See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/320003238

The problem of identifying arsenic anomalies in the basin of Sahand dam

through risk-based 'soft modelling'

Article  in  Science of The Total Environment · September 2017

DOI: 10.1016/j.scitotenv.2017.08.027

CITATIONS READS

4 312

4 authors:

Ata Nadiri Fariba Sadeghi Aghdam

University of Tabriz University of Tabriz
98 PUBLICATIONS   597 CITATIONS    10 PUBLICATIONS   14 CITATIONS   

SEE PROFILE SEE PROFILE

Rahman Khatibi Asghar Asghari Moghaddam

GTEV-ReX University of Tabriz
71 PUBLICATIONS   714 CITATIONS    106 PUBLICATIONS   1,076 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

A new framework for mapping subsidence vulnerability indices (SVIs): ALPRIFT View project

Investigation of arsenic concentration anomalies in groundwater resources of Chahardoly area, Kordestan View project

All content following this page was uploaded by Ata Nadiri on 26 September 2017.

The user has requested enhancement of the downloaded file.

 Science of the Total Environment 613–614 (2018) 693–706

Contents lists available at ScienceDirect

Science of the Total Environment

journal homepage: www.elsevier.com/locate/scitotenv

The problem of identifying arsenic anomalies in the basin of Sahand dam

through risk-based ‘soft modelling’
Ata Allah Nadiri a,⁎, Fariba Sadeghi Aghdam a, Rahman Khatibi b, Asghar Asghari Moghaddam a
a
Department of Earth Sciences, Faculty of Natural Sciences, University of Tabriz, 29 Bahman Boulevard, Tabriz, East Azerbaijan, Iran
b
GTEV-ReX Limited, Swindon, UK

H I G H L I G H T S G R A P H I C A L A B S T R A C T

• Arsenic levels at a 10 year old dam sys-

tem exceed its allowed limit by 28 times.
• Arsenic hotspots at Sahand Dam impact
human health and its ecology
• Studied Origins, Sources, Pathways, Re-
ceptors and Consequences (OSPRC)
• Assembled 6 levels of techniques and
gained insights into six OSPRC cells
• Some of these OSPRC cells have local but
mostly have system-wide impacts.

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

Article history: An investigation is undertaken to identify arsenic anomalies at the complex of Sahand dam, East Azerbaijan,
Received 10 May 2017 northwest Iran. The complex acts as a system, in which the impounding reservoir catalyses system components
Received in revised form 30 July 2017 related to Origin-Source-Pathways-Receptor-Consequence (OSPRC) viewed as a risk system. This ‘conceptual
Accepted 2 August 2017
framework’ overlays a ‘perceptual model’ of the physical system, in which arsenic with geogenic origins diffused
Available online xxxx
into the formations through extensive fractures swept through the region during the Miocene era. Impacts of ar-
Editor: D. Barcelo senic anomalies were local until the provision of the impounding reservoir in the last 10 years, which trans-
formed it into active system-wide risk exposures. The paper uses existing technique of: statistical, graphical,
Keywords: multivariate analysis, geological survey and isotopic study, but these often seem ad hoc and without common
Arsenic anomalies knowledgebase. Risk analysis approaches are sought to treat existing fragmentation in practices of identifying
Graphical method and mitigating arsenic anomalies. The paper contributes towards next generation best practice through:
Isotopic study (i) transferring and extending knowledge on the OSPRC framework; (ii) introducing ‘OSPRC cells’ to capture
Multivariate statistics analysis unique idiosyncrasies at each cell; and (iii) suggesting a ‘soft modelling’ procedure based on assembling
Sahand dam
knowledgebase of existing techniques with partially converging and partially diverging information levels,
Soft modelling
where knowledgebase invokes model equations with increasing resolutions. The data samples from the study
area for the period of 2002–12 supports the study and indicates the following ‘risk cells’ for the study area:
(i) local arsenic risk exposures at south of the reservoir, (ii) system-wide arsenic risks at its north; and (iii)
system-wide arsenic risk exposures within the reservoir even after dilution.
© 2017 Elsevier B.V. All rights reserved.

⁎ Corresponding author.
E-mail addresses: nadiri@tabrizu.ac.ir (A.A. Nadiri), Fariba_Sadeghi@tabrizu.ac.ir (F. Sadeghi Aghdam), gtev.rex@gmail.com (R. Khatibi), Moghaddam@tabrizu.ac.ir
(A. Asghari Moghaddam).

