space composed of PC 1
and PC 2 best explains the mixing in the study area; therefore, we consider that the groundwater chem-
istry mainly evolved by mixing among three end-members (i.e., paleo-seawater, infiltrating rain, and the
K-rich groundwater). The distinct depletion of sulfate in groundwater, likely due to bacterial sulfate
reduction, can also be explained by EMMA. The evaluation of mass balances using geochemical modeling
supports the explanation that the freshening process accompanying direct cation exchange occurs
through mixing among three end-members involving the K-rich groundwater. This study shows that
the multiple end-members mixing model is useful to more successfully assess complex hydrogeochem-
ical processes occurring in a salinized aquifer under freshening, as compared to the conventional inter-
pretation using the theoretical mixing line based on only two end-members (i.e., seawater and
rainwater).
Ó 2017 Elsevier B.V. All rights reserved.

1. Introduction been thoroughly examined in many studies (Andersen et al.,

2005; Antonellini et al., 2015; Appelo, 1994; Blaser et al., 2010;
Many low-lying coastal plains over the world recently have de Montety et al., 2008; Fidelibus, 2003; Gattacceca et al., 2009;
experienced significant environmental changes due to human Ghabayen et al., 2006; Lambrakis, 2006; Mondal et al., 2010;
development as well as natural processes such as seawater intru- Pulido-Leboeuf, 2004; Russak and Sivan, 2010; Werner et al.,
sion. In such areas, groundwater is generally saline but has been 2013). The saline groundwater generally originates from seawater
undergone freshening and anthropogenic contamination. Freshen- intrusion or significant evaporation. In particular, seawater intru-
ing process is defined as a phenomenon where saline groundwater sion into a fresh aquifer in many low-lying coastal plains occurred
is slowly flushed by freshwater recharge; this phenomenon has extensively at the beginning of Last Interglacial period by costal
progradation due to a rapid sea-level rise. After the stabilization
⇑ Corresponding author.
of the sea level to the present sea level, the intruded seawater
remained in the aquifer as saline groundwater (i.e., fossil seawater)
E-mail address: [email protected] (S.-T. Yun).

http://dx.doi.org/10.1016/j.jhydrol.2017.04.040
0022-1694/Ó 2017 Elsevier B.V. All rights reserved.
704 J.-H. Kim et al. / Journal of Hydrology 549 (2017) 703–714

and has been successively freshened. Freshening of a saline aquifer calculation (M3) model (Laaksoharju et al., 1999; Laaksoharju
depends on a complex of natural and/or artificial factors, such as et al., 2008) and the model MIX (Carrera et al., 2004; Tubau
the continental recharge system (e.g., recharge by precipitation), et al., 2014; Vàzquez-Suñé et al., 2010) have been recently
regional geology, sea level regression (Blaser et al., 2010; Post proposed.
et al., 2013), tidal fluctuations (Wang and Tsay, 2001), and artifi- A better understanding of freshening process of an aquifer in
cial recharge (Lambrakis, 2006). Thus, we need to a better under- low-lying coastal plains which experienced paleo-seawater intru-
standing of the complex hydrogeochemical processes occurring sion is crucial to exploit groundwater sustainably for human devel-
during freshening. opment. However, such task is challenging because of the
The freshening process generally accompanies two major complexities of hydrogeologic and hydrochemical environments.
hydrogeochemical processes: (1) dilution (i.e. mixing) and (2) Therefore, the present study aims to examine the usefulness of
cation exchange. During freshening, the concentrations and behav- the multiple end-members mixing model to quantitatively assess
iors of cations in water are strongly influenced by the surface of complex hydrochemical processes occurring during freshening of
minerals in an aquifer. Cation exchange is an important process salinized aquifer. For this assessment, we used two approaches of
occurring at the mineral surfaces that proceeds with a sequence mixing and mass balance calculation, namely: 1) end-member
during the progressive freshening. Direct cation exchange is repre- mixing analysis (EMMA) in conjunction with the principal compo-
sented by the displacement of Na ion at mineral surfaces by other nent analysis (PCA) and 2) forward mixing modeling using
cations, such as Ca and Mg ions in the solution as below (Appelo, PHREEQC. The EMMA was developed by Christophersen and
1994; Appelo and Postma, 2005): Hooper (1992) to estimate the number of end-members involved
in the mixing to form stream water. In the present study, we
applied EMMA to estimate the conservative mixing space and the
2NaX þ Ca2þ ! 2Naþ þ CaX2
number of end-members involved in a saline groundwater system
Thus, an initial Na-Cl type saline groundwater progressively under alluvial setting. Mass balances of major cations during sev-
evolves towards the Na-HCO3 type water and, eventually, changes eral processes, such as cation exchange during the freshening, were
to the Ca-HCO3 type fresh groundwater over a long period of time also evaluated by the forward mixing modeling (Appelo, 1994;
through the direct cation exchange during freshening (Appelo, Appelo et al., 1990; Beekman, 1991) with the end-members
1994; Appelo and Postma, 2005; Appelo and Willemsen, 1987; deducted from EMMA. The methods used in this study can be
Beekman, 1991; Lambrakis, 2006). Cation exchange can also be applied to interpret the hydrogeochemical processes during fresh-
used as a criterion of freshening or salinization in an alluvial aqui- ening of a heterogeneous aquifer.
fer and, therefore, several indices were proposed to identify the
type of mixing processes (Giménez-Forcada, 2010, 2014; Russak
and Sivan, 2010). 2. Study area
Several hydrogeochemical processes, such as dissolution/pre-
cipitation of minerals and redox reactions, can also occur during The study area is located in the northeastern part of the Kien
the freshening of a saline aquifer. Dissolution or precipitation of Xuong district of the Thai Binh Province, Vietnam, which belongs
secondary minerals, such as calcite, dolomite, and gypsum, can to the Red River delta and borders the Gulf of Tonkin (Fig. 1a).
occur because the mixing of groundwater leads to dis- There is a hydrologic network comprising a main river (named
equilibrium states between aquifer minerals and groundwater Tra-Ly) as a tributary of the Red River, which forms an important
(Chae et al., 2012; Garing et al., 2013; Gomis-Yagües et al., 2000; drainage system used for pervasive agricultural activities (Fig. 1a).
Plummer, 1975; Wigley and Plummer, 1976). As alluvial sediments The area is under a subtropical climate, with nearly 80% of annual
in floodplains undergoing freshening usually contain abundant precipitation between June and July. The mean annual rainfall and
organic matters, diverse redox reactions, such as bacterial reduc- temperature are ca. 1800 mm and 23 °C, respectively. Due to high
tions of Fe- and Mn-oxyhydroxides, nitrate, and sulfate, can occur. soil fertility, flat topography, and favorable irrigation condition,
Thus, distinct sulfate depletion is frequently found in saline agricultural activity, mainly rice cropping, is intensive throughout
groundwater, although sulfate is very high in seawater (Andersen the delta plain.
et al., 2005; Chapelle and McMahon, 1991; Kim et al., 2014). Fur- The Red River delta is surrounded by hills that are geologically
thermore, leaching of redox-sensitive, toxic elements, such as composed of Precambrian crystalline rocks and Paleozoic to Meso-
arsenic, has been reported in many delta floodplains of Southeast zoic sedimentary rocks. The delta was initially formed at the south-
Asia, including the Red River, Mekong, and Bengal (Buschmann west of the present Red River channel during the Last Glacial
and Berg, 2009; Fendorf et al., 2010; Winkel et al., 2008; Winkel Maximum (LGM) (Tanabe et al., 2003a, 2003b). Quaternary sedi-
et al., 2011). ments of the Red River delta overlie the Neogene deposits and
To assess several hydrochemical processes during freshening, are ca. 100 m thick with the increasing thickness (up to 200 m)
the differences of hydrochemical parameters between the towards the coast. After the LGM period, global deglaciation led
observed and the theoretical concentrations based on the conser- to a rapid rise of the sea level (i.e., transgression) up to ca. 1–3 m
vative mixing between seawater and freshwater have been used above the present sea level and resulted in the inundation of the
in many previous studies (Appelo and Postma, 2005; de Montety study area by seawater due to coastal progradation. Following
et al., 2008; Fidelibus et al., 1993; Russak and Sivan, 2010; the stability of the sea level to the present level, regression of the
Yamanaka and Kumagai, 2006). This conventional method is based Red River delta resulted in the emergence of the Haihung Forma-
on the linear regression of the chemical compositions of two end- tion (7000–4000 years BP) and the Thai Binh Formation (after
members (i.e., seawater and freshwater), as the samples with a 3000 years BP) throughout the Holocene (see Fig. 1b). The Holo-
simple and conservative mixing follow the theoretical mixing line. cene alluvial sedimentary sequence is composed of the interbed-
However, due to the uncertainties in the number of end-members ded layers of sand and gravel with subordinate lenses of silt and
involved in the mixing and selection of conservative tracers, this clay (Mathers and Zalasiewicz, 1999; Tanabe et al., 2003a, 2003b,
conventional interpretation using two end-members cannot effec- 2006). Seawater intruded into the Holocene alluvial sedimentary
tively interpret a heterogeneous aquifer system under freshening. layers during the regression of the Red River delta has remained
Therefore, alternative multiple-dimensional and mathematical as fossil seawater, causing the salinity problems in the study area.
approaches, such as the multivariate mixing and mass balance Following the paleo-seawater intrusion, the alluvial aquifer has
 J.-H. Kim et al. / Journal of Hydrology 549 (2017) 703–714 705

