Geochemical associations for evaluating the availability …users.uoa.gr/~kelepert/pdf/Geochemical associations for... ·  · 2015-04-22Geochemical associations for evaluating the

Applied Geochemistry 59 (2015) 63–73

Contents lists available at ScienceDirect

Applied Geochemistry

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

Geochemical associations for evaluating the availability of potentiallyharmful elements in urban soils: Lessons learnt from Athens, Greece

http://dx.doi.org/10.1016/j.apgeochem.2015.03.0190883-2927/� 2015 Elsevier Ltd. All rights reserved.

⇑ Corresponding author. Tel.: +30 210 7274867.E-mail address: [email protected] (E. Kelepertzis).

Efstratios Kelepertzis ⇑, Ariadne ArgyrakiFaculty of Geology and Geoenvironment, National and Kapodistrian University of Athens, Panepistimiopolis, Zographou, 157 84 Athens, Greece

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

Article history:Available online 6 April 2015Editorial handling by M. Kersten

The estimation of potentially harmful element (PHE) availability in urban soil is essential for evaluatingimpending risks for human and ecosystem health. In the present study five single extraction procedureswere evaluated based on the analysis of 45 urban top-soil samples from Athens, Greece. The pseudototal(aqua regia), potentially phytoavailable (0.05 M EDTA), mobilizable (0.43 M HAc), bioaccessible (0.4 Mglycine) and reactive pools (0.43 M HNO3) of PHEs were determined. In general, geogenic elements inAthens soil (Ni, Cr, Co, As) are relatively less available than typical tracers of anthropogenic contam-ination (Pb, Zn, Cu, Cd). Results of principal component analysis (PCA) indicate an association betweenavailable fractions of Pb, Cu, Zn, Cd and amorphous Fe oxides, whereas amorphous Mn oxides accountfor the available concentrations of Mn, Ni and Co. Empirical multiple linear regression models demon-strate that pseudototal concentration is the predominant explanatory factor of variability for the avail-able pools of the anthropogenic elements. Major elemental composition and total organic carbon (TOC)improve the predictions for the geogenic group of elements, although the explained variability remainslow. Dilute HNO3 is a better predictor of Zn, Ni, As and Mn availability, whereas Pb and Cu available frac-tions are predicted more accurately by the classical aqua regia protocol. This study contributes to theinternational database on the environmental behavior of PHEs and provides additional knowledge thatcan be used toward the harmonization of chemical extraction methodology in urban soil.

� 2015 Elsevier Ltd. All rights reserved.

1. Introduction

Urban soil is a complex component of the urban landscape thatis influenced by both natural and anthropogenic factors.Potentially harmful elements (PHEs), such as heavy metals andmetalloids, are among the most common chemical constituentsin soil that are associated with human activities (RodríguezMartín et al., 2015). As a result elevated concentrations of PHEsin urban soil have been reported for various cities around theworld (e.g. Shi et al., 2008; Cannon and Horton, 2009). SomePHEs, for example zinc (Zn), copper (Cu), manganese (Mn), chro-mium (Cr), nickel (Ni) and cobalt (Co), are essential for livingorganisms in trace amounts while others such as lead (Pb), arsenic(As) and cadmium (Cd) are considered toxic even at lowconcentrations.

Determination of total or pseudototal concentrations of PHEs inurban soil is of little significance for evaluating potential risks forthe environment because it is assumed that 100% of the contami-nant is released from the soil and subsequently involved in chemi-cal and biological processes. PHEs in soil are distributed among

different fractions and forms with different solubility and reactiv-ity in relation to plant uptake, leaching to groundwater andabsorption by the human body (Rodrigues et al., 2010). These so-called plant available, mobilizable and oral bioaccessible fractionsdefine the availability of PHEs in urban soil (Luo et al., 2012).

Chemical methods for evaluating the availability of PHEsinclude single extraction procedures and in vitro digestion modelsthat have been developed to provide a conservative measure of therelative hazard to ecological and human receptors. Potential phy-toavailability and mobility are commonly determined by chelatingagents including ethylenediamine tetraacetic acid (EDTA) anddilute acetic acid (HAc), respectively (Madrid et al., 2008; Poggioet al., 2009; Li and Zhang, 2013). Human oral bioaccessibility isassessed by physiologically based extraction tests (PBET) originallydeveloped by Ruby et al. (1996) simulating the chemical environ-ment of the human gastrointestinal system. Due to its complexity,a simplified extraction test (SBET) has been alternatively applied toseveral studies providing useful information on solubility of PHEsin a medium compositionally similar to acid gastric fluids(Morman et al., 2009; Poggio et al., 2009). Moreover, Römkenset al. (2009) introduced the concept of reactive metal fractionrepresenting the pools of PHEs that are associated with adsorp-tion–desorption phenomena and might be taken up by plants

64 E. Kelepertzis, A. Argyraki / Applied Geochemistry 59 (2015) 63–73

and organisms. Determination of this fraction can be achieved by a0.43 M nitric acid (HNO3) extraction test (Römkens et al., 2009).Interestingly, Rodrigues et al. (2013) concluded that the reactivepools of Cd, Cu, Pb and Zn were robust predictors of their respec-tive availabilities in urban soils from Porto, Portugal. Followingfrom these findings, the applicability of the 0.43 M HNO3 soilextraction for assessing the availability of a considerable suite ofPHEs in urban soil requires further investigation.

Previous research in PHEs availability has shown that varioussoil parameters such as pH, organic carbon and soil texture, as wellas the bulk chemical composition and mineralogy typically controlthe available PHEs fractions (Kelepertzis and Stathopoulou, 2013;Römkens et al., 2009), especially the orally bioaccessible pools(Cox et al., 2013; Palmer et al., 2014). However, the majority ofurban soil investigations that aimed to evaluate the potentialtransfer of pollutants to the groundwater, the plants and the healthrisks to humans by combining major soil properties and chemicalextractions has principally focused on the risks associated withPHEs of anthropogenic origin (Luo et al., 2012; Rodrigues et al.,2013); there is undoubtedly a gap of knowledge regarding theavailability of naturally enriched PHEs in urban soil. Recentinvestigations highlighted the significant contribution of geogenicPHEs in the chemistry of topsoil in many European cities, for exam-ple Athens, Greece (Argyraki and Kelepertzis, 2014), Thiva, Greece(Kelepertzis, 2014), Northampton, UK (Cave et al., 2013) andGrugliasco, Italy (Poggio et al., 2009).

In this article, we provide insights in the reactivity and availabil-ity of PHEs of both natural and anthropogenic origin using an urbandata set of Athens topsoil (Argyraki and Kelepertzis, 2014). Wedetermine the pseudototal (aqua regia), potentially phytoavailable(0.05 M EDTA, pH 7.0), mobilizable (0.43 M HAc), bioaccessible(0.4 M glycine, pH 1.5) and reactive (0.43 M HNO3) pools of PHEs.It is noted that the bioaccessible pool measured in this study refersonly to the oral exposure pathway. Soil properties were also exam-ined including pH, total organic carbon, soil texture and amorphousiron (Fe) and Mn oxides content. The various concentrations ofextractable PHEs are expressed as a function of pseudototal contentand geochemical variables. The results were statistically inter-preted with a view to better understand the soil factors that influ-ence PHEs available fractions and possibly link the chemical formsof the studied elements to their availability.

The specific objectives of the present study are: (a) to evaluatethe availability and reactivity of various PHEs in selected urban soilsfrom Athens and to compare these to the pseudototal content, (b) toassess the soil factors that control the PHEs available fractions byapplying multivariate principal component analysis (PCA) and mul-tiple linear regression, (c) to explore the possibility of implement-ing the fast and easily applicable 0.43 M HNO3 leaching methodfor characterizing the availability of PHEs in Athens soil.