http://dx.doi.org/10.1016/j.scitotenv.2017.08.027
0048-9697/© 2017 Elsevier B.V. All rights reserved.
694 A.A. Nadiri et al. / Science of the Total Environment 613–614 (2018) 693–706
1. Introduction
A study is presented in this paper to the problem of identifying arse-
nic anomalies at a study area, where measured concentrations can ex-
ceed the WHO (2004) limit of (0.01 mg/L) N 200 times. The study area
comprises the complex of the Sahand dam basin, East Azerbaijan, north-
west Iran. The treatment of the problem encompasses hydrosphere,
lithosphere and anthroposphere and the paper seeks risk-based
approaches to the identification problem to overcome existing frag-
mentation on the contributing disciplines. The identification problem
is akin to hazard identification, as in risk analysis, which also covers
risk assessment, risk management and risk communication, see Pizzol
et al. (2015) and Tiedeken et al. (2017) but the paper leans towards
risk analysis more than that towards traditional identification of arsenic
anomalies.
The rationale for innovations in the paper derives from integrating
the following contributions: (i) a risk analysis approach is promoted to-
wards the problem of identifying of arsenic anomalies; (ii) the problem
is treated through the dimensions of Origins, Sources, Pathways, Recep-
tors and Consequence (OSPRC) and these dimensions together form a
framework, where a framework is consensual agreement on the dimen-
sions without any empirical-theoretical basis; (ii) the system is divided
into a number of OSPRC cells (risk cells) as per Khatibi (2008), each of
which captures independently varying aspects of arsenic anomalies of
the study area; (iii) the concept of “soft modelling” is introduced to in-
tegrate diverse range of techniques (including: statistical, graphical,
multivariate analysis, geological survey and isotopic study) used both
in practice and research for identifying arsenic anomalies; and (iv) the
rationale in an integrated fashion forms a conceptual model with a full
picture on arsenic anomalies and this builds on a perceptual model to
formalise the description of the scientific understanding of the process-
es within physical systems largely driven by the geology of the area. The
paper shows that risks in the study area were of local scope but this was
transformed into a system-wide risk exposure by the provision of the
impounding reservoir.
The practice of identifying arsenic anomalies is quite idiosyncratic
for a range of reasons including variations in OSPRC dimensions and
the diversity of techniques used to identify anomalies The authors are
not aware of any critical thinking against existing methodologies. The
paper seeks to formalise the methodology on the identification of arse-
nic anomalies through the three elements of the OSPRC framework, risk
cells and soft modelling procedure, the following review is focussed on
the main contributions of the paper.
Origins express a potential for the release of arsenic loads from geo-
logical formations, ores or chemical compounds but existence of arsenic
at a place is not necessarily exposure to risks. Origins of arsenic anomalies
are diverse and Bundschuh et al. (2011) show that arsenic anomalies are
generally found in the atmosphere, water, soil, rock, and organisms in var-
ious organic or inorganic compounds. The origins of arsenic anomalies in
water include: geogenical (Smedley and Kinniburgh, 2005; Alonso et al.
2014; Beiyuan et al., 2017), biological (Mahimairaja et al., 2005), anthro-
pogenic (Smedley and Kinniburgh, 2005; Bundschuh et al., 2011;
Muhammad et al., 2016; Martin et al., 2017), geogenical and anthropogenic
(Johnson et al., 2014; Lapworth et al., 2017; Kazakis et al., 2017). High
arsenic anomalies in groundwater are often geogenic, also called natu-
ral (Mahimairaja et al., 2005). Main natural origins of arsenic include
basin-fill deposits (J. Wang et al., 2017, S. Wang et al., 2017), geothermal
and volcanic activities (Nriagu and Pacyna, 1988; Bondu et al., 2017).
The paper is focussed on geogenic origins, which is mostly a diffuse
source and its identification from this origin is often more challenging
than other origins.
Source: This refers to the processes of arsenic ions leaching out of
geological formations through hydrogeochemical processes. Leaching
out sorbed arsenic ions at sources are associated with triggering risks,
which expresses the likelihood of adverse effects. If the loads remain
local at the source, the risk is likely to remain local and even may not
trigger risk exposures. Notably, the interchangeable use of the terms or-
igin and source in some of the published works is not uncommon. The
literature on identifying arsenic anomalies at source is vast (Chuah
et al., 2016; Bondu et al., 2017) and the various techniques are outlined
in Section 3, which include: wide applications of statistical method e.g.
Nadiri et al. (2013), Graphical method e.g. the Piper diagram (Piper,
1944), Durov diagram (Durov, 1948; Lloyd and Heathcote, 1985); and
Stiff diagram (Stiff, 1951; Hem, 1989), multivariate analysis (J. Wang
et al., 2017, S. Wang et al., 2017; Uher et al., 2017), geological and
hydrogeological studies (Hounslow, 1995), isotopic studies (Clark and
Fritz, 1997). These techniques serve as tools for both practice and re-
search to investigate hydrochemical and hydrogeochemical processes.
It is not common practice to compare these techniques but published
works seems uncritical on the absence of their inter-comparisons.
Pathways and Receptors: The focus of the paper is on either leached
out arsenic ions triggering local risk exposures at the source or system-
wide risk exposures propagating through pathways to each remote re-
ceptors. The authors are not aware of systematic research works on
pathways whereas research on receptors includes studies on the social
dimension in terms of vulnerability and resilience but the authors are
not aware of studies in risk exposure related to arsenic anomalies.
Consequence: Arsenic is carcinogenic (Pershagen, 1981) and the
consequences of exposures to arsenic in water have recently been eval-
uated (Kapaj et al., 2006) including cancers, several dermatologic and
vascular diseases, cerebrovascular disease, infant mortality, and reduc-
tion in birth weight (Tseng, 1977; Bhattacharya et al., 2007). Exposures
of this nature occur in many parts of the world with each occurrence
often seems quite idiosyncratic and recent studies have focused on arse-
nic anomalies in various parts of the world, e.g. Bangladesh (Anawar
et al. 2011), China (Ning et al. 2007), Northern Greece (Kouras et al.
2007), Pakistan (Nickson et al. 2005), India (Rahman et al. 2005),
Latin America (Bundschuh et al. 2011), Costa Rica (Hammarlund and
Pinones 2009), Guatemala (Cardoso et al. 2010), Colombia (Tassinari
et al., 2008), west Africa (Bretzler et al., 2017). There is no published
work related to the study area on any specific consequences of arsenic
anomalies other than stating here that approx. the 60,000 population
of Hashtrud is likely to be exposed to the risks; as well as an unknown
number of people downstream of the dam. Already some of at-risk vil-
lages are abandoned by migrating from the area to avoid health impacts
but these have not been documented yet.
Soft Modelling: A set of overlapping techniques, referred to above
and to be detailed in due course, have been developed over the years
since the 1940s and applied to diverse sites. Whilst the authors promote
the techniques listed above, to their knowledge, a critical view on these
techniques remains outstanding and therefore both research and prac-
tice is based on selecting a number of them for analysis. Arguably,
these techniques overlap and their solutions display both convergent
and divergent features and this is similar to soft systems, in which the
term ‘soft’ underpin convergent and divergent human behaviours, for
more information see (Checkland and Scholes, 1990; French et al.,
2005; Nidumolu et al., 2006). Thus, the paper formulates a ‘soft model-
ling’ procedure to exploit convergence/divergence behaviours among
different techniques.
Shortfalls in existing techniques include: (i) absence of any concept
on convergence/divergence of these techniques; (ii) serious problems
in knowledge integration among different techniques; and (iii) absence
of modelling procedure similar to those in in physical sciences. Argu-
ably, shortfalls retard the practice of identifying arsenic anomalies but
soft modelling procedures would fill these gaps.
2. Specification of basic information on the study area
2.1. Geographical location
The study area covers approximately 384 km2 of land, in the East
Azerbaijan province, Iran, and located between cities of Maragha and
 A.A. Nadiri et al. / Science of the Total Environment 613–614 (2018) 693–706
the west of Hashtrud (also known as Heshderi), see Fig. 1. The main fea-
ture within the study area is Sahand Dam, which impounds water in a
reservoir storing 165 Mm3 of water. This was constructed in 2001–
2005 to serve as water resource for approx. 60,000 people. Its construc-
tion led to the submergence of the 6000-year-old archaeological Kul
Tepe site (http://www.stonepages.com/news/archives/001949.html).
The construction works of the project was announced after the comple-
tion of half of its dam structure (http://www.archaeology.ws/2004-11-
29.htm).
The impounding reservoir of the dam provides potable, agricultural
and industrial water for the demand from Hashtrud and some villages
in the vicinity. The reservoir is at the confluence of the Rivers Qaranqu
and Almalu (normally known as Almali) and from here the River
Qaranqu flows to Qizil Ozen (the River Qizil) some 150 km downstream
of Sahand dam, and that flows to the Caspian Sea some 765 km down-
stream of the confluence with the River Qaranqu. The basins upstream
of the impounding reservoir have each different basin characteristics
and different water quality.
Based on Emberger (1930), the prevailing climate in this region is
semiarid-cold; the average annual precipitation is approximately
350 mm (East Azerbaijan Regional Water Authority, 2010). Highest
and lowest precipitation occurs in the spring and summer seasons, re-
spectively. The mean annual temperature is 11 °C, whereas mean
