Assessing background values of chloride, sulfate and fluoride in groundwater: A geochemical-statistical approach at a regional scale

Accepted Manuscript Assessing background values of chloride, sulfate and fluoride in groundwater: A geochemical-statistical approach at a regional scale

R. Biddau, R. Cidu, M. Lorrai, M.G. Mulas PII: DOI: Reference:

S0375-6742(17)30071-7 doi: 10.1016/j.gexplo.2017.08.002 GEXPLO 5971

To appear in:

Journal of Geochemical Exploration

Received date: Revised date: Accepted date:

31 January 2017 22 June 2017 1 August 2017

Please cite this article as: R. Biddau, R. Cidu, M. Lorrai, M.G. Mulas , Assessing background values of chloride, sulfate and fluoride in groundwater: A geochemicalstatistical approach at a regional scale, Journal of Geochemical Exploration (2017), doi: 10.1016/j.gexplo.2017.08.002

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT Assessing background values of chloride, sulfate and fluoride in groundwater: a geochemicalstatistical approach at a regional scale

R. Biddaua, R. Cidua*, M. Lorraib and M.G. Mulasb a

Dipartimento Scienze Chimiche e Geologiche, Università di Cagliari, via Trentino 51, Cagliari 09127, Italy b

Regione Autonoma della Sardegna-ADIS-Servizio tutela e gestione delle risorse idriche, via Mameli 88,

PT

Cagliari 09100, Italy

RI

* Corresponding Author e-mail: [email protected]

Abstract

SC

The Sardinia island (Italy) is one of the European areas least affected by potentially anthropogenic impacts, such as spreading urbanization, intensive agriculture and regional atmospheric contamination. Such

NU

characteristics allow to consider Sardinia a good site for testing an approach that integrates geochemical tools, hierarchical cluster and geographical information system, aimed at estimating background concentrations of chloride, sulfate and fluoride at the regional scale.

MA

Analytical data were obtained from several hydrogeochemical surveys and from the groundwater-monitoring program established by the Sardinian Regional Government. Groundwater samples were grouped according to their circulation in the predominant hydrogeologic complex: Quaternary sediments, Quaternary basalts,

D

Tertiary sediments, Tertiary volcanic rocks, Mesozoic carbonatic rocks, Paleozoic carbonatic rocks, granitic

PT E

rocks and metamorphic rocks.

Samples surely affected by anthropogenic inputs, thermal waters, waters collected at wells with unknown construction details and poor quality analyses were excluded. The resulting dataset included 1414 groundwater sampling sites distributed over an area of 24,090 km2 (All data). Another dataset comprised of

CE

641 sampling sites (Selected data) was derived by All data excluding the groundwater with NO3- > 10 mg/L. Hierarchical clustering analysis was performed on both datasets considering Ca2+, Mg2+, Na+, K+, Cl-, HCO3-,

AC

SO42-, NO3- and F-. The values of total dissolved solids (TDS) were a major distinguishing factor among clusters, but distinct signatures related to the median nitrate and fluoride concentrations were also recognized. The geographic distribution of clusters reflected the role of geological and geographic characteristics on the geochemistry of groundwater. Background ranges of the regulated parameters chloride, sulfate and fluoride in each cluster, identified either using All data or Selected data, were calculated using the median±2MAD. Although results were found in general agreement, the threshold using the median+2MAD was calculated using the Selected data only, because the Selected data better represents near pristine conditions. Chloride threshold values above the drinking water limit were mainly observed in groundwater located in western Sardinia, where sediments and volcanic rocks prevalently outcrop, and also in some coastal areas. Threshold values of sulfate and fluoride above the limit were related to local conditions. Specifically, high threshold values of sulfate were observed

ACCEPTED MANUSCRIPT in groundwater interacting with the Tertiary volcanic rocks that host known sulfide mineralization and at sites where evaporitic deposits occur. Threshold values of fluoride above the limit were often observed in the areas where fluoride mineralization occurs. High fluoride values may also result from cation exchange and/or supersaturation with respect to calcite. The results of this study indicate that the integration of hierarchical clustering analysis with the geochemical characteristics of groundwater, also taking into account the geological context, allow the repartition of groundwater samples in distinct hydrogeochemical groups, which in turn allow to calculate the background

PT

ranges and reliable threshold values in groundwater. This approach can be applied to assess the background

RI

concentrations of chemical parameters at a regional scale when a large dataset is available.

