Science of the Total Environment 484 (2014) 27–35

Contents lists available at ScienceDirect

Science of the Total Environment

j ourna l homepage: www.e lsev ie r .com/ locate /sc i totenv

Metal contamination in campus dust of Xi'an, China: A study based onmultivariate statistics and spatial distribution

Hao Chen a, Xinwei Lu a,⁎, Loretta Y. Li b,⁎⁎, Tianning Gao a, Yuyu Chang a

a School of Tourism and Environment, Shaanxi Normal University, Xi'an 710062, PR Chinab Department of Civil Engineering, University of British Columbia, Vancouver V6T 1Z4, Canada

H I G H L I G H T S

• Metal content in dust from schools was determined by XRF.• Spatial distribution of metals in urban dust was focused on campus samples.• Multivariate statistic and spatial distribution were used to identify metal sources.• Pb, Zn, Cu, Co and Cr mainly originate from anthropogenic sources.• As, Mn, Ni, V and Ba were mainly influenced by natural sources.

⁎ Corresponding author. Tel.: +86 29 85310525; fax: +⁎⁎ Corresponding author. Tel.: +1 604 822 1820; fax: +

E-mail addresses: luxinwei@snnu.edu.cn (X. Lu), lli@c

http://dx.doi.org/10.1016/j.scitotenv.2014.03.0260048-9697/© 2014 Elsevier B.V. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 21 December 2013Received in revised form 8 March 2014Accepted 8 March 2014Available online 28 March 2014

Editor: Xuexi Tie

Keywords:MetalsMultivariate statisticsSpatial analysisSourceDust

The concentrations of As, Ba, Co, Cr, Cu, Mn, Ni, Pb, V and Zn in campus dust from kindergartens, elementaryschools, middle schools and universities of Xi'an, China were determined by X-ray fluorescence spectrometry.Correlation coefficient analysis, principal component analysis (PCA) and cluster analysis (CA) were used toanalyze the data and to identify possible sources of these metals in the dust. The spatial distributions of metalsin urban dust of Xi'an were analyzed based on the metal concentrations in campus dusts using the geostatisticsmethod. The results indicate that dust samples from campuses have elevatedmetal concentrations, especially forPb, Zn, Co, Cu, Cr and Ba, with the mean values of 7.1, 5.6, 3.7, 2.9, 2.5 and 1.9 times the background values forShaanxi soil, respectively. The enrichment factor results indicate that Mn, Ni, V, As and Ba in the campus dustwere deficiently to minimally enriched, mainly affected by nature and partly by anthropogenic sources, whileCo, Cr, Cu, Pb and Zn in the campus dust and especially Pb and Zn were mostly affected by human activities. Asand Cu, Mn and Ni, Ba and V, and Pb and Zn had similar distribution patterns. The southwest high-tech industrialarea and south commercial and residential areas have relatively high levels of most metals. Three main sourceswere identified based on correlation coefficient analysis, PCA, CA, as well as spatial distribution characteristics.As, Ni, Cu,Mn, Pb, Zn and Cr havemixed sources— nature, traffic, aswell as fossil fuel combustion andweatheringof materials. Ba and V are mainly derived from nature, but partly also from industrial emissions, as well asconstruction sources, while Co principally originates from construction.

© 2014 Elsevier B.V. All rights reserved.

1. Introduction

Atmospheric pollution constitutes amajor challenge inmany dense-ly populated cities inmany countries, in particular those under rapid in-dustrialization and urbanization which face poor air quality and heavydust deposition (Hien et al., 1999; Tanner et al., 2008; Schleicher et al.,2011). Dust, containing trace metals, is released to the atmosphereduring combustion of fossil fuels and wood, as well as from high-

86 29 85303883.1 604 822 6901.ivil.ubc.ca (L.Y. Li).

temperature industrial processes, waste incineration and traffic (Allenet al., 2001; Thakur et al., 2004). Dust particulates have beenwell recog-nized to influence human health (Li et al., 2008; Ruiz-Jimenez et al.,2012), in particular due to trace metals (Chillrud et al., 2004; Khairyet al., 2011; Lu et al., 2014)which are toxic to humans through ingestionor inhalation. For example, As, Cd and their chemical compounds arehighly carcinogenic, while low concentrations of Pb in blood can affectchildren's mental development, an effect that persists into adulthood(Needleman, 1990; Laidlaw and Tayor, 2011). Mn is considered toxicif taken up through inhalation, causingmovement disorders, respiratoryeffects and reproductive dysfunction (WHO, 2000; USEPA, 2003).Although Zn is an essential nutrient for human organs, it is toxic athigh concentrations (Adamson et al., 2000).

28 H. Chen et al. / Science of the Total Environment 484 (2014) 27–35

Numerous studies on street dust have been conducted onmetal con-centrations, distribution, and source identification in the past decade(Bennett et al., 2006; Tanner et al, 2008; Lu et al., 2010; Laidlaw andTayor, 2011; Glorennec et al., 2012). While there has been some recentinformation related to dust in the workplace and in residential houses(Abdul-Wahab and Yaghi, 2004), very few studies have been reportedwithin sensitive environments such as nursery schools (Tong andLam, 1998; Lu et al., 2014). Xi'an, the biggest city in northwesternChina, has experienced rapid urbanization and industrialization inrecent decades causing metal contamination in urban soil and streetdust (Han et al., 2008; Chen et al., 2012). Despite these serious effectson health development, especially for children and young adults, stud-ies in these areas are lacking, and information about metal contamina-tion in the academic urban environment is limited. Our work wascarried out to assess pollution of metals in dust sampled from a widerange of educational campuses including kindergartens, elementaryschools, middle schools and universities of Xi'an. The main objectiveswere to determine the campus concentrations of Cu, Pb, Zn, As, Ba, Co,Cr, Mn, Ni and V (metals which are potentially harmful to the environ-ment and human health); to investigate the spatial distribution ofmetals in urban dust of Xi'an; and to identify the sources of metals incampus dust based on multivariate statistical methods and spatialanalysis.

