Pa

Ya

b

a

ARAE

KQDPHAD

1

AjHwh

gDtiLw

UT

h0

Chemie der Erde 75 (2015) 197–205

Contents lists available at ScienceDirect

Chemie der Erde

j o ur na l ho mepage: www.elsev ier .de /chemer

rincipal component analysis and hierarchical cluster analyses ofrsenic groundwater geochemistry in the Hetao basin, Inner Mongolia

uxiao Jianga,b, Huaming Guoa,b,∗, Yongfeng Jiaa,b, Yongsheng Caoa,b, Chao Hub

State Key Laboratory of Biogeology and Environmental Geology, China University of Geosciences, Beijing 100083, PR ChinaSchool of Water Resources and Environment, China University of Geosciences, Beijing 100083, PR China

r t i c l e i n f o

rticle history:eceived 23 August 2014ccepted 26 December 2014ditorial handling – J.C.J. Petit

eywords:uantitative investigationseep aquifersrincipal component analysisierarchical cluster analysiss speciesominant factor

a b s t r a c t

Although high As groundwater has been observed in shallow groundwater of the Hetao basin, little isknown about As distribution in deep groundwater. Quantitative investigations into relationships amongchemical properties and among samples in different areas were carried out. Ninety groundwater sampleswere collected from deep aquifers of the northwest of the basin. Twenty-two physicochemical param-eters were obtained for each sample. Statistical methods, including principal component analysis (PCA)and hierarchical cluster analysis (HCA), were used to analyze those data. Results show that As specieswere highly correlated with Fe species, NH4-N and pH. Furthermore, result of PCA indicates that highAs groundwater was controlled by geological, reducing and oxic factors. The samples are classified intothree clusters in HCA, which corresponded to the alluvial fans, the distal zone and the flat plain. Moreover,the combination of PCA with HCA shows the different dominant factors in different areas. In the alluvialfans, groundwater is influenced by oxic factors, and low As concentrations are observed. In the distal

zone, groundwater is under suboxic conditions, which is dominated by reducing and geological factors.In the flat plain, groundwater is characterized by reducing conditions and high As concentrations, whichis dominated by the reducing factor. This investigations indicate that deep groundwater in the alluvialfans mostly contains low As concentrations but high NO3 and U concentrations, and needs to be carefullychecked prior to being used for drinking water sources.

© 2015 Elsevier GmbH. All rights reserved.

. Introduction

In China, about 15 million people have been affected by highs groundwater with As concentration >10 �g/L, especially in Xin-

iang, Inner Mongolia and Shanxi provinces (Guo et al., 2014a). Theetao plain is one of the well-known high As groundwater areas,here local residents were exposed to drinking groundwater withigh As concentration (Guo et al., 2014a; Luo et al., 2012).

High As concentration was frequently observed in shallowroundwater in the flat plain of the Hetao basin (Guo et al., 2008;eng et al., 2009). Previous investigations indicated that distribu-

ion of high As groundwater exhibited a distinct spatial zonation,

ncluding Shuangmiao-Sandaoqiao, Shahai-Manhui, Bainaobao-angshan, Taerhu, and Shengfeng (Yang et al., 2008), whichere coincident with the subsidence center of the basin. Locally,

∗ Corresponding author at: School of Water Resources and Environment, Chinaniversity of Geosciences (Beijing), Beijing 100083, PR China.el.: +86 10 8232 1366; fax: +86 10 8232 1081.

E-mail address: hmguo@cugb.edu.cn (H. Guo).

ttp://dx.doi.org/10.1016/j.chemer.2014.12.002009-2819/© 2015 Elsevier GmbH. All rights reserved.

biogeochemical and hydrogeological conditions also affectedgroundwater As concentrations (Guo et al., 2011). High As ground-water mainly occurred at depths between 20 and 30 m below landsurface. However, few studies were carried out along the piedmontareas of the Langshan Mountains, which are located in the north ofthe Hetao basin. In the areas, many wells with various screeningdepths (20–100 m) are used for irrigation and domestic watersupply. Jia et al. (2014) showed that low As concentrations wereobserved in deep groundwater near the mountains, while high Asconcentrations in deep groundwater of the flat plain. Since ground-waters are used for both drinking water resources and agriculturalirrigation in this region, As concentrations of these waters wouldnot only affect quality of food products, but also determine drinkingwater quality.

In recent years, multivariate statistical methods, including prin-cipal component analysis (PCA) and hierarchical cluster analysis(HCA), have been applied to investigate the factors controlling

As mobilization from large sets of groundwater chemistry dataand to classify the groundwater areas (Andrade and Stigter, 2013;Mukherjee et al., 2012). Based on PCA and HCA methods Andradeand Stigter (2013) established FRA explanatory models, which

1 der Er

ifnscfitPwdAeacS

gtgtiacg

2

2

cts

ptb

2

s(cp2bseofmfltb2

htidrdA>

98 Y. Jiang et al. / Chemie

mproved the understanding of the role of land use and intrinsicactors, such as aquifer lithology and water depth, on As contami-ation. Mukherjee et al. (2012) studied the relationships among theamples in the areas between the Himalayan foothills and Indianraton (including central Gangetic plain) using HCA and PCA, whichnally explained the influence of geology and geomorphology onhe As fate in aquifers. Another study in Taiwan, which also used theCA and HCA, indicated that the reductive dissolution of Fe mineralsas prerequisite for mobilization of As, and the shift of redox con-itions from Fe- to As-reducing led to the accumulation of dissolveds in aquifers of Choushui River alluvial fan and Chianan Plain (Lut al., 2012). Moreover, an investigation on the sources of As by PCAnd HCA showed that groundwater As might be from geologic andlimatic origins, instead of anthropogenic sources (Andrade andtigter, 2013).

In this study, PCA and HCA methods are applied to investi-ate groundwater As distribution and geochemical processes inhe piedmont areas. Objectives are to (1) investigate distribution ofroundwater As along the piedmont areas of the Langshan Moun-ains; (2) evaluate factors contributing to high As concentrationsn groundwater by means of PCA; (3) locate low As groundwaterreas by means of HCA in combination with PCA; (4) assess geo-hemical processes controlling occurrence and mobility of As inroundwater.