Fig. 1. (a) Location map of the study area in Vietnam and sampling points and (b) a representative geologic profile of the Red River delta, showing the general sequence of
different layers (modified after Tanabe et al., 2006; Mathers and Zalasiewicz, 1999).

been undergone progressive freshening and land reclamation to bottles. The samples for the cation analysis were acidified to
enable settlements and farming. The average cation exchange pH < 2 by adding a few drops of ultrapure nitric acid. All samples
capacity (CEC) of alluvial soils in the Red River delta was reported were refrigerated at 4 °C until the analyses. Alkalinity was mea-
to amount to 10.71 cmolc/kg (Nguyen et al., 2005). sured by the titration method using 0.05 N HNO3 solution in the
The water use in the study area relies on alluvial groundwater. field (Stumm and Morgan, 1996).
Therefore, there are many groundwater wells for the water supply Chemical compositions of the collected water samples were
for drinking, agricultural, and industrial purposes. The preexisting analyzed at the Center for Mineral Resource Research, Korea
groundwater wells in the study area have the depths between University. Major cations (Na+, K+, Ca2+, and Mg2+), Sr2+, Fetotal,
1.5 m and 45 m below the ground surface (Table 1); therefore, Mntotal, and dissolved SiO2 were measured by the inductively cou-
the wells in use are mostly completed within the sand layer under- pled plasma-atomic emission spectrometry (ICP-AES: Perkin Elmer
lying clay/silt layer (lenses) of the Thai Binh and Haihung forma- OPTIMA 3000 XL). Major anions (Cl-, SO2- - - -
4 , NO3, and F ) and Br were
tions of Holocene age (see Fig. 1b). Alluvial aquifer in the study measured by the ion chromatography (IC: Dionex 120). The calcu-
area is generally semi-confined and leaky due to the discontinuity lated charge balance errors (C.B.E.) of the analytical results were
of a clay-silt layer at the surface. Although groundwater levels mostly below 5%, indicating the reliability of the analytical data.
could not be measured at each groundwater wells, local groundwa- During the second field campaign in June 2009, 17 groundwater
ter flow in the study area is considered to overall direct northeast- samples were selected for the analysis of sulfur and oxygen iso-
ward towards the Tra-Ly River. topes of sulfate. The collected samples were acidified (pH 2–3)
after filtering; then, SO4 was precipitated in the field as BaSO4 by
addition of BaCl2 (Clark and Fritz, 1997). The samples were ana-
3. Materials and methods lyzed in the Isotope Science Laboratory at the University of Calgary
using the high-temperature conversion elemental analyzer (TC/EA)
3.1. Sampling and analysis coupled to the Delta Plus XL isotope-ratio mass spectrometer with
SO2 as the measuring gas. Stable isotope data of the samples are
The results of the chemical analysis of 23 groundwater samples reported in the d notation relative to the Canyon Diablo Troilite
from preexisting wells, together with one rainwater sample, are (V-CDT) standard for d34S values and the Standard Mean Oceanic
shown in Table 1. Sampling was conducted twice in the same Water (V-SMOW) standard for d18O values. The precision and accu-
selected wells during two periods: October 2006 and June 2009. racy were below 0.3‰ for d34S and 0.5‰ for d18O.
One rainwater sample was also collected in October 2006. The
sampling procedure and field measurements followed the standard
method (APHA et al., 2001). Groundwater sampling for the subse- 3.2. End-member mixing analysis (EMMA)
quent analyses was performed after sufficient purging of at least
three well volumes. The temperature, pH, Eh, electrical conductiv- The dataset (X) used in EMMA consists of the collected ground-
ity (EC), and dissolved oxygen (DO) were measured in the field water samples and a rainwater sample. The average seawater com-
using potable multi-parameter meters (Thermo Scientific Orion position was also used (Hem, 1985). Twelve hydrochemical
Co.). The collected samples were filtered through 0.45 lm cellulose parameters, including EC, Na+, K+, Ca2+, Mg2+, Cl, NO 2 
3 , SO4 , HCO3 ,
membranes and then transferred into pre-cleaned 60 ml HDPE SiO2, Fe, and Mn, were used as input data. The hydrochemical data
706 J.-H. Kim et al. / Journal of Hydrology 549 (2017) 703–714

Table 1
Summary of hydrochemical parameters of groundwater samples from the study area.

Samples Well depth pH EC DO Eh Dissolved constituents (mmol/L)

