Environmental Forensics

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/uenf20

Determination of water quality assessment in

wells of the Göksu Plains using multivariate
statistical techniques

Esra Deniz Güner , Hatice Oncel Cekim & Galip Seçkin

To cite this article: Esra Deniz Güner , Hatice Oncel Cekim & Galip Seçkin (2020): Determination
of water quality assessment in wells of the Göksu Plains using multivariate statistical techniques,
Environmental Forensics, DOI: 10.1080/15275922.2020.1834025

To link to this article: https://doi.org/10.1080/15275922.2020.1834025

Published online: 17 Nov 2020.

Submit your article to this journal

Article views: 5

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=uenf20
ENVIRONMENTAL FORENSICS
https://doi.org/10.1080/15275922.2020.1834025

Determination of water quality assessment in wells of

€ ksu Plains using multivariate statistical techniques
the Go
€nera
Esra Deniz Gu , Hatice Oncel Cekimb, and Galip Seçkina
a
Environmental Engineering Department, Cukurova University, Adana, Turkey; bDepartment of Statistics, Hacettepe University,
Ankara, Turkey

ABSTRACT KEYWORDS
Groundwater is one of the most important sources of freshwater globally and has a key role Seawater intrusion; coastal
in sustaining the ecological value of many areas. These valuable and vulnerable ground- aquifer; multivariate
water resources are under the pressure of many external pollution threats, such as industrial, analysis; hierarchical cluster
analysis; factor analysis
domestic, and agricultural chemicals. To prevent undesirable consequences to future
groundwater resources, comprehensive and proper knowledge of groundwater quality and
contamination levels is vital at the local/regional scale. In this study, multivariate statistical
techniques, such as a correlation matrix, hierarchical cluster analysis, and factor analysis,
were applied to water quality data sets obtained from 24 wells located in the Go €ksu Plain.
These techniques were applied primarily to nineteen hydrochemical parameters collected
between May 2011 and April 2012. The results obtained from the correlation matrix showed
that the seawater descriptors such as EC, TDS, Cl
, Naþ, and Kþ were strongly correlated.
Principal component analysis indicated that most of the variations in groundwater were
caused by four factors that were responsible for the water quality, which represented more
than 80.22% of the total data variance. The hierarchical average linkage cluster analysis pro-
duced two major clusters that reflected seawater salinity.

Introduction fresh water (Kang and Jackson, 2016; Ros and

Zuurbier, 2017; Sae-Ju et al., 2020). In particular,
Groundwater resources are precious and particularly
because of excess groundwater extraction and increas-
vulnerable resources used in many areas such as irri-
ing salinization, non-renewable groundwater in the
gation of crops and pastures, industrial and drinking
coastal aquifer zone is endangered. The consequence
water consumption, and ecological services (Siebert
is that people face a serious drinking water crisis due
et al., 2010; De Stefano et al., 2015). Groundwater
to freshwater salinization in coastal aquifers (Talukder
resources are important alternative water sources in
et al., 2016; Rakib et al., 2020).
coastal areas. Coastal areas generally attract the major- Various methodologies such as numerical model-
ity of the population due to favorable living condi- ing, geophysical and hydrochemical techniques, obser-
tions and economic activities (agricultural, trade, and vations, lab experiments, and theoretical studies have
tourism). This usually causes an increase in water been conducted to assess seawater intrusion in coastal
demand for consumption in coastal areas. Therefore, aquifers (Jeong et al., 2010; Gong and Shen, 2011;
groundwater resources in coastal areas are under Werner et al., 2011; Zhang et al., 2011; Cheng et al.,
growing pressure from excessive groundwater extrac- 2012; Cai et al., 2015; Eissa et al., 2016; Kazakis et al.,
tion (Shi and Jiao, 2014; Morgan and Werner, 2015; 2016; Fadili et al., 2017; Najib et al., 2017). In add-
Qi et al., 2019). Thus, over-pumping of coastal aqui- ition, many different statistical analyses, such as multi-
fers may damage the seawater–freshwater interface variate statistical analysis, cluster analysis, principal
equilibrium, which in turn may cause seawater intru- component analysis (PCA), and factor analysis (FA)
sion (Huang et al., 2015; Liu et al., 2018; Tully et al., have been used successfully to determine the changes
2019; Hounsinou, 2020). Seawater intrusion in coastal in quality and pollution in surface water and ground-
alluvial aquifers causes groundwater quality deterior- water (Bhuiyan et al., 2011; Arslan, 2013; Belkhiri and
ation, salinization, low productivity, and reduction of Narany, 2015; Telahigue et al., 2018; Sae-Ju et al.,

CONTACT Esra Deniz G€

uner eguner@cu.edu.tr
ß 2020 Informa UK Limited, trading as Taylor & Francis Group
2 E. D. GÜNER ET AL.

2020). Additionally, statistical techniques can be suc- HCO3
-Cl
(slightly impacted), and (3) Naþ-Cl

cessfully applied to identify anthropogenic and geo- (highly impacted). Machiwal and Jha (2015) identified
genic phenomena influencing groundwater spatio-temporal variations in fifteen groundwater
hydrochemistry (Moyo, 2013; Seanego and Moyo, quality parameters in Udaipur district (India) by using
2013; Palma et al., 2014; Dragon and Gorski, 2015; PCA scores, box-and-whisker plots, and GIS-based
Souid et al., 2018). geostatistical modeling. Box-and-whisker plots indicate
Rakib et al. (2020) used instrumental and statistical that the hardness in groundwater exceeds its permis-
tools to determine the basic pollution parameters in sible limit at most sites in the aquifer system. The
26 groundwater samples from Shyamnagar Sub- results of HCA successfully classified 53 sites into two
District (Bangladesh). The results showed that the pol- clusters, in which cluster II is affected by natural geo-
lution could be divided into two main groups of fac- logic processes, and the groundwater of cluster I is
tors. One group contained salinization factors such as contaminated by anthropogenic sources. Telahigue
EC, TDS, Cl
, SO42
, NO2
, Br
, F
, Pb, Fe, As, et al. (2018) explained salination and saltwater intru-
Ca2þ, and temperature, while the other group sion using hydrogeochemical and multivariate statis-
included Zn, Mn, PO43
, and Li related to ground- tical analysis in the coastal aquifer. These authors
water quality. Busico et al. (2020) used multivariate identified two groups of water types as Ca-Mg-Cl-SO4
statistical and regression analyses to delineate the spe- in the middle of the island and Na-K-Cl on the
cific vulnerability to anthropogenic pollution (SVAP) nearby coast. Statistical analysis verified that the high
and weighting anthropogenic factors in the Florina groundwater mineralization was the result of seawater
Basin (Greece) and the Garigliano River Basin (Italy). intrusion and gypsum and carbonate dissolution.
Using these techniques, they successfully separated Magyar et al. (2013) applied multivariate statistical
groundwater resources into those related to seawater methods to determine water quality in Neusiedler See,
intrusion, interaction with igneous sediments, and in Austria. They stated that most of the variance in
anthropogenic impact by grouping the water samples the dataset could be explained by the main compo-
according to their characteristics. Selle et al. (2013) nents (ions) and eight different areas that were distin-
examined spatial and temporal patterns of principal guished. Using Wilk’s lambda statistics, it was verified
component (PC) scores to improve the understanding that the ions were mainly responsible for the forma-
of processes governing groundwater quality in the tion of the previously discussed cluster groups.
Ammer catchment located in southwest Germany. Previous studies in this area have focused mainly
The results delineated the influence of land use and on evaluating the groundwater chemical status in a
geology on groundwater quality. Kim et al. (2014) water framework directive (Demirel, 2007; Seckin
used model-based cluster analysis to understand the et al., 2010; Yilmaz et al., 2010). These authors con-
contributions of natural and anthropogenic factors to cluded that the deterioration in water quality in some
groundwater quality in South Korea. The results areas close to the sea was due to seawater intrusion
showed that the bivariate normal mixture model was and the intense use of groundwater for irrigation.
more valid than multivariate analysis, and provided However, these data have not been assessed using stat-
better identification between anthropogenic and nat- istical methods, and water quality parameters have not
ural groundwater groups. Souid et al. (2018) stated been classified according to importance levels.
that there is a strong relationship between seawater As a specially protected environmental area, the
intrusion in coastal aquifers and lithium by geochem- G€oksu Plain is an important natural reserve and one
ical modeling and multivariate statistical analysis. of the most important natural wetlands in the
According to the study results, seawater intrusion, Mediterranean region considering Turkey’s historical,
evaporation, carbonate dissolution, and cation cultural, and social values. The G€ oksu Plain was
exchange interactions between groundwater and inter- included in the 1994 statute of the Convention on
calated clay layers are the dominant factors in the Wetlands of International Importance (Ramsar) as a
aquifer materials. Sae-Ju et al. (2020) applied HFE- water habitat in order to aim to ensure the sustainable
Diagram and PCA to successfully explain intrusion use of these areas. Considering the bird species hosted
and freshening mechanisms in coastal aquifers in the in the past, this area is classified as a class “A” wet-
Cha-Am District in Phetchaburi Province (Thailand). land (according to the Ramsar Convention, waterfowl
According to the level of impact of seawater intrusion, wetlands hosting more than 25,000 species are class
the geochemical evolution of groundwater is governed “A” wetlands). Similar to the G€ oksu Plain, Silifke is
by three factors: (1) Ca2þ-Naþ-HCO3
and (2) Ca2þ- located on the Turkish Mediterranean coast and
 ENVIRONMENTAL FORENSICS 3