. Study area

.1. Location and climate

The study area is located in the northwest of the Hetao basin. Itovers the piedmont of the Langshan Mountains, and extends fromhe front of the mountain ranges in the north to the plain in theouth (Fig. 1).

The climate in the region is arid, with the annual averagerecipitation between 130 and 220 mm, annual average evapora-ion between 2000 and 2500 mm; and the average temperatureetween 5.6 and 7.8 ◦C.

.2. Geology and hydrogeology

The Langshan mountain ranges are mainly composed of Juras-ic to Cretaceous metamorphic rocks (slate, gneiss and marble)Guo et al., 2008). Basement bedrocks in the Hetao basin areharacterized by Early Archean and Proterozoic metamorphic com-lexes (gneiss, schist and slate), granite, and quartzite (Hu et al.,013). Inland lacustrine sediments with fine clast have locallyeen deposited in the Quaternary period and thick Mid-Cenozoicedimentary formation has developed (Deng et al., 2011; Lint al., 2002). The thickness of the sediment in the southeastf the basin ranges from 500 to 1500 m, and in the northwestrom 7000 to 8000 m (Guo et al., 2008, 2011). The sediments are

ainly composed of alluvial–pluvial sand, sandy silt, lacustrine anduvial–lacustrine sandy silt, silty clay and clay rich in organic mat-er in the central part of the basin, fluvial sand and fine sand on theanks of rivers, and alluvial sand in the fan areas (Guo et al., 2008,011).

There are evident zonation characteristics in geological andydrogeological conditions in the study area. From the foothills tohe plain, grain-size of sediments decreases, with the decreasingn permeability and hydraulic conductivity. Hydraulic conductivityecreases along the flow path from the alluvial fans to the flat plain,

anging from ∼2.0 m/d along the mountain front to ∼0.2 m/d in theowndip region (Inner Mongolia Institute of Hydrogeology, 1982).ccording to hydrogeological conditions, groundwater at depths40 m is usually considered as deep groundwater (Jia et al., 2014).

de 75 (2015) 197–205

Shallow groundwater usually occurs at depths <10 m. Deep ground-water is mainly recharged by lateral flow from mountains andvertical infiltration from shallow groundwater, and discharged byartificial extraction. Shallow groundwater is mainly recharged fromprecipitation, irrigation water, and surface water (from lakes andwater channels), which is mainly discharged by artificial extrac-tion, evaporation, and vertical flow into deep groundwater. Surfacetopography strongly affects directions of groundwater flow. Thedirection of groundwater flow is from north to south (Fig. 1) withthe flow rate generally higher in the alluvial fans than in the flatplain. Due to the relatively flat terrain, fine grain of the aquifer, aridclimate and low hydraulic conductivity, the groundwater flow ratein the plain is very low (Guo et al., 2012).

3. Materials and methods

3.1. Sample collection and analysis procedures

Ninety groundwater samples were collected from tube wells in the study areain the years of 2011 and 2012 (Fig. 1). Depths of sampling wells varied from 50 to110 m below ground level (bgl), thus essentially representing deep groundwaters.

In the field, groundwater was sampled after pumping sufficiently until the flow-ing water showed a stabile temperature, pH, EC (specific conductance), and ORP.Parameters (including ORP, temperature, pH, and EC) were measured in an in-line flow cell under minimal atmospheric contact using a multiparameter portablemeter (HANNA, HI 9828), which was calibrated using standard solution beforeuse. Redox-sensitive parameters (including Fe(II), NH4-N, H2S) were determinedusing a portable spectrophotometer (HACH, DR2800). Alkalinity was determinedby titration with 0.8 M H2SO4 using a Model 16900 digital titrator (HACH) withphenolphthalein and methyl-orange indicators.

All groundwater samples were filtered through 0.22 �m filters (Sartorius). Sam-ples for cation and trace element analysis were collected in HDPE bottles, andacidified to pH <2 using ultrapure HNO3. Those for anion analysis were not acid-ified. Samples for dissolved organic carbon (DOC) analysis were collected in amberglass bottles and acidified with H2SO4 to pH <2. Sub-samples for As species analysiswere preserved with 0.25 M EDTA (10%) in 2.0 mL amber glass bottles. All ground-water samples were stored in an ice box, and delivered to the laboratory, wherethey were immediately preserved in a refrigerator at 4 ◦C until analysis within twoweeks.

Major cations and trace metal elements were measured using an inductivelycoupled plasma atomic emission spectrometer (iCAP6300, Thermo) and ICP-MS(7500C, Agilent), respectively. The analysis precision of ICP-AES and ICP-MS was0.5%. The detection limit for As was 0.01 �g L−1. Concentrations of Cl, NO3, and SO4

were determined using an ion chromatography system (ICS2000, Dionex), with theanalysis precision less than 3.0%. Arsenic species (including As(III), As(V), MMA,and DMA) were analyzed by high-performance liquid chromatography-atomic flu-orescence spectrometry (HPLC-AFS) (AFS9130, Titan), with the relative standarddeviation (RDS) <±5% and the analysis precision of 5.0% (Guo et al., 2014b). Dis-solved organic carbon (DOC) was determined by total carbon analyzer (TOC-Vwp,Shimadzu). For most samples, ion charge imbalances were less than 5%.

3.2. Statistical methods

Multivariate statistics (including principal component analyses (PCA) andhierarchical cluster analyses (HCA)) were employed to quantitatively investigaterelationships among the dataset of 1980 values of the investigated samples (22parameters determined in 90 samples).

Principal component analysis (PCA) is one of the most commonly used multi-variate statistical methods in natural sciences, which was developed by Hotelling(1933) in the thirties from original work of Pearson. The main objective of thismethod is to simplify data structure by reducing the dimension of the data. The orig-inal parameters would be rearranged into several new uncorrelated comprehensivecomponents (or factors) without losing significant information (Brown and Brown,1998; Pereira and Sousa, 1988). Every new component is the linear combination ofthe original variables and unrelated, which makes it accurate to describe the charac-teristics of the analyzed data. The calculated factors are rotated with the method ofVarimax rotation, thus making the loadings of closely related variables in each factorstrengthened. Each component describes a certain amount of the statistical varianceof the analyzed data and is interpreted according to the inter-correlated variables.Variable loadings are defined by the orthogonal projection of the variables on eachof the factors. The selection of the factors is based on both the significance (eigenval-ues >1) of the factor and the cumulative percentage of data variance explained. The

last and significant step is to interpret each factor in association with the studiedissue.