2. Materials and methods

2.1. Area Description

Xi'an, the capital of Shaanxi province, is located in the central WeiRiver valley (107°40′–109°49′E and 33°39′–34°45′N). It has a typicaltemperate continental semi-arid climate, with a monthly average tem-perature of 0.9 °C in January, 26.4 °C in July and an annual average tem-perature of ~13 °C (Cao et al., 2011). Xi'an is located in a loess plateauthat is the major source of Asian dust (Zhang et al., 2001), with seriouscontamination from airborne particulate matter (PM), especially in thespring when frequent dust storms occur (Han et al., 2008). Xi'an has ahistory of more than two thousand years. With its rapid industrializa-tion, urbanization and high-tech development in recent decades, thiscity is the center of the economy, culture, manufacturing and educationin northwest China (Chen et al., 2012). The total urban area of Xi'an cityis ~3580 km2, and its population was 6,470,000 in 2009. The motorizedvehicle density has grown from 0.52 million in 2004 to 0.98 million in

Fig. 1. Sketch indicating sa

2009 (XAMBS, 2011). Presently, Xi'an has 1004 kindergartens, 1531 el-ementary schools, 660 middle schools and 50 universities, hosting700,000 nursery and primary school students, 779,000 middle-schoolstudents and 733,000 university students (XAMBS, 2011).

2.2. Dust sampling and analytical method

A total of 157 kindergartens, primary schools, middle schools anduniversities of Xi'an were selected for the collection of dust samples(Fig. 1). Sampling was conducted during the dry season betweenOctober 2011 andOctober 2012. Dust samples in each campuswere col-lected in the same dry season by sweeping, using a clean plastic dustpanand a brush (Akhter and Madany, 1993; Lu et al., 2009) from windowsills, balconies, doorsteps and school playgrounds most accessible tostudents. The collected dust sampleswere stored in sealed polyethylenebags for transport and storage. All samples were air-dried in the labora-tory for two weeks and then sieved through a 1.0 mm nylon mesh toremove large particles such as tree leaves, refuse and small stones,before splitting the samples by halving and mixing. 50 g of each sieveddust sample was quartered and then ground with a vibration mill andsieved through a 0.075 mm nylon mesh.

X-ray fluorescence (XRF) samples were prepared by weighing 4.0 gof milled dust sample and 2.0 g of boric acid, placed in a mold, andpressed into a 32 mm diameter pellet under 30 t pressure. The concen-trations of As, Ba, Co, Cr, Cu, Mn, Ni, Pb, V and Zn in dust samples werethen measured by wavelength dispersive X-ray fluorescence spectrom-etry (XRF, PANalytical PW2403 apparatus) (Lu et al., 2010).

For quality assurance and control (QA/QC), duplicate samples andstandard reference materials (GSS1 and GSD12) (Lu et al., 2010),purchased from the Center of National Standard Reference Material ofChina, were prepared and analyzed using the same procedure. Theprecision, calculated from the relative standard deviation of duplicatesamples, was routinely 3–5%. The accuracy, based on the relative errorof standard reference materials, was b5%.

2.3. Outlier detection

Outliers, often observed in environmental geochemical datasets(Zhang et al., 1998), arise due to human error, instrument error, naturaldeviations in populations, fraudulent behavior, changes in behavior ofsystems or faults (Hodge and Austin, 2004). Outliers are dealt withbefore statistical analysis by the most common method — the range

mpling sites in Xi'an.

29H. Chen et al. / Science of the Total Environment 484 (2014) 27–35

method: Values lower than the average values minus three standarddeviations and those higher than the average plus three times the stan-dard deviation are considered to be outliers. Here, the outlying valueswere deleted and then replaced by the highest values in datasets afterdeletion (Zhang et al., 1998).

2.4. Enrichment factor analysis

To assess the degree of anthropogenic influence, enrichment factors(EFs) were calculated. They provide a measure of the extent to whichtrace elements are enriched or reduced relative to a specific source(Odabasi et al., 2002). They can be used to differentiate betweenmetalsoriginating from human activities and those from natural sources(Meza-Figueroa et al., 2007). The EF of an element in a studied sampleis based on the standardization of a measured element against a refer-ence element. The reference element is often characterized by lowoccurrence variability, with the most commonly used elements beingAl, Fe, Ti, Si, Sr, and K (Tasdemir and Kural, 2005; Han et al., 2006;Turner and Simmonds, 2006; Hao et al., 2007; Meza-Figueroa et al.,2007). EF is calculated by

EF ¼ Cx=Cref

h isample

= Cx=Cref

h iBackground

ð1Þ

where Cx is the concentration of the element of interest and Cref is theconcentration of the reference element for normalization. EF analysiscan assist in differentiating anthropogenic sources from natural ones.A value of EF close to 1 indicates natural origin, whereas values N10are considered to originate mainly from anthropogenic sources (Hanet al., 2006; Turner and Simmonds, 2006). EF analysis can also assist indetermining the degree of metal contamination (Meza-Figueroa et al.,2007). EF≤ 2 means deficient to minimal enrichment, 2 b EF≤ 5 corre-sponds tomoderate enrichment, 5 b EF≤ 20 signifies significant enrich-ment, 20 b EF ≤ 40 indicates very high enrichment, and 40 b EFmeansextremely high enrichment (Lu et al., 2009).

