Chemosphere 138 (2015) 96–103

Contents lists available at ScienceDirect

Chemosphere

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

Prediction of the fate of Hg and other contaminants in soil arounda former chlor-alkali plant using Fuzzy Hierarchical Cross-Clusteringapproach

http://dx.doi.org/10.1016/j.chemosphere.2015.05.0700045-6535/� 2015 Elsevier Ltd. All rights reserved.

⇑ Corresponding author at: 11 Arany Janos Street, 400028 Cluj-Napoca, Romania.E-mail address: csarbu@chem.ubbcluj.ro (C. Sârbu).

Tiberiu Frent�iu, Michaela Ponta, Costel Sârbu ⇑Faculty of Chemistry and Chemical Engineering, Babes-Bolyai University, Arany Janos 11, 400028 Cluj-Napoca, Romania

h i g h l i g h t s

� Fuzzy clustering of the Hg contaminated soil samples and associated characteristics.� Hg fractionation as water leachable, mobile, semi-mobile and non-mobile species.� Soil samples clustering based on the nature of contaminant and contamination level.� In-deep understanding of fate of Hg species in soil in the surrounding of a chlor-alkali plant.

a r t i c l e i n f o

Article history:Received 14 January 2015Received in revised form 9 May 2015Accepted 19 May 2015Available online 4 June 2015

Keywords:Hg fate in soilHg speciationChlor-alkali industryFuzzy Hierarchical Cross-Clustering

a b s t r a c t

An associative simultaneous fuzzy divisive hierarchical algorithm was used to predict the fate of Hg andother contaminants in soil around a former chlor-alkali plant. The algorithm was applied on several nat-ural and anthropogenic characteristics of soil including water leachable, mobile, semi-mobile,non-mobile fractions and total Hg, Al, Ba, Ca, Cr, Cu, Fe, K, Li, Mg, Mn, Na, Sr, Zn, water leachable fractionof Cl�, NO3

� and SO42�, pH and total organic carbon. The cross-classification algorithm provided a divisive

fuzzy partition of the soil samples and associated characteristics. Soils outside the perimeter of the for-mer chlor-alkali plant were clustered based on the natural characteristics and total Hg. In contaminatedzones Hg speciation becomes relevant and the assessment of species distribution is necessary. Thedescending order of concentration of Hg species in the test site was semi-mobile > mobile >non-mobile > water-leachable. Physico-chemical features responsible for similarities or differencesbetween uncontaminated soil samples or contaminated with Hg, Cu, Zn, Ba and NO3

� were also high-lighted. Other characteristics of the contaminated soil were found to be Ca, sulfate, Na and chloride, someof which with influence on Hg fate. The presence of Ca and sulfate in soil induced a higher water leach-ability of Hg, while Cu had an opposite effect by forming amalgam. The used algorithm provided anin-deep understanding of processes involving Hg species and allowed to make prediction of the fate ofHg and contaminants linked to chlor-alkali-industry.

� 2015 Elsevier Ltd. All rights reserved.

1. Introduction

Mercury in any form, as elemental species, organic and inor-ganic compounds, is regarded as one of the most toxic elementswith high impact on the environment and human health (Graniteet al., 2015; Amos et al., 2014; Horvat et al., 2014, 2013; Kalishet al., 2014; Llop et al., 2014; Visnjevec et al., 2014). The main nat-ural sources of mercury include volcanoes and volatilization fromoceans, while the anthropogenic origin is associated to local

fossil-fuel fired power plants, mining (cinnabar, gold and complexores), chlor-alkali and cement plants, waste incinerators, or diffusesources such as road traffic (Pirrone et al., 2010). The contributionof mercury released from contaminated sites to the global mercurybudget was summarized for the first time by Kocman et al. (2013).Studies on the contamination of soil and river sediment and Hgtransfer to plants were conducted on several areas related to Hgmining and chlor-alkali plants in Spain (Fernandez-Martinezet al., 2014; Higueras et al., 2013; Rocio et al., 2013; Schmidet al., 2013), Slovenia (Miklavcic et al., 2013), United States ofAmerica (Donovan et al., 2014; Richardson et al., 2013), China(Meng et al., 2014; Dai et al., 2013; Qiu et al., 2013), Netherlands

Fig. 1. Soil sampling sites.

T. Frent�iu et al. / Chemosphere 138 (2015) 96–103 97

(Bernaus et al., 2006), Italy (Vaselli et al., 2013) and South-Africa(Lusilao-Makiese et al., 2013). These studies highlightedhot-spots around the contamination source with elevated Hg levelup to 5350 mg kg�1 in soil, 2580 mg kg�1 in sediment and50 lg m�3 in air. In contaminated areas it was observed an ele-vated Hg accumulation in vegetable foodstuffs, up to 20 lg kg�1

in tomato and 80 lg kg�1 in cabbage (Inacio et al., 2014), 7 lg kg�1

in carrots (Ding et al., 2014) and 584 lg kg�1 in rice, of which up to132 lg kg�1 as methylmercury (Meng et al., 2014) resulting in ahigh risk of exposure through food consumption. The Hg transferfrom soil to plants is controlled by pH, free Al oxide and level ofwater leachable fraction of Hg rather than its total concentration(Ding et al., 2014; Inacio et al., 2014; Meng et al., 2014).

Methods for Hg speciation in soil based on 3 or 4 step-selectiveextractions (Fernandez-Martinez et al., 2014; Lusilao-Makieseet al., 2013; Neculita et al., 2005; USA EPA 3200 method, 2005)or non destructive techniques, such as X-ray absorption spec-troscopy (Bernaus et al., 2006) revealed semi-mobile andnon-mobile Hg species as major fractions. The water leachablefraction with the highest toxicity was much lower.

Evaluation of results for contaminated sites is commonly madeby relating to quality standards in force and at most using a statis-tical approach such as Pearson’s correlation, Principal ComponentAnalysis (PCA) and Cluster Analysis (CA) tools (Abollino et al.,2011; Zhou et al., 2007). Although these statistical approachesenable recognition of the main contributors to variability of thesites, they do not provide a simultaneous classification of samplesand associated characteristics to identify the specific contaminantsand the dominant factors predicting their fate. Different Fuzzyalgorithms were always designed in order to improve the patternrecognition for partitioning variables but not yet used for sitesand contaminants classification (Niros et al., 2015;Hemmateenejad et al., 2013; Bang and Lee, 2011).