2. Materials and methods

2.1. Sample selection and preparation for analysis

Athens is a European city with a very long history concentratingabout one third of the Greek population (around 3,000,000 cityresidents), as well as a major part of economic and commercialactivities of the whole country. Forty-five composite soil samples(0–10 cm depth) were selected from the sample database of an ear-lier systematic soil geochemical survey. Details on the samplingmethodology are provided in the study of Argyraki andKelepertzis (2014). The criteria for sample selection were the totalcontent of PHEs as determined by a strong acid mixture dissolutionand the spatial variability of soil chemical composition. Theselected samples included low, medium and high levels of PHEs

covering both the periphery and the city core of Athens (Fig. 1).The importance of local geology has been shown to control the dis-tribution of a specific group of elements including Ni, Cr, Co, Mnand As; anthropogenic soil enrichment was identified for Pb, Zn,Cu, Cd in the urban environment of Athens (Argyraki andKelepertzis, 2014).

2.2. Laboratory experimental work

The soil samples were air-dried, disaggregated and sieved to<2 mm fraction. Representative portions of each soil sample werefurther sieved through a nylon 100-lm sieve and stored in roomtemperature. Major physicochemical properties including pH, totalorganic carbon (TOC) and texture (sand, silt, clay) were deter-mined. Soil pH was measured in a soil to deionized water suspen-sion of 1:2.5 (w/v) based on the <2 mm sample fraction (ISO, 1994).Total organic carbon (TOC%) was determined on the <100 lm frac-tion according to the volumetric method described by Walkley andBlack (1934). Grain size distribution (vol.%) in the sand, silt andclay fractions was determined using the hydrometer sedi-mentation method (Bouyoucos, 1962). Amorphous Fe oxides(Feox) and Mn oxides (Mnox) were determined by the acid ammo-nium oxalate extraction in the dark (Schwertmann, 1964). Ironand Mn concentrations in the filtrates were analyzed by flameatomic absorption spectrometry (FAAS) at the Laboratory ofEconomic Geology and Geochemistry, University of Athens, andthe results are presented in mg/kg. Scanning electron microscopy(SEM) and energy dispersive spectra (EDS) analysis were carriedout on free surfaces of carbon-coated soil grains, using a Jeol JSM5600 SEM instrument, equipped with an Oxford ISIS 300 micro-analytical device. The SEM study was performed on the high-den-sity (specific gravity >2.96) fraction of selected soil samples aftergravity separation in sodium polytangstate. Examination in thebackscattered electron (BSE) mode permitted the localization ofareas where heavy metals were concentrated.

Chemical analysis was performed on the <100 lm fractionbecause it has been demonstrated that this soil particle fractionis of major significance for assessing potential environmental risksin urban areas (Luo et al., 2011). The pseudototal content of PHEswas determined after digestion by aqua regia at the AcmeAnalytical Laboratories Ltd of Canada using ICP-MS and was usedto establish the availability and reactivity ratio, i.e. per cent avail-ability and reactivity. Results for the major elements (calcium (Ca),magnesium (Mg), aluminum (Al), Fe) are also presented and theconcentrations are expressed in %. Replicates, in-house referencematerials and reagent blanks were used for quality control. Theavailability of PHEs was assessed by applying the widely adoptedextraction methods of EDTA, HAc and SBET. The metal fractionsobtained by these analytical procedures were operationallydefined as potentially phytoavailable (Madrid et al., 2008; Li andZhang, 2013), mobilizable (Gupta et al., 1996; Sahuquillo et al.,2003) and orally bioaccessible (Luo et al., 2012; Popescu et al.,2013), respectively. Following results recently presented byRodrigues et al. (2013), the soil reactive content of PHEs was deter-mined after extraction with dilute HNO3.

The EDTA extractable content was obtained from subsamples of4 g of soil leached with 40 ml of 0.05 M EDTA solution (adjusted topH 7) for 1 h at room temperature (Ure et al., 1993). The mobilizablepools of PHEs were determined by treating 1 g of soil sample with40 ml of 0.43 M HAc solution for 16 h at room temperature(Houba et al., 1996). The reactive forms of PHEs were extractedby mixing 1 g of soil material with 40 ml of a 0.43 M HNO3 solutionand shaken for 2 h at room temperature. We modified the proposed1/10 (w:v) ratio of Rodrigues et al. (2010) because of the calcareousnature of Athens soil ensuring that pH values of the final extractionfluids were within the range 0.8–1.0. All the mixtures from these

Fig. 1. Topographical map showing the soil sampling locations within the urban net of the Greater Athens and Piraeus area (Athens Ring corresponds to the city center wheretraffic restrictions have been enforced by allowing alternatively odd/even plate number vehicles to enter on subsequent days).

E. Kelepertzis, A. Argyraki / Applied Geochemistry 59 (2015) 63–73 65

extractions were shaken in a mechanical and end-over-end shakerat 25 rpm. The extracts were separated from the solid residue bycentrifugation at 3000 rpm for 10 min. A 10-ml portion of thesupernatant was subsequently removed by using a disposable syr-inge and filtered through a 0.45-lm filter. All chemical extractionswere performed at the Laboratory of Economic Geology andGeochemistry, University of Athens. Analar�-grade chemicals anddeionized water were used throughout the analysis. All glass lab-oratory utensils were washed with a detergent, then soaked for24 h in 10% HNO3 acid solution and rinsed repeatedly with deion-ized water. Concentrations of Pb, Zn, Cu, Ni, Cr, Co, Mn, As and Cdwere measured by ICP-MS at the analytical facilities of School ofEarth and Environmental Sciences, University of Portsmouth, UK.

Procedural blanks and five analytical duplicates were added toeach analytical batch (batches were based on chemical extraction)for quality control purposes. The relative percent difference (RPD)was calculated for each pair of duplicates as an indication ofmethod precision revealing mean RPD values lower than 20% forall elements and extractions except for As in the EDTA test. Toassess the accuracy of the EDTA and HAc extractions, two certifiedreference materials (BCR 483 and BCR 484) for extractable traceelement concentrations (Quevauviller et al., 1997) were includedin duplicate in each batch. Average recovery percentages of the dif-ferent elements were 123% (Ni), 110% (Cu), 107% (Zn), 95% (Cd),78% (Pb) for the EDTA extraction, and 120% (Ni), 96% (Cu), 134%(Zn), 127% (Cd), 88% (Pb) for the HAc extraction. The HAc extracta-ble Cr was below the detection limit for the majority of samples. As

a consequence the results of this extraction were not further pro-cessed for the specific element.

The bioaccessible forms of PHEs were determined by the SBETmethod simulating the low pH conditions of the human gastric flu-ids (Oomen et al., 2002). Fifty ml of the 0.4 M glycine extractionsolution (adjusted to pH 1.5 with concentrated HCl) were addedto 0.5 g of soil. The suspensions were quickly placed in a waterbath and agitated at 37 �C for 1 h. The pH of the unfiltered soil sus-pensions was measured by immersing a pH electrode calibratedwith standards of pH 1 and 4, verifying that all pH values werewithin 0.5 units of the starting pH (1.5). A portion of the super-natant was removed with a disposable syringe and filtratedthrough a 0.45-lm filter. Chemical determination of Pb, Zn, Cu,Ni, Cr and Mn was performed by flame atomic absorption spec-trometry (FAAS). Method limitations relevant to the capabilitiesof the used instrument as well as research logistics did not allowthe measurement of Co, Cd and As concentrations. Three extractionblanks and five analytical duplicates were used as part of the qual-ity assurance protocol. The RPD values were below 10% on averagefor all elements except for Cd (23%), Ni (15%) and Cr (13%).

2.3. Statistical analysis

The statistical analysis described below was performed usingthe IBM SPSS v.22 (2013) software. Principal component analysis(PCA) with the varimax rotation method was applied to the log-transformed chemical data with the aim to explore the association

Table 1Statistical summary of physicochemical properties and aqua regia extractableconcentrations of major and trace elements for the investigated topsoil samplesfrom Athens city (n = 45).