monthly temperature varies from −2 °C (in January) to 24 °C (in Au-
gust). In general, mean monthly relative humidity is fairly high and
ranges from 37.1% (in July) to 62.9% (in February).
2.2. Geology and hydrogeology – the basis for a perceptual model
The geological formations in the study area comprise: (i) the older
Miocene Upper Red Formation (URF), and (ii) the younger Miocene
Sahand volcanic pyroclastic deposits, and (iii) recent Quaternary de-
posits and formation, see Fig. 1. The Miocene succession includes the
Fig. 1. Geological map of the study area.
695
Upper Red Formation and Sahand volcanic deposits. The older URF is
composed of conglomerate, marl, siltstone, sandstone and limestone,
which have undergone extensive fracturing and have joints and fault
lines; whereas fracturing in the younger pyroclastic deposits is not ex-
tensive. It is flagged at this stage that the composition of this formation
has a potential role in arsenic anomalies of the study area, as it com-
prises gypsum, anhydrite and halite deposits, where they are located
in the south of the impounding reservoir and its northwest along the
River Almalu. The younger Miocene and Pliocene volcanic formations
are composed of Sahand pyroclastic deposits and ash flows, where the
River Qaranqu flows through. The Quaternary deposits are composed
of old and young alluvial terraces, limestone and recent alluvial
deposits.
Faulting, folding, uplifting and landslides have contributed to forma-
tion processes of the area. There are several faults in the study area
shown in Fig. 1, which occurred often during or after the Miocene
Epoch and gave rise to the connection between surface water with
groundwater but the extent of such fracturing and connections in the
pyroclastic deposits are believed to be very limited. It is flagged that
this connection during older Miocene era but its absence within the
younger Miocene deposits is a key information for formulating a per-
ceptual model of the study area to be presented later in the paper.
The study area has also arsenic-bearing ores, e.g. orpiment (As2S3),
realgar (As2S2) and arsenopyrite (FeAsS). When the conditions are rife
for active hydrogeochemical processes, sorbed arsenic loads are actively
leached out in the area and evidence for this is visible around old springs
dried up now. This process is likely within the geological formations be-
longing to the older Miocene era but not in younger pyroclastic deposits
with limited fracturing.
The hydrogeological conditions of the study area are directly related
to the lithology of geological formations and its past intense tectonic ac-
tivities. The connectivity of surface water and groundwater is under in-
vestigations with the focus on the pathways of rain and melting snow
696 A.A. Nadiri et al. / Science of the Total Environment 613–614 (2018) 693–706
and their percolation to ground through the soil media, joints and fault
lines. Attention will also be given to the interactions of the water perco-
lating or seeping through these pathways with the Upper Red Forma-
tion and younger pyroclastic deposits. Hydrogeologically, it is flagged
that aquifers in the study area are patchy, shallow and heterogeneous,
none of which are large or extensive. In spite of the lack of an extensive
aquifer in the study area, there are wells, springs, qanats, and surface
water ponds/lakes, which serve as water resources for drinking and
agriculture.
2.3. Data strategy for water sampling
A data strategy was formulated at a later stage to capture that arsenic
anomalies at the study area could be investigated in a comprehensive
manner. It was found that a combination of the available data from dif-
ferent programs could profitably be used for identifying the anomalies.
These data are specified in Table A.2, which comprise 149 samples; as
well as an additional 144 samples from the embankment reservoir, col-
lected from 2002 to 2012 (Tables A.1 and A.2). The collected samples
were analysed by the East Azerbaijan Regional Water Authority and
the hydrochemistry laboratory of Tabriz University.
Various techniques to be used in identifying arsenic anomalies make
appropriate selections from these data, which comprise:
(i) physicochemical parameters, e.g. Electrical Conductivity (EC), alka-
linity (pH); (ii) major ions (Ca2 + , Mg2 + , Na+ , K+ , CO23 − , HCO− 3 ,
SO24 − , Cl−); (iii) minor ions (N – NO− − 2− −
3 , NO4 , PO4 , F , SiO2 ,
Fe2 + , 3 + , Al3 +); (iv) trace elements (As 3 + , 5 +, Mn2 + , Zn2 + ,
Pb2+ ,4+ , Cu2+ , Cr3+ , Co2+): and (v) isotopic data determined by stan-
dard methods (American Public Health Association, 1998).
Measurement of the total concentration of major, minor and trace el-
ements carried out using standard methods (ASTM, 2002) including Ca2+
and Mg2+ by complexometric titration, Na+ and K+ by Flame photome-
− −
try, CO2−3 and HCO3 by acid base titration, Cl by precipitation titrations,
SO4 by the turbidimetric method, PO4 , N − NO−
2− 2−
3 , SiO2 , F
−
by
spectophotometer, and As3+ ,As5+, Fe2+ , Fe3+ and other heavy metals
by the Atomic absorption method.
According to Hounslow (1995), accuracy analysis should be
checked for water samples by using charge balance and TDS/EC ratio.
Also, ion balance errors should be b 5% (Freeze and Cherry, 1979),
in which the total concentration of cations and anions are expressed
in meq/L.
P P
Cations− Anions
%Ion−balance error ¼ P P  100
Cations þ Anions
An analysis of the data samples used in this study shows that the
charge balance values for all the samples are b5%, and hence the data
are acceptable. TDS/EC ratios for most of the samples range from 0.55
to 0.75, which are within acceptable ranges.
Several datasets are available, each related to different parts of the
study area. The latest complete dataset available for this study goes
back to June 2012, which comprises: 50 samples (samples from 26
springs, 6 wells, 8 samples from rivers and 10 from the impounding res-
ervoir). The full sampling details are given Fig. 2 with respect to their lo-
cation and sample number.
3. Methodology – the basis for a conceptual model
Existing methodologies for the identification of arsenic anomalies
for each study area are largely based on ad hoc selections from a large
array of techniques but this retards knowledge integration. The
contribution of the paper to the methodology is as follows:
(i) transferring knowledge on OSPRC; (ii) dividing a study area into
a number of OSPRC cells; and (iii) introducing a soft modelling pro-
cedure by putting each technique into a level of information with a
focus on their comparative performances falling into the following
possibilities:
i. A higher level technique corroborates with a lower one by
overlapping its solutions – the common area may be referred
to as convergent solutions, inclusive solution, working model
or living model;
ii. A higher level technique challenges a lower one by conflicting
it – discordant aspects may be referred to as divergent solu-
tions or exclusion features of each technique; and
iii. A technique at one level provides more insight than another
level by revealing new effects – this may be referred to as
knowledge creation.
3.1. Outline of OSPRC
Risk communicated through the SPRC framework is discussed by
Khatibi (2008), who states that the framework is believed to have
emerged in 2000, although receptor and pathway terms are in common
usage in medical science for much longer. The paper extends this frame-
work to cover Origins and hence OSPRC. Origins of arsenic anomalies are
active research areas and include: (i) Nriagu and Pacyna (1988), who
state that volcanoes are the most important natural source for arsenic
anomalies and volcanic activities are responsible for arsenic releases;
(ii) Gurung et al. (2005) argue that high concentrations of arsenic on
surface and in groundwater may also be related to individual geological
and climatic conditions; and (iii) Bundschuh et al. (2011) recommend
the consideration of a wide range of possible origins for the identifica-
tion of arsenic pollution. Generally, high concentrations of arsenic in
water resources originate from geological formations through sodium
salts to leach out arsenic and this has greater effects than other salts
(Anwar et al., 2003).
Whilst the paper reserves the term Origin to the identification of
potential occurrences, the term ‘Source’ signifies the activation of the
process for arsenic releases. There are various mechanisms for this,
but active arsenic anomalies require the presence of a range of min-
erals mentioned above. A broad overview of OSPRCs dimensions is as
follows. Origin: refers to potential occurrence of arsenic without
necessarily triggering any risk processes. Sources: refer to point-
sources (springs) or diffuse-sources over regions (aquifers) where
hydrogeochemical processes are leaching out arsenic ions and this
creates the potential for harm. Pathways: these are the routes for
transferring risks from one location to another and can be water-
ð1Þ courses, aquifers, joints, fault lines and/or manmade conduits such
as reservoirs. Receptors: these are endpoints, where the arsenic
sorbed ions are collected and come to contact with human beings
to pose risks with a likelihood to harm. Consequence: This is
outlined in the Introduction section and carcinogenicity is just one
prime consequence. Both receptor and consequence normally refer
to the social dimensions. Thus, OSPRC can differentiate risk impacts
across societies and countries.
3.2. OSPRC cells
Khatibi (2008) suggests the application of SPRC framework for inte-
grating risks by using one SPRC to represent each generic variation in
the overall risk and this is taken up in the paperas well by the authors
elsewhere, e.g. Nadiri et al. (2017a, b, and c). To this end, an OSPRC
cell (or risk cell) is a new concept, according to which each significant
generic variation is captured by one OSPRC cell. Each study area may
be divided into n number of OSPRC cells, where n is an integer value
from 1 to N. The paper suffices to six such cells, i.e. N = 6. Although a
greater number is possible, the aim at this stage is a contribution to
the methodology of identifying arsenic anomalies and therefore a
broad identification is sufficient.
 A.A. Nadiri et al. / Science of the Total Environment 613–614 (2018) 693–706 697