supports the production of more than twenty types of The G€ oksu Plain was formed by the deposition of
agricultural crops due to its high fertility characteris- alluvial material carried by the G€ oksu River. Alluvial
tics. With its productive alluvial soils and favorable soils show that the rocks were primarily transported.
climatic conditions, the G€ oksu Plain can produce a River plains such as the G€ oksu Plain usually consist of
wide range of products twice or thrice yearly. The high levels of calcareous structures. Therefore, the
general trends that provide insight into the structure majority of alluvial soil is limestone.
of the variables must be measured continuously and The Mediterranean climate is dominant in this
monitored systemically (Sargaonkar and Deshpande, region. Summer is hot and arid, and winter is rainy
2003; Simeonov et al., 2003; Giardino et al., 2007). and mild. The total annual precipitation is 606.8 mm,
In this study, the data set in the study by G€ uner the average temperature is 21  C, and the average
et al. (2018) was used. In this data set, the water qual- evaporation is 1745 mm. The average climatic data for
ity was analyzed using samples collected from 24 wells the study area are presented in Table 2.
in G€ oksu Plain. Over the period of a year (May 2011
- April 2012) a total of nineteen physical and chemical
Physical and chemical analyses
water quality parameters were analyzed. However,
when the seasonal and agricultural impacts of the Samples were collected monthly between May 2011
region are taken into consideration, the average and April 2012 from 24 wells located in the G€ oksu
annual monthly data collected in this study were used Plain. At each sampling station, the temperature (T;
to reflect the veracity of the results. 
C), pH, and electrical conductivity (EC; lS cm
1) of
Determining the status of water resources and devel- the water samples were measured in the field using
oping solutions to problems that arise will be easier WTW pH340 and 2510-A Orion instruments. The
using multivariate statistics. The intended purposes of amount of total dissolved solids (TDS) was determined
the study were as follows: (1) to collect monthly water by filtering the samples through Whatman (0.45 lm)
samples from the wells and classify the wells by water filter paper, followed by evaporation for 24 h at 150  C
quality according to their change throughout one year in pre-conditioned and pre-weighed crucibles.
and also classify hydrochemical parameters for changes The change in weight was used to determine the dis-
in the wells using cluster analysis. (2) To determine the solved salts. The alkalinity was analyzed by titration
basic parameters affecting the water quality in the wells (CO32
, HCO3
) using the alkalinity (2320)/titration
using FA and to identify these parameters in the saline S.2-35 standard method. For ion analysis, nitric acid
fields in the study area where they are more dominant was added to each water sample to ensure a pH <2.
using cluster analysis. (3) To identify field properties The bottles were tightly capped to prevent the intrusion
with the aid of diagrams through an evaluation of basic of atmospheric CO2. All samples were transported to
parameter fractions to determine whether there is any the Environmental Chemistry Laboratory of Cukurova
seawater intrusion. University and stored in a refrigerator at 4  C until ana-
lysis. All major cations, including Ca2þ, Mg2þ, Naþ,
Materials and methods Kþ, Si, and Ba were analyzed using a Perkin Elmer
highly sensitive inductively coupled plasma spectrom-
Study area eter (ICP), and anion concentrations for SO42
, NO3
,
The G€ oksu Plain is an important ecological and aes- and Cl
, were measured using ion chromatography
thetic resource along the Mediterranean coast. This plain (IC), following the Water Environment Federation and
provides fish and wildlife nurseries and habitat, forms American Public Health Association (2005). While per-
storm buffers, filters many waterborne contaminants, forming the analysis, each sample was analyzed in tripli-
and is important to coastal food chains. The G€ oksu cate and two sets of internal standards were run at the
Plain, which has a drainage area of 10,000 km2, is beginning and during the analysis, to check the accur-
located south of Silifke (Southern Anatolia). The loca- acy and precision of the results. For most of the ele-
tions of samples shown in Figure 1 and the geographical ments, the detection limits were approximately 1 ppb
positions of the wells are also given in Table 1. and the estimated analytical errors were within 5%.
The geological formations in this area, which have The sodium adsorption ratio (SAR) was used in the
different lithological properties, are aged from the first classification of water, and the residual sodium car-
to the third phase of its geological history, and most bonate (RSC) and %Na were calculated according to
of the area is covered by sedimentary rocks (medium the formulas below. All ionic values used in these for-
hard-soft) that were deposited during the Miocene. mulas are in meq/L.
4 E. D. GÜNER ET AL.

Figure 1. The locations of samples in the G€oksu Plains.

The SAR was calculated using the following Statistical analysis

formula (Raghunath, 1987):
The data were transformed to ensure suitability for
Naþ proper statistical tests. In many statistical tests, the
SAR ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (1)
Ca2þ þMg 2þ assumption of normal distribution should be pro-
2
vided; therefore, the data were logarithmically trans-
The RSC was determined as follows (Eaton, 1950): formed to eliminate the measurement unit differences

between the data. The Shapiro-Wilk test indicated
3 þ HCO3 Þ
ðCa þMg Þ
RSC ¼ ðCO2
2þ 2þ
(2)
that the P-value (¼99%) exceeded 0.01, indicating that
The %Na was determined as follows (Wilcox, 1955): the data were normally distributed. Similarly,
100  Naþ Kaiser–Meyer–Olkin (KMO) and Bartlett’s tests were
%Na ¼ (3) used to determine whether the data were suitable for
Naþ þ Ca2þ þ K þ þ Mg 2þ
FA analysis. The KMO value was 0.50, and Bartlett’s
 ENVIRONMENTAL FORENSICS 5

Table 1. Geographical positions of the wells in the G€oksu Plain.

Location-well Elevation (m) Latitude (North) Longitude (East)
G-01 6.0 36˚ 210 41.01ʺ 34˚ 10 4.04ʺ
G-02 7.0 36˚ 210 10.04ʺ 34˚ 10 8.15ʺ
G-03 43.0 36˚ 200 59.56ʺ 34˚ 10 18.99ʺ
G-04 44.0 36˚ 210 1.55ʺ 34˚ 10 54.37ʺ
G-05 42.0 36˚ 200 32.61ʺ 34˚ 20 30.95ʺ
G-06 8.0 36˚ 210 48.83ʺ 34˚ 20 44.69ʺ
G-07 32.0 36˚ 220 13.45ʺ 34˚ 20 24.56ʺ
G-08 17.5 36˚ 220 15.98ʺ 34˚ 20 45.88ʺ
G-09 40.0 36˚ 220 22.35ʺ 34˚ 20 56.72ʺ
G-10 41.0 36˚ 200 44.84ʺ 34˚ 10 59.84ʺ
G-11 43.0 36˚ 220 18.55ʺ 34˚ 20 58.6ʺ
G-12 65.0 36˚ 200 11.47ʺ 34˚ 30 36.5ʺ
G-13 60.0 36˚ 200 11.16ʺ 34˚ 30 37.07ʺ
S-01 55.0 36˚ 210 50.77ʺ 33˚ 590 17.01ʺ
S-02 30.0 36˚ 210 50.66ʺ 33˚ 590 48.69ʺ
S-03 75.0 36˚ 200 17.81ʺ 33˚ 590 56.73ʺ
S-04 35.0 36˚ 190 49.19ʺ 34˚ 00 16.49ʺ
S-05 49.0 36˚ 180 25.09ʺ 34˚ 00 3.74ʺ
S-06 30.0 36˚ 200 53.71ʺ 33˚ 570 29.18ʺ
S-07 38.0 36˚ 200 31.59ʺ 33˚ 570 39.08ʺ
S-08 30.0 36˚ 190 59.36ʺ 33˚ 560 39.26ʺ
S-09 42.0 36˚ 200 9.02ʺ 33˚ 550 18.82ʺ
S-10 4.0 36˚ 190 1.88ʺ 33˚ 550 10.45ʺ
S-11 12.5 36˚ 190 13.33ʺ 33˚ 540 55.87ʺ

Table 2. The annual average values of meteorological param- analysis (HCA) was used in this study, the Euclidean
eters in the G€oksu Plain between May 2011 and April 2012. distance and Ward’s clustering algorithm were
G€uner et al. (2018). selected, and the drainage channels of the data were
Temperature ( C) Relative humidity (%) Precipitation (mm) converted into standardized Z scores. All of
Maximum 25.4 77.1 6.8 the mathematical and statistical computations were
Minimum 17.0 30.1 0.3
Average 21.0 55.5 1.8 performed using IBM SPSS Statistics 23 version
(License server: spss.hacettepe.edu.tr).