(m, bgl)1 (lS/cm) (mg/L) (mV)
Na K Ca Mg Cl NO3 SO4 HCO3 SiO2 Fe Mn
Groundwater
1 1.5 7 453 1.84 506.4 1.16 0.29 2.04 0.23 1.01 0.03 0.27 4.37 0.41 0.0001 0.0001
2 2 6.6 1020 1.92 512 3.99 0.44 1.69 0.84 3.66 0.04 0.17 4.5 0.76 0.3381 0.0696
3 8 6.3 1080 0.7 212 3.88 1.96 1.37 2.04 5.26 0.07 0.01 7.62 0.84 0.0115 0.0626
4 8 6.6 1050 0.85 475.1 3.36 1.76 1.27 1.24 2.81 1.01 0.65 4.4 0.79 0.0002 0.006
5 10 5.4 2050 0.36 110.2 7.53 4.24 3.2 2.24 9.83 0.03 1.8 10 0.79 0.0288 0.0348
6 10 5.6 2330 1.97 85.2 11.8 1.99 2.86 3.09 16.68 0.05 0.01 11.35 0.73 0.1036 0.0359
7 10 5.4 1515 1.65 371 7.31 0.72 2.88 1.49 7.97 0.07 1.86 5.17 0.8 0.0008 0.0029
8 10 6.8 1828 1.11 537.2 6.89 3.56 2.54 1.65 7.02 0.3 1.73 7.4 0.73 0.0658 0.0305
9 10 7.1 3430 1.68 509.2 15.64 1.75 4.46 3.48 21.08 0.08 0.26 9.25 0.7 0.1701 0.0516
10 10 7 2060 4.82 375 10.09 0.64 2.64 1.53 8.86 0.07 2.16 4.5 0.71 0.0014 0.0022
11 11 6.4 506 1.15 77.2 1.27 0.98 0.94 0.7 1.41 0.02 0.01 5.5 0.79 0.2823 0.0251
12 11 6.7 770 0.95 544.6 2.15 0.76 1.14 0.84 2.14 0.01 0.04 4.25 0.73 0.3056 0.0367
13 12 6.9 1025 1.24 76.7 6.05 0.64 0.9 1.15 6.97 0.02 0.01 5.55 0.63 0.0704 0.0112
14 12 6.6 2990 0.9 60.2 19.62 2.77 2.31 3.46 26.71 0.07 0.23 11.37 0.86 0.1643 0.0206
15 12 6.5 5190 0.91 461 34.69 0.7 1.52 3.7 42.1 0 0.41 2.5 0.69 0.0244 0.0165
16 12 6.9 3090 0.88 632.5 16.26 1.77 1.47 2.61 17.05 0 0.21 8.85 0.79 0.1065 0.0132
17 39 5.3 3380 2.28 244.5 23 0.65 1.52 3.37 32.19 0.04 0.22 2.67 0.8 0.0132 0.0174
18 39 6.5 4340 1.01 401.5 27.99 0.53 1.48 3.31 34.11 0 0.25 2.5 0.75 0.0117 0.0174
19 40 5.4 5480 1.25 141 44.12 1.42 1.22 4.24 53.52 0.09 0.21 4.35 0.58 0.0821 0.0152
20 40 6.8 5290 0.85 505.8 37.62 0.86 0.98 3.25 41.35 0 0.27 3.5 0.59 0.0615 0.0127
21 45 5.4 4610 1.23 194.1 38.37 1.19 1.2 3.71 46.02 0.07 0.33 3.6 0.66 0.0207 0.0126
22 45 6.7 4440 2.29 320 36.3 1.14 1.66 4.4 46.84 0.08 0.33 3.25 0.69 0.0105 0.016
23 45 6.7 5230 0.75 450.1 37.24 0.81 1.07 3.38 41.9 0 0.37 3.15 0.64 0.0113 0.0118
Average 19.2 6.4 2746 1.42 339.2 17.23 1.37 1.84 2.43 20.72 0.09 0.51 5.64 0.71 0.082 0.0227
S.D.2 15.6 0.61 1723 0.91 184.9 14.47 1.01 0.9 1.26 17.66 0.21 0.66 2.81 0.1 0.103 0.0185
Rainwater - 5.78 67.4 4.96 283.8 0.06 0.03 0.28 0.01 0.1 0.05 0.03 0.52 0.044 0 0.0001
1
Depth below ground level; 2Standard deviation.

were standardized into the correlation matrix (Z) by using Eq. (1), 3.3. Forward mixing modeling with PHREEQC
as each parameter has the different scale and variance:
The PHREEQC code (version 3.1.2) was used to perform one-
ðxij 
xj Þ
Z ij ¼ ð1Þ dimensional forward transport modeling to simulate the freshen-
sj
ing process (Appelo, 1994; Appelo et al., 1990; Appelo and
where xij is the original value of the chemical parameter (j) in the Postma, 2005; Beekman, 1991). Mass transfer was calculated
observed data (i),
xj is the mean value of the parameter (xj) for all using synthetic waters (Appelo and Postma, 2005). The parame-
data, and sj is the standard deviation of the parameter for all data. ters used in the modeling were temperature, pH, Ca, Na, Mg, K,
The EMMA model was used to determine both the conservative Cl, HCO3, SO4, and Si. Here, chloride was considered as a conser-
tracer and the number of end-members to better explain the vative element which is essential to determine the mixing pro-
hydrochemistry of saline groundwater in the aquifer. The model portions. In the forward mixing modeling, mixing, dissolution of
was established following the proposed procedure based on the minerals, such as calcite, dolomite, and gypsum, and cation
orthogonal regression analysis (Christophersen and Hooper, exchange were considered to simulate the chemical reactions in
1992). The conservative mixing space formed by a space of PCA groundwater. To consider the diffusion in the simple reaction
score of Z was projected back onto the original data space. path modeling, the TRANSPORT code was used. The time duration
The diagnostic tools of Hooper (2003) are most commonly used of the simulation was set up as 100 years and the diffusion coef-
in the EMMA model to determine the dimensionality of the conser- ficient was determined up to the degree of dilution of seawater to
vative mixing space and the number of end-members by retaining the level of Cl concentration of freshwater. In the EXCHANGE
the number of eigenvector (i.e., the number of principal compo- code, the Gains-Thomas convention was used to the cation
nent). A common criterion for determining the number of eigen- exchange coefficients.
vector to retain for further analysis should be those with an
eigenvalue above 1. In the case of determining the number of
4. Results and discussion
end-members, the number of retained eigenvectors (n) necessi-
tates n + 1 end-members. Then, the residual analysis was per-
4.1. General water chemistry
formed by calculating the differences between the measured
concentrations and the concentrations expected from the
The results of the chemical analysis of 23 groundwater samples
conservative mixing space. Then, the residuals were evaluated by
are summarized in Table 1. Fig. 2a shows the Durov diagram of the
errors statistics. A random pattern (i.e., approach to zero) of the
hydrochemical data. The hydrochemical phases of groundwater in
residuals indicates the well-considered mixing space. The relative
the study area gradually changed from Na-Cl through Na(-Ca)-
root-mean-square error (RRMSE) was calculated using Eq. (2) to
HCO3 to Ca-HCO3 types with the concomitant decrease of the
represent the standard deviation of residual values:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi TDS value from 3450 to 481 mg/L. The observed hydrochemical
Pn change generally tended to follow the theoretical mixing line
^
i¼1 ðxij  xij Þ
2

rj ¼ ð2Þ (TML) between rainwater (TDS = 59 mg/L; Ca-HCO3 type, Table 1)

n
xj
and seawater (TDS = 34,575 mg/L; Na-Cl type); however, the plots
 J.-H. Kim et al. / Journal of Hydrology 549 (2017) 703–714 707

Fig. 2. (a) Durov diagram showing general groundwater chemistry in the study area, (b) plots of the relative concentrations of major constituents (in meq/L) of samples on the
hydrochemical facies evolution diagram (Giménez-Forcada, 2010), and (c) the calculated salinization index (Russak and Sivan, 2010) versus Cl diagram indicating the
occurrence of progressive freshening in groundwater. RW = rainwater, SW = seawater, TDS = total dissolved solid, TML = theoretical mixing line.

