Science of the Total Environment 484 (2014) 19–26

Contents lists available at ScienceDirect

Science of the Total Environment

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

Electrical conductivity and emerging contaminant as markers of surfacefreshwater contamination by wastewater

Diana Nara Ribeiro de Sousa a, Antonio Aparecido Mozeto a, Renato Lajarim Carneiro b, Pedro Sergio Fadini a,⁎a Laboratório de Biogeoquímica Ambiental (LBGqA), Núcleo de Diagnósticos e Intervenções Ambientais (NEDIA), Departamento de Química, Universidade Federal de São Carlos — UFSCar,Rodovia Washington Luís, km 235, SP-310, São Carlos 13.565-905, SP, Brazilb Grupo de Quimiometria Aplicada (GQA), Departamento de Química, Universidade Federal de São Carlos— UFSCar, Rodovia Washington Luís, km 235, SP-310, São Carlos 13.565-905, SP, Brazil

H I G H L I G H T S G R A P H I C A L A B S T R A C T

• Emerging contaminants as markers ofwastewater contamination

• Linear correlation between electricalconductivity and pharmaceuticalproducts

• Principal component analysis as a toolfor the identification of polluted sites

• Unambiguous anthropogenic pollutionmarkers

⁎ Corresponding author at: Laboratório de BiogeoquímiQuímica, Universidade Federal de São Carlos, Brazil. Tel1633518212.

E-mail address: psfadini@ufscar.br (P.S. Fadini).

http://dx.doi.org/10.1016/j.scitotenv.2014.02.1350048-9697/© 2014 Elsevier B.V. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 25 November 2013Received in revised form 28 February 2014Accepted 28 February 2014Available online xxxx

Keywords:Emerging contaminantsElectrical conductivityWastewatersPrincipal component analysis

The use of chemical markers of undoubted anthropogenic sources for surface freshwater contamination by waste-waters was evaluated employing correlations observed between measured physico-chemical parameters as theelectrical conductivity and the concentration of different emerging organic compounds. During the period fromApril/2011 to April/2012 spatial–temporal variations and contamination patterns of two rivers (Piraí and Jundiaírivers), São Paulo state, Brazil were evaluated. Seven physico-chemical parameters and concentrations of differentclasses of emerging contaminants were determined in samples collected in seven field campaigns. The high linearcorrelation coefficients obtained for the compounds diclofenac (r = 0.9085), propanolol (r = 0.8994), ibuprofen(r = 0.8720) and atenolol (r = 0.7811) with electrical conductivity, also corroborated by principal componentanalysis (PCA), point to the potential use of these compounds asmarkers of investigated surfacewater contamina-tion by wastewaters. Due to specific inputs, these environmental markers showed very good effectiveness for theidentification and differentiation of water body contamination by discharges of treated and untreated urbansewage.

© 2014 Elsevier B.V. All rights reserved.

ca Ambiental, Departamento de.: +55 1633518065; fax: +55

1. Introduction

The contamination of aquatic ecosystems with discharges of treatedand/or non-treated wastewaters is a historical challenge to civilization(Schwarzenbach et al., 2010). The changes in the environmental context

Table 1Description of the sampling sites and their geographical location coordinates.

Samplingsite

Description Location Latitude (S) Longitude (O)

S1 Piraí River Near the river source,Itupeva

23.26201° −47.05790°

S2 Piraí River Near the river mouth,Indaiatuba

23.18542° −47.23906°

S3 Jundiaí River Campo Limpo Paulista 23.20717° −46.78303°S4 Jundiaí River Downstream WWTP,

Jundiaí23.14015° −47.03800°

S5 Jundiaí River Industrial district,Indaiatuba

23.13860° −47.21755°

S6 Jundiaí River River Mouth, Salto 23.21055° −47.29176°

20 D.N.R. de Sousa et al. / Science of the Total Environment 484 (2014) 19–26

caused by the strong development in the economic activities and urbangrowth have been assigned as one of the main effects on the accelera-tion of the water resource degradation. Regions with high industrializa-tion and strong demand for water for multiple use often present waterquality and availability problems.

Usually, wastewater treatment plants (WWTP) are designed aimingat the removal of solids, dissolved organic matter and nutrients, where-as specific classes of chemical compounds are not always targeted in thetreatment systems. Among these compounds there are the emergingcontaminants, which include those of different classes such as pharma-ceuticals, steroids, hormones, personal care products, flame retardants,drugs of abuse, and others (Daughton and Ternes, 1999; Kolpin et al.,2002). These compounds have been of great environmental concernwhich has encouraged the development of an increasing number ofstudies and publications which have appeared in the current literature.

Many of these substances, obeying their degradation characteristics,have presented a regular and constant consumption pattern givingthem behavior of pseudo-persistence in the environment (Daughton,2002). These contaminants have been detected in a wide variety of ma-trices that include environmentalmatrices as surfacewater and ground-water (Grujic et al., 2009 and Yoon et al., 2010)wastewater (Lacey et al.,2008 and Yu et al., 2011) soils and sediments (Tadeo et al., 2012; Xuet al., 2008) and biological materials (Fick et al., 2010; Gelsleichter andSzabo, 2013; Subedi et al., 2012). The concentration values detected inthese matrices are currently measured using the LC–MS/MS techniquethat allows the use of these compounds as chemical markers of effluentdischarge from domestic waste.