In this study, after the interpretation of each factor, the contribution of each fac-tor (factor scores) at each monitoring well was computed. The factor scores showedthe relation between samples and the components quantitatively, which indicated

Y. Jiang et al. / Chemie der Erde 75 (2015) 197–205 199

ea an

ttfa

bcmpsctdsa

Z

wtd

ng

4

4

iA<Wiitac

Fig. 1. Location map of study ar

he extent of the impact of each factor on samples. Besides, the factor scores showedhe similarity of the individual samples. To synthetically analyze the effect of eachactor on samples and quantify the similarity of the samples, hierarchical clusternalysis (HCA) was applied subsequently.

Hierarchical cluster analysis uses a variance approach to define a distanceetween variable clusters, attempting to minimize the sum of squares of any twolusters that could be formed at each step. In this study, factor scores at eachonitoring well were used as variables in the HCA for the grouping following the

rocedure of HCA (Cloutier et al., 2008; Yidana, 2010; Yidana et al., 2010). Finally,amples with similar factor scores were grouped into the same cluster. To avoid mis-lassifications arising from the different orders of magnitude of variables or fromhe effect of parameters with the highest variances on the calculation of Euclideanistance (Cloutier et al., 2008; Güler et al., 2002), the variance for each variable istandardized to their corresponding Z scores, which are calculated by Eq. (1) (Davisnd Sampson, 2002).

i = (Xi − mean)s

(1)

here Zi is the standardized Z scores; Xi is the value of each variable; mean ishe mean value of the normal distribution from each datum, and s is the standardeviation of the distribution.

Furthermore, the dominant factors in each group would be revealed in combi-ation with the results of PCA, which, in this study, reflects different behaviors ofroundwater in different areas.

. Results

.1. Descriptive statistics and scatter plots

Table 1 shows the descriptive statistics for As and other phys-cochemical parameters measured in site and in the laboratory.rsenic concentrations of deep groundwaters ranged between0.01 and 342 �g/L. Approximately 37.8% of samples exceeded theHO guideline value of 10 �g/L. In this area, groundwaters are

ntensively used for irrigation. Studies have shown that As can be

ntroduced into food chain by irrigation with high As groundwa-er (Tong et al., 2014; Panaullah et al., 2009), which has become

new health threat to local residents. In addition to As, physico-hemical parameters, including ORP, U, NO3, SO4, Fe(II), and NH4-N,

d details of sampling locations.

which are related to redox environment, also show large variations(Table 1).

It can be noted that, groundwater As is mainly composed ofAs(III) species, which is more toxic to the health. Fig. 2 showsthe scatter plots of As(total) versus other physicochemical param-eters. It is observed that As is closely related to redox sensitiveparameters. Negative correlation exists between As and ORP value(Fig. 2a). Arsenic is also negatively correlated with U, NO3, and SO4(Fig. 2b–d). However, As is positively correlated with NH4-N, Fe(II),and total dissolved Fe. It indicated that high As groundwater isassociated with reducing environment.

4.2. Principal component analysis (PCA)

The PCA results of 22 geochemical parameters from 90 ground-water samples are shown in Table 2 and Fig. 3. Four major principlecomponents (F1, F2, F3, and F4) affecting quality of deep ground-water are identified, which explains 78.2% variance of the originaldata structure (Table 2). The data shown in bold in Table 2 indi-cate relatively higher loading and contribution to the respondingcomponents. F1, which explains 28.7% of the total variance, hashigh loadings on HCO3, Cl, Na, EC, TDS, NH4-N, K, Mg, and Sr.The high contribution of EC and TDS to this factor is likely dueto dissolved ions in the groundwater. The major ions (i.e., Cl,Na, K, and Mg) are associated with hydrochemical variables frommineral weathering and water–rock interactions in the aquifer,which indicates that F1 would reflect climatic and geological(lithological) effects. Other studies also showed that major ionsreflected effects of geothermal system and mineral weathering(Al-Farraj et al., 2013; Brahman et al., 2013; Chapagain et al.,

2010; Purushotham et al., 2011). However, the water quality vari-ables, like HCO3 and NH4-N, are closely related to organic mattercontents in the aquifers (Jia et al., 2014; Oremland and Stolz,2005).

200 Y. Jiang et al. / Chemie der Erde 75 (2015) 197–205

Table 1Descriptive statistics for As and other physicochemical parameters.

Parameter Units N Min Max Mean SD Variance

As(total) �g/L 90 <0.01 342 74.8 94.5 8925As(III) �g/L 90 <0.01 342 74.7 103 10,624Ca mg/L 90 55.0 300 112 43.5 1896Cl mg/L 90 40.8 966 209 212 45,004EC �S/cm 90 504 4230 1400 986 971,869Fe(II) mg/L 90 <0.002 3.53 0.56 0.67 0.44Fe(total) mg/L 90 4.92 3580 608 728 529,980HCO3 mg/L 90 157 678 316 113 12,671K mg/L 90 3.02 14.7 6.20 2.50 6.25Mg mg/L 90 14.9 225 47.0 38.5 1484Mn mg/L 90 0.77 295 103 61.6 3797Na mg/L 90 25.8 734 152 162 26,386NH4-N mg/L 90 <0.01 4.90 1.03 1.27 1.62NO3 mg/L 90 <0.01 188 22.1 33.6 1130ORP mV 90 −194 243 −68.2 98.4 9677pH 90 6.67 8.34 7.63 0.33 0.11SO4 mg/L 90 30.2 1110 280 248 61,666Sr �g/L 90 380 3750 1290 825 681,302H2S �g/L 90 <1.0 66.0 7.88 10.8 117TDS mg/L 90 252 2120 701 492 242,127DOC mg/L 90 0.76 12.9 3.86 2.38 5.67U �g/L 90 <0.01 323 36.5 57.5 3305

SD = standard deviation, N = number of samples, EC = electrical conductivity compensate at 25 ◦C.