2.5. Statistical analysis

Experimental data were analyzed by correlation coefficient analysis,principal component analysis (PCA) and cluster analysis (CA) to deter-mine the relationship among the metals investigated in the dusts.They are widely used in environmental studies (Han et al., 2006;Tokalıoğlu and Kartal, 2006; Lu et al., 2010). Pearson's correlation coef-ficient, a measure of the strength of a linear relationship between twoquantitative variables (Tokalıoğlu and Kartal, 2006), was applied.

PCA is widely used to reduce the data and to extract a small numberof latent factors (principal components, PCs) for analyzing the relation-ships among the observed variables (Han et al., 2006). In PCA, the prin-cipal components are calculated based on the correlation matrix.Varimax with Kaiser Normalization was used as the rotation method(Lee et al., 2006). PCA can reduce the number of correlated variablesto a smaller set of orthogonal factors, making it easier to interpret a

Table 1Metal concentrations in campus dust of Xi'an and reference values (mg kg−1).

Element As Ba Co Cr

Max 29.7 2195.9 81.1 402.4Min 1.4 542.7 19.3 77.4Mean 11.5 958.9 39.6 154.2GM 10.2 914.6 37.5 145.4Median 10.3 869.5 37.6 134.4SD 5.8 336.8 13.4 64.0CV(%) 50 35 34 41Skewness 1.2 2.0 0.8 2.5Reference valuea 11.1 516.0 10.6 62.5

a CNEMC (1990).

given multidimensional system by displaying the correlations amongthe original variables (Lu et al., 2010). PCA was applied in this study toidentify the possible sources of metals in the dust by applying varimaxrotation with Kaiser Normalization (Lu et al., 2010). The Kaiser–Meyer–Olkin (KMO)measure of sampling adequacy is used to comparethemagnitudes of the observed correlation coefficients with themagni-tudes of the partial correlation coefficients. Large KMO (0–1) values areuseful because correlations between pairs of variables can be explainedby the other variables (Yuan et al., 2014). Bartlett's test of sphericity isused to test the hypothesis that the correlation matrix is an identitymatrix with suitable significance (P b 0.05) (Yuan et al., 2014). In thisstudy, the KMO (0.747) and Bartlett's test (P b 0.001) show that themeasured metal concentrations for campus dust from Xi'an are suitablefor PCA (Chen et al., 2012; Yuan et al., 2014).

CA is often coupled to PCA to check results and to help group individ-ual parameters and variables (Facchinelli et al., 2001). CA was devel-oped according to Ward's method. Euclidean distance was employedto measure the distance between clusters of similar metal contents.The purpose of CA is to organize observations where a number ofgroups/variables share observed properties (Lu et al., 2010). Statisticalanalyses were performed with SPSS 19.0 for Windows.

2.6. Spatial analysis

The spatial distributions of metals in urban dust of Xi'an were ana-lyzed based on the metal concentrations in campus dusts using thegeostatistics method. Geostatistics is a tool for studying and predictingthe spatial structure of geo-referenced variables, focused on spatialobjects and spatial correlation (Chen et al., 2008). Kriging is based onthe assumption that the parameter being interpolated can be treatedas a regionalized variable (Xie et al., 2011). It is regarded as the best spa-tial covariance interpolation method, providing optimal interpolation(Chen et al., 2008). The spatial distribution maps of all metals studiedin Xi'an urban dust were generated by the Kriging interpolation ofgeostatistics method with ArcGIS software.

3. Results and discussion

3.1. Metal concentrations in campus dust

The descriptive statistical results of metal concentrations in campusdust of Xi'an after outlier treatment, as well as background values forShaanxi soil (CNEMC, 1990), are shown in Table 1. This table showsthat the arithmetic means of As, Ba, Co, Cr, Cu, Mn, Ni, Pb, V and Zn inthe studied dust samples are 11.5, 958.9, 39.6, 154.2, 62.1, 546.2, 32.2,151.6, 68.7 and 390.7 mg kg−1, respectively. Compared with the back-ground values of Shaanxi soil, the arithmetic means, geometric meansand medians of all analyzed metals in campus dust are clearly higherthan the corresponding background values of Shaanxi soil, except forAs, Mn, Ni and V. The maximum concentrations of Cu, Pb, Zn, Cr, Coand Ba in the dust are 6.5, 23.1, 26.5, 6.4, 7.7 and 4.3 times the

Cu Mn Ni Pb V Zn

138.3 795.8 64.2 494.1 99.3 1838.322.3 349.5 16.8 37.2 50.2 65.962.1 546.2 32.2 151.6 68.7 390.757.4 538.9 31.2 131.8 68.0 319.559.2 527.6 31.5 132.4 67.6 319.124.9 94.7 8.2 93.2 10.0 299.340 17 26 61 15 771.0 0.7 0.6 2.1 0.9 2.8

21.4 557 28.8 21.4 66.9 69.4

30 H. Chen et al. / Science of the Total Environment 484 (2014) 27–35

background values of Shaanxi soil, respectively, and their mean valuesin the dust are 2.9, 7.1, 5.6, 2.5, 3.7 and 1.9 times the correspondingbackground values, respectively. The mean values of metals in campusdust divided by the corresponding background value of Shaanxi soil de-crease in the order Pb N Zn N Co N Cu N Cr N Ba N Ni N As N V NMn. Therange and themean of Cu, Pb, Zn, Cr andMn in the studied dust samplesare all lower than their ranges andmeans in the road dust of Xi'an (Hanet al., 2008), while the concentration of As in the campus dust is similarto that in the road dust (Han et al., 2008). This may be because mostschools are far from main streets.