Among a wide range of compounds, caffeine has been one of themost studied compounds and most frequently indicated as a goodchemical marker of anthropogenic contamination of surface waters bywastewaters, mainly due to its correlation of average per capita con-sumptions (Buerge et al., 2003; Buerge et al., 2006), microbiologicalmarkers (Kurissery et al., 2012; Peeler et al., 2006), and nitrate (Seileret al., 1999). Currently, other emerging contaminants as human phar-maceuticals have been studied as potential indicators of contaminationby the discharge of domestic sewage (Clara et al., 2004; Gasser et al.,2010; Gasser et al., 2011; Vystavna et al., 2012; Young et al., 2008).

More recently artificial low-calorie sweeteners have been pointed asgood markers of wastewater contamination (Harwood, 2014;Robertson et al., 2013). The presence of these substances in the environ-ment is associated with sewage discharges, due the poor alteration bythe human metabolism, after their ingestion and persistence in the en-vironment. Buerge et al. (2009) verified the presence of four sweetenersin differentmatrices, such aswastewaters, surfacewaters, groundwaterand drinkingwaters, in the Switzerland regions. This study demonstrat-ed the potential use of the acesulfame-K as a marker. Similar resultsconcerning acesulfame-K and sucralose were reported by Scheureret al. (2009) in Germany.

As markers of pollution, pharmaceuticals and other emerging con-taminants aremore effective thanmicrobiologicalmarkers as the chem-ical analyses are time consuming and do not discriminate betweenhuman and animal sources of contamination. An ideal marker shouldpromote the identification of the kind of contamination source andthe degree of pollution. In this context, the electrical conductivity hasbeen investigated as a marker of pollution by wastewater discharges(Chalupová et al., 2012; Stewart, 2001; Thompson et al., 2012). Theelectrical conductivity measurement by itself is useful in providing ascreening of the pollution level but, when in association with emergingcontaminant concentrations, can provide unambiguous informationabout anthropogenic sources of contaminant discharges.

In the present study the occurrences and concentrations of 11emerging contaminants, including caffeine, were studied aiming at theevaluation of the potential of these compounds as chemical markers ofsurface water contamination by wastewaters. To investigate thishypothesis, these compounds were analyzed in surface waters duringa period of one year of sampling totaling 7 campaigns in the

abovementioned rivers. Concentration values of these chemical com-pounds and physico-chemical characteristics were also evaluated byprincipal component analysis seeking correlations between measuredemerging contaminant concentrations and their occurrence and classi-cal pollution indicators like electrical conductivity.

2. Experimental

2.1. Sampling area

Located within the metropolitan region of São Paulo and Campinascity, the Jundiaí River Basin, with its 1.114 km2, is characterized bywater scarcity as a result of high urbandensity and intense industrial ac-tivities in the region. This watershed is also impacted by the discharge,into the water bodies, of treated and untreated domestic and industrialwastewaters generated by about 1 million of inhabitants. Description ofthe sampling site locations is presented in Table 1. Samplings were car-ried out during the months of April/2011 to April/2012, every twomonths (7 different dates). For the analysis of emerging contaminants,a total of 42 samples were collected from two sites along the PiraíRiver (Sites S1 and S2) and four sampling sites along the Jundiaí River(Sites S3–S6). These sampling sites are illustrated in Fig. 1.

Sites S1 and S2 are considered low pollution degree areas belongingto a relatively cleanwater body used as awater source for public supply.From Site S3 up to Site S6, the pollution level increases from a relativemedium pollution level in Site S3 that evolves into Sites S4 and S5,until a high degradation level at Site S6. The last three sites are in regionsof high degree of pollution and intensive industrial activity. The rivermouth region, located in the city of Salto (São Paulo state), representsthe zone with the worst water quality of all the extension of the JundiaíRiver.

The pollution degree of rivers in the São Paulo state is evaluated onthe basis of periodicmonitoring of several physical, chemical andmicro-biological parameters in different sampling sites. Results are publishedby the Environmental Company of the São Paulo state (CETESB) in an-nual reports including water quality indices for the different aquaticecosystems (São Paulo, 2013). Nationally, the National Council of Envi-ronment (CONAMA) makes a similar work and in both evaluations, theJundiaí River is classified in the worst possible pollution level.

2.2. Determination of physico-chemical parameters

The physico-chemical parameters were monitored using a multi-parameter probe model YSI 6820 V2-2 (Yellow Springs, Ohio, USA).Electrical conductivity (EC), dissolved oxygen (DO), turbidity (TRB),pH, temperature (TPR) and redox potential (RP) were also determinedin this study as potential chemical markers, all being determined as insitu parameters during the monitoring period. Total organic carbon(TOC) was performed by a Shimadzu TOC-L CPH/CPN analyzer (Tokyo,Japan).

Fig. 1.Map of the catchment of the Jundiaí River Basin showing the location of sampling sites.

21D.N.R. de Sousa et al. / Science of the Total Environment 484 (2014) 19–26

2.3. Chemical reagents

All standard compounds usedwere of high purity grade (N96%). Caf-feine (CFN), ibuprofen (IBP), naproxen (NPX), diclofenac (DCF), carba-mazepine (CBZ), atenolol (ATL), propanolol (PPL), triclosan (TCS),estrone (E1), 17-β-estradiol (E2), and 17-α-ethinylestradiol (EE2)were supplied from Dr. Ehrenstorfer (Augsburg, Germany). Isotopicallylabeled compounds, used as internal standards, caffeine-d3, naproxen-d3, atenolol-d7, carbamazepine-d10, ibuprofen-d3, diclofenac-d4,triclosan-d3, propanolol-d7, estrone-d4, 17-β-estradiol-d5, and 17-α-ethinylestradiol-d4 were obtained from CDN Isotopes (Quebec,Canada). Table 2 shows some characteristics of these compounds.

