Environ Earth Sci (2016) 75:1020

DOI 10.1007/s12665-016-5823-y

ORIGINAL ARTICLE

Assessment of groundwater quality of Lakshimpur district

of Bangladesh using water quality indices, geostatistical methods,
and multivariate analysis
Mohammad Amir Hossain Bhuiyan1,6 • Md. Bodrud-Doza1 • A. R. M. Towfiqul Islam2,3 •
M. A. Rakib3 • M. Safiur Rahman4,5 • A. L. Ramanathan6

Received: 10 April 2015 / Accepted: 6 June 2016

Ó Springer-Verlag Berlin Heidelberg 2016

Abstract Groundwater evaluation indices, multivariate quality indices and PCA components. The results of
statistical techniques, and geostatistical models are applied semivariogram models have shown that most of the vari-
to assess the source apportionment and spatial variability of ables have weak spatial dependence, indicating agricultural
groundwater pollutants at the Lakshimpur district of Ban- and residential/domestic influences. The spatial distribution
gladesh. A total of 70 groundwater samples have been maps of water quality parameters have provided a useful
collected from wells (shallow to deep wells, i.e., and robust visual tool for decision makers toward defining
10–375 m) from the study area. Groundwater quality index adaptive measures. This study is an implication to show the
reveals that 50 % of the water samples belong to good- multiple approaches for quality assessment and spatial
quality water. The degrees of contamination, heavy metal variability of groundwater as an effort toward a more
pollution index, and heavy metal evaluation index present effective groundwater quality management.
diversified results in samples even though they show sig-
nificant correlations among them. The results of principal Keywords Geostatistical methods  Groundwater quality
component analysis (PCA) show that groundwater quality index  Multivariate analysis  Bangladesh
in the study area mainly has geogenic (weathering and
geochemical alteration of source rock) sources followed by
anthropogenic source (agrogenic, domestic sewage, etc.). Introduction
Cluster analysis and correlation matrix also supported the
results of PCA. The Gaussian semivariogram models have Metals’ contamination of groundwater is of great concern
been tested as the best fit models for most of the water on lives owing to their toxicity, persistence, and extensive
bioaccumulation. Groundwater is an important resource for
& Mohammad Amir Hossain Bhuiyan agriculture, industrial, and other economic sectors in
[email protected]; [email protected]; Bangladesh. Rapid urbanization, agricultural, and indus-
[email protected]
trial activities are affecting groundwater day by day. A
1
Department of Environmental Sciences, Jahangirnagar wide range of public health issues such as cancer, hyper-
University, Dhaka 1342, Bangladesh tension, hyperkeratosis, peripheral vascular disease,
2
School of Atmospheric Sciences, Nanjing University of restrictive lung disease, and gangrene occurs due to the
Information Science and Technology, Nanjing 210044, China consumption of contaminated water (Smith et al. 2000).
3
Department of Disaster Management, Begum Rokeya Approximately 17 % of groundwater in Bangladesh exhibit
University, Rangpur 5400, Bangladesh arsenic (As) concentrations beyond the acceptable limit
4
Department of Environmental Engineering, Dalhouise (10 lg/L) of DoE (1997) for drinking water.
University, Halifax, NS B3J 1Z1, Canada The interpretation of water quality data sets for pollution
5
Chemistry Division, Atomic Energy Center, Dhaka, evaluation is quite difficult by only elemental concentrations
Bangladesh (Nimic and Moore 1991). However, WQIs have huge scope
6
School of Environmental Sciences, Jawaharlal Nehru to analyze the data sets for better interpretation of pollution.
University, New Delhi 110067, India There exists a wide range of WQIs; however, the choice

123
1020 Page 2 of 23 Environ Earth Sci (2016) 75:1020

depends upon the input variables and the desired results Bangladesh, has been selected for this study. Geographi-
(Handa 1981; Zou et al. 1988; Sahu et al. 1991; Li et al. cally, the study area is positioned between 22°490 –22°030 N
2009). Due to some limitations, WQI values provide better and 90°430 –92°000 E (Fig. 1). It is bounded by Raipur,
results together with the chemometric techniques. It has Ramganj, and Chatkhil upazilas on the north; Daulatkhan,
been found that chemometric methods are the most reliable Kamalnagar, and Noakhali Sadar upazilas on the south;
approaches for data mining of matrices from environmental Begumganj and Sonaimuri upazilas on the east; and Raipur
quality assessment (Astel et al. 2007, 2008). Among the upazila and Meghna River on the west. Lakshimpur upazila
available chemometric methods, multivariate statistical (subdistrict) has an area of 514.78 sq km with a total
analysis has been widely used for source apportionment of population of 575,278 (Banglapedia 2006). The sites are
metals in soil and water in different parts of the world (e.g., chosen mainly based on their proximity of suspected pol-
Singh et al. 2005; Halim et al. 2010; Bhuiyan et al. 2010; Li lution sources and ecological and environmental impor-
et al. 2013; Machiwal and Jha 2015). tance. Dalal Bazar, Parbatinagar, Dattapara, Hajipara,
On the other hand, geostatistical method has gained Jacksinhat sites are densely populated areas. The ground-
importance to evaluate the spatial distribution of pollutant in water quality at Mazuchowdhurihat and Shackchar (west-
soils and water. It is also an important tool for spatial ern part of the area) is highly dominated by Meghna River.
dependence/autocorrelation among the sampling points. Physiographically, it is a coastal floodplain that experi-
Subsequently, this type of information is important in esti- ences tide actions regularly. About 74 % of this area is
mating the pollutant migration history and spatial distribution under water supply coverage provided by several local and
of the pollutant at different sites. However, the integrated national NGOs.
approaches of multivariate analysis and geostatistics may
provide a holistic approach on the complex pollution system. Sample collection and preparation
As the spatial distribution of heavy metal contamination
in groundwater is controlled by the geological/geochemical Groundwater samples are collected from 70 preselected
heterogeneity, the spatial interpolation technique has been sampling points at the Sadar upazila of Lakshimpur dis-
used to estimate the concentration at unmeasured locations trict of Bangladesh (Fig. 1). The sampling locations are
and devise points to show groundwater contamination recorded by a GPS device (Explorist model: 200). The
(Webster and Oliver 2001). Detailed and extensive expla- information regarding well depths is collected from the
nations of geostatistical method have been reported in record preserved by the well owners and local government
different literatures (Isaaks and Srivastava 1989; Goovaerts offices. Three types of tubewells such as (1) shallow wells
1997; Webster and Oliver 2001). The cross-validation (10–60 m depth), (2) deep wells (80–375 m depth), and
results from geostatistics represent that the ordinary kriging (3) dug wells (7–318 m) have been selected on the basis
technique can predict spatial variability more accurately. of the availability at the study area. Samples are collected
The ordinary kriging method deals with spatial correlations in pre-washed high-density polypropylene (HDPP) bottles
between the sample points and has been widely used for following the standard method of APHA-AWWA-WEF
mapping spatial variability of elements. An assessment of (2005). For metal ions and dissolved organic carbon
drinking and irrigation water quality is very essential for (DOC) analysis, water samples are preserved following
understanding the suitability of groundwater for different the standard procedures of Rahman and Gagnon (2014).
purposes. In the study area, a limited work has been con- All analytical procedures of groundwater samples are
ducted on groundwater quality. Hence, the integrated conducted following the standard methods (Table 1). The
approaches of different chemometric methods are consid- accuracy and precision of analysis are tested through
ered as important tool for pollution evaluation in this study running duplicate analysis on selected samples, and the
area. Considering all these aspects, Lakshimpur district of average results for all analyses are used to represent the
Bangladesh has been selected as the study area for a data.
comprehensive study using the integrated approaches of
multivariate analysis and geostatistical methods. Groundwater pollution evaluation indices

Groundwater quality index (GWQI)

Materials and methods
Groundwater quality index (GWQI) method reflects the
Study area composite influence of the different water quality param-
eters on the suitability for drinking purposes (Sahu and
Lakshimpur Sadar upazila (a subdistrict, a small adminis- Sikdar 2008). The groundwater quality has been measured
trative unit), located in southeastern Lakshimpur district of by using the following equation for GWQIs Vasanthavigar

123
Environ Earth Sci (2016) 75:1020 Page 3 of 23 1020

Fig. 1 Location map showing the sample sites in the study area

Table 1 List of chemical elemental analysis, methods, and equipments

Variables Method Units Equipment

pH pH meter Accumet electrode and Accumet Excel, XL50

(Dual channel pH/ion/conductivity) meter
EC Conductivity meter Accumet electrode and Accumet Excel, XL50
(Dual channel pH/ion/conductivity) meter
DOC Total organic carbon analyzer TOC-VCHP Total Organic Carbon Analyzer (Shimadzu, Japan)
HCO3 Field titration HACH digital titrator (HACH 1690, USA)
Cl, F, SO4, NH4–N, Na, K, Ca, Mg Chromatography Ion chromatograph (761 Compact IC, Metrohm)
As, Pb, Fe, Mn, Ni, Sb, Ba, Mo, Inductively coupled plasma Thermo Scientific X-Series2 ICP-MS
Al, Zn, B, P, Si mass spectrometry

et al. (2010) with respect to WHO (2011) and Bangladesh where Ci is the concentration of each parameters, Si is the
standards (1997). limit values, wi is the assigned weight according to its
X relative importance in the overall quality of water for
GWQI ¼ SIi
drinking purposes (Table 2), qi is the water quality rating,
X X  wi  Ci 
¼ ðWi  qi Þ ¼ Pn   100 Wi is the relative weight, and SIi is the subindex of ith
i¼1 wi Si parameter.The heavy metal pollution index (HPI) method