of most hydrochemical data fall slightly above the TML. Such 4.2. Major hydrogeochemical processes
hydrochemical evolution indicates the occurrence of the freshen-
ing process accompanying direct cation exchange (i.e., the release Before the principal component analysis (PCA), the Bartlett’s
of Na+ from the surface of mineral particles to solution while sphericity test was conducted on data matrix to test whether our
Ca2+ in solution is absorbed onto the surface instead). The calcula- data are adequate to be evaluated by PCA. The results yielded the
tion of the percentage of seawater fraction in water using the con- associated probability of 0.000, indicating that the variables in cor-
centration of Cl ion shows that ca. 10% of seawater are still relation matrix are strongly correlated. Therefore, our data are
maximally remained in the groundwater bodies. enough to proceed PCA even with the small number of samples
There is a general trend of decreasing EC values (r2 = 0.64) (23). Three major hydrogeochemical processes of groundwater
with the decreasing depth of groundwater wells (see Table 1). were identified by PCA of 12 analyzed parameters (EC, Na+, Ca2+,
This suggests that salinized groundwater has been undergone K+, Mg2+, Cl-, HCO-3, SiO2, SO2- -
4 , NO3, Fe, and Mn). These are: (1) pro-
progressive freshening by the inflow of freshwater slowly through gressive freshening, represented by PC 1; (2) water–rock interac-
semi-confined and leaky aquifer. In addition, the hydrochemical tion, represented by PC 2; and (3) redox process, represented by
facies evolution diagram (HFE-D) and the calculated salinization PC 3 (Fig. 3). Each process (component) explains 37.9%, 20.1%,
index involving the cation exchange reaction also indicate the and 16.2% of the total variance of groundwater chemistry in the
freshening process. In the HFE-D proposed by Gimenez-Forcada study area, respectively (Table 2).
(2010), groundwater samples with relatively high TDS values fol-
low the trend of intrinsic freshening during which the initial 4.2.1. Freshening
phase of saline groundwater (Na-Cl type) shifts to the Ca(-Na)- PC 1 has positive correlations with EC, Na+, Mg2+, and Cl-
HCO3 phase due to the dilution (mixing) and direct cation (Fig. 3a) and the samples with lower Cl- concentrations generally
exchange (Fig. 2b). The other group of freshwater (i.e. Na(-Ca)- have the more negative PC 1 scores (Fig. 3b). These trends indicate
HCO3 type groundwater) plots above the mixing line and is char- the progressive freshening towards the negative PC 1 direction. The
acteristically HCO3-rich (Fig. 2b); this groundwater represents water with negative PC 1 scores are typically of Na(-Ca)-HCO3 type,
another path of freshening (see below). Calculation of the salin- while those with positive scores are of Na-Cl type. Thus, direct
ization index proposed by Russak and Sivan (2010), based on cation exchange accompanied by freshening is also expected to
the opposite behavior between Ca2+ and K+ during cation result in the enrichment of Na+ with the concomitant depletion
exchange, yielded negative values. This also indicates the occur- of Ca2+ in groundwater. However, a careful examination of the vari-
rence of freshening (Fig. 2c). ation of the concentrations of major cations shows that most ions
708 J.-H. Kim et al. / Journal of Hydrology 549 (2017) 703–714

Fig. 3. Scatter plots of the results of principal component analysis (PCA). (a and b) PC 1 vs. PC 2 relationships and (c and d) PC1 vs. PC 3 relationships. The concentrations of Cl-
and SO2-4 of the samples are also shown in (b) and (d), respectively. The identified three major hydrochemical processes of groundwater are: PC 1 = progressive freshening, PC
2 = water-rock interaction, and PC 3 = redox process. RW = rain water, SW = seawater, TML = theoretical mixing line.

interactions (i.e., dissolution of silicate and/or carbonate minerals)

Table 2
are the second important hydrochemical process (Fig. 3a). A few
The results of principal component analysis (PCA). samples including sample #5 with the lower concentrations of
Cl- show such hydrochemical characteristics (Fig. 3b). The calcu-
Eigenvalue Total variance (%) Cumulative variance (%)
lated saturation index shows that groundwater is under-
PC1 4.54 37.86 37.86 saturated with respect to calcite, dolomite, and gypsum (Fig. 5).
PC2 2.41 20.07 57.93
However, the Quaternary alluvial aquifer in the Kien Xuong district
PC3 1.95 16.24 74.17
PC4 1.09 9.07 83.24 consists mainly of silicate sediments without carbonate minerals
PC5 0.68 5.69 88.93 or gypsum. Potassium ion is the most influential factor of PC 2, sug-
Eigenvector PC1 PC2 PC3 gesting the dissolution of K-bearing silicate minerals. Thus, we
EC 0.4326 0.2107 0.0531
consider that water-rock interaction in the aquifer is characterized
Na 0.4527 0.1338 0.0563
K 0.1664 0.5118 0.0735
by the dissolution of silicate minerals. It is also noticeable that high
Ca 0.1725 0.4218 0.1138 concentrations of K+, Ca2+, and HCO-3 occur in the groundwater
Mg 0.3703 0.3601 0.1218 with low chloride concentrations (below 30 mM; Fig. 4).
Cl 0.4477 0.1601 0.0811 Therefore, we can now suggest that there are two types of fresh
NO3 0.1183 0.0117 0.3490
groundwater (Cl- < 30 mM) in the study area. The groundwater
SO4 0.1201 0.2107 0.5209
HCO3 0.2539 0.4313 0.1351 with the positive factor scores on PC 2 represents the fresh ground-
SiO2 0.1945 0.2707 0.1175 water, possibly with a shorter residence time after infiltration,
Fe 0.1997 0.1067 0.5308 while groundwater with the negative scores on PC 2 possibly rep-
Mn 0.2061 0.1653 0.4973 resents their long residence time accompanying larger degrees of
dissolution of silicate minerals, resulting in high concentrations
of K+, Ca2+, and HCO-3.
do not simply follow the TML line (Fig. 4). The reason of this obser-
vation is considered as the unsuitability of the use of conventional 4.2.3. Redox process
TML to interpret complex hydrochemical data. Thus, it is needed to The positive factor scores of Fe and Mn, but the negative scores
find another reliable conservative mixing space instead of the TML. 
of SO2
4 and NO3 on PC 3 (Fig. 3c) indicate that the redox process
represents the third major process regulating the groundwater
4.2.2. Water-rock interaction chemistry. A few samples with the negative PC3 scores showed
PC 2 negatively correlates with the parameters such as K+, the higher concentrations of sulfate than the samples with the
HCO-3, Ca2+, Mg2+ and aq. SiO2, implying that water-rock positive PC3 scores (Fig. 3d). We consider that organic carbon in
 J.-H. Kim et al. / Journal of Hydrology 549 (2017) 703–714 709

Fig. 4. Plots of the concentrations of major hydrochemical parameters versus Cl- in groundwater. Solid lines indicate the theoretical mixing line (TML). Shaded triangles
indicate the mixing diagrams (spaces) drawn by linear combinations between EMs.