The novelty of this work consists in providing an in-deep under-standing of the fate of Hg and other contaminants in soil around aformer chlor-alkali plant using Fuzzy Hierarchical Cross-Clustering(FHCC), a tool designed for the first time by Sârbu and Pop (2000)and Sârbu et al., 1996 and used for modeling and predicting thefate of contaminants near an abandoned uranium mine inGermany (Pourjabbar et al., 2014). The capability and effectivenessof the FHCC tool was demonstrated within a prospective studyconducted in the surroundings of a former mercury-cellchlor-alkali plant in Turda town, north-western Romania. In a pre-vious study, Frentiu et al. (2013b) used PCA and CA tools to char-acterize soil in the same area, but these approaches were notable to predict the fate of Hg and other associated contaminants.To recognize the physico-chemical features responsible for thesimilarities or differences between groups of soil samples andmake prediction about Hg fate the FHCC algorithm was conductedon data related to total mercury, Hg species (water leachable,mobile, semi-mobile and non-mobile), elemental composition(Al, Ba, Ca, Cr, Cu, Fe, K, Li, Mg, Mn, Na, Sr and Zn), water leachableanions (Cl�, NO3

� and SO42�), soil pH and total organic carbon (TOC).

2. Materials and methods

2.1. Site description and sampling

The Chemical Plant in Turda, a town in north-western Romania,was founded in 1911 to produce sodium hydroxide and chloridefrom mercury-cell chlor-alkali electrolysis, hydrochloric acid, cop-per and zinc pesticides, calcium hypochlorite and several inorganicsalts. After 1998, the plant activity has ceased and the industrialfacilities have been partially demolished or abandoned. Study ofresidual pollution is needed since no measures have been

Fig. 2. Scheme of soil sample preparation for the determination and speciation ofHg.

98 T. Frent�iu et al. / Chemosphere 138 (2015) 96–103

undertaken aiming soil decontamination. To predict Hg fate andother contaminants a number of 38 soil samples were collectedwithin the test site from the topsoil layer (5–20 cm depth) from3 locations (Fig. 1): (i) former chlor-alkali plant (7 samples; sites16–22); (ii) former waste landfills (5 samples; sites 25, 29, 35–37) and (iii) residential area (26 samples). The coordinates of sam-pling points were recorded using a 310 Magellan GPS. The positioncorresponding to former pollution sources are indicated on themap.

2.2. Soil sample preparation for Hg determination and speciation

Soil samples were air dried after collection to avoid Hg loss byvaporization. Moisture was determined on a parallel sample driedin oven at 105 ± 5 �C for 3 h. After drying, samples were crushedand sieved to <2 mm to remove stones and roots and the fraction<2 mm was ground and sieved to <250 lm. An amount of250 mg sample was subjected to microwave assisted mineraliza-tion in a MW3S + Berghof microwave digester (Berghof,Germany) with 12 mL aqua regia following a previously establishedprotocol (Frentiu et al., 2013a). The resulted digest was filtered,diluted to 100 mL and used for total Hg determination.

The analytical protocol for Hg speciation involved sequentialextractions to establish the distribution as water leachable, mobile,semi-mobile and non-mobile species (Fig. 2). The Hg water leach-able fraction was extracted following a standard procedure givenin (SR EN 12457-1, 2003). The leaching of Hg was performed in aREAX 20 overhead mixer Heidolph (Schwabach, Germany) on

wet soil sample (<4 mm) corresponding to 175 g dry mass at asolid-to-liquid ratio of 1:2.

Fractions of species exhibiting different solubility were sepa-rated by a 3-step sequential extraction procedure 3200 (US EPA,2005).

The sequential extraction was conducted on 1.5 g test soil sam-ple (<250 lm). After each extraction the supernatant was sepa-rated by centrifugation and the residue was washed with 5 mLultrapure water. The extracts and rinse were combined and usedfor analysis. A detailed description of the extraction procedurewas provided in Frentiu et al. (2013b). Extractions were carriedout in an ultrasonic bath Nahita 610/10 (Navara, Spain). Aliquotvolumes of extracts were subjected to oxidation with 0.5 mL 10%(v/v) BrCl and the excess was reduced with 0.5 mL solution of12% hydroxylamine hydrochloride. Then samples were diluted to50 mL in 5% (v/v) HCl. Cold vapor generation capacitively coupledplasma microtorch optical emission spectrometry (CV-lCCP-OES)was used for Hg determination (Frentiu et al., 2013a). The methodconsisted in chemical derivatization of Hg2+ from aqueous solutionto CV by mixing sample with 20% (w/v) SnCl2 in 15% (v/v) HCl med-ium in the CV generator, purging of Hg vapor via an Ar streamtoward microplasma (operated at 10 W) and emission measure-ment at 253.652 nm using a low resolution QE65 ProSpectrometer, Ocean Optics (Dunedin, USA).

2.3. Soil sample preparation to determine chemical composition andgeneral characteristics

Total metal content was determined in the solution resulted afteraqua regia digestion by inductively coupled plasma optical emissionspectrometry (ICP-OES) using a Spectro CirosCCD instrument (Kleve,Germany). The concentrations of anions in water leachate weredetermined using a 761 Compact IC Methrom ion chromatograph(Herisau, Switzerland) following a standard procedure (ISO10304-1, 2007), while pH in a 1:5 suspension of soil in water (ISO10390, 2005) with a 350i Multiparameter WTW (Wilhelm,Germany). TOC was determined according to the Schollenberger’sprocedure (1945) based on oxidation with a K2Cr2O7 in H2SO4 solu-tion and back titration of oxidant excess with a Mohr’s salt solution.

2.4. Quality assurance/Quality control procedures

Five topsoil sub-samples (2 m distance) were collected at eachsampling site for a composite sample of around 2 kg. The compos-ite sample was placed in a labeled plastic bag, transported to lab-oratory, homogenized and prepared for physical and chemicalanalysis. Three parallel samples were run to evaluate the standarddeviation of repeatability.

