Journal of Geochemical Exploration 107 (2010) 305–316

Contents lists available at ScienceDirect

Journal of Geochemical Exploration

j ourna l homepage: www.e lsev ie r.com/ locate / jgeoexp

Investigation of the hydrogeochemistry of some bottled mineral waters in Hungary

Ubul Fugedi, Laszlo Kuti, Gyozo Jordan ⁎, Barbara KerekGeological Institute of Hungary, H-1143 Budapest, Stefánia út 14. Hungary

⁎ Corresponding author. Geological Institute of HunBudapest 1143, Hungary. Tel.: +36 1 251 0999; fax: +3

E-mail address: jordan@mafi.hu (G. Jordan).

0375-6742/$ – see front matter © 2010 Elsevier B.V. Aldoi:10.1016/j.gexplo.2010.10.011

a b s t r a c t

a r t i c l e i n f o

Article history:Received 9 March 2010Accepted 20 October 2010Available online 5 November 2010

Keywords:Mineral waterStatistical analysisWater qualityPannonian BasinHungary

Bottled drinking water constitutes a significant part of total water consumption in developed countries andnational and EU legislation regulates their market production. In the framework of an international projectcarried out by the EuroGeoSurveys Geochemistry Expert Group 36 bottled waters were obtained from publicmarkets in Hungary in order to determine their hydrogeochemical composition. The objective of this study isto investigate the possible relationship between groundwater aquifer lithology and the processed andmarketed bottled waters, and to develop a classification of bottled waters, based on their dissolved mineralcontent. Analytical results of this study are compared with the composition shown on bottle labels, and witharchive hydrochemical data from the producing wells. Results show that, while processing of originalgroundwater, such as oxygen addition, iron or hydrogen-sulphide removal can significantly alter watercomposition, bottled water composition can be used for selection of sites for detailed hydrogeochemical andhydrogeological characterization. A simple and useful classification of bottled water quality is also presentedthat is based on natural groups of sampled waters derived by means of statistical data analysis methods.

gary (MAFI), Stefania ut 14,6 1 251 0703.

l rights reserved.

© 2010 Elsevier B.V. All rights reserved.

1. Introduction

Bottled drinkingwater production has steadily increased in the lastfew decades in industrialised countries, and the global consumption ofbottled water reached 154 billion litres in 2004, up 57% from the98 billion litres consumed five years earlier (Arnold, 2006). It is themost dynamic sector of all the food and beverage industry:consumption in the world increases by an average of 12% each year,in spite of its excessively high price compared to tap water. The percapita consumption of bottled water in the European Union variesenormously from one country to another with the averageconsumption at 105 l per year which is about 25% of the total drinkingwater consumption in Europe. 85% of total bottled water consumptionis mineral water, followed by spring water (12%) and table water (3%)(Kay-Shuttleworth, 2009). In Hungary, since the transition to marketeconomy in 1990, mineral water consumption has boomed from3.7 l/person in 1991 to 110 l/person in 2009. Among the reasons forthis increase are the lack of good quality potable tap water (pipedwater) in some regions, like large urban areas, or in the natural arseniccontaminated aquifers of the Great Hungarian Plain, aswell as the easymarket availability of a large selection of bottled waters with differentcomposition and taste, and the marketing pressure of advertising bymulti-national companies. There are, however, differences among the

countries, both in the consumed bottled water quantities and in thequality preferences. Finland has the lowest consumption level with 16 l ayear per inhabitant, and Italy has the highest at just under 200 l a year perinhabitant (Italy is theworld leader in bottledwater consumption, too). Interms of water quality, sparkling (CO2 gaseous) water is preferred to stillwater in Hungary, as is in Germany, Austria and Czechia, while still wateris favoured in France, Italy and Spain. In terms of market selection, thereare 45–50 different brands of bottled mineral water available on themarket in Hungary alone (Hungarian Mineral Water Association andProduct Council, 2009), while in Germany there are more than 500brands.

In order to protect drinking water, one of the most fundamentalnatural resources, the European Union (EU) introduced stringentregulations for water quality control, and for the management ofwater resources (Directive 2000/60/EC). Due to the huge consumedquantities, and large number of marketed brands, the EU specificallyregulates the quality standards of bottledwaters (Directive 80/777/EEC;Directive 96/70/EC). There are, however, different legal classificationsand health standards for bottled waters in some countries. According toEU Directive 80/777/EC, “natural mineral water” is water that ismicrobiologically wholesome, “originating in an underground watertable or deposit and emerging from a spring tapped at one or morenatural or bore exits”. Treatment of mineral water is restricted toremoval of unstable elements, such as iron and sulphur compounds. InHungary, for example, legislation designates natural groundwaters as‘drinking water’, ‘mineral water’ and ‘medicinal water’, based on themaximum admissible concentration of certain components (drinkingwater), on its subsurface origin and bottling at the location of extraction

306 U. Fugedi et al. / Journal of Geochemical Exploration 107 (2010) 305–316

(mineral water), and based on proven positive health effect (medicinalwater) (Government Decree 201/2001; Government Decree 65/2004).

Bottled mineral waters are derived from groundwater aquifersand, therefore, their chemical composition is originally defined bygeochemical water-rock interaction processes. In fact, some of theirunique composition features, such as elevated carbon dioxide (CO2),hydrogen sulphide (H2S), sulphate (SO4

2−), iron and high salt contentare specific to the geological and hydrogeological position of theexploited springs and producing wells. However, these naturallyoccurring waters are often chemically processed before bottling inorder to adjust to market needs. Processing can include de-gassing orjust the addition of CO2 gas, removal of iron to improve flavour,removal of reducing hydrogen sulphide for better consumptionexperience, etc. Manipulation of carbon dioxide content changes pH,removal of iron by precipitation can co-precipitate several traceelements, and the removal of reducing hydrogen sulphide increasesEh of the mineral water. The final bottled water product can have,therefore, little resemblance to the original groundwater composition.

Nevertheless, Misund et al. (1999) found a traceable link betweenbottledwater composition and aquifer lithology in 66 European bottledwaters. Groselj et al. (2008) used neural networks to arrive at the sameconclusion. The natural mineral content of bottled waters originatingfrom water-rock interaction in the exploited groundwater aquifers hasbeen the subject of numerous studies, too (Garzon and Eisenberg, 1998;Ikemet al., 2002; Fiket et al., 2007; Kochubovski et al., 2007).Others, likeChiarenzelli and Pominville (2008) and Saleh et al. (2008), went furtherand evaluated the chemical, microbial and physical composition ofcommercial bottled waters for health considerations.

In this study, 36 randomly selected bottled waters were obtainedfrom the public market and their hydrochemical composition wasdetermined. The objective of this study is to investigate the linkbetween the composition of marketed bottled water products andtheir geological–geochemical origin. Another aim of the study is todevelop a classification of bottled waters, based on their dissolvedmineral content. The laboratory analysis results are compared toarchive hydrochemical data from the producing wells, and alsocompared to the chemical composition reported on bottle labels.When available, gaseous and non-gaseous varieties of marketedwaters are compared in order to estimate the impact of waterprocessing. It is not the objective of this study to evaluate or tocompare the studied waters' drinking quality by any means, andneither to characterise bottled waters available on the market. Rather,the present study is a preliminary methodological investigation of thelimitations and possibilities of using bottled waters for mapping thegeochemistry of groundwater.

2. Study area

Hungary is situated in the Pannonian Basin surrounded by theCarpathian Mountains chain. Convergent motion between the Africanand European plates in the Alpine period (Late Jurassic–Neogene)resulted in the southerly subduction of the European plate under thePannonian continental fragment between the two continents. This ledto the formation of calcalkaline volcanic arcs (Inner-CarpathianVolcanic Belt) during the Eocene in the north, while the emergenceof the Alpine-Carpathian chains, and the fast thermal sinking of thePannonian fore-arc basin, led to the accumulation of very thick (often6–7000 m; Fig. 1) marine and then fluvial clastic sedimentarysequences over the Palaeozoic and Mezozoic basement (Fig. 2)(Trunko, 1996). Some parts of the basement remained in elevatedposition exposing the dominantly Mesozoic carbonates (mostlylimestone) of the closing Thethys Sea (Fodor et al., 2005). Besidesthe important carbonate groundwater aquifers, the majority ofgroundwater resources are stored in the thick clastic porous waterreservoirs (Figs. 1 and 2). The country is rich in water resources,including mineralised and thermal waters, due to the high thermal

gradient under the thin crust of the Pannonian Basin. Somlai et al.(2002) investigated the Ra content while Szanto et al. (2008) studiedstable isotopes of Hungarian bottled waters. They both concluded thatthere exists a traceable link between the studied bottled watercomposition and aquifer lithology.