123
1020 Page 4 of 23 Environ Earth Sci (2016) 75:1020

has been developed by assigning rating or weightage (Wi) Heavy metal evaluation index (HEI) method provides an
for each chosen parameter (As, Pb, Fe, Mn, Ni, Sb, Ba, Mo, insight into the overall quality of the groundwater with
Al, Zn) and selecting the groundwater parameter on which respect to heavy metals and metaloids (As, Pb, Fe, Mn, Ni,
the index has to be based on (Bhuiyan et al. 2010). The Sb, Ba, Mo, Al, Zn (Edet and Offiong 2002). It has been
rating is nearly zero to one, and its selection reveals the calculated by Prasad and Jaiprakas (1999) as follows:
significance of each water quality parameter. It has been Xn
Hc
developed based on the monitored values, ideal values, and HEI ¼
Hmac
recommended standard values of the studied parameters. It i¼1

can be defined as inversely proportional to the recom- where Hc is the monitored value and Hmac is the maximum
mended standard (Si) for each parameter (Horton 1965; admissible concentration (MAC) of ith parameter.
Reddy 1995; Mohan et al. 1996). The concentration limits The degree of contamination (Cd/CD) has been adopted
(i.e., the highest permissible value for drinking water (Si) from Backman et al. (1997). Prasad and Bose (2001)
and maximum desirable value (Ii) for each parameter) are evaluate the combined effects of several quality parameters
taken from the Indian drinking water specification stan- which are considered detrimental to household water. The
dards of 2012 (BIS 2012) for this study. Heavy metal CD/Cd is determined by:
pollution index (HPI) has been used for assigning rating or
Xn
weightage (Wi) for each selected parameter and can be Cd ¼ Cfi
computed using the following expression (Mohan et al. i¼1
1996; Bhuiyan et al. 2015):
Pn where
i¼1 Wi Qi Cai
HPI ¼ P n
i¼1 wi
Cfi ¼ 1
Cni
where Qi is the subindex of the ith parameter, Wi is the unit Cfi is the contamination factor, Cai is the analytical value,
weight of the ith parameter, and n is the number of and Cni is the upper permissible concentration for the ith
parameters. The subindex Qi is computed by component and n indicates the normative value. Here, Cni
X
n
fMi ðÞIi g is taken as maximum admissible concentration (MAC).
Qi ¼  100
i¼1
ðSi  Ii Þ
Multivariate statistical analysis
where Mi, li, and Si stand for the monitored values, ideal
values, and standard values of the ith parameter, respec- Principal component analysis (PCA) reduces the dimen-
tively. The negative sign (-) denotes numerical difference sionality of data by a linear combination of original data
in the two values ignoring algebraic sign. to generate new latent variables which are orthogonal and

Table 2 List of parameters,

Parameters Units Weight (wi) Relative weight (Wi) Limit values
weight factors, and limit values
for the water quality index after pH 4 0.078 6.5–8.5
Vasanthavigar et al. (2010)
HCO3 mg/L 1 0.020 600
Cl mg/L 3 0.059 250
F mg/L 4 0.078 1.5
SO4 mg/L 4 0.078 400
NH4–N mg/L 3 0.059 0.5
Na mg/L 4 0.078 200
K mg/L 2 0.039 12
Ca mg/L 2 0.039 75
Mg mg/L 2 0.039 30
As lg/L 4 0.078 50
Pb lg/L 4 0.078 10
Fe lg/L 4 0.078 1000
Mn lg/L 4 0.078 300
Al lg/L 3 0.059 200
Zn lg/L 3 0.059 5000
P P
wi ¼ 51 Wi ¼ 1

123
Environ Earth Sci (2016) 75:1020 Page 5 of 23 1020

uncorrelated to each other (Nkansah et al. 2010). It interpolation techniques among other techniques. It inte-
extracts the eigenvalues and eigenvectors from the grates both spatial correlation and the dependence in the
covariance matrix of original variables (Chabukdhara and prediction of a known variable. Estimations of nearly all
Nema 2012). The eigenvalues of the PCs are the measure spatial interpolation methods can be represented as
of their associated variance, the participation of original weighted averages of sampled data. The equation can be
variables in the PCs is given by the loadings, and the written as follows (Delhomme 1978):
coordinates of the objects are called scores (Helena et al.
X
n
2000; Wunderlin et al. 2001; Heberger et al. 2005). PCA z^ðxo Þ ¼ ki zðxi Þ
provides an objective way of finding indices of this type i¼1
so that the variation in the data can be accounted for as
where z^ is the estimated value of an attribute at the point of
concisely as possible (Sarbu and Pop 2005). PCA has
interest xo , z is the observed value at the sampled point xi,
been performed to extract principal components (PC)
ki is the weight assigned to the sampled point, and n rep-
from the sampling points and to evaluate spatial varia-
resents the number of sampling points used for the esti-
tions and possible sources of heavy metals in
mation (Webster and Oliver 2001). The attribute is usually
groundwater.
called the primary variable, especially in geostatistics. The
Factor analysis is similar to PCA except for the prepa-
semivariance can be estimated from groundwater data by
ration of observed correlation matrix for extraction and the
the following equation:
underlying theory (Tabachnick and Fidell 2007; Bhuiyan
n h i
et al. 2011a). The goal of FA can be achieved by rotating 1 X
the axis defined by PCA, according to the well-established cðhÞ ¼ zðxi Þ  zðxi þ hÞ2
2n i¼1
rules, and constructing new variables, also called varifac-
tors (VF) (Shrestha and Kazama 2007). where n is the number of pairs of sample points separated
The correlation coefficient matrix measures how well by the standard distance calls lag h (Burrough and
the variance of each constituent can be explained by rela- McDonnell 1998), and zðxi Þ is the value of variable z at
tionship with each other (Liu et al. 2003). According to the location xi. Variogram modeling and estimation are
approach of Liu et al. (2003), the terms ‘‘strong,’’ ‘‘mod- important for structural analysis and spatial interpolation.
erate,’’ and ‘‘weak’’ are applied to factor loadings and refer Among the different kriging techniques, OK has been used
to absolute loading values as [0.75, 0.75–0.50, and in this study because of its easy calculation and prediction
0.50–0.30, respectively. accuracy compared to the other kriging methods (Gorai and
The cluster analysis (CA) is applied to identify groups or Kumar 2013). Recently, different variograms or semivari-
clusters of similar sites on the basis of similarities within a ogram models such as linear, exponential, and spherical
class and dissimilarities between different classes (Lattin models are very popular worldwide for spatial analysis of
et al. 2003). Hierarchical cluster analysis (HCA) studies geochemical data sets (Kitanidis 1997; Elogne et al. 2008;
distance between parameters of samples. The most similar Varouchakis and Hristopulos 2013). In this study, circular,
points are grouped forming one cluster, and the process is spherical, exponential, and Gaussian models have been
repeated until all points belong to one cluster (Danielsson used to measure spatial autocorrelation or dependence of
et al. 1999; Birth 2003). The result obtained is shown in a the groundwater data. The best fit theoretical semivari-
2D plot called dendrogram. In the study, Ward’s method ogram models are prepared based on selecting the trial-
with squared Euclidean distances is used. The experimental and-error basis. Predictive performances of fit models are
groundwater data were subjected to statistical analysis checked on the basis of cross-validation tests (Gorai and
using SPSS software (version 22.0). Pearson’s correlation Kumar 2013). The mean error (ME), mean square error
matrix is used to identify the relationship among the pairs (MSE), root mean error (RMSE), average standard error
of parameters. (ASR), and root mean square standardized error (RMSSE)
values are assessed to establish the fit models performance.
Geostatistical modeling Hu et al. (2004) have discussed several criteria for using
error measurements to judge the performance of spatial
Ordinary kriging (OK) and semivariogram models are interpolation methods. Models attain the best goodness-of-
applied for spatial distribution of groundwater parameters fit results in minimum mean error (ME), root mean error
which are related to groundwater application in hydrolog- (RME), and mean squared error (MSE), attain root mean
ical studies. These interpolation techniques are well doc- squared error (RMSE) and average squared error (ASE)
umented in the recent literature (e.g., Masoud and Atwia close to unity and are considered as the best fit models
2010; Ahmed et al. 2011; Masoud 2014; Marko et al. performance (ESRI 2009). The errors are estimated by the
2014). Kriging is one of the most popular and robust following equations:

123
1020 Page 6 of 23 Environ Earth Sci (2016) 75:1020

1X n values of metal content in groundwater samples. The range

ME ¼ ðpi  oi Þ and mean values of HPI in groundwater samples are
n i¼1
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2.19–59.01 and 10.41, respectively. Among the total sam-
u" n # ples, 77 % of the samples fall above the critical values, i.e.,
u 1X
RMSE ¼ t ðpi  oi Þ 2
CD 3. According to Edet and Offiong (2002), most of the
n i¼1
samples are considered as highly polluted water. Besides
the CD values, the water samples are further analyzed by
1X n
MSE ¼ ðpsi  osi Þ HPI and HEI methods to compare with the results of CD.
n i¼1
However, HPI and HEI values for all sample locations fall
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u" n # below the critical values prescribed for drinking purposes
u 1X
RMSSE ¼ t ðpsi  osi Þ 2 (Table 5). The heavy metal evaluation index (HEI) is used
n i¼1 to synchronize the criteria for various pollution indices.
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi The HEI criteria for groundwater samples are thus classi-
u" n #ffi
u 1X Xn . fied as low (HEI \ 40), medium (HEI = 40–80), and high
ASE ¼ t ðpi  i¼1 i
p nÞ2 (HEI [ 80). It is observed that groundwater in the study
n i¼1
area exhibits low level of pollution. The results of GWQIs,
where n is the number of observation points or samples; o CD, HPI, and HEI methods show more or less similar
and p are the observed value and predicted or estimated trends for most of the samples (Fig. 2). The GWQI values
values at location i; os is the standardized observed value; have shown higher spatial variation, whereas HEI values
ps is the standardized predicted value. After completing the have depicted lower variation. We have also assessed the
cross-validation process, kriging offers graphical repre- relationship between metal concentration with the com-
sentation of the distribution of groundwater quality. In this puted indices (GWQIs, CD, HPI, and HEI in Table 5). The
study, Arc GIS (10.2 version) has been used for this GWQIs show positive significant correlations with CD,
interpolation technique. HPI, and HEI. As, Fe, Mn, and Ni show strong positive
correlation (Table 6) with the indices values indicating the
metals are the major factors for the pollution in this region.