Fig. 2. Scatter plots of As(total) versus ORP (a), U (b), NO3 (c), SO4 (d), NH4-N (e), Fe(II) (f), Fe(total) (g), HCO3 (h).

Y. Jiang et al. / Chemie der Er

Table 2Rotated factor loadings of principal components on variables.

Variables F1 F2 F3 F4

Sr 0.887 0.299 0.221 −0.08HCO3 0.835 0.124 0.346 0.035Cl 0.826 0.441 0.222 −0.062Na 0.82 0.44 0.198 −0.093EC 0.767 0.596 0.154 0.1TDS 0.765 0.595 0.152 0.104NH4-N 0.664 −0.083 0.532 −0.403K 0.648 0.237 0.178 0.344Mg 0.64 0.632 0.009 0.2H2S −0.443 −0.057 0.082 −0.419Mn 0.171 0.87 0.172 −0.197SO4 0.508 0.768 −0.004 0.206Ca 0.313 0.662 0.026 0.553DOC 0.188 0.508 0.055 −0.097Fe(II) 0.114 0.396 0.855 −0.177Fe(total) 0.15 0.438 0.811 −0.142As(III) 0.454 −0.121 0.777 −0.314As(total) 0.395 −0.122 0.764 −0.335NO3 0.124 −0.116 −0.112 0.827ORP −0.004 −0.12 −0.179 0.818U −0.1 0.134 −0.285 0.715pH 0.022 −0.008 0.362 −0.428Eigenvalue (%) 28.743 18.797 15.88 14.742Cumulative (%) 28.743 47.54 63.42 78.162

cmia1Afet2eroNwaAUto

et al., 2009). To represent the relative importance of factors in dif-

Fig. 3. Plots of PCA scores for As species and other parameters.

F2 accounts for 18.8% of the total variance, which is positivelyorrelated with Mn, SO4, and Ca, indicating impact of Mn-bearinginerals on high As groundwater. In addition, the loading on DOC

s relatively high in F2, which indicates that the Mn-bearing miner-ls effect is accompanied by organic matters. Among total variance,5.9% is explained by F3. It is carried by NH4-N, Fe(II), dissolved Fe,s(III) and total As. This factor is highly weighted by As among the

our factors. The reductive dissolution of minerals is responsible forlevated Fe and Mn (Fig. 3). Therefore, As species are mainly con-rolled by the redox conditions of groundwater (Chapagain et al.,009; Guo et al., 2008). It indicates that both F2 and F3 could reflectffects of reducing factors. Furthermore, As species are highly cor-elated to Fe species, which is consistent with the scatter plotsf As versus Fe(II) and total dissolved Fe (Fig. 2). The loadings ofH4-N and pH are relatively higher on F3 than on others (Table 2),hich reflects that As may be released and mobilized in the alkaline

quifers, in addition to reductive dissolution of Fe oxyhydroxides.dditionally, 17.7% of total variance is exhibited by ORP, NO3 and

under F4, which are associated with oxic groundwater condi-ions. Moreover, the negative contribution of H2S also supports thisbservation. Therefore, F4 indicates effects of the oxic conditions.

de 75 (2015) 197–205 201

As mentioned above, F2 and F3 can be reflected by F1 and F4to some degree. Therefore, F1 and F4 are considered in the plot ofPCA scores (Fig. 3). The plot shows characteristics of the compo-nents and helps to understand the relative importance of variablesin components. The negative direction of F4 has the meaning ofreducing conditions for F2 and F3, with the predominance of NH4-N, DOC, H2S, Mn, As and Fe species (Table 2), while the positivedirection indicates oxic conditions with the predominance of NO3,U and ORP. In addition, the positive direction of F1 (Fig. 3) reflectsthat the climatic and geological (lithological) effects. The plot of Mn,As and Fe species shows that climatic and geological (lithological)effects may contribute to their release to groundwater.

4.3. Hierarchical cluster analyses (HCA)

To quantify the similarity of the samples, hierarchical clusteranalyses (HCA) was then performed based on the principal com-ponent analysis (PCA). In HCA, factor scores obtained in PCA wereused as variables for statistical grouping. The dendrogram of HCAis presented in Fig. 4. The phenon line is defined as the distance of20, by which all samples are classified into three statistically signif-icant clusters. Cluster I is composed of the wells mainly located inthe flat plain (Fig. 1). Most samples in Cluster II lie in the transitionzone (the distal of alluvial fans), except several wells in foothills.Cluster III mainly includes wells from the alluvial fans of the Lang-shan Mountains. To quantify the variation between clusters andsummarize the characteristics of samples in different clusters, theminimum, the maximum and the mean values of physicochemicalparameters are presented in Table 3.

In the study area, pH values have the range between 7 and 8.Groundwater has slightly higher mean pH value in Cluster I thanClusters II and III (Table 3). Bicarbonate, NH4-N and DOC concen-trations are higher in Clusters I and II than Cluster III, with themean values of 331 mg/L, 1.64 mg/L and 4.07 mg/L in Cluster I, and355 mg/L, 0.83 mg/L and 4.95 mg/L in Cluster II, respectively. Sam-ples in Cluster III are characterized by higher ORP values, higherNO3 and U concentrations, in comparison with Clusters I and II(Table 3), with mean values of 26.6 mV, 46.8 mg/L and 90.4 �g/L,respectively. Therefore, Cluster III shows the relatively more oxicstate than other two clusters. Mean concentrations of As(III), Fe(II)and H2S are much higher in Cluster I than Clusters II and III, whichreflects more reducing conditions in Cluster I.

In Cluster I, As(III) concentration ranges between <0.01 and342 �g/L. Around 89.6% of groundwater samples have As(III) con-centration >10 �g/L (the WHO drinking water guideline value),and 68.8% samples >50 �g/L (the Chinese drinking water guidelinevalue in rural areas). Apparently, groundwater in Cluster I wouldpose a health risk to local residents.