test of sphericity resulted in a chi-square value of Results and discussion

723.401 (P value ¼ 0.000). A descriptive statistical overview of the hydrochemical
Combining many variables associated with each data of the groundwater, which was sampled from
other, FA is a statistical method that allows for the iden- May 2011 to April 2012 and includes major cations
tification of few, new, unrelated, or conceptually signifi- and anions, is presented in Table 3. The annual distri-
cant variables (factors, dimensions). Here, the most butions of the water quality were evaluated with the
important factor is determining the number of PCs. The minimum, maximum, mean, and standard deviation
calculation of the stated variance and deciding on the values for the considered variables. The water samples
number of components (factors) with eigenvalues of 1 obtained from the wells in the study area had pH
or larger than 1 are both considered important factors. values ranging from 7.5 to 8.19 with mean ± SD
FA is a multivariate statistical technique to obtain a of 7.82 ± 0.2, which indicates slight alkalinity. It is
smaller number of meaningful variables from a larger understood that the pH of seawater is >7, which is
number of variables and is intended to measure the weakly alkaline. The relatively low alkalinity associated
same structure or property. The factors obtained in this with groundwater in the G€ oksu Plain zone is caused
analysis are assessed as strong if f > 0.75, medium if by seawater intrusion. The well water samples were
0.50 < f < 0.75, and weak if 0.30 < f < 0.50. The import- moderately hard with high EC, and TDS varied over
ant factors obtained in this study were subjected to a wide range of 265.5 to 5677.5 mS cm
1 and 156.6
Varimax rotation to provide independence, significance, to 3941.3 mg/L with an overall mean ± SD
and clarity in the interpretation. of 807.8 ± 839.6 mg/L and 1471.7 ± 1289.9 mS cm
1,
The Pearson coefficient (r) was used to evaluate the respectively. The average annual values Ca2þ and
statistical significance of the measured correlation val- Mg2þ ranged from 15.83 to 136.43 mg/L and 13.4 to
ues and to provide information on the strength and 125.6 mg/L, with mean ± SD of 62.44 ± 36.05 and
direction of linear relationships. Hierarchical cluster 39.53 ± 26.79 mg/L, respectively. The Ca2þ content was
6 E. D. GÜNER ET AL.

Table 3. Analytical results of the water quality parameters from the G€oksu Plain.
Sample point T PH EC TDS Ca2þ Mg2þ Kþ Naþ Cr SO42
NO

3 HCO3
CO32
Ba Si B
G-01 20.69 7.68 1014 605.15 88.61 51.91 5.05 31.6 94.8 228.78 12.39 347.9 5.93 0.72 2.39 0.78
G-02 20.18 7.66 1324.58 796.88 100.43 48.04 5.69 77.5 127.74 251.29 12.2 381.62 2.58 0.5 2.56 0.89
G-03 20.28 8 667.42 574.69 85.52 45.52 4.89 88.2 212.77 169.26 12.13 191.13 17.6 0.19 2.12 0.65
G-04 20.38 7.96 883.67 392.82 78.23 20.56 4.88 108.7 94.02 226.57 12.14 211.25 17.12 0.26 2.37 0.64
G-05 21.2 7.6 1632 817.65 15.83 14.77 7.43 282.03 308.89 171.98 12.36 311.03 5.58 0.23 3.87 0.88
G-06 22.75 7.63 1208.5 728.74 63.66 48.29 14.66 89.34 98.5 253.67 12.21 453.72 2.05 0.46 2.92 0.88
G-07 21.23 8.02 1106.17 555.87 43.61 63.39 11.69 191.67 171.61 321.53 12.12 285.65 11.43 0.23 2.85 0.83
G-08 20.81 8.19 1014 523.32 21.39 20.46 13.59 134.25 136.69 138.67 12.42 240.43 22.42 0.22 3.15 0.88
G-09 20.35 8.02 1171.43 548.93 19.36 21.55 13.85 149.1 154.93 150.99 12.06 251.86 11.78 0.06 3.38 0.8
G-10 20.39 7.99 832.08 405.14 21.55 19.28 6.19 110.4 111.92 124.11 12.1 229.03 14.43 0.25 2.63 0.68
G-11 20.92 8.12 1104.92 551.18 28.07 18.22 11.3 155.93 123.13 245.67 12.51 264.87 21.38 0.2 3.33 0.87
G-12 22.3 7.68 5677.5 3941.3 124.8 125.64 34.46 880.11 1597.57 299.78 12.41 211.58 2.33 0.49 2.26 1.31
G-13 22.33 7.75 4629.17 2695.07 100.82 96.22 31.76 741.76 1279.42 207.76 12.94 233.98 4.98 0.47 2.14 1.72
S-01 20.74 7.75 753.33 322.7 56.28 31.54 3.27 25.72 112.64 105.61 12.56 233.55 9.55 0.4 3.47 0.64
S-02 20.71 7.71 949.75 488.31 136.43 64.44 4.2 42.05 185.13 249.97 12.73 289.22 5.08 0.94 2.24 0.69
S-03 20.53 7.83 1429.33 610.05 44.36 44.39 3.02 169.25 332.06 123.28 13.02 205.2 7.63 0.69 2.61 0.64
S-04 22.68 7.94 1277.58 568.92 37.35 27.65 5.99 149.6 254.34 130.78 13.31 196.45 13.8 0.38 2.58 0.72
S-05 22.04 8.18 3352.5 1517.11 36.68 29.98 20.44 451.96 543.51 223.46 13.34 267.6 23.1 0.37 3.89 1.44
S-06 21.6 7.84 265.57 156.61 31.57 13.4 4.66 411.14 100.51 302.31 12.33 77.75 9.88 0.14 0.8 0.41
S-07 20.91 7.91 601.25 256.4 45.06 25.96 2.41 31.8 72.02 114.17 12.47 203.53 9.97 0.21 2.36 0.55
S-08 20.68 7.57 633.42 308.15 36.86 16.64 2.71 53.44 77.32 115.7 12.54 221.73 2.6 0.7 1.67 0.64
S-09 23.13 7.56 924.75 528.17 96.81 30.69 6.32 19.51 123.47 160.68 12.68 260.67 2.57 0.43 2.03 0.55
S-10 21.34 7.67 2041.58 994.67 74.45 34.29 5.11 255.51 362.65 205.84 12.5 273.57 2.58 0.29 1.99 0.78
S-11 22.37 7.5 827.33 501.29 110.75 35.84 2.51 23.54 145.61 181.38 12.32 305.5 2.48 0.29 1.66 0.58
The water temperature (T) is in  C, electrical conductivity (EC) is in mS cm
1 and the other parameters are in mg/L.

higher in the limestone bedrocks and the Mg2þ Naþ, Cl
, and NO3
in the study area were classified as
content was higher in the dolomite area (Jiang et al., class 3 (polluted status) (Table 4). The TDS and SO42

2009; Sae-Ju et al., 2020). As shown in Table 3, there values included low pollution (class 2), while the other
was more limestone than dolomite in the study area parameters showed clean water quality (class 1). The
because Ca2þ and Mg2þ exhibited the same trends in presence of Naþ, Cl
, NO3
, and SO42
can be
concentration. The Naþ and Kþ values ranged from explained as the result of agricultural, domestic waste-
19.51 to 880.1 mg/L and 2.41 to 34.46 mg/L, with water, and possibly seawater intrusion.
mean ± SD of 194.75 ± 222.5 and 9.42 ± 8.65 mg/L, Box-and-whisker plots (Figure 2) were used to
respectively. Naþ was relatively abundant compared to display statistical summaries of the major elements in
Ca2þ. This may indicate very high TDS, hardness, and the groundwater samples. The central box represents
alkalinity of the water samples. the values from the lower to upper quartile (25th to
From a careful observation of these data, it is clear 75th percentile). The middle line represents the
that the Naþ, Mg2þ, Cl
, and SO42
ions showed median. The vertical line extends from the minimum
a wide range of distributions and higher standard to the maximum values.
deviations (SO42
and Cl
have mean ± SD of The relationships between the measured
195 ± 64.83 mg/L and 284.2 ± 375.4 mg/L, respectively). hydrochemical parameters were obtained using correl-
This suggests a possible incursion of nearby saline ation analysis (Table 5). Pearson correlation coefficients
water, which has comparatively higher concentrations of 0.9–1, 0.7–0.89, 0.5–0.69, 0.26–0.49, and 0–0.25
of Naþ, Mg2þ, Cl
, and SO42
(Table 3). Higher EC between the two variables indicate very good, good,
and TDS values were recorded in wells located near moderate, weak, and very weak correlations, respectively.
the sea. However, another reason for the high concen- In general, high temperatures tend to increase
trations of Mg2þ is that limestone, which is dominant the dissolution quantities of certain ions, such as Kþ,
in this region, contains high amounts of Mg2þ. Naþ, Cl
, and SiO2 (Zhu and Yu, 1995; Edmunds
Water quality was classified into four classes accord- et al., 2006; Sung et al., 2012). However, in this study,
ing to the Turkish Water Pollution Control Regulation there was a weak and wery weakly correlation between
(Anonymous, 1988). The water quality of class 1 is con- the water temperature and the other parameters (EC,
sidered to be clean; the water quality of class 2 corre- Kþ, Mg2þ, Naþ, and Cl
); the absolute value of the
sponds to a low pollution status; the water quality of correlation coefficient was usually less than 0.4. These
class 3 corresponds to a polluted status, and the water results are consistent with those of previous studies.
quality of class 4 is considered to be highly polluted. Rakib et al. (2020), Altın et al. (2009), Yusuf et al.
According to the water quality based on the class guide- (2013) found that the water temperature was very
lines and the annual average concentrations (Table 4), weakly correlated with other water quality parameters
 ENVIRONMENTAL FORENSICS 7