alluvial sediments facilitates the reduction of NO3 , Mn(IV), Fe(III), Sasaki, 1996; Soler et al., 2002; Vitòria et al., 2004). We calculated

and SO24 , resulting in the depletion of NO3 and SO4
2
and the the sulfur isotope enrichment factor (e) using the Rayleigh distilla-
enrichment of dissolved Fe and Mn. tion model (Eq. (3)) under a closed system condition (Fritz and
The relationships between redox-sensitive ions and Cl also Fontes, 1980):
explain that the aquifer is affected by the redox process (Fig. 4).
The considerable depletion of SO2 4 implies the occurrence of sul-
d ¼ d0 þ e  lnf ð3Þ
fate reduction. The concentrations of other redox-sensitive param-
eters, such as Fe2+ and Mn2+, are locally very high (up to 0.10 mM
and 0.02 mM, respectively) in the fresh groundwater (Cl < 30 where d and do are the residual and initial d34S values of sulfate, f is
mM). It is noticeable that such Fe- and Mn-rich samples are the fraction of the remaining sulfate, and e is the isotopic enrich-
observed in the shallow wells with the depths of 2–12 m below ment factor. The f values were calculated using the SO2- -
4 /Cl values
the land surface. to consider the effect of freshening (Yamanaka and Kumagai,
To further evaluate the sulfate reduction, several samples were 2006) (see Eq. (4)):
analyzed for sulfur and oxygen isotope compositions of sulfate
(Table 3; Fig. 6). The samples with high sulfur and oxygen isotope  
f ¼ ðSO2
4 sample =Cl
2
sample Þ=ðSO4 initial =Clinitial Þ ð4Þ
values of sulfate have very low sulfate concentrations, indicating
the occurrence of bacterial sulfate reduction in the study area. Even with a few samples considered, the calculated e values for
The plots of the dual isotope compositions (Fig. 6) suggest the exis- the groundwater in the study area were 28.71‰ for the reduction
tence of two kinds of sulfate origin in the study area: 1) paleo- path from seawater sulfate and 29.25‰ for the path from an
seawater (d34SSO4 = 21‰ and d18OSO4 = 9.5‰) which intruded into anthropogenic sulfate. These estimates fall within the general
the aquifer and 2) anthropogenic sulfate (e.g., sulfate from fertiliz- range (8 to 90‰) reported for bacterial sulfate reduction
ers) (Clark and Fritz, 1997; Krouse and Mayer, 2000; Mizota and (BSR; Kim et al., 2014; Massmann et al., 2003).
710 J.-H. Kim et al. / Journal of Hydrology 549 (2017) 703–714

Fig. 6. Sulfur and oxygen isotopic compositions of sulfate in the groundwater

samples, suggesting the potential occurrence of two pathways of bacterial sulfate
Fig. 5. Relationships between Cl- concentrations and calculated saturation indices reduction in the study area. The boxes of different origins of sulfate are from Krouse
of groundwater samples, suggesting that groundwater was mostly undersaturated and Mayer (2000) and Vitòria et al. (2004).
with respect to calcite, dolomite, and gypsum during the freshening process.
TML = theoretical mixing line.
the two-dimensional mixing space have the lower RRMSE values
for major parameters, such as Na+, K+, Ca2+, Mg2+, Cl, HCO 3 , and
SiO2 than those from the one-dimensional mixing space. As
Table 3 expected, chloride ion has the lowest RRMSE (6.02%) from the
Sulfur and oxygen isotopic compositions of sulfate two-dimensional mixing space. Therefore, the results of EMMA
in groundwater.
suggest the following: 1) the hydrochemical evolution model with
Sample d34S (‰) d18O (‰) the two-dimensional mixing space composed of PC 1 and PC 2 bet-
2 32.9 16.5 ter explains the mixing phenomenon in the study area and 2) the
4 9.3 5.8 evolution of groundwater chemistry is controlled by the mixing
8 3.1 7.8 among three end-members. Combined with the observations as
9 27.5 17.6 described above, it is obvious that these three end-members are
10 0.8 8.7
12 22.5 –
paleo-seawater (EM1), infiltrated freshwater as rainwater (EM2),
15 45.5 16.5 and groundwater evolved through water-rock interaction (EM3).
16 – 15.8 Herein, the third end-member (EM3) is selected as the sample
18 40.6 15.7 #5 with the lowest factor score on PC 2 and is named the ‘K-rich
20 47.5 15.2
groundwater’. Here, the notation ‘K-rich groundwater’ means the
23 48.4 16.3
groundwater containing relatively high concentrations of K, even
though such samples (e.g. sample #5) hydrochemically belong to
the NaHCO3 type groundwater. The K-rich groundwater is also
4.3. Multiple end-member mixing and geochemical reaction modeling shown in the opposite directions to the seawater (SW) on PC 1
and to the rainwater sample (RW) on PC 2 (Fig. 3b), which indi-
4.3.1. End-member mixing analysis: Evaluation of end members cates the suitability of the sample #5 to be selected as the third
In order to determine the conservative mixing space with end-member involved in the mixing process. Therefore, we con-
appropriate end-members, the end-member mixing analysis sider that freshening process in the Red River delta aquifer has
(EMMA) was conducted. The residuals for each parameter were occurred through the freshwater recharge by both slow direct infil-
compared between the results from the one-dimensional and tration (of rainwater and irrigation water) and K-rich groundwater.
the two-dimensional mixing space to check the reliability of
the projection of principal components (Fig. 7). Here, the one-
dimensional mixing space consists of PC 1 only, while the two- 4.3.2. Residuals of sulfate reduction: evaluation of sulfate reduction
dimensional mixing space is made up of both PC 1 and PC 2 and, Another major process of hydrochemical evolution in the study
thus, represents the compositional changes by both freshening area, the redox process represented as PC 3, was further evaluated
and water-rock interaction. It is noticeable that, in the two- by the calculated residuals. The remarkable depletion of sulfate,
dimensional mixing space, the fluctuations of the residuals for compared to the ideal compositional change along the TML line,
major parameters, such as Na+, Ca2+, Mg2+, K+, and HCO3 , are signif- was observed in groundwater with very high salinity (Fig. 4). This
icantly reduced as compared to those of the one-dimensional mix- trend is certainly the result of the reduction of seawater sulfate, as
ing space, especially in the fresh groundwater with low chloride suggested by the S-O isotopic composition of sulfate in groundwa-
concentrations (<30 mM) (Fig. 7). This indicates that two types of ter (Fig. 6). However, it is noteworthy that the residuals for the sul-
groundwater whose chemistry was controlled mainly by freshen- fate calculated from the two-dimensional mixing space (Fig. 7b)
ing process (PC 1) and water-rock interaction (PC 2) are both were quite different from those from TML (Fig. 4).
involved in the mixing. There are distinct two groups of sulfate residuals in Fig. 7b. The
To further quantitatively identify the conservative mixing space strong depletion of sulfate is observed in the samples with the Cl-
and the number of end-members, the residuals were concentrations of < 30 mM, which possibly indicates that fresh
mathematically evaluated by the residual analysis using the rela- groundwater experienced larger degrees of BSR during freshening.
tive root-mean-square error (RRMSE) (Table 4). The results from Moreover, the comparison between the depletion of SO2- 4 and the
 J.-H. Kim et al. / Journal of Hydrology 549 (2017) 703–714 711

Fig. 7. Plots of the Cl concentrations vs. residuals relationships of groundwater samples in the study area. Residuals denote the difference between the original data (i.e.,
measurements) and the projected data onto (a) the one-dimensional mixing space and (b) the two-dimensional mixing space using orthogonal regression analysis.

Table 4
Relative root square errors (RRMSE; %) of the projection of the groundwater samples onto different mixing spaces of dimensionality.