Results and discussion Source identification of groundwater pollutants

Groundwater quality Principal component analysis (PCA) is used for source

identification of heavy metals following the standard pro-
GWQI values are calculated using different International cedures (Dragovı́c et al. 2008; Franco-Urı́a et al. 2009).
Standards and BMAC values to determine the suitability of Varimax rotation is used to maximize the sum of variance
groundwater quality for drinking purposes (Table 3). The of the factor coefficients which better explains the possible
GWQI values range from 24.73 to 430.94 with the mean of groups/sources that influence water systems (Gotelli and
113.56. The critical limit (100) for drinking water purposes Ellison 2004). Six factors with eigenvalues [1 are
has been proposed by Vasanthavigar et al. (2010). The extracted for groundwater data sets which represent
results in Table 4 show that 50 % of samples exceed the 80.61 % of the total variance. The scree plot is used to
critical limit (100) of GWQIs. Among the total samples, identify the number of PCs to be retained to understand the
17 % of samples belong to excellent water quality and underlying parameters’ structure (Fig. 3a). The calculated
33 % represent good water quality, 40 % exhibit poor factor loadings together with cumulative percentage and
water quality, 7 % of water is of very poor quality, and the percentages of variance are explained by each factor as
rest of 3 % indicate unsuitable water for drinking purposes listed in Table 7. The positive and negative scores in PCA
(Table 4). indicate that most of the water samples are either essen-
The degree of contamination (CD) has been used for tially affected or unaffected by the presence of extracted
estimating the extent of metal pollution (Al-Ami et al. loads on a specific factor/component, respectively. About
1987; Bhuiyan et al. 2010). Table 5 shows that range and 58.38 % of the total variance is represented in the first
mean values of CD for groundwater samples are three loading factors (Fig. 3b). In this study, PC1, PC2,
0.17–66.23 and 11.17, respectively. CD may be catego- PC3, PC4, PC5, and PC6 explain more than 26, 17, 14, 8,
rized (Backman et al. 1997; Edet and Offiong 2002) as 7, and 6 % of the total variance, respectively.
follows: low (CD \ 1), medium (CD = 1–3), and high The first principal component (PC1) in the groundwater
(CD [ 3). The heavy metal pollution indices (HPI) are data sets explains more than 26.87 % of the total variance.
computed using the International Standard (BIS 2012) It is loaded with Na, EC, Cl, B, Mg, K, Ca, SO4, and

123
Environ Earth Sci (2016) 75:1020 Page 7 of 23 1020

Table 3 Descriptive statistics of physiochemical parameters and heavy metals in the study area
Parameters Minimum Maximum Mean SD WHO limits BMAC Acceptable limit Permissible limit
(2011) (1997) of Indian (2012) of Indian (2012)

pH 6.53 7.59 7.0369 0.24661 6.5–8.5 6.5–8.5

a
EC (ls/cm) 384 4640 1135.086 954.7244 700
DOC (mg/L) BDL 14 3.5386 3.06104
HCO3- (mg/L) 125 1140 430.18 217.5315 600a
-
Cl (mg/L) 2.7 1550 227.1904 364.6709 250 150–600 250 1000
F- (mg/L) 0.08 0.45 0.2149 0.07277 1.5 1 1.5
SO42-(mg/L) BDL 265 16.1429 45.05423 400 200 400
NH4–N (mg/L) BDL 17.8 1.9434 2.73216 0.5 0.5
Na? (mg/L) 7.9 1090 159.7871 225.6383 200
K? (mg/L) 3.1 41.3 10.8943 7.75802 12
Ca2? (mg/L) 16.6 186 55.7614 34.72118 75 75 200
2?
Mg (mg/L) 15.7 154 46.1486 31.55469 30–35 30 100
As (lg/L) BDL 404 85.5443 98.45996 10 50 10 50
Pb (lg/L) 0.04 9.23 0.3667 1.09535 10 50 10
Fe (lg/L) 122 24,800 3235.286 3868.188 300–1000 300
Mn (lg/L) 51 3830 652 583.4879 100 100 300
Ni (lg/L) BDL 6.8 1.9086 1.23038 70 100 20
Sb (lg/L) BDL 1.18 0.0673 0.15305 20
Ba (lg/L) 5 657 77.1857 107.0766 700 10 700
Mo (lg/L) BDL 20 2.9157 3.67043 70
Al (lg/L) BDL 93 6.3143 12.61646 200 30 200
Zn (lg/L) 2 82 21.6429 19.36756 5000 5000 15,000
B (mg/L) BDL 0.82 0.1716 0.1799 2.4 1 0.5 1
P (mg/L) BDL 3.4 0.7729 0.78407 0
Si (mg/L) 8.37 31.6 17.4703 5.55114
BDL below detection limit, BMAC Bangladesh maximum admissible concentration
a
FAO standard for irrigation purposes

HCO3. This factor was brought under the purview of many as a mineral from sedimentary rocks or from mining and
natural hydrogeochemical evolution of groundwater industrial waste products (Bhuiyan et al. 2010). Bacterial
through groundwater–geological medium interaction activities on Fe and Mn are also responsible for releasing
(Omo-Irabor et al. 2008). PC2, explaining 17 % of total Fe and Mn (Bromfield 1978; Bhattacharya et al. 2002;
variance, is loaded with P, As, DOC, NH4–N, Mo, Fe, and Naujokas et al. 2013; Islam et al. 2013). PC5 has strong
HCO3 (Table 5). PC2 explains leaching of materials from loadings of Al and Pb with 8 % of total variance, indicating
soil horizon to the aquifer which are basically trace ele- anthropogenic pollution from domestic and agricultural
ments and are regarded as nonpoint pollutant sources along sources. Natural causes like geogenic process along with
with partial natural weathering processes. Arsenic comes anthropogenic activities (like industrialization) control
under this category, i.e., geogenic processes augmented these enrichments and are shown in high scores of PC5 and
through human-induced activities such as fertilizers and S12 and S63 samples. PC6 has moderate to strong loadings
animal feedings (Bhuiyan et al. 2011b). R-mode cluster of Ni and Sb which is linked to release from stainless steel
analyses thus are influenced by both geogenic and and alloy product industries.
anthropogenic activities. Excessive presence of NH4–N has
found to contribute by the use of chemical fertilizer in Spatial similarities and sampling site grouping
agricultural fields (Backman et al. 1998). PC3, accounting
for 14 % of total variance, is associated with Fe, Ba, and In this study, the CA results strongly agree with the PCA
Si, indicating geogenic factors. PC4 is moderately loaded results. The R-mode CA retains four main clusters for
with Mn, accounting for 8.5 % of total variance and is analyzed parameters (Fig. 4a). Elements belonging to the
related to geogenic factors. Generally, Mn occurs naturally same cluster are likely to have originated from a common

123
 1020

123
Table 4 Pollution potential of groundwater samples of the study area based on different quality indices
Index methods Index values/categories Degree of pollution Number of samples % of samples Samples
Page 8 of 23

GWQI 50–100 Good water 23 33 S1, S3,S7,S13, S14, S16, S21, S32, S35–S39, S41, S43, S46, S49, S53, S56, S62,
S66, S68, S69
100.1–200 Poor water 28 40 S5, S11, S12, S15, S19, S20, S22–S26, S28, S29, S33, S34, S40, S42, S44, S50–
S52, S55, S58, S59, S63, S65,S67, S70
200.1–300 Very poor water 5 7 S6, S8, S45, S60, S64
[300 Water unsuitable for 2 3 S10, S18
drinking purposes
Cd \10 Low 42 60 S1–S4, S7, S9, S13, S14, S16, S17, S19, S21, S25, S27, S30–S32, S35–S39, S41,
S43, S46–S51, S53–S57, S61, S62, S65–S69
10–20 Medium 17 24.29 S5, S11, S12, S15, S20, S22, S24, S26, S33, S34, S40, S44, S52, S58, S59, S63, S70
[20 High 11 15.71 S6, S8, S10, S18, S23, S28, S29, S42, S45, S60, S64
HPI with As \25 Low 47 67.14 S1, S2, S4, S7, S9, S13, S14, S17, S20, S21, S24–S32, S34–S44, S47–S50, S53,
S54, S56–S59, S61, S62, S65, S66, S68–S70
25–50 Medium 13 18.57 S3, S5, S12, S16, S18, S19, S33, S46, S51, S55, S60, S64, S67
[50 High 10 14.29 S6, S8, S10, S11, S15, S22, S23, S45, S52, S63
HPI without As \10 Low 42 60 S1–S5, S7, S9, S13–S17, S21–S22, S26–S27, S30–S37, S39, S41, S46–S50, S53–
S57, S59, S61–S62, S66, S69–S70
10–20 Medium 24 34.29 S6, S8, S11–S12, S18–S20, S23–S25, S28–S29, S38, S40, S42–S45, S52, S58, S60,
S65, S67–S68
[20 High 4 5.71 S10, S51, S63–S64
HEI with As \10 Low 53 75.71 S1–S5, S7, S9, S12–S17, S19, S21, S22, S25–S32, S34–S39, S41–S43, S46–S50,
S53–S57, S59–S62, S65–S70
10–20 Medium 16 22.86 S6, S8, S11, S18, S20, S23, S24, S33, S40, S44, S45, S51, S52, S58, S63, S64
[20 High 1 1.43 S10
\50 Excellent water 12 17 S2, S4, S9, S17, S27, S30, S31, S47, S48, S54, S57, S61
HEI without As \5 Low 39 55.71 S1–S4, S9, S13–S17, S21–S22, S26–S27, S29–S32, S34, S36–S37, S41–S42, S46–
S48, S50, S53–S57, S59, S61–S62, S67–S70
5–10 Medium 21 30 S5–S7, S11–S12, S19, S23, S25, S28, S33, S35, S38–S39, S43–S45, S49, S51, S60,
S65–S66
[10 High 10 14.29 S8, S10, S18, S20, S24, S40, S52, S58, S63–S64
Environ Earth Sci (2016) 75:1020
 Table 5 Groundwater pollution indices
Sample GWQI CD HPI HPI HEI HEI Sample GWQI CD HPI HPI HEI HEI Sample ID GWQI CD HPI HPI HEI HEI
ID with without with without ID with without with without with without with without
As As As As As As As As As As As As

S1 84.4 6.16 8.43 7.83 3.87 3.61 S31 35.3 0.17 4.72 3.2 1.65 1.64 S61 49 2.22 19.4 5.42 3.8 2.19
S2 27.4 0.3 4.25 2.51 1.32 1.31 S32 51.2 1.43 5.35 4.72 1.62 1.32 S62 64.3 3.19 20.6 7.89 5.47 3.95
S3 78.4 5.46 36.4 8.48 6.59 3.59 S33 116 10.1 28.1 9.94 10.2 8.17 S63 139 10.8 67 59 12.2 10.7
Environ Earth Sci (2016) 75:1020