Keywords: hydrogeochemistry; hierarchical clustering analysis; background range; threshold value;

SC

Geographical Information System; Sardinia

1. Introduction

NU

Chemical compositions of groundwater depend on reactions between the water and minerals and gases with which it comes in contact from recharge to discharge areas (Appelo and Postma, 2005). The chemistry of

MA

groundwater is the result of several factors, such as rainfall input and rate, chemical and biological processes in the unsaturated zone, mineralogy of aquifers and relative reactivity of minerals, residence time and mixing (Edmunds and Shand, 2008). The natural composition of groundwater, also called natural or geochemical

D

background, defines its natural quality and is characterized by large spatial variations at a range of scales

PT E

(Edmunds and Shand, 2008). In addition to natural factors, human activities may affect the quality of groundwater due to pollution from urban, agricultural, industrial and mining activities. Whereas the occurrence of specific components (e.g. pesticides) are direct indicators of human impacts, inorganic components may originate from both natural and anthropogenic sources. In this context, understanding the

CE

processes that control groundwater quality and assessing natural background ranges is crucial in terms of sustainable management of groundwater resources. From a regulatory point of view, the definition of

AC

threshold values, i.e., the upper limits of background variation, is mandatory for the regional authorities. Different approaches have been developed to determine background concentrations. Direct methods use historical data and groundwater dating, which are particularly helpful to identify concentration ranges that would have existed prior to human influence (Edmunds and Shand, 2008; Daughney et al., 2012; Morgenstern et al., 2015). Indirect methods, based on univariate and multivariate statistics, have been widely applied to separate natural from anthropogenic populations (e.g., Matschullat et al., 2000). Once the natural population has been identified, the background may be expressed as a range of values, calculated using the mean±2SD (SD = standard deviation) or the median±2MAD (MAD = Median Absolute Deviation) or identifying inflection points in cumulative probability plots. Calculations using the mean±2SD implicitly assume a normal distribution of values (Matschullat et al., 2000; Reimann and Garrett, 2005). However, geochemical data are usually heavily skewed and rarely follow a normal distribution, therefore, using the

ACCEPTED MANUSCRIPT mean±2SD is considered inappropriate (Reimann and Filzmoser, 1999; Edmunds et al., 2003). The use of the median±2MAD has been suggested either for normal or not-normal distributions (Reimann et al., 2005). The use of probability plots is grounded on the principle that different sources of components generate different water populations separated by inflection points or threshold values between populations (Edmunds et al., 2003; Panno et al., 2006; Morgenstern and Daughney, 2012; Ducci et al., 2016 and references therein). However, breakpoints in probability plots may also correspond to natural variations of geochemical facies and/or local geochemical anomalies (e.g. Preziosi et al., 2014).

PT

Methods based on multivariate statistics assume that simultaneously evaluating the distributions of several parameters will provide more discrimination to define background ranges than is possible by only looking at

RI

one parameter at a time. Cluster analysis has received increasing attention, being considered efficient tools in processing hydrogeochemical data (Templ et al., 2008). Cluster analysis comprises a range of methods for

SC

classifying multivariate observations (data) into meaningful, multivariate, homogeneous groups. In studies at a regional scale (500 to 500,000 km2; Reimann at al., 2009), the large number of data may lead to difficulties in the interpretation and representation of the results; by organizing multivariate data into groups, clustering

NU

can help revealing the characteristics of any structure or pattern present (Everitt et al., 2011). Among cluster analysis, hierarchical clustering has been applied to: i) identify distinct hydrogeochemical groups and factors

MA

affecting the chemical composition (Güler et al., 2002; Singh et al., 2004, 2005; Cloutier et al., 2008; Belkhiri et al., 2010; Yidana, 2010; Monjerezi et al., 2011; Golzar et al., 2013;), ii) provide information about hydrogeochemical processes (Yidana et al., 2010; Li et al., 2012; Jiang et al., 2015; Voutsis et al.,

D

2015) and iii) evaluate the representativeness of a monitoring network for assessing the baseline quality of

PT E

groundwater samples (Daughney et al., 2012). Several studies have demonstrated the convenience of integrating different methods to provide more reliable background and threshold estimates (Matschullat et al. 2000; Urresti-Estala et al., 2013; Preziosi et al., 2014; Hernández-Crespo and Martín, 2015; Rothwell and Cooke, 2015).