HPLC-grade methanol and acetonitrile were supplied by JT Baker(Ecapetec, Mexico), HPLC-grade acetone was from Mallinckrodt(Phillipsburg, USA), and NH4OH was from Sigma-Aldrich (Schweiz,

Table 2Characteristics of the compounds studied.

Compound Compound class Log Kow pKa

Atenolol β-Blocker 0.16 9.6Caffeine Stimulant −0.07 6.1Carbamazepine Antiepileptic 2.47 7.0Diclofenac Anti-inflammatory 4.51 4.14Ibuprofen Analgesic 3.97 4.91Naproxen Anti-inflammatory 3.18 4.15Propanolol β-Blocker 3.48 9.42Triclosan Antiseptic 4.76 8.0Estrone Hormone 3.13 10.317-α-Estradiol Hormone 3.94 10.417-β-Ethinylestradiol Hormone 4.01 10.5

Gros, M.; Petrovic, M.; Barceló, D. Development of a multi-residue analytical methodologybased on liquid chromatography–tandem mass spectrometry (LC–MS/MS) for screeningand trace level determination of pharmaceuticals in surface and wastewater. Talanta2006; 70:678–690.Rodil, R.; Quintana, J. B.; López-Mahía, P.; Muniategui-Lorenzo, S.; Padra-Rodríguez, D. J.Multi-residue analytical methods for the determination of emerging pollutants in waterby solid-phase extractions and liquid chromatography–tandem mass spectrometry. J.Chromatogr. A, 2009; 1216:2958–2969.Yoon, Y.; Ryu, J.; Oh, J.; Choi, B.-G.; Snyder, S. A. Occurrence of endocrine disruptingcompounds, pharmaceutical, and personal care products in the Han River (Seoul, SouthKorea). Sci. Total Environ. 2010; 408: 636–643.

Switzerland). Ultrapure water used (18.2 MΩ·cm) was obtained froma Milli-Q device manufactured by Millipore (Merck Millipore, SãoPaulo, Brazil). The 99.9% pure nitrogen used for drying was suppliedfrom White Martins (São Paulo, Brazil).

Most of the investigated pharmaceuticals are available in the Popu-lar Pharmacy Program of the Ministry Health, which offers drugs oflow cost to population (Brasil, 2013). Although Brazil is among the larg-est consumers of drugs in the world according to IMS Health (2011),accurate consumption data are scarce due mainly to ease of purchaseof medicines, without prescription, even if it is obligatory, as the caseof diclofenac and other compounds widely consumed.

2.4. Sample preparation

Samplesweremanually collected in amber glass bottles and vacuumfiltered through 1.2 μmand0.7 μmglass fiberfilters followedby 0.45 μmnylonmembrane filters fromWhatman™ (Little Chalfont, Buckingham-shire, UK). Following this step, theywere stored at−20 °C in a refriger-ator for the subsequent extraction by solid phase extraction (SPE) usingOasis HLB (200 mg, 6 cm3) cartridges supplied by the Waters Corpora-tion (Milford, Massachusetts, USA). A vacuum manifold processingstation (Agilent Technologies, USA) was used in the concentrationstep. Firstly, the cartridges were conditionedwith 2 × 5mL of methanolfollowed by 2×5mLof ultrapurewater. Then, 500mL of thewater sam-pleswas spiked at 100 ng L−1 of each internal standard and then passedthrough the Oasis HLB cartridges at a flow rate of 5mLmin−1. After thisstep, the cartridges were rinsedwith 2 × 4mL of ultrapure water. To re-move the excess water, the cartridges were dried under vacuum for40 min. Elution was performed with 2 × 3 mL of methanol and 3 mLof methanol:acetone (1:1, v/v). The extract was evaporated under a ni-trogen flow stream in a Dry-Block SL-22/50 system from SolabEquipamentos para Laboratório Ltda (Piracicaba, SP, Brazil) andreconstituted with 1 mL of methanol:water (25:75, v/v). All samplesare extracted and injected in triplicate.

2.5. Analysis of emerging contaminants

Analysis of the target compounds was performed by WatersACQUITYUPLC™ system comprising an ACQUITYUPLC™ binary solvent

22 D.N.R. de Sousa et al. / Science of the Total Environment 484 (2014) 19–26

manager and an ACQUITY UPLC™ sample manager (USA). AnAcquity UPLC BEH C18 (1.7 μm, 2.1 × 50 mm) column, equippedwith a pre-column with similar material, both supplied by Waters(USA), was used in the chromatographic method. The mobile phaseswere ultrapurewater with 0.05% of ammonium hydroxide (Eluent A)and methanol (Eluent B). The gradient used was 0–0.5 min (95% A–5% B), 0.5–1.5 min (55% A–45% B), 1.5–4.0 min (5% A–95% B), 4.0–4.5 min (5%A–95% B), 4.5–4.55 min (95% A–5% B) and 4.55–6.0 min(95% A–5% B). The initial conditions were then re-establishedand held for 6 min. Temperature was kept at 40 °C, the mobilephase flow rate was 0.45 mL/min and the injection volume was10 μL.

A triple-quadrupole mass spectrometer (TQD, Waters, UK)equipped with an electrospray ionization (ESI) source was used.Analyses were performed in positive and negative modes with acapillary voltage optimized to 3000 V for the ESI interface in positiveionization and 2500 V in negative ionization, source temperature of150 °C and desolvation temperature of 500 °C. A cone gas flow of20 L/h and desolvation gas flow of 750 L/h were used. Nitrogenwas used as nebulizing and desolvation gas. The mass spectrometerwas operated in the multiple reaction monitoring (MRM) mode,monitoring two transitions for each compound.