5. Discussion

5.1. Factors affecting groundwater chemistry

The multivariate statistical analysis, including PCA and HCA,indicates that As concentration is associated with redox-sensitiveparameters (Fig. 2, and Table 3). Effects of reducing conditionshave been studied and the similar relations are found (Andradeand Stigter, 2013; Kim et al., 2009; Phuong et al., 2012; Smedleyand Kinniburgh, 2002). Surface geology has been considered as animportant effect on As concentration in alluvial groundwater (Kim

ferent clusters, all groundwater samples of the three clusters areallocated in the factors according to their factor scores obtained inPCA (Fig. 5).

202 Y. Jiang et al. / Chemie der Erde 75 (2015) 197–205

Fig. 4. Dendrogram for groundwater samples.

Table 3Basic description of physicochemical parameters of groundwater in three clusters.

Variables Units I (N = 48) II (N = 15) III (N = 27)

Min Max Mean Min Max Mean Min Max Mean

As(III) �g/L <0.01 342 133 <0.01 137 19.7 <0.01 7.00 1.18Fe(II) mg/L <0.002 3.53 0.91 <0.002 1.00 0.13 <0.002 1.00 0.17HCO3 mg/L 179 678 331 203 464 355 157 500 266NH4-N mg/L <0.01 4.90 1.64 <0.01 2.00 0.83 <0.01 <0.01 <0.01NO3 mg/L <0.01 56.0 7.92 1.00 98.0 23.2 1.00 188 46.8ORP mV −165 −61.0 −120 −170 145 −74.5 −194 243 26.6pH 7.00 8.34 7.77 7.00 8.00 7.56 6.67 8.00 7.43H2S �g/L 2.0 66.0 12.5 <DOC mg/L 1.28 11.4 4.07

U �g/L <0.01 62.0 4.53 <

tStOCaIpF

Fig. 5. Distribution of factors on samples of the three clusters.

Figs. 4 and 5 clearly reveal the dominant factors in different clus-ers. In general, Cluster III has F4 values >0, while Clusters II and I <0.amples in Cluster III are mostly plotted along the positive direc-ion of Fig. 5, which are enriched in U and NO3 with relatively higherRP values (Fig. 3). It indicates that groundwater geochemistry ofluster III is controlled by oxic factors, that is, groundwaters in the

lluvial fans of the Langshan Mountains corresponding to ClusterII (Fig. 1) are under oxic state. Samples in Cluster II are mostlylotted in the positive direction of F1 and the negative direction of4, which indicates that they are affected by both lithological and

1.0 6.00 1.13 <1.0 10.0 3.480.87 12.9 4.95 0.76 6.27 2.870.01 233 41.7 14.0 323 90.4

redox factors. Cluster II is located in the transitional zone betweenalluvial fans and the plain. Furthermore, it also indicates that As,Fe and Mn in Cluster II should be released (lithological factors) tothe groundwater under reducing conditions (reducing factors). Sta-tistical results show that SO4, EC, As and Mn concentrations arethe highest in Cluster II (Fig. 6). Many researchers have also indi-cated that As was released with the reductive dissolution of Fe/Mnminerals (Deng et al., 2009; Guo et al., 2008). Although similar situa-tions occur in Cluster I, effects of reducing conditions are relativelymore apparent (Fig. 5) in the flat plain. This is supported by thedata presented in Table 3, showing that redox sensitive compo-nents (including As and Fe) are elevated in this cluster. However,the significant difference is that samples of Cluster II are mainlylocated in the second quadrant of Fig. 5, enriched in Fe(II), whilemost of samples in Cluster I are in the third quadrant of Fig. 5, withhigh H2S concentration (Fig. 3 and Table 3).

5.2. Controls of redox conditions on groundwater Asconcentration

Samples in Cluster III are mainly located in the alluvial fans(Fig. 1), where groundwater flow is generally high, greater than

20 cm/d with relative high permeability and hydraulic gradients(Jia et al., 2014). The high hydraulic gradient and conductivitylead to the great recharge of atmospheric precipitation and sur-face water, and thus enrichment in dissolved oxygen. Moreover,

Y. Jiang et al. / Chemie der Erde 75 (2015) 197–205 203

b), As

trtvdtrtfatisae2obm(

fiiaIlalr

wtlScIcoioimM

Fig. 6. Box-whisker plots of EC (a), Mn (

he high flow rate leads to rapid flushing of groundwater, whichesults in low EC values of groundwater in Cluster III (Fig. 6a). Withhe effect of oxic factors, this zone is characterized by high ORPalues, high concentrations of NO3 and U (Table 3). Even thoughissolved organic matters are recharged locally in oxic conditions,hey are oxidized immediately by free oxygen or NO3. Therefore,elatively low DOC is observed (Table 3). Meanwhile, in oxic condi-ions, adsorption of As on aquifer sediments occurs in the alluvialans, which leads to low As concentrations in Cluster III (Fig. 6c),lthough the grains are coarse in the alluvial fans with low adsorp-ion capacity for As (Stumm, 1992). Other researchers have alsondicated that the reductive dissolution of Fe (hydr)oxides is largelyuppressed because that NO3 strongly buffers the redox potentialnd As would be adsorbed on Fe minerals in the NO3 reductionnvironment (Guo et al., 2011; Kim et al., 2009; Nickson et al.,000; Senn and Hemond, 2002; Gibney and Nüsslein, 2007). As freexygen is consumed along the flow path, NO3 becomes the first toe electron acceptor in the process of organic matter oxidation byicroorganisms, which leads to the decrease in NO3 concentration

Guo et al., 2011).Along the groundwater flow, grain-size changes from coarse to

ne in the down-gradient of the alluvial fans, which leads to thencrease in the retention time of groundwater. ORP value is gener-lly lower with longer retention time of groundwater in the aquifer.t is observed that the ORP values are lower in Clusters II and I withower concentrations of NO3 and U (Table 3). Wells in Clusters IInd I (Fig. 1) are affected by lithological and reducing factors, whicheads to reductive dissolution of Fe–Mn (oxyhydr) oxides. As NO3 iseduced to a certain extent, Fe(III) becomes the oxidant in Cluster II.