The CV-lCCP-OES method was previously validated for total Hgdetermination in microwave digested soil samples and comparedwith cold vapor atomic fluorescence spectrometry (CV-AFS)(Frentiu et al., 2013a). Analysis of certified reference materials(CRMs) of soil containing 0.22–99.8 mg kg�1 Hg revealed recoveryof 97 ± 7% for 95% confidence interval. The Bland and Altman testcarried out on 31 test soil samples containing 0.26–55.30 mg kg�1

Hg showed no statistical difference against CV-AFS for 95% confi-dence level. The method detection limit (MDL) and practical quan-tification limit (PQL) for total Hg considering the whole protocolwere 0.005 and 0.015 mg kg�1 respectively. MDLs and PQLs forHg corresponding to sequential extraction protocols were:3 � 10�5 and 15 � 10�5 mg kg�1 (water leachable) and 0.0001 and0.0003 mg kg�1 (mobile, semi-mobile and non-mobile). The overallrelative standard deviation found from pooled standard deviationsfor Hg determination was in the range 0.3–15.0% (n = 3 parallelsamples). As there are no certified reference materials of soil to val-idate the speciation scheme, the sequential extraction procedure

Table 1Concentrations (mg kg�1) of total Hg and water-leachable, mobile, semi-mobile and non-mobile fractions in soil.

Sample Total mercury in aqua regiaa Water leachablea Mobile fractiona Semimobile fractiona Non-mobile fractiona Fraction sumb

1 0.72 0.0080 0.034 0.39 0.230 0.652 0.10 0.0050 0.008 0.08 0.003 0.093 0.46 0.0080 0.040 0.26 0.120 0.424 0.47 0.0130 0.053 0.39 0.041 0.485 0.24 0.0070 0.012 0.22 0.022 0.256 1.99 0.0009 0.070 1.42 0.570 2.067 0.13 0.0090 0.022 0.06 0.043 0.138 0.67 0.0030 0.009 0.57 0.073 0.659 1.27 0.0060 0.094 0.79 0.370 1.2510 0.45 0.0070 0.008 0.34 0.041 0.3911 0.23 0.0003 0.028 0.17 0.035 0.2312 0.38 0.0060 0.019 0.33 0.022 0.3713 0.35 0.0090 0.024 0.25 0.120 0.3914 0.08 0.0040 0.018 0.04 0.036 0.0915 0.42 0.0040 0.022 0.25 0.140 0.4116 85.50 0.0170 5.600 65.60 20.200 91.4017 92.60 0.0250 31.200 33.30 20.000 84.5018 26.30 0.3700 4.780 16.70 5.200 26.7019 19.30 0.0070 0.850 20.10 1.790 22.7020 114.00 0.0510 7.800 104.00 1.760 114.0021 35.90 0.0190 0.220 42.70 0.500 43.4022 54.80 1.2000 9.310 31.30 6.180 46.8023 0.57 0.0110 0.017 0.37 0.067 0.4524 0.09 0.0070 0.015 0.06 0.015 0.0925 16.80 0.0890 0.540 15.5 2.920 19.0026 0.70 0.0040 0.020 0.57 0.150 0.7427 0.28 0.0220 0.036 0.23 0.032 0.3028 0.20 0.0060 0.017 0.12 0.090 0.2329 4.44 0.0970 0.210 3.27 0.240 3.7230 0.41 0.0110 0.012 0.36 0.046 0.4231 0.12 0.0030 0.021 0.07 0.035 0.1332 0.42 0.0070 0.021 0.15 0.160 0.3333 0.11 0.0040 0.014 0.07 0.021 0.1134 0.84 0.0050 0.098 0.48 0.240 0.8235 20.00 0.0130 0.810 18.10 0.640 19.6036 6.46 0.0130 0.530 4.87 0.450 5.8537 8.88 0.0010 0.390 7.80 1.000 9.1938 0.62 0.0070 0.042 0.51 0.120 0.67

Mean 13.09 0.0547 1.658 9.78 1.677 13.13Median 0.60 0.0070 0.035 0.39 0.130 0.56Std. Dev. 27.78 0.2004 5.385 21.27 4.603 27.57Min 0.07 0.0003 0.008 0.04 0.003 0.09Max 114.00 1.2000 31.200 104.00 20.200 114.00

a n = 3 replicates.b Sum of mobile, semi-mobile and non-mobile fractions.

T. Frent�iu et al. / Chemosphere 138 (2015) 96–103 99

and analytical technique was assessed using the Bland and Altmantest by comparing the sum of the Hg fractions with total Hgextracted in aqua regia (Frentiu et al., 2013b).

The ICP-OES technique was validated for metals determinationwith recovery in the range 91–109% and 0.8–4.6% repeatability. Inthis study, Cd, Co, Ni and Pb were not considered, since in most ofsamples their contents were below the detection limits in ICP-OES.

The chromatographic method for anions quantification was val-idated by analyzing SPS-NUTRWW1 Waste water (Oslo, Norway).For certified values (mg L�1) of 5.00 ± 0.05 Cl�, 1.00 ± 0.01 NO3

�

and 20.0 ± 0.2 SO42� recovery was 99.2 ± 2.0, 100.0 ± 2.0 and

101.0 ± 2.0 respectively, for n = 5 and 95% confidence interval.The detection limits (mg kg�1) were: 0.70 Cl�; 0.05 NO3

� and 0.90SO4

2�, while the corresponding standard deviations of repeatability(n = 3 replicate) were (%): 4.6–10.0; 3.1–8.0 and 3.6–7.1.

2.5. Fuzzy Hierarchical Cross-Clustering

The application of Fuzzy sets in a clustering function causes theclass membership to become a relative one and consequently anobject or sample can belong to several classes at the same timebut with different degrees. The FHCC approach allows the

identification of qualitative and quantitative characteristicsresponsible for the observed similarities and dissimilaritiesbetween samples, as well as the association of characteristics ateach division level of the hierarchy. The method is the straightfor-ward way of developing a hierarchical algorithm that should use ateach node of the clustering tree the algorithm described in (Popand Sârbu, 1997). The nodes of the tree are labeled with a pair(C, D), where C is a fuzzy set from a fuzzy partition of objectsand D is a fuzzy set from a fuzzy partition of characteristics. Theroot node corresponds to the pair (X, Y). In the first step the twosub-nodes (A1, B1) and respectively (A2, B2) will be computed byusing the cross-classification algorithm. These two nodes will beeffectively built only if the fuzzy partitions {A1, A2} and {B1, B2}describe real clusters. For each of the terminal nodes of the tree,partitions having the form {A1, A2} and {B1, B2} are determinedand in this way the binary clustering tree is extended with twonew nodes, (A1, B1) and (A2, B2). The process continues until forany terminal node it is no longer possible to obtain a structure ofreal clusters, either for a set of objects, or of characteristics. Thefinal fuzzy partitions will contain the fuzzy sets corresponding tothe terminal nodes of the binary classification tree. The advantagesof the FHCC algorithm include the ability to observe, on the one

Table 2Total concentration of metals in soil, concentration of water-leachable anions, pH and TOC (n = 3 replicates).