Table 4. Statistical summary of the hydrochemical parameters of groundwater from the G€oksu Plain (N ¼ 24).
Water quality guideline
Class
Parameter Mean Max. Min. SD I II III IV
T (˚C) 21.27 23.13 20.18 0.9 25 25 30 >30
pH 7.82 8.19 7.50 0.2 6.5–8.5 6.5–8.5 6.0–9.0 <6.0–9.0<
EC (mS/cm) 1471.74 5677.50 265.57 1289.9
TDS (mg/L) 807.88 3941.30 156.61 839.66 500 1500 5000 >5000
Ca2þ (mg/L) 62.44 136.43 15.83 36.05 — — — —
Mg2þ (mg/L) 39.53 125.64 13.40 26.79 — — — —
Kþ (mg/L) 9.42 34.46 2.41 8.65 — — — —
Naþ (mg/L) 194.75 880.11 19.51 222.53 125 125 250 >250
Cl
(mg/L) 284.22 1597.57 72.02 375.40 25 200 400 >400
SO42
(mg/L) 195.97 321.53 105.61 64.83 200 200 400 >400
NO3
(mg/L) 12.49 13.34 12.06 0.36 5 10 20 >20
HCO3
(mg/L) 256.20 453.72 77.75 73.22 — — — —
CO32
(mg/L) 9.54 23.10 2.05 6.88 — — — —
Ba (mg/L) 0.38 0.94 0.06 0.21 1 2 2 >2
Si (mg/L) 2.55 3.89 0.80 0.73 — — — —
As (mg/L) 0.02 0.03 <LD 0.01
B (mg/L) 0.81 1.72 0.41 0.3 — — — —
SAR 5.13 17.41 0.45 4.97 — — — —
RSC
1.93 3.26
13.16 3.73 — — — —
%Na 47.59 86.27 10.71 24.60 — — — —

Figure 2. Box-and-whisker plots of variability in trace element concentrations in the water samples (EC, TDS, Cl
, CO32
, alkalinity,
Ca2þ, Mg2þ, Hardness, Kþ, Naþ, SO42
, SAR, Ba, Si, and B, respectively).

such as alkalinity, EC, TOC, turbidity, total Kjeldahl The strong positive correlations between TDS,
nitrogen (TKN), DOC, salinity, and BOD5, which is hardness, Ca2þ, and Mg2þ indicate that the bedrock
similar to our findings. was derived from the dissolution of carbonate and
 8 E. D. GÜNER ET AL.

NO3
RSC dolomite (Aiuppa et al., 2003; Jiang et al., 2009; Qin

0.613
0.880
0.859
0.146
0.503 0.497
0.090 1

et al., 2013). Similar to the results of this study, Qin
et al. (2013) reported relatively weak positive correla-

0.086
0.204 0.460 0.111 1

tions between HCO3
, Ca2þ, and Mg2þ, which were
Si

due to the lamination and heterogeneous distribution

0.325
0.325
0.160
0.204
0.040 1

of carbonate rock and dolostone around the sampling
Ba

0.574 0.566 0.035 1

areas. Additionally, the strong positive correlation
between TDS and Naþ and Cl
can be interpreted as
SO42

1
paleo-seawater intrusion, and an important factor in

0.417
the assessment of groundwater chemistry is the stor-
Hard

age of minerals with precipitation/evaporation proper-

0.938 1
ties (Rakib et al., 2020; Sae-Ju et al., 2020).
Ca2þ

0.364

0.059
The correlation matrix analysis showed that there
was very good correlation between Naþ and Kþ (0.702).

0.548 1

0.508

0.488
This positive correlation may indicate that these ions in
CO32

0.175

0.088
the groundwater originated from the same sources. The
1

reason that Naþ and Kþ are introduced into grassland,

0.612
0.923
0.370
0.220
0.292
0.098
0.338

0.066
0.104
0.362
0.383

forest, and cultivated land is due to the use of fertilizers

Alk

for agricultural activities and anthropogenic causes, such

as wastewater discharge from residential areas.

0.512

0.410

0.505

0.277

0.041

0.366
Ph

The EC of water depends on the total concentra-

1
Table 5. Correlation coefficient matrix of the chemical data for groundwater from the G€oksu Plain (N ¼ 24).

tion, mobility, valence, and temperature of the ions.

0.427

0.316
0.045

0.382
0.247
0.271
0.219
0.171

0.163

0.522

The EC of water increases as T increases, and the

T

concentration of the ions in the water increases. In

0.146
0.433
0.306 0.410

0.148
0.472
0.254 0.093

0.366
0.371
0.004
0.410
0.358 0.295

0.147
0.314
0.156 0.292
0.701 0.097
0.348
0.101
0.009 0.053
0.302
0.386 0.152
0.187
0.699
0.594 0.658

this study, there was a weak correlation between EC

Sar

0.620 0.964 0.472
0.040
0.205 1

and T (0.348). Therefore, the EC value is a good indi-

Correlation is significant at the 0.05 level (2-tailed).  Correlation is significant at the 0.01 level (2-tailed).

cator of the total ion concentration in water.

Zn

Accordingly, the EC represents a qualitative estimate

of inorganic pollution. Information on TDS can be
0.200

0.071

0.258

0.257
0.339
0.018
Mn

considered a good indicator of pollution. The EC was

1

positively correlated with water salinity, ions such as

0.052
0.093
0.175
0.120

0.151

0.168
0.298
0.198

0.084
0.064
0.239
0.009

0.170
0.077
0.180
0.135
0.208
0.207
0.033
0.027

Cl
, Mg2þ, B, Kþ, Naþ, Si, and parameters such as

As

0.559 0.627 0.682 0.534 1

SAR, hardness, and TDS (see Table 5). The high cor-
0.415

relation implies that groundwater chemistry in the

0.054

0.168
0.322

0.173

0.165
0.053

0.029
0.184
0.470
Naþ

0.732 0.607 0.702 1

study area was mainly controlled by these parameters.