Quantifications of the target compounds was performed accord-ing to criteria defined in the EU Commission Decision, 2002/657/ECfor identification and confirmations this compounds for LC–tandemMS as the instrumental technique. The most abundant transitionwas used for quantification, with a second transition for confirma-tion according to criterion of EC. For the determination of the limitdetection of the analytical method and of the limit quantification,the approach of signal-to-noise was used. These limits wereestablished using spiked water samples, as the minimum detectableamount of analyte with 3 and 10 signal-to-noise, for detection andquantification limits, respectively.

Table 3Occurrence and concentration of studied compounds in the samples collected from thePiraí River (n = 14).

Compounds LOD(ng L−1)

OccurrencesR. Piraí(%)

Mean detected(ng L−1)

Range(ng L−1)

Atenolol 3.38 43 11.0 5.83–16.4Caffeine 1.72 79 80.4 11.1–213.6Carbamazepine 1.12 50 8.53 5.35–14.4Diclofenac 1.33 43 15.9 9.11–29.2Ibuprofen 0.89 7 6.48Naproxen 2.28 0Propanolol 1.56 0Triclosan 1.32 7 4.75Estrone 1.31 017-β-Estradiol 1.81 017-α-Ethinylestradiol 2.14 0

2.6. Principal component analysis (PCA)

Principal component analyses (PCA) were used to evaluate the spa-tial and temporal variations of the physico-chemical parameters andconcentration of emerging contaminants in water samples from thePiraí and Jundiaí Rivers. Firstly, the data set was auto-scaled to expresseach observation in terms of variations intrinsic to the system. Thispre-treatment consisted of the transformation of the data set, so thateach variable presents mean zero and variance one (Moita Neto andMoita, 1998).

PCA is a multivariate method usually applied to identify the correla-tion amongst the studied variables. New variables, the principal compo-nents (PCs), are created in an orthogonal space aiming at themaximization of the variance explained by each PC (Vieira et al.,2012). Then, in the first factors from the PCA, are concentrated the rel-evant information of the data set, and as consequence the later factorscan be excluded without significant loss of data (Tauler, 2000). Theloadings of the PCA show how the variables are correlated (or uncorre-lated) by analyzing the distance between them in the loading plot, thecloser being the more correlated. The scores present the influence ofeach loading in each sample. If two samples present similar value ofscores, this mean that the samples are similar, because they will showsimilar values for the original variables.

The PCA was initially applied to physicochemical parameters only,resulting in a 42 × 7 matrix. Following that, the concentrations of stud-ied compounds were included in the data set in a 42 × 15 matrix. Thehormoneswere excluded from this analysis because their concentrationdid not present significant spatio-temporal variations.

Chemometric analyses were performed using Pirouette 4.0(Infometrix, Inc., Bothell, WA) software.

3. Results and discussion

3.1. Distribution of emerging contaminants in river water samples

In the Piraí River samples, only six of the eleven analyzed com-pounds were analytically detected. The concentration did not exceed30 ng L−1, except for the caffeine. Table 3 shows the detection frequen-cies and mean concentration for each compound of these samples. Caf-feinewas the substance that showed the highest frequency of detection(79%) and the largest mean concentration (80.4 ng L−1) among the an-alyzed compounds. Other compounds more often detected were carba-mazepine, atenolol and diclofenac, with frequencies of 50%, 43% and43%, respectively. Ibuprofen and triclosan were detected only once foreach one in the Piraí River in the sample from the river mouth region.Analyzed hormones were not detected in any of the collected samples,probably due to their relatively high detection limits or because of theinteraction with particulate matter (Duong et al., 2009; Petrovic et al.,2002) while naproxen and propanolol showed trace levels only, withconcentrations below the detection limits of the method. Only caffeineand atenolol were present in both sampling sites, source (S1) andmouth (S2) of the Piraí River. Other compounds were detected only inthe mouth region.

Detection frequencies and average concentrations of each com-pound for 28 samples in the Jundiaí River are shown in Table 4. Exceptthe hormones 17-β-estradiol and 17-α-ethinylestradiol, all the othercompounds were detected in these samples. For the collected set ofwater samples, caffeine was also the compound that presented thehighest detection frequencies, followed by atenolol, carbamazepineand ibuprofen. All the other compounds presented frequencies up to75%,with the exception of estrone. Compounds detectedmost frequent-ly in this study did not necessarily have the highest or most variableconcentration ranges.

Variation inmeasured concentrations reflects the seasonal influenceand heterogeneity of the sampling sites aswell as variations on the tem-poral consumption patterns. Fig. 2 are relative to the sumof the concen-trations for all compounds, obtained from all sampling sites by thesampling campaign. Firstly, the effect of the dilution phenomenon duethe increased river flow is clearly observed in the first month of samplecollection (April/2011), when the concentrations were strongly influ-enced by the rainfall of the previous months.

Fig. 2 shows that the concentration of emerging compounds ishigher in the dry season than in the rainy season. The concentrationvalues shown in Fig. 2 are relative to the sum of the concentrations forall compounds, obtained from all sampling sites by the sampling cam-paign. Firstly, the effect of the dilution phenomenon due the increasedriver flow is clearly observed in the first month of sample collection

Table 4Occurrence and concentrations of studied compounds in the samples collected from theJundiaí River (n = 28).

Compounds LOD(ng L−1)

OccurrencesR. Jundiaí(%)

Mean detected(ng L−1)

Range(ng L−1)

Atenolol 3.38 100 189.6 15.2–412.9Caffeine 1.72 100 6550.9 993.9–19329.7Carbamazepine 1.12 100 130.6 5.76–659.5Diclofenac 1.33 96 108.7 37.3–328.5Ibuprofen 0.89 100 74.2 3.33–208.2Naproxen 2.28 93 29.3 5.14–98.6Propanolol 1.56 86 23.3 3.62–52.7Triclosan 1.32 75 69.5 4.95–323.5Estrone 1.31 25 6.02 4.75–8.0117-β-Estradiol 1.81 017-α-Ethinylestradiol 2.14 0

23D.N.R. de Sousa et al. / Science of the Total Environment 484 (2014) 19–26

(April/2011), when the concentrations were influenced by the rainfallof the previous months.