S4 42.7 1.66 4.89 3.22 3.01 3 S34 110 10.5 22 7.05 4.95 3.23 S64 286 36.6 43.1 34.9 17.9 16.6
S5 109 10.2 45.2 9.43 9.14 5.36 S35 90.3 6.93 6.29 5.95 6.85 6.84 S65 100 7.34 16.5 12 6.63 5.9
S6 257 35.5 81.5 10.2 15.4 8.13 S36 65.4 2.35 12.8 5.95 3.62 2.7 S66 98.6 8 12.8 8.52 7.37 6.68
S7 79.5 5.9 15.2 9.45 8.07 7.23 S37 92.8 7.6 17 6.98 3.54 2.3 S67 109 7.64 26.2 11 6.55 4.78
S8 230 27.1 93.3 17.2 19.5 11.7 S38 77.6 4.51 17.5 11.7 7.35 6.49 S68 56.5 3.26 12.5 12.3 3.91 3.72
S9 28 0.4 4.17 2.19 1.46 1.45 S39 87.3 5.88 10.8 9.7 7.23 7.22 S69 53.3 3.55 4.34 2.97 2.93 2.89
S10 394 51.5 81.2 30.6 34.4 29 S40 149 14.5 13.5 12.6 13.7 13.7 S70 107 10 23 6.16 4.01 2.12
S11 134 13.8 58.7 11 11.2 6.28 S41 84.7 5.53 18.2 9.8 5.25 4.15 Min 24.7 0.17 4.17 2.19 1.11 1.1
S12 137 13.8 36.7 11.3 9.29 6.51 S42 171 24.7 14.9 14.1 4.35 4.34 Max 431 66.2 93.3 59 34.4 29
S13 96 7.25 20.2 5.99 4.09 2.45 S43 91.2 6.77 33.6 15.1 9.01 6.87 Mean 114 11.2 26.1 10.4 7.44 5.73
S14 87.9 7.37 20.7 6.57 4.11 2.47 S44 123 10.3 29.1 19.5 10.7 9.37 CV 100 [3 100 100
S15 114 10.4 57.5 8.02 8.55 3.43 S45 221 27.9 64.6 11.2 14.6 9.05 % exceeding 50 % 77 % 0% 0%
CV
S16 85.9 6.53 38.7 7.69 5.83 2.53 S46 76 5.93 25.4 4.55 5.09 2.81
S17 48.3 2.91 6.21 4.65 4.12 4.11 S47 27.7 0.49 5.07 3.41 1.64 1.63
S18 431 66.2 47.1 13.5 16.5 12.9 S48 32.1 0.84 4.91 3.08 2.05 2.04
S19 111 8.71 29.8 15.5 9.47 7.75 S49 78.3 5.47 6.48 4.94 5.51 5.5
S20 139 13.1 14.6 13.7 12.8 12.7 S50 103 8.98 18.8 8.24 4.35 3.05
S21 56.2 2.44 9.07 3.66 1.99 1.23 S51 114 9.22 49 24.1 11.6 8.77
S22 127 13.2 54.2 3.06 6.79 1.55 S52 162 17.8 54.1 18 15.5 11.6
S23 185 20 90.3 10.9 16.2 8.15 S53 65.1 4.14 6.76 5.25 4.37 4.36
S24 136 14.2 14.7 13.6 11.5 11.3 S54 49.2 1.29 18 9.83 4.11 3.03
S25 116 9.82 24.6 10.2 7.08 5.38 S55 102 8.48 36.2 7.69 6.86 3.8
S26 105 10.2 24 6.54 3.83 1.87 S56 58.3 2.67 22.8 8.43 5.34 3.66
S27 43.8 2.17 5.79 4.17 3.54 3.53 S57 43.2 0.93 7.89 4.08 2.22 1.61
S28 173 23.5 20.2 19.3 7.57 7.13 S58 151 16.9 12.3 11.3 10.1 10.1
S29 180 26.4 13.5 12.5 4.83 4.82 S59 118 10.6 16 8.06 4.77 3.72
Page 9 of 23

S30 24.7 0.2 4.46 2.74 1.11 1.1 S60 279 39.8 33.5 14.5 9.07 6.89

Bold and italic numbers indicate relatively higher values

1020

CV critical value

123
1020 Page 10 of 23 Environ Earth Sci (2016) 75:1020

1000 Q-mode CA has been applied to detect the spatial sim-

ilarities and site grouping among the sampling points
(Table 8). Samples are clustered in a particular group
100 which share similar characteristics with respect to the
analyzed parameters. The 70 sampling sites fall into four
clusters (Fig. 4b). Cluster 1 consists of 37 sampling points.
10 These 41 sampling points are S4–S12, S15–S18, S20, S23,
S24, S27, S33, S35, S38–S41, S43–S49, S51–S55, S58,
S62, S63, S65, S66, and S69. Among the sites, strong
1 correlation exists among the element pairs: Ca with Cl
S1 S9 S17 S25 S33 S41 S49 S57 S65 (r = 0.679, p \ 0.01) and Mg with Na (r = 0.786,
GWQI Cd HPI HEI
p \ 0.01). It is worth noting that moderate concentrations
Fig. 2 Spatial variation in groundwater quality index values in the of Ca, Mg, K, SO4, Cl, HCO3, EC, B, and Na have been
study area observed at most of the stations. Cluster 2 contains only
four samples such as S19, S56, S64, and S67. Cluster 3
includes the following 12 sample sites: S1–S3, S14, S22,
Table 6 Correlation coefficient matrix for indices values and metal
concentration S25, S26, S30, S31, S34, S36, and S37. Cluster 4 consists
of 13 sites which are S13, S21, S28, S29, S32, S42, S50,
HPI HEI CD GWQI
S57, S59, S60, S61, S68, and S70. The relationships among
HPI 1 0.774** 0.599** 0.661** the analyzed parameters are also visualized in the factor
HEI 0.774** 1 0.760** 0.830** loading plots of PC1 versus PC2 and PC1 versus PC3
Cd 0.599** 0.760** 1 0.988** (Figs. 5, 6). Five major clusters are obtained for all
GWQI 0.661** 0.830** 0.988** 1 parameters from the plotting of PC1 versus PC2 (Fig. 5a).
As 0.941** 0.627** 0.498** 0.544** Cluster 1 contains parameters As, P, DOC, Mo, and Fe.
Pb 0.235 0.116 -0.002 0.042 Cluster 2 consists of F, Al, pH, Ni, Zn, Mn, Ba, Al, Sb, and
Fe 0.423** 0.850** 0.615** 0.661** Pb. Cluster 3 includes B, Mg, Be, Na, Cl, K, and SO4.
Mn 0.349** 0.429** 0.366** 0.419** NH4–N, HCO3, Ca belong to cluster 4, whereas Si inde-
Ni 0.350** 0.580** 0.506** 0.539** pendently remains in cluster 5. Near similar groupings of
Sb 0.118 0.1 -0.007 0.016 parameters are observed in the plot of PC1 versus PC3
Ba -0.108 0.314** 0.261* 0.259*
(Fig. 5b) with some exceptions of Si, pH’s groupings. For
Mo 0.445** 0.176 0.309** 0.313**
sampling sites, the PC1 versus PC2 and PC1 versus PC3
Al 0.281* 0.250* 0.097 0.144
plots show same common sources having four distinct
clusters (Fig. 6a, b) which differ from the clustering phe-
Zn 0.145 0.164 0.331** 0.312**
nomenon that are shown in Fig. 5a, b.
* Correlation is significant at the 0.05 level (two-tailed)
** Correlation is significant at the 0.01 level (two-tailed) Correlation matrix analysis

Pearson’s correlation coefficient matrix is generated in

order to identify the rotations among the parameters and
source. The R-mode CA performed on the groundwater sources of groundwater pollution (Table 9). Correlation
samples produces four clusters. Cluster 1 includes EC, Na, matrix shows that inter-parameter relationships agree with
Cl, Mg, SO4, K, HCO3, and B which may be explained by the results obtained from PCA. It also shows some new
combining nonpoint sources and leaching of fertilizers associations between the parameters that are not adequately
from the soil horizon to the aquifer. Cluster 2 consists of reported in the previous sections. Strong (p \ 0.01) and
Mn, Zn, Ca, Ni, Fe, and Ba reflecting the influence of significant correlations (p \ 0.05) are observed in most of
domestic and agricultural pollution (Omo-Irabor et al. the parameters of groundwater samples. EC has a strong
2008). Cluster 3 includes As, DOC, NH4–N, P, Mo, and pH positive correlations at p \ 0.01 with Na? (r = 0.959), K?
and is explained by the dissolution of minerals under basic (r = 0.682), Mg2? (r = 0.847), Ca2? (r = 0.645), NH42-–
condition. Cluster 4 includes Pb, Al, Si, and Sb which N (r = 0.538), SO42- (r = 0.685), HCO3- (r = 0.608),
represent the presence of anthropogenic and geogenic Cl- (r = 0.939), and B (r = 0.842), and these parameters
activities. Although there are some variations between the are also positively correlated with each other (Table 8).
results of CA and PCA, a good agreement between the two These associations indicate mixed sources of geogenic/
statistical techniques is evident in all the data sets analyzed. anthropogenic origin which are described in PC1

123
Environ Earth Sci (2016) 75:1020 Page 11 of 23 1020

(a) (b)

Fig. 3 a Scree plot of the characteristic roots (eigenvalues) of principal component analysis, b Component plot in rotated space of principal
component analysis

section. Na?, Cl-, Mg2?, K? are the main constituents of the correlations between As, Fe, and Mn in groundwater
groundwater as a result of interaction with minerals or are due to the precipitation of dissolved Fe as siderite solids
trapped saline fluids in aquifers and chemical weathering of (FeCO3) or rhodochrosite (MnCO3) under reduced envi-
catchment rocks. The partially acidic nature of ground- ronmental conditions (Reza et al. 2010a). High As, low Fe,
water is due to leaching of altered rocks and anthropogenic low Mn in groundwater of the Meghna flood plain in
sources. DOC shows significant correlations with As southeastern parts of Bangladesh are evaluated by Reza
(r = 0.617, p \ 0.01), Mo (r = 0.479, p \ 0.05), Fe et al. (2010b). However, their findings quite differ from the
(r = 0.414, p \ 0.05), NH42-–N (r = 0.422, p \ 0.05), observations in this study where very high concentrations
and P (r = 0.583, p \ 0.01). These results are also show- of Fe, Mn, and As are observed. A positive strong corre-
ing similarity with PC2. These correlation results are lation is observed between Pb and Al (r = 0.843,
attributed to geogenic sources from the basement rocks. p \ 0.01), indicating similar sources as observed in PC5.
Harvey et al. (2002) have reported that DOC in ground- Ni exhibits significant correlation with Sb (r = 0.480,
water in Bangladesh is positively correlated with As and p \ 005) as observed in P62 section.
has concluded that the mobility of As has been closely
related to the recent inflow of carbon or desorption of As Geostatistical modeling
by carbonate rock. pH shows negative significant correla-
tions with Fe (r = -0.488, p \ 0.05) and Si (r = -0.696, Semivariogram models are employed after normalizing the
p \ 0.01). The metal pairs Fe–Ni, Fe–Ba, and Fe–P show data using ArcGIS (version 10.2). Among other models,
significant correlations (at p \ 0.05) with correlation OK (ordinary kriging) is applied in this study. The nugget,
coefficients (r) 0.493, 0.609, and 0.458, respectively, sill, and the range values of the best fit semivariogram
depicting a similar sources as mentioned in PC3 sec- models for quality parameters are shown in Table 10.
tion. Similar observations are made by Chapagain et al. Extremely low nugget effects and the absence of variability
(2010) in the deep groundwater quality in Kathmando, in groundwater elevation at short distances demonstrate
Nepal, where the occurrence of heavy metals is possibly that there is an insignificant small-scale variability mea-
influenced by redox levels and nature of underlying sedi- surement error. Thus, the fit semivariogram represents very
ment (i.e., mineral composition and organic matter con- well the spatial structure of these variables in the ground-
tents) of groundwater. A poor correlation exists between water. The error statistics are estimated to ascertain the
As with Fe (r = 0.305, p \ 0.05) and Mn (r = 0.142, reproducibility of observed values by the theoretical
p \ 0.05), whereas As shows significant correlations with semivariogram models and developed spatial maps
P (r = 0.773, p \ 0.01) and Mo (r = 0.543, p \ 0.01). So, (Table 10). The best fit semivariogram model is chosen