CE

In this paper, the use of direct methods was hampered by the lack of historical data and groundwater dating. The background ranges of the regulated components chloride, sulfate and fluoride in groundwater have been

AC

evaluated integrating geochemical and hierarchical cluster analysis. The chemistry of groundwater was assessed on the basis of ions Ca2+, Mg2+, Na+, K+, HCO3-, Cl-, SO42-, NO3- and F-. Hierarchical cluster analysis was used to group samples into statistically distinct hydrogeochemical groups, then background concentrations of chloride, sulfate and fluoride were estimated for each group. Mapping and spatial visualization were carried out by GIS. The study was performed in Sardinia (Italy). Because industrial activities and intensive agriculture occupy only minor parts in Sardinia, and the population is mostly concentrated in few towns, large areas are potentially unaffected by anthropogenic impacts. Such characteristics allow to consider Sardinia a good site for assessing near-pristine conditions of groundwater quality. The geological heterogeneity of the study area, i.e. aquifer lithology from siliciclastic to carbonate predominant rocks, with ages spanning from Cambrian to Quaternary, and a comprehensive dataset acquired from regional and local surveys, provided an additional value in estimating the background concentrations.

ACCEPTED MANUSCRIPT

2. Study area The Sardinia island (24,090 km2) is located in the Mediterranean Sea between 38° 51' 52'' N and 41° 15' 42'' N, and between 08° 08' 10'' E and 09° 50' 08'' E (Fig. 1). Sardinia hosts a population of about 1,672,000 inhabitants mainly concentrated in few metropolitan areas (ISTAT, 2011). The average altitude is 334 m above sea level (asl), the maximum elevation is 1,834 m asl and mountains (19 % of land) are mainly located in the eastern part of the island. Hills (68 % of land) prevail over flat areas that occur in the Campidano Plain

PT

and around the river mouths (Fig. 1a). A sketch map representing the land use (Fig. 1b) shows predominant semi-natural areas and forests (53 % of land), followed by pastures and forage systems (43 % of land).

RI

Marginal or poorly-developed rural areas occupy 81 % of the regional territory. Farming activities are mainly located in western Sardinia, where hills and flat areas prevail. Intensive dairy cattle and horticulture

SC

systems occur locally in the Campidano Plain. The mountainous areas have great relevance for grazing and forestry activities (Aru et al., 2006). Industrial activities (metallurgical and petrochemical poles) are mainly concentrated in coastal areas (Fig. 1b).

NU

The main geological complexes are comprised of: i) the Paleozoic basement, ii) the Mesozoic carbonate platform, and iii) the Cenozoic to Quaternary volcanic and sedimentary cover (Oggiano and Di Pisa, 2001).

MA

A sketch geological map of Sardinia is shown in Fig. 1c. The Paleozoic basement extends nearly continuously in the eastern part of the island and consists of rocks metamorphosed during the Hercynian Orogeny widely intruded by granitic rocks. Metamorphic rocks are mainly comprised of metasandstone,

D

metaconglomerate, quartzite and phyllite. Cambrian metalimestone and metadolostone occur in the south-

PT E

western part of the island. Mesozoic formations lay discontinuously on the Paleozoic basement and consist mainly of carbonate shelf deposits: limestone, dolostone, marly limestone and, in western Sardinia, marls with gypsum and mudstones. In Tertiary times, the island was partially affected by the Alpine Orogeny that deformed Mesozoic formations, allowed deepening of the western territory (Sardinian Rift) with effusion of

CE

calc-alkaline magmas. Tertiary sediments consist of continental conglomerate, lacustrine deposits and shallow-water carbonates. The Quaternary cover consists of sediments and basaltic rocks.