3. Materials and methods

3.1. Sampling, analysis and datasets

In order to characterise the composition of the studied bottledwaters, three independent data sources were used: (1) laboratoryanalysis of bottled waters purchased from the public market carriedout by the Federal Institute for Geosciences and Natural Resources(BGR) (‘BGR Dataset’), (2) the chemical composition reported onbottle labels (‘Bottle Dataset’), and (3) archive hydrochemical datafrom the producing wells in the National Borehole Database of theGeological Institute of Hungary (‘MAFI Dataset’). The present study isbased on the new BGR Dataset, and the two other datasets are usedonly for comparison.

The BGR Dataset was generated from 36 different brands ofcommercially available bottled waters, purchased from supermarketsin Budapest during 2009 (Table 1). When available, the smaller 0.5 lbottles were preferred to the larger bottles. Also, when available, thesparkling, still and medium varieties of the same water product werepurchased (Table 1). In total, 51 bottles were purchased and sent tothe BGR laboratory for analysis. The bottled waters were purchasedwithin a period of 3 weeks, and they were transported to Germanyeither personally or by freight mail.

Themineral water bottles were first opened in the laboratory and a100 cm3 plastic beaker was filled with mineral water. The cap wasthen loosely replaced on the mineral water bottle, and the bottle wasdegassed for about 20 min in an ultrasound bath. The electricalconductivity and pH of the mineral water in the beaker weremeasured. After degassing, 25 cm3 were withdrawn from the mineralwater bottle for the bicarbonate titration. A 250 cm3 HDPE bottle wasrinsed with the degassed mineral water and then filled with themineral water. The anions were determined by ion chromatographyand ammonium ion was determined by a photometric method usingthe water in the HDPE bottle. A 125 cm3 FLPE bottle was rinsed withthe degassed bottled water and then filled with the bottled water. Inthe clean room, 1.5 cm3 ultrapure 69% HNO3 was added and the bottleshaken. The main cations were determined by ICP-OES and the traceelements by ICP-QMS. The ICP-QMS analyses weremade using 13 cm3

PP tubes (Sarstedt) with LDPE stoppers. This database is uniform andhomogeneous due to the use of a single analytical method under well-controlled conditions of a single laboratory. The following parameterswere determined: Ag, Al, As, B, Ba, Be, Bi, Cd, Ce, Co, Cr, Cs, Cu, Dy, Er,Eu, Fe, Ga, Gd, Ge, Hf, Hg, Ho, I, K, La, Li, Lu, Mg, Mn, Mo, Na, Nb, Nd, Ni,Pb, Pr, Rb, Sb, Sc, Se, Sm, Sn, Ta, Tb, Te, Th, Ti, Tl, Tm, U, V, W, Y, Yb, Zn,Zr, Br, HCO3, Cl, F, NH4, NO2, NO3, PO4, SO4, SiO2, pH, electricalconductivity and total alkalinity.

Eight bottled waters (Ave, Fonyódi, Ferenc József, Hunyadi János,Bükkszéki Salvus, Germán Aqua, Parádi I., Mira), because of theirunusual high salt content, were purchased a second time and sent toBGR for re-analysis.

More details about the analytical methods, quality control results,and re-analysis of bottled waters with unusually high values is foundin Reimann and Birke (2010) and Birke et al. (2010).

The Bottle Dataset was very simply generated from the chemicalcomposition displayed on bottle labels. The name of the 36 producingwells, if recorded, was also added to the database, together with the gascontent of thewater as sparkling (blue cap), still (pink cap) andmedium(green cap). Information about bottle materials, such as plastic (PET) orglass bottle, bottle colour, and the plastic ormetal capwas also includedin thedatabase. It is noted that the chemical compositiononbottle labels

Fig. 1. A. Bedrock depth of the groundwater flow system in the Pannonian Basin in Hungary inmetres above sea level (after Korossy, 1968).White dots show location of bottled waterproducing wells investigated in this study (see Table 1 for brand names). Inset: dark polygon shows location of Hungary in Europe. B. Hydrogeological cross-section showing majorgroundwater flow directions in the Quaternary sedimentary layers. Note marked regional recharge and discharge areas (after Erdelyi, 1975).

307U. Fugedi et al. / Journal of Geochemical Exploration 107 (2010) 305–316

is the result of various laboratories, using different water analysismethods and unknown quality control procedures. According tolegislation, the chemical analysis can be performed only by accreditedchemical laboratories in Hungary. However, this database is, for theaforementioned reasons, considered inhomogeneous and should betreatedwith caution. In termsof theparameters recorded in this dataset,the total dissolved solids content and the concentration of major ionsappear on each bottle. Ions appearingmost frequently on the studied 36bottled waters are Na+ (36), HCO3

− (36), Ca2+ (35), Mg2+ (35), Cl−

(17), SO42− (16) and K+ (14). For bottled waters with higher salt

content, additional ions are shown, such as Cl−, Li+, Br−, I− and F−. Forexample, mineral water Salvus (this is the only water with Ca2+ andMg2+not shown), the concentration of Li+, Br−, I− and F− are indicatedon the label. Sometimes, alkali elements (Na+ and K+) are simplysummed up.

The third data source, the MAFI Dataset, was extracted from archivehydrochemical data of water producing wells stored in the MAFI

National Borehole Database. Most importantly, the lithology at themineral water producing wells was taken from this database (Table 1).Also, the geographical co-ordinates of the wells, producing the studiedbottled waters, were obtained from this data source (see Figs. 1 and 2).When no co-ordinates were provided, the locations of the wells wereidentified by the nearest settlement's geographical centre point. At thecoarse spatial scale of the study (Figs. 1 and 2), this decision did notcause problems indata interpretation. It is noted that this database is themost heterogeneous. Some water samples were analysed when theproducing well was developed 50 years ago, using a multitude ofanalytical methods (including old semi-quantitative methods),laboratories, sample storage and treatment procedures. Some wellsproducing the studied bottled mineral waters have no data available inthis database (8 of 36 wells havemissing data). If missing in the archivedatabase, some water quality data was obtained from the literature(Schulhoff, 1957; Borszeki, 1979). Since the water analyses in theboreholes were carried out for industrial purposes, the analysed

Fig. 2. Generalised surface lithology of the Pannonian Basin, Hungary. White dots show location of bottled water producing wells investigated in this study (see Table 1 for brandnames).

308 U. Fugedi et al. / Journal of Geochemical Exploration 107 (2010) 305–316

hydrochemical parameters also have a great diversity. The followingcomponents were available in this dataset: Ca2+, Na+, K+, Mg2+, SiO2,HCO3

−, SO42−, S2, Cl−, F−, I−, Li+, Br−, NO2

−, NO3−, CO2.

Finally, the used 1:500,000 scale geological and hydrogeologicalmapswere obtained from the digital spatial database of the GeologicalInstitute of Hungary (MAFI).

3.2. Statistical methods of data analysis

Statistical analysis of bottled water hydrochemical data consideredVersari et al. (2002) chemometric studies and results of Guler (2007).The numerical analysis started with the investigation of the univariatedistribution of parameters using histograms and cumulative distribu-tion functions (CDF). Groups of samples were delineated by visualinspection of significant minima and breaks in the histograms and CDFcurves, respectively. Due to the obvious heterogeneity of samplepopulations, caused by multi-modality and presence of outliers,statistical tests for distribution fitting, including checking for normality,were not performed. Under these conditions median was taken as thedistribution central value, and the variability or spread of thedistribution was characterised by the median absolute deviation(mad.): mad=Me|xi−Mex|, whereMex is the median (Me) of variablexi. In addition, twoother statistics are used for the characterisation of thevariability: (1) the ratio of the highest and lowest values (maximum/minimum), and (2) the ratio of the interquartile range (IQR) and themedian (IQR/Me).