123
1020 Page 12 of 23 Environ Earth Sci (2016) 75:1020

Table 7 Varimax rotated

Parameters PC1 PC2 PC3 PC4 PC5 PC6
principal component analysis
(R-mode) for groundwater R-mode
samples
pH 0.212 0.265 -0.732 -0.26 -0.058 -0.249
EC 0.957 0.158 0.161 0.047 -0.015 0.023
DOC -0.016 0.792 0.035 0.021 -0.041 -0.056
HCO3 0.627 0.5 -0.355 -0.059 0.03 -0.022
Cl 0.917 -0.029 0.271 0.141 -0.035 0.016
F -0.16 0.233 -0.366 -0.702 -0.032 0.009
SO24 0.762 -0.269 -0.117 0.24 -0.039 0.008
NH4_N 0.416 0.701 0.109 -0.059 0.012 -0.009
Na 0.977 0.025 0.017 -0.044 -0.046 0.01
K 0.782 -0.01 -0.254 0.208 -0.025 -0.066
Ca 0.529 0.332 0.517 0.486 0.059 0.007
Mg 0.839 0.056 0.139 0.413 0.056 -0.006
As -0.105 0.865 -0.189 0.11 0.007 0.077
Pb -0.012 -0.101 -0.003 0.101 0.947 0.034
Fe 0.006 0.517 0.74 0.165 0.076 0.077
Mn 0.267 0.094 -0.207 0.77 0.182 0.208
Ni 0.333 0.215 0.376 0.304 0.23 0.637
Sb -0.067 -0.054 -0.021 0.019 -0.038 0.944
Ba 0.16 -0.032 0.91 -0.077 -0.071 -0.042
Mo 0.153 0.663 -0.314 -0.49 0.076 0.065
Al -0.015 0.089 0.054 0.041 0.955 0.013
Zn 0.361 0.177 0.004 0.01 0.118 0.171
B 0.84 0.19 -0.325 -0.202 -0.018 0.007
P 0.029 0.888 0.029 -0.097 0.008 0.019
Si -0.326 -0.477 0.692 -0.057 0.1 -0.074
Eigenvalues 7.468 4.596 3.42 2.073 1.438 1.158
% of variance 26.869 17.288 14.224 8.546 7.82 5.865
Cumulative % 26.869 44.157 58.382 66.928 74.748 80.613

(a) (b)

Fig. 4 a Dendrogram showing the hierarchical clusters of analyzed parameters. Dashed lines in dendrogram represent Phenon lines.
b Dendrogram showing the hierarchical clusters of analyzed samples site. Dashed lines in dendrogram represent Phenon lines

based on ME, MSE, RMSE, RMSSE, ASE criterion. A RMSSE is close to 1 (Adhikary et al. 2010). Ranges of
model is considered robust and accurate when ME and semivariograms are a measure of extension where auto-
MSE are close to zero, RMSE and ASE are small, and correlation existed (Li et al. 2009). The distance where the

123
Environ Earth Sci (2016) 75:1020 Page 13 of 23 1020

Table 8 Varimax rotated principal component (Q-mode) analysis for groundwater samples
Sample PC1 PC2 PC3 PC4 PC5 PC6 Sample PC1 PC2 PC3 PC4 PC5 PC6

S1 0.43 -0.21 -0.49 -0.88 0.11 0.23 S36 -0.26 -0.12 -0.71 -1.25 -0.04 -0.21
S2 -0.62 -0.51 0.16 -0.19 -0.40 -0.88 S37 -0.12 -0.04 -0.70 -0.35 -0.03 -0.29
S3 -0.49 0.75 -0.82 1.14 -0.40 -0.56 S38 -0.83 -0.31 -0.40 1.29 -0.51 -0.43
S4 -0.75 -0.62 0.56 -0.21 -0.45 -0.73 S39 -0.36 -0.97 1.33 -0.72 -0.37 1.37
S5 -0.77 0.55 -0.21 0.44 -0.47 1.40 S40 -0.26 -0.48 2.20 -0.14 -0.40 0.05
S6 1.09 3.09 -0.06 -2.04 -0.04 -0.16 S41 -0.45 -0.38 -0.30 0.85 -0.32 0.95
S7 -0.80 -0.30 0.07 0.49 0.16 -0.40 S42 3.57 -1.26 -0.82 0.67 -0.14 0.27
S8 -0.54 2.64 0.10 1.20 -0.12 0.09 S43 -0.53 -0.29 -0.35 0.64 0.46 0.36
S9 -0.62 -0.79 0.33 -0.54 -0.14 -0.32 S44 -0.16 0.03 -0.68 1.98 -0.18 0.02
S10 0.12 3.51 2.62 1.35 1.11 -0.16 S45 -0.46 1.87 0.35 0.52 -0.39 -0.26
S11 -0.62 1.59 -0.48 -0.19 0.01 -0.18 S46 -0.80 0.21 -0.39 -0.04 -0.46 -0.35
S12 -0.09 1.25 -0.45 -0.86 1.35 0.62 S47 -0.74 -0.97 0.35 -0.47 -0.13 -0.68
S13 0.29 0.04 -0.44 -0.68 -0.05 -0.17 S48 -0.71 -0.98 0.31 -0.52 -0.15 -0.43
S14 0.53 0.66 -0.93 -2.58 0.45 0.51 S49 -0.32 -0.81 1.23 -0.30 -0.53 -0.13
S15 -0.77 0.75 -0.54 0.80 -0.23 -0.41 S50 0.55 0.19 -0.75 -1.13 0.27 0.45
S16 -0.67 0.40 -0.92 0.07 0.02 -0.35 S51 -0.79 -0.41 -0.36 0.04 -0.23 7.25
S17 -0.75 -0.91 0.81 -0.60 -0.23 -0.38 S52 -0.85 1.12 0.36 1.29 -0.28 0.40
S18 2.30 2.53 1.63 -1.02 -0.23 -0.23 S53 -0.53 -0.86 0.87 -0.32 -0.31 -0.11
S19 -0.45 -0.15 -0.67 1.36 -0.32 -0.33 S54 -0.41 -0.45 -0.46 -0.66 1.17 0.60
S20 -0.35 -0.26 1.96 0.09 -0.32 -0.39 S55 -0.43 0.58 -0.53 -0.71 -0.24 0.11
S21 -0.08 -0.19 -0.77 -1.13 -0.18 -0.37 S56 -0.78 -0.61 -0.41 0.47 -0.35 0.16
S22 -0.16 0.89 -1.30 -1.34 -0.35 -0.37 S57 -0.03 -0.70 -0.94 -0.82 -0.20 -0.58
S23 -0.83 1.33 -0.73 0.85 -0.53 -0.41 S58 1.23 -0.93 3.68 -0.43 -0.51 -0.59
S24 0.48 -0.31 2.99 -0.24 -0.06 1.29 S59 0.76 -0.42 -0.30 -0.39 -0.19 -0.03
S25 0.08 0.31 -0.69 -0.12 -0.19 -0.20 S60 2.59 0.25 0.07 0.20 -0.18 -0.22
S26 -0.02 0.54 -0.90 -1.89 0.40 0.28 S61 -0.67 -0.37 -0.66 0.40 -0.31 -0.74
S27 -0.73 -0.84 0.89 -0.69 -0.25 -0.32 S62 -0.44 -0.33 -0.66 -0.01 0.19 -0.46
S28 2.36 -0.19 -0.22 1.36 -0.05 0.45 S63 -0.19 -0.84 0.19 0.69 7.68 -0.30
S29 3.93 -0.99 -0.68 0.27 -0.13 0.15 S64 1.00 -0.07 0.00 3.84 -0.20 0.24
S30 -0.67 -0.85 0.01 -0.26 0.04 -0.81 S65 0.46 -0.51 -0.79 0.72 -0.13 -0.26
S31 -0.50 -0.95 0.26 -0.26 -0.37 -0.29 S66 0.20 -0.73 -0.17 0.88 -0.38 -0.08
S32 0.10 -0.29 -1.03 -0.86 -0.14 -0.21 S67 0.01 -0.20 -0.72 1.31 -0.36 -0.47
S33 -0.61 0.29 0.03 0.51 -0.39 -0.39 S68 -0.13 -0.67 -0.54 0.90 0.22 -0.34
S34 0.27 0.36 -0.78 -0.89 0.09 -0.14 S69 -0.67 -0.91 0.70 -0.62 -0.03 -0.63
S35 -0.15 -0.77 1.64 -0.38 0.23 -0.06 S70 1.61 -1.01 -0.98 0.11 -0.37 -0.46

models first flatten out is known as the range which varies more than 75 %, it represents weak spatial dependence.
in each variable of groundwater quality indices. The range Figures 7 and 8 show experimental semivariograms (bin-
of semivariograms for all the variables is from 0.5 to ned sign) around the omnidirectional semivariogram model
22 km showing the lowest range for PC1 and highest range (blue line) and average of semivariogram models (plus
for GWQIs. sign). The best fit semivariogram models for different
The nugget/sill ratio represents spatial dependence of groundwater quality parameters are shown in Figs. 7 and 8,
groundwater quality parameters (Nayanaka et al. 2010). and Table 10. The Gaussian semivariogram model is
There are three classifications used for model explanation: identified to be the best fit model for CD, HPI, PC3, PC4,
If the ratio is less than 25 %, it shows strong spatial PC5, PC6, while the circular semivariogram model fit best
dependence; if the ratio is in between 25 and 75 %, it for GWQIs. The exponential semivariogram model is
indicates moderate spatial dependence; and if the ratio is observed to be the best fit model for HEI and PC2, whereas

123
1020 Page 14 of 23 Environ Earth Sci (2016) 75:1020

(a) (b)

Fig. 5 Plots of first three principal component loadings. a PC1 versus PC2 and b PC1 versus PC3 for all analyzed parameters

(a) (b)