AC

Mining was a significant economic activity in Sardinia, with Pb(Ag)-Zn-Cu-Sb-F-Ba deposits exploited intensively from 1880 to 1980. In southwest Sardinia massive sulfide deposits mostly occur together with barite-fluorite in Paleozoic carbonatic rocks (De Vivo et al., 1997). In central and southeast Sardinia, stratabound deposits consist of scheelite, arsenopyrite, antimonite, and massive chalcopyrite-sphaleritegalena minerals (De Vivo et al., 1997). Epithermal high- and low-sulfidation gold deposits occur in Tertiary volcanic rocks (Rayner and Manis, 2001; Cidu et al., 2013). The environments affected by mining underwent severe deterioration, particularly due to the weathering of mining-related wastes disposed nearby the mines, and highly contaminated waters flowing out of flooded mines (Cidu et al., 2012 and references therein). Climatic conditions range from sub-arid to semi-humid, with mean annual temperature of 15 °C. Mean annual precipitation is 780 mm: below 500 mm per year on the south-western coast, up to 1250 mm per year in the mountainous areas (Chessa and Delitala, 1997). River waters collected in reservoirs supply about 70%

ACCEPTED MANUSCRIPT of water for domestic, agricultural and industrial uses. The storage capacity in reservoirs strongly depends on rainfall. Because consecutive years of drought caused severe restrictions in water supply, groundwater resources have been receiving increasingly attention. The mean annual effective infiltration has been estimated at 938 Mm3/year (R.A.S., 2006). The groundwater bodies are fed by rain water, with local contribution of marine aerosols (Cidu et al., 2008; Ghiglieri et al., 2009). Moderately high tritium values indicated that groundwater samples generally have less than 50 years recharging time (Dettori, 1978; Caboi and Noto, 1982).

PT

Hydrogeologic complexes with relatively low permeability (Paleozoic granitic and metamorphic rocks and Tertiary volcanic rocks) cover about 60% of the Sardinian territory. Water in the Paleozoic metamorphic

RI

complex usually circulates in fractured zones and underground paths are generally short. Groundwater in granitic rocks circulates in fractured zones and in the weathered, porous parts (Ciappeddu et al., 1981;

SC

Barroccu, 2007). The Cambrian and Mesozoic limestone and dolostone, the Quaternary alluvia and scoriaceous basalts (about 18 % of land) host the aquifers exploited for domestic uses. According to requirements of the European Union Water Framework Directive (EU, 2000), 38

NU

hydrogeologic complexes subdivided into 114 groundwater bodies were recognized in Sardinia (R.A.S., 2011). In this study the groundwater bodies were grouped according to a homogeneous lithology, namely:

MA

Quaternary sediments, Quaternary basalts, Tertiary sediments and volcanic rocks, Mesozoic and Paleozoic carbonatic rocks, Paleozoic granites and metamorphic rocks.

D

3. Materials and methods

PT E

3.1. Groundwater dataset

Data for this study derive from several hydrogeochemical surveys carried out at the University of Cagliari since 1986 to 2012 (partial results were already published in AAVV, 1986; Musio, 1990; Cidu et al., 1991, 1997, 2007, 2008, 2012; Da Pelo, 1993; Venturelli et al., 1999; Caboi et al., 2001; Biddau et al.; 2002; Cidu

CE

and Mulas, 2003; Lorrai, 2004; Biddau and Cidu, 2005; Pichiri, 2007; Melis, 2010; Ibba, 2011; KNOW, 2012) and from the groundwater-monitoring program established by the Sardinian Regional Government

AC

(R.A.S., 2011; 2014). Sampling density and measured chemical parameters were dependent on specific objectives established in each survey. Hydrogeochemical surveys were focused either in mine affected areas for assessing the impact of mining on the water quality, or in nearly pristine areas for investigating waterrock interaction processes. The R.A.S. groundwater-monitoring program is a long-term activity aimed to identify temporal trends in groundwater quality at the regional scale. Monitoring sites include fresh groundwater from relevant aquifers, as well as groundwater in areas of environmental relevance such as wetlands, and target areas such as industrial sites. For this study, analyses derived from the R.A.S. surveys carried out in 2008-2009 and 2011-2014 with sampling repeated under different seasonal conditions were available. The dataset inspection revealed samples surely affected by anthropogenic inputs. Specifically, groundwater sampled at industrial sites, in the nitrate vulnerable zone (NVZ Arborea, Fig. 1b; R.A.S., 2005), in areas

ACCEPTED MANUSCRIPT affected by past-mining activities and coastal groundwater affected by the intrusion of modern seawater caused by over-exploitation were discarded. Thermal waters, waters collected at wells with unknown construction details and poor quality analyses were also excluded. The resulting dataset included 1414 groundwater sampling sites geo-referenced, with locations shown in Fig. 1c. Repeated analyses were available at 351 sites, between 2 and 8 samples being collected per site (in total 3933 analyses). Springs (518 sites), with flow range from <0.1 to 100 L/s, were often used as drinking water. Groundwater in wells (896 sites) occurred at depths of 13, 45 and 90 m below ground level, respectively corresponding to the 25th, 50th

PT

and 75th percentiles.