Mixing subspace dimensionality EC Na K Ca Mg Cl NO3 SO4 HCO3 SiO2 Fe Mn

1 10.67 6.72 13.54 9.18 7.28 7.20 47.52 17.26 9.93 3.23 26.42 17.00
2 10.33 5.70 8.67 7.06 5.54 6.02 47.52 17.59 7.28 3.09 26.26 16.40
3 10.35 5.74 8.54 7.00 6.08 6.15 41.81 19.00 6.97 3.04 17.37 11.22

enrichment of HCO-3 for the groundwater samples with the Cl- con- from EMMA, we could identify the existence of the reduction of
centrations of <30 mM shows that the residuals of SO2- 4 are inver- anthropogenic sulfate, as we have proposed by S-O isotopic com-
sely correlated with those of HCO-3. This observation supports the positions of sulfate.
explanation that BSR was more distinct in groundwater with lower Therefore, our results suggest that the occurrence of BSR in the
salinity (cf. Kim et al., 2014). It is also noticeable that two outlier study area is identified by evaluating the residuals of sulfate from
samples with positive residuals of sulfate were collected from the two-dimensional mixing space. Bacterial sulfate reduction pos-
the same borehole and have the sulfate-sulfur isotopic composi- sibly occurred during both the seawater mixing in the past and the
tions falling in the general range of anthropogenic origin (Fig. 6). anthropogenic contamination in recent years (cf. Kim et al., 2014).
Through careful examination of the calculated residuals of sulfate Moreover, it is likely that BSR was more distinct during the
712 J.-H. Kim et al. / Journal of Hydrology 549 (2017) 703–714

recharge of salinized groundwater by fresh groundwater being end-members (i.e., seawater, infiltrated rainwater, and the K-rich
loaded with anthropogenic sulfate (possibly, fertilizers). groundwater).
The results of the mass balance calculation show that Na is sig-
nificantly released into the solution during progressive freshening,
4.3.3. Geochemical modeling of mixing and cation exchange whereas Mg and Ca are mostly absorbed onto the exchanger in
The compositional changes for cations in Fig. 4 were further alluvial sediments (Fig. 9a). It is noteworthy that the mass balance
evaluated by geochemical modeling using the identified three of each cation during mixing is best explained by the direct cation
end-members (EMs) of mixing. By drawing linear combinations exchange if the aquifer condition with the seawater fraction is ca.
between EMs (Mas-Pla et al., 2013), the mixing diagrams (spaces) 60%. Potassium shows an opposite behavior, falling between the
were added in Fig. 4 as shaded triangles. It is noticeable that plots results of Simulation 1 (i.e., releasing into water as the saline
of Ca2+ and K+ mostly fall within the mixing space. This clarifies groundwater mixes with rain infiltration) and Simulation 2 (i.e.,
that the third end-member (i.e. K-rich groundwater) should be sorbing onto sediment as the saline groundwater mixes with the
considered to successfully explain the freshening process in the K-rich groundwater). However, the behavior of K better follows
study area. the result of Simulation 2, implying that the K-rich groundwater
Therefore, the forward reaction modeling using the PHREEQC is involved in the direct cation exchange during progressive fresh-
code (Appelo and Postma, 2005) was implemented to elucidate ening. It is also likely that the extent of the direct cation exchange
the geochemical reaction path during the mixing among the sug- becomes minor and weaker, if the aquifer is significantly freshened
gested three end-members. For this purpose, we simulated the towards the seawater fraction of 0.3 to 10% (Fig. 9b).
freshening of a salinized aquifer with seawater component (EM1) Fig. 9c shows the calculated equivalent fractions (denoted as
by mixing with either rainwater (EM2, Simulation 1) or the K- lines) of cations sorbed onto the exchanger in sediments during
rich groundwater (EM3, Simulation 2). For the calculation of cation the aquifer freshening. The simulation was conducted using the
exchange, we used the average cation exchange capacity value PHREEQC code (Andersen et al., 2005). For comparison, actual
(10.71 cmolc/kg) reported for alluvial soils in the Red river delta observations in groundwater are shown as dots. The results show
(Nguyen et al., 2005). the following (Fig. 9c): 1) during progressive mixing, the cations
The modeling results (Fig. 8) show the concentration changes of with the maximum fractions in exchanger vary in the sequence
major cations, such as Na+, Mg2+, Ca2+, and K+, during the progres- of Na+, Mg2+, and Ca2+; 2) concomitantly with progressive freshen-
sive freshening (i.e., decreasing paleo-seawater fraction in ground- ing, especially Na is gradually released into groundwater from
water). Except for Na ion, our actual data do not follow the sorption sites, while Ca is typically absorbed onto sediments; these
conventional TMLs, implying the limitation of the use of the simple simulation trends are highly consistent with our actual data and
mixing model using two end-members (i.e., seawater and rainwa- the results of other studies (Appelo, 1994; Appelo and Postma,
ter) to properly explain the hydrochemical evolution. Furthermore, 2005; Beekman, 1991; Giménez-Forcada, 2010; Lambrakis, 2006;
the data of other cations, such as Mg2+, Ca2+, and K+, fall between Walraevens et al., 2001); and 3) our actual data for cations tend
the results of two simulations, although the observed data of Na to be more adequately explained by Simulation 2 than by Simula-
ion fit well with the results of Simulation 1 (Fig. 8). Therefore, there tion 1. Therefore, it can be concluded that 1) the groundwater
is convincing evidence that the groundwater chemistry in the chemistry in alluvial sediments of the study area in the Red
study area has evolved through the mixing among the three River delta is controlled by progressive mixing accompanying

Fig. 8. Plots of the relationships between seawater fraction and the concentrations of major cations in groundwater, compared with the results of end-member mixing
models using PHREEQC: Simulation 1 for the chemical change of seawater (EM1) during the freshening by infiltrating rainwater (EM2); and Simulation 2 for the chemical
change of seawater (EM1) during the freshening by the influx of the K-rich groundwater (EM2). TML = theoretical mixing line.
 J.-H. Kim et al. / Journal of Hydrology 549 (2017) 703–714 713

1) The observed hydrochemical data indicate the occurrence of

overall, progressive freshening of the salinized aquifer. Up to
ca. 10% of seawater still remained in groundwater. The
results of PCA indicate three major hydrogeochemical pro-
cesses: 1) progressive freshening (PC 1), which explains
37.9% of the total variance of groundwater chemistry; 2)
water-rock (silicates) interaction (PC 2), which results in
the generation of the K-rich groundwater and explains
20.1% of the total variance; and 3) redox process (PC 3),
which involves sulfate reduction preferentially in shallower
depths (<12 m below the ground surface) and explains 16.2%
of the total variance. The bacterial sulfate reduction is evi-
denced by heavy sulfur and oxygen isotopic compositions
of sulfate.
2) The results of EMMA reveal the following: 1) the hydrogeo-
chemical model with the two-dimensional mixing space
composed of PC 1 and PC 2 better explains the mixing in
the study area and 2) the mixing model with three end-
members (i.e., seawater, infiltrating rainwater, and the K-
rich groundwater) adequately explains the evolution of
groundwater chemistry. On the other hand, the use of con-
ventional theoretical mixing lines (TMLs) is unsuitable to
interpret complex hydrochemical data in the study area. In
particular, the one-dimensional mixing model (i.e., conven-
tional TML) is not appropriate to explain the hydrochemistry
of the significantly freshened groundwater with low chlo-
ride concentrations (<30 mM).
3) The use of the hydrogeochemical model with the two-
dimensional mixing space also better explains the occur-
rence of bacterial sulfate reduction (BSR). Two sources of
sulfate are likely to exist in the study area: seawater sulfate
and anthropogenic (possibly, fertilizers-derived) sulfate. The
evaluation of the residuals of sulfate from the two-
dimensional mixing space indicates that BSR is more distinct
during the inflow of the fresh, K-rich groundwater loaded
with anthropogenic sulfate.
4) The results of the forward mixing modeling using PHREEQC
show that the groundwater chemistry in the study area is
controlled by progressive mixing of both K-rich groundwa-
ter and infiltrating rainwater into the saline aquifer, which
accompanies dilution and direct cation exchange. The extent
of direct cation exchange becomes minor and weaker
towards the significantly freshened aquifer with the seawa-
ter fraction of 0.3 to 10%. During progressive freshening, Na
Fig. 9. The results of geochemical modeling as a function of seawater fraction (%) in gradually releases into groundwater from sorption sites,
the simulated groundwater. (a) Calculated mass balance (meq/l) for major cations while Ca is absorbed onto sediments.
as a result of direct cation exchange during freshening, (b) Enlarged figure of the 5) In summary, this study shows that the multiple end-
shaded part of (a), and (c) Calculated equivalent fractions of major cations existing
on the exchanger. Dots represent the observed compositions of the groundwater
members mixing modeling is very useful to assess complex
samples. The simulation processes 1 and 2 are the same as those in Fig. 8. hydrogeochemical processes occurring in an aquifer. To bet-
ter understand the hydrogeochemical processes of an allu-
vial aquifer under freshening, the combined use of EMMA
and forward geochemical modeling is highly recommended.
dilution and direct cation exchange and 2) in addition to rain infil- The approach used in the present study can also be very
tration, the K-rich groundwater with high concentrations of K+, helpful to assess desalinization (artificial freshening) tech-
Ca2+, and HCO-3 is involved in the mixing and dilution of saline nology as a part of the solutions of seawater intrusion prob-
groundwater. lems in many sites over the world.