In addition, this correlation may be considered as due
0.466
0.082

0.370
0.303

0.082

0.351
0.087
0.032
Kþ

to the influence of seawater intrusion. Similarly, the

0.638 0.835 1

strong correlation between Naþ and Cl
and that

between Naþ and B may be interpreted as a result of
0.579 0.488

0.412

0.507
0.316
0.127
0.212
0.027

0.230
0.159

CO32

0.203
0.273
0.403
0.128
0.022

0.307
0.309

0.290
0.017
B

seawater intrusion. Water containing Naþ and Cl

0.740 1

ions in high concentrations is characterized as saline

0.531 0.926 0.426

0.443
0.220

0.375

0.026
0.091
0.251
0.012
0.327
0.018

0.742 0.227

0.255
0.174
0.136

0.332
Cl

water, whereas water containing relatively high Ca2þ

0.872 0.884 0.590 1

and HCO3
concentrations in addition to other major

0.904 0.891 0.515

0.420

0.502

0.031
0.721
Mg2þ

ions is characterized as freshwater (Appelo and

0.725 0.750 0.387
0.591 0.582 0.157
0.566 0.575 0.345

0.406
0.099
0.258

0.402

0.040
0.127

Postma, 2005; Sae-Ju et al., 2020). Naþ and Cl
were

Mg2þ 0.602 0.683 1

the dominant ions in areas close to the coast. The

0.302
0.003

0.376

0.371

0.331

0.350
0.270
0.328
0.281

strong correlation between Naþ and Cl
showed the

TDS

0.969 1

effect of seawater–groundwater mixing. As shown in

the correlation matrix, Kþ can be considered as it is
0.439

0.431

0.438
0.304

0.092

0.348

0.382

0.223

SO42
0.251
0.273

NO3
0.387
0.096
EC

also associated with seawater intrusion and it has a

1

positive correlation with B and Cl
. Because seawater

Ca2þ
Hard
Naþ
TDS

RSC
Cl

Mn

Alk
Sar

intrusion is the result of cation exchange, the Ca2þ

Kþ

Zn

Ph
EC

Ba
As

Si
B

T
 ENVIRONMENTAL FORENSICS 9

concentration was lower than that of Naþ in the quality. While determining the number of PCs to be
groundwater where the seawater intruded. rotated, one of the methods other than eigenvalue sta-
According to the correlation coefficient matrix, tistics is the scree plot (Figure 3). In this graph, the
there was a weak correlation between Si and pH, number of PCs determined by the point at which the
which may indicate that silicate mineral weathering slope begins to disappear is determined. Accordingly,
processes are related to increases in pH. This conclu- four PCs were identified.
sion is supported by the statistical and graphical The high correlation between NO3
and T may be
methods and hydrochemical analyses performed in 23 associated with pollution that occurred as a result of
wells in the Shagari area by Owoseni et al. (2013). intense agricultural activities in the surrounding
The variance explanation rates of the hydrochemi- region. The groundwater quality in the areas sur-
cal components are listed in Table 6. For example, it rounding Kayseri agricultural activities demonstrated
indicates that 95% of the variable TDS is explained by high concentrations of NO3
and NaCl that caused
the PCs. Here, variables with low common variance contamination due to untreated sewage and agricul-
(below 0.50) were excluded from the analysis and FA tural waste during spring (Elhatip et al., 2003; Baba
was re-performed. In this case, both the KMO value and Ayyildiz, 2006).
and the value of the variance explanation statistic Although Re et al. (2013) and El Yaouti et al.
reached higher values. Accordingly, T and Mn compo- (2009) detected relatively low Cl concentrations in
nents were excluded from the analysis and when the their underground water studies; they noted that there
analysis was repeated, the variance explanation rate was extremely high enrichment of NO3
ions. The
for all variables in the analysis was greater than 0.5. samples from our study could be affected by the input
In this study, FA was applied to detect the parame- of manure-derived NO3
or septic effluents.
ters that primarily contributed to the changes in water Alternatively, high Cl
and low NO3
contents were
measured in a specific area of the aquifer and were
attributed to seawater intrusion. It was observed that
Table 6. Variance explanation rates using PCA for hydrochem-
ical data from the G€oksu Plain. there was statistically moderate correlation (r ¼ 0.443)
Initial Extraction Initial Extraction Initial Extraction between the changes in the Cl
and NO3
contents
EC 1 .94 As 1 .50 CO32
1 .68 measured in the wells throughout the year in the
TDS 1 .95 Mn 1 .48 Ca2þ 1 .81 study conducted in the G€ oksu Plain.
Mg2þ 1 .79 Zn 1 .52 Hardness 1 .83
Cl
1 .90 SAR 1 .83 SO42
1 .56 For the scree plot and eigenvalues greater than or
B 1 .94 T 1 .26 Ba 1 .81 equal to 1, four PCs were selected that explained
Kþ 1 .89 pH 1 .67 Si 1 .91
Naþ 1 .87 Alk 1 .92 NO3
1 .83
80.22% of the total variance of the data set. The par-
RSC 1 .77 ameter loadings for the four PCs from the FA are pre-
Extraction method: principal component analysis. The bold values indicate sented in Table 7, and those marked in bold are
variables with low common variance (below 0.50).
considered strong values. The terms strong, moderate,

Figure 3. Scree plot of eigenvalues using the principal components for the hydrochemical data from the G€oksu Plain.
10 E. D. GÜNER ET AL.

Table 7. Varimax rotated factor matrix for hydrochemical data Table 8. Component score covariance matrix.
from the G€oksu Plain. Component 1 2 3 4
PC 1 1.000 0.000 0.000 0.000
2 0.000 1.000 0.000 0.000
1 2 3 4 3 0.000 0.000 1.000 0.000
Kþ 0.899
0.097 0.211
0.172 4 0.000 0.000 0.000 1.000
Naþ 0.874
0.254
0.254 0.042
Extraction method: principal component analysis. Rotation method:
Cl
0.860 0.252
0.017 0.324
Varimax with Kaiser Normalization. Component Scores.
TDS 0.847 0.363 0.311 0.103
B 0.844 0.088 0.454 0.104
EC 0.825 0.268 0.382 0.229
SAR 0.756
0.475
0.263 0.040 Table 9. Component score coefficient matrix.
As 0.708
0.009
0.023 0.001 Component
SO42
0.474 0.269
0.197
0.473
Ca2þ 0.034 0.918
0.073
0.020 1 2 3 4
Hard 0.269 0.892 0.030
0.016 EC 0.102 0.024 0.122 0.084
CO32
0.042 20.777
0.027 0.244 TDS 0.113 0.042 0.088 0.018
Mg2þ 0.487 0.749 0.140 0.011 Mg2þ 0.054 0.134 0.014
0.001
pH 0.242 20.743 0.026 0.152 Cl
0.119 0.037
0.052 0.135
Ba
0.082 0.728 0.230 0.374 B 0.114
0.019 0.160 0.011
RSC
0.601 20.613 0.245
0.031 Kþ 0.153
0.057 0.061
0.144
Alk
0.003 0.314 0.885
0.233 Naþ 0.160
0.065
0.135
0.030
Si 0.205
0.314 0.862 0.129 As 0.119
0.020
0.037
0.038
NO3
0.173 0.101 0.080 0.857 Zn 0.003 0.019 0.116
0.356
Zn
0.104 0.221 0.270 20.682 SAR 0.147
0.107
0.126
0.030
Eigenvalues 6.539 5.164 2.410 1.931 pH 0.052
0.155 0.029 0.048
Total % of Vaiance 32.694 25.821 12.050 9.657 Alk
0.031 0.025 0.374
0.121
Cumulative % 32.694 58.515 70.565 80.222 CO32
0.016
0.151 0.014 0.108
The bold values indicate PCs to which hydrochemical parameters in terms Ca2þ
0.016 0.188
0.067 0.013
of loading scores. Hard 0.020 0.172
0.031 0.001
SO42
0.104 0.036
0.112
0.271
Ba
0.063 0.152 0.074 0.224
and weak, which were applied to the factor loadings, Si 0.002
0.098 0.377 0.045
refer to the absolute loading values of >0.75, 0.75–0.5, NO3

0.023 0.036 0.020 0.454
RSC
0.091
0.118 0.153
0.002
and 0.5–0.3, respectively. From Table 7, it can be seen
that the first PC (PC1) shows a strong and positive
correlation with K (0.899), Naþ (0.874), Cl
(0.860),
TDS (0.847), B (0.844), EC (0.825), and SAR (0.756), presence of rocks, such as carbonate and dolostone
a moderate positive correlation with As (0.708), and a around the sampling areas, can be attributed to the
weak positive correlation with SO42
(0.474); these dissolution of soil and minerals. pH and CO32
are
variables explain 32.694% of the total variance. The two important variables controlled by nat-
second PC explained 25.821% of the total variance, ural processes.
while having a strong positive relationship with Ca2þ, Several different studies have shown that the source
hardness, and Mg2þ, and a moderate positive relation- of nitrate pollution may be the result of different agri-
ship with Ba. It has a strong negative relationship cultural practices and hydrogeological structures of a
with CO32
, and a moderate negative relationship region (Kim et al., 2005; Grassi et al., 2007; Stigter
with pH and RSC. Three PCs explained 70.565% of et al., 2008; Zhao et al., 2010; Landon et al., 2011;
the variance, and PC3 is correlated with Alk and Si G€uler et al., 2012; Qin et al., 2013). Rekha et al.
with a high positive degree. The variance explanation (2011) characterized the leaching of NO3–N into shal-
rate of the last PC (PC4), which was positive for low groundwater from agricultural fields under differ-
NO3
and negative for Zn, was 9.657%. ent cultivation practices. Taking into consideration
According to the results in Table 8, the fact that the intensive agricultural activities in this area and the
the covariance values are zero provides the assump- commercial composites that have been applied for a
tion that the factors are unrelated. The factor scores long period of time, the high score of PC4 may be
in the component column in Table 9 have a normal due to agricultural fertilizers. In an area similar to our
distribution and do not have multicollinearity prob- study area (Figure 4) (a landscape with intensive agri-
lems. The obtained factor scores can be used as varia- culture, close to the sea and river), Kim et al. (2005)
bles in other analyses. reported that under toxic conditions, the main com-
It is well known that hardness is correlated with ponent of fertilizers, NH4þ, is easily oxidized into
Ca2þ and Mg2þ contents (Reghunath et al., 2002; NO3
by nitrification. These authors stated that the
Papatheodorou et al., 2006). This factor accounted for part that constituted 6.3% of the total variance was
the temporary water hardness. Additionally, the primarily associated with a high positive loading for
 ENVIRONMENTAL FORENSICS 11

Figure 4. A map of land use in the G€oksu Plain. Source: G€uner, 2016.