Comparison of different data sources and modelling of parameterinterdependence were carried out by bivariate linear regressionanalysis. Variability of sample populations was stabilised bylogarithmic transformation of parameters. Linear correlations, andcalculated regression lines, discussed in this paper were all significantat the 0.05 confidence level. The robust non-parametric Spearmanrank correlation coefficients were used in this study. Effect of outlierson the regression line fitting was checked by subsequent exclusion ofoutliers from the regression models. Results show that outliers inmeasured parameters in the studied datasets do not have a significant

impact on the calculated regression lines and, therefore, they werekept in the models.

Classification of bottled water samples into groups in multivariatespace was performed by cluster analysis (CA) using Euclideandistance and simple linkage method, and standard factor analysis(FA). FA was based on the Spearman rank correlation coefficient.Three independent methods were used to identify groups among thebottled mineral waters: (1) visual inspection of univariate histogramsand CDF curves of major components, (2) identification of groups ofparameters with anomalous values using histograms, and (3) CA andFA studies. Groups discussed in this paper were identified by all thesemethods to confirm their existence. Statistical analysis used the SPSSPC+ statistical software package (SPSS® Base 8.0).

4. Results and discussion

4.1. Comparing different water treatments

Carbon-dioxide gas content is a critical issue in bottledmineral waterquality. EU legislation (Directive 80/777/EEC; Directive 96/70/EC;Directive 2009/54/EC) state that if natural mineral is effervescent, itmust be labelled accordingly, depending on the origin of the carbondioxide: (a) naturally carbonated naturalmineralwater (no introductionof CO2); (b) natural mineral water fortified with gas from the spring(reintroduction of CO2), and (c) carbonated natural mineral water (CO2

added following strict guidelines). In the present study, comparison ofdissolved bicarbonate content in the still and sparklingwater varieties bymeans of simple linear regression showed no significant differencebetween the waters treated and untreated with CO2 gas. Differencesbetween thepHvaluesmeasured in still and sparklingwaters of the samebrand varies widely, with pH systematically being lower in the sparklingvarieties due to the formation of weak bicarbonate acid. Since, partialpressure of dissolved CO2 and therefore the concentration of bicarbonateacid changes fast after opening the bottle, stable pH values can beobtained only from the stillwaters. Conductivity valuesmeasured for stilland sparkling waters are very similar to each other. In two sample-pairs

Table 1Lithology at wells and chemical classification of studied bottled waters. First column gives the sample code of the 36 bottled waters in the BGR Dataset (Bottled water Vivien withsample code 22 was not analysed by BGR). See Figs. 1, 2 and 6 for their location and classification.

Sample code Name of bottled mineral water CO2 content Lithology at wells Chemical class*

1 Harmatvíz Sparkling, still Dolomite 3b2 Margitszigeti Kristályvíz Sparkling, still (Fissured) marl 2b3 Theodora Kereki Sparkling Sandstone, clay marl (carbonate rocks) 2b4 Theodora Kékkúti Still Sandstone, clay marl (carbonate rocks) 2b5 Szentkirályi Sparkling, still, medium Sandstone 3c6 Balfi Still Sandstone 2b7 Mohai Moha II Still Sandstone 2c8 Mohai Ágnes Still Sandstone 2c9 Ave Sparkling, still, medium Sandstone 3b10 Nestlé Aquarel Sparkling, still Dolomite 3d11 Natur Aqua Sparkling, still Limestone 3c12 Visegrádi Still Limestone 2b13 Fonyódi Sparkling, still Sandstone 3a14 Ferenc József Medium Weathered clay 1a15 Hunyadi János Medium Weathered clay 1a16 Parádi Csevice II Medium Andezite (sandstone) 2a17 Pávai Vajna Sparkling Sandstone 3b18 Aqua Friss Still Sandstone 3c19 Bükkszéki Salvus Sparkling Limestone 1b20 Germán Aqua Sparkling, still, medium Sandstone 3d21 Santa Aqua Sparkling Sandstone 3d23 Parádi Csevice I Medium Andezite (sandstone) 2a24 AquaSol Sparkling Sandstone 3b25 Mizse Still Sandstone 3c26 Primavera Sparkling, still Sandstone 3c27 Mira Medium Weathered clay 1a28 Pannon Aqua Sparkling Sandstone 3b29 Aquarius Sparkling, still Sandstone 3b30 Apenta Optima Medium Sandstone 3c31 Veritas Gold Still Sandstone 3d32 Óbudai Gyémánt Still Limestone 3c33 Fonte Verde Sparkling Sandstone 3c34 Aqua Mathias Still Sandstone 2c35 Dogerita Sparkling, still Sandstone 3b36 Aqua+ Oxigén plusz Medium Sandstone 3c37 Aqua Natural Still Sandstone 2c

*Notation: Group 1: ECN19,000 μS/cm, TDSN15,000 mg/l, Ca2+bMg2+, Na+N10xCa2+; Sub-group 1a: Ca2+-Mg2+-SO42− type; Sub-group 1b: Na+-HCO3

− type; Group 2: TDS:900–4500 mg/l, Ca2+NMg2+ (Ca2+:Mg2+ ratio=1.7–3.5), dominance of HCO3

−, low SO42− and Cl−; Sub-group 2a: Ca2+:Mg2+ ratio=2.2, TDSN1400 mg/l, Na2+-Ca2+-HCO3

− type; Sub-group 2b: Ca2+-HCO3

− type, Ca2+:Mg2+ ratio=1.7–2.5; Sub-group 2c: Ca2+-HCO3− typewith Ca2+:Mg2+=2.6–3.5; Group 3: TDSb900 mg/l, HCO3

− type, Ca:Mg ratio=1.2–3.8 andwithlow sulphate and chloride contents; Sub-group 3a: Na+-HCO3

− type with Na+:Ca2+ ratio~4; Sub-group 3b: same level in dissolved Ca2+, Mg2+ and Na2+ and with no trace elementanomalies; Sub-group 3c: same Ca2+ and Mg2+ (Ca2+:Mg2+ ratio=1–1.5) content but with much lower dissolved Na+ concentrations (Ca2+:Na+ ratio=3.9–23); Sub-group 3d:dominance of dissolved Ca2+ (Ca2+:Mg2+ratioN1.6; Ca2+:Na2+ ratio=3.7–8.1).

309U. Fugedi et al. / Journal of Geochemical Exploration 107 (2010) 305–316

(Harmatvíz and Nestlé Aquarel) the difference in conductivity is morethan 10% (in one case the sparkling variety has a higher value, and in theother case the still one). Due to the lack of an obvious reason for this

Fig. 3. Comparison of Ca and Na from different data sources. A. Scattergram of dissolved Cabetween the BGR and Bottle Datasets.

difference for Nestlé Aquarel, the sparkling variety was included in theanalysis, since its bottle label compositionwas closer to theMAFI Dataset(original groundwater) composition. For bottledwaterHarmatvíz, nitrite

contents between the BGR and MAFI Datasets. B. Scattergram of dissolved Na contents

Table 3Range of values, equation of linear regression line between the major ions and otherparameters, and coefficient of determination in the MAFI and BGR Datasets(concentration values in mg/l and EC in μS/cm).