Sample pH TOCa(%) Total metal content Water leachable content(mg kg�1)

(%) (mg kg�1)

Al Ca Fe K Mg Mn Na Ba Cr Cu Li Sr Zn Cl� NO3� SO4

2�

1 8.0 2.78 2.72 2.70 2.75 1.51 0.69 0.07 0.08 197.0 28.4 51.7 55.3 189 156 21.50 43.50 19.22 9.1 1.29 1.48 1.82 2.49 0.63 0.54 0.07 0.04 116.0 45.4 48.5 21.7 63 146 3.02 3.34 23.03 8.1 0.24 1.23 4.56 1.95 0.48 0.51 0.07 0.06 160.0 2.0b 165. 28.3 168 590 14.00 150.00 23.44 9.3 0.55 1.14 2.76 3.15 0.55 0.61 0.08 0.03 131.0 32.5 54.1 22.9 114 208 3.33 6.69 24.95 8.9 1.52 1.93 4.58 2.86 1.14 0.97 0.09 0.06 86.1 48.2 45.9 53.2 199 139 3.41 149.00 13.56 9.2 1.35 1.21 6.26 1.98 0.77 0.48 0.07 0.04 174.0 39.2 67.7 15.3 137 303 6.58 36.30 30.17 8.3 1.31 1.52 4.15 2.15 1.03 0.54 0.09 0.06 178.0 51.1 58.0 23.6 143 180 7.41 2.44 38.68 8.3 1.46 1.52 3.34 2.07 0.79 0.50 0.07 0.07 152.0 55.9 85.9 16.1 137 200 6.89 24.90 31.39 8.8 1.68 1.29 6.01 2.07 0.69 0.48 0.06 0.09 160.0 36.7 76.4 91.2 108 366 2.71 0.19 19.110 8.6 1.32 1.60 3.95 2.58 0.93 0.66 0.08 0.05 169.0 48.1 66.0 30.2 141 253 3.53 0.20 25.111 8.5 0.85 1.19 2.39 2.35 0.48 0.60 0.08 0.03 112.0 33.6 47.3 23.7 58 162 4.84 3.29 24.212 8.9 0.29 0.81 5.38 1.69 0.56 0.48 0.07 0.04 127.0 32.6 57.5 11.0 137 166 9.03 0.19 14.213 8.8 1.64 1.06 5.59 1.98 0.48 0.47 0.05 0.06 84.7 42.6 90.3 9.6 97 236 7.75 2.70 48.714 8.0 1.26 1.72 4.05 2.84 0.93 0.97 0.08 0.05 82.7 33.3 46.0 50.9 176 117 4.59 2.64 14.915 8.1 0.33 2.43 3.99 2.57 0.56 0.62 0.08 0.06 225.0 2.0b 158.0 28.4 128 798 79.60 93.50 338.016 8.0 1.07 1.31 3.01 4.58 0.56 0.51 0.07 0.08 240.0 45.8 478.0 21.7 103 418 45.70 37.70 1339.017 8.3 1.20 1.21 2.91 2.99 0.41 0.48 0.07 0.09 127.0 11.3 127.0 27.4 110 223 7.19 89.90 370.018 8.3 0.33 0.83 6.58 1.52 0.36 0.45 0.04 1.24 78.3 5.25 54.7 16.6 140 77 12388 94.70 1405.019 8.2 0.47 1.20 7.25 2.32 0.52 0.49 0.06 0.10 168.0 17.9 294.0 21.3 114 234 4.04 46.40 151.020 8.5 0.13 1.00 5.27 2.35 0.37 0.52 0.07 0.07 122.0 11.7 120.0 27.1 127 180 217.00 71.00 187.021 8.4 1.19 1.39 9.24 2.36 0.66 0.65 0.08 0.12 151.0 24.5 209.0 33.5 174 598 49.10 21.20 339.022 8.2 1.77 0.86 13.4 1.67 0.59 0.39 0.04 0.19 92.0 17.3 111.0 34.0 163 76 363.00 40.80 336.023 8.9 1.28 1.06 3.15 2.01 0.58 0.44 0.07 0.03 139.0 42.3 75.4 10.7 90 171 152.00 61.80 203.024 8.7 1.49 0.55 3.81 1.95 0.19 0.30 0.06 0.05 92.7 58.3 80.7 10.5 71 211 5.69 6.86 43.325 8.3 1.25 1.06 2.27 2.27 0.56 0.53 0.06 0.03 101.0 31.5 80.0 18.8 76 188 2.52 0.14 21.326 8.8 1.65 1.28 3.25 2.53 0.60 0.49 0.06 0.06 138.0 53.3 70.1 13.4 81 241 9.47 0.19 58.727 7.7 0.68 1.58 1.15 2.79 0.61 0.59 0.07 0.03 127.0 55.3 48.8 24.6 51 145 3.92 1.96 14.728 8.4 0.81 1.29 0.97 3.10 0.44 0.44 0.12 0.02 139.0 40.3 40.7 8.0 45 111 6.42 0.24 88.329 8.1 1.64 1.16 2.84 2.16 0.50 0.54 0.06 0.04 95.2 35.6 67.1 13.4 61 182 2.94 13.70 24.330 8.4 1.40 1.97 1.28 2.54 0.54 0.59 0.04 0.04 156.0 51.9 45.8 27.4 49 139 3.49 0.19 22.931 7.8 1.01 1.26 0.94 2.79 0.49 0.34 0.14 0.02 133.0 44.4 31.3 6.31 43 108 2.79 0.20 13.332 8.2 1.55 1.17 2.00 2.50 0.45 0.58 0.10 0.02 106.0 24.1 63.9 14.4 50 212 3.33 0.20 15.133 8.2 1.44 1.51 1.04 3.15 0.49 0.41 0.10 0.03 145.0 47.2 31.8 10.0 43 93 3.79 0.60 13.234 8.5 0.15 2.10 0.95 2.64 0.50 0.41 0.06 0.06 150.0 2.0b 203.0 28.1 138 565 3.43 273.00 35.835 8.5 0.29 1.02 1.95 2.29 0.38 0.57 0.08 0.06 107.0 3.3 87.2 14.7 71 203 8.49 128.00 41.636 8.3 1.02 1.41 6.70 1.30 1.37 0.47 0.03 0.08 122.0 18.2 114.0 33.7 208 500 111.00 129.00 402.037 8.4 2.41 1.31 1.48 2.45 0.41 0.61 0.12 0.04 89.1 13.9 288.0 25.9 87 274 18.50 12.00 843.038 8.4 2.20 0.93 2.30 2.21 0.14 0.49 0.06 0.06 96.9 2.0b 197.0 28.1 138 505 14.40 260.00 78.1