Parameter Mean Minimum Maximum Standarddeviation

Median

pH 8.32 7.7 9.0 0.31 8.3TOC (%) 2.15 0.6 4.49 0.93 2.15Sand (%) 56.4 25 75 9 57Silt (%) 29.5 15 48 6.18 29Clay (%) 14 10 27 3.58 13Feox (mg/kg) 1340 342 3630 684 1220Mnox (mg/kg) 271 52.1 2050 303 183Ca (%) 12.6 1.53 25.9 6.07 13.4Mg (%) 0.83 0.37 2.86 0.51 0.69Al (%) 1.29 0.43 2.75 0.49 1.22Fe (%) 2.40 0.99 4.06 0.71 2.35Mn (mg/kg) 662 246 2810 380 564Pb (mg/kg) 157 9.6 823 159 106Zn (mg/kg) 174 37.2 783 138 146Cu (mg/kg) 72.3 15.1 316 52 59.2Ni (mg/kg) 131 25.4 762 135 94.5Cr (mg/kg) 95.4 21.1 558 84.7 82.4Co (mg/kg) 17.2 8.7 52.8 8.25 14.5Cd (mg/kg) 0.45 0.09 2.35 0.35 0.37As (mg/kg) 43.2 4.4 227 43.6 27.1

66 E. Kelepertzis, A. Argyraki / Applied Geochemistry 59 (2015) 63–73

of the various PHEs pools with specific chemical forms. We incor-porated in the PCA the key soil properties of TOC as well as amor-phous Fe and Mn oxides content as the geochemical parametersmost likely to exert influence on PHEs availability. The soil textureand pH as well as the pseudototal Ca, Al and Mg contents were notconsidered because their contribution to the PCA appeared to beinsignificant. Aqua regia extractable Fe was also not includedbecause amorphous Fe oxides will be inherently controlled by theirpseudototal content leading to adverse influence on the outputresults. The selection of the optimum principal components wasbased on the scree plot (Cattell, 1966) which is the graphicalvisualization of the relationship between the eigenvalues and thenumber of components. In this case, the cut-off is chosen at thepoint where the function displayed by the scree plot shows anelbow allowing the division of the major components from thetrivial components.

To further investigate the influence of pseudototal element con-tent and the basic soil properties on PHEs availabilities, multiplelinear regression was employed to develop a model explainingthe highest proportion of variance in the data. To overcome theinfluence of extreme values on the regression analysis, the loga-rithms of the original chemical data were used (Reimann et al.,2008). The first step of analysis consisted of selecting the most sig-nificant parameters to describe the PHEs availability. The Pearsoncorrelation coefficients between each of the explanatory soil chem-istry variables, i.e. aqua regia extractable concentration and soilproperties (pH, TOC, sand, silt, clay, Feox, Mnox) and the dependentvariable (available or reactive concentration of PHE) was tested toselect the soil parameters with the strongest linear relationshipwith the available or reactive PHEs pools. Empirical regressionmodels were derived by stepwise selection of explanatory vari-ables. The least significant parameters, defined by their largest pvalue, were discarded until all the remaining variables (includingthe intercepts) contributed significantly to the regression modelat the 95th percentile confidence level (p < 0.05).

Once the regression model was determined, a normal probabil-ity plot for residuals was used for suggesting the presence of out-liers in the data and checking that residuals were approximatelynormally distributed (Reimann et al., 2008). The co-linearity inthe model was examined by the variance inflation factor (VIF) ofeach explanatory variable (Rawlings et al., 1998). Variables withVIF > 2 were excluded from the model to avoid the redundancyof the independent variables. The root mean square error (RMSE)and the adjusted coefficient of determination (R2

adj) were used toprove a regression model is robust. Regression models withR2

adj < 50% were considered as insignificant and are not presented.

3. Results

3.1. Soil characteristics and pseudototal elemental concentrations

The physicochemical characteristics and concentrationsextracted by aqua regia for the selected soil samples are presentedin Table 1. The alkaline soil pH values, ranging from 7.7 to 9.0, arereflective of the abundant presence of calcite in bedrock (Argyrakiand Kelepertzis, 2014). The selected soil samples exhibit a widerange of TOC values (0.6–4.5%), clay (10–27%), silt (15–58%) andsand (25–75%). Significant variations were observed in Fe andMn amounts extracted by the ammonium oxalate dissolution(342 to 3630 mg/kg and 52.1 to 2050 mg/kg respectively).Amorphous Fe and Mn oxides account for about 1.2–18.1% and17.7–73% respectively of the iron and manganese contentextracted by aqua regia. Results of SEM–EDS analysis of isolatedmetal bearing soil grains verified the presence of iron oxide phasesenriched in Pb (1.41 wt%) and Zn (2.32 wt%) as well as

aluminosilicate mineral grains encrusted by Mn–Fe oxide coatingsenriched in Pb (14.2 wt%) and Sb (2 wt%) (Fig. 2). All elements exhi-bit a large range in their concentration values reflecting the selec-tion criteria of the investigated soils. Of the PHEs, Mn and Pbdisplay the highest concentration ranges (2560 and 813 mg/kgrespectively). Median values of the studied elements follow thedecreasing order of Ca > Fe > Al > Mg > Mn > Zn > Pb > Ni > Cr >Cu > As > Co > Cd.

Comparing the results from this work to other urban geochemi-cal studies is challenging because of the different objectives of eachsurvey. However, compared to similar targeted studies levels ofanthropogenic Pb and Cu in this urban data set were considerablyhigher than those from Porto (Rodrigues et al., 2013) and HongKong (Luo et al., 2012). Zinc loadings are similar to levels measuredin Hong Kong and notably higher compared to those determined inPorto soils. Arsenic levels found in our study are identical to thevalues observed in soils from Glasgow and London (Appletonet al., 2012a) whereas Cr and Ni are somewhat similar to therespective aqua regia extractable content in Grugliasco (Italy)(Poggio et al., 2009). Cobalt and Mn have not yet been investigatedin urban geochemical studies of this kind except for Mn content(median of 425 mg/kg) measured in urban soils from the city ofGlasgow (Sialelli et al., 2010).

3.2. Single extraction efficiency for the investigated PHEs

The availability of PHEs in this study was defined as the ratio ofamounts dissolved by the EDTA, HAc and SBET chemical proce-dures and the aqua regia extractable content. Likewise, the reactiv-ity was expressed as the percentage of the fraction solubilized bythe dilute HNO3 relative to the aqua regia. The ranges of availabil-ity and reactivity varied widely among the investigated elements(Fig. 3 and Table 2). The median availability ratios in decreasingorder are Cd (44%) > Pb (28%) � Cu (27%) > Zn (18%) > Mn(14%) > As (4%) � Co (4%) � Ni (3%) > Cr (1%) for the EDTA extrac-tion; Cd (74%) > Mn (38%) > Zn (16%) > Co (8%) > Pb (6%) � As(6%) � Ni (5%) > Cu (1%) for the HAc extraction; and Pb(58%) > Mn (37%) > Zn (29%) > Cu (21%) > Ni (14%) > Cr (5%) forthe SBET extraction. Considering their reactivity ratio, PHEs fol-lowed a decreasing order of Pb (76%) > Mn (68%) � Cd (68%) > Co(32%) � Cu (31%) � Zn (31%) > Ni (12%) � As (11%) > Cr (5%).

Fig. 2. SEM microphotographs in back scatter mode showing PHE enriched soil grains. (a) Bright Pb (1.41 wt%), Zn (2.32 wt%)-bearing iron oxide phase, (b) aluminosilicategrain encrusted by Mn–Fe oxide coating on its edge (white frame). (c) Detail of (b) showing Mn–Fe oxide phase (bright) which is enriched in Pb (14.2 wt%) and Sb (2 wt%) andthe respective EDS spectrum.