Fig. 6 Plots of first three principal component loadings. a PC1 versus PC2, b PC1 versus PC3 for all analyzed sampling sites

the spherical model fit well for PC1. In this study, the southeast to northwest direction of the study area (Fig. 9).
results show that most of the variables are weakly spatial In general, contamination of groundwater with metal pol-
dependent in semivariogram shapes except PC1 and HEI. lution is attributed to anthropogenic sources. Other than
PC1 represents strong spatial dependence, whereas HEI this, both geogenic and anthropogenic sources are likely to
shows moderate spatial dependence (Fig. 7). Mostly, the be contributed by the high scores of GWQIs, CD, HPI, and
weak spatial dependence has been demonstrated in the HEI in the study area. High scores are observed at the
large nugget effect in semivariogram shape (Fig. 8) and is northern and northwestern parts of the sampling area, and
due to the high variability of topography of groundwater low scores are observed at the southeastern part, suggesting
which varies with the agriculture, residential, and industrial the existence of similar point sources. The strong signifi-
areas. cant correlations among GWQIs, CD, HPI, and HEI show
The OK interpolation techniques are applied to develop similar distribution pattern of the previous findings.
spatial distribution maps of groundwater data set (n = 70) Spatial distribution of PC1 score in Fig. 10a reveals
for each groundwater pollution index and each PC score. complex pattern where high scores (i.e., values from 0.42
After conducting the cross-validation test of best fit mod- to 3.93) are mostly observed at the southern parts and low
els, maps of kriged estimates are generated which provide a scores are found at the northern and central parts. The
visual representation of the distribution of groundwater higher score of PC1 is probably due to the rock–water
samples. GWQIs, CD, HPI, and HEI exhibit more or less interaction and ion exchange in groundwater (Guler et al.
similar distribution patterns with an increasing trend at the 2012). Similar findings are obtained from the recent studies

123
 Table 9 Pearson’s correlation matrix among physicochemical parameters and metals in the analyzed samples
pH EC DOC HCO3 Cl F SO4 NH4–N Na K Ca Mg As

pH 1
EC 0.098 1
DOC 0.193 0.14 1
HCO3 0.458** 0.608** 0.374** 1
Cl -0.02 0.939** -0.023 0.367** 1
F 0.484** -0.215 0.126 0.098 -0.319** 1
Environ Earth Sci (2016) 75:1020

SO4 0.124 0.685** -0.146 0.349** 0.687** -0.343** 1

NH4–N 0.198 0.538** 0.422** 0.512** 0.428** 0.114 -0.042 1
Na 0.2 0.959** 0.032 0.615** 0.922** -0.158 0.749** 0.432** 1
K 0.321** 0.682** -0.013 0.570** 0.645** -0.169 0.632** 0.279* 0.705** 1
Ca -0.246* 0.645** 0.243* 0.293* 0.679** -0.485** 0.329** 0.458** 0.478** 0.0385** 1
Mg -0.004 0.847** -0.007 0.431** 0.894** -0.372** 0.635** 0.427** 0.786** 0.697** 0.756** 1
As 0.318** 0.005 0.617** 0.372** -0.123 0.21 -0.2 0.510** -0.064 -0.019 0.166 -0.001 1
Pb -0.115 -0.025 -0.132 -0.054 -0.008 -0.14 0.004 -0.046 -0.05 -0.01 0.056 0.087 -0.016
Fe -0.488** 0.213 0.414** 0.021 0.182 -0.252* -0.14 0.397** 0.014 -0.1 0.620** 0.172 0.305*
Mn -0.081 0.270* 0.002 0.223 0.306* -0.349** 0.317** 0.161 0.204 0.320** 0.469** 0.592** 0.143
Ni -0.345** 0.425** 0.176 0.16 0.433** -0.308** 0.214 0.240* 0.302* 0.195 0.631** 0.475** 0.145
Sb -0.222 -0.044 -0.075 -0.071 -0.035 -0.063 -0.011 -0.047 -0.039 -0.068 -0.052 -0.069 0.059
Ba -0.543** 0.270* 0.018 -0.231 0.371** -0.301* 0.042 0.062 0.156 -0.121 0.536** 0.234 -0.193
Mo 0.509** 0.152 0.479** 0.567** -0.047 0.582** -0.153 0.480** 0.153 0.071 -0.073 -0.043 0.543**
Al -0.075 -0.005 0.085 0.062 -0.038 -0.079 -0.029 0.034 -0.048 -0.03 0.121 0.044 0.059
Zn 0.098 0.358** 0.111 0.187 0.313** -0.017 0.22 0.292* 0.299* 0.223 0.269* 0.302* 0.099
B 0.435** 0.766** 0.118 0.853** 0.617** 0.123 0.560** 0.426** 0.842** 0.663** 0.221 0.563** 0.103
P 0.227 0.179 0.583** 0.471** -0.01 0.194 -0.17 0.633** 0.075 -0.041 0.257* 0.016 0.773**
Si -0.696** -0.296* -0.373** -0.664** -0.11 -0.267* -0.246* -0.340** -0.322** -0.420** -0.001 -0.19 -0.492**
Pb Fe Mn Ni Sb Ba Mo Al Zn B P Si

pH
EC
DOC
Page 15 of 23

HCO3
Cl
1020

F
SO4
NH4–N
Na

123
 Table 9 continued
1020

Pb Fe Mn Ni Sb Ba Mo Al Zn B P Si

123
K
Ca
Mg
Page 16 of 23

As
Pb 1
Fe 0.022 1
Mn 0.216 0.025 1
Ni 0.21 0.493** 0.425** 1
Sb 0.036 0.037 0.133 0.480** 1
Ba -0.061 0.609** -0.212 0.330** -00.055 1
Mo -0.06 0.01 -0.081 -0.019 -0.019 -0.205 1
Al 0.843** 0.194 0.17 0.278* -0.005 -0.03 0.072 1
Zn 0.054 0.082 0.178 0.300* -0.012 0.005 0.172 0.067 1
B -0.075 -0.156 0.178 0.129 -0.059 -0.163 0.474** -0.034 0.223 1
P -0.079 0.458** -0.019 0.165 -0.013 -0.022 0.603** 0.086 0.134 0.213 1
Si 0.12 0.196 -0.257* 0.003 -0.041 0.551** -0.513** 0.077 -0.219 -0.551** -0.377** 1

* Correlation is significant at the 0.05 level (two-tailed)

** Correlation is significant at the 0.01 level (two-tailed)
Environ Earth Sci (2016) 75:1020
Environ Earth Sci (2016) 75:1020 Page 17 of 23 1020

Table 10 Indices and PCs of best fit semivariogram models for groundwater parameters and their variance
Indices and PCs Fit model types Nugget Major range Sill Nugget/ Lag size ME RMSE MSE RMSSE ASE
Sill

GWQI Circular 5911.38 22,381.50 6055.74 0.976 2797.69 -0.826 80.49 -0.01 1.01 79.52
CD Gaussian 147.90 22,381.49 147.90 1.00 2797.69 -0.044 12.68 -0.003 1.01 12.55
HPI Gaussian 411.10 8719.78 411.10 1.00 1089.98 -0.627 21.61 -0.029 1.03 20.95
HEI Exponential 19.42 8142.02 25.72 0.76 1017.75 -0.139 5.45 -0.025 1.07 5.04
PC1 Spherical 0.20 5304.55 1.07 0.19 663.07 -0.004 0.98 -0.001 1.09 0.89
PC2 Exponential 0.73 8354.05 0.86 0.84 1044.26 -0.023 0.95 -0.023 1.01 0.94
PC3 Gaussian 0.91 6743.03 0.91 1.00 842.88 -0.001 1.01 -0.001 1.01 0.99
PC4 Gaussian 1.04 11,276.01 1.039 1 1409.50 -0.011 1.01 -0.009 0.96 1.05
PC5 Gaussian 0.93 9418.91 0.927 1 1177.37 -0.009 1.03 -0.009 1.03 0.99
PC6 Gaussian 1.04 13,156.66 1.036 1 1644.58 0.005 1.01 0.004 0.97 1.05
ME mean error, RMSE root mean square error, MSE mean standardized error, RMSSE root mean square standardized error, ASE average standard
error

Fig. 7 Best fit semivariogram models of groundwater quality evaluation indices in the study area. a GWQI, b CD/Cd, c HPI, and d HEI

conducted in various regions of the world (e.g., Fernandes Figure 10c shows the distribution of PC3 scores, where
et al. 2008; Guler et al. 2012). The spatial distribution of moderately higher scores (i.e., values from 0.23 to 1.91) are
PC2 scores shows high scores (values from 0.02 to 3.52) at observed at the northern corner of the study area and the
the northern part and the low scores (i.e., values from lower scores are found at the southern edge. PC3 contains
-1.26 to -0.38) at the southern part (Fig. 10b). This very high scores of Fe, Ba, and Si which are likely due to
variation is influenced by the dominance of nonpoint geogenic origin. Figure 10d showing the distribution of
sources such as fertilizer and sewage discharges. For PC4 scores reverses the distribution pattern of PC3 score,
example, high NH4 and P in PC2 of groundwater samples where the higher scores (from 0.12 to 2.18) are observed at
may be attributed to the use of fertilizers and urban runoff. the southern part and the lower scores are observed at the

123
1020 Page 18 of 23 Environ Earth Sci (2016) 75:1020

Fig. 8 Best fit semivariogram models of groundwater parameters in the study area. a PC1, b PC2, c PC3, d PC4, e PC5, and f PC6

northern part. PC4 comprises of the high positive score of The best fit semivariogram models are selected based on
Mn and negative score of F reflecting the geogenic origin. the criterion of ME, MSE, RMSE, RMSSE, and ASE
The distribution of PC5 component score in Fig. 10e shows which varied for each PC and index. The Gaussian semi-
higher score (values from 0.12 to 2.18) at the eastern part variogram model is identified to be the best fit model for
and the lower score at the western part of the study area. CD, HPI, PC3, PC4, PC5 and PC6 while the circular
Overall, the map shows a decreasing trend toward the east. semivariogram model fit best for GWQIs. The exponential
In PC5, high score of Al is strongly correlated with Pb, semivariogram model is observed to be the best fit model
indicating the anthropogenic origin. Figure 10f shows the for HEI and PC2, whereas the spherical model fit well for
spatial distribution of PC6 component score. The highest PC1. In the study, most of the variables have weak spatial
score (i.e., values from 1.00 to 7.25) is found at the central dependence except PC1 and HEI. PC1 represents strong
part which comprises of high score of Sb and Ni. This spatial dependence, whereas HEI shows moderate spatial
component is contributed by the wastewater discharge dependence. However, the weak spatial dependence indi-
from urban areas. cates high variability of topography in groundwater system
Among the geostatistical models, OK is considered as an (e.g., agriculture, residential, and industrial areas). The
effective tool for preliminary decision makers of ground- cross validation indicates that kriging method is the most
water quality management at the southeastern Bangladesh. accurate interpolation technique for the study. The spatial

123
Environ Earth Sci (2016) 75:1020 Page 19 of 23 1020

(a) (b)