However, in Cluster I, SO4 reduction occurs in reduction statesith the formation of H2S in groundwater, which is confirmed by

he analysis result that wells of Cluster I are plotted in the lower-eft of Fig. 5 with high loadings of H2S (Fig. 3). Moreover, the lowestO4 concentrations in Cluster I are observed among the threelusters (Fig. 6d), also indicating that SO4 is consumed in Cluster, with the highest H2S concentration (Table 3). Furthermore, Asoncentration is high in Clusters II and I, with high concentrationsf Fe(II), Mn, DOC, NH4-N, H2S and HCO3 (Table 3 and Fig. 6). Itndicates that reductive dissolution of Fe–Mn (oxyhydr)oxides

ccurs in Clusters II and I, which leads to As release under thempact of lithological and reducing factors. This process would be

ediated by microorganism (Guo et al., 2008; Islam et al., 2004;ukherjee et al., 2008; Postma et al., 2007; Wang et al., 2012a,b).

(c) and SO4 (d) values in three clusters.

The relationships between As and SO4, NH4-N, and Fe(II) shown inFig. 2 also support this observation. In addition, groundwater As isdominated by As(III), which reflects the effect of reduction factors.

In addition to the reductive dissolution of Fe(III) minerals, Asis mobilized in alkaline conditions via competitive adsorption inCluster I. The pH value and HCO3 concentration are higher (Table 3)in Cluster I, where As concentration is the highest with the higherratio of As(III) to total As.

5.3. Sources of groundwater As

Although As in groundwater has two sources, natural andanthropogenic, groundwater As would originate from naturalsources and its release would be promoted by human activities inthis study area. The negative correlation between As and NO3 isobserved (Fig. 2c). In nature, the mobility and fate of NO3 and Asare driven by different conditions. Arsenic concentrations can belimited by the introduction of NO3 from the land surface (Kim et al.,2009), that is, As from nature is negatively related to NO3. Besides,results of PCA show that F1 geological (lithological) factor is thefirst principle component (Table 2). It indicates that the dominantfactor controlling As concentration is geologically dependent.

In addition, the mean P values in the groundwater of the flatplain, the distal zone and the alluvial fans, corresponding to Clus-ters I, II and III, are 60.7, 29.5 and 14.1 �g/L, respectively. It showsthat high As concentrations correspond to high P concentrations inthe flat plain (Cluster I) and low As concentrations to low P concen-trations in the alluvial fans (Cluster III), indicating that the mobilityof As may be influenced by P in the study area. Although the useof chemical fertilizers in agricultural activity is not the source ofAs, it influences the mobility of As from soil to groundwater (Zahidet al., 2008) with the introduction of P. Phosphorous (P) fertilizers,which may increase As solubility due to the competitive PO4–AsO4exchange in the soils (Campos, 2002; Smedley and Kinniburgh,2002). In NPK fertilizers, which was used in basal and top dress-ing in extensive farming and horticulture in an area of Spain, Aswas detected in groundwater (Otero et al., 2005).

5.4. Water safety

According to this study, it can be noted that not all deep ground-water is safe for drinking. Arsenic concentration is relatively low indeep groundwater of the alluvial fans, which can be considered to

2 der Er

ba

tttif2ABtahr

Tts

6

tpoIfgfirwanaiatA

A

rtPN2

R

A

A

A

E

B

B

04 Y. Jiang et al. / Chemie

e the safe drinking water sources. However, concentrations of NO3nd U are relatively high in the alluvial fans.

Moreover, As concentrations in the flat plain correspondingo Cluster I is mostly higher than the World Health Organiza-ion drinking-water guideline value. Although the groundwater inhis region is not used for drinking, it is utilized for agriculturalrrigation. Investigation has shown that As can be introduced intoood chain by irrigation with high As groundwater (Tong et al.,014; Panaullah et al., 2009), which is one of the main pathways ofs exposure (Aitio and Becking, 2001; Authority, 2009; Carbonell-arrachina et al., 2009). Irrigation with high As groundwater leadso elevated As concentration in wheat (Tong et al., 2014), and maizend sunflowers (Neidhardt et al., 2012). Those crops irrigated withigh As groundwater would be a potential health risk to the localesidents (Tong et al., 2013).

As mentioned above, groundwater As is geologically dependent.herefore, it is urgent to find effective economically viable methodso forecast the safe aquifers as drinking water resources, rather thanimply switching wells from shallow to deep aquifers.

. Conclusions

The statistics methods (PCA and HCA) show that the main fac-ors controlling concentrations of As and others physicochemicalarameters in deep groundwater include lithological, reducing andxic factors. The relative importance is different in different zones.n the alluvial fans, groundwater chemistry is controlled by the oxicactor, where low As concentrations are observed. In the distal zone,roundwater is suboxic and influenced by geological and reducingactors. The positive relationship between As and Mn/Fe in this zonendicates the release of As occurred in suboxic conditions with theeductive dissolution of Fe/Mn minerals. In the flat plain, ground-ater is under reducing conditions and characterized by high As

nd H2S concentrations. Geological (lithological) effect is the domi-ant factor in high As groundwater, which provides the source of As,nd leads to the release of As to the groundwater under the reduc-ng conditions. Besides, this release is influenced by agriculturalctivities in the plain (irrigation and fertilizer usage). It suggestshat deep groundwater in the alluvial fans mostly containing lows concentrations.

cknowledgements

The study has been financially supported by National Natu-al Science Foundation of China (nos. 41222020 and 41172224),he program of China Geology Survey (12120113103700), therogram for New Century Excellent Talents in University (No.CET-07-0770), and the Chinese Universities Scientific Fund (No.652013028).

eferences

itio, A., Becking, G., 2001. Arsenic and Arsenic Compounds. World Health Organi-zation, Geneva.

l-Farraj, A.S., Al-Wabel, M.I., El-Saeid, M.H., El-Naggar, A.H., Ahmed, Z., 2013.Evaluation of groundwater for arsenic contamination using hydrogeochemicalproperties and multivariate statistical methods in Saudi Arabia. J. Chem. 2013,1–9.

ndrade, A., Stigter, T., 2013. The distribution of arsenic in shallow alluvial ground-water under agricultural land in central Portugal: insights from multivariategeostatistical modeling. Sci. Total Environ. 449, 37–51.

uropean Food Safety Authority, 2009. Panel on contaminants in the food chain(CONTAM): scientific opinion on arsenic in food. EFSA J. 7.

rahman, K.D., Kazi, T.G., Afridi, H., Naseem, S., Arain, S.S., Wadhwa, S.K., Shah, F.,

2013. Simultaneously evaluate the toxic levels of fluoride and arsenic species inunderground water of Tharparkar and possible contaminant sources: a multi-variate study. Ecotoxicol. Environ. Saf. 89, 95–107.

rown, C.E., Brown, C.E., 1998. Applied Multivariate Statistics in Geohydrology andRelated Sciences. Springer, Berlin.

de 75 (2015) 197–205

Campos, V., 2002. Arsenic in groundwater affected by phosphate fertilizers at SaoPaulo, Brazil. Environ. Geol. 42 (1), 83–87.