E. Kelepertzis, A. Argyraki / Applied Geochemistry 59 (2015) 63–73 67

Amounts extracted with 0.43 M HNO3 are significantly higherthan those extracted by the other extraction schemes for Pb, Cu,Mn, Co and As (p < 0.05 in all cases). This indicates that diluteHNO3 solubilizes a specific soil constituent that is not attackedeven by the acid conditions prevailing during the SBET extraction.For Zn, no statistically significant differences were observedbetween the SBET and HNO3 pools (p > 0.05). EDTA extractablecontents are typically lower than those obtained by the stronglyacidic procedures of HNO3 and SBET except for Cu which showscomparable concentrations (p > 0.05) for the EDTA and SBET soiltests. This may be due to the strong affinity of Cu for the chelatingagent via the release of soluble organically bound Cu complexes(Römkens et al., 2009). EDTA-extracted concentrations of Pb andCu are higher than those determined by the HAc protocol becauseof the higher sensitivity of these elements to complexation(Sahuquillo et al., 2003). Nickel, As, Co and Zn exhibit similarextraction characteristics by both (EDTA and HAc) extracting solu-tions while a higher HAc extraction efficiency compared to EDTA isobserved for Cd and Mn.

3.3. Results of PCA and multiple linear regression

The results of PCA including key soil properties and the variousPHEs pools are shown in Table 3. A two-component solution for theHAc and HNO3 chemical extractions was identified based on thescree plot criterion. Three major components were extracted forthe EDTA and SBET methods. In the EDTA extraction, Ni, Co andMn are closely associated with Mnox in PC1. Lead, Cu, Zn and Cdare associated with Feox in PC2, while Cr is univocally isolated inPC3 with high loadings. In the HAc extraction, Pb, Cu, Zn and Cd

are heavily loaded in PC1, while TOC, Mn and Co are included inPC2. For the HNO3 extraction, Pb, Cu, Zn and Cd are associated withFeox in PC1, while Ni, Co and Cr showed high loadings in PC2.Medium loadings of Cu (0.59) also occur in PC2. Finally in theSBET experiment, Pb, Cu and Zn are linked to Feox in PC1, whileMn, Mnox, Ni and TOC are included in PC2. Chromium occurs inPC3 with high loadings.

Relationships between availabilities and reactivities of PHEs inAthens soil and their pseudototal content, general soil characteris-tics and concentrations of major elements were further investi-gated by means of multiple linear regression analysis (Table 4and Fig. 4). All regression models are statistically significant(p < 0.005). The regression models shown in Table 4 have passedthe residual and co-linearity checks providing geochemically justi-fied explanations for the available fraction of PHEs in Athens soil.When the coefficients of the intercept values were found to beinsignificant for the predictions (indicated by p values higher than0.05), the equations have been omitted from Table 4, for examplethe predictions of the reactive pools of Mn and Cu. In other cases,serious violations of the normality of residuals were observed as inthe case of Pb reactive pools and oral bioaccessible pools of Ni andCr. The ability of the regression models to simulate PHEs availabil-ity and reactivity was quantified using RMSE. The RMSE values ran-ged from 0.11 (Cd) to 0.21 (As) for the predictions of potentiallyphytoavailable pools, 0.10 (Mn) to 0.31 (Cu) for the predictionsof mobilizable pools, 0.15 (Zn) to 0.20 (Cr) for the predictions ofthe reactive pools and 0.09 (Mn) to 0.15 (Cu) for the predictionsof the oral bioaccessible pools. Such low RMSE values indicate thatthe available and reactive pools of PHEs were predicted well by theproduced models. The extracted equations (Table 4) demonstrated

Fig. 3. Extractability ratio of the HAc, EDTA, SBET and HNO3 pools relative to the aqua regia extractable content for Athens topsoil (n = 45). In the box plots, the length of thebox indicates the interquartile range whereas the horizontal line within each box represents the median. Whiskers extend to the maximum and minimum data point within1.5 box heights from the top and the bottom of the box. The star characters are the maximum and minimum values and the square sign represents the mean value.

68 E. Kelepertzis, A. Argyraki / Applied Geochemistry 59 (2015) 63–73

that the variability in the concentration of the available and reac-tive pools of the anthropogenic elements is mostly explained bytheir respective aqua regia extracted content (75.5% for CuHAc upto 91.9% for PbEDTA). The only exceptions were the reactive poolsof Cu and Pb for which aqua regia was not found to be a robust pre-dictor. In most cases, the stepwise incorporation of aqua regiaextractable Ca and Mg for individual anthropogenic elementsallowed improvement of predictions, indicated by increased R2

adj

values. With respect to the geogenic elements aqua regia is not a pri-mary controlling factor in determining their availability and reactiv-ity, as revealed by the lower values of correlations coefficients(Fig. 4) and the low R2

adj values (Table 4).

4. Discussion

4.1. Evaluation of extractability patterns of PHEs in Athens soil andpossible links to their geochemical forms

In this study four single extraction schemes were applied to 45urban soil samples from Athens city with the aim to evaluate theavailability and reactivity of PHEs. Overall, large differences inextractability patterns for the studied PHEs between the variouschemical procedures are observed (Fig. 3). It may be inferred thatthe observed extraction efficiencies are due to the chemical

conditions during the extraction tests as well as the variety ofchemical forms defining the speciation of the studied elements insoil.

Specifically, there is a differentiation in the extractability ofanthropogenic (Pb, Zn, Cu, Cd) and geogenic (Mn, Ni, Cr, Co, As)group of elements. Lead was found to exhibit high oral bioac-cessibility and reactivity with median actual values of 60 and67 mg/kg respectively (Table 2). The SBET extractable Pb of thisstudy (median 58%) was comparable to the median bioaccessibilityof 68% in London (Appleton et al., 2012b), 59% in Hong Kong (Luoet al., 2012) and 45% in Lisbon, Portugal (Reis et al., 2014). Leadreactive data for Porto soils (Rodrigues et al., 2013) and diversesoils from Portugal (Rodrigues et al., 2010) were substantiallylower than the reactive pools (median 76%) determined in the pre-sent study. Such leachate amounts indicate that a considerablefraction of Pb in Athens soil remains in a highly bioaccessibleand reactive form, probably reflecting its fixation by Fe oxides thatare effectively dissolved by the stronger acid extractants. Suchphases were identified in selected soil samples studied by SEM–EDS (Fig. 2a). By contrast, Zn and Cu exhibited lower availabilitiesand reactivities (Fig. 3), in line with findings by other authors(Gasparatos et al., in press; Poggio et al., 2009) who highlighted alarge residual pool of these metals. Nonetheless, it is clear that asmall percentage of both Zn and Cu (up to 31% for reactive Cuand up to 30% for reactive Zn) is consistently available and reactivereflecting the presence of metal-host phases of high lability.

Table 2An overview of the PHEs pools (mg/kg) as determined by 0.05 M EDTA, 0.43 M HAc,0.4 M glycine (SBET) and 0.43 M HNO3 chemical extractions.

Mean Minimum Maximum Median

Pb PbHAc 16.4 0.12 251 5.5PbEDTA 49.1 2.14 310 29.9PbSBET 91.2 10 520 60.4PbHNO3 114 1.4 532 66.8

Zn ZnHAc 40.1 0.33 416 27.5ZnEDTA 36.9 1.37 309 23.7ZnSBET 59.9 1 544 33ZnHNO3 59.8 3.8 453 38.2

Cu CuHAc 2.1 0.03 29.9 0.848CuEDTA 21.9 1.63 115 15.2CuSBET 19.8 1.99 159 11CuHNO3 27 1.79 182 19.5

Cd CdHAc 0.37 0.09 1.58 0.3CdEDTA 0.21 0.04 1.18 0.16CdHNO3 0.3 0.06 1.25 0.25

Mn MnHAc 233 94.6 590 219MnEDTA 102 33.3 473 78.2MnSBET 223 84.8 896 211MnHNO3 441 90.8 1950 382

Ni NiHAc 6.15 1.92 20.6 5.44NiEDTA 3.86 0.72 17.1 3.18NiSBET 13.6 3.39 42.5 13.7NiHNO3 13.2 0.97 81.7 10.5

Co CoHAc 1.64 0.28 10 1.21CoEDTA 1.05 0.2 13.6 0.65CoHNO3 5.64 1.33 22.6 4.63

As AsHAc 2.47 0.16 13 1.88AsEDTA 1.73 0.34 6.7 1.15AsHNO3 4.28 0.49 23.6 3.19

Cr CrEDTA 0.66 0.14 1.71 0.59CrSBET 4.19 0.5 8.77 4.19CrHNO3 6.14 0.62 27.4 4.7

E. Kelepertzis, A. Argyraki / Applied Geochemistry 59 (2015) 63–73 69

Cadmium was found to exhibit high availability and reactivity(Fig. 3) indicating that significant proportions of Cd are associatedwith labile ion-exchangeable soil components and easily solublehydrous oxides (Rodrigues et al., 2013). Results of PCA (Table 3)provide evidence that the extractable fraction of the anthropogenicgroup of elements is closely related to Feox and reveals the effectivebinding of Cu, Zn, Pb and Cd to amorphous Fe oxides by adsorptionprocesses.