3.3. Soft modelling procedure
Studies of arsenic anomalies are faced with the problem that full at-
risk areas are inaccessible but this is partially compensated for by over-
lapping complementary techniques to firm up the information sug-
gested by telltale signs based on samples. Sources of arsenic anomalies
are studied by using the measured data samples. These studies by-
and-large are driven by data and arguably, they are condensed into
information and this in turn is condensed into knowledge
(knowledgebase). Knowledgebase is analogous to laws of nature or em-
pirical rules in physical sciences. Conversely, knowledgebase obtained
through data-driven techniques is soft, as they can be divergent and
care is needed for transferring knowledge from one study to another.
Modelling practices in science are formalised by modelling proce-
dures but the knowledgebase for identifying arsenic anomalies lacks a
formal procedure. The soft modelling procedure aims to fill this gap

Fig. 2. Overview of samples and the OSPRC cells: (a) major ions concentration; (b) arsenic concentration in the water samples in June 2012; and the OSPRC cells overlaying Fig. 2b.
698 A.A. Nadiri et al. / Science of the Total Environment 613–614 (2018) 693–706
Fig. 2 (continued).
through a procedure outlined below, which categorises existing tech-
niques to six levels with overlapping but increasing information con-
tents, as follows:
Level 0: A simple parsing out of the modelling data is a presumed
practice. Also in soft modelling, data samples should be used to identify
possible patterns (or lack of them) to draw telltale signs on arsenic
anomalies of the study area. The activities at this level help formulate
a preliminary model.
Level 1: The findings at Level 0 on arsenic anomalies are refined by
knowledgebase using statistical analysis of cross-correlation coefficients
among the various groups of ions to reveal dominant ions within the
study area.
Level 2: The findings at Level 1 on arsenic anomalies can be refined
by the knowledgebase of graphical methods to interpret hydrogeo-
chemical processes through the following methods: (i) Piper diagram
(Piper, 1944) to identify groundwater types (Todd and Mays, 2005);
Durov diagram (Durov, 1948) to improve on the Piper diagrams by a
better display of different types of water, as detailed by Lloyd and
Heathcote (1985); and Stiff diagram (Stiff, 1951; Hem, 1989) to
infer types of source rocks. The subsequent refinements are still subject
to several limitations as they cannot analyse some of the physical-
chemical parameters, e.g. heavy metals, SiO2 , NO−
3 , pH and
temperature.
Levels 1 and 2 invoke top-down modelling activities similar to de-
terministic models, whereby instead of laws of nature, the employed
techniques are top-down knowledgebase as the consolidation of past
research and years of professional experience. Level 0 would not show
any pattern but the techniques at Levels 1 and 2 identify hydrogeo-
chemical processes towards the identification of arsenic anomalies.
Level 3: The findings at Levels 1–2 on arsenic anomalies are refined
by knowledgebase built on multivariate analysis methods capable of
treating limitations of graphical methods, see among others Singh
et al. (2004) and Khan (2011). Such analyses include Hierarchical Clus-
tering Analysis (HCA) and Factor Analysis (FA) able to explain the corre-
lation among a large number of variables and reduce the number of
variables into a smaller number of factors without a loss of essential in-
formation (Cloutier et al., 2008). Refining the understanding provides a
better focus on sources, pathways and receptors. The techniques at
Level 3 invoke bottom-up modelling activities similar to data-driven
statistical models, but the information gain is site-specific and not trans-
ferrable to other sites.
Hierarchical Clustering Analysis (HCA) is a classification technique
for investigating hydrogeochemical types and the interpretation of
their origin (Cloutier et al., 2008), which uses (i) the Euclidean distance
(or similarity measurement) between samples (Deza and Deza, 2009),
and (ii) Ward's method as a linkage rule (Ward, 1963). The data
grouped together by HCA use a threshold value for the linkage distance
but expert opinion is used to select its value (Nadiri et al., 2013).
FA identifies factors influencing effectively the hydrogeochemistry
of study areas and the significance of the correlation between factors
and variables of sample data. The principle aim in FA is to reduce com-
plex patterns of correlations among many parameters to simpler sets
of ‘factors’ to make them amenable to interpretations. The analysis re-
quires the standardisation of the correlation coefficient matrixes, their
eigenvalues and eigenvectors. The data can then be transformed into
factors, which are extracted in terms of the minimum eigenvalue
being N 1 (Kaiser, 1958). Next the factor loading matrix is rotated ac-
cording to the varimax rotation technique (Davis, 1986) to distinguish
variables with high factor loadings (close to 1 or − 1) from variables
with low factor loadings (close to 0). This procedure renders a new fac-
tor loading matrix, which provides a greater ease of interpretation. Fi-
nally, the contribution of each factor at every site leads to computing
the values of the factor scores to map geographical distributions of
each factor (Nadiri et al., 2013).
 A.A. Nadiri et al. / Science of the Total Environment 613–614 (2018) 693–706
Level 4: The inclusive features of Levels 1–3 on arsenic anomalies are
refined by a further knowledgebase using geological study to identify
source, pathways and receptor in three stages: (i) field study: to corrob-
orate the published geological formations based on material and geo-
logical time and refine them, (ii) further sampling: select some
formations for a further sampling to focus more on local information,
(iii) prepare samples to microscopic analysis. These activities at Level
4 are tantamount to ground truthing techniques and invoke model cal-
ibration processes.
Level 5: Isotopic studies are used to detect sources and age of
groundwater through seeking consistency between the shared informa-
tion at Levels 1–4 and inner properties of water in terms of the
knowledgebase for their isotopes at the study area, the regional and
global isotopic measurements. This activity invokes model validation
in general modelling procedure.
Science base is used by the paper as discussed by Friedman (1953),
Craig (1961a) and Criss (1999) to study isotopes to elicit the origin,
phase transitions, and transport of water by graphically displaying
“δD” against “δ” 18O for the isotopic compositions of precipitation
along the Global Meteoric Water Line (GMWL); evaporation line, ther-
mal line, and local groundwater line.
Level 6 invokes model prediction activities as normally practiced in
modelling. A comparison of time variations of arsenic anomalies and
flood flows at specific observation stations exploits the knowledgebase
that during a flooding event arsenic concentration ought to reduce due
to the dilution and washing out effects. This exploitation gives a sense of
model prediction.
4. Results of analysis
Arsenic anomalies at the study area were detected by Nadiri et al.
(2012). As outlined by Criss (1999), the following bodies of water play
role in activating and maintaining arsenic anomalies: ocean, hydrother-
mal fluids, magmatic waters, metamorphic waters, juvenile (volcanic)
waters and meteoric (groundwater and surface water). Although direct
anthropogenic origins at the study area are likely to be negligible, the
provision of the dam is an anthropogenic activity, which has trans-
formed local scale impacts into system-wide impacts. Such a diverse
range of origins triggers hydrogeochemical processes and the onset of
arsenic anomalies with exposures to health risks. Although the study
area is at a volcanic region, it has not been particularly active in recent
geological times.
4.1. Parsing out the data samples – Level 0 techniques
Available samples of data specified in Tables A.1–A.2 are parsed out
for existence of ion exchange processes, their reversal, the identification
of dominant ions and the groundwater types. An overview of the values
of anions and cations and Electrical Conductivity (EC) corroborated the
past studies reported by Nadiri et al. (2012). Evidence of this nature
identifies: (i) the dominance of bicarbonate and chlorine anions in
most samples from the southern part of the impounding reservoir and
from Eynabad Wadi (north of the reservoir). (ii) Nitrates originate
from anthropogenic traditional agricultural activities.
Arsenic concentrations at the study area vary from b 0.001 mg/L (10
times lower than the MPL value) to 2.009 mg/L (200 times higher than
the MPL value) with a median value of 0.0215 mg/L, which is twice the
MPL value (Table A.1). As shown in Fig. 2b and c, high arsenic concentra-
tions occur at the following cells: (i) Cell 1: along the springs at loca-
tions to the south of the impounding reservoirs; (ii) Cell 2: the
groundwater at the northwest of the impounding reservoir along the
River Almalu; (iii) Cell 3: at the north of the impounding reservoir but
in the vicinity of the Qopuz fault line and a landslide in the vicinity;
(iv) Cell 4: along the Shor Dere fault line associated with the Upper
Red Formation and other locations; (v) Cell 5: within the embankment
reservoir; and (vi) Cell 6: along the watercourse downstream of Sahand
699
dam. Thus, the data confirms existence of arsenic anomalies in the study
area and it may be flagged that the overall information invokes the pos-
sibility of dividing the study area into the above six OSPRC cells to study
arsenic anomalies through Levels 1–6 techniques.
4.2. Using statistical analysis to identify telltale signs – Level 1 rule of thumb
methods
The data are analysed for a more quantitative assessment of exis-
tence of hydrogeochemical processes, or its reversal and on sources of
major ions and groundwater types.
Basic cross-correlation analysis: The correlation matrix of hydro-
geochemical parameters (Table A. 3) provides evidence as per
Hounslow (1995) that: (i) high positive correlation between chloride
and sodium shows the main source of chloride to be halite; (ii) high
positive correlation between bicarbonate and sulphate with calcium
and magnesium shows groundwater sources may be from calcite, dolo-
mite or gypsum; (iii) high positive correlation is observed between ar-
senic and trace elements of magnesium, zinc and aluminium; as well
as high negative correlation between arsenic and iron, cobalt and chro-
mium; (iv) the same effective process and/or source is indicated with
the following ions: high negative correlation between bicarbonate and
pH confirms that bicarbonate ions will be removed by increasing pH
(N8.3) through the hydrogeochemical process but the carbonate ions
will increase, (v) the weak cross-correlation between arsenic and bicar-
bonates rules out the role of bicarbonates (−0.01)in the release of arse-
nic anomalies in the study area. Also, high positive cross-correlation
coefficients of arsenic values with the fluoride (0.39), nitrate (0.42), sul-
phate (0.57), and silica (0.22), EC (0.34) parameters confirm solution
processes (Section 4.5), in which the roles of nitrate and sulphate are
corroborated but those of fluoride and silica are under an ongoing
investigation.
Complex cross-correlation analysis using combined ions: Based
on McLean and Jankoswski (2000), a visual display of the collective
−
ions of (Ca2+ + Mg2+) against (SO2− 4 + HCO3 ) is shown in Fig. 3a for
the water samples, which identified calcium in the water exchanged
for sodium as a common ion exchange process. This confirms the disso-
lution of gypsum/anhydrite and calcite/dolomite processes as also con-
firmed by the Hounslow ionic ratios, Hounslow (1995). For all the water
samples, the HCO− 3 /SiO2 ratio is N10, and the Mg
2+
/(Mg2 + + Ca2 +)
equivalent ratio is lower than 0.5 which indicates carbonate solution.
Based on the Hounslow (1995) ionic ratios, approximately 58% of
the water samples have Na+/(Na+ + Cl−) ratio within the range 0.5 ±
0.1 or plot along or close to Line 1:1 in Log Na+ against Log Cl− which
indicates that either saline water or halite dissolution is partially re-
sponsible for Na+ in the groundwater mediated by halite within the
Upper Red Formation in the study area. However, approximately 42%
of the water samples have Na+/(Na+ + Cl−) ratios significantly N 0.5,
which indicates another sodium source. This can be either due to ion ex-
change or the weathering of alumino-silicate (Na-plagioclase) from the
Pyroclastic Formation in the study area. Datapoints in red on Fig. 3b are
sample points from locations at Eynabad Wadi, which have high salinity
values. Arsenic anomalies at the sample points related to the Miocene
Formations are observed along this tributary, after joining the main trib-
utary of the River Almalu.
As per Jankoswski et al. (1998), the study of the ion groupings of
(Ca2 + + Mg2 +) − (HCO− 2−
3 + SO4 ) against the halite ion group of
(Na+ − Cl−) indicates the dominance of ion exchange (Ca/Na) for the
samples of the study area.
4.3. Graphical analysis – Level 2
The data are analysed by a higher knowledgebase using various
graphical techniques but only that of the Piper diagram is elaborated.
Bodies of groundwater can be discriminated into different types as out-
comes of different hydrogeochemical processes using the Piper
700 A.A. Nadiri et al. / Science of the Total Environment 613–614 (2018) 693–706
Fig. 3. Studies on ion compositions: (a) Ion state of groundwater: Log (Ca2+ + Mg2+)
versus Log (HCO− 2− + −
3 +SO4 ), (b) Na versus Cl for representative water samples.
diagram, which is a diamond plot divided into five zones (A, B, C, D and
E). The Piper diagram for the study area is shown in Fig. 4a and these are
interpreted as follows. Zone A: the majority of the sample points are
clustered at this zone towards the dominant (Ca2 + + Mg2 +) and
(HCO− 2−
3 + CO3 ) with a good water quality recharged from carbonated
formations comparable to the quality of surface waters. Zone C: few
of the samples (16%) are clustered towards (Na+ + K+) and
−
(SO2−4 + Cl ) zone, likely originated from halite and gypsum dissolu-
tions initiated within the older Miocene Formation at the sources main-
ly in the Eynabad Wadi springs. Zone E: the relevant samples show the
characteristics of mixed water types with neither dominant cation nor
anion and these are identified in the samples from the impounding res-
ervoir but the identification of this type is problematic. These results
corroborate the different OSPRC paths, see also Fig. 2c.
Similar Hydrogeochemical process analyses carried out by the ex-
panded Durov diagram, which confirm Piper diagram results and indi-
cate that the dominant processes in the study area are: (i) the
Chebotarev sequence responsible for dissolution reaction, and (ii) ion
exchange processes responsible for Ca/Na exchange between water
and soil, the correlation coefficient between Ca and Na (0.57) supported
this processes (Table A.3). These results corroborate the different OSPRC
pathways, see also Fig. 2c.
Different sources of groundwater are identified by the Stiff diagram,
as it provides a better display of different rocks which impact on water,
as displayed in Fig. 4b based on Hounslow (1995), as follows: Class C1:
samples in this class indicate saline water; Class C2: samples in this class
identify a dominant groundwater feeding from limestone and
dolostone; Class C3: samples in this field identify groundwater feeding
gypsiferous formations; Class C4 and C5 samples in this field identify a
mixing process affecting groundwater; Class C6: samples in this field
identify groundwater to be influenced by ion exchange and this may
be feeding from shale. All of these waters from different rocks shows
high arsenic concentration in the study area.
All of these diagrams help refining the convergence of the results
that water samples in the vicinity of the River Qaranqu have been influ-
enced by bicarbonate and dolomite formations and also the samples
from the impounding reservoir (OSPRC Cell 5) indicate mixing process-
es. Water samples in the vicinity of the River Almalu at Eynabad wadi
(OSPRC Cell 2) and Southeast part of study area (OSPRC Cell 1) are influ-
enced by the halite and gypsiferous Miocene formations, which are re-
lated to high amount of arsenic values, see also Fig. 2.
4.4. Multivariate statistical analysis – Level 3
Hierarchical Cluster Analysis (HCA) is carried out for the 50 ground-
water samples obtained at the same time (January 2012), where these
samples are representative of the impounding reservoir, groundwater
and surface water. The results are shown in the dendrogram in Fig. 4c.
The method identifies 5 clusters, denoted from HC1 to HC5, in which
HC1, HC2 and HC3 have similar linkage distance and thus have greater
hydrogeochemical similarities. Samples with linkage distances shorter
than 4 are grouped within one cluster.
HC1: The HC1 samples relate to the impounding reservoir (OSPRC
Cell 5) with the designation of Quaternary Formation, where mean elec-
trical conductivity is approx. 2800 μs/cm and arsenic value is higher
than 0.01 mg/L. Most of the samples in this class are located in Zone E
in the Piper diagram with the exception of samples, 26 and 35 located
in Zone C in the Piper diagram.
HC2: The HC2 samples relate to mainly Eynabad Wadi (OSPRC Cell
2) with the designation of Upper Red Formations. It corresponds to
Zone C of the Piper diagram with a high value of electrical conductivity
and arsenic value higher than of 0.01 mg/L.
HC3: The HC3 samples relate to a different designation, which is re-
ferred to as the Upper Red Miocene Formation and limestone (OSPRC
Cel 3 and 6). It Piper's zones vary from sample to sample, e.g.: Zones A
(Samples: 8, 15 and 39), B (Sample: 2), C (Sample: 19), E (Sample:
37) with a high value of arsenic in the range of 0.02–2 mg/L and electri-
cal conductivity in the range of 840–3800 μs/cm.
HC4 and HC5: Hydrogeochemical similarity between the geochemis-
try of HC4 and HC5 samples are strong. The samples of HC4 and HC5 have
a relatively low value of electrical conductivity and arsenic where HC4
corresponds to Zone A. These match to OSPRC Cells 1 and 4.
Factor Analysis (FA) was carried out to interpret hydrochemical pro-
cesses in the study area, which identified three dominant factors. These
together explain 74% of the variance in the data. FA results are presented
in Table 1, see also Fig. 2c.
Factor 1: This shows water-rock interactions and explains 45% of the
total variance. It shows positively high factor loadings
for Ca2+ , Mg2+ , Na+ , K+ , HCO− − 2−
3 , Cl , SO4 , EC and pH. Most of these
variables are agents of dissolution reaction in the presence of water am-
plified by water-rock interactions leading to the salinity of water
through natural hydrogeochemical processes. As per spatial distribution
of the factor scores, this factor influences: (i) the south of the study area
draining dissolution from saline formations with halite and gypsum
(OSPRC Cell 1); (ii) Eynabad wadi at the River Almalu (OSPRC Cell 2);
(iii) the north of impounding reservoir (OSPRC Cell 3); and (iv) down-
stream of the impounding reservoir (OSPRC Cell 6).
Factor 2: This detects geogenic and anthropogenic origins and ac-
counts for 16% of the total variance as characterised by high positive
loading of N − NO− 3 ,F
−
and As. This factor is challenging for hinting
multiple origins, geogenic and anthropogenic, where arsenic can be
from both origins (its anthropogenic origin is associated with fertilisers,
which can act as source for both N − NO− 3 and arsenic anomalies). So the
 A.A. Nadiri et al. / Science of the Total Environment 613–614 (2018) 693–706 701