The observed concentration range indicates that the two studiedwater bodies are exposed to different levels of environmental impact.Although used as a strategic water source by some cities within thebasin, the Piraí River receives the discharge of treated sewage and sur-face runoff of pluvialwaters from the rural zone. For the Piraí River sam-ples, most of the target compounds were detected only at the rivermouth, different from the Jundiaí River, that is strongly affected bywastewater discharge through its course.

3.2. Principal component analysis (PCA)

As a first approach, the PCA was applied using the data set samplesversus physicochemical parameters, which resulted in a 42 × 7 matrix(samples x variables). PCA results showed that 80.9% of the varianceof the original data may be explained by the four main components.Fig. 3 shows scores and loadings for the physicochemical variables.The PC1 represents 33.1% of the data variability and themost influentialvariables in this PC are DO, TOC, EC and TPR. PC2 explained 21.9% of thevariance, where pH and RP are the variables which presented more in-fluence. TRB is the most important factor in the PC3, which explains15.8%. This PCA analysis shows clearly that TOC and EC are highly corre-lated as well as the RP and pH variables.

The score graph shows the separation of Sites S1, S2 and S3 into onegroup, and Sites S4, S5 and S6 into a second group which differentiates

Fig. 2. Total concentrations of

the less and the most impacted sites, respectively. However, these sitesdo not result in a clear cluster because of the lack of similarity betweenthese samples, since they differ in space and time. The water samplesfrom Sites S4, S5 and S6 showed the highest values of the EC, TOC andTPR, and consequently, the lowest concentrations of DO. An inversetrend is observed for the other sample group. These conclusions canbe observed in the scores and loading graphics. Since DO showed highvalues and EC, TOC and TRP showed low values in samples from SitesS1, S2 and S3, these samples presented a negative score and DO valueswith respect to PC1. The same interpretation is valid for samples fromSites S4, S5 and S6 which presented positive PC1 score values and EC,TOC and TPR presented positive loading values, so these parameterspresent high levels for these samples. The other variables do not presentsignificant enough PC1 score values to have an influence.

EC, TOC andDO are extremely important parameters in the analysis ofwater quality. Electrical conductivity (EC) is indeed an indirect indicatorof pollution because it presents a close relationship with the dissolvedsalt content present in the water column of continental water bodiesthat often is associated to sewage discharge and is therefore a well-established water quality parameter (Chalupová et al., 2012; Thompsonet al., 2012). High EC values can usually be associated to the presence ofdomestic sewage due to an increase in the chloride ion concentration,mostly coming from the humandiet, as each person consumes an averageof 6 g of chloride per day (WHO, 2003). This increase results in chlorideconcentrations that exceed 15 mg L−1 in raw sewage.

The TOC content of natural waters represents a direct measurementof the organic matter and therefore also constitutes an excellent indica-tor ofwater quality (Tchobanoglous and Burton, 1991). The obtained re-sults show that the EC and TOC values are strongly correlated and have,consequently, an inverse relationship with the DO values. Sites S4, S5and S6 presented DO concentrations that ranged from 2.45 to 7.09 mgL−1, where the highest concentrations were observed during the rainyseason. The low DO levels are a consequence of organic matter dis-charges in water that demands a high oxygen uptake during its decom-position (Bhuiyan et al., 2011). Dissolved oxygen is an essentialprerequisite to the survival of aquatic organisms and therefore is avery useful indicator of the aquatic ecosystem quality (Akkoyunlu andAkiner, 2012; Kannel et al., 2007; Markfort and Hondzo, 2009).

In the second PCA performed, the concentrations of the eight com-pounds in every sample were included in the previous data set,resulting in a 42 × 15matrix. This PCA aimed at looking for correlationsamong the high concentration of studied compounds and the physico-chemical parameters (see Fig. 4). Results showed that 82.5% of the

compounds and rainfall.

Fig. 3. Scores and loadings for physico-chemical variables.

24 D.N.R. de Sousa et al. / Science of the Total Environment 484 (2014) 19–26

variance can be explained by the first five PCs. The PC1 axis represents44.6% of the data variance and its loadings show that variables DO, TOC,EC and all emerging contaminants present high influence on the PC1.The variables DO, RP, pH and TPR presented high influence on the PC2,which explains 16.0% of the data variance. Through an analysis of thisloading graph, it is possible to see that the high concentrations of allemerging contaminants are correlated with high EC and TOC values. Inaddition to this, high concentrations of the compounds can also be relatedto low DO values which corroborates with the issues considered above.

The score plot presented in Fig. 4 shows the formation of three clus-ters. Themajority of the samples in Group 1 (G1) represent the samplesof Sites S1 and S2 that correspond to the Piraí River, while Group 2 (G2)correspondsmainly to the samples of Site 3 of the Jundiaí River; Group 3(G3) contains the samples from other sites of the Jundiaí River (S4, S5and S6). PC1 was responsible for the observed clustering, because thesamples are scattered when they are projected in PC2.