The geogenic elements, Ni, Cr, Co and As are characterized bylow availability and reactivity with the exception of Co which dis-played a median value for its reactive pool equal to 32% of the aquaregia extractable content. The low lability of geogenic Ni, Cr, Co andAs suggests that these contaminants in Athens soil are present in

Table 3Results of principal component analysis (rotated matrix) for log-transformed data includincomponent loadings are indicated in italics. Correlations between HAc extracted concentr

Extraction method EDTA HAc

PC1 PC2 PC3 PC1 P

TOC 0.403 0.312 0.485 0.162Mnox 0.822 0.140 �0.314 –Feox 0.452 0.705 �0.057 –Pb �0.080 0.816 0.164 0.844Cu �0.180 0.885 0.260 0.837 �Zn �0.038 0.837 0.394 0.871Cd 0.439 0.637 0.357 0.777Ni 0.715 0.083 0.458 0.139 �Co 0.832 �0.274 0.249 0.122Mn 0.953 �0.028 0.041 0.137Cr 0.182 0.177 0.879 –As �0.130 0.206 0.405 0.508Eigenvalues 4.64 3.04 1.21 3.61% of variance 29.46 27.71 16.93 34.58

rather immobile and chemically inert fractions. Similarly,Kelepertzis and Stathopoulou (2013) reported the immobilizationof large pools of geogenic Ni, Cr and Co in soils from Thiva town(central Greece) due to their incorporation in the crystal structureof recalcitrant minerals. Low bioaccessible As has also beenreported in Glasgow and Northampton areas with medians of19% and 9% respectively (Appleton et al., 2012a) whereas the med-ian of As reactivity was rather low (6.2%) in various contaminatedand non-contaminated soils from Portugal (Rodrigues et al., 2010).Despite its natural origin, Mn is characterized by high mobility,oral bioaccessibility and reactivity (Fig. 3 and Table 2). This isreflective of the abundant presence of amorphous Mn oxides in soil(Villalobos et al., 2005a) that are unstable even under slightlyacidic conditions. Manganese oxide phases as soil grain coatingshave been identified by SEM–EDS in this study (Fig. 2b). The sig-nificant role of layer-type, biogenic and synthetic manganese oxi-des in the sequestration of PHEs and specifically Pb has also beendemonstrated experimentally by Villalobos et al. (2005b). Theseauthors reported relatively fast sorption, with maximum saturat-ing concentrations reached within 1 day and extremely high sorp-tion capacities, suggesting Pb incorporation into the structure ofthe oxides. Results of PCA also showed that Mn is strongly boundto organic carbon in Athens soil resulting to its high mobilizablepools (Table 2). Furthermore, PCA results revealed a major associa-tion of Ni and Co with Mn and Mnox (Table 3) indicating thepreferential binding of the soluble fraction of Ni and Co with amor-phous Mn oxides. Indeed, relationships of Ni with hydrous Mn oxi-des have been reported for natural soils from northeast Portugal(Alves et al., 2011). The existence of Co-bearing Mn oxides is alsocommon in the literature (Kelepertzis and Stathopoulou, 2013).On the other hand, Cr and As do not show a clear relationship withthe considered geochemical parameters reflecting their fixation tomore stable chemical forms, such as well crystallized Fe oxides oracid soluble phyllosilicate minerals.

At this point, it should be emphasized that the application ofPCA requires some statistical criteria to be fulfilled for obtainingscientifically stable results. One requirement is driven by thedimensionality of the geochemical data determined by the avail-ability of a sufficient number of samples for the number of vari-ables (Reimann et al., 2002). In the present study, the number ofsamples (n = 45) is rather low in relation to the number of variablesentered in the PCA (p = 13), either when tolerant rules are used, forexample n > p2 or just n > 8p (Reimann et al., 2008). The aboverules do not necessarily weaken the strength of the results fromthe PCA, but they do remind us that geochemical interpretationsshould be treated with caution and that geochemical reasoning is

g TOC, Mnox, Feox and various PHEs pools for Athens soil (n = 45). Significant principleations and Mnox, Feox are not considered.

HNO3 SBET

C2 PC1 PC2 PC1 PC2 PC3

0.848 0.433 0.064 0.151 0.734 0.386– �0.097 �0.033 �0.236 0.796 �0.379– 0.697 0.044 0.642 0.437 �0.3210.167 0.826 �0.164 0.847 �0.011 0.1470.372 0.669 0.587 0.902 �0.035 0.2700.336 0.875 0.337 0.879 0.227 0.2680.310 0.819 0.201 – – –0.017 0.041 0.877 0.144 0.669 0.5140.636 �0.137 0.775 – – –0.909 0.199 0.470 0.128 0.875 0.171– 0.352 0.861 0.403 �0.196 0.7420.230 0.510 0.082 – – –1.78 4.87 2.38 3.63 2.27 1.17

25.13 30.64 23.93 31.95 26.55 20.04

Table 4The stepwise linear regression model for the reactive and available pools of the investigated PHEs against the respective aqua regia content and a range of elemental compositionsand soil properties. All parameters shown are based on log-transformed values. R2

adj(a) corresponds to the variance explained only with respect to the aqua regia content andR2

adj(b) corresponds to the variance explained taking into account all the considered variables. Only the most significant robust models are shown.

Regression equation R2adj(a) R2

adj(b)

EDTAPb PbEDTA = �0.936 + 1.17PbAR 91.9Zn ZnEDTA = �3.43 + 1.24ZnAR + 0.426CaAR 79.2 86.9Cu CuEDTA = �1.41 + 1.45CuAR 90Cd CdEDTA = �0.424 + 0.955CdAR + 0.201TOC 84.1 85.3Mn MnEDTA = 2.46 + 0.584MnAR � 0.457CaAR + 0.473TOC 54.9 79.6Ni NiEDTA = �1.23 + 0.552NiAR + 0.539TOC + 0.205Mnox 31.2 58.6Co CoEDTA = 1.497 + 0.717CoAR � 0.533CaAR + 0.597TOC 34.2 61.4As AsEDTA = 2.80 + 0.609AsAR � 0.822FeAR 41.6 54.4

HAcPb PbHAc = �4.696 + 1.503PbAR + 0.466CaAR 83.3 86.1Zn ZnHAc = �5.07 + 1.618ZnAR + 0.576CaAR 77 85Cu CuHAc = �7.85 + 2.26CuAR + 0.918MgAR 75.5 81.4Cd CdHAc = �0.22 + 0.798CdAR 82.4Mn MnHAc = 1.49 + 0.261MnAR + 0.428TOC 19.3 56Ni NiHAc = �1.34 + 0.498NiAR + 0.280MgAR 55.9 63.2As AsHAc = �5.325 + 0.524AsAR + 0.951CaAR 14.2 64.6

HNO3

Zn ZnHNO3 = �4.55 + 1.18ZnAR + 0.535CaAR + 0.289Feox 80.4 89.9Cd CdHNO3 = �0.161 + 1.05CdAR 89.6Ni NiHNO3 = �1.60 + 0.726NiAR + 0.304MgAR 55.2 58.8Co CoHNO3 = �0.633 + 1.093CoAR 55.1As AsHNO3 = �3.41 + 0.714AsAR + 0.566CaAR 45 68.9Cr CrHNO3 = �5.14 + 0.761CrAR + 0.585CaAR + 0.370MgAR 26.6 62

SBETPb PbSBET = �1.75 + 0.769PbAR + 0.394CaAR 81.2 88.9Zn ZnSBET = �4.81 + 1.37ZnAR + 0.675CaAR 76.3 90.9Cu CuSBET = �3.774 + 1.117CuAR + 0.401CaAR + 0.2225MgAR 83.8 91.1Mn MnSBET = 0.889 + 0.474MnAR + 0.380TOC 40.9 63.7

70 E. Kelepertzis, A. Argyraki / Applied Geochemistry 59 (2015) 63–73

required for the appropriate variable selection when the number ofsamples is not considerably higher than the number of variables.