– 0 2.5 5

Kilometers
10

– 0 2.5 5

Kilometers
10

Cd
GWQI

(c) (d)

– 0 2.5 5

Kilometers

HPI
10

– 0 2.5 5

Kilometers
10

HEI

Fig. 9 Maps showing the spatial distribution of four index scores obtained by the indices of quality evaluation of the groundwater samples.
a GWQI, b CD/Cd, c HPI, and d HEI. Here, both HPI and HEI are plotted considering with As and other metals

distribution maps are generated using OK method showing salinity observed at the northwestern part of PC1 of the
the variation in different indices in the study area. Con- study area is alarming, as households depend upon
sidering the distribution maps, six PC groups (PC1–PC6) groundwater for domestic purposes.
and four indices (GWQIs, CD, HPI, and HEI) with dif-
ferent spatial variability patterns can be distinguished
(Figs. 9, 10). Indices scores exhibit more or less similar Conclusions
distribution pattern with an increasing trend at the south-
east to northwest direction where high scores of GWQIs, Integrated approaches of different multivariate statistical
CD, HPI, and HEI are observed at the northern and and geostatistical techniques are employed in this study to
northwestern parts of the study area (Fig. 9a–d). The low evaluate the variations in groundwater quality of the Lak-
scores are found at the southeastern part which suggests the shimpur district of Bangladesh. Cluster analysis groups 70
existence of similar point sources. However, spatial anal- sampling points into four clusters of similar water quality
ysis of GWQIs depicts that water quality is poor at the characteristics. Based on the obtained information, it may
western and northwestern parts of the study area. Different be easier to design the study area further where optimal
distribution patterns of the PC scores have implied the sampling strategies can reduce the number of sampling
existence of different sources. Distribution maps of PC1 to sites and associated costs. Principal component analysis
PC6 have shown some anomalies regarding the mean assists for identifying the factors or sources responsible for
scores of groundwater variables (Fig. 10a–f). In fact, water quality variations. This study illustrates the

123
1020 Page 20 of 23 Environ Earth Sci (2016) 75:1020

(a) (b)

0 2.5 5 10 0 2.5 5 10

Kilometers Kilometers

PC1 PC2
-0.85 – -0.77 -1.26 – -0.87

-0.77 – -0.72 -0.87 – -0.63

-0.72 – -0.64 -0.63 – -0.47

-0.64 – -0.52 -0.47 – -0.38

-0.52 – -0.33 -0.38 – -0.22

-0.33 – -0.04 -0.22 – 0.02

-0.04 – 0.42 0.02 – 0.41

0.42 – 1.13 0.41 – 1.02

1.13 – 2.22 1.02 – 1.98

2.22 – 3.93 1.98 – 3.52

(c) (d)

0 2.5 5 10 0 2.5 5 10

Kilometers Kilometers

PC3
PC4
-1.30 – -0.93
-2.58 – -1.55
-0.93 – -0.71
-1.55 – -0.91
-0.71 – -0.58
-0.91 – -0.52
-0.58 – -0.50
-0.52 – -0.28
-0.50 – -0.37
-0.28 – -0.13
-0.37 – -0.14
-0.13 – 0.12
-0.14 – 0.23
0.12 – 0.51
0.23 – 0.86
0.51 – 1.15
0.86 – 1.91 1.15 – 2.18
1.91 – 3.68 2.18 – 3.87

(e) (f)

0 2.5 5 10 0 2.5 5 10

Kilometers Kilometers

PC5 PC6
-0.53 – -0.39 -0.88 – -0.54

-0.39 – -0.31 -0.54 – -0.40

-0.31 – -0.28 -0.40 – -0.34

-0.28 – -0.20 -0.34 – -0.32

-0.20 – -0.06 -0.32 – -0.26

-0.06 – 0.22 -0.26 – -0.12

0.22 – 0.76 -0.12 – 0.22

0.76 – 1.79 0.22 – 1.0

1.79 – 3.80 1.0 – 2.86

3.80 – 7.68 2.86 – 7.25

Fig. 10 Maps showing the spatial distribution of six PC scores. a PC1: Na, K, Ca, Mg, Cl, SO4, HCO3, EC, and B; b PC2: P, Mo, As, NH4N,
DOC, and Fe; c PC3: Fe, Ba, and Si; d PC4: Mn and Fe; e Pb and Al; and f Ni and Sb

123
Environ Earth Sci (2016) 75:1020 Page 21 of 23 1020

usefulness of multivariate statistical and geostatistical Backman B, Bodis D, Lahermo P, Rapant S, Tarvainen T (1997)
techniques for the analysis and interpretation of complex Application of a groundwater contamination index in Finland
and Slovakia. Environ Geol 36:55–64
data sets, identification of pollution sources, and under- Backman B, Bodiš D, Lahermo P, Rapant S, Tarvainen T (1998)
standing spatial variations in water quality for effective Application of a groundwater contamination index in Finland
groundwater management. Moreover, the chemometric and Slovakia. Environ Geol 36(1–2):55–64
studies enable us to demonstrate the similarities and dif- Banglapedia (2006) National encyclopedia of Bangladesh. Asiatic
Society of Bangladesh, Dhaka. https://en.wikipedia.org/wiki/
ferences among the examined variables that are not evi- Banglapedia
dently visible from an examination of the analytical data as Bhattacharya P, Frisbie SH, Smith E, Naidu R, Jacks G, Ahmed KM,
shown in the tables. The ordinary kriging is an effective Khan AA, Routh J (2002) Arsenic in groundwater of the Bengal
tool for preliminary decisions makers of groundwater Delta plain aquifers in Bangladesh. Bull Environ Contam
Toxicol 69:538–545
quality management in southeastern Bangladesh. Except Bhuiyan MAH, Islam MA, Dampare SB, Parvez L, Suzuki S (2010)
As, other geogenic pollutants are not much alarming for the Evaluation of hazardous metal pollution in irrigation and
groundwater consumption at the study area. However, drinking water systems in the vicinity of a coal mine area of
some anthropogenic processes are quite unpleasant. northwestern Bangladesh. J Hazard Mater 179:1065–1077.
doi:10.1016/j.jhazmat.2010.03.114
The results of principal component/factor analysis indi- Bhuiyan MAH, Rakib MA, Dampare SB, Ganyaglo S, Suzuki S
cate that anthropogenic (agrogenic, surface runoff, and (2011a) Surface water quality assessment in the central part of
domestic sewage) and natural/geogenic sources (weathering Bangladesh using multivariate analysis. KSCE J Civil Engineer
of source rock) are responsible for variation in physico- 15(6):995–1003. doi:10.1007/s12205-011-1079-y
Bhuiyan MAH, Suruvi NH, Dampare SB, Islam MA, Quraishi SB,
chemical parameters and metal contents in groundwater Ganyaglo S, Suzuki S (2011b) Investigation of the possible
systems at the southeastern coastal region of Bangladesh. sources of heavy metal contamination in lagoon and canal water
The resulting spatial distribution maps provide a helpful and in the tannery industrial area in Dhaka, Bangladesh. Environ
robust visual tool for researchers and policy makers toward Monit Assess 175:633–649. doi:10.1007/s10661-010-1557-6
Bhuiyan MAH, Dampare SB, Islam MA, Suzuki S (2015) Source
defining adaptive measures. The study provides background apportionment and pollution evaluation of heavy metals in water
information on physiochemical parameters, harmful metals, and sediments of Buriganga River, Bangladesh, using multivari-
possible sources, and spatial variation in groundwater sys- ate analysis and pollution evaluation indices. Environ Monit
tems in Lakshimpur district of Bangladesh. Assess 187:4075. doi:10.1007/s10661-014-4075-0
Birth G (2003) A scheme for assessing human impacts on coastal
aquatic environments using sediments. In: Woodcoffe CD,
Acknowledgments The authors would like to acknowledge Chem- Furness RA (eds) Coastal GIS. Wollongong University Papers
istry Division, Atomic Energy Center, Dhaka. The authors gratefully in Center for Maritime Policy, Australia
acknowledge the authority of the Department of Environmental Sci- BIS (Bureau of Indian Standards) (2012) Indian Standard drinking
ences, Jahangirnagar University, Bangladesh, for logistic supports for water-specification, 1st rev., pp 1–8
this study. Bromfield SM (1978) The oxidation of manganous ions under acidic
conditions byan acidophilousactinomycete from acid soil. Aust J
Soil Res 16:91–100
References Burrough PA, McDonnell RA (1998) Principles of geographical
information systems. Oxford University Press, Oxford
Adhikary PP, Chandrasekharan H, Chakraborty D, Kamble K (2010) Chabukdhara M, Nema AK (2012) Assessment of heavy metal
Assessment of groundwater pollution in West Delhi, India using contamination in Hindon River sediments: a chemometric and
geostatistical approach. Environ Monit Assess geochemical approach. Chemosphere 87:945–953
167(1–4):599–615 Chapagain SK, Pandey VP, Shrestha S, Nakamura T, Kazama F
Ahmed ZU, Panaullah GM, DeGloria SD, Duxbury JM (2011) (2010) Assessment of deep groundwater quality in Kathmandu
Factors affecting paddy soil arsenic concentration in Bangladesh: valley using multivariate statistical techniques. Water Air Soil
prediction and uncertainty of geostatistical risk mapping. Sci Pollut 210:277–288. doi:10.1007/s11270-009-0249-8
Total Environ 412–413:324–335 Danielsson A, Cato I, Carman R, Rahm L (1999) Spatial clustering of
Al-Ami MY, Al-Nakib SM, Ritha NM, Nouri AM, Al-Assina A metals in the sediments of Skagerrak/Kattegat. Appl Geochem
(1987) Water quality index applied to the classification and 14:689–706
zoning of Al-Jaysh canal, Bagdad, Iraq. J Environ Sci Health Delhomme JP (1978) Kriging in the hydrosciences. Adv Water Res
A22:305–319 1:251–266
APHA-AWWA-WEF (2005) Standard methods for the examination DoE (Department of Environment) (1997) The environment conser-
of water and wastewater, 20th edn. APHA, AWWA and WEF, vation rules 1997. Government of the People’s Republic of
Washington DC, USA Bangladesh, Dhaka
Astel A, Tsakovski S, Barbieri P, Simeonov V (2007) Comparison of Dragovı́c S, Mihailovı́c N, Gajı́c B (2008) Heavy metals in soils:
self-organizing maps classification approach with cluster and distribution, relationship with soil characteristics and radionu-
principal components analysis for large environmental data sets. clides and multivariate assessment of contamination sources.
Water Res 41:4566–4578 Chemosphere 72:491–549
Astel A, Tsakovski S, Simeonov V, Reisenhofer E, Piselli S, Barbieri Edet AE, Offiong OE (2002) Evaluation of water quality pollution
P (2008) Multivariate classification and modeling in surface indices for heavy metal contamination monitoring. A study case
water pollution estimation. Anal Bioanal Chem 390:1283–1292 from Akpabuyo-Odukpani area, Lower Cross River Basin
(southeastern Nigeria). GeoJournal 5:295–304