5. Conclusions
Acknowledgments
The combined use of principal component analysis (PCA),
end-member mixing analysis (EMMA), and geochemical reaction This study was initiated for the Master of Science study of the
modeling was very successful to explain the evolution of salinized, third author as one of collaboration studies between Korea Univer-
alluvial groundwater in the Kien Xuong district in the Red River sity and Vietnamese Institute of Geological Sciences by the recom-
delta, Vietnam. Major findings of this study can be summarized mendation of Dr. Nguyen Trung Minh, currently at Vietnam
as follows: National Museum of Nature. The completion of this work was
714 J.-H. Kim et al. / Journal of Hydrology 549 (2017) 703–714

supported by KEITI through the Korea CO2 storage Environmental Kim, D.M., Yun, S.T., Kwon, M.J., Mayer, B., Kim, K.H., 2014. Assessing redox zones
and seawater intrusion in a coastal aquifer in South Korea using
Management (K-COSEM) Research Center.
hydrogeological, chemical and isotopic approaches. Chem. Geol. 390, 119–134.
Krouse, H.R., Mayer, B., 2000. Sulphur and oxygen isotopes in sulphate. In: Cook, P.
References G., Herczeg, A.L. (Eds.), Environmental Tracers in Subsurface Hydrology. Kluwer
Academic, pp. 195–231.
Andersen, M.S., Nyvang, V., Jakobsen, R., Postma, D., 2005. Geochemical processes Laaksoharju, M., Gascoyne, M., Gurban, I., 2008. Understanding groundwater
and solute transport at the seawater/freshwater interface of a sandy aquifer. chemistry using mixing models. Appl. Geochem. 23, 1921–1940.
Geochim. Cosmochim. Acta 69, 3979–3994. Laaksoharju, M., Skårman, C., Skårman, E., 1999. Multivariate Mixing and Mass-
Antonellini, M., Allen, D.M., Mollema, P.N., Capo, D., Greggio, N., 2015. Groundwater balance (M3) calculations, a new tool for decoding hydrogeochemical
freshening following coastal progradation and land reclamation of the Po Plain. information. Appl. Geochem. 14, 861–871.
Italy. Hydrogeol. J. 23, 1009–1026. Lambrakis, N., 2006. Multicomponent heterovalent chromatography in aquifers.
APHA, AWWA, WEF, 2001. Standard Methods for the Examination of Water and Modelling salinization and freshening phenomena in field conditions. J. Hydrol.
Wastewater. American Public Health Association, Washington, DC. 323, 230–243.
Appelo, C.A.J., 1994. Cation and proton exchange, pH variations, and carbonate Mas-Pla, J., Menció, A., Marsiñach, A., 2013. Basement groundwater as a
reactions in a freshening aquifer. Water Resour. Res. 20, 2793–2805. complementary resource for overexploited stream-connected alluvial
Appelo, C.A.J., Postma, D., 2005. Geochemistry, Groundwater and Pollution. aquifers. Water Resour. Manag. 27, 293–308.
Balkema, Rotterdam. 536 p.. Massmann, G., Tichomirowa, M., Merz, C., Pekdeger, A., 2003. Sulfide oxidation and
Appelo, C.A.J., Willemsen, A., 1987. Geochemical calculations and observations on sulfate reduction in a shallow groundwater system (Oderbruch Aquifer,
salt water intrusions. I. A combined geochemical/mixing cell model. J. Hydrol. Germany). J. Hydrol. 278, 231–243.
94, 313–330. Mathers, S., Zalasiewicz, J., 1999. Holocene sedimentary architecture of the Red
Appelo, C.A.J., Willemsen, A., Beekman, H.E., Griffioen, J., 1990. Geochemical River Delta, Vietnam. J. Coastal Res. 15, 314–325.
calculations and observations on salt water intrusions. II. Validation of a Mizota, C., Sasaki, A., 1996. Sulfur isotope composition of soils and fertilizers:
geochemical model with laboratory experiments. J. Hydrol. 120, 225–250. differences between Northern and Southern Hemispheres. Geoderma 71, 77–
Beekman, H.E., 1991. Ion Chromatography of Fresh- and Seawater Intrusion. 93.
Multicomponent Dispersive and Diffusive Transport in Groundwater. H.E. Mondal, N.C., Singh, V.P., Singh, V.S., Saxena, V.K., 2010. Determining the interaction
Beekman, Leiderdorp. between groundwater and saline water through groundwater major ions
Blaser, P.C., Coetsiers, M., Aeschbach-Hertig, W., Kipfer, R., van Camp, M., Loosli, H. chemistry. J. Hydrol. 388 (1–2), 100–111.
H., Alraevens, K., 2010. A new groundwater radiocarbon correction approach Nguyen, Q.H., Kazuhiko, E., Pham, Q.H., Ho, Q.D., 2005. Selected chemical properties
accounting for paleoclimate conditions during recharge and hydrochemical of alluvial soils in Vietnam. J. Faculty Agric. (Kyushu Univ.) 50, 233–242.
evolution: Ledo-Paniselian Aquifer. Belgium. Appl. Geochem. 2, 437–455. Plummer, L.N., 1975. Mixing of sea water with calcium carbonate ground water.
Buschmann, J., Berg, M., 2009. Impact of sulfate reduction on the scale of arsenic Geol. Soc. Am. Memoir 142, 219–236.
contamination in groundwater of the Mekong, Bengal and Red River deltas. Post, V.E., Groen, J., Kooi, H., Person, M., Ge, S., Edmunds, W.M., 2013. Offshore fresh
Appl. Geochem. 24, 1278–1286. groundwater reserves as a global phenomenon. Nature 504, 71–78.
Carrera, J., Vàzquez-Suñé, E., Castillo, O., Sanchez-Vila, X., 2004. A methodology to Pulido-Leboeuf, P., 2004. Seawater intrusion and associated processes in a small
compute mixing ratios with uncertain end-members. Water Resour. Res. 40, coastal complex aquifer (Castell de Ferro, Spain). Appl. Geochem. 19, 1517–
W12101. 1527.
Chae, G.T., Yun, S.T., Yun, S.M., Kim, K.H., So, C.S., 2012. Seawater-freshwater mixing Russak, A., Sivan, O., 2010. Hydrogeochemical tool to identify salinization or
and resulting calcite dissolution: an example from a coastal alluvial aquifer in freshening of coastal aquifers determined from combined field work,
eastern South Korea. Hydrol. Sci. J. 57, 1672–1683. experiments, and modelling. Environ. Sci. Technol. 44, 4096–4102.
Chapelle, F.H., McMahon, P.B., 1991. Geochemistry of dissolved inorganic carbon in Soler, A., Canals, A., Goldstein, S.L., Otero, N., Antich, N., Spangenberg, J., 2002. Sulfur