Mean 8.4 1.17 1.35 3.82 2.42 0.61 0.54 0.07 0.09 133.0 34.7 106.0 25.0 111.0 255.0 358.00 47.60 177.0Median 8.4 1.26 1.26 3.20 2.36 0.55 0.51 0.07 0.06 129.0 35.2 75.9 23.2 111.0 202.0 7.04 17.50 37.2Std. Dev. 0.4 0.63 0.44 2.56 0.57 0.28 0.13 0.02 0.19 38.5 15.4 90.5 16.1 47.4 169.0 2006.0 69.40 331.0Min 7.7 0.13 0.13 0.94 1.30 0.14 0.30 0.03 0.02 78.3 2.0b 31.3 6.31 43.0 76.0 2.52 0.14 13.2Max 9.3 2.78 2.78 13.40 4.58 1.51 0.97 0.14 1.24 240.0 58.3 478.0 91.2 208.0 798.0 12388 273.00 1405.0

a Total organic carbon.b Limit of detection.

100 T. Frent�iu et al. / Chemosphere 138 (2015) 96–103

hand, the fuzzy partitions obtained and the relations betweenthem, and, on the other hand, the characteristics correspondingto each final partition of objects which have contributed to the sep-aration of the respective group.

3. Results and discussions

The concentration of total Hg and weights of fractions in soilsamples demonstrated a high variability (Table 1). The analysisshowed alkaline pH, low TOC and elevated concentrations of cal-cium, iron, aluminum, magnesium, potassium, sodium, and occa-sionally water-leachable sulfate and chloride (Table 2). Highconcentration of Cu and Zn were found in the area of the formerchlor-alkali plant (sites 16–22) and waste deposits (sites 34–38).

The hard partition corresponding to the fuzzy divisive partitionof the soil samples using fuzzy divisive hierarchicalcross-clustering (autoscaled data) was obtained by defuzzificationof fuzzy partition. The samples and characteristics are assigned tothe cluster with the highest membership degree and are ranked indescending order of their membership degree corresponding to the

final partition of soil samples and their characteristics(Supplementary material 1 and Supplementary material 2).Comparing the hard partitions considering all soil parameters(Table 3) or only those associated to Hg (Table 4) and the member-ship degrees to the final fuzzy partitions one can observe a goodagreement of results related to the nature and location of samples.Much more, the soil samples were better separated with member-ship degrees close to 1 for most samples when only Hg character-istics were considered.

Considering all parameters, soils were divided into 12 partitionlevels, while considering only parameters associated to mercurydivision partition is limited to 4 levels. In both cases the first levelwas split in two groups based on the Hg contamination level. Thefirst group A1 contained most of the soil samples collected outsidethe contaminated area for which the linked characteristics relate tonatural attributes and total Hg concentration. The second class A2

encompassed soil samples heavily contaminated by Hg and hencethe associated characteristics were the different Hg species. Mostof soil samples in class A1 exhibited low contamination with totalHg in the range 0.08–0.84 mg kg�1 not exceeding the alert

Table 3The hard fuzzy divisive partition of soil samples and their characteristics (ranked in descending order of their membership degree).

Partitionlevel

Fuzzydivisivepartition

Samples Associated characteristics

Location Membershipdegree range

Characteristica Membershipdegree range

0 A 1,. . ., 38 Hgaq, Hgw, Hgm, Hgsm, Hgnn, Hgs, pH, TOC, Ba, Cr, Cu, Li,Sr, Zn, Al, Ca, Fe, K, Mg, Mn, Na, Cl�, NO3

�, SO42�

1 A1 38 26 8 3 4 2 23 30 6 33 28 7 34 27 32 11 31 1314 10 37 29 9 36 25 12 35 15 24 1 5 19 21

0.998–0.518 Zn, Mg, K, TOC, Ba, pH, Al, Mn, Cr, Li, Fe, NO3�, Sr 0.999–0.557

A2 17 20 22 18 16 0.996–0.830 SO42�, Hgnn, Hgm, Hgw, Hgsm, Hgs, Ca, Hgaq, Na, Cl�, Cu 0.999–0.664

2 A11 14 34 3 9 36 15 10 7 1 5 21 38 0.933–0.512 Zn, K, Al, Mg, Ba, NO3�, Li, Sr 0.992–0.501

A12 23 32 26 11 33 2 28 13 31 25 29 4 12 27 37 24 358 30 6 19

0.987–0.665 pH, Mn, TOC, Cr, Fe 0.980–0.744

3 A111 34 3 15 38 21 0.883–0.359 Zn, NO3�, Ba 0.957–0.765

A112 14 9 10 5 1 7 36 0.852–0.471 K, Mg, Li, Al, Sr 0.974–0.378

4 A1121 9 10 7 36 0.684–0.436 K, Li, Al, Sr 0.971–0.298A1122 14 5 1 0.820–0.320 Mg 0.928

5 A11211 9 0.684 Li 0.812A11212 10 7 36 0.552–0.402 K, Al, Sr 0.966–0.280

6 A121 32 33 28 31 37 27 11 35 4 19 0.952–0.381 Mn, Fe 0.952; 0.767A122 23 13 26 2 8 24 30 12 6 29 25 0.978–0.494 pH, Cr, TOC 0.950–0.848

7 A1211 27 11 35 4 19 0.584–0.242 Mn 0.952A1212 28 31 37 32 33 0.802–0.458 Fe 0.676

8 A1221 8 30 29 25 0.599–0.431 TOC, Cr 0.810;0.731A1222 23 2 13 26 6 24 12 0.961–0.355 pH 0.949

9 A21 17 20 16 0.982–0.808 Hgnn, Hgm, Hgsm, Hgs, Hgaq, Cu 0.986–0.490A22 22 18 0.943–0.858 Hgw, Ca, Na, Cl�, SO4

2� 0.948–0.595

10 A211 17 16 0.964–0.692 Hgm, Hgnn, Cu 0.970–0.294A212 20 0.951 Hgsm, Hgs, Hgaq 0.913–0.886

11 A2111 17 0.964 Hgm 0.970A2112 16 0.692 Hgnn, Cu 0.946; 0.218

12 A221 22 0.940 Hgw, Ca 0.935–0.933A222 18 0.857 Na, Cl�, SO4

2� 0.918–0.534

a Hgaq – total Hg extracted in aqua regia; Hgw – fraction of water leachable species; Hgm – fraction of mobile species; Hgsm – fraction of semi-mobile species; Hgnn – fractionof non-mobile species; Hgs– sum of fractions; TOC – total organic carbon.