3.2. Enrichment factor analysis results

Enrichment factors (EFs) of all analyzed metals were calculated foreach dust sample relative to the background value of the local soil(CNEMC, 1990), with Al chosen as the reference element (Turner andSimmonds, 2006). Fig. 2 shows the cumulative frequency distributionanalysis of EF of metals in the dust. The EF values of As, Ba, Co, Cr, Cu,Mn, Ni, Pb, V and Zn are in the ranges of 0.09–2.35, 0.60–3.73,1.29–5.37, 0.97–5.26, 0.75–5.28, 0.40–1.17, 0.40–1.82, 0.88–18.98,0.52–1.12 and 0.48–23.22, with averages of 0.74, 1.32, 2.66, 1.76, 2.08,0.70, 0.80, 5.08, 0.73 and 4.05, respectively. EF of Mn, Ni and V in alldust samples, As in 98% of the samples and Ba in 92% of the dust samplesare b2, showing the lack of pollution of these metals in Xi'an campusdust. The mean EF and the finding of 73% of EF values for Co in therange of 2–5 reveal that Co of campus dust corresponds to moderatepollution. For Cr 83% of EF values and 16% of EF values were b2 and inthe range 2–5, respectively, with amean EF b 2, indicating that Cr corre-sponds to deficient to moderate pollution. The mean EF and 46% of EFvalues of Cu are in the 2–5 range, indicating moderate pollution, while53% of EF values are b2 showing deficient to minimal enrichment. Pbhas 57% EF and 37% EF in the ranges of 2–5 and 5–20, respectively,with a mean EF N 5, indicating moderate to significant pollution. Thevariation of EF is comparatively larger for Zn than for the other metalsanalyzed. The mean EF and 57% of EF values for Zn are in the range 2–5 showing moderate pollution, while 22% of EF values were N5 and21% b2, indicating significant pollution and deficient tominimal enrich-ment respectively. The above results indicate that Mn, Ni and V in thecampus dust of Xi'an were affected by natural sources, Ba and As weremainly caused by nature and partly influenced by human activities,while Co, Cr, Cu, Pb and Zn in the campus dust and especially the lattertwo, were mostly affected by human activities. Compared with the EFvalues of Cu, Pb, Zn, Cr and As in street dust of Xi'an, reported by Hanet al. (2006), the mean EF and maximum EF values of Cu, Pb, Zn andCr in the campus dust are well below their levels in street dust ofXi'an, implying that their enrichment degrees and contamination levelsin campus dust are lower than in the Xi'an street dust.

3.3. Correlation coefficient analysis results

Pearson's correlation coefficients of metals in the campus dust arepresented in Table 2. A significantly positive correlation at the P b 0.01significance level was found for As, Ba, Cu, Mn, Ni, Pb and Zn. V is signif-icantly positively correlated with As, Ba, Mn, Ni and Zn at P b 0.01whereas, Cr is significantly positively correlated with Cu, Mn and Ni atP b 0.01, and is positively correlated with Pb at P b 0.05. Co is only pos-itively correlated with Ba and negatively correlatedwith Pb at a P b 0.05significance level. Inter-element relationships provide interesting infor-mation on the sources and pathways of the metals (Lu et al., 2010).

3.4. Multivariate statistical analysis

In order to gain some insight into the sources of metals and majorcorrelations among them, PCA and CA were applied to each set ofdata. PCA results shown in Table 3 display factor loadings with avarimax rotation, as well as eigenvalues and commonalities calculated

using the software package SPSS 19.0. The results show that there arefive eigenvalues N1 and that these five factors explain 80% of the totalvariance. The first factor explains 24.4% of the total variance, heavilyloaded on As, Cu, Mn and Ni. Factor 2 is loaded primarily by Ba and V,and also moderately by Mn, accounting for 17.6% of the total variance.The Mn loading (0.442) is not as high as for V and Ba (0.878 and0.814, respectively), which may imply quasi-independent behaviorwithin the group (Lu et al., 2010). Factor 3, dominated by Pb and Zn,accounts for 14.3% of the total variance. Factors 4 and 5 explain 12.5%and 11.2% of the total variance and loading of Cr and Co, respectively.

Results for CA are shown in Fig. 3 as a dendrogram. The metal con-centration data were standardized by means of z-scores before CA andEuclidean distances for similarities in the variables were calculated(Tokalıoğlu and Kartal, 2006; Lu et al., 2010). The hierarchical clusteringby applying Ward's method was then performed on the standardizeddataset. Fig. 3 displays five clusters: (1) As–Ni–Cu–Mn; (2) Pb–Zn; (3)Cr; (4) Ba–V; (5) Co, which is in close agreement with the PCA results.It is observed, however, that Clusters 1, 2 and 3 join together at a rela-tively higher level, implying a possible common source.

3.5. Spatial distribution of metals in urban dust

The spatial distributions of all analyzedmetals in urban dust of Xi'anare presented in Fig. 4. Fig. 4a shows that the concentrations of As in theurban dusts from the southwest and the middle zones extending fromnorth to south of the study area were 12.0 to 29.7 mg kg−1, 1.1 to 2.7times the backgroundvalue of local soil, while in other areas, As concen-trations were lower than, or close to, the background value. The south-west high-value area is a high-tech industrial area, and themiddle high-value area includes commercial centers and residential areas. Thesouthwest high-tech industrial area was built during the 1990s. Beforethe 1990s, that area consisted of agricultural land and a village. Thehigh-value areas of Ba concentrations (N2 times the backgroundvalue) were found in the southwest, south and east of the study area(Fig. 4b). The east high-value area is an old industrial zone of Xi'anwhere steel mills, railway signal factories and machinery plants arelocated.