The Group 3 (G3) samples are characterized by high EC values andhigh concentrations of the analyzed compounds. In this group, a highlyscattered number of samples could be observed, probably due to amoreintense seasonal influence on the levels of contaminants in comparisonto the other sample groups. This aspect may be emphatically observedin samples from Sites S4, S5 and S6 of April/2011 which presented sim-ilar behavior to the Group 2 samples (triangles in the ellipse in the G2).For the sampling Sites S1 and S2, all the samples were grouped regard-less of the sampling period in which all compounds were detected inlow concentrations and presented high DO concentration and low ECand TOC values. These characteristics confirm the low level of anthropo-genic impact on this river when compared to Jundiaí River samples.

The samples of Sites S4, S5 and S6 showed high dispersion, if com-pared with other sites. This behavior can be mainly related to the fact

Fig. 4. Scores and loadings for all s

that the WWTP, located upstream from these sites, receives and pro-cesses different kinds of residues, such as sanitary and industrial waste-waters, leachates, liquid wastes from different sources and regionstransported by trucks fromdifferentmunicipalities,mostly fromoutsidethe watershed. It is postulated that these characteristics resulted in animpact on the composition uniformity of treated effluents that aredischarged into the Jundiaí River. In addition, Sites S5 and S6 are greatlyinfluenced by the remaining organic load from the treated wastewatergenerated by the municipalities as well as by different kinds of non-treated wastewater discharges.

The loading graphic permits to observe that all pharmaceutical com-pounds showed high correlation among themselves and individuallywith respect to EC. Fig. 5 shows the four best significant correlation co-efficients for diclofenac (r= 0.9085), propanolol (r = 0.8995), ibupro-fen (r = 0.8720), and atenolol (r = 0.7812) presented with respect tothe EC. Except for the caffeine and triclosan, all other linear correlationcoefficients were close to or higher than 0.70. With respect to TOC, thecorrelation coefficients were lower than 0.57. This difference of the cor-relation between each emerging contaminant and EC in comparisonwith the correlation between TOC and EC allows us to postulate thatthe persistence of the studied emerging contaminants is greater thanthe persistence of the total organic matter budget.

In recent years, it has been widely shown that caffeine is a goodchemical marker of water body contamination by anthropogenicsources of contaminants (Buerge et al., 2003; Buerge et al., 2006;Kurissery et al., 2012; Peeler et al., 2006; Young et al., 2008). In thepresent study it could be verified that other indicators, such as the com-pounds diclofenac, propanolol, ibuprofen, atenolol and carbamazepinemay be more effective than caffeine as chemical indicators of pollution,due to their restriction to anthropogenic origins. In addition, the

amples and variables studied.

Fig. 5. Correlation to EC and diclofenac, propanolol, ibuprofen and atenolol.

25D.N.R. de Sousa et al. / Science of the Total Environment 484 (2014) 19–26

correlation coefficients for univariate analysis, mainly for diclofenac,propanolol and ibuprofen, were much higher than for caffeine.

Although caffeine is one of themost studied and promisingmarkers,a great majority of studies related to caffeine are performed in lenticsystems (lakes), which makes the extension of the use of this markerto lotic systems (rivers) that present a quite singular dynamic nature.This issue was raised by Peeler et al. (2006) who observed great con-trasts in the correlation values to rivers which have their peculiarcharacteristics.

Triclosan is the most hydrophobic compound of this study what isdecisive in terms of its partition and transference to the solid phase(suspended particulate material or sediment). This aspect was reportedby Karnjanapiboonwong et al. (2010) who pointed out the strong sorp-tion tendency and, consequently, the low desorption tendency of triclo-san with respect to different soils. In addition to this, Ying and Kookana(2007) observed high triclosan removal in an Australian activatedsludge WWTP denoting a strong interaction between this drug andthe sludge. Another factor that can be related to its poor stabilityin the environment is its degradation when exposed to sunlight(Lindstrom et al., 2002; Sabaliunas et al., 2003).

In this context, othermarkers such as carbamazepine and diclofenachave been investigated as good markers alternatively or in conjunctionwith caffeine. Carbamazepine has been indicated as a promisingmarkerdue to its high refractory characteristics (Clara et al., 2004; Gasser et al.,2010; Gasser et al., 2011). Recently, Vystavna et al. (2012) pointed outthe potential use of diclofenac, caffeine and carbamazepine in a studyconducted in two different socioeconomic and geographical regions(France and Ukraine). In this report, these compounds were suggestedas potential markers of wastewaters in rivers for identification of differ-ent uncontrolled wastewater discharges.

4. Conclusions

Through the specialized literature it may be seen that a great deal ofeffort has been spent in the search for pollution indicators, and amongthe so-called emerging compounds, caffeine is often used within thistask. In thiswork, it is demonstrated that other compounds of anthropo-genic origins proved indeed to be undoubtedly better than caffeine. Theresults reveal that diclofenac, propanolol, ibuprofen, atenolol and

carbamazepine have shown correlation coefficients with electrical con-ductivity that overcomes the one of caffeine, a compound has beenwidely used as an anthropogenic indicator of sewage discharge intowater bodies over the last 10 years.We conclude, therefore, that the es-tablishment of a single marker for anthropogenic contamination bysewage is not adequate and conclusive so far. This difficulty arisesfrom the large complexity of the wastewater composition, wide globaldiversity in the lotic and lentic aquatic ecosystems and climatic condi-tions such as temperature, rainfall and sunlight incidence that havestrong influence on the biogeochemical behavior of studied compounds.Additionally, the methods developed are frequently performed by anal-ysis of a group or different classes of these compounds. In this context,the use of classes or groups ofmarkers becomes a very important aspectconsidering the geographic differences and the use pattern of thesecontaminants.