Table 4The hard divisive fuzzy partition of soil samples and their Hg characteristics (ranked in descending order of their membership degree).

Partitionlevel

Divisive fuzzypartition

Samples Associated characteristics

Location Membershipdegree range

Characteristica Membershipdegree range

0 A 1,. . ., 38 Hgaq, Hgw, Hgm, Hgsm,Hgnn, Hgs

1 A1 29 6 36 9 34 26 8 38 1 23 4 30 10 3 15 12 13 32 27 5 11 28 7 31 33 224 14 37 25 35 19 18 21

0.998–0.531 Hgs, Hgaq, Hgsm 0.992–0.918

A2 16 17 20 22 0.999–0.592 Hgnn, Hgm, Hgw 0.997–0.9902 A11 35 19 25 18 21 0.916–0.437 Hgsm 0.919

A12 9 34 26 8 1 38 23 4 3 30 10 15 12 13 32 27 5 11 28 6 7 31 33 2 24 1429 36 37

0.998–0.738 Hgs, Hgaq 0.978; 0.969

3 A121 36 37 29 0.908–0.666 Hgaq 0.970A122 4 23 3 30 10 15 38 8 12 13 32 1 26 27 5 11 28 34 7 31 33 2 24 14 9 6 0.998–0.892 Hgs 0.978

4 A21 17 0.975 Hgm, Hgnn 0.977; 0.974A22 20 16 22 0.907–0.590 Hgw 0.990

a Hgaq – total Hg extracted in aqua regia; Hgw – fraction of water leachable species; Hgm – fraction of mobile species; Hgsm – fraction of semi-mobile species; Hgnn – fractionof non-mobile species; Hgs– sum of fractions; TOC – total organic carbon.

T. Frent�iu et al. / Chemosphere 138 (2015) 96–103 101

threshold for soil of sensitive use (1 mg kg�1) given in theRomanian legislation, Ministerial Order No. 756 (1997). These soilspose any direct danger in terms of water leachable species(0.0003–0.0220 mg kg�1 Hg), which are below the limit of0.05 mg kg�1 for non-hazardous waste in the guideline on wasteacceptance criteria at landfills, Ministerial Order No. 95 (2005).The group A1 contained also soil samples with relatively high totalHg concentration (4.44–20.00 mg kg�1) coming from sites eitherunder the direct influence of the contamination source (sites 19,21) or situated in the zone of the former waste landfills (25, 29,

35–37). Both intervention threshold for Hg in soil and leachabilitylimit for non-hazardous wastes were exceeded for these samples.

The non-contaminated soil samples or containing<20.00 mg kg�1 Hg were associated with pH-independent naturalcharacteristics of soil such as K, Mg, Li and Sr-bearing aluminosil-icate (levels 2–5, partitions A11, A112, A1121, A1122, A11211, A11212,Table 3), but also with pH-sensitive characteristics like TOC, oxidiz-able organic species of Cr and reducible compounds of Mn and Fe(level 2, class A11; levels 6–8, partitions A121, A122, A1211, A1212,A1221 and A1222; Table 3). The fuzzy partition relating Hg species

102 T. Frent�iu et al. / Chemosphere 138 (2015) 96–103

(Table 4) showed that also the non-contaminated soil samples orwith Hg below 10 mg kg�1 had the total Hg as characteristicparameter (level 1, partition A1; level 2 partition A12; level 3 parti-tions A121 and A122). In addition, soil samples containing 10–20 mg kg�1 Hg exhibited as characteristic the Hg semi-mobile frac-tion (level 2, partition A11, Table 4).

Hierarchical cross-clustering allowed also the recognition of themost Hg polluted zone and clustering of the associated parametersof anthropogenic origin (Table 3, partition A2). The zone of the for-mer chemical plant was characterized by elevated concentration ofHg under different forms, Ca, Na, Cu, water-leachable Cl� and SO4

2�

(level 2, partition A2; levels 9,10,11, Table 3, and level 1, partitionA1 and level 4, partitions A21 and A22, Table 4). These soils werefound to exhibit the highest Hg contamination (19.30–114.00 mg kg�1 total Hg). The concentration of individual speciesin descending order was (mg kg�1): semi-mobile (16.70–104.00) > mobile (0.220–31.200) > non-mobile (0.500–20.200) > water-leachable (0.0070–1.200). In particular cases, soilcould be regarded as hazardous waste, since the leachability limitin water was exceeded. The soil characteristic related to high Caconcentration in this area was consistent with the anthropogenicinput from the production of calcium hypochlorite in the past.The coexistence of Na and chloride in the same cluster results fromthe use of NaCl as raw material for electrolysis. The Cu concentra-tion in soil beyond the intervention threshold (200 mg kg�1) wasthe consequence of Cu-pesticides production starting fromCuSO4. The water leachability of Hg from soil in the contaminatedarea was found to be influenced by Ca and SO4

2�, while Cl� and NO3�

had not a significant role. The association of the water leachablefraction of Hg with sulfate was in agreement with the results ofBernaus et al. (2006) who identified HgSO4 as the major waterleachable species from soil in the surrounding of a chlor-alkaliplant. The enhancement of water leachability of Hg in the presenceof Ca (level 12, partition A222, Table 3) was previously remarked insoil where calcium hypochlorite was used in strategy to remediateHg contamination (Renneberg and Dudas, 2002). The high concen-tration of Cu in soil resulted in the increase of the fraction ofnon-mobile Hg species (level 11, partition A2112, Table 3) likelydue to amalgam formation as observed also by Bernaus et al.(2006). As a general remark, the natural components of the soil likeorganic matter, Fe and Mn oxyhydroxides, minerals containing Al,K, Mg and Sr and soil pH did not influence the fate of Hg in soil. Thisis an evidence of the anthropogenic origin of Hg found in thesemi-mobile fraction mostly as elemental species.

Hierarchical cross-clustering highlighted a cluster of soil sam-ples contaminated with Zn as nitrate (level 3, partition A111,Table 3) of concentration close or above the alert threshold for soilof sensitive use (600 mg kg�1) given in Ministerial Order No. 756(1997). Elevated concentrations of Zn, NO3

� and Ba in these samplesare the result of using fertilizers and Zn and Ba containing pesti-cides as well as their production in the former chemical plant.