4.2. Significance of soil factors in PHEs availability

In this section, we discuss the most significant factors control-ling the release or retention of PHEs in Athens soil and, thereby,having an effective role to their availability and reactivity. Thestrong influence of aqua regia analyte values on various PHEs avail-abilities has been expressed widely in the literature (e.g. Rousselet al., 2010; Sialelli et al., 2010; Li and Zhang, 2013) and this pat-tern was also shown in Athens soil (Fig 4). However, a clear differ-entiation was seen between the anthropogenic and geogenic groupof PHEs. In particular, relationships between availability andreactivity values and pseudototal contents were clearly strongerfor the anthropogenic Pb, Zn, Cu and Cd (Fig. 4). The availableand reactive pools of these metals were significantly predicted bytheir respective pseudototal content (Table 4). In contrast, predic-tions were substantially less accurate for the geogenic group of ele-ments. Several soil parameters including pH, organic carbon, grainsize and major element composition have been also shown to exertsignificant control on PHEs availability (Cox et al., 2013; Pelfrêneet al., 2012; Roussel et al., 2010). Results of both PCA and regres-sion analysis demonstrated the high sensitivity of the geogenicgroup of elements to TOC. The effect of organic carbon is far morepronounced for controlling the availability of Mn highlighting thatsoil organic matter effectively binds some Mn that is readily avail-able in all the applied leaching procedures. Furthermore, TOC wasfound to exert significant controls on predicting the potentiallyphytoavailable pools of Ni and Co. Some statistical relationships,for example the influence of Mg content on PHEs availabilitiesare difficult to explain highlighting the difficulties of statisticalmodels to overcome the complex interactions of PHEs with the soilcomponents.

Apparently, the parameters determined in this study are not theonly ones controlling the available and reactive pools of geogenicPHEs. The amount of Fe associated with well-crystallized oxideswas not specifically measured in this study. This may explain thelow values of R2

adj determined for the geogenic PHEs. Incorporationof metals to aged crystalline forms of Fe oxides by adsorption pro-cesses are known to restrict their availability, as it has been demon-strated in the case of As (Appleton et al., 2012a), Ni (Massoura et al.,2006) and Cr (Kelepertzis and Stathopoulou, 2013). In the presentstudy, aqua regia extractable Fe was accounted for the potential phy-toavailability of As (Table 4). The negative model coefficients inregression equations suggest that Fe restricts As release from the soilmatrix. Negative correlations between extractable As and Fe contentare common in the literature (Appleton et al., 2012a; Das et al., 2013)and have been attributed to adsorption of As onto less soluble Feoxyhydroxides.

The Ca content of Athens soil is revealed in the present study asan important factor controlling the available and reactive pools ofvarious PHEs of both natural and anthropogenic origin (Table 4).The fact that partial correlations are positive is reflective of theadsorption of contaminants on the abundant carbonate surfacesof Athens calcareous soil. Additionally, carbonates are easilydegraded by acid digestion, causing the release of adsorbed ele-ments to solution. Adsorption mechanisms of As to carbonate sur-faces or the presence of Ca–As precipitates were found to controlAs release from alkaline soils in Madrid (Mingot et al., 2011). TheCa content was the only significant predictor of Pb mobility(HAc-extracted) in mildly acidic and alkali soils from Hangzhoucity (Li and Zhang, 2013). Nevertheless, the exact influence of Cain Athens soil is unclear because it seems to beneficially impacton the potentially phytoavailable pools of Co and Mn.Relationships between extractabilities of PHEs and Ca content inAthens soil are highly variable and need further clarificationbecause metal binding competition between carbonates and

Fig. 4. Available and reactive concentrations of PHEs in Athens soil as a function of geochemical values obtained by the aqua regia extraction. The Pearson correlationcoefficients (r) between extractable and pseudototal PHEs pools are also shown based on their log-transformed values.

E. Kelepertzis, A. Argyraki / Applied Geochemistry 59 (2015) 63–73 71

ligands in the extracted solutions may affect the analytical results,as in the case of glycine (De Miguel et al., 2012). The influence ofpH on the availability and reactivity of PHEs in Athens soil is notso crucial as in other studies (Rodrigues et al., 2013; Luo et al.,2012). This may be due to the limited range of pH values character-izing the Athens soil that does not allow the importance of this soilproperty in governing contaminant availability to be implicated.Finally, the soil clay fraction (<2 lm) which is typically enrichedin Fe, Mn and Al oxides and organic coatings on mineral surfacesdoes not influence the PHEs availability and reactivity in Athenssoil.

4.3. Is the dilute HNO3 appropriate for evaluating the availability ofPHEs in Athens soil?

Recently, Römkens et al. (2009) introduced the concept of reac-tive fraction, determined by dilute nitric acid (0.43 M), for eval-uating the directly available pools of Cd, Zn and Ni in paddyfields in Taiwan. This fraction incorporates the sorption of PHEson surfaces of organic carbon, amorphous metal oxides and clayand is distinguished from the geochemically inert fraction that isembedded in crystalline metal oxides and stable silicate and sul-fide mineral phases. Moreover, Rodrigues et al. (2013) used thereactive pools of Pb, Cu, Zn and Cd for risk assessment in the urbanarea of Porto. They concluded that reactive concentrations are

more suitable to effectively assess metal pools available for trans-fer to plants, leaching to groundwater and uptake by the humanbody.

In Table 5, results of multiple regression analyses for the inves-tigated PHEs are presented using the 0.43 M HNO3 extraction con-centrations as the predictor variable instead of the classical aquaregia protocol results. The use of HNO3 results in a higher R2

adj forZn, Ni, As and Mn indicating that the reactive pools provided betterpredictions of the available contents of these elements. The onlyexception is Mn for which the potential phytoavailability was betterexpressed by aqua regia. Nonetheless, it should be emphasized thatNi, Cr and Mn availabilities were still not well described by the equa-tions with maximum R2

adj values reaching 75% for Ni mobility and60% for Mn oral bioaccessibility (Table 5) highlighting that moreinformation on soil properties and crystalline forms of Fe oxidesare needed to estimate more accurately their availability. In the caseof As, the estimations based on HNO3 were substantially improvedreaching 67% for As potential phytoavailability and 75% for As mobil-ity. Interestingly, all the available pools of Pb and Cu were describedmore accurately by the aqua regia data. This observation indicatesthat pseudototal fraction of Pb and Cu includes a geochemical phasethat is in equilibrium with their available pools. For Cd, the estima-tions based on HNO3 are comparable to those obtained by using aquaregia, showing that the reactive pools are closely related to thepseudototal content in the soil.

Table 5The stepwise linear regression model for the available pools of the investigated PHEs against the respective reactive content and a range of elemental compositions and soilproperties. All parameters shown are based on log-transformed values. R2

adj(a) corresponds to the variance explained only with respect to the respective reactive content andR2

adj(b) corresponds to the variance explained taking into account all the considered variables. Only the most significant robust models are shown.