NO3
and negative moderate loadings for pH concentrations if the water contains a low concentra-
and HCO3
. tion of sulfate.
Ba is present as a trace element in both igneous It has been stated that the primary reason for
and sedimentary rocks. This element is not found in a nitrate intrusion into shallow water is the fertilizer
free form in nature and occurs in a number of com- used to increase crop productivity (Krapac et al.,
pounds, most commonly in the form of barite 2002; Psychoyou et al., 2007; Jalali, 2011). The wetland
(BaSO4) and, to a lesser extent, witherite (BaCO3). of the G€oksu Plain has a particularly high amount of
The acetate, nitrate, and salts of barium are soluble in green vegetable production.
water, but the carbonate, chromate, fluoride, oxalate, Two separate dendrograms were created for the
phosphate, and sulfate salts are quite insoluble. The water quality parameters considered and the hydro-
aqueous solubility of barium compounds increases as chemical sampling sites (Figure 5). The clusters are
the pH decreases. The concentration of barium ions obtained using Ward’s method with Euclid’s metric of
in natural aquatic systems will be limited by naturally the square. From Figure 5, according to the CA, two
occurring anions, such as sulfates and carbonates, and major clusters are identified for the measured water
by the possible adsorption of barium ions onto metal quality parameters. In Figure 5, it is clear that the first
oxides and hydroxides. Barium can only reach toxic cluster contained pH, NO3
, RSC, T, Kþ, CO32
,
12 E. D. GÜNER ET AL.

Figure 5. A dendrogram showing the water quality parameters of the G€oksu Plain (y-axis).

Ca/Mg ratios in the G-01, G-02, G-03, G-04, S-01, S-

Table 10. Geochemical criteria for distinguishing the origin
02, S-06, S-07, S-08, S-09, S-10, and S-11 wells were
of saltwater.
higher because Mg2þ was considerably more abundant
Measurement Criteria
¼0.86 seawater than Ca2þ (Figure 6).
<0.86 seawater intrusion According to the values in Table 10 and Figure 6,
Na/Cl ratio >1.0 anthropogenic sources (e.g., sewage)
Ca/Mg >1.0 seawater intrusion
the Na/Cl ratios of the G-01, G-03, G-12, S-01, S-02,
Ca/(HCO3 þ SO4) >1.0 seawater intrusion S-03, S-07, S-09, and S-11 wells implied greater sea-
water intrusion compared to the other sampling wells.
Mg2þ, Ca2þ, B, Ba, SAR, Si, Mn, Zn, and As, and the In addition, the other sampling stations were slightly
second cluster contained EC, TDS, alkalinity, SO42
, below the 1:1 line in the Na/Cl plot, indicating the
hardness, Cl
, and Naþ. absence of clear geochemical evidence of seawater
To distinguish the seawater intrusion from the intrusion. It should be noted that the G-01, G-03, S-
other salinity resources in coastal aquifers, the geo- 01, S-02, and S-03 wells are closer to the G€oksu River,
chemical ratios of Na/Cl (a limit of 0.86), Ca/Mg (a and the other wells (G-12, S-07, S-09, and S-11) may
limit of 1.00), and Ca/(HCO3þSO4) (a limit of 1.00) show that seawater intrusion has extended inland. The
have been used (Al-Khatib and Al-Najar, 2011; Sherif Ca/Mg ratio may indicate the possibility of greater
et al., 2011; Re et al., 2013) (Table 10). The values of seawater intrusion than on the Silifke side.
the geochemical ratios calculated for this study are The SAR ratio is an important reference for the
shown in Figure 6. In general, freshwater was domi- classification of water regarding the salinity effect.
nated by calcium, and seawater was dominated by Most of the samples were in the best class for the
magnesium. The Ca/Mg ratio provided an indicator of SAR ratio. In contrast to the general trend of the sam-
the contaminated seawater–freshwater interface. The pling area, the G-05, G-11, G-12, S-05, and S-06 wells
 ENVIRONMENTAL FORENSICS 13

Figure 6. Geochemical ratios for 24 wells.

Table 11. Classification of water according to the %Na value in Table 4, based on the RSC, only G-05 Class II is
(Sahinci, 1991). suitable for irrigation, and the other samples are first-
%Na Irrigation water classification class irrigation water. The %Na value is an important
<20 Very well
20–40 Well
parameter for groundwater, which indicates whether
40–60 Good the water is suitable for use as irrigation water in agri-
60–80 Questionable use cultural areas.
According to the classification provided in Table 11,
were higher than the SAR ratio of 12, which can cause the maximum %Na value was identified at the G-05
serious physical problems and difficulties for plants to sampling point in the study area of G€ oksu, which was
absorb water. classified as “not suitable.” The other sampling points
It has been concluded from studies that water con- were classified as “questionable use” (G-04, G-07, G-08,
sistently containing more than 2.5 meq/L sodium car- G-09, G-10, G-11, G-12, G-13, S-03, S-04, S-05, S-08,
bonate is unsuitable for irrigation. Water with RSC and S-10), “good” (G-02, G-03, G-06, and S-07), and
values ranging from 1.25 to 2.5 meq/L or less than “well” (G-01, S-01, S-02, S-09, and S-11).
1.25 meq/L can be used safely using a proper grading
system. Water with values higher than 2.5 meq/L is
Conclusions
not suitable for irrigation. Because wells with a nega-
tive RSC value do not cause any problems, the RSC This study has demonstrated that multivariate statistical
value of these wells is assumed to be zero (Lloyd and techniques are useful for water quality assessment and
Heathcote, 1985). According to the values presented investigation of salinity in coastal aquifers near wetland
14 E. D. GÜNER ET AL.

areas. A total of 24 artesian well water samples were col- ORCID

lected between May 2010 and April 2011 from the G€ oksu Esra Deniz G€
uner http://orcid.org/0000-0002-0492-2999
Plain, which is one of the coastal aquifers located on the
Turkish Mediterranean coast. In the assessment of
groundwater for salinity, TDS can be used, which is calcu- References
lated as the sum of the major ions. None of the water sam- Aiuppa, A., Bellomo, S., Brusca, L., D’Alessandro, W., and
ples collected from the sampling points in the G€ oksu Federico, C. 2003. Natural and anthropogenic factors
Plain were classified as freshwater (TDS <500 mg/L), and affecting groundwater quality of an active volcano (Mt.
Etna, Italy). Applied Geochemistry 18(6):863–882.
all of the points were considered as brackish water
Al-Khatib, M., and Al-Najar, H. 2011. Hydro-geochemical
(Custodio and Llamas, 1976). As an increase in the salinity characteristics of groundwater beneath the Gaza Strip.
of groundwater in coastal areas reveals the presence of Journal of Water Resource and Protection 335043(5):
saline water intrusion, the quality of freshwater aquifers is 341–348.
reduced for their usage for irrigation. Altın, A., Filiz, Z., and Filik Iscen, C. 2009. Assessment of
seasonal variations of surface water quality characteristics
It has been determined that the potential cause of salin-
for Porsuk Stream. Environmental Monitoring and
ity in coastal aquifer systems and seawater intrusion is due Assessment 158(1–4):51–65.
to an increase in the concentrations of Naþ, Mg2þ, Ca2þ, Anonymous. 1988. Turkish Water Pollution Control
and HCO3
. The presence of NO3
, which, together with Regulation. Official Gazette 19919.
seawater intrusion, represents an issue for groundwater Appelo, C. A. J., and Postma, D. 2005. Geochemistry,
Groundwater and Pollution. Boca Raton, FL: CRC Press.
management in the region, may depend on the use of fer-
Arslan, H. 2013. Application of multivariate statistical tech-
tilizers. FA and correlation analysis grouped the 24 sta- niques in the assessment of groundwater quality in sea-
tions into four clusters with similar characteristics based water intrusion area in Bafra Plain, Turkey.
on each factor obtained by FA. The statistical results of FA Environmental Monitoring and Assessment 185(3):
indicate that the high factor loadings of EC, Naþ, Kþ, Cl
, 2439–2452.
Baba, A., and Ayyildiz, O. 2006. Urban groundwater pollu-
and TDS typically originate from marine sources, seawater tion in turkey. In Urban Groundwater Management and
intrusion into aquifers, or distribution from marine aero- Sustainability, eds. Tellam, J. H., Rivett, M. O., Israfilov,
sols in rainfall. The results of the cluster analysis demon- R. G., and Herringshaw, L. G. Basel, Switzerland:
strated that the groundwater quality was affected by Springer, 93–110.
seawater intrusion (Naþ, Cl
). Belkhiri, L., and Narany, T. S. 2015. Using multivariate stat-
istical analysis, geostatistical techniques and structural
For the assessment of the origin of groundwater, the equation modeling to identify spatial variability of
Na/Cl, Ca/Mg, and Ca/(HCO3þSO4) values were meas- groundwater quality. Water Resources Management 29(6):
ured. These results may indicate that seawater intrusion 2073–2089.
causes a change in the water quality of the area. In add- Bhuiyan, M. A., Rakib, M. A., Dampare, S. B., Ganyaglo, S.,
ition, according to the SAR, %Na, and RSC analyses and Suzuki, S. 2011. Surface water quality assessment in
the central part of Bangladesh using multivariate analysis.
results, the use of underground water for irrigation is an KSCE Journal of Civil Engineering 15(6):995–1003.
important factor affecting salinization. Busico, G., Kazakis, N., Cuoco, E., Colombani, N., Tedesco,
A comprehensive groundwater monitoring system D., Voudouris, K., et al. 2020. A novel hybrid method of
and protection and control systems should be estab- specific vulnerability to anthropogenic pollution using
lished in such a valuable wetland area. Thus, actual data multivariate statistical and regression analyses. Water
Research 171:115386.
can be obtained from this system. In this way, more Cai, H., Savenije, H. H., Zuo, S., Jiang, C., and Chua, V. P.
comprehensive suggestions can be made for further 2015. A predictive model for salt intrusion in estuaries
research. In future studies, methods that include deep applied to the Yangtze estuary. Journal of Hydrology 529:
learning and algorithms using large data sets may be 1336–1349.
preferred for classification, feature selection and extrac- Cheng, X.-J., Zhan, W., Guo, Z.-R., and Yuan, L.-R. 2012. A
modeling study on saltwater intrusion to western four
tion. Among these methods that have been used fre- watercourses in the Pearl River estuary. China Ocean
quently in recent years, one of the methods such as Engineering 26(4):575–590.
decision tree (DT), k-nearest neighbors (KNN), support Custodio, E., and Llamas, M. R. 1976. Hidrologıa sub-
vector machines (SVM) and artificial neural networks terranea [Groundwater hydrology]. Vol. 2. Barcelona,
(ANN) can be chosen for future analysis. Spain: Editorial Omega.
De Stefano, L., Fornes, J. M., L opez-Geta, J. A., and
Villarroya, F. 2015. Groundwater use in Spain: An over-
Acknowledgment view in light of the EU Water Framework Directive.
International Journal of Water Resources Development
The authors thank Cukurova University for their support. 31(4):640–656.
 ENVIRONMENTAL FORENSICS 15