4. Conclusions

It was demonstrated the excellent ability of the fuzzy associativesimultaneous FHCC algorithm for the characterization and cluster-ing of soil samples and contaminants around a former chlor-alkaliplant. This approach allowed also the prediction of Hg fate in thezone, recognition of the most Hg contaminated sites and of natu-ral/anthropogenic characteristics responsible for the similarities ordifferences between groups of soil samples. Non- and low Hg con-taminated soils situated outside the perimeter of the chlor-alkaliplant were clustered based on natural characteristics (K, Mg and Sraluminosilicate, pH, Mn, Cr, Fe and TOC) and total Hg, while Hg spe-ciation had weak relevance. Highly contaminated soil samples

(>20 mg kg�1 Hg) collected in the perimeter of the former plantand waste deposits were described by the Hg distribution aswater-leachable, mobile, semi-mobile and non-mobile species, aswell as other characteristics linked to anthropogenic origin, likeCa, Na, chloride, sulfate and Cu. In these zones, several characteris-tics linked to anthropogenic origin had a major effect on the Hg fate.Thus, Ca and sulfate anion influenced mainly the increase of waterleachability of Hg, while Cu induced the decrease of mobility, likelydue to formation of amalgam as non-mobile species. The naturalcomponents of soil were found to have no influence on Hg fate.Unlike clustering based on Hg species when samples were groupedaccording to collection area and contamination degree, the consider-ation of chemical characteristics resulted in a cluster of soils comingfrom different locations described by Zn, NO3

� and Ba of anthro-pogenic origin, namely former production of pesticides and theirused in agriculture. The FHCC approach is able to manipulate largeset of characteristics with high variability, high efficient for simulta-neous clustering and the achieved models are predictive and provideinformation on the fate of contaminants.

Acknowledgments

This work was supported by a Grant of the Romanian NationalAuthority for Scientific Research, CNDI–UEFISCDI, Project numberPN-II-PT-PCCA-2011-3.2-0219 (Contract no. 176/2012).

Appendix A. Supplementary material

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.chemosphere.2015.05.070.

References

Abollino, O., Malandrino, M., Giacomino, A., Mentasi, E., 2011. The role ofchemometrics in single and sequential extraction assay: a review: part I.Extraction procedures, uni- and bivariate techniques and multivariatereduction techniques for pattern recognition. Anal. Chim. Acta 688, 104–121.

Amos, H.M., Jacob, D.J., Kocman, D., Horowitz, H.M., Zhang, Y.X., Dutkiewicz, S.,Horvat, M., Corbitt, E.S., Krabbenhoft, D.P., Sunderland, E.M., 2014. Globalbiogeochemical implications of mercury discharges from rivers and sedimentburial. Environ. Sci. Technol. 48, 9514–9522.

Bang, Y.K., Lee, C.H., 2011. Fuzzy time series prediction using hierarchical clusteringalgorithms. Exp. Syst. Appl. 38, 4312–4325.

Bernaus, A., Gaonaq, X., Van Ree, D., Valente, M., 2006. Determination of mercury inpolluted soils surrounding a chlor-alkali plant, direct speciation by X-rayabsorption spectroscopy techniques and preliminary geochemicalcharacterization of the area. Anal. Chim. Acta 565, 73–80.

Dai, Z.H., Feng, X.B., Zhang, C., Shang, L.H., Qiu, G.L., 2013. Assessment of mercuryerosion by surface water in Wanshan mercury mining area. Environ. Res. 125,2–11.

Ding, C.F., Zhang, T.L., Taolin, X.G., Wang, X.X., 2014. Major controlling factors andprediction models for mercury transfer from soil to carrot. J. Soil Sediment 14,1136–1146.

Donovan, P.M., Blum, J.D., Demers, J.D., Gu, B.H., Brooks, S.C., Peryam, J., 2014.Identification of multiple mercury sources to stream sediments near Oak Ridge,TN, USA. Environ. Sci. Technol. 48, 3666–3674.

Fernandez-Martinez, R., Loredo, J., Ordonezb, A., Rucandio, I., 2014. Mercuryavailability by operationally defined fractionation in granulometricdistribution of soils and mine wastes from an abandoned cinnabar mine.Environ. Sci. Proc. Impacts 16, 1069–1075.

Frentiu, T., Mihaltan, A.I., Senila, M., Darvasi, E., Ponta, M., Frentiu, M., Pintican, B.P.,2013a. New method for mercury determination in microwave digested soilsamples based on cold vapor capacitively coupled plasma microtorch opticalemission spectrometry: comparison with atomic fluorescence spectrometry.Microchem. J. 110, 545–552.

Frentiu, T., Pintican, B.P., Butaciu, S., Mihaltan, A.I., Ponta, M., 2013b. Determination,speciation and distribution of mercury in soil in the surroundings of a formerchlor-alkali plant: assessment of sequential extraction procedure and analyticaltechnique. Chem. Cent. J. 7, article number 178.

Granite, E.J., Pennline, H.W., Sennior, C. (Eds.), 2015. Mercury Control for Coal-Derived Gas Streams. Wiley-VCH, Verlag GmbH & Co. KGaA, Weinheim. pp. 3–8.

Hemmateenejad, B., Karimi, S., Mobaraki, N., 2013. Clustering of variables inregression analysis: a comparative study between different algorithms. J.Chemometr. 27, 306–317.

T. Frent�iu et al. / Chemosphere 138 (2015) 96–103 103

Higueras, P., Esbri, J.M., Oyarzun, R., Llanos, W., Martinez-Coronado, A., Lillo, J.,Lopez-Berdonces, M.A., Garcia-Noguero, E.M., 2013. Industrial and naturalsources of gaseous elemental mercury in the Almadén district (Spain): anupdated report on this issue after the ceasing of mining and metallurgicalactivities in 2003 and major land reclamation works. Environ. Res. 125, 197–208.

Horvat, M., Sakamoto, M., Chan, L., Faganeli, J., 2013. Mercury in contaminated sites:Identification, characterization, impacts, and remediation. Environ. Res. 121, 1.1–1.

Horvat, M., Degenek, N., Lipej, L., Tratnik, J.S., Faganeli, J., 2014. Trophic transfer andaccumulation of mercury in ray species in coastal waters affected by historicmercury mining (Gulf of Trieste, northern Adriatic Sea). Environ. Sci. Pollut. R21, 4163–4176.

Inacio, M., Neves, O., Pereira, V., Da Silva, E.F., 2014. Levels of selected potentialharmful elements (PHEs) in soil and vegetables used in the diet of thepopulation living in the surroundings of the Estarreja Chemical Complex(Portugal). Appl. Geochem. 44, 38–44.