Regression equation R2adj(a) R2

adj(b)

EDTAPb PbEDTA = 0.2 + 0.717PbHNO3 83.4Zn ZnEDTA = �0.008 + 0.882ZnHNO3 88.2Cu CuEDTA = �4.14 + 1.15CuHNO3 + 0.582Feox + 0.369CaAR 65.1 73.1Cd CdEDTA = �0.380 + 0.778CdHNO3 + 0.278TOC 74.2 76.7Mn MnEDTA = 4.30 + 0.444MnHNO3 + 0.355TOC � 0.281MgAR � 0.505CaAR 33.8 81.8Ni NiEDTA = 1.772 + 0.531NiHNO3 + 0.541TOC � 0.351CaAR 45 69.1Co CoEDTA = 3.355 + 0.635CoHNO3 � 0.510CaAR � 0.381MgAR + 0.341TOC 36.2 70.6As AsEDTA = �0.255 + 0.758AsHNO3 67.1

HAcPb PbHAc = �4.45 + 0.855PbHNO3 + 0.727CaAR 64.8 72Zn ZnHAc = �0.538 + 1.18ZnHNO3 89.1Cu CuHAc = �5.01 + 3.26lCuHNO3 71Cd CdHAc = �0.127 + 0.714CdHNO3 80.7Mn MnHAc = 1.39 + 0.331MnHNO3 + 0.318TOC 50.2 66.3Ni NiHAc = 0.122 + 0.602NiHNO3 74.8As AsHAc = �2.64 + 0.820AsHNO3 + 0.492CaAR 74.6 85.3

SBETPb PbSBET = �1.61 + 0.402PbHNO3 + 0.538CaAR 52.4 66.9Zn ZnSBET = �1.15 + 0.979ZnHNO3 + 0.228CaAR 92.4 93.6Cu CuSBET = �4.52 + 1.07CuHNO3 + 0.608CaAR + 0.328Feox 65.4 77.2Mn MnSBET = 1.10 + 0.442MnHNO3 + 0.258TOC 59.7 68.6Cr CrSBET = �1.70 + 0.419CaAR + 0.260CrHNO3 44.6 68.7

72 E. Kelepertzis, A. Argyraki / Applied Geochemistry 59 (2015) 63–73

5. Conclusions

The pseudototal, potentially phytoavailable, mobilizable, orallybioaccessible and reactive pools of PHEs were successfully deter-mined on 45 urban top-soil samples from Athens, Greece. Theresponse of chemical elements during the extraction tests com-bined with the variety of their speciation in soil are reflected onlarge differences in the observed extractability patterns. Despitethe elevated pseudototal concentrations of geogenic elements (Ni,Cr, As and Co) in Athens soil, their availability is limited becauseof their sequestration in stable mineral phases. Amorphous Mn oxi-des and TOC influence the available fractions of Mn, Ni and Co how-ever additional geochemical variables need to be identified for abetter prediction of their availability. The pseudototal content ofthe anthropogenic group of elements (Pb, Zn, Cu, Cd) is the pre-dominant factor controlling their availability. An associationbetween available fractions of this group of elements and amor-phous Fe oxides has also been observed. The applicability of diluteHNO3 for evaluating the availability of PHEs needs furtherinvestigation in order to clarify its effectiveness with respect to ele-ments of diverse origin and speciation in urban soil.

Acknowledgments

This project was funded by the John S. Latsis Public BenefitFoundation (Scientific Projects 2013, Grant Number: 70/4/12000).The sole responsibility for the content lies with its authors. TheMSc student Dimitrios Katritsis is sincerely thanked for his meticu-lous experimental work within the frame of his Master Thesis inApplied Environmental Geology course at the Laboratory ofEconomic Geology and Geochemistry, University of Athens.Special thanks are given to Principal Lecturers Nick Walton andMike Fowler for their involvement in chemical determinations ofgeochemical solutions at the analytical facilities of School ofEarth and Environmental Sciences, University of Portsmouth, UK.Finally, we acknowledge the two anonymous reviewers for theirconstructive comments that clarified the statistical analysis ofour work.

References

Alves, S., Trancoso, M.A., Goncalves, M., Correia dos Santos, M.M., 2011. A nickelavailability study in serpentinized areas of Portugal. Geoderma 164, 155–163.

Appleton, J.D., Cave, M.R., Wragg, J., 2012a. Anthropogenic and geogenic impacts onarsenic bioaccessibility in UK topsoils. Sci. Total Environ. 435–436, 21–29.

Appleton, J.D., Cave, M.R., Wragg, J., 2012b. Modelling lead bioaccessibility in urbantopsoils based on data from Glasgow, London, Northampton and Swansea, UK.Environ. Pollut. 171, 265–272.

Argyraki, A., Kelepertzis, E., 2014. Urban soil geochemistry in Athens, Greece. Theimportance of local geology in the distribution of potentially harmful traceelements. Sci. Total Environ. 482–483, 366–377.

Bouyoucos, G.J., 1962. Hydrometer method improved for making particle sizeanalysis of soils. Agron. J. 54, 464–465.

Cannon, W.F., Horton, J.D., 2009. Soil geochemical signature of urbanization andindustrialization – Chicago, Illinois, USA. Appl. Geochem. 24, 1590–1601.

Cattell, R.B., 1966. The scree test for the number of factors. Multivar. Behav. Res. 1,245–276.

Cave, M.R., Wragg, J., Harrison, H., 2013. Measurement modeling and mapping ofarsenic bioaccessibility in Northampton, United Kingdom. J. Environ. Sci.Health, Part A 48, 629–640.

Cox, S., Chelliah, M.C.M., McKinley, J.M., Palmer, S., Ofterdinger, U., Young, M.E.,Cave, M.R., Wragg, J., 2013. The importance of solid-phase distribution on theoral bioaccessibility of Ni and Cr in soils overlying Palaeogene basalt lavas,Northern Ireland. Environ. Geochem. Health 35, 553–567.

Das, S., Jean, J.-S., Kar, S., 2013. Bioaccessibility and health risk assessment of arsenicin arsenic-enriched soils, central India. Ecotoxicol. Environ. Saf. 92,252–257.

De Miguel, E., Mingot, J., Chacón, E., Charlesworth, S., 2012. The relationshipbetween soil geochemistry and the bioaccessibility of trace elements inplayground soil. Environ. Geochem. Health 34, 677–687.

Gasparatos, D., Mavromati, G., Kotsovilis, P., Massas, I., in press. Fractionation ofheavy metals and evaluation of the environmental risks for the alkaline soils ofthe Thriassio plain: a residential, agricultural and industrial area in Greece.Environ. Earth Sci., doi: http://dx.doi.org/10.1007/s12665-015-4096-1.

Gupta, S.K., Vollmer, M.K., Krebs, R., 1996. The importance of mobile, mobilisableand pseudo total heavy metal fractions in soil for three-level risk assessmentand risk management. Sci. Total Environ. 178, 11–20.

Houba, V.J.G., Lexmond, T.M., van der Novozamsky, I., Lee, J.J., 1996. State of the artand future developments in soil analysis for bioavailability assessment. Sci.Total Environ. 178, 21–28.

IBM Corp. Released, 2013. IBM SPSS Statistics for Windows, Version 22.0. Armonk,NY, IBM Corp.

ISO 10390, 1994. Soil quality-determination of pH.Kelepertzis, E., 2014. Investigating the sources and potential health risks of

environmental contaminants in the soils and drinking waters from the ruralclusters in Thiva area (Greece). Ecotoxicol. Environ. Saf. 100, 258–265.

Kelepertzis, E., Stathopoulou, E., 2013. Availability of geogenic heavy metals in soilsof Thiva town (central Greece). Environ. Monit. Assess. 185, 9603–9618.