Demirel, Z. 2007. Monitoring of heavy metal pollution of Hounsinou, S. P. 2020. Assessment of potential seawater
groundwater in a phreatic aquifer in Mersin-Turkey. intrusion in a coastal aquifer system at Abomey - Calavi,
Environmental Monitoring and Assessment 132(1–3): Benin. Heliyon 6(2):e03173.
15–23. Huang, W., Hagen, S., Bacopoulos, P., and Wang, D. 2015.
Dragon, K., and Gorski, J. 2015. Identification of ground- Hydrodynamic modeling and analysis of sea-level rise
water chemistry origins in a regional aquifer system impacts on salinity for oyster growth in Apalachicola
(Wielkopolska region, Poland). Environmental Earth Bay, Florida. Estuarine, Coastal and Shelf Science 156:
Sciences 73(5):2153–2167. 7–18.
Eaton, F. M. 1950. Significance of carbonates in irrigation Jalali, M. 2011. Nitrate pollution of groundwater in
waters. Soil Science 69(2):123–134. Toyserkan, western Iran. Environmental Earth Sciences
Edmunds, W. M., Ma, J., Aeschbach-Hertig, W., Kipfer, R., 62(5):907–913.
and Darbyshire, D. P. F. 2006. Groundwater recharge his- Jeong, S., Yeon, K., Hur, Y., and Oh, K. 2010. Salinity intru-
tory and hydrogeochemical evolution in the Minqin sion characteristics analysis using EFDC model in the
Basin, North West China. Applied Geochemistry 21(12): downstream of Geum River. Journal of Environmental
2148–2170. Sciences 22(6):934–939.
Eissa, M. A., Mahmoud, H. H., Shouakar-Stash, O., El- Jiang, Y., Wu, Y., Groves, C., Yuan, D., and Kambesis, P.
Shiekh, A., and Parker, B. 2016. Geophysical and geo- 2009. Natural and anthropogenic factors affecting the
chemical studies to delineate seawater intrusion in groundwater quality in the Nandong karst underground
Bagoush area, Northwestern coast, Egypt. Journal of river system in Yunan, China. Journal of Contaminant
African Earth Sciences 121:365–381. Hydrology 109(1–4):49–61.
El Yaouti, F., El Mandour, A., Khattach, D., Benavente, J., Kang, M., and Jackson, R. B. 2016. Salinity of deep ground-
and Kaufmann, O. 2009. Salinization processes in the water in California: Water quantity, quality, and protec-
unconfined aquifer of Bou-Areg (NE Morocco): A geo- tion. Proceedings of the National Academy of Sciences
statistical, geochemical, and tomographic study. Applied 113(28):7768–7773.
Kazakis, N., Pavlou, A., Vargemezis, G., Voudouris, K. S.,
Geochemistry 24(1):16–31.
Elhatip, H., Afşin, M., Dirik, K., Kurmaç, Y., and Soulios, G., Pliakas, F., et al. 2016. Seawater intrusion
mapping using electrical resistivity tomography and
Kavurmacı, M. 2003. Influences of human activities and
hydrochemical data. An application in the coastal area of
agriculture on groundwater quality of Kayseri-Incesu-
eastern Thermaikos Gulf, Greece. The Science of the Total
Dokuzpınar springs, central Anatolian part of Turkey.
Environment 543:373–387.
Environmental Geology 44(4):490–494.
Kim, J.-H., Kim, R.-H., Lee, J., Cheong, T.-J., Yum, B.-W.,
Fadili, A., Najib, S., Mehdi, K., Riss, J., Malaurent, P., and
and Chang, H.-W. 2005. Multivariate statistical analysis
Makan, A. 2017. Geoelectrical and hydrochemical study
to identify the major factors governing groundwater qual-
for the assessment of seawater intrusion evolution in
ity in the coastal area of Kimje, South Korea.
coastal aquifers of Oualidia, Morocco. Journal of Applied
Hydrological Processes 19(6):1261–1276.
Geophysics 146:178–187. Kim, K.-H., Yun, S.-T., Park, S.-S., Joo, Y., and Kim, T.-S.
Giardino, C., Brando, V. E., Dekker, A. G., Str€ ombeck, N., 2014. Model-based clustering of hydrochemical data to
and Candiani, G. 2007. Assessment of water quality in demarcate natural versus human impacts on bedrock
Lake Garda (Italy) using Hyperion. Remote Sensing of groundwater quality in rural areas, South Korea. Journal
Environment 109(2):183–195. of Hydrology 519:626–636.
Gong, W., and Shen, J. 2011. The response of salt intrusion Krapac, I. G., Dey, W. S., Roy, W. R., Smyth, C. A.,
to changes in river discharge and tidal mixing during the Storment, E., Sargent, S. L., et al. 2002. Impacts of swine
dry season in the Modaomen Estuary, China. Continental manure pits on groundwater quality. Environmental
Shelf Research 31(7–8):769–788. Pollution 120(2):475–492.
Grassi, S., Cortecci, G., and Squarci, P. 2007. Groundwater Landon, M. K., Green, C. T., Belitz, K., Singleton, M. J.,
resource degradation in coastal plains: The example of and Esser, B. K. 2011. Relations of hydrogeologic factors,
the Cecina area (Tuscany–Central Italy). Applied groundwater reduction-oxidation conditions, and tem-
Geochemistry 22(11):2273–2289. poral and spatial distributions of nitrate, Central-Eastside
G€uler, C., Kurt, M. A., Alpaslan, M., and Akbulut, C. 2012. San Joaquin Valley, California, USA. Hydrogeology
Assessment of the impact of anthropogenic activities on Journal 19(6):1203–1224.
the groundwater hydrology and chemistry in Tarsus Liu, B., Peng, S., Liao, Y., and Long, W. 2018. The causes
coastal plain (Mersin, SE Turkey) using fuzzy clustering, and impacts of water resources crises in the Pearl River
multivariate statistics and GIS techniques. Journal of Delta. Journal of Cleaner Production 177:413–425.
Hydrology 414:435–451. Lloyd, J. W., and Heathcote, J. A. A. 1985. Natural
G€uner, E. D., 2016. Determination of groundwater quality Inorganic Hydrochemistry in Relation to Ground Water.
and distribution levels in G€oksu Delta, PhD diss., Oak Ridge, TN: OSTI.
University of Çukurova. Machiwal, D., and Jha, M. K. 2015. Identifying sources of
G€uner, E. D., Tekin, S., and Seçkin, G. 2018. An assessment groundwater contamination in a hard-rock aquifer system
of shallow groundwater wells in an agricultural and using multivariate statistical analyses and GIS-based geo-
coastal area in G€ oksu Delta, Turkey. Bilge International statistical modeling techniques. Journal of Hydrology:
Journal of Science and Technology Research 2(1):9–16. Regional Studies 4:80–110.
16 E. D. GÜNER ET AL.