Fig. 4. Identification of types and sources of groundwater using water samples of the study area, (a) Piper diagrams (b) Stiff diagrams, (c) Five clusters displayed in the dendrogram.

challenge is to identify origins of the ions making up Factor 2. A pre-
requisite for this problem-solving is local knowledge that upstream of
the impounding reservoir: (i) is an sparsely populated area; (ii) indus-
try is literally non-existent; and (iii) agriculture is normally traditional
with limited usage of fertilisers. These are reflected in the concentra-
tions of these ions given in Table A.1, according to which N − NO− 3 is
often lower than its WHO standard (hence it is anthropogenic) but arse-
nic concentrations are much higher than the WHO standard (hence it
can only be geogenic). The geogenic origin of arsenic anomalies is ex-
plained by patchiness of the aquifers in the study area. Notably, both an-
thropogenic N − NO− 3 and genogenic arsenic anomaly are triggered by
common surface water but undergo different processes. As per spatial
distribution of the factor scores, it influences mainly samples in (i) the
southern part of the study area bearing saline water (OSPRC Cell 1);
(ii) Eynabad wadi (OSPRC Cell 2); and (iii) the springs at Qopuz
(OSPRC Cell 3). Factor 2 is triggered by surface water, which contains
significant amount of nitrate prior to penetration and then maintains
the processes on arsenic releases within Miocene Formation through
fractures.
Factor 3: This corresponds partially to OSPRC Cell 1 and 4 and shows
impacts of pyroclastic formation (tuff) at the study area. It accounts for
13% of the total variance and is characterised by the highly positive load-
ing of: CO23 −, SiO2, pH and Fe, where its influence increases from the
west to the east of the study area as per spatial distribution of the factor
scores. This factor is associated with pyroclastic formations (tuff), which
results in rich iron and silica contents and this is the case in the study
area around southwest and northeast of the impounding reservoir. It
creates a trend increase in the concentration of iron, silica, and carbon-
ate at the areas of recharge (low concentration) towards those of dis-
charge (high concentration).
702 A.A. Nadiri et al. / Science of the Total Environment 613–614 (2018) 693–706

Table 1 Leaching of arsenic from geological formations to groundwater can

Results of factor analysis for the study area. explain various processes discussed above and are corroborated by the
Factors field evidence as follows:
1 2 3
(i) As noted earlier in the paper, high levels of arsenic occasionally
have been reported in areas with high salinity. During the field studies,
Ca2+ 0.756⁎ 0.335 −0.204
arsenic sulphide was observed in limestone fractures, marl, sandstone
Mg2+ 0.682⁎ 0.478 −0.042
Na+ 0.953⁎ −0.115 0.002 and Upper Red conglomerate. Fig. 5 presents some evidence from
K+ 0.918⁎ 0.168 −0.080 these field surveys, in which realgar and orpiment minerals are
CO2−3 −0.206 −0.099 0.768⁎ scattered and interspersed within the pores of polymictic sandstone
HCO− 3 0.916⁎ 0.003 −0.232 and limestone (containing respectively 70% and 61% of arsenic) in
Cl− 0.950⁎ −0.130 0.005
SO2− 0.786⁎ 0.519 0.102
their joints. Sulphide minerals such as arsenopyrite oxidise in contact
4
N − NO− 3 0.020 0.797⁎ −0.023 with the atmosphere release soluble arsenate, sulphate and ferrous
F −
0.267 0.682⁎ −0.363 iron (Mahimairaja et al., 2005; Chen et al., 2013) as follows:
SiO2 −0.202 0.274 −0.611⁎
As3+ ,5+ −0.051 0.706⁎ −0.041 Orpiment : 32H2 O þ 24As2 S3 ðsÞ þ 80Hþ ðaqÞ
Fe2 + ,3+ −0.040 0.555⁎
0.061
þ 80NO3 − ðaqÞ→48H3 AsO4 ðaqÞ þ 9S8 ðsÞ þ 80NOðgÞ
EC 0.969⁎ −0.053 −0.006
pH −0.651⁎ −0.138 0.579⁎
Realgar : As2 S2 þ 14H2 O→3H3 AsO3 þ 2HSO− þ
4 þ 20H þ 18e
−
⁎ Signifies loadings N 0.5.

4.5. Geological study: Level 4
Level 4 activities invoke ground trothing, similar to calibration, using
petrography study and geological field surveys. The geological study
used heavy metals in the study area to provide further clues for the
identification of the location of the origins for arsenic anomalies. Studies
have shown that depositions of arsenic-contaminated soils are associat-
ed with dried springs to infer that arsenic-bearing materials in their vi-
cinity are leftovers of similar past activities. The evidence for the
association of arsenic and heavy metals are found mainly in: (i) Shor
Dere and Zulbin (OSPRC Cell 1); (ii) Eynabad wadi (OSPRC Cell 2);
(iii) the Qopuz valley (OSPRC Cell 3); (iv) the proximity of the
impounding reservoirs (OSPRC Cell 5 and 6), see also Nadiri et al.,
(2012) for some coverage. Based on Fig. 1, the older Upper Red Forma-
tions in the study area are diffused by-and-large with arsenic minerals,
where their arsenic origin is attributed to hydrogeochemical processes
activated by hydrothermal flows within older fractures during the
past geological times; whereas younger pyroclastic formations were
often not affected by these fracturing processes. As such, their content
is often low in arsenic and sulphides. Geochemical analysis of pyroclas-
tic deposits supports these results as reported by Nadiri et al. (2012).
Based on the Anwar et al. (2003), high concentration of arsenic in
water resources originates generally from geological formations. Since
the effect of sodium salts in leaching arsenic is greater than other salts,
high levels of arsenic have been reported in areas with high salinity
and correlation coefficient of 0.34 (Table A.3), which is not the domi-
nant process in the study area. These are observed at the south of the
impounding reservoir (OSPRC Cell 1) and downstream of Eynabad
wadi and Almalu River (OSPRC Cell 2) related to the Upper Red
Formation.
þ
Arsenopyrite : FeAsS þ 13Fe3þ þ 8H2 O→14Fe2þ þ SO2−
4 þ 13H
þ H3 AsO4 ðaqÞ
(ii) The outcomes of geochemical, microscopic and field studies con-
ducted in the study area support the controlling role of fractures in arse-
nic concentration which is related to magmatism, alteration, tectonic
activities and porosity.
(iii) Epithermal processes feed concentrated arsenic to fractures in
the Miocene Formation and as such it is a major source of arsenic anom-
alies. Deposition of arsenic gives rise to increased EC and pH variations
due to epithermal fluids but their dissolution releases arsenic during
the weathering of rocks and the subsequent flow of water passing
through the tectonic fractures facilitates the spread of arsenic within
the reach of the flowing water.
The geological study considered the role of anthropogenic activities
on arsenic anomalies in groundwater samples. Land use in the study
area is by-and-large rural with no significant industrial activities and
the role of subsequent anthropogenic activities on water-soil natural
processes is limited to low level usage of chemical fertilisers and pesti-
cides in recent years, but this is negligible and hence the focus is on
geogenic origins and not on anthropogenic pollutions.
4.6. Isotopic study: Level 5
The Level 5 knowledgebase on water isotopes is presented in
Section 3 and both global and sample data are shown in Fig. 6. It displays
δD against δ18O and serves as the basis for identifying the origins and
types of the measured arsenic concentrations in groundwater. The fig-
ure shows fives lines, as follows: (i) the line representing the measure-
ments from groundwater samples; (ii) Local Meteoric Line, in which
δ18O ranges from −6.99‰ to −12.52‰ and δ2H from −80.24‰ to −
42.14‰; (iii) Evaporative Trend Line; (iv) Thermal Water Line and (v)