The concentrations of Co (19.3 to 81.1 mg kg−1) in the urban dustare significantly higher than the background value of Shaanxi soil(10.6 mg kg−1), with the highest concentration of Co (N4.1 times thebackground value) in the southwest high-tech industrial area andsoutheast Qujiang new district (Fig. 4c). Qujiang new district, builtafter 2000, is designed as a tourism, cultural, commercial and residentialzone, with many construction sites. Before the 2000s, this area was alsoagricultural land and a village. Fig. 4d indicates that the higher concen-tration areas of Cr (N149.3 mg kg−1 = 2.4 times the background valueof Shaanxi soil) are located in the north between the inner ring road andsecond ring road and the south area around the second ring road. A busstation and auto repair factories are located in the north high-value area.The south high-value area is a commercial and residential area, withdense roads and heavy traffic. The spatial distribution of Cu in theurban dust (Fig. 4e) is similar to that of As (Fig. 4a). The concentrationsof Cu in the dusts of the southwest and the middle zones, extendingfrom north to south of the study area, are N3.1 times the backgroundvalue of Shaanxi soil, while in other areas they are 1.0–3.1 times thebackground value.

The concentrations of Mn in the dusts of the south and southwestare higher than in other areas (Fig. 4f). The Ni concentrations in thedusts of most study areas are lower than, or close to, the backgroundlevel for Shaanxi soil. Ni high-value areas (N1.2 times the backgroundvalue, Fig. 4g) are similar to those for Mn (Fig. 4f). The concentrationsof Pb in the dusts of the study area are significantly higher than its back-ground value. There is one high-value zone (N143.4 mg kg−1) in themiddle region, extending from north to south of the study area, espe-cially in the north between the inner ring road and second ring roadand in the south around the second ring road, the concentrations of Pb

Fig. 2. Cumulative frequency distribution of EF of metals in campus dust.

31H. Chen et al. / Science of the Total Environment 484 (2014) 27–35

Table 2Pearson's correlation matrix for metal concentrations.

As Ba Co Cr Cu Mn Ni Pb V Zn

As 0.004 0.773 0.052 0.000 0.000 0.000 0.000 0.000 0.000Ba 0.229⁎⁎ 0.045 0.170 0.001 0.000 0.001 0.002 0.000 0.000Co −0.023 0.160⁎ 0.840 0.859 0.617 0.780 0.023 0.453 0.561Cr 0.155 0.110 −0.016 0.000 0.000 0.000 0.013 0.330 0.291Cu 0.579⁎⁎ 0.256⁎⁎ 0.014 0.354⁎⁎ 0.000 0.000 0.000 0.066 0.000Mn 0.409⁎⁎ 0.385⁎⁎ −0.040 0.321⁎⁎ 0.528⁎⁎ 0.000 0.000 0.000 0.001Ni 0.630⁎⁎ 0.252⁎⁎ −0.023 0.369⁎⁎ 0.653⁎⁎ 0.548⁎⁎ 0.000 0.000 0.000Pb 0.356⁎⁎ 0.251⁎⁎ −0.182⁎ 0.197⁎ 0.447⁎⁎ 0.368⁎⁎ 0.450⁎⁎ 0.002 0.000V 0.332⁎⁎ 0.573⁎⁎ −0.060 0.078 0.147 0.407⁎⁎ 0.472⁎⁎ 0.241⁎⁎ 0.005Zn 0.465⁎⁎ 0.350⁎⁎ 0.047 0.085 0.348⁎⁎ 0.269⁎⁎ 0.322⁎⁎ 0.396⁎⁎ 0.255⁎⁎

The left lower part is correlation coefficient; the right upper part is significant level.⁎⁎ Correlation is significant at P b 0.01 level (2-tailed).⁎ Correlation is significant at P b 0.05 level (2-tailed).

32 H. Chen et al. / Science of the Total Environment 484 (2014) 27–35

were 177.6 to 494.1 mg kg−1, 8.3 to 23.1 times the background level(Fig. 4h). The concentrations of V in the dusts of most area are lessthan, or close to, the background value of Shaanxi soil, with its high-value areas (1.1–1.5 times the background value, Fig. 4i) similar tothose for Ba (Fig. 4b). It can be seen from Fig. 4j that the high-valuearea (N386.6 mg kg−1) of Zn is located in the southwest to the middlewhere there is a high-tech industrial and commercial area with denseroads and heavy traffic.

3.6. Metal source identification

Three main sources can be identified according to correlation coeffi-cient analysis, PCA, CA and the spatial distribution characteristics ofmetals in the dust: (1) As, Ni, Cu, Mn, Pb, Zn and Cr have mixed sourcesof nature, traffic, aswell as fossil fuel combustion andweathering ofma-terials; (2) Ba and V mainly originate from natural sources, but partlyalso from industrial activities, as well as construction; (3) Co comesfrom construction.

As, Ni, Cu,Mn, Pb, Zn and Cr have significantly positive correlation inthe correlation coefficient analysis. As, Ni, Cu andMn belong to Factor 1in PCA and are classified together in CA (Cluster 1). Pb and Zn belong toanother principal component (Factor 3) in PCA and are classified to-gether in CA (Cluster 2). Cr belongs to Factor 4 in PCA and Cluster 3 inCA. Elements in the As, Ni, Cu and Mn cluster, elements in the Pb andZn cluster, and cluster Cr join together at a relatively higher level inCA, implying a possible common source. The coefficients of variationof these metals are relatively large except for Mn, revealing the influ-ence of human activity on their concentrations. The concentrations ofAs,Mn andNi inmost dust samples are lower than, or close to, their cor-responding background levels, indicating that they largely originatednaturally (from local soil). The higher-value (N1.2 times the backgroundvalue) areas for As, Mn and Ni are distributed in a similar manner, i.e.

Table 3Rotated componentmatrix for data of Xi'an campus dust (PCA loadings N0.4 are shown inbold).