Acknowledgments

The authors are grateful to the FAPESP, Grant # 2010/01731-0, to theCNPq Grant # 142495/2011-5 for the PhD scholarship granted to thefirst author and to PETROBRAS-ANP Grant # 0050.0043180.08.4 thatfunded the purchase of the LC–MS/MS employed in the present study.

References

Akkoyunlu A, Akiner ME. Pollution evaluation in streams using water quality indices: acase study from Turkey's Sapanca Lake Basin. Ecol Indic 2012;18:501–11.

Bhuiyan MAH, Rakib MA, Dampare SB, Ganyaglo S, Suzuki S. Surface water quality assess-ment in the central part of Bangladesh using multivariate analysis. J Civ Eng 2011;15:995–1003.

Brasil. Portal da Saúde. Ministério da Saúde. www.saude.gov.br, 2013.Buerge IJ, Poiger T, Muller MD, Buser H-R. Caffeine, an anthropogenic marker for waste-

water contamination of surface waters. Environ Sci Technol 2003;37:691–700.Buerge IJ, Poiger T, Muller MD, Buser H-R. Combined sewer overflows to surface waters

detected by the anthropogenic marker caffeine. Environ Sci Technol 2006;40:4096–102.

Buerge IJ, Buser H-R, Kahle M, Muller MD, Poiger T. Ubiquitous occurrence of the artificialsweetener acesulfame in the aquatic environment: an ideal chemical marker of do-mestic wastewater in groundwater. Environ Sci Technol 2009;43:4381–5.

Chalupová D, Havlíková P, Janský B. Water quality of selected fluvial lakes in the contextof the Elbe River pollution and anthropogenic activities in the floodplain. EnvironMonit Assess 2012;184:6283–95.

26 D.N.R. de Sousa et al. / Science of the Total Environment 484 (2014) 19–26

Clara M, Strenn B, Kreuzinger N. Carbamazepine as a possible anthropogenic marker inthe aquatic environment: investigations on the behavior of carbamazepine in waste-water treatment and during groundwater infiltration. Water Res 2004;38:947–54.

Commission Decision. Of 12 August 2002 Implementing Council Directive 96/23/ECconcerning the performance of analytical methods and the interpretation of results.Official Journal of the European Communities, L221; 2002/657/ECC. p. 8–36 [Brussels,Belgium].

Daughton CG. Environmental stewardship and drugs as pollutants. Lancet 2002;360:1035–6.

Daughton CG, Ternes TA. Pharmaceuticals and personal care products in the environ-ment: agents of subtle change? Environ Health Perspect 1999;107:907–38.

Duong CN, Schlenk D, Chang NI, Kim SD. The effect of particle size on the bioavailability ofestrogenic chemicals from sediments. Chemosphere 2009;76:395–401.

Fick J, Lindberg RH, Parkkonen J, Arvidsson B, Tysklind M, Larsson DGJ. Therapeutic levelsof levonorgestrel detected on blood plasma of fish: results from screening rainbowtrout exposed to treated sewage effluents. Environ Sci Technol 2010;44:2661–6.

Gasser G, Rona M, Voloshenko A, Shelkov R, Tal N, Pankratov I, et al. Quantitative evalu-ation of tracers for quantification of wastewater contamination of potable watersources. Environ Sci Technol 2010;44:3919–25.

Gasser G, Rona M, Voloshenko A, Shelkov R, Lev O, Elhanany S, et al. Evaluation ofmicropollutant tracers. II Carbamazepine tracer for wastewater contamination froma nearby water recharge system and from non-specific sources. Desalination 2011;273:398–404.

Gelsleichter J, Szabo NJ. Uptake of human pharmaceuticals in bull sharks (Carcharhinusleucas) inhabiting a wastewater-impacted river. Sci Total Environ 2013;456–457:196–201.

Gros M, Petrovic M, Barceló D. Development of a multi-residue analytical methodologybased on liquid chromatography-tandem mass spectrometry (LC–MS/MS) forscreening and trace level determination of pharmaceuticals in surface and wastewa-ter. Talanta 2006;70:678–90.

Grujic S, Vasiljevic T, LausevicM. Determination ofmultiple pharmaceutical classes in sur-face and ground waters by liquid chromatography–ion trap–tandemmass spectrom-etry. J Chromatogr A 2009;1216:4989–5000.

Harwood JJ. Molecular markers for identifying municipal, domestic and agriculturalsources of organic matter in natural waters. Chemosphere 2014;95:3–8.

IMS Health. The global use of medicines: outlook through 2015; 2011.Kannel PR, Lee S, Lee Y-S, Kanel SR, Khan SP. Application of water quality indices and dis-

solved oxygen as indicators for river water classification and urban impact assess-ment. Environ Monit Assess 2007;132:93–110.

Karnjanapiboonwong A,Morse AN, Maul JD, Arderson TA. Sorption of estrogens, triclosan,and caffeine in a sandy loam and a silt loam soil. J Soils Sediments 2010;10:1300–7.

Kolpin DW, Furlong ET, Meyer MT, Thurman EM, Zaugg SD, Barber LB, et al. Pharmaceu-ticals, hormones, and other organic wastewater contaminants in US streams,1999–2000: a national reconnaissance. Environ Sci Technol 2002;36:1202–11.

Kurissery S, Kanavillil N, Verenitch S, Mazumder A. Caffeine as an anthropogenicmarker of domestic waste: a study form Lake Simcoe watershed. Ecol Indic2012;23:501–8.

Lacey C, McMahon G, Bones J, Barron L, Morrissey A, Tobin JM. An LC–MSmethod for thedetermination of pharmaceutical compounds in wastewater treatment plant influentand effluent samples. Talanta 2008;75:1089–97.