Fig. 5. Realgar and Orpiment images taken from the site interspersed in the pores of URF, (a) polymictic sandstone (thin section) and, (b) limestone (polished section.
 A.A. Nadiri et al. / Science of the Total Environment 613–614 (2018) 693–706
Fig. 6. Stable isotopic composition of various water samples in the study area (Data from
the East Azerbaijan Regional Water Authority). Note 1: Water sample line (δ2H = 7.88
δ18O + 3. r2 = 0.86) and Note 2: Global Meteoric Water Line (GMWL), Evaporative
Trend Line and Thermal Water Line are as per Craig (1961a, b).
Global Meteoric Water Line (GMWL). The data for (i)–(ii) were mea-
sured by the East Azerbaijan Regional Water Authority and the data
for (iii)–(v) as per Craig (1961a, b).
Attention is drawn to Fig. 6, which displays the δD–δ18O space. It
provides the basis for comparing local conditions with global conditions
and learns from the underlying science. The focus is on GMWL and the
local line with respect to sample data. The figure may be divided into
three sub-spaces: (i) the distributions of data samples are close to
GMWL and this indicates a meteoric origin; (ii) data samples deviate
Fig. 7. Mapping mutual behaviour of hydrological variations and arsenic concentrations using the June 2006 data samples: (a) Arsenic concentrations during a flooding event,
(b) comparing arsenic concentrations at the impounding reservoir with those at Rivers Almalu/Qaranqu.
703
from GMWL towards lower δD and this would be indicative of either
water-rock interactions under high temperature during geological
time scales, or groundwater would have been within shallow aquifer
systems under low temperatures; and (iii) data samples deviate from
GMWL towards higher δD, often associated with hydration of silicate
minerals, as well as with mineral alterations, for more detail see Clark
and Fritz (1997).
Building on the above science, evidence is sought for the behaviour
of data points around the local line, as follows. Whilst F2 confirms in
Section 4.4 that surface water drives arsenic anomalies of geogenic ori-
gins, the points close to the local line in Fig. 6 captures an identical infor-
mation that the bulk of groundwater is of meteoric origins. As the
groundwater at the basin is predominantly shallow and the regional
groundwater temperature is low, this local condition is reflected in
data samples by deviating from appropriate points towards lower δD
of the local data. Similarly, as some of the formations are known to
bear silicate minerals, this local condition is reflected in data samples
by deviating from appropriate points towards higher δD of the local
data.
4.7. Temporal study: Level 6
The analysis for arsenic identification used a further knowledgebase
related to the temporal variability of the minerals in comparison with
the flow hydrographs. Fig. 7 presents the variations of arsenic concen-
trations over flood events with discharge measurements provided by
the East Azerbaijan Regional Water Authority. The behaviour is studied
at four stations: (i) the confluence of Rivers Almalu and Qaranqu (Sta-
tion 46 – Fig. 2a); (ii) just before the entry to the impounding reservoirs
(Station 48 – Fig. 2a); and (iii) within the impounding reservoir and at
704 A.A. Nadiri et al. / Science of the Total Environment 613–614 (2018) 693–706
the dam outlet (Station 1 – Fig. 2a). The results in Fig. 7b show that ar-
senic load is not affected by flood peaks but by baseflow and this finding
can certainly be inferred from the findings of Levels 1–5.
Under high precipitation seasons, river flows are largely fresh water,
but during low precipitation seasons baseflows from springs and aqui-
fers contribute to river flows. The figure also suggests that recharges
to shallow aquifers control the concentrations of arsenic in groundwa-
ter. The implication of the control is that by increasing recharge to the
aquifer, arsenic concentration is reduced and therefore higher arsenic
concentrations are only expected in water with older age. The study
provides evidence that surface waters are linked to shallow groundwa-
ter systems.
5. Discussion
The contributions of the paper reflect some of the necessary steps
towards an interconnected discipline, in which the problem of identify-
ing arsenic anomalies becomes a risk analysis problem. These steps
comprise: (i) the OSPRC framework; (ii) OSPRC cells; and (iii) soft
modelling for the identification of arsenic anomalies.
The perceptual model of the physical system developed for the study
area is as follows: (i) as the study area underwent the process of exten-
sive tectonic activities during and after the older Miocene times, frac-
tures, joints and faults were developing, whereby hydrothermal
minerals diffused volcanic arsenic loads to the various formations in
the area in the form of arsenic sulphide minerals – hence these forma-
tions have become origins for a potential to risk exposures; (ii) at the
presence of water, arsenic minerals within the geological formations
may undergo hydrogeochemical processes to leach out arsenic ions –
hence, risk exposures are likely (note the emphasis in the terms poten-
tial and likely); (iii) arsenic risk existed in the study area prior to the
provision of the impounding reservoir but at a local scale; whereas
since its provision, the risk has been transformed into a system-wide
scale; (iv) pyroclastic formations/deposits are products of younger Mio-
cene era and have not yet undergone fracturing and hence their role in
risk exposures is negligible.
The conceptual model of the study area is communicated using
OSPRC cells, the salient features of which include: (i) spring waters in
the area are cold, fresh and can be very salty depending on the forma-
tion at its origin; (ii) springs are generally located along fractures; (iii)
the process of mixing at the impounding reservoir is complex, as there
are complex thermocline patterns but these are yet to be studied. The
evidence base for the conceptual model through Level 0–Level 6 tech-
niques and the outcome is summarised in Fig. 2c in terms of risk cells
1 to 6 and these are summarised as follows. Cell 1 integrates: Piper
(Zone A), Stiff (C3), HCA (HC4–HC5), FA (1&3); Cell 2 integrates: Piper
(Zone C), Stiff (C1), HCA (HC2), FA (1&2); Cell 3 integrates: Piper
(Zone A and C), Stiff (C1), HCA (HC3), FA (1&2); Cell 4 integrates:
Piper (Zone A and C), Stiff (C1 and C2), HCA (HC5), FA (Factor 3); Cell
5 integrates: Piper (Zones E), Stiff (C4, C5, and C6), HCA (HC1) and Cell
6 integrates: Piper (Zones A & E), Stiff (C4, C2, and C6), HCA (HC1).
The study is conducted in Phases, in which Phase 1 (Nadiri et al.,
2012) and Phase 2 (this paper) are largely focussed on developing a
risk-based best practice procedure to include on the identification of ar-
senic anomalies. Phase 3 will study the protection of the health of the
residents by seeking remediation measures to ensure safe water in the
Hashtrud area. Phase 4 is planned to gather evidence to develop a
geospatial distribution of sources, pathways and receptors. Following
this study, it was realised that the role of the fissures, joints and fault
can be a controlling factor at the origins and sources of arsenic anoma-
lies in the study area. Thus, a further study is envisaged to directly take
samples at these locations and identify their roles.
A broad knowledge of the arsenic anomalies of the study area is
known to the relevant authorities and recently they have cut off the
pathways from OSPRC Cell 3 at the village of Qopuz towards the
impounding reservoir. Also signs have been displayed warning the
public against using the water from the local springs. Further responses
by the appropriate authorities remain to be seen. Phase 3 of the present
study is ongoing to identify mitigation measures, which should comple-
ment the measures to be developed by the relevant authorities, if any.
There are international procedures to mitigate the problems and some
of the measures are outlined by Alaerts et al. (2001).
The perceptual model based on geogenic origins is corroborated by
its convergence with the conceptual model, as follows. (i) local informa-
tion is used to rule out the introduction of arsenic to the study area
through anthropogenic or biological origins; (ii) extensive detection of
orpiment (As2S3), realgar (As2S2) and arsenopyrite (FeAsS) have been
document in fractured and soils in the study area, which would only in-
voke economical or scientific curiosity; (iii) fractures play an important
role in two ways, either they allow hydrothermal flows rich in arsenic to
precipitate and form minerals containing arsenic or isotopic evidence
suggest that meteoric water infiltrating from surface water activates hy-
drogeochemical processes of arsenic anomalies. If aquifers were deep
and long, the area would have been a hotspot for adverse health effects
but this role is now played by the provision of the impounding reservoir.
6. Conclusions
The problem of identifying arsenic anomalies in the complex of
Sahand dam is investigated, where the complex includes: the dam, its
impounding reservoir, tributaries (Rivers Qaranqu and Almalu), a
whole set of springs, baseflows, geological fissures, joints and fault
lines and formations undergoing arsenic leaching. The paper adopts