RI

3.2. Sampling and laboratory analyses

The sampling protocol adopted at the University of Cagliari was unchanged over time and similar to that

SC

adopted in the R.A.S. groundwater-monitoring program. Wells were purged 30 min prior to sample collection. Flow, temperature, pH, redox potential (Eh), dissolved oxygen (DO), electrical conductivity (EC) and alkalinity (reported as HCO3-) were measured on site. Alkalinity was determined by acidimetric titration

NU

using methyl orange as indicator. The water samples were filtered through 0.45 μm pore-size filters immediately upon collection into pre-cleaned high-density polyethylene bottles and taken refrigerated until

MA

analysis. An aliquot was acidified with supra pure HNO3 (1 %, v/v). At the University of Cagliari, Ca2+, Mg2+, Na+, K+, Cl-, SO42-, NO3-, PO42-, NO2-, NH4+ and F- were analyzed in the filtered aliquot by ion chromatography (IC, Dionex ICS3000). The cations and total sulfur were also

D

analyzed in the filtered and acidified aliquot by inductively coupled plasma optical emission spectrometry

PT E

(ICP-OES, ARL3520). Chemical analyses of the R.A.S. groundwater-monitoring program were carried out at certified laboratories (R.A.S., 2011). Dissolved oxygen, reduced nitrogen species and phosphate were not determined in specific surveys. The quality of chemical analyses was carefully monitored by analyzing blanks solutions, duplicate samples and standard reference solutions. The calculated charge balance errors

CE

were ≤ 5 % in 86 % of samples and between 6 % and 10 % in 14 % of samples.

AC

3.3. Geochemical and statistical analysis Time series values at each monitoring site did not reveal specific trends. Therefore, prior to data processing, they were converted to median values to ensure an equal contribution at each sampling site. Summary statistics included the calculation of minimum, maximum, median, percentiles and MAD values. For assessing the main geochemical features, groundwater samples were grouped according to their circulation in the predominant hydrogeologic complex: Quaternary sediments, Quaternary basalts, Tertiary sediments, Tertiary volcanic rocks, Mesozoic carbonatic rocks, Paleozoic carbonatic rocks, granitic rocks and metamorphic rocks. The Piper diagram (Piper, 1953) was used to identify the predominant hydrogeochemical facies in each hydrogeologic complex. The Stiff diagram (Stiff, 1951) was used to show the median hydrogeochemical facies in each cluster. The computer program PHREEQC was used for calculations of the activity and saturation index (SI) values (Parkhurst and Appelo, 1999). The saturation

ACCEPTED MANUSCRIPT index with respect to a mineral phase is equal to log IAP − log K, where IAP is the ionic activity product and K is the equilibrium constant. In hierarchical clustering analysis, two datasets were considered: i) the 1414 groundwater samples (hereafter All data), thus leaving to the clustering algorithm to identify water chemistries that may not be related to natural processes, such as groundwater affected by relative high nitrate (Daughney and Reeves, 2005) and ii) 641 groundwater samples (hereafter Selected data) with NO3- concentrations ≤ 10 mg/L (Wendland et al., 2008; Cruz and Andrade, 2015; Rotiroti et al., 2015). Concentrations of Ca2+, Mg2+, Na+, K+, Cl-, HCO3-,

PT

SO42-, NO3-, and F- in both datasets were considered. The non detected values were substituted using the Robust Regression on Order Statistics method (Lee and Helsel, 2005a, b), which imputes a value for each

RI

non detect using a distributional model fitted to the observed data. Calculations were performed using logtransformed (ln) data. Concentrations have been standardized by subtracting the mean of each parameter and