Lindstrom A, Buerge IJ, Poiger T, Bergqvist P-A, MullerMD, Buser H-R. Occurrence and en-vironmental behavior of the bactericide triclosan and its methyl derivative in surfacewaters and in wastewater. Environ Sci Technol 2002;36:2322–9.

Markfort CD, Hondzo M. Dissolved oxygen measurements in aquatic environments: theeffects of changing temperature and pressure on three sensor technologies. J EnvironQual 2009;38:1766–74.

Moita Neto JM, Moita GC. Uma introdução à análise exploratória de dados multivariados.Quím Nova 1998;21(4):467–9.

Paulo São. Companhia Ambiental do Estado de São Paulo, CETESB. Qualidade das águassuperficiais no estado de São Paulo 2012. São Paulo: CETESB; 2013 [370p. SérieRelatórios/CETESB, ISSN 0103–4103].

Peeler KA, Opsahl SP, Chanton JP. Tracking anthropogenic inputs using caffeine, indicatorbacteria, and nutrients in rural freshwater and urban marine systems. Environ SciTechnol 2006;40:7616–22.

Petrovic M, Solé M, López de Alda MJ, Barceló D. Endocrine disruptors in sewage treat-ment plants, receiving river waters, and sediments: integration of chemical analysisand biological effects on feral carp. Environ Toxicol Chem 2002;21(10):2146–56.

Robertson WD, Van Stempvoort DR, Solomon DK, Homewood J, Brown SJ, Spoelstra J,et al. Persistence of artificial sweeteners in a 15-year-old septic system plume. JHydrol 2013;477:43–54.

Rodil R, Quintana JB, López-Mahía P, Muniategui-Lorenzo S, Padra-Rodríguez DJ. Multi-residue analytical methods for the determination of emerging pollutants in waterby solid-phase extractions and liquid chromatography-tandem mass spectrometry.J Chromatogr A 2009;1216:2958–69.

Sabaliunas D, Webb SF, Hauk A, Jacob M, Eckhoff WS. Environmental fate of triclosan inthe River Aire Basin, UK. Water Res 2003;37:3145–54.

Scheurer M, Brauch H-J, Lange FT. Analysis and occurrence of seven artificial sweetenersin German waste water and surface water and in soil aquifer treatment (SAT). AnalBioanal Chem 2009;394:1585–94.

Schwarzenbach RP, Egli T, Hofstetter TB, von Gusten U, Wehrli B. Global water pollutionand human health. Annu Rev Environ Res 2010;35:109–36.

Seiler RL, Zaugg SD, Thomas JM, Howcroft DL. Caffeine and pharmaceuticals as indicatorsof waste water contamination in wells. Ground Water 1999;37:405–10.

Stewart AJ. A simple stream monitoring technique based on measurements of semicon-servative properties of water. Environ Manage 2001;27:37–46.

Subedi B, Du B, Chambliss CK, Koschorreck J, Rüdel H, Quack M, et al. Occurrence of phar-maceuticals and personal care products in German fish tissue: a national study. Envi-ron Sci Technol 2012;46:9047–54.

Tadeo JL, Sánchez-Brunete C, Albero B, García-Valcárcel AI, Pérez RA. Analysis of emergingorganic contaminants in environmental solid samples. Cent Eur J Chem 2012;10(3):480–520.

Tauler R. Interpretation of environmental data using chemometrics. In: Barceló D, editor.Sample handling and trace analysis of pollutants: techniques, applications and qual-ity assurance. Amsterdam: Elsevier Science BV; 2000. p. 689–736.

Tchobanoglous G, Burton FL. Wastewater engineering: treatment, disposal, and reuse. 3rded. Metcalf & Eddy, Inc.; 1991

Thompson MY, Brandes D, Kney AD. Using electronic conductivity and hardness data forrapid assessment of stream water quality. J Environ Manage 2012;104:152–7.

Vieira JS, Pires JCM, Martins FG, Vilar VJP, Boaventura RAR, Botelho CMS. Surface waterquality assessment of Lis River using multivariate statistical methods. Water AirSoil Pollut 2012;223:5549–61.

Vystavna Y, Huneau F, Grynenko V, Vergeles Y, Celle-Jeanton H, Tapie N, et al. Pharmaceu-ticals in rivers of two regions with contrasted socio-economic conditions: occurrence,accumulation, and comparison for Ukraine and France. Water Air Soil Pollut 2012;223:2111–24.

WHO. Chloride in Drinking-water: Background document for development. WHO Guide-lines for Drinking-water Quality; 2003.

Xu J, Wu L, Chen W, Chang AC. Simultaneous determination of pharmaceuticals, endo-crine disrupting compounds and hormone in soils by gas chromatography–massspectrometry. J Chromatogr A 2008;1202:189–95.

Ying G-G, Kookana RS. Triclosan in wastewater and biosolids from Australian wastewatertreatment plants. Environ Int 2007;33:199–205.

Yoon Y, Ryu J, Oh J, Choi B-G, Snyder SA. Occurrence of endocrine disrupting compounds,pharmaceuticals, and personal care products in the Han River (Seoul, South Korea).Sci Total Environ 2010;408:636–43.

Young TA, Heidler J, Matos-Pérez CR, Sapkota A, Toler T, Gibson KE, et al. Ab initio and insitu comparison of caffeine, triclosan, and triclocarban as indicators of sewage-derived microbes in surface waters. Environ Sci Technol 2008;42:3335–40.

Yu Y, Huang Q, Wang Z, Zhang K, Tang C, Cui J, et al. Occurrence and behavior of pharma-ceuticals, steroid hormones, and endocrine-disrupting personal care products inwastewater and the recipient river water of the Pearl River Delta, South China. J En-viron Monit 2011;13:871–8.