the following innovatory approaches: (i) it seeks a risk analysis ap-
proach to explain arsenic anomalies at the study area, in which the pre-
sentation in the paper is akin to hazard identification; (ii) it uses the
OSPRC (Origins, Sources, Pathways, Receptors, Consequences) frame-
work to integrate information, where the dimension of Origins is
added by the paper to cover the potential for the occurrence of arsenic
and this differs from the Source dimension referring to the actuation
of processes for releasing arsenic ions through hydrogeochemical pro-
cesses; (iii) the study area is divided into 6 risk cells, where risks cells
integrate local knowledge with each of them capturing unique features;
(iv) it uses a set of the techniques serving as tools to practitioners and
researchers but the paper organises their knowledgebase in levels
through “soft modelling,” to integrate emerging local knowledge on
convergent/divergent feature among these techniques; and (v) a com-
plete picture emerges by streamlining and integrating perceptual and
conceptual models. These are summarised as:
1. As per graphical methods (Piper, Stiff, and Durov diagrams) results in
Level 2, arsenic anomalies detected existence of different types of
waters (Ca2 + + Mg2 +) and (HCO− 2− + +
3 + CO3 ), (Na + K ) and
2− −
(SO4 + Cl ), and mixed water affected by limestone, gypsium, ha-
lite, gypsiferous formations, mixing process affecting groundwater
with Chebotarev sequence and ion exchange (Ca/Na) processes.
These results show that the water and formation types are not con-
trolling arsenic anomalies at origins and sources.
2. The controlling role of fractures come to light when the locations of
fractures, joints and faults overlap with the locations of detected ar-
senic anomalies (Level 4). These are supported by correlation coeffi-
cients between sulphate and arsenic (0.57) and this strengthens the
perceptual model on the geogenic origin of arsenic loads diffused
into the three fractured formations of older Miocene by hydrother-
mal flows; as such, springs are now the main sources of arsenic
anomalies but patchy aquifers have some role.
3. This research provides evidence that the process for arsenic anoma-
lies are triggered by water originated in surface water through isoto-
pic studies (Level 5) and the FA method (Level 3) and this
information is the key to understand pathways.
4. Temporal variations of arsenic concentrations in Level 6 show that
arsenic concentration of the impounding reservoir is affected by
 A.A. Nadiri et al. / Science of the Total Environment 613–614 (2018) 693–706
dilution with the highest risk being instigated during low rainfall sea-
sons to receptors of the water supply system used for drinking, in-
dustry and agriculture.
5. The above emerging knowledge is better communicated through
OSPRC cells and the paper identified six such cells to capture the con-
vergence and divergence between the findings of each technique.
Acknowledgments
The authors acknowledge gratefully the advice by Mr. Oroji as direc-
tor and expert of hydrogeological laboratory at the University of Tabriz
for his help in hydrochemical analysis. Additionally, gratitude is
expressed to Mr. Armanfar, of the East Azerbaijan Water Authority, for
providing some of the data.
Appendix A. Supplementary data
Supplementary data to this article can be found online at http://dx.
doi.org/10.1016/j.scitotenv.2017.08.027.
References
http://www.archaeology.ws/2004-11-29.htm archived news item Archaeology News and
Research (http://www.archaeology.ws/index.html) dated: 29/11/20104 (accessed
on 05/04/2017).
http://www.stonepages.com/news/archives/001949.html archived new item by Archaeo
News (http://www.stonepages.com/) on 8 July 2004 (accessed on 05/04/2017).
Alaerts, G.J., Khouri, N., Kabir, B., 2001. Strategies to Mitigate Arsenic Contamination of
Water Supply. Chapter 8. The World Bank, Washington, DC, USA http://
www.who.int/water_sanitation_health/dwq/arsenicun8.pdf.
Alonso, D.L., Latorre, S., Castillo, E., Brandao, P.F.B., 2014. Environmental occurrence of ar-
senic in Colombia: a review. Environ. Pollut. 186:272–281. http://dx.doi.org/10.1016/
j.envpol.2013.12.009.
American Public Health Association, 1998. Standard Method for the Examination of Water
and Wastewater. 17th edition. (Washington, DC).
Anawar, H.M., Akai, J., Mihaljevic, M., Sikder, A.M., Ahmed, G., Tareq, S.M., Rahman, M.M.,
2011. Arsenic contamination in groundwater of Bangladesh: perspectives on geo-
chemical, microbial and anthropogenic issues. Water 3, 1050–1076.
Anwar, F., Bhanger, M.I., Kazi, T.G., 2003. Relationships of Rancimat and AOM values at
varying temperatures for several oils and fats. J. Am. Oil Chem. Soc. 80 (2):
151–155. http://dx.doi.org/10.1007/s11746-003-0668-2.
ASTM, 2002. Annual Book of ASTM Standards. Water and environmental technology. 11:
p. 01. http://www.worldcat.org/title/annual-book-of-astm-standards/oclc/52526499.
Beiyuan, J., Li, J., Tsang, D.C.W., Wang, L., Poon, C., Li, X., Fendorf, S., 2017. Fate of arsenic
before and after chemical-enhanced washing of an arsenic-containing soil in
Hong Kong. Sci. Total Environ. 599–600:679–688. http://dx.doi.org/10.1016/
j.scitotenv.2017.04.208.
Bhattacharya, P., Welch, A.H., Stollenwerk, K.G., McLaughlin, M.J., Bundschuh, J.,
Panaullah, G., 2007. Arsenic in the environment: Biology and Chemistry. Sci. Total En-
viron. 379, 109–120.
Bondu, R., Cloutier, V., Rosa, E., Benzaazoua, M., 2017. Mobility and speciation of geogenic
arsenic in bedrock groundwater from the Canadian Shield in western Quebec, Canada.
Sci. Total Environ. 574:509–519. http://dx.doi.org/10.1016/j.scitotenv.2016.08.210.
Bretzler, A., Lalanne, F., Nikiema, J., Podgorski, J., Pfenninger, N., Berg, M., Schirmer, M.,
2017. Groundwater arsenic contamination in Burkina Faso, West Africa: Predicting
and verifying regions at risk. Sci. Total Environ. 584–585, 958–970.
Bundschuh, J., Litter, M.I., Parvez, F., Román-Ross, G., Nicolli, H.B., Jean, J.Sh., Liu, Ch., López,
D., Armienta, M.A., Guilherme, L.R.G., Cuevas, A.G., Cornejo, L., Cumbal, L., Toujaguez,
L.R., 2011. One century of arsenic exposure in Latin America: a review of history and
occurrence from 14 countries. Sci. Total Environ. 429:2–35. http://dx.doi.org/
10.1016/j.scitotenv.2011.06.024.
Cardoso, S., Grajeda, C., Argueta, S., Garrido, S.E., 2010. Experiencia satisfactoria para la
remocion de Arsenico en Condado Naranjo, Mixco, Guatemala. In: Litter, M.I.,
Sancha, A.M., Ingallinella, A.M. (Eds.), Tecnologías economicas para el abatimiento
de Arsenico en aguas. Buenos Aires, Argentina: Editorial Programa Iberoamericano
de Ciencia. Tecnologia para el Desarrollo, pp. 211–222.
Checkland, P., Scholes, J., 1990. Soft Systems Methodology in Action. John Wiley,
Chichester.
Chen, P., Yan, L., Wang, Q., Li, H., 2013. Arsenic precipitation in the bioleaching of realgar
using Acidithiobacillus ferrooxidans Hindawi Publishing Corporation. J. Appl. Chem.
http://dx.doi.org/10.1155/2013/424253.
Chuah, C.J., RuiLye, H., Ziegler, A.D., Wood, H., Kongpun, C., Rajchagool, S., 2016. Fluoride:
a naturally-occurring health hazard in drinking-water resources of Northern
Thailand. Sci. Total Environ. 545–546:266–279. http://dx.doi.org/10.1016/
j.scitotenv.2015.12.069.
Clark, I.D., Fritz, P., 1997. Environmental Isotopes in Hydrogeology. CRC Press, p. 352.
Cloutier, V., Lefebvre, R., Therrien, R., Savard, M.M., 2008. Multivariate statistical analysis
of geochemical data as indicative of the hydrogeochemical evolution of groundwater
in a sedimentary rock aquifer system. J. Hydrol. 353:294–313. http://dx.doi.org/
10.1016/j.jhydrol.2008.02.015.
705
Craig, H., 1961a. Standard for reporting concentrations of deuterium and oxygen-18 in natural
waters. Science 133:1833–1834. http://dx.doi.org/10.1126/science.133.3467.1833.
Craig, H., 1961b. Isotopic variations in meteoric waters. Science 133:1702–1708. http://
dx.doi.org/10.1126/science.133.3465.1702.
Criss, R.E., 1999. Principles of Stable Isotope Distribution. Oxford University Press, DOI
http://dx.doi.org/10.1017/S0016756803257848.
Davis, J.C., 1986. Statistics and Data Analysis in Geology. John Wiley and sons, New York:
p. 646. http://eu.wiley.com/WileyCDA/WileyTitle/productCd-0471172758.html.
Deza, E., Deza, M.M., 2009. Encyclopaedia of Distances. Springer, p. 94.
Durov, S.A., 1948. Natural Waters and Graphical Representation of Their Composition:
Dokl. Akad. Nauk. USSR. 59 pp. 87–90.
East Azerbaijan Regional Water Authority, 2010. Compressive groundwater investigation
report of Hashtrud area. p. 210.
Emberger, L., 1930. La vegetation de la region mediterraneenne. Essai d'une classification
des groupments vegetaux. Rev. Gen. Bot. 42 (641–662), 705–721.
Freeze, R.A., Cherry, J.A., 1979. Groundwater. Prentice-Hall Inc.
French, S., Maule, A.J., Mythen, G., 2005. Soft modelling in risk communication and man-
agement: examples in handling food risk. J. Oper. Res. Soc. 56 (8):879–888. http://
dx.doi.org/10.1057/palgrave.jors.2601901.
Friedman, I., 1953. Deuterium content of natural waters and other substances. Geochim.
Cosmochim. Acta 4, 89–103.
Gurung, K., Hiroaki, I., Khadka, M.S., 2005. Geological and geochemical examination of ar-
senic contamination in groundwater in the Holocene Terai Basin, Nepal. Environ.