SC

dividing by the standard deviation, to ensure that each parameter was equally weighted. The Euclidean distance and the Ward's method were used for assessing the similarity and linkage rule, respectively (Güler et al., 2002). In fact, the classification scheme using Euclidean distance together with Ward's method for

NU

linkage, produces the most distinctive cluster where each sample within the group is more similar to its fellow members than to any member outside the cluster (Güler et al., 2002). The classification of

MA

groundwater samples into clusters was based on the linkage distance observed in the dendrogram. The median values of TDS, Ca2+, Mg2+, Na+, K+, Cl-, HCO3-, SO42-, NO3- and F- in groundwater samples belonging to each cluster were calculated in order to identify any relation between the cluster and

D

geochemical features and to determine a median groundwater type according to the classification approach

PT E

reported in Cloutier et al. (2008). The non-parametric Kruskal-Wallis test (Helsel and Hirsch, 2002; Cerar and Mali, 2016) was performed for Ca2+, Mg2+, Na+, K+, Cl-, HCO3-, SO42-, NO3-, and F-, considering the pvalue at the significance level of 0.05. The results of the test revealed significant differences in the median values of each parameters, with p-value always below 0.00001. Background ranges of Cl-, SO42- and F- were

CE

calculated for each cluster using the median±2MAD, the upper limit of the background range (median+2MAD) was assumed as threshold value (Matschullat et al., 2000; Reimann et al., 2005). Statistical

AC

analysis were carried out using the R language of statistical computing (R Development Core Team, 2013). Geochemical maps were drown by ArcGIS 10.2 (ESRI, 2013).

4. Results 4.1. Groundwater geochemistry and descriptive statistics Main geochemical features of groundwater samples in All data are summarized in Table 1 reporting the minimum, maximum, MAD, selected percentiles and median (50th percentile) values for chemical-physical parameters, major ions, nitrogen species and phosphate. The pH and Eh values ranged from 4.2 to 9.3 and from -0.14 to 0.86 V, with median values of 7.2 and 0.37 V, respectively. Values of pH below 6 (< 5 % of samples) were measured in groundwater from Tertiary volcanic rocks that host sulfide-mineralized bodies;

ACCEPTED MANUSCRIPT the low amount of carbonate minerals in these rocks does not allow buffering the acidity produced by the oxidative dissolution of sulfide minerals (Biddau and Cidu, 2005; Da Pelo et al., 2009). Groundwater samples exhibit considerable variation in total dissolved solids (TDS range: 0.04 - 8.6 g/L, Table 1). Fig. 2 shows the distribution of EC values in groundwater from the different hydrogeologic complexes. Notwithstanding the large dispersion, relatively low median EC values (0.35 to 0.57 mS/cm) were measured in groundwater interacting with basaltic, granitic and metamorphic rocks, i.e. when waterrock interaction processes involved silicate minerals with low solubility. Relatively high EC values were

PT

generally observed in western Sardinia, where groundwater mainly interacted with Quaternary sediments, Tertiary sediments and volcanic rocks, and in coastal areas. Relatively high EC values in groundwater of

RI

Sardinia may have been derived from natural sources, such as the inland transport of marine aerosol (Lorrai et al., 2004), the occurrence of relict seawater trapped in Quaternary sediments (Ardau and Barbieri, 2000)

SC

and leaching of thin layers of evaporitic minerals hosted in Tertiary sediments (Barbieri et al., 1991). The composition of groundwater is shown in the Piper diagrams of Fig. 3; considerable overlaps among samples were observed. Many groundwater samples belong to the hydrogeochemical facies Na-Cl and Ca-

NU

(Mg)-HCO3, regardless the rock type and mineralogy of aquifers. Some groundwater samples from Tertiary sediments, Tertiary volcanic rocks and metamorphic rocks showed a relative enrichment in sulfate (Fig. 3).

MA

Sulfate and chloride concentrations in most groundwater showed an alignment along the sulfate:chloride ratio observed in seawater (Fig. 4), indicating that marine salts provide a significant contribution of these anions to the groundwater. The sulfate enrichments observed in Figs. 3 and 4 suggest additional sulfate other

D

than marine sources, likely derived from the oxidation of sulfide minerals.