E. Kelepertzis, A. Argyraki / Applied Geochemistry 59 (2015) 63–73 73

Li, Y., Zhang, M.-k., 2013. A comparison of physiologically based extraction test(PBET) and single-extraction methods for release of Cu, Zn, and Pb from mildlyacidic and alkali soils. Environ. Sci. Pollut. Res. 20, 3140–3148.

Luo, X.-s., Yu, S., Li, X.-d., 2011. Distribution, availability, and sources of trace metalsin different size fractions of urban soils in Hong-Kong: implications forassessing the risk to human health. Environ. Pollut. 159, 1317–1326.

Luo, X.-s., Yu, S., Li, X.-d., 2012. The mobility, bioavailability and humanbioaccessibility of trace metals in urban soils of Hong Kong. Appl. Geochem.27, 995–1004.

Madrid, F., Biasioli, M., Ajmone-Marsan, F., 2008. Availability and bioaccessibility ofmetals in fine particles of some urban soils. Arch. Environ. Contam. Toxicol. 55,21–32.

Massoura, S.T., Echevarria, G., Becquer, T., Ghanbaja, J., Leclerc-Cessac, E., Morel, J.-L., 2006. Control of nickel availability by nickel bearing minerals in natural andanthropogenic soils. Geoderma 136, 28–37.

Mingot, J., De Miguel, E., Chacón, E., 2011. Assessment of oral bioaccessibility ofarsenic in playground soil in Madrid (Spain). At three-method comparison andimplications for risk assessment. Chemosphere 84, 1386–1391.

Morman, S.A., Plumlee, G.S., Smith, D.B., 2009. Application of in vitro extractionstudies to evaluate element bioaccessibility in soils from a transect across theUnited States and Canada. Appl. Geochem. 24, 1454–1463.

Oomen, A.G., Hack, A., Minekus, M., Zeijdner, E., Cornelis, C., Schoeters, G.,Verstraete, W., Van de Wiele, T., Wragg, J., Rompelberg, C.J.M., Sips, A.J.A.M.,Van Wijner, J.H., 2002. Comparison of five in vitro digestion models to study thebioaccessibility of soil contaminants. Environ. Sci. Technol. 36, 3326–3334.

Palmer, S., Cox, S.F., McKinley, J.M., Ofterdinger, U., 2014. Soil geochemical factorscontrolling the distribution and oral bioaccessibility of nickel, vanadium andchromium in soil. Appl. Geochem. 51, 255–267.

Pelfrêne, A., Waterlot, C., Mazzuca, M., Nisse, C., Cuny, D., Richard, A., Denys, S.,Heyman, C., Roussel, H., Bidar, G., Douay, F., 2012. Bioaccessibility of traceelements as affected by soil parameters in smelter-contaminated soils: astatistical approach. Environ. Pollut. 160, 130–138.

Poggio, L., Vršcaj, B., Schulin, R., Hepperle, E., Marsan, F.A., 2009. Metals pollutionand human bioaccessibility of topsoils in Grugliasco (Italy). Environ. Pollut. 157,680–689.

Popescu, I., Biasioli, M., Ajmone-Marsan, F., Stanescu, R., 2013. Lability of potentiallytoxic elements in soils affected by smelting activities. Chemosphere 90, 820–826.

Quevauviller, P., Rauret, G., Rubio, R., López-Sánchez, J.-F., Ure, A., Bacon, J., Muntau,H., 1997. Certified reference materials for the quality control of EDTA-and aceticacid-extractable contents of trace elements in sewage sludge amended soils(CRMs 483 and 484). Fresenius J. Anal. Chem. 357, 611–618.

Rawlings, J.O., Pantula, S.G., Dickey, D.A., 1998. Applied Regression Analysis: AResearch Tool, second ed. Springer-Verlag, New York.

Reimann, C., Filzmoser, P., Garrett, R.G., 2002. Factor analysis applied to regionalgeochemical data: problems and possibilities. Appl. Geochem. 17, 185–206.

Reimann, C., Filzmoser, P., Garrett, R., Dutter, R., 2008. Statistical Data AnalysisExplained: Applied Environmental Statistics with R. Wiley-Blackwell,Chichester.

Reis, A.P., Patinha, C., Wragg, J., Dias, A.C., Cave, M., Sousa, A.J., Batista, M.J., Prazeres,C., Costa, C., Ferreira da Silva, E., Rocha, F., Urban geochemistry of lead in

gardens, playgrounds and schoolyards of Lisbon, Portugal: assessing exposureand risk to human health. Appl. Geochem. 44, 45–53.

Rodrigues, S.M., Henriques, B., Ferreira da Silva, E., Pereira, M.E., Duarte, A.C.,Römkens, P.F.A.M., 2010. Evaluation of an approach for the characterization ofreactive and available pools of twenty potentially toxic elements in soils: Part I– The role of key soil properties in the variation of contaminants’ reactivity.Chemosphere 81, 1549–1559.

Rodrigues, S.M., Cruz, N., Coelho, C., Henriques, B., Carvalho, L., Duarte, A.C., Pereira,E., Römkens, P.F.A.M., 2013. Risk assessment for Cd, Cu, Pb, and Zn in urbansoils: chemical availability as the central concept. Environ. Pollut. 183, 234–242.

Rodríguez Martín, J.A., De Arana, C., Ramos-Miras, J.J., Gil, C., Boluda, R., 2015.Impact of 70 years urban growth associated with heavy metal pollution.Environ. Pollut. 196, 156–163.

Römkens, P., Guo, H.-Y., Chu, C.-L., Liu, T.-S., Chiang, C.-F., Koopmans, G., 2009.Characterization of soil heavy metal pools in paddy fields in Taiwan: chemicalextraction and solid-solution partitioning. J. Soils Sediments 9, 216–228.

Roussel, H., Waterlot, C., Pelfrêne, A., Pruvot, C., Mazzuca, M., Douay, F., 2010. Cd, Pband Zn oral bioaccessibility of urban soils contaminated in the past byatmospheric emissions from two lead and zinc smelters. Arch. Environ.Contam. Toxicol. 58, 945–954.

Ruby, M.V., Davis, A., Schoof, R., Eberle, S.S.C.M., 1996. Estimation of lead andarsenic bioavailability using a physiologically based extraction test. Environ.Sci. Technol. 30, 422–430.

Sahuquillo, A., Rigol, A., Rauret, G., 2003. Overview of the use of leaching/extractiontests for risk assessment of trace metals in contaminated soils and sediments.Trends Anal. Chem. 22, 152–159.

Schwertmann, U., 1964. Differenzierung der Eisenoxide des Bodens durchphotochemische Extraktion mit saurer Ammoniumoxalat-Lösung. ZPflanzenernaehr Bodenkd 105, 194–202.

Shi, G., Chen, Z., Xu, S., Zhang, J., Wang, L., Bi, C., Teng, J., 2008. Potentially toxicmetal contamination of urban soils and roadside dust in Shanghai, China.Environ. Pollut. 156, 251–260.

Sialelli, J., Urquhart, G.J., Davidson, C.M., Hursthouse, A.S., 2010. Use of aphysiologically based extraction test to estimate the human bioaccessibilityof potentially toxic elements in urban soils from the city of Glasgow, UK.Environ. Geochem. Health 32, 517–527.

Ure, A.M., Quevauviller, P., Muntau, H., Griepink, B., 1993. Speciation of heavymetals in soils and sediments. An account of the improvement andharmonization of extraction techniques undertaken under the auspices of theBCR of the Commission of the European Communities. Int. J. Environ. Anal.Chem. 51, 135–151.

Villalobos, M., Bargar, J., Sposito, G., 2005a. Trace metal retention on biogenicmanganese oxide nanoparticles. Elements 1, 223–226.

Villalobos, M., Bargar, J., Sposito, G., 2005b. Mechanisms of Pb(II) sorption on abiogenic manganese oxide. Environ. Sci. Technol. 39, 569–576.

Walkley, A., Black, I.A., 1934. An examination of the Degtjareff method fordetermining soil organic matter and a proposed modification of the chromicacid titration method. Soil Sci. 37, 29–37.