Carbonell-Barrachina, Á.A., Signes-Pastor, A.J., Vázquez-Araújo, L., Burló, F.,Sengupta, B., 2009. Presence of arsenic in agricultural products from arsenic-endemic areas and strategies to reduce arsenic intake in rural villages. Mol. Nutr.Food Res. 53 (5), 531–541.

Chapagain, S., Shrestha, S., Nakamura, T., Pandey, V., Kazama, F., 2009. Arsenic occur-rence in groundwater of Kathmandu Valley, Nepal. Desalin. Water Treat. 4 (1–3),248–254.

Chapagain, S.K., Pandey, V.P., Shrestha, S., Nakamura, T., Kazama, F., 2010. Assess-ment of deep groundwater quality in Kathmandu Valley using multivariatestatistical techniques. Water Air Soil Pollut. 210 (1–4), 277–288.

Cloutier, V., Lefebvre, R., Therrien, R., Savard, M.M., 2008. Multivariate statisticalanalysis of geochemical data as indicative of the hydrogeochemical evolu-tion of groundwater in a sedimentary rock aquifer system. J. Hydrol. 353 (3),294–313.

Davis, J.C., Sampson, R.J., 2002. Statistics and Data Analysis in Geology, vol. 3. Wiley,New York.

Deng, Y.M., Wang, Y.X., Ma, T., 2009. Isotope and minor element geochemistry ofhigh arsenic groundwater from Hangjinhouqi, the Hetao plain, Inner Mongolia.Appl. Geochem. 24, 587–599.

Deng, Y., Wang, Y., Ma, T., Yang, H., He, J., 2011. Arsenic associations in sedimentsfrom shallow aquifers of northwestern Hetao Basin, Inner Mongolia. Environ.Earth Sci. 64 (8), 2001–2011.

Gibney, B.P., Nüsslein, K., 2007. Arsenic sequestration by nitrate respiring microbialcommunities in urban lake sediments. Chemosphere 70, 329–336.

Güler, C., Thyne, G.D., McCray, J.E., Turner, K.A., 2002. Evaluation of graphicaland multivariate statistical methods for classification of water chemistry data.Hydrogeol. J. 10 (4), 455–474.

Guo, H.M., Wen, D., Liu, Z., Jia, Y., Guo, Q., 2014a. A review of high arsenic groundwaterin Mainland and Taiwan, China: distribution, characteristics and geochemicalprocesses. Appl. Geochem. 41, 196–217.

Guo, H.M., Zhang, D., Wen, D.G., Wu, Y., Ni, P., Jiang, Y.X., Guo, Q., Li, F.L., Zheng,H., Zhou, Y.Z., 2014b. Arsenic mobilization in aquifers of the southwest Songnenbasin, P.R. China: evidences from chemical and isotopic characteristics. Sci. TotalEnviron. 490, 590–602.

Guo, H.M., Yang, S., Tang, X., Li, Y., Shen, Z., 2008. Groundwater geochemistry andits implications for arsenic mobilization in shallow aquifers of the Hetao Basin,Inner Mongolia. Sci. Total Environ. 393 (1), 131–144.

Guo, H.M., Zhang, B., Li, Y., Berner, Z., Tang, X.H., Norra, S., 2011. Hydroge-ological and biogeochemical constrains of arsenic mobilization in shallowaquifers from the Hetao basin, Inner Mongolia. Environ. Pollut. 159 (4),876–883.

Guo, H.M., Zhang, Y., Xing, L.N., Jia, Y.F., 2012. Spatial variation in arsenic and fluorideconcentrations of shallow groundwater from the Shahai town of the Hetao basin,Inner Mongolia. Appl. Geochem. 27, 2187–2196.

Hotelling, H., 1933. Analysis of a complex of statistical variables into principal com-ponents. J. Educ. Psychol. 24 (6), 417.

Hu, S., Luo, T., Jing, C., 2013. Principal component analysis of fluoride geochemistryof groundwater in Shanxi and Inner Mongolia, China. J. Geochem. Explor. 135,124–129.

Inner Mongolia Institute of Hydrogeology, 1982. Hydrogeological setting and reme-diation: Approaches of soil salinity in the Hetao basin, Inner Mongolia. Scientificreport (in Chinese).

Islam, F.S., Gault, A.G., Boothman, C., Polya, D.A., Charnock, J.M., Chatterjee, D., Lloyd,J.R., 2004. Role of metal-reducing bacteria in arsenic release from Bengal deltasediments. Nature 430 (6995), 68–71.

Jia, Y., Guo, H., Jiang, Y., Wu, Y., Zhou, Y., 2014. Hydrogeochemical zonation and itsimplication for arsenic mobilization in deep groundwaters near alluvial fans inthe Hetao Basin, Inner Mongolia. J. Hydrol. 518, 410–420.

Kim, K., Moon, J.-T., Kim, S.-H., Ko, K.-S., 2009. Importance of surface geologiccondition in regulating As concentration of groundwater in the alluvial plain.Chemosphere 77 (4), 478–484.

Lin, N.F., Tang, J., Bian, J.M., 2002. Characteristics of environmental geochemistry inthe arseniasis area of the Inner Mongolia of China. Environ. Geochem. Health 24(3), 249–259.