123
1020 Page 22 of 23 Environ Earth Sci (2016) 75:1020

Elogne S, Hristopulos D, Varouchakis E (2008) An application of Liu CW, Lin KH, Kuo YM (2003) Application of factor analysis in
Spartan spatial random fields in environmental mapping: focus the assessment of groundwater quality in a black foot disease
on automatic mapping capabilities. Stochastic Environ Res Risk area in Taiwan. Sci Total Environ 313:77–89
Assess 22(5):633–646 Machiwal D, Jha MK (2015) Identifying sources of groundwater
ESRI (2009) ArcGIS desktop software. ArcGIS Desktop 9.3., contamination in a hard-rock aquifer system using multivariate
Redlands statistical analyses and GIS-based geostatistical modeling tech-
Fernandes PG, Carreira P, da Silva MO (2008) Anthropogenic niques. Region Stud, J Hydrol. doi:10.1016/j.ejrh.2014.11.005
sources of contamination recognition—Sines coastal aquifer Marko K, Al-Amri NS, Elfeki AMM (2014) Geostatistical analysis
(SW Portugal). J Geochem Explor 98:1–14 using GIS for mapping groundwater quality: case study in the
Franco-Urı́a AC, López-Mateo E, Roca ML, Fernández-Marcos (2009) recharge area of Wadi Usfan, western Saudi Arabia. Arab J
Source identification of heavy metals in pastureland by multi- Geosci 7:5239–5252. doi:10.1007/s12517-013-1156-2
variate analysis in NW Spain. J Hazard Mater 165:1008–1015 Masoud AA (2014) Groundwater quality assessment of the shallow
Goovaerts P (1997) Geostatistics for natural resource evaluation. aquifers west of the Nile Delta (Egypt) using multivariate
Oxford University Press, New York statistical and geostatistical techniques. J Afr Earth Sci
Gorai AK, Kumar S (2013) Spatial distribution analysis of ground- 95:123–137. doi:10.1016/j.jafrearsci.2014.03.006
water quality index using GIS: a case study of Ranchi Municipal Masoud AA, Atwia MG (2010) Spatio-temporal characterization of
Corporation (RMC) area. Geoinfor Geostat Overview 1:2 the Pliocene aquifer conditions in Wadi El-Natrun area, Egypt.
Gotelli NJ, Ellison AM (2004) A primer of ecological statistics, 1st J Environ Earth Sci 62(7):1361–1374
edn. Sinauer Associates, Sunderland Mohan SV, Nithila P, Reddy SJ (1996) Estimation of heavy metal in
Guler C, Kurt MA, Alpaslan M, Akbulut C (2012) Assessment of the drinking water and development of heavy metal pollution index.
impact of anthropogenic activities on the groundwater hydrology J Environ Sci Health A31:283–289
and chemistry in Tarsus coastal plain (Mersin, SE Turkey) using Naujokas MF, Anderson B, Ahsan H, Aposhian HV, Graziano JH,
fuzzy clustering, multivariate statistics and GIS techniques. Thompson C, Suk WA (2013) The broad scope of health effects
J Hydrol 414–415:435–451 from chronic arsenic exposure: update on a worldwide public
Halim MA, Majumder RK, Nessa SA, Oda K, Hiroshiro Y, Jinno K health problem. Environ Health Persp 121(3):295–302
(2010) Arsenic in shallow aquifer in the eastern region of Nayanaka VGD, Vitharana WAU, Mapa RB (2010) Geostatistical
Bangladesh: insights from principal component analysis of analysis of soil properties to support spatial sampling in a paddy
groundwater compositions. Environ Monit Assess 61:453–472. growing alfisol. Trop Agric Res 22(1):34–44. doi:10.4038/tar.
doi:10.1007/s10661-009-0760-9 v22i1.2668
Handa BK (1981) An integrated water-quality index for irrigation use. Nimic DA, Moore JN (1991) Prediction of water-soluble metal
Indian J Agric Sci 51:422–426 concentrations in fluvially deposited tailings sediments, Upper
Harvey CF, Swartz CH, Badruzzaman ABM, Keon-Blute (2002) Clark Fork Valley, Montana, U.S.A. Appl Geochem 6:635–646
Arsenic mobility and groundwater extraction in Bangladesh. Nkansah K, Dawson-Andoh B, Slahor J (2010) Rapid characterization
Science 22:1602–1606 of biomass using near infrared spectroscopy coupled with
Heberger K, Milczewska K, Voelkel A (2005) Principal component multivariate data analysis: part 1 yellow-poplar (Liriodendron
analysis of polymer-solvent and filler-solvent interactions by tulipifera L.). Bioresour Technol 101(2):4570–4576
inverse gas chromatography. Colloids Surf A 260:29–37 Omo-Irabor OO, Olobaniyi SB, Oduyemi K, Akunna J (2008) Surface
Helena B, Pardo R, Vega M, Barrado E, Fernández JM, Fernández L and groundwater water quality assessment using multivariate
(2000) Temporal evolution of groundwater composition in an analytical methods: a case study of the Western Niger Delta,
alluvial aquifer (Pisuerga River, Spain) by principal component Nigeria. Phys Chem Earth 33:666–673
analysis. Water Res 34(3):807–816 Prasad B, Bose JM (2001) Evaluation of the heavy metal pollution
Horton RK (1965) An index systems for rating water quality. J Water index for surface and spring water near a limestone mining area
Pollut Control Fed 37:300–306 of the lower Himalayas. Environ Geol 41:183–188
Hu K, Li B, Lu Y, Zhang F (2004) Comparison of various spatial Prasad B, Jaiprakas KC (1999) Evaluation of heavy metals in ground
interpolation methods for non-stationary regional soil mercury water near mining area and development of heavy metal
content. Environ Sci 25(3):132–137 pollution index. J Environ Sci Health A34:91–102
Isaaks EH, Srivastava RM (1989) An introduction to applied Rahman MS, Gagnon GA (2014) Bench-scale evaluation of drinking
geostatistics. Oxford University Press, New York water treatment parameters on iron particles and water quality.
Islam IR, Rahman M, Reza AHMS, Rahman M (2013) Groundwater Water Res 48:137–147. doi:10.1016/j.watres.2013.09.018
geochemistry and its implication for arsenic enrichment and Reddy SJ (1995) Encyclopaedia of environmental pollution and
mobilization in shallow alluvial aquifers of Pakshi Union, control, vol 1. Karlia, Environmental Media, p 342
Ishwardi, Pabna, Bangladesh. Int J Chem Mater Sci Reza AHMS, Jean J-S, Lee M-K, Liu C-C, Bundschuh J, Yang H-J,
1(4):069–078 Lee J-F, Lee Y-C (2010a) Implications of organic matter on
Kitanidis PK (1997) Introduction to geostatistics: applications in arsenic mobilization into groundwater: evidence from north-
hydrogeology. Cambridge University Press, Cambridge western (Chapai-Nawabganj), central (Manikganj) and south-
Lattin J, Carrol D, Green P (2003) Analyzing multivariate data. western (Chadpur) Bangladesh. Water Res 44:5556–5574
Duxbury Press, Belmont Reza AHMS, Jean J-S, Yang H-J, Lee M-K, Woodall B, Liu C-C, Lee
Li Z, Fang Y, Zeng G, Li J, Zhang Q, Yuan Q, Wang Y, Ye F (2009) JF, Luo S-D (2010b) a). Occurrence of arsenic in core sediments
Temporal and spatial characteristics of surface water quality by and groundwater in the Chapai-Nawabganj district, Northwest-
an improved universal pollution index in red soil hilly region of ern Bangladesh. Water Res 44(6):2021–2037
South China: a case study in Liuyanghe River watershed. Sahu P, Sikdar PK (2008) Hydrochemical framework of the aquifer in
Environ Geol 58:101–107 and around East Kolkata Wetlands, West Bengal. India Environ
Li F, Huang J, Zeng G, Yuan X, Li X, Liang J, Wang X, Tang X, Bai Geol 55:823–835
B (2013) Spatial risk assessment and sources identification of Sahu BK, Panda RB, Sinha BK, Nayak A (1991) Water quality index
heavy metals in surface sediments from the Dongting Lake, of the River Brahmani at Rourkela industrial complex of Orissa.
Middle China. J Geochem Explor 132:75–83 J Ecotoxicol Environ Monit 1:169–175

123
Environ Earth Sci (2016) 75:1020 Page 23 of 23 1020

Sarbu C, Pop HF (2005) Principal component analysis versus fuzzy Vasanthavigar M, Srinivasamoorthy K, Vijayaragavan K, Ganthi RR,
principal component analysis. A case study: the quality of Chidambaram S, Anandhan P, Manivannan R, Vasudevan S
Danube water (1985e 1996). Talanta 65(5):1215–1220 (2010) Application of water quality index for groundwater
Shrestha S, Kazama F (2007) Assessment of surface water quality quality assessment: Thirumanimuttar sub-basin, Tamilnadu,
using multivariate statistical techniques: a case study of the Fuji India. Environ Monit Ass. doi:10.1007/s10661-009-1302-1
river basin, Japan. Environ Model Soft 22:464–475 Webster R, Oliver M (2001) Geostatistics for environmental scien-
Singh KP, Malik A, Sinha S (2005) Water quality assessment and tists. Wiley, Chichester
apportionment of pollution sources of Gomti River (India) using WHO (2011) Guidelines for drinking-water quality, 3rd ed, vol. 1,
multivariate statistical techniques: a case study. Anal Chim Acta Recommendations. WHO, Geneva
538:355–374 Wunderlin DA, Diaz M, Ame MMV, Pesce SF, Hued AC, Bistoni M
Smith AH, Lingas EO, Rahman M (2000) Contamination of drinking- (2001) Pattern recognition techniques for the evaluation of
water by arsenic in Bangladesh: a public health emergency. Bull spatial and temporal variations in water quality. A case study:
World Health Org 78:1093–1103 Suquia River basin (Cordoba-Artgentina). Water Res
Tabachnick BG, Fidell LS (2007) Using multivariate statistics. 35:2881–2894
Pearson/Allyn and Bacon, London Zou J, Zhang J, Wu J, Zhong F, Gu T (1988) On organic pollution and
Varouchakis EA, Hristopulos DT (2013) Improvement of groundwa- its control in the Haibe estuarine area of the Bohai Bay. Stud
ter level prediction in sparsely gauged basins using physical laws Mar Sin 29:1–20
and local geographic features as auxiliary variables. Adv Water
Resour 52:34–49

123