Parameter Rangeofvalues

Equation of linearregression line

Numberof sample-pairs(outliers)

Coefficient ofdetermination(r2)

Ca2+ 7–405 CaM=0.919 * CaBGR 25 (1) 0.92Mg2+ 5–3,150 MgM=0.911 * MgBGR 26 0.98Na+ 3–6,200 NaM=0.925 * NaBGR 19 0.97K+ 0.8–63 KM=1.154 * KBGR 19 0.89Fetotal 0.156–160 FeM=0.026 * FeBGR 28 0.22B3+ 0.006–1,450 BM=1.655 * BBGR 18 (1) 0.91Mntotal b0.001–0.9 MnM=0.281 * MnBGR 27 0.16HCO3

− 250–1,700 HCO3M=0.975 * HCO3

BGR 24 (2) 0.94SO4

2− b0.005–20,500 SO4M=1.014 * SO4

BGR 28 0.95SO4

− b0.005–230 SO4M=1.288 * SO4

BGR 24 (4) 0.85Cl− 0.98–2400 ClM=0.909 * ClBGR 28 0.98F− 0.031–2,8 FM=0.969 * BGR 19 0.62Br− 0.009–16 BrM=0.643 * BrBGR 18 (1) 0.99I− 0.001–2,500 IM=0.829 * BGR 18 (1) 0.96PO4

3− b0.002–0.66 PO4M=0.884 * PO4

BGR 11 0.80SiO2 8–51 SiO2

M=0.765 * SiO2BGR 20 0.87

EC 250–26,000 ECM=0.889 * ECBGR 12 0.99pH 5.3–8.3 pHM=1.035 * pHBGR 27 0.97

Ca2+ and Br− without bottled water Mira.HCO3

−: without bottled waters Aqua Friss and Bükkszéki Salvus.SO4

2−: without bottled waters Ferenc József, Hunyadi János, Aqua Friss and Harmatvíz.M: MAFI Dataset.BGR: concentration measured in the BGR Laboratory.

Table 4Range of values, equation of linear regression between major ions and coefficient ofdetermination in the Bottle and MAFI Datasets.

310 U. Fugedi et al. / Journal of Geochemical Exploration 107 (2010) 305–316

and nitrate seem to be higher in the still water variety, while ammoniumhas ahigher concentration in the sparklingbottledwater sample. It is alsointeresting that the amount of aluminium is six times higher in the stillvariety than in the sparkling. Based on the above findings, it seems thatthe original water is oxidised in the still water during processing and,thus, the more reductive sparkling variety is considered to have morestable and reliable chemical composition. Therefore, the sparkling watertypewas used in all the subsequent discussion for Harmatvíz. Apart fromHarmatvíz and Nestlé Aquarel, since there is little difference betweenwater samples from the same aquifer, but with different CO2 gastreatment, chemical concentrations in the two- or three-types of thesame water sample (sparkling, still, medium) were averaged for eachmeasured parameter in the BGR Dataset. All the discussion below usesthese data.

4.2. Comparing different data sources

When comparing the BGR laboratory results with the chemicalcomposition displayed on bottle labels, a very good correspondence ofalmost all analytical results with a coefficient of determination (r2)above 0.95 can be observed, apart from F− (r2=0.85) (Fig. 3B; Table 2).Wide concentration ranges (2−4 orders of magnitude) and thepresence of outliers could cause this excellent match of the twodatasets. In our case, however, the regressions, calculated without theoutliers, still showed that the strong correlations remainedunaffected. Asingle outstanding mismatch is the 0.031 mg/l and 1.02 mg/l fluorideconcentrations in the BGR laboratory and Bottle label results,respectively, for bottled water Theodora Kékkúti. Nevertheless, theregression results suggest that there is an overall very good agreementbetween the official andmeasured concentrations ofmajor componentsin the studied bottled waters over all concentration ranges.

An interesting aspect of water chemistry is revealed by thecomparison of bottled water composition shown by the BGR laboratoryresults, and the original pumped groundwater composition taken fromthe MAFI archive groundwater database. It is noted that while the BGRanalytical results come fromuniformwell-controlled analyticalmethods,the MAFI chemical analyses are from different dates using varioussampling, sample storage and preparation and analytical methods.Differences inmeasured parameter values between the two datasets cancome from(1)water treatment and analyticalmethods, (2) the chemicalcomposition of original pumped groundwater and processed bottledwater, and (3) changes over time in groundwater composition due togroundwater extraction. According to the results in Table 3, the poorestmatch of values is found for Fe andMn (r2=0.22 and 0.16, respectively),that ismetalswithmultiple-valences. This is readily explained by sampletreatment procedures,where aeration ofwaters along theprocessing line(affecting BGR Dataset of bottled waters) and during sample storage(affecting MAFI Dataset containing extracted groundwater data) cancause the oxidation and subsequent precipitation of thesemetals. Similareffects are reflected by the moderate correlations found for nitrogen

Table 2Range of values, equation of linear regression line betweenmajor ions and coefficient ofdetermination in the Bottle and BGR Datasets.

Parameter Range ofvalues(mg/l)

Equation of linearregression line

Number ofsample-pairs

Coefficient ofdetermination(r2)

Ca2+ 7–405 CaB=0.949 * CaBGR 35 0.99Mg2+ 5–3,150 MgB=0.807 * MgBGR 35 0.97Na+ 3–6,200 NaB=1.088 * NaBGR 35 0.99K+ 0.8–63 KB=0.883 * KBGR 15 0.95HCO3

− 250–1,700 HCO3B=1.028 * HCO3

BGR 35 0.99SO4

2− b0.05–20.5 SO4B=0.967 * SO4

BGR 16 0.99Cl− 0.98–24.0 ClB=1.018 * ClBGR 17 0.99F− 0.031–2.8 FB=0.856 * FBGR 13 0.85

BGR: concentration measured in the BGR Laboratory.B: Bottle Dataset.

compounds (ammonium, nitrite and nitrate), measured in the twodatasets (not shown in Table 3). Fluoride, too, seems to be anoutstandingelement with a moderately good match in the two datasets (r2=0.62).The highest similarity is obviously for EC, a parameter easy to measureand it simply represents the TDS content. The relatively poor linearcorrelation for PO4

3− is due to its usual low concentration and varyinganalytical methods used for its determination. However, despite theabove differences, there is a very good overall match between the twodatasets (r2N0.9; Fig. 3A; Table 3). This shows that applied industrialwater treatment technologies (including CO2 addition and removal),apart from specific treatments, such as Fe, Mn and H2S removal, do notalter significantly the major ion composition of the pumpedgroundwaters.

The difference of Ca and HCO3− concentrations between bottled

and original groundwaters is more obvious in this comparison(Table 4). This might be attributable to the impact of CO2 changeson carbonate equilibrium during bottled water production. Oxidation

Parameter Range ofvalues (mg/l)

Equation oflinear regressionline

Number ofsample-pairs(outliers)

Coefficient ofdetermination(r2)

Ca2+ 25–370 CaB=0.948 * CaM 25 (1) 0.91Mg2+ 10–2,900 MgB=0.879 * MgM 26 0.96Mg2+* 10–85 MgB=0.809 * MgM 23 (3) 0.90Na+ 7–5,800 NaB=1.187 * NaM 19 0.97Na+* 7–230 NaB=0.778 * NaM 15 (4) 0.83HCO3

− 200–1,700 HCO3B=0.933 * HCO3

M 24 (2) 0.90SO4

2− 5–290 SO4B=0.688*SO4

M 10 (4) 0.86Cl− 5–2,200 ClB=1.111 * ClM 14 0.99Cl−* 5–75 ClB=1.273 * ClM 9 (5) 0.95F− 0.031–1.5 FB=0.894 * FM 6 (1) 0.95

*Range of values and equation of regression line after removal of outliers.B: Bottle Dataset.M: MAFI Dataset.Ca: without bottled water Mira.F−: without bottled water Harmatvíz.

Fig. 4. TDS (logarithm of original concentration values in mg/l) histogram of the BGRDataset (n=31, excluding the 5 highest values).

311U. Fugedi et al. / Journal of Geochemical Exploration 107 (2010) 305–316

due to water aeration (similarly to the removal of sulphur) can alsoexplain the somewhat lower match of original groundwater andbottled water sulphate contents (r2=0.86; Table 4).