Element Component Commonalities

1 2 3 4 5

As 0.845 0.115 0.267 −0.117 0.007 0.812Ba 0.015 0.814 0.334 0.114 0.220 0.837Co 0.005 0.029 −0.022 0.010 0.956 0.915Cr 0.169 0.023 0.024 0.908 0.008 0.855Cu 0.742 0.004 0.102 0.345 0.057 0.759Mn 0.515 0.442 0.102 0.386 −0.069 0.625Ni 0.816 0.268 0.100 0.247 −0.070 0.813Pb 0.269 0.132 0.677 0.249 −0.330 0.719V 0.275 0.878 −0.010 −0.061 −0.135 0.868Zn 0.260 0.149 0.828 −0.084 0.130 0.799Eigenvalue 2.44 1.76 1.43 1.25 1.12% of variance 24.4 17.6 14.3 12.5 11.2% of cumulative 24.4 42.0 56.3 68.8 80.0

centered in the southwest high-tech industrial area and the southcommercial and residential areas.

The pre-1990 agricultural activities (sewage irrigation, chemicalpesticide usage, etc.) in the southwest high-tech industrial area causedmetal accumulation in the local soil, which is the main source of thehigh-value of As, Mn and Ni concentrations in this area's dust. Thehigher concentrations of As,Mn andNi in the south commercial and res-idential areasmainly relate to fossil fuel combustion. The concentrationsof Cu, Pb and Zn exceed 1.5 times the corresponding background valuein 95% of dust samples. Their higher-value areas are concentrated in thesouthwest high-tech industrial area and the middle part where thereare dense roads andheavy traffic. Zinc alloy and galvanized componentsare widely used in motor vehicles. Zinc compounds are also employedextensively as antioxidants and as detergent/dispersant improvers forlubricating oils (De Miguel et al., 1997). Zn, added to tires during thevulcanizing process, comprises from 0.4% to 4.3% of the resulting tiretread (Chen et al., 2012). The wear and tear of vulcanized vehicle tiresand corrosion of galvanized automobile parts are the main sources ofZn in urban environments (Han et al., 2006; Lu et al., 2010).

Although the use of leaded petrol has been banned in Xi'an since2000, the content of Pb in urban soil still reflects the significant degreeof historical Pb contamination and the long half-life of Pb in soil (Yanget al., 2011). Pb in the bare soil could enter the urban dust by resuspen-sion. In addition, Pb contained in paints and coatings of buildings andsome urban facilities could enter the urban dust due to the effects of

Fig. 3. Dendrogram results from Ward's method of hierarchical cluster analysis for 10elements.

33H. Chen et al. / Science of the Total Environment 484 (2014) 27–35

weather (rain, sun, etc.) on the buildings and urban facilities (Chenet al., 2012).

Copper alloy is a material used in mechanical parts due to its desir-able qualities, such as corrosion resistance and strength (Chen et al.,2012). Cu is also used in Cu–brass automotive radiators due to its highcorrosion resistance and high thermal conductivity (Yang et al., 2011).It is also often used in car lubricants (Lu et al., 2010). The deteriorationof themechanical parts in vehicles over time results in Cu being emittedto the surrounding environment (Li et al., 2004). Oxidation of lubricat-ing oils upon exposure to air at high temperature results in the forma-tion of organic compounds which are corrosive to metal (Chen et al.,2012). Cu can be released to the urban environment as a result ofwear of automobile oil pumps or corrosion of metal parts which comeinto contact with the oil (Lu et al., 2010).

The concentrations of Cr in the dust are significantly higher than thebackground value of Shaanxi soil and its coefficient of variation is great-er, showing that Cr concentration in the dust was mainly caused byhuman activities. Elevated concentrations of Cr were reported in the

Fig. 4. Spatial distribution of metal con

dusts collected around the coal-fired power plant (Ren et al., 2011). Cris extensively used in automobile parts, aluminum alloys and titaniumalloys. Considering the concentration, coefficient of variation andhigh-value area distribution, one can conclude that Cr in the dustmainlyoriginates from traffic emissions and fossil fuel combustion.

The second group of elements consisting of Ba and V is stronglypositively correlated in PCA and correlation coefficient analysis, and isclassified together in CA. The coefficient of variation of V concentrationin the dusts is smaller and its concentrations in the dusts from mostareas are lower than, or close to, the background value for Shaanxisoil, indicating that V in the urban dust mainly originated from localsoil. V is often incorporated in stainless steel and alloys. The high-value area distribution of V concentration indicates that V in the dustpartly originated from industrial sources. The enrichment factor resultsindicate that Ba was mainly affected by nature, but also partly influ-enced by human activities. Ba is widely used in alloys, paints, ceramics,plastic cements, and glass (Monaci and Bargagli, 1997). From the spatialdistribution characteristics of Ba concentration in the dust, it can be

centrations in urban dust of Xi'an.

Fig. 4 (continued).

34 H. Chen et al. / Science of the Total Environment 484 (2014) 27–35

concluded that Ba in the dust mainly originated from natural sources,but with some contribution from industrial emissions, as well as con-struction. Element cluster Ba andV, and element cluster Co join togetherat a relatively higher level in CA, indicating Ba and V in the dust partlyoriginated from construction.

The third group of element is Co. The coefficient of variation of Co isrelatively large, and its concentrations in the urban dust are clearlyhigher than the corresponding background value for Shaanxi soil, dem-onstrating that Co in the urban dust ismainly governed by human activ-ities. Co is extensively used in coating materials, paints and pigments.These Co-containing materials are widely used in modern buildingsdue to their gloss, faultless color and visual impact. This is true for thesouthwest high-tech industrial area and southeast Qujiang new districtconstruction the high-value areas of dust Co concentration. This sug-gests that Co in urban dust of Xi'an predominantly originated frombuilding construction or renovation, leading to weathering and corro-sion of building materials.