ISO 10304-1:2007 Water quality. Determination of dissolved anions by liquidchromatography of ions. Part 1: Determination of bromide, chloride, fluoride,nitrate, nitrite, phosphate and sulfate.

ISO 10390:2005 Soil quality. Determination of pH.Kalish, B.T., Rifas-Shiman, S.L., Wright, R.O., Amarasiriwardena, C.J., Jayawardene, I.,

Gillman, M.W., Lipshultz, S.E., Oken, E., 2014. Associations of prenatal maternalblood mercury concentrations with early and mid-childhood blood pressure: aprospective study. Environ. Res. 133, 327–333.

Kocman, D., Horvat, M., Pirrone, N., Cinnirella, S., 2013. Contribution ofcontaminated sites to the global mercury budget. Environ. Res. 125, 160–170.

Llop, S., Murcia, M., Aguinagalde, X., Vioque, J., Rebagliato, M., Cases, A., Iniguez, C.,Lopez-Espinosa, M.J., Amurrio, A., Navarrete-Munoz, E.M., Ballester, F., 2014.Exposure to mercury among Spanish preschool children: trend from birth to agefour. Environ. Res. 132, 83–92.

Lusilao-Makiese, J.G., Cukrowska, E.M., Tessier, E., Amouroux, D., Weirsbye, I., 2013.The impact of post gold mining on mercury pollution in the West Rand region,Gauteng, South Africa. J. Geochem. Explor. 134, 111–119.

Meng, M., Li, B., Shao, J.J., Wang, T., He, B., Shi, J.B., Ye, Z.H., Jiang, G.B., 2014.Accumulation of total mercury and methylmercury in rice plants collected fromdifferent mining areas in China. Environ. Pollut. 184, 179–186.

Miklavcic, A., Mazej, D., Jacimovic, R., Dizdarevic, T., Horvat, M., 2013. Mercury infood items from the Idrija Mercury Mine area. Environ. Res. 125, 61–68.

Ministerial Order No 95/2005 for the approval of waste acceptance criteria andnational lists of waste accepted in each landfill class.

Ministerial Order 756/1997 for the Regulation approval concerning theenvironmental pollution assessment.

Neculita, C.M., Zagury, G.J., Deschenes, I., 2005. Mercury speciation in highlycontaminated soils from chlor-alkali plants using chemical extractions. J.Environ. Qual. 34, 255–262.

Niros, A.D., Tsekouras, G.E., Tsolakis, D., Manousakis-Kokorakis, A., Kyriazis, D.,2015. Hierarchical fuzzy clustering in conjunction with particle swarm

optimization to efficiently design RBF neural networks. J. Intell. Robot. Syst.78, 105–125.

Pirrone, N., Cinnirella, S., Feng, X., Finkelman, R.B., Friedl, H.R., Leaner, J., Mason, R.,Mukherjee, A.B., Stracher, G.B., Streets, D.G., Telmer, K., 2010. Global mercuryemissions to the atmosphere from anthropogenic and natural sources. Atmos.Chem. Phys. 10, 5951–5964.

Pop, H., Sârbu, C., 1997. The fuzzy hierarchical cross-classification algorithm.Improvements and comparative study. J. Chem. Inform. Comput. Sci. 37, 510–516.

Pourjabbar, A., Sârbu, C., Kostarelos, K., Einax, J.W., Buchel, G., 2014. Fuzzyhierarchical cross-clustering of data from abandoned mine site contaminatedwith heavy metals. Comput. Geosci.-UK 72, 122–133.

Qiu, G.L., Feng, X.B., Meng, B., Zhang, C., Gu, C.H., Du, B.Y., Lin, Y., 2013.Environmental geochemistry of an abandoned mercury mine in Yanwuping,Guizhou Province, China. Environ. Res. 125, 124–130.

Renneberg, A.J., Dudas, M.J., 2002. Calcium hypochlorite removal of mercury andpetroleum hydrocarbons from co-contaminated soils. Waste Manage. Res. 20,468–475.

Richardson, J.B., Friedland, A.J., Engerbretson, T.R., Kaste, J.M., Jajson, H.P., 2013.Spatial and vertical distribution of mercury in upland forest soils across thenortheastern United States. Environ. Pollut. 182, 127–134.

Rocio, M., Elvira, E., Pilar, Z., Maria-Jose, S., 2013. Could an abandoned mercury minearea be cropped? Environ. Res. 125, 150–159.

Sârbu, C., Pop, H., 2000. Fuzzy classification of the first 10 MEIC. Chemosphere 40,513–520.

Sârbu, C., Horovitz, O., Pop, H., 1996. A fuzzy cross-classification of chemicalelements, based on their physical, chemical and structural features. J. Chem.Inform. Comput. Sci. 36, 1098–1108.

Schmid, T., Rico, C., Rodriguez-Rastrero, M., Sierra, M.J., Diaz-Puente, F.J., Pelayo, M.,Milian, R., 2013. Monitoring of the mercury mining site Almaden implementingremote sensing technologies. Environ. Res. 125, 92–102.

Scollenberger, C.I., 1945. Determination of soil organic matter. Soil Sci. 59, 53–56.SR EN 12457-1:2003 Characterization of waste. Leaching. Compliance test for

leaching of granular waste materials and sludges. One stage batch test at aliquid to solid ratio of 2 l/kg for materials with high solid content and withparticle size below 4 mm.

U.S. Environmental Protection Agency, EPA 3200 method, 2005. Mercury speciesfractionation and quantification by microwave assisted extraction, selectivesolvent extraction and/or solid phase extraction. <http://www.epagov/osw/hazard/testmethods/pdfs/3200> (accessed 01.05.13).

Vaselli, O., Higueras, P., Nisi, B., Esbri, J.M., Cabassi, J., Martinez-Coronado, A., Tassi,F., Rappuoli, D., 2013. Distribution of gaseous Hg in the Mercury mining districtof Mt. Amiata (Central Italy): a geochemical survey prior the reclamationproject. Environ. Res. 125, 179–187.

Visnjevec, A.M., Kocman, D., Horvat, M., 2014. Human mercury exposure and effectsin Europe. Environ. Toxicol. Chem. 33, 1259–1270.

Zhou, F., Guo, H., Liu, Z., Jiang, Y., 2007. Chemometrics data analysis of marine waterquality and source identification in Southern Hong Kong. Mar. Pollut. Bull. 54,745–756.