4.3. Classification of bottled waters (with some aspects to their relationto hydrogeological settings)

Pattern recognition in the BGR Dataset concentration spatialdistributions is made difficult by the small sample set (n=36) andhigh scatter of measured parameter values. In general, NO2

−, NO3−,

NH4+, PO4

3− and F− are present in trace levels according to theirmedian concentrations (0.005–0.22 mg/l) (Table 5). Potassium andCl− have similar, but more than one order of magnitude highermedian concentrations (2.5 and 4.6 mg/l, respectively). The nextgroup of parameters that have increasing median concentrations inthe following order are SO4

2−, SiO2, Na+, Mg2+, Ca2+, with theoutlying high HCO3

− value (Table 5). On the contrary, the relativevariation given by IQR/Me and mad/Me shows that the least variableparameters are Cl−, F−, SiO2, Na+, Mg2+, Ca2+ and HCO3

−. Togetherwith their high concentrations, these elements represent the overallgeochemical background in the studiedwells, and their concentrationsarenot seriously affectedby technologicalwaterprocessing. Thevery lowvariability and low concentration of phosphate anion can be caused bymicrobes living in the surficial layers and using up the phosphate contentof infiltrating water. Nitrite, NO3

− and NH4+ are among the parameters

with the highest overall variation, despite of their trace concentrations(Table 5). These nitrogen compounds are essentially of surfacecontamination origin, and they are sensitive to oxidation during variousprocessing methods, explaining the high diversity among the studiedbottled water samples. Besides potassium, sulphate has the singleoutlying highest variation. Sulphate often has high concentrations in thestudied bottled waters, coming from highly mineralised mineral andmedicinal waters, while it can have trace concentrations in other waters(see minimum values in Table 5). The ratio of maximum and minimumvalues highlights the effect of extreme values on variability. Besidessulphate, Cl− and Na+ have the most extreme differences inconcentrations among the studied bottled waters, for reasons similarto sulphate.

In general, the majority of drinking water is of Ca2+-Mg2+−HCO3−

type in Hungary (Borszeki, 1979). Mineral waters typically haveapproximately equal concentrations of dissolved Ca2+, Mg2+ and Na+

cations, and about equal anion concentrations of HCO3−, SO4

2− and Cl−.Medicinal waters are characterised by dominance of Na+ (andrarely Mg2+) and SO4

2− (and rarely Cl−) (Borszeki, 1979). The TDShistogram of the BGR Dataset (Fig. 4) displays characteristicgrouping of samples: lg TDSN4 (TDSN10,000 mg/l; 4 samples),3.3b lg TDSb4 (2,000bTDSb10,000 mg/l; 1 sample), 3.1b lg TDSb3.3

Table 5Summary statistics of the major ions and other parameters in the BGR Dataset (n=36).

Parameter min. median max. mad

pH 5.4 6.2 8.3 0.56EC (μS/cm) 254 732 26,000 314Ca2+ (mg/l) 32 75 405 43K+ (mg/l) 0.8 2.5 62.3 1.6Mg2+ (mg/l) 5.2 39.2 3,150 32Na+ (mg/l) 4 33.7 6,200 28.8HCO3

− (mg/l) 93 464 13,450 366Cl − (mg/l) 0.98 4.6 2,360 3.41F− (mg/l) 0.031 0.22 2.71 0.19PO4

3− (mg/l) b0.02 0.18 0.66 0.07SO4

2− (mg/l) b0.05 16.3 20,300 16SiO2 (mg/l) 8.8 22.2 50.9 5.65NH4

+ (mg/l) b0.005 0.147 1.64 –

NO2− (mg/l) b0.005 b0.005 0.988 –

NO3− (mg/l) b0.02 0.198 9.63 0.193

min.: minimum; max.: maximum; Me.: median; mad: median absolute deviation.

(1,200bTDSb2,000 mg/l; 7 samples), and lg TDSb2.9 (TDSb800 mg/l;22 samples). Two samples (Theodora Kereki, Margit-szigeti Kristály-víz) in the 2.9b lg TDSb3.1 interval cannot be classified into eithergroups. The majority of samples (20 of 36) have TDS in the narrowrange of 390–700 mg/l, independent of aquifer lithology. Anotherlocal maximum of the histogram consists of 9 samples with elevatedTDS between 900 and 2000 mg/l. This feature is explained by the factthat bottledwaters were allowed to be soldwith the precious ‘mineralwater’ label only with TDSN1000 mg/l concentration in the past inHungary. Thus, companies focused on the utilisation of these waterreservoirs. The four anomalous samples of TDSN10,000 mg/l representhistoric medicinal waters in Hungary. Similarly to TDS, groupscorresponding to the above three histogrammaxima can be identifiedfor the other components as well (Fig. 5). The threemajor populationsdisplayed in the histograms can be identified as background, elevatedand anomalous concentrations that correspond to the group of lowestconcentrations that includes the majority of samples, a smaller groupwith samples from aquifers with elevated total dissolved concentrations,and finally to the group with anomalously high values definedstatistically as outliers.

Identification of similarities among the studied bottled waters wascarried out by means of univariate histogram analysis (based on TDSprimarily), ratios of major cations, and factor analysis (FA) using theSpearman rank correlation coefficients. Factor analysis was completedin two steps. The aim of the first step was the determination ofrelevant group-forming factors. The result was that the Bi, NO3

−, Sn

mad/Me Interquartile Range (IQR) max./min. IQR/Me

0.067 1.7 1.5 0.270.428 1,270 102 1.730.573 133 12.6 1.770.64 7.8 78 3.120.82 30.1 606 0.770.855 66.2 1.55 1.960.789 771 145 1.660.741 17.5 2.41 1.630.864 0.366 87 1.660.222 0.16 N33 0.890.982 104 N406 6.380.255 13.3 5.8 0.6– 0.32 N328 2.18– 0.0155 N198 3.10.975 0.958 N482 4.84

Table 6Rotated component matrix (2nd step). Extraction method is Principal ComponentAnalysis, using the varimax rotation method with Kaiser normalization. Rotationconverged in 9 iterations.

Component

1 2 3 4 5 6 7

RANK of B 0.663 0.211 0.274 0.229 0.469 0.168 −0.128RANK of Br 0.607 0.730RANK of Ca 0.470 0.788 −0.172 0.179RANK of Cd 0.160 0.132 −0.189 0.861 −0.209RANK of Ce 0.868 0.134RANK of Co 0.413 0.291 0.269 0.217 0.687RANK of Cl 0.774 0.157 0.131 −0.226 0.385 0.147RANK of Eu 0.278 −0.112 0.303 0.566 0.181 0.350RANK of F 0.663 −0.263RANK of HCO3 0.684 0.125 0.495 0.203 0.151 0.280 0.143RANK of Hf 0.288 0.716 0.150 0.149 0.118 −0.104RANK of Ho 0.264 0.766 0.152 0.126 0.320RANK of I 0.121 0.133 0.901RANK of K 0.872 0.226 0.162 0.124 0.177RANK of Mg 0.473 0.770 −0.225 0.151 0.249RANK of Mn 0.478 0.513 0.344 0.266 0.122RANK of Na 0.528 0.312 0.142 0.341 0.554 0.120RANK of Ni 0.228 0.368 0.380 0.715 −0.149RANK of Sc 0.179 0.409 0.443 0.135 0.353 0.123 0.545RANK of Sr 0.573 0.708 0.276RANK of SO4 0.614 0.299 0.128 −0.581 0.167 −0.148RANK of Ta 0.573 0.103 0.138 0.535 0.348 −0.214RANK of Tb 0.164 0.725 −0.100 0.483 0.223RANK of U 0.151 0.114 0.302 −0.605 0.215 −0.453RANK of Al 0.118 0.769 −0.389 −0.257RANK of Be 0.798 0.333 0.167 0.104 0.211RANK of Fe −0.158 0.609 0.196 0.491 0.316RANK of Ge 0.733 0.375 0.310 −0.101RANK of Li 0.862 0.170 0.303 0.109 0.186 0.214RANK of Mo −0.480 −0.148 −0.303 0.529 0.202 −0.458RANK of NH4 0.262 0.107 0.769RANK of PO4 0.772 0.172RANK of Pb 0.665 0.463 −0.295 0.269RANK of Rb 0.880 0.198 −0.161 0.168RANK of Th 0.236 0.817 −0.136 −0.232RANK of Zr 0.395 0.654 0.232 0.396RANK of SIO2 −0.193 0.176 −0.220 0.862

Fig. 5.Histograms of Ca and Na in the studied bottled waters, based on the BGR Dataset. A. Dissolved Ca concentration. B. Dissolved Na concentration. For clarity of the histogram, fouroutliers for Na are not shown. Note the presence of distinct groups.