PT E

Fluoride concentrations were in the range of <0.01 to 13 mg/L, with median of 0.20 mg/L; the 95% of samples showed concentration ≤ 1.0 mg/L. Fluoride concentrations higher than 1.0 mg/L were not dependent on potential contaminants from fertilizer application, such as nitrate, potassium and phosphate fertilizers. Nitrate concentrations, commonly used for assessing widespread groundwater contamination, had a median

CE

value of 13 mg/L. Groundwater samples with relatively high nitrate were observed in springs with flow below 0.1 L/s and in shallow wells. Nitrate above 10 mg/L showed a regional random distribution, whereas

AC

concentrations above 50 mg/L frequently occurred in the more cultivated areas of western Sardinia (Fig. 5). More than 80 % of samples showed NO2- and NH4+ below detection limits (0.02, and 0.01 mg/L, respectively) and PO4-3 in 66 % of samples was below 0.10 mg/L. 4.2 Hierarchical clustering analysis The dendrogram resulting from hierarchical clustering analysis carried out on All data is shown in Fig. 6a: the main water groups were represented by 9 clusters. The number of groundwater samples in each hydrogeologic complex per cluster are reported in Table 2. In order to identify any relation between clusters and geochemical features, median values of TDS and dissolved components per cluster were calculated (Table 3) with corresponding chemistries shown in the Stiff diagrams (Fig. 7a).

ACCEPTED MANUSCRIPT Groundwater samples in clusters G1 and G2 (9.8 % and 15 % of total samples) were characterized by predominant Na-Cl composition (Fig. 7a); cluster G1 showed the lowest median TDS of 0.18 g/L (Table 3) with groundwater samples mainly belonging to granitic and metamorphic rocks and Quaternary basalts, whereas about 50% of groundwater in cluster G2 were hosted in Quaternary sediments (Table 2). Groundwater samples in clusters G3, G4 and G6 (8.5, 15 and 12 % of total samples, respectively) had predominant Ca-HCO3 composition (Fig. 7a); they were distinguished by different TDS and nitrate with lower and higher values in clusters G3 and G6, respectively (Table 3); most of them were hosted in

PT

carbonatic rocks (Table 2). Groundwater samples in clusters G5, G7, G8 and G9 (21, 6.6, 2 and 11 % of total samples, respectively) had

RI

relatively high TDS and predominant Na-Cl composition; they were distinguished by different TDS and nitrate, the highest values being observed in cluster G9 (Table 3); most of them were hosted in Quaternary

SC

sediments, Tertiary sediments, Tertiary volcanic rocks and Mesozoic carbonatic rocks (Table 2). The small cluster G8 was characterized by low concentrations of Ca and Mg and the highest median fluoride concentration (Table 3). It is worth noting that the high nitrate waters did not group in one specific cluster. In

NU

fact, median nitrate above 20 mg/L were observed in clusters G5, G6 and G9, and values above 10 mg/L occurred in clusters G2 and G4, likely reflecting the widespread distribution of nitrate in Sardinian

MA

groundwater (Fig. 5).

The dendrogram resulting from hierarchical clustering analysis carried out on the Selected data is shown in Fig. 6b: 8 clusters represented the main water groups, with corresponding chemistries shown in Fig. 7b.

D

Median values of TDS and dissolved components and the number of groundwater samples in each cluster are

PT E

reported in Table 3 together with median values calculated on similar clusters identified using All data. Notwithstanding the number of samples in the Selected data much smaller (641) than the All data (1414), the Stiff diagrams and median values were in good agreement for both datasets, indicating that removing samples with high nitrate do not influence significantly the features of clusters identified using All data. One

CE

cluster in Selected data (G4Sel) was an exception showing intermediate composition between clusters G2

AC

and G4 identified using All data (Table3).