a coastal plain aquifer. 1. Sulfate from confining beds as an oxidant in microbial and strontium isotope composition of the Llobregat River (NE Spain): tracers of
CO2 production. J. Hydrol. 127, 85–108. natural and anthropogenic chemicals in stream waters. Water Air Soil Pollut.
Christophersen, N., Hooper, R.P., 1992. Multivariate analysis of stream water 136, 207–224.
chemical data: the use of principal components analysis for the end-member Stumm, W., Morgan, J.J., 1996. Aquatic Chemistry: Chemical Equilibria and Rates in
mixing problem. Water Resour. Res. 28, 99–107. Natural Waters. Wiley, New York.
Clark, I., Fritz, P., 1997. Environmental Isotopes in Hydrogeology. CRC Press, Lewis Tanabe, S., Hori, K., Saito, Y., Haruyama, S., Doanh, L.Q., Sato, Y., Hiraide, S., 2003a.
Pubishers, p. 328p. Sedimentary facies and radiocarbon dates of the ND (Nam Dinh)—1 core
de Montety, V., Radakovitch, O., Vallet-Coulomb, C., Blavous, B., Hermitte, D., Valles, sediments obtained from the Song Hong (Red River) delta, Vietnam. J. Asian
V., 2008. Origin of groundwater salinity and hydrogeochemical processes in a Earth Sci. 21, 503–513.
confined coastal aquifer: case of the Rhone delta (Southern France). Appl. Tanabe, S., Hori, K., Saito, Y., Haruyama, S., Vu, V.P., Kitamura, A., 2003b. Song Hong
Geochem. 23, 2337–2349. (Red River) delta evolution related to millennium-scale Holocene sea level
Fendorf, S., Michael, H.A., van Geen, A., 2010. Spatial and temporal variations of changes. Quat. Sci. Rev. 22, 2345–2361.
groundwater arsenic in South and Southeast Asia. Science 328, 1123–1127. Tanabe, S., Saito, Y., Vu, Q.L., Hanebuth, T.J., Ngo, Q.L., Kitamura, A., 2006. Holocene
Fritz, P., Fontes, J.Ch., 1980. Handbook of Environmental Isotope Geochemistry. The evolution of the Song Hong (Red River) delta system, northern Vietnam. Sed.
Terrestrial Environment. Elsevier, New York. Geol. 187, 26–61.
Fidelibus, M.D., 2003. Environmental tracing in coastal aquifers: old problems and Tubau, I., Vàzquez-Suñé, E., Jurado, A., Carrera, J., 2014. Using EMMA and MIX
new solutions. Coastal Aquifers Intrusion Technology: Mediterranean analysis to asses mixing ratios and to identify hydrochemical reactions in
Countries, vol. II. IGME Pub, Madrid, pp. 79–111. groundwater. Sci. Total Environ. 470–471, 1120–1131.
Fidelibus, M.D., Giménez, E., Morell, I., Tulipano, L., 1993. Salinization processes in Vàzquez-Suñé, E., Carrera, J., Tubau, I., Sánchez-Vila, X., Soler, A., 2010. An approach
the Castellon Plain aquifer (Spain). In: Custodio, E., Galofré, A. (Eds.), Study and to identify urban groundwater recharge. Hydrol. Earth Syst. Sci. Discuss. 7,
Modelling of Saltwater Intrusion Into Aquifers. Centro Internacional de Métodos 2543–2576.
Numéricos en Ingeniería, Barcelona, pp. 267–283. Vitòria, L., Otero, N., Soler, A., Canals, A., 2004. Fertilizer characterization: isotopic
Garing, C., Luquot, L., Pezard, P.A., Gouze, p., 2013. Geochemical investigations of data (N, S, O, C and Sr). Environ. Sci. Technol. 38, 3254–3262.
saltwater intrusion into the coastal carbonate aquifer of Mallorca. Spain. Appl. Walraevens, K., Van Camp, M., Van Der Kemp, W.J.M., Loosli, H.H., 2001. Pleistocene
Geochem. 39, 1–10. and Holocene groundwaters in the freshening Ledo-Paniselian aquifer in
Gattacceca, J.C., Vallet-Coulomb, C., Mayer, A., Claude, C., Radakovitch, O., Flanders, Belgium. Geol. Soc. Lond. Spec. Publ. 189, 49–70.
Conchetto, E., Hamelin, B., 2009. Isotopic and geochemical characterization of Wang, J., Tsay, T.K., 2001. Tidal effects on groundwater motions. Trans. Porous
salinization in the shallow aquifers of a reclaimed subsiding zone: The southern Media 43, 159–178.
Venice Lagoon coastland. J. Hydrol. 378, 46–61. Werner, A.D., Bakker, M., Post, V.E., Vandenbohede, A., Lu, C., Ataie-Ashtiani, B.,
Ghabayen, S.M.S., McKee, M., Kemblowski, M., 2006. Ionic and isotopic ratios for Simmons, C.T., Barry, D.A., 2013. Seawater intrusion processes, investigation
identification of salinity sources and missing data in the Gaza aquifer. J. Hydrol. and management: recent advances and future challenges. Adv. Water Resour.
318, 60–73. 51, 3–26.
Giménez-Forcada, E., 2010. Dynamic of sea water interface using hydrochemical Wigley, T.M.L., Plummer, L.N., 1976. Mixing of carbonate waters. Geochim.
facies evolution diagram. Ground Water 48, 212–216. Cosmochim. Acta 40, 989–995.
Giménez-Forcada, E., 2014. Space/time development of seawater intrusion: a study Winkel, L., Berg, M., Amini, M., Hug, S.J., Johnson, C.A., 2008. Predicting groundwater
case in Vinaroz coastal plain (Eastern Spain) using HFE-Diagram, and spatial arsenic contamination in Southeast Asia from surface parameters. Nat. Geosci.
distribution of hydrochemical facies. J. Hydrol. 517, 617–627. 1, 536–542.
Gomis-Yagües, V., Boluda-Botella, N., Ruiz-Bevia, F., 2000. Gypsum precipitation/ Winkel, L.H., Pham, T.K., Vi, M.L., Stengel, C., Amini, M., Nguyen, T.H., Pham, H.V.,
dissolution as an explanation of the decrease of sulphate concentration during Berg, M., 2011. Arsenic pollution of groundwater in Vietnam exacerbated by
seawater intrusion. J. Hydrol. 228, 48–55. deep aquifer exploitation for more than a century. Proc. Natl. Acad. Sci. USA 108,
Hem, J.D., 1985. Study and Interpretation of the Chemical Characteristics of Natural 1246–1251.
Water (3rd ed.). U.S. Geol. Surv. Water-Supply Paper 2254, U.S. Government Yamanaka, M., Kumagai, Y., 2006. Sulfur isotope constraint on the provenance of
Printing Office, Washington, DC, 263 p. salinity in a confined aquifer system of the southwestern Nobi Plain, central
Hooper, R.P., 2003. Diagnostic tools for mixing models of stream water chemistry. Japan. J. Hydrol. 325, 35–55.
Water Resour. Res. 39, W01055.