312 U. Fugedi et al. / Journal of Geochemical Exploration 107 (2010) 305–316

and Ti are not group-forming components (i.e., they are not takingpart in the formation of any significant factors), so these elementswere excluded from the factor analysis dataset. In the second part ofthe analysis, the factors became well-defined (Table 6). However, ithas to be noted that factor weights still remain quite small, due to thestrong heterogenity and low sample number of the dataset: the firstfactor has a loading of 22%, while the second 18%. The within groupscatter increases with increasing TDS, while the most coherent groupsof samples are formed from waters with low overall TDS. Thefollowing classification of bottled waters was derived.

The outlying TDS group consists of four bottled medicinal watersFerenc József, Hunyadi János, Mira and Bükkszéki Salvus (group 1;Table 1)with ECN19,000 μS/cm and TDSN15,000 mg/l, respectively. Inthese waters Ca2+bMg2+, and Na+ concentration is about ten timesmore than Ca2+. The first three waters (sub-group 1a; Table 1) are ofCa2+-Mg2+-SO4

2− type, due to their potassium-sulphate (Na2SO4) andmagnesium-sulphate (MgSO4) content. These three waters (Fig. 2)originate from Oligocene clastic sedimentary rock aquifers. FerencJózsef, andHunyadi János has anomalous trace element concentrationsin B, Bi, Br, Cd, Ce, Co, Cr, Cu, In, Mn, Ni, Se, Sr, Ta, Te, Th, U, Zn; REE: Ce,Dy, Ho, Lu, Sm, Tb, Tm, Y, Yb,whileMira is anomalouswith respect to B,Bi, Br, Cd, Co, Cr, I,Mn,Mo, Ni, Sc, Se, Ta, Te, Th, U, V; REE:Ho, Lu, Sm, Tb,Tm. Bottled waters Bükkszéki Salvus is of the Na+-HCO3

− type, due toits potassium-sulphate (Na2SO4) and magnesium-sulphate (MgSO4)content (sub-group 1b; Table 1). This water originates from Eocenelimestone formations. Bükkszéki Salvus has anomalous trace elementconcentrations in B, Bi, Br, Cd, Cr, Cs, Ge, I, Li, Mn, Ni, Rb, Te, Th, V, Zr;REE: Eu, Hf, Tm. The high borate content is explained by the geologicalposition of the aquifer near a small producing oil field.

The second group consists of eleven bottled waters with TDS:900–4,500 mg/l (group 2; Table 1). In these waters Ca2+NMg2+ (Ca2+:Mg2+=1.7–3.5, the largest of all groups), while Na+ concentration hasa high variation. The dominant anion is HCO3

−, whereas sulphate andchloride have low concentrations. Three water quality sub-groups areobvious among these samples. The first sub-group (sub-group 2a;Table 1) includes the two special mineral waters from the ore depositareas in the Matra Mts. in the north central part of Hungary, ParádiCsevice I and Parádi Csevice II (Table 1) with Ca:Mg=2.2. The localname ‘csevice’of Slavic originmeans ‘sour’ suggesting their elevatedCO2

content. These waters have high TDS content (4,445 and 1,464 mg/l)and an overall Na+-Ca2+-HCO3

− character (Fig. 6), with elevatedconcentrations in such trace elements as Be, Fe, Mn, Ta, Pb, Sn, Ti, REE:Dy, Er, Gd, Ho, La, Lu, Nd, Pr, Sm, Y, Yb. Both waters have a characteristicH2S odour, which explains the elevated amount of reduced formof NH4

+

in Parádi Csevice I. Due to ore mining activities in the vicinity of the

Parádi Csevice II well, however, it has lost its reducing chemistry and itscomposition is now characterised by dissolved Ag, As, Cu, Mo, Sb, U,SO4

2− and SiO2, which is typical to sulphide ore minerals leachate.The second sub-group (sub-group2b;Table1)hasfivebottledmineral

waters of Ca2+-HCO3− character with elevated Mg2+ content: Ca2+:

Fig. 6. Chemical characterisation of studied bottled waters, based on the BGR Dataset. A. The dominant cation and anion composition of studied bottled waters. B. Piper diagram ofstudied bottled waters.

313U. Fugedi et al. / Journal of Geochemical Exploration 107 (2010) 305–316

Table 7Summary statistics of selected trace elements in the BGR Dataset (n=36), and comparison with international water quality standards (EU, Codex, WHO). Bold numbers denotevalues above one of the standards.

Element min. median max. mad EU standards Codex standard WHO standards

Antimony (μg/l) 0.094 0.365 0.65 0.102 5 5 20Arsenic (μg/l) 0.17 0.62 9.64 3.27 10* 10* 10 (P)Barium (mg/l) b0.002 0.097 0.577 0.096 1 0.7 0.7Boron (μg/l) 5.9 45.05 40,300 438 – 5000 500 (T)Cadmium (μg/l) b0.001 0.0023 0.051 0.001 3 3 3Chromium (μg/l) b0.05 0.0652 0.854 0.0386 50 50* 50Copper (μg/l) 0.097 0.438 14.3 0.284 1000 1000 2000Fluoride (mg/l) 0.031 0.22 2.71 0.19 5 1.5 1.5Lead (μg/l) 0.0037 0.023 1.14 0.0192 10 10 10Manganese (mg/l) b2 17.5 884 17 0.5 0.4 0.4Nickel (μg/l) b0.02 0.137 6.67 0.09 20 20 70Nitrate (mg/l) b0.02 0.198 9.63 0.193 50 50 50Nitrite (mg/l) b0.005 b0.005 0.988 – 0.1 0.1 3 (P)Selenium (μg/l) b0.01 0.0064 0.754 – 10 10 10

min.: minimum, max.: maximum, Me.: median, mad: median absolute deviation.P: Provisional guideline value, as there is evidence of a hazard, but the available information on health effects is limited.T: Provisional guideline value, because calculated guideline value is below the level that can be achieved through practical treatment methods, source protection, etc.* Calculated as total.Source for EU standards: Commission Directive 2003/40/EC of 16 May 2003 establishing the list, concentration limits and labelling requirements for the constituents of naturalmineral waters and the conditions for using ozone-enriched air for the treatment of natural mineral waters and spring waters.http://www.ec.europa.eu/food/food/labellingnutrition/water/index_en.htm.Source for Codex standards: The revised Codex standard for ‘natural mineral waters’(4), Seventh Session of the Codex Committee on Natural Mineral Waters (CCNMW) to be held inFribourg, Switzerland, 30 October1 November, 2000.http://www.fsis.usda.gov/OPPDE/rdad/FRPubs/00-041n.htm.Source for WHO standards: WHO and FAO limits for mineral and drinking waters: http://www.ec.europa.eu/food/fs/ifsi/eupositions/ccnmw/cl07_25e.pdf.

314 U. Fugedi et al. / Journal of Geochemical Exploration 107 (2010) 305–316

Mg2+=1.7–2.5. Accordingly, these waters (Margitszigeti Kristályvíz,Theodora Kereki, Theodora Kékkúti, Balfi and Visegrádi) originate fromcarbonate rock aquifers. The groundwater extracted from the Balfi wellhas hydrogen-sulphide content, but it is eliminated during technologicalprocessing, leaving the water with a Ca2+-HCO3

− character. Sulphideelimination, and accompanying oxidation, explains the lowNH4

+ content.Slight anomalies are found in such trace elements as Be and Cs in thisgroup.