4. Conclusions

The content measurement results of As, Ba, Co, Cr, Cu, Mn, Ni, Pb, Vand Zn show that kindergarten, elementary school, middle school anduniversity dusts of Xi'an have elevated metal concentrations, especiallyof Pb, Zn, Co, Cu, Cr and Ba, these being 1.7–23.1, 0.9–26.5, 1.8–7.7, 1.0–6.5, 1.2–6.4 and 1.1–4.3 times the corresponding background values forShaanxi soil, respectively. Themean concentrations ofmetals in thedustdivided by the corresponding background Shaanxi soil values decreasein the order Pb N Zn N Co N Cu N Cr N Ba NAs NNi NV NMn. Enrichmentfactor analysis indicates thatMn, Ni, V, As andBa in campus dust of Xi'anwere deficient as a whole, while Co, Cr, Cu, Pb and Zn in the campusdust, especially Pb and Zn,weremostly caused by human activities. Spa-tial distributions of metals in Xi'an urban dust were determined basedon the campus dust samples using the geostatistics method. Differentdistribution patterns were found among the investigated metals. Asand Cu,Mn andNi, Ba andV, Pb andZn had similar distribution patterns.

35H. Chen et al. / Science of the Total Environment 484 (2014) 27–35

The southwest high-tech industrial area and south commercial andresidential areas were high-value areas for most metals. Three mainsources of metals in urban dust of Xi'an were identified according tocorrelation coefficient analysis, PCA and CA, coupled with the spatialdistribution characteristics of metals. As, Ni, Cu, Mn, Pb, Zn and Crhavemixed sources fromnature, traffic, aswell as fossil fuel combustionand weathering of materials. Ba and V are mainly natural in origin, butalso partly from industrial emissions, as well as construction, while Coprincipally arises from construction.

Conflict of interest

Work on this manuscript was supported by the funding of thesecond author (Xinwei Lu) and the research results by Dr Lu's group.There is no conflict among the authors, and there is also no conflictbetween the authors and the organizations.

Acknowledgments

The research was supported by the National Natural ScienceFoundation of China through Grant 41271510 and the Fundamental Re-search Funds for the Central University through Grants GK201305008,GK201104002 and GK201101002. Guang Yang and Caifeng Zhaoassistedwith sample preparation.Wealso thank local school authoritiesfor their cooperation. Appreciation is expressed to Associate Editor PhDXuexi Tie and the anonymous reviewers for insightful suggestions andcritical reviews of the manuscript. The authors sincerely thank localschool authorities for their cooperation and Dr. J.R. Grace for editorialassistance.

References

Abdul-Wahab SA, Yaghi B. Total suspended dust and heavy metal levels emitted from aworkplace compared with nearby residential houses. Atmos Environ 2004;38:745–50.

Adamson IYR, Prieditis H, Hedgecock C, Vincent R. Zinc is the toxic factor in the lungresponse to an atmospheric particulate sample. Toxicol Appl Pharmacol 2000;166:111–9.

Akhter MS, Madany IM. Heavy metals in street and house dust in Bahrain. Water Air SoilPollut 1993;66:111–9.

Allen AG, Nemitz E, Shi JP, Harrison RM, Greenwood JC. Size distributions of trace metalsin atmospheric aerosols in the United Kingdom. Atmos Environ 2001;35:4581–91.

Bennett CM, McKendry IG, Kelly S, Denike K, Koch T. Impact of the 1998 Gobi dust eventon hospital admissions in the Lower Fraser Valley, British Columbia. Sci Total Environ2006;366:918–25.

Cao Z, Yang Y, Lu J, Zhang C. Atmospheric particle characterization, distribution, anddeposition in Xi'an, Shaanxi Province, Central China. Environ Pollut 2011;159:577–84.

Chen T, Liu X, Zhu M, Zhao K, Wu J, Xu J, et al. Identification of trace element sources andassociated risk assessment in vegetable soils of the urban–rural transitional area ofHangzhou, China. Environ Pollut 2008;151:67–78.

Chen X, Lu X, Yang G. Sources identification of heavy metals in urban topsoil from insidethe Xi'an Second Ringroad, NW China using multivariate statistical methods. Catena2012;98:73–8.

Chillrud SN, Epstein D, Ross J, Ramstrom S, Pederson D, Spengler JD, et al. Elevatedairborne exposures to manganese, chromium and iron of teenagers from steel dustand New York City's subway system. Environ Sci Technol 2004;38:732–7.

CNEMC (China National Environmental Monitoring Center). The background values ofelements in Chinese soils. Beijing: Environmental Science Press of China; 1990[in Chinese].

De Miguel E, Llamas JF, Chacón E, Berg T, Larssen S, Røyset O, et al. Origin and patterns ofdistribution of trace elements in street dust: unleaded petrol and urban lead. AtmosEnviron 1997;31:2733–40.

Facchinelli A, Sacchi E, Mallen L. Multivariate statistical and GIS-based approach toidentify heavy metal source in soils. Environ Pollut 2001;114:313–24.

Glorennec P, Lucas JP, Mandin C, Bot BL. French children's exposure to metal via ingestionof indoor dust, outdoor playground dust and soil: contamination data. Environ Int2012;45:129–34.

Han Y, Du P, Cao J, Posmentier ES. Multivariate analysis of heavy metal contamination inurban dusts of Xi'an, Central China. Sci Total Environ 2006;355:176–86.

Han Y, Cao J, Posmentier ES, Fung K, Tian H, An Z. Particulate-associated potentiallyharmful elements in urban road dusts in Xi'an, China. Appl Geochem 2008;23:835–45.

Hao Y, Guo Z, Yang Z, Fang M, Feng J. Seasonal variations and sources of various elementsin the atmospheric aerosols in Qingdao, China. Atmos Res 2007;85:27–37.