Geol. 49, 98–113.
Hammarlund, L., Pinones, J., 2009. Arsenic in geothermal waters of Costa Rica: A minor
field study. MSc Thesis. Department of Land and Wayer Resources Engineering,
KTH Royal Institute of Technology, Sweden ISSN 1651-064X.
Hem, J., 1989. Study and interpretation of the chemical characteristics of natural water. U.
S. Geol. Surv. Water Supply Pap. 2254, 263.
Hounslow, A.W., 1995. Water Quality Data: Analysis and Interpretation. Lewis Publisher,
p. 397.
Jankoswski, J., Acworth, R.I., Shekarforoush, S., 1998. Reverse Ion-exchange in Deeply
Weathered Porphyritic Dacite Fractured Aquifer System, Yass, New South Wales,
Australia.
Johnson, C.A., Berg, M., Sabatini, D., 2014. Towards sustainable safe drinking water supply
in low- and middle-income countries: the challenges of geogenic contaminants and
mitigation measures. Sci. Total Environ. 488–489:475–476 http://dx.doi.org/
10.1016/j.scitotenv.2014.01.131.
Kaiser, H.F., 1958. The Varimax criterion for analytic rotation in factor analysis.
Psychometrika 23:187–200. http://dx.doi.org/10.1007/BF02289233.
Kapaj, S., Peterson, H., Liber, K., Bhattacharya, P., 2006. Human health effects from chronic
arsenic poisoning: a review. J. Environ. Sci. Health., Part A 41:2399–2428. http://
dx.doi.org/10.1080/10934520600873571.
Kazakis, N., Kantiranis, N., Kalaitzidou, K., Kaprara, E., Mitrakas, M., Frei, R., Vargemezis, G.,
Tsourlos, P., Zouboulis, A., Filippidis, A., 2017. Origin of hexavalent chromium in
groundwater: the example of Sarigkiol Basin, Northern Greece. Sci. Total Environ.
593–594, 552–566.
Khan, T.A., 2011. Multivariate analysis of hydrochemical data of the groundwater in parts of
Karvan-Sengar sub-basin central Ganga basin, India. 3. Global NEST:pp. 229–236.
https://journal.gnest.org/sites/default/files/Journal%20Papers/229-236_653_KHAN_13-
3.pdf.
Khatibi, R., 2008. Systemic nature of, and diversification in systems exposed to, flood
risk, WIT transactions on ecology and the environment. WIT Press Flood Recov-
ery, Innovation and Response I. Vol 118:p. 91. http://dx.doi.org/10.2495/
FRIAR080091.
Kouras, A., Katsoyiannis, I., Voutsa, D., 2007. Distribution of arsenic in groundwater in the
area of Chalkidik, Northern Greece. J. Hazard. Mater. 147, 890–899.
Lapworth, D.J., Krishan, G., MacDonald, A.M., Rao, M.S., 2017. Groundwater quality in the
alluvial aquifer system of northwest India: new evidence of the extent of anthropo-
genic and geogenic contamination. Sci. Total Environ. 599–600:1433–1444. http://
dx.doi.org/10.1016/j.scitotenv.2017.04.223.
Lloyd, J.W., Heathcote, J.A., 1985. Natural Inorganic Hydrochemistry in Relation to
Groundwater: An Introduction, Oxford: Clarendon Press. Oxford University Press,
New York:p. 1985. http://trove.nla.gov.au/version/45045800.
Mahimairaja, S., Bolan, N.S., Adriano, D.C., Robinson, B., 2005. Arsenic contamination
and its risk management in complex environmental settings. Advances in
Agronomy. 86. Elsevier Inc. pp. 1–82. www.kiwiscience.com/JournalArticles/
AdvancesInAgronomy05.pdf.
Martin, K.J.W., Mailloux, B.J., van Geen, A., Bostick, B.C., Ahmed, K.M., Choudhury, I., Slater,
G.F., 2017. Human and livestock waste as a reduced carbon source contributing to the
release of arsenic to shallow Bangladesh groundwater. Sci. Total Environ. 595:63–71.
http://dx.doi.org/10.1016/j.scitotenv.2017.03.234.
McLean, W., Jankoswski, J., 2000. Groundwater quality and sustainability in an alluvial
aquifer, Australia. In: Sililo, et al. (Eds.), Proc. IAH congress on Groundwater: Past
Achievements and Future Challenges. Cape Town South Africa 26th November–1st
December 2000. AA Balkema, Rotterdam, Brookfield.
Muhammad, A.M., Zhonghua, T., Sissou, Z., Mohamadi, B., Ehsan, M., 2016. Analysis of
geological structure and anthropological factors affecting arsenic distribution in the
Lahore aquifer, Pakistan. Hydrogeol. J. 24 (7):1891–1904. http://dx.doi.org/10.1007/
s10040-016-1453-4.
Nadiri, A.A., Asghari Moghaddam, A., Sadeghi Aghdam, F., Aghaee, H., 2012. Investigation
of arsenic anomalies in water resources of Sahand Dam. Journal of Environmental
Studies 38, 61–74.
Nadiri, A.A., Asghari Moghaddam, A., Tsai, F.T.-C., Fijani, E., 2013. Hydrogeochemical anal-
ysis for Tasuj Plain Aquifer, Iran. J. Earth Syst. Sci. 22:1091–1105. http://dx.doi.org/
10.4172/978-1-63278-061-4-062.
706 A.A. Nadiri et al. / Science of the Total Environment 613–614 (2018) 693–706
Nadiri, A.A., Gharekhani, M., Khatibi, R., Sadeghfam, S., Asghari Moghaddam, A., 2017a.
Groundwater vulnerability indices conditioned by Supervised Intelligence Commit-
tee Machine (SICM). Sci. Total Environ. 574:691–706. http://dx.doi.org/10.1016/
j.scitotenv.2016.09.093.
Nadiri, A.A., Sedghi, Z., Khatibi, R., Gharekhani, M., 2017b. Mapping vulnerability of
multiple aquifers using multiple models and fuzzy logic to objectively derive model
structures. Sci. Total Environ. 593–594:75–90. http://dx.doi.org/10.1016/
j.scitotenv.2017.03.109.
Nadiri, A.A., Gharekhani, M., Khatibi, R., Asghari Moghaddam, A., 2017c. Assessment of
groundwater vulnerability using supervised committee to combine fuzzy logic
models. J. Environ. Sci. Pollut. Res. 24 (9), 8562–8577.
Nickson, R.T., McArthur, J.M., Shrestha, B., Kyaw-Myint, T.O., Lowry, D., 2005. Arsenic and
other drinking water quality issues, Muzaffargarh district, Pakistan. Appl. Geochem.
20, 55–68.
Nidumolu, U.B., de Bie, C.A.J.M., Keulen, H.V., Skidmorec, A.K., Harmsena, K., 2006. Review
of a land use planning programme through the soft systems methodology. Land Use
Policy 23, 187–203.
Ning, Z., Lobdell, D.T., Kwok, R.K., Liu, Z., Zhang, Sh., Macheng, L., 2007. Residential expo-
sure to drinking water arsenic in Inner Mongolia, China. Toxicol. Appl. Pharmacol.
222, 351–356.
Nriagu, J.O., Pacyna, J.M., 1988. Quantitative assessment of worldwide contamination of
air, water and soils by trace metals. Nature 333:134–139. http://dx.doi.org/10.1038/
333134a0.
Pershagen, G., 1981. The carcinogenicity of arsenic. Environ. Health Perspect. 40:93–100.
http://www.biomedsearch.com/attachments/00/07/02/39/…/envhper00467-
0097.pdf.
Piper, A.M., 1944. A graphical procedure in the geochemical interpretation of water analyses
25. American Geophysics:pp. 914–923. http://dx.doi.org/10.1029/TR025i006p00914.
Pizzol, L., Zabeo, A., Critto, A., Giubilato, E., 2015. Risk-based prioritization methodology
for the classification of groundwater pollution sources. Sci. Total Environ. 506–507:
505–517. http://dx.doi.org/10.1016/j.scitotenv.2014.11.014.
Rahman, M.M., Sengupta, M.K., Ahamed, S., Chowdhury, U.K., Amir Hossain, M.D., Das, B.,
2005. The magnitude of arsenic contamination in groundwater and its health effects
to the inhabitants of the Jalangi - one of the 85 arsenic affected blocks in West Bengal,
India. Sci. Total Environ. 338, 189–200.
Singh, K., Malik, A., Mohan, D., Sinha, S., 2004. Multivariate statistical techniques for the
evaluation of spatial and temporal variations in water quality of Gomti River
View publication stats
(India)–a case study. Water Res. 38:3980–3992. http://dx.doi.org/10.1007/s10653-
005-9001-5.
Smedley, P.L., and Kinniburgh, D.G., 2005. Arsenic in groundwater and the environmental,
essential of medical geology. Elsevier Academic Press. 11, 263–299.
Stiff, H.A., 1951. The interpretation of chemical water analysis by means of patterns. Pe-
troleum Technology 3:60–62. http://dx.doi.org/10.2118/951376-G.
Tassinari, G.C.C., Pinzon, D.F., Ventura, B.J., 2008. Age and sources of gold mineralization in
the Marmato mining district, NW Colombia: a Miocene–Pliocene epizonal gold de-
posit. Ore Geol. Rev. 33, 505–518.
Tiedeken, E.J., Tahar, A., McHugh, B., Rowan, N.J., 2017. Monitoring, sources, receptors, and
control measures for three European Union watch list substances of emerging con-
cern in receiving waters – a 20 year systematic review. Sci. Total Environ. 574:
1140–1163. http://dx.doi.org/10.1016/j.scitotenv.2016.09.084.
Todd, D.K., Mays, L.W., 2005. Groundwater hydrology. John Wiley and Sons, New York,
p. 535.
Tseng, C.H., 1977. Effects and dose-response relationships of skin cancer and blackfoot
disease with arsenic. Environ. Health Perspect. 19:109–119. https://
www.ncbi.nlm.nih.gov/pmc/articles/PMC1637425/.
Uher, E., Mirande-Bret, C., Gourlay-Francé, C., 2017. Assessing the relation between an-
thropogenic pressure and PAH concentrations in surface water in the Seine River
basin using multivariate analysis. Sci. Total Environ. 557–558, 551–561.
Wang, J., Liu, G., Liu, H., Lam, P.K.S., 2017. Multivariate statistical evaluation of dissolved
trace elements and a water quality assessment in the middle reaches of Huaihe
River, Anhui, China. Sci. Total Environ. 583, 421–431.
Wang, S., Luo, K., Ni, R., Tian, Y., Gao, X., 2017. Assessment of elemental background values
and their relation with lifespan indicators: a comparative study of Jining in Shandong
Province and Guanzhong area in Shaanxi Province, northern China. Sci. Total Environ.
595:315–324. http://dx.doi.org/10.1016/j.scitotenv.2017.03.265.
Ward Jr., J.H., 1963. Hierarchical grouping to optimize an objective function. J. Am. Stat.
Assoc. 58, 236–244.
World Health Organization (WHO), 2004. Guidelines for Drinking-water Quality. Third
Edition Vol. 1, Recommendations, WHO Press World Health Organization, Geneva.
p. 515.