Finally, the third sub-group (sub-group 2c; Table 1) has four bottledmineral waters of Ca2+-HCO3

− character with Ca2+:Mg2+=2.6–3.5,including Mohai Moha II, Mohai Ágnes, Aqua Mathias and Aqua Natural.TheMohaiÁgnes is the oldestmineralwater site inHungary, it has naturalCO2 content and the lowest sodium concentration among all the bottledwaters. Typical trace components are Mn, SiO2 and Eu in this group.

The last group consists of twenty-one bottled waters with(TDSb900 mg/l) (group 3; Table 1). They are all of HCO3

− type withlow sulphate and chloride content, and Ca2+:Mg2+ ratio of 1.2–3.8.Further sub-groups can be identified based on Ca2+:Na+ ratio. The firstsub-group (sub-group 3a; Table 1) has one bottled water (Fonyódi) ofNa+-HCO3

− typewithNa+:Ca2+ ~ 4. The lack of trace element anomaliesis readily explained by the applied iron-removing technologicalprocessing. Sub-group 3b (Table 1) includes 8 samples (Harmatvíz,Ave, Pávai Vajna, AquaSol, Pannon Aqua, Aquarius, Apenta Optima andDogerita) with about the same level in dissolved Ca2+, Mg2+ and Na+,and with no trace element anomalies. The third sub-group (sub-group3c; Table 1) includes 7 samples (NestléAquarel,Natur Aqua, SantaAqua,Mizse, Primavera, Óbudai Gyémánt, and Fonte Verde) having about thesame Ca2+ and Mg2+ (Ca2+:Mg2+=1–1.5) content but with muchlower dissolved Na+ concentratrions (Ca2+:Na+=3,9–23). Anoutstanding feature is the extreme Tl content of bottled water ÓbudaiGyémánt (0.611 μg/l), which is 16 times more than the second largestconcentration (Hunyadi János: Tl=0.0367 μg/l). The very last sub-group (sub-group 3d; Table 1) is characterised by the dominance ofdissolved Ca2+ (Ca2+:Mg2+N1.6; Ca2+:Na2+=3.7–8.1) for 5 bottledwater samples (Szentkirályi, Aqua Friss, Germán Aqua, Veritas Gold,Aqua+Oxigén plusz).

Summarising the above findings, it can be concluded that thecharacter and classes of bottled waters, identified from the laboratoryanalysis and from the bottle label chemical composition, reflect

primarily the quality of the bottled water rather than the geochemicalcomposition of the extracted mineral water. Iron is precipitated fromalmost all waters coming from Pliocene–Pleistocene lithologies in thesouth-western part of the country, like bottled water Fonyódi, forexample (Fig. 2). Similarly, reductive hydrogen-sulphide waters areaerated and sulphur is removed, like from bottled water Balfi. Carbon-dioxide is commonly added to bottled waters or, just on the contrary, itmay be removed. The Aqua+Oxigénpluszwater is treated, for example,with oxygen addition. During the bottling procedure, waters are simplyaerated and, thus, oxidised through which hydrogen-sulphide oxidisesinto sulphate, ammonium turns into nitrite and nitrate, while multi-valent metals (iron and manganese, and to a lesser extent cobalt andchromium) precipitate, as their oxyhydroxide colloids often absorbingand coprecipitating other trace elements, such as arsenic, nickel andzinc, for example. It can be stated, therefore, that the industrialtreatment of extracted mineral waters tends to eliminate the originalgeochemical differences among the natural groundwater sources and itmakes the composition of the marketed waters uniform.

4.4. Trace element content

Comparing the analytical results of this study with European,international and national drinking water standards, it seems thatsome bottled waters contain trace elements above the legislativelimits (Table 7). This is similar to the findings of Naddeo et al. (2008)for Italian bottled waters. For example, the EU limit for B is 1 mg/l,while it is 0.7 mg/l for the Codex standard (see Table 7). In Hungary,5 mg/l of dissolved B is classified as ‘tolerable’. A problem arises withrespect to boron, because it cannot easily be removed from naturalwaters by means of common technological processing. Therefore,some of the studied bottled waters have a boron content above thehealth standards (Table 7). Since boron tends to remain in solution,and accumulate in evaporating alkaline (pHN9.2) brine waters asborate (B(OH)4), it can impact underlying groundwater composition.The studied medicinal bottled waters Mira (B=0.9 mg/l), FerencJózsef and Hunyadi János (both with B≅1.4 mg/l), coming from thistype of geological setting (Fig. 2), have elevated B content. Theoutlying B content in bottled water Salvus (40.3 mg/l) is explained by

315U. Fugedi et al. / Journal of Geochemical Exploration 107 (2010) 305–316

its origin from an aquifer close to hydrocarbon reservoir where boronis associated with organic compounds.

Another example is fluoride, where five of the studied waters(Margitszigeti Kristályvíz, Theodora Kereki, Nestlé Aquarel, Visegrádi,Salvus) exceed the 1.5 mg/l limit (Table 7). Two of them (MargitszigetiKristályvíz, Theodora Kereki) state fluoride content on bottle labels,while the others show F− concentrations below the internationallimits.Manganese concentration is above the 0.5 mg/l EU limit only forbottled water Parádi Csevice I (Mn=0.884 mg/l) originating from anaquifer with ore mineralization. If the recent WHO recommendation(WHO, 2008) is followed, and limits are lowered to the 0.4 mg/l level,further three studied bottled waters have Mn content that exceed thisnew limit. While thallium does not have defined legislative healthstandards, this element is very toxic. The exceptional Tl level in bottledwater Óbudai Gyémánt (Tl=0.611 μg/l) exceeds by 16 times thesecond highest value (Hunyadi János: Tl=0.0367 μg/l). Finally, theselenium content of the studied waters (≤0.754 μg/l) remains well-below the health limits of 10 μg/l, especially in the south-western partof the country (Bogye, 2000). Thus, since Se is an essential element,groundwater tends to be depleted with respect to selenium inHungary. Selenium is above 0.5 μg/l only in five of the studied bottledwaters. Besides the three medicinal waters, Mira, Ferenc József andHunyadi János, bottled waters Santa Aqua and Óbudai Gyémánt havethe desirable dissolved Se concentrations (Bogye, 2000).

5. Conclusions

Bottled mineral waters are extracted, not from randomly selectedwells, but from producing wells carefully designed for maximumwater yield from specific aquifers. Therefore, their chemical compo-sition is limited by business and marketing objectives and, thus,represent only a few aquifers in the country. In this way, these watersamples cannot be used for the regional mapping of groundwaterquality, due to the biased spatial representation. The technologicalprocessing line from the producing well to the packing in PET bottlesseriously alters the original water composition, rendering the bottledwater chemical composition different from the natural source waterchemistry. According to EU legislation (Directive 80/777/EEC;Directive 96/70/EC), treatment of mineral water is restricted toremoval of unstable elements, such as iron and sulphur compounds.Treatment of such mineral waters is restricted to only filtration ordecanting with oxygenation. Free carbon dioxide may be removedonly by physical methods, and the regulations for introduction (orreintroduction) of CO2 are strictly defined. Disinfection of naturalmineral water is completely prohibited, including the addition of anyelement that is likely to change bacterial colony counts. The impact oftechnology onwater chemistry, therefore, differs fromwater to water,which makes the evaluation of bottled mineral water compositioneven harder. Technological processing seems to “uniform” the originaldiverse natural composition of bottled waters.

Results of this study, suggest that bottled mineral water concen-trations of major dissolved components (Ca2+, Mg2+, K+, Na+, Cl−),are unaffected by technological processing, together with some traceelements, can be considered to partially represent and characterisethe original groundwater in the exploited aquifers. This study mayinspire more detailed hydrogeochemical and hydrogeological studies.