Lu, K.L., Liu, C.W., Jang, C.S., 2012. Using multivariate statistical methods to assess thegroundwater quality in an arsenic-contaminated area of Southwestern Taiwan.Environ. Monit. Assess. 184 (10), 6071–6085.

Luo, T., Hu, S., Cui, J., Tian, H.X., Jing, C.Y., 2012. Comparison of arsenic geochemicalevolution in the Datong Basin (Shanxi) and Hetao Basin (Inner Mongolia), China.Appl. Geochem. 27 (12), 2315–2323.

Mukherjee, A., von Brömssen, M., Scanlon, B.R., Bhattacharya, P., Fryar, A.E., Hasan,M.A., Ahmed, K.M., Chatterjee, D., Jacks, G., Sracek, O., 2008. Hydrogeochemi-cal comparison and effects of overlapping redox zones on groundwater arsenicnear the Western (Bhagirathi sub-basin, India) and Eastern (Meghna sub-basin, Bangladesh) margins of the Bengal Basin. J. Contam. Hydrol. 99 (1),31–48.

Mukherjee, A., Scanlon, B.R., Fryar, A.E., Saha, D., Ghosh, A., Chowdhuri, S.,Mishra, R., 2012. Solute chemistry and arsenic fate in aquifers between theHimalayan foothills and Indian craton (including central Gangetic plain):

influence of geology and geomorphology. Geochim. Cosmochim. Acta 90,283–302.

Neidhardt, H., Norra, S., Tang, X., Guo, H., Stüben, D., 2012. Impact of irrigation withhigh arsenic burdened groundwater on the soil–plant system: results from acase study in the Inner Mongolia, China. Environ. Pollut. 163, 8–13.

der Er

N

O

O

P

P

P

P

P

S

S

Y. Jiang et al. / Chemie

ickson, R.T., McArthur, J.M., Ravenscroft, P., Burgess, W.G., Ahmed, K.M., 2000.Mechanism of arsenic release to groundwater, Bangladesh and West Bengal.Appl. Geochem. 15, 403–413.

remland, R.S., Stolz, J.F., 2005. Arsenic, microbes and contaminated aquifers. TrendsMicrobiol. 13 (2), 45–49.

tero, N., Vitoria, L., Soler, A., Canals, A., 2005. Fertiliser characterisation: major,trace and rare earth elements. Appl. Geochem. 20 (8), 1473–1488.

anaullah, G.M., Alam, T., Hossain, M.B., Loeppert, R.H., Lauren, J.G., Meisner, C.A.,Ahmed, Z.U., Duxbury, J.M., 2009. Arsenic toxicity to rice (Oryza sativa L.) inBangladesh. Plant Soil 317 (1–2), 31–39.

ereira, H.G., Sousa, A., 1988. Análise de dados para o tratamento de quadros multi-dimensionais. Textos de apoio ao Curso Intensivo Análise Dados.

huong, N.M., Kang, Y., Sakurai, K., Sugihara, M., Kien, C.N., Bang, N.D., Ngoc, H.M.,2012. Arsenic contamination in groundwater and its possible sources in Hanam,Vietnam. Environ. Monit. Assess. 184 (7), 4501–4515.

ostma, D., Larsen, F., Minh Hue, N.T., Duc, M.T., Viet, P.H., Nhan, P.Q., Jessen, S.,2007. Arsenic in groundwater of the Red River floodplain, Vietnam: controllinggeochemical processes and reactive transport modeling. Geochim. Cosmochim.Acta 71 (21), 5054–5071.

urushotham, D., Prakash, M., Rao, A.N., 2011. Groundwater depletion and qualitydeterioration due to environmental impacts in Maheshwaram watershed of RR

district, AP (India). Environ. Earth Sci. 62 (8), 1707–1721.

enn, D.B., Hemond, H.F., 2002. Lake nitrate controls on iron and arsenic in an urbanlake. Science 296, 2373–2376.

medley, P., Kinniburgh, D., 2002. A review of the source, behaviour and distributionof arsenic in natural waters. Appl. Geochem. 17 (5), 517–568.

de 75 (2015) 197–205 205

Stumm, W., 1992. Chemistry of the Solid–water Interface: Processes at theMineral-water and Particle-water Interface in Natural Systems. John Wiley& Sons.

Tong, J.T., Wei, C., Guo, H.M., 2013. Detection and health risk of arsenic species incrops from arsenic-affected areas of Hetao Plain, InnerMongolia. Asian J. Eco-toxicol. 8 (3), 426–434 (in Chinese with English Abstract).

Tong, J.T., Guo, H.M., Wei, C., 2014. Arsenic contamination of the soil-wheat systemthrough irrigation with high arsenic groundwater in Inner Mongolia, China. Sci.Total Environ. 496, 479–487.

Wang, S.-W., Liu, C.-W., Lu, K.-L., Lin, L.-H., 2012a. Biogeochemical cycling offerric oxyhydroxide affecting As partition in groundwater aquitard. Environ.Geochem. Health 34 (4), 467–479.

Wang, Y., Jiao, J.J., Cherry, J.A., 2012b. Occurrence and geochemical behavior ofarsenic in a coastal aquifer–aquitard system of the Pearl River Delta, China. Sci.Total Environ. 427, 286–297.

Yang, S.Z., Guo, H.M., Tang, X.H., Shen, Z.L., 2008. Distribution of abnormal ground-water arsenic in Hetao Plain, Inner Mongolia. Earth Sci. Front. 15 (1), 242–249(in Chinese with English Abstract).

Yidana, S.M., 2010. Groundwater classification using multivariate statistical meth-ods: Southern Ghana. J. Afr. Earth Sci. 57 (5), 455–469.

Yidana, S.M., Banoeng-Yakubo, B., Akabzaa, T.M., 2010. Analysis of groundwater

quality using multivariate and spatial analyses in the Keta basin, Ghana. J. Afr.Earth Sci. 58 (2), 220–234.

Zahid, A., Hassan, M.Q., Balke, K.-D., Flegr, M., Clark, D.W., 2008. Groundwater chem-istry and occurrence of arsenic in the Meghna floodplain aquifer, southeasternBangladesh. Environ. Geol. 54 (6), 1247–1260.