Magyar, N., Gabor Hatvani, I., Szekely, I. K., Herzig, A., hydrogeochemical studies: An example from Karnataka,
Dinka, M., and Kovacs, J. 2013. Application of multivari- India. Water Research 36(10):2437–2442.
ate statistical methods in determining spatial changes in Rekha, P. N., Kanwar, R. S., Nayak, A. K., Hoang, C. K.,
water quality in the Austrian part of Neusiedler See. and Pederson, C. H. 2011. Nitrate leaching to shallow
Ecological Engineering 55:82–92. groundwater systems from agricultural fields with differ-
Morgan, L. K., and Werner, A. D. 2015. A national inven- ent management practices. Journal of Environmental
tory of seawater intrusion vulnerability for Australia. Monitoring 13(9):2550–2558.
Journal of Hydrology: Regional Studies 4:686–698. Ros, S. E. M., and Zuurbier, K. G. 2017. The impact of inte-
Moyo, N. A. G. 2013. An analysis of the chemical and grated aquifer storage and recovery and brackish water
microbiological quality of ground water from boreholes reverse osmosis (ASRRO) on a coastal groundwater sys-
and shallow wells in Zimbabwe. Physics and Chemistry of tem. Water 9(4):273.
the Earth, Parts A/B/C 66:27–32. Sae-Ju, J., Chotpantarat, S., and Thitimakorn, T. 2020.
Najib, S., Fadili, A., Mehdi, K., Riss, J., and Makan, A. 2017. Hydrochemical, geophysical and multivariate statistical
Contribution of hydrochemical and geoelectrical investigation of the seawater intrusion in the coastal
approaches to investigate salinization process and sea- aquifer at Phetchaburi Province, Thailand. Journal of
water intrusion in the coastal aquifers of Chaouia, Asian Earth Sciences 191:104165.
Morocco. Journal of Contaminant Hydrology 198:24–36. Sahinci, A. 1991. Geochemistry of Natural Waters. Izmir,
Owoseni, J. O., Tamarautobou, E. U., and Asiwaju-Bello, Turkey: Reform Printing Office, 548.
Y. A. 2013. Application of sequential analysis and geo- Sargaonkar, A., and Deshpande, V. 2003. Development of
graphic information systems for hydrochemical evolution an overall index of pollution for surface water based on a
survey, Shagari Environ, Southwestern Nigeria. Amerikan general classification scheme in Indian context.
International Journal of Contemporary Reserch 3(3): Environmental Monitoring and Assessment 89(1):43–67.
38–48. Seanego, K. G., and Moyo, N. A. G. 2013. The effect of sew-
Palma, P., Ledo, L., Soares, S., Barbosa, I. R., and age effluent on the physico-chemical and biological char-
Alvarenga, P. 2014. Spatial and temporal variability of the acteristics of the Sand River, Limpopo, South Africa.
Physics and Chemistry of the Earth, Parts A/B/C 66:
water and sediments quality in the Alqueva reservoir
75–82.
(Guadiana Basin; southern Portugal). Science of the Total
Seckin, G., Yilmaz, T., Sari, B., and Ersu, C. B. 2010.
Environment 470:780–790.
Groundwater hydrochemistry at the Mediterranean
Papatheodorou, G., Demopoulou, G., and Lambrakis, N.
coastal plains—The case of Silifke, Turkey. Desalination
2006. A long-term study of temporal hydrochemical data
253(1):164–169.
in a shallow lake using multivariate statistical techniques.
Selle, B., Schwientek, M., and Lischeid, G. 2013.
Ecological Modelling 193(3):759–776.
Understanding processes governing water quality in
Psychoyou, M., Mimides, T., Rizos, S., and Sgoubopoulou,
catchments using principal component scores. Journal of
A. 2007. Groundwater hydrochemistry at Balkan coastal
Hydrology 486:31–38.
plains—The case of Marathon of Attica, Greece. Sherif, M., Mohamed, M., Kacimov, A., and Shetty, A. 2011.
Desalination 213(1):230–237. Assessment of groundwater quality in the northeastern
Qi, H., Ma, C., He, Z., Hu, X., and Gao, L. 2019. Lithium coastal area of UAE as precursor for desalination.
and its isotopes as tracers of groundwater salinization: A Desalination 273(2):436–446.
study in the southern coastal plain of Laizhou Bay, Shi, L., and Jiao, J. J. 2014. Seawater intrusion and coastal
China. The Science of the Total Environment 650: aquifer management in China: A review. Environmental
878–890. Earth Sciences 72(8):2811–2819.
Qin, R., Wu, Y., Xu, Z., Xie, D., and Zhang, C. 2013. Siebert, S., Burke, J., Faures, J.-M., Frenken, K., Hoogeveen,
Assessing the impact of natural and anthropogenic activ- J., D€
oll, P., et al. 2010. Groundwater use for irrigation—
ities on groundwater quality in coastal alluvial aquifers of A global inventory. Hydrology and Earth System Sciences
the lower Liaohe River Plain, NE China. Applied 14(10):1863–1880.
Geochemistry 31:142–158. Simeonov, V., Stratis, J. A., Samara, C., Zachariadis, G.,
Raghunath, H. M. 1987. Ground Water. New Delhi, India: Voutsa, D., Anthemidis, A., et al. 2003. Assessment of
New Age International. the surface water quality in Northern Greece. Water
Rakib, M. A., Sasaki, J., Matsuda, H., Quraishi, S. B., Research 37(17):4119–4124.
Mahmud, M. J., Bodrud-Doza, M., et al. 2020. Souid, F., Agoubi, B., Telahigue, F., Chahlaoui, A., and
Groundwater salinization and associated co-contamin- Kharroubi, A. 2018. Groundwater salinization and sea-
ation risk increase severe drinking water vulnerabilities in water intrusion tracing based on Lithium concentration
the southwestern coast of Bangladesh. Chemosphere 246: in the shallow aquifer of Jerba Island, southeastern
125646. Tunisia. Journal of African Earth Sciences 138:233–246.
Re, V., Sacchi, E., Martin-Bordes, J. L., Aureli, A., El Stigter, T. Y., Ribeiro, L., and Dill, A. C. 2008. Building fac-
Hamouti, N., Bouchnan, R., et al. 2013. Processes affect- torial regression models to explain and predict nitrate
ing groundwater quality in arid zones: The case of the concentrations in groundwater under agricultural land.
Bou-Areg coastal aquifer (North Morocco). Applied Journal of Hydrology 357(1):42–56.
Geochemistry 34:181–198. Sung, K.-Y., Yun, S.-T., Park, M.-E., Koh, Y.-K., Choi, B.-
Reghunath, R., Murthy, T. S., and Raghavan, B. R. 2002. Y., Hutcheon, I., et al. 2012. Reaction path modeling of
The utility of multivariate statistical techniques in hydrogeochemical evolution of groundwater in granitic
 ENVIRONMENTAL FORENSICS 17

bedrocks, South Korea. Journal of Geochemical to an Australian case study. Water Resources
Exploration 118:90–97. Management 25(7):1831–1853.
Talukder, M. R. R., Rutherford, S., Phung, D., Islam, M. Z., Wilcox, LVj. 1955. Classification and Use of Irrigation
and Chu, C. 2016. The effect of drinking water salinity Waters. Washington, DC: USDA.
on blood pressure in young adults of coastal Bangladesh. Yilmaz, T., Seckin, G., and Sari, B. 2010. Trace element lev-
Environmental Pollution 214:248–254. els in the groundwater of Mediterranean Coastal Plains—
Telahigue, F., Agoubi, B., Souid, F., and Kharroubi, A. 2018. The case of Silifke, Turkey. CLEAN - Soil, Air, Water
Assessment of seawater intrusion in an arid coastal aqui- 38(3):221–224.
fer, south-eastern Tunisia, using multivariate statistical Yusuf, K. A., Surukite, O., Abdusalam, I. O., and Majolagbe,
analysis and chloride mass balance. Physics and A. O. 2013. Assessment of spatial variation of surface
water quality in Lagos, using multivariate statistical
Chemistry of the Earth, Parts A/B/C 106:37–46.
techniques. Journal of Environment 2(4), 94-102.
Tully, K., Gedan, K., Epanchin-Niell, R., Strong, A.,
Zhang, E., Savenije, H. H., Wu, H., Kong, Y., and Zhu, J. 2011.
Bernhardt, E. S., BenDor, T., et al. 2019. The invisible
Analytical solution for salt intrusion in the Yangtze Estuary,
flood: The chemistry, ecology, and social implications of China. Estuarine, Coastal and Shelf Science 91(4):492–501.
coastal saltwater intrusion. BioScience 69(5):368–378. Zhao, M., Zeng, C., Liu, Z., and Wang, S. 2010. Effect of
Water Environment Federation and American Public different land use/land cover on karst hydrogeochemistry:
Health Association. 2005. Standard methods for the exam- A paired catchment study of Chenqi and Dengzhanhe,
ination of water and wastewater. Washington, DC: Puding, Guizhou, SW China. Journal of Hydrology
American Public Health Association (APHA). 388(1):121–130.
Werner, A. D., Alcoe, D. W., Ordens, C. M., Hutson, J. L., Zhu, B., and Yu, H. 1995. The use of geochemical indicator
Ward, J. D., and Simmons, C. T. 2011. Current practice elements in the exploration for hot water sources within
and future challenges in coastal aquifer management: geothermal fields. Journal of Geochemical Exploration
Flux-based and trigger-level approaches with application 55(1):125–136.