Acknowledgements

We thank Manfred Birke for the generous laboratory free analysesof bottled water samples at the Federal Institute for Geosciences andNatural Resources in Germany. We also thank the EuroGeoSurveysGeochemistry Expert Group for the research opportunity, especiallyClemens Reimann, Benedetto de Vivo and Alecos Demetriades.Review of the manuscript and useful suggestions by György Tóth ofthe Geological Institute of Hungary are gratefully acknowledged. The

thorough reviews by Peter Hayoz and an anonymous reviewer werevery helpful. The assistance of the U.S.A. Fulbright Programme grant,the Hungarian–American Enterprise and Scholarship Fund (HAESF)grant, The Eotvos Grant of the Hungarian Scholarship Board, theBolyai Janos Research Grant of the Hungarian Academy of Science andthe Norwegian Financial Mechanism – Hungarian Science Fund OTKAare gratefully acknowledged.

References

Arnold, E., 2006. Bottled Water: Pouring Resources Down the Drain. Copyright © 2006Earth Policy Institute. 2006. February 2. http://www.bvsde.paho.org/bvsacd/cd47/bottled.pdf Last accessed on 29/09/2010.

Birke, M., Reimann, C., Demetriades, A., Rauch, U., Lorenz, H., Harazim, B., Glatte, W.,2010. Determination of major and trace elements in European bottled mineralwater – analytical methods. In: Birke, M., Demetriades, A., De Vivo, B. (Eds.),Mineral waters of Europe: Special Issue, Journal of Geochemical Exploration, 107 (3),pp. 217–226 (this volume).

Bogye, G., 2000. Selenium deficiency in Hungary - nutritional and clinical aspects. PhDThesis, SOTE University, Budapest 93pp. (in Hungarian).

Borszeki, B., 1979. Mineral waters, medicinal waters. Mezőgazdasági Kiadó, Budapest,371pp. (in Hungarian).

Chiarenzelli, J., Pominville, C., 2008. Bottled water selection and health considerationsfrom multi-element analysis of products sold in New York state. Journal of Waterand Health 6 (4), 505–512.

Directive 2000/60/EC, 2000. Directive 2000/60/EC of the European Parliament and ofthe Council establishing a framework for the Community action in the field of waterpolicy. Official Journal of the European Communities L327, 1–72.

Directive 2009/54/EC, 2009. Directive 2009/54/EC of the European Parliament and ofthe Council on the exploitation and marketing of natural mineral waters. OfficialJournal of the European Union L164, 45–58 26/6/2009.

Directive 80/777/EEC, 1980. Council Directive 80/777/EEC of 15 July 1980 on theapproximation of the laws of the Member States relating to the exploitation andmarketing of natural mineral waters. Official Journal of the European CommunitiesL229, 11–29.

Directive 96/70/EC, 1996. Directive 96/70/EC of the European Parliament and of theCouncil of 28 October 1996 amending Council Directive 80/777/EEC on theapproximation of the laws of the Member States relating to the exploitation andmarketing of natural mineral waters. Official Journal of the European CommunitiesL299, 26–28 23/11/1996.

Erdelyi, M., 1975. Hydrodynamics of the Hungarian Basin. Hidrológiai Közlöny 55,147–156 (in Hungarian).

Fiket, Z., Roje, V., Mikac, N., Kniewald, G., 2007. Determination of arsenic and other traceelements in bottled waters by high resolution inductively coupled plasma massspectrometry. Croatia Chemica Acta 80 (1), 91–100.

Fodor, L., Bada, G., Csillag, G., Horvath, E., Ruszkiczay-Rüdiger, Z., Sikhegyi, F., 2005. Newdata on neotectonic structures and morphotectonics of the western and centralPannonian Basin. Proceedings of the workshop on Application of GPS in platetectonics. Occasional Papers of the Geological Institute of Hungary 204, 35–44.

Garzon, P., Eisenberg, M.J., 1998. Variation in the mineral content of commerciallyavailable bottled waters: implications for health and disease. The American Journalof Medicine 105 (2), 125–130.

Government Decree 201/2001. (X. 25.) on the standards and inspection procedures ofdrinking water quality Magyar Közlöny, 2001/118, 8188.

Government Decree 65/2004. (IV. 27.) FVM-ESzCsM-GKM joint decree on the bottlingand marketing rules of natural mineral water, spring water, drinking water, waterwith added-minerals and flavoured water. Magyar Közlöny, 2004/57, 5916.

Groselj, N., Van der Veer, G., Tusar, M., Vracko, M., Novic, M., 2008. Verification of thegeological origin of bottled mineral water using artificial neural networks. FoodChemistry 118, 941–947.

Guler, C., 2007. Characterization of Turkish bottled waters using pattern recognitionmethods. Chemometrics and Intelligent Laboratory Systems 86 (1), 86–94.

Hungarian Mineral Water Association and Product Council, 2009. http://www.asvanyvizek.hu Last accessed on 2010/09/29.

Ikem, A., Odueyungbo, S., Egiebor, N.O., Nyavor, K., 2002. Chemical quality of bottledwaters from three cities in eastern Alabama. The Science of the Total Environment285 (1–3), 165–175.

Kay-Shuttleworth, R., 2009. The ebb andflowofwater. Trends across the EU. Summaryof Canadian's presentation for EFBW, June 2009. http://www.asvanyvizek.hu/js/tinymce/plugins/filemanager/files/efbw/Summary_of_canadeans_presenta-tion_for_EFBWN2_06_09-1.doc Last accessed on 2010/09/29.

Kochubovski, M., Aleksoski, B., Kendrovski, V., Dimovska, M., 2007. Review of bacterio-logical andphysicochemical quality of naturalmineralwater produced in theRepublicof Macedonia. Journal of Environmental Protection and Ecology 8 (4), 950–955.

Korossy, L., 1968. The paleographic evolution in the Lower-Pannonian in Hungary. ActaGeologica Hungarica 12, 1–4.

Misund, A., Frengstad, B., Siewers, U., Reimann, C., 1999. Variation of 66 elements inEuropean bottled mineral waters. The Science of the Total Environment 243–244,21–41.

Naddeo, V., Zarra, T., Belgiorno, V., 2008. A comparative approach to the variation ofnatural elements in Italian bottled waters according to the national andinternational standard limits. Journal of Food Composition and Analysis 21 (6),505–514.

316 U. Fugedi et al. / Journal of Geochemical Exploration 107 (2010) 305–316

Reimann, C., Birke, M. (Eds.), 2010. Geochemistry of European bottled water. BorntraegerScience Publishers, Stuttgart, p. 268. URL: http://www.schweizerbart.de/publications/detail/isbn/9783443010676/Geochemistry-of-European-Bottled-Water - Lastaccessed on 24/9/2010.

Saleh, M.A., Abdel-Rahman, F.H., Woodard, B.B., Clark, S., Wallace, C., Aboaba, A., Zhang,W., Nance, J.H., 2008. Chemical, microbial and physical evaluation of commercialbottled waters in greater Houston area of Texas Journal of Environmental Scienceand Health part A. Toxic/Hazardous Substances & Environmental Engineering 43(4), 335–347.

Schulhoff, O., 1957. Mineral and medicinal waters of Hungary. Akadémiai Kiadó,Budapest. 269pp.

Somlai, J., Horvath, G., Kanyar, B., Kovacs, T., Bodrogi, E., Kavasi, N., 2002. Concentrationof (super 226) Ra in Hungarian bottled mineral water. Journal of EnvironmentalRadioactivity 62 (3), 235–240.

SPSS® Base 8.0, 1998. Application Guide. SPSS Inc, Chicago.Szanto, Z., Futo, I., Braun,M., 2008. Analysing theoriginof bottledmineralwaters bymeans

of stable isotopes. In: Orosz, Z. (Ed.), Proceedings of the 4th Carpathian EnvironmentalScience Conference. Hungary, Debrecen, pp. 256–262 (in Hungarian).

Trunko, L., 1996. Geology of Hungary. Gebrüder Borntraeger, Berlin. 240pp.Versari, A., Parpinello, G.P., Galassi, S., 2002. Chemometric survey of Italian bottled

mineral waters by means of their labelled physico-chemical and chemicalcomposition. Journal of Food Composition and Analysis 15 (3), 251–264.

WHO, 2008. Guidelines for drinking-water quality, third edition, incorporating first andsecond addenda. Volume 1. Recommendations. World Health Organisation, Geneva,515 pp.