Hien PD, Binh NT, Truong Y, Ngo NT. Temporal variations of source impacts at the recep-tor, as derived from air particulate monitoring data in Ho Chi Minh City, Vietnam.Atmos Environ 1999;33:3133–42.

Hodge VJ, Austin J. A survey of outlier detection methodologies. Artif Intelli Rev 2004;22:85–126.

Khairy MA, Barakat AO, Mostafa AR, Wade TL. Multielement determination by flameatomic absorption of road dust samples in Delta Region, Egypt. Microchem J 2011;97:234–42.

Laidlaw MAS, Tayor MP. Potential for childhood lead poisoning in the inner cities ofAustralia due to exposure to lead in soil dust. Environ Pollut 2011;159:1–9.

Lee CS, Li XD, Shi WZ, Cheung SC, Thornton I. Metal contamination in urban, suburban,and country park soils of Hong Kong: a study based on GIS and multivariate statistics.Sci Total Environ 2006;356:45–61.

Li XD, Lee SL, Wong SC, Shi WZ, Thornton I. The study of metal contamination in urbansoils of Hong Kong using a GIS-based approach. Environ Pollut 2004;29:113–24.

Li N, Xia T, Nel AE. The role of oxidative stress in ambient particulate matter-induced lungdiseases and its implications in the toxicity of engineered nanoparticles. Free RadicBiol Med 2008;44:1689–99.

Lu X, Li LY, Wang L, Lei K, Huang J, Zhai Y. Contamination assessment of mercury andarsenic in roadway dust from Baoji, China. Atmos Environ 2009;43:2489–96.

Lu X, Wang L, Li LY, Lei K, Huang L, Kang D. Multivariate statistical analysis of heavymetals in street dust of Baoji, NW China. J Hazard Mater 2010;173:744–9.

Lu X, Zhang X, Li LY, Chen H. Assessment of metals pollution and health risk in dust fromnursery schools in Xi'an, China. Environ Res 2014;128:27–34.

Meza-Figueroa D, De la O-Villanueva M, De la Parra ML. Heavy metal distribution in dustfrom elementary schools in Hermosillo, Sonora, México. Atmos Environ 2007;41:276–88.

Monaci F, Bargagli R. Barium and other trace metals as indicators of vehicle emissions.Water Air Soil Pollut 1997;100:89–98.

Needleman HL. The future challenge of lead toxicity. Environ Health Perspect 1990;89:85–9.

Odabasi M, Muezzinoglu A, Bozlaker A. Ambient concentrations and dry deposition fluxesof trace elements in Izmir, Turkey. Atmos Environ 2002;36:5841–51.

Ren C, Lu X, Yang L, Chen C, Yang G. Contamination assessment of heavy metal in dustaround Changqing coal-fired power plant of Baoji. Environ Chem 2011;30(10):1820–1. [in Chinese].

Ruiz-Jimenez J, Parshintsev J, Laitinen T, Hartonen K, Petaja T, Kulmala M, et al. Influenceof the sampling site, the season of the year, the particle size and the number of nucle-ation events on the chemical composition of atmospheric ultrafine and totalsuspended particles. Atmos Environ 2012;49:60–8.

Schleicher NJ, Norra S, Chai F, Chen Y, Wang S, Cen K, et al. Temporal variability of tracemetal mobility of urban particulate matter from Beijing — a contribution to healthimpact assessments of aerosols. Atmos Environ 2011;45:7248–65.

Tanner PA, Ma H, Yu P. Fingerprinting metals in urban street dust of Beijing, Shanghai,and Hong Kong. Environ Sci Technol 2008;42:7111–7.

Tasdemir Y, Kural C. Atmospheric dry deposition fluxes of trace elements measured inBursa, Turkey. Environ Pollut 2005;138:462–72.

Thakur M, Deb MK, Imai S, Suzuki Y, Ueki K, Hasegawa A. Load of heavy metals in theairborne dust particulates of an urban city of central India. Environ Monit Assess2004;95:257–68.

Tokalıoğlu Ş, Kartal Ş. Multivariate analysis of the data and speciation of heavy metals instreet dust samples from the Organized Industrial District in Kayseri (Turkey). AtmosEnviron 2006;40:2797–805.

Tong STY, Lam KC. Are nursery schools and kindergartens safe for our kids? The HongKong study. Sci Total Environ 1998;216:217–25.

Turner A, Simmonds L. Elemental concentrations and metal bioaccessibility in UK house-hold dust. Sci Total Environ 2006;371:74–81.

USEPA, United States Environmental Protection Agency. Health effects support documentfor Manganese; 2003 [EPA 822-R-03-003].

WHO. Air quality guidelines for Europe; 2000.XAMBS (Xian Municipal Bureau of Statistics). Xi'an statistical yearbook in 2011. Beijing:

China Statistics Press; 2011 [in Chinese].Xie YF, Chen TB, Lei M, Yang J, Guo QJ, Song B, et al. Spatial distribution of soil heavymetal

pollution estimated by different interpolation methods: accuracy and uncertaintyanalysis. Chemosphere 2011;82:468–76.

Yang Z, Lu W, Long Y, Bao X, Yang Q. Assessment of heavy metals contamination in urbantopsoil from Changchun City, China. J Geochem Explor 2011;108:27–38.

Yuan GL, Sun TH, Han P, Li J, Lang XX. Source identification and ecological risk assessmentof heavy metals in topsoil using environmental geochemical mapping: typical urbanrenewal area in Beijing, China. J Geochem Explor 2014;136:40–7.

Zhang C, Selinus O, Schedin J. Statistical analyses for heavy metal contents in till and rootsamples in an area of southeastern Sweden. Sci Total Environ 1998;212:217–32.

Zhang X, Arimoto R, Cao J, An Z, Wang D. Atmospheric dust aerosol over the TibetanPlateau. J Geophys Res 2001;106:18471–6.