4.3 Background and threshold values for chloride, sulfate and fluoride The ranges of background values, calculated considering the median±2MAD (back-transformed from the corresponding ln values reported) for chloride, sulfate and fluoride in each cluster identified using All data and Selected data, are reported in Table 4. Background ranges derived from both datasets were generally in good agreement considering the different dispersion of values in the two datasets (Table 4). Nevertheless, threshold values derived from the Selected data will be only discussed, because they reflect less contaminated environments. Threshold values were compared with the limits established for drinking water: 250 mg/L for both chloride and sulfate and 1.5 mg/L for fluoride (Table 5). Their geographical representation is shown in Fig. 8. Chloride threshold values above the limit (51% of Selected data) were mainly observed in groundwater

ACCEPTED MANUSCRIPT located in western Sardinia, where sediments and volcanic rocks prevalently outcrop, and also in some coastal areas (Fig. 8a). Threshold values of sulfate above the limit (24% of Selected data) were observed in: i) groundwater interacting with Tertiary volcanic rocks that host known sulfide mineralization (Biddau and Cidu, 2005; Da Pelo et al., 2009); ii) groundwater interacting with Mesozoic carbonatic rocks in northwestern Sardinia where evaporitic deposits occur (Ghiglieri et al., 2009), and iii) groundwater located in some coastal areas (Fig. 8b). Threshold values of fluoride above the limit (29% of Selected data were

PT

randomly observed (Fig. 8c).

5. Discussion and Conclusions

RI

Results of this study suggest that background ranges can be calculated on a large dataset, without preselection. However, median nitrate in All data were relatively high (10 to 36 mg/L, Table 3), especially in

SC

groundwater samples located in areas with diffused agricultural activities (Fig.s 1b and 5). Median nitrate ≤ 2 mg/L was observed in eastern Sardinia where the land use shows predominant forests and pastures in granitic and metamorphic environments (Fig. 1b; R.A.S., 2013b). This result was in agreement with the nitrate

NU

background previously estimated in groundwater samples from metamorphic rocks in Sardinia (1.9 mg/L; Biddau et al., 2017). Therefore, more reliable threshold values, better representing near-pristine conditions,

MA

should be estimated following a pre-selection aimed at excluding those groundwater samples with nitrate above 10 mg/L.

In clusters identified on the Selected data, TDS values were a major distinguishing factor, reflecting the

D

concentrations of dissolved components. Relatively more saline, Na-Cl groundwater samples interacting with

PT E

sediments and Tertiary volcanic rocks were mostly located in western Sardinia and in coastal areas. The less saline, Na-Cl groundwater samples mainly interacting with granitic, metamorphic and basaltic rocks were mostly located in eastern Sardinia. Groundwater samples mainly interacting with carbonatic rocks were distinguished by a predominant Ca-HCO3 composition. Distinct signatures related to the median values of

CE

nitrate and fluoride were recognized in some clusters (Table 3). Relatively high concentrations of fluoride were observed in clusters G4Sel, G8Sel and G9Sel (Tables 3 and

AC

5). Groundwater samples in cluster G4Sel mainly occurred in areas hosting known fluoride mineralization (De Vivo et al., 1998). In cluster G8Sel high fluoride is associated with low calcium, which is consistent with thermodynamic control of dissolved fluoride with respect to the fluorite mineral (Stumm and Morgan, 1996). The low calcium in G8Sel groundwater could result from cation exchange between Ca2+ and Na+, consistently with high sodium being observed in this cluster (Fig. 6b). In cluster G9Sel high fluoride is associated with high TDS (Table 3). To investigate such trend, the activity values of fluoride were plotted versus the saturation indexes with respect to calcite. Groundwater samples in clusters G4Sel, G8Sel and G9Sel, showing threshold values above the fluoride limit (Table 5), were selected with results shown in Fig. 9. The more saline groundwater samples in cluster G9Sel were saturated with respect to calcite, and reached super saturation in some cases. The eventual precipitation of calcite would cause the removal of calcium from solution, in turn allowing the increase of fluoride. Thus, relatively high fluoride concentrations in saline

ACCEPTED MANUSCRIPT groundwater could be controlled by the solubility of calcite (Appelo and Postma, 2005). The above considerations highlight the contribution of hierarchical cluster analysis in understanding geochemical processes (Yidana et al., 2010; Li et al., 2012). The integration of results derived by hierarchical cluster analysis with those derived by geochemical analyses allowed to estimate the background ranges and threshold values of chloride, sulfate and fluoride in groundwater samples at a regional scale. This integrated approach provided the classification of groundwater samples in hydrogeochemical groups more efficiently distinguished than either using statistical or

PT

geochemical methods alone, which result in frequent overlapping among groundwater samples hosted in different hydrogeologic complexes. The spatial distribution of threshold values revealed that different values

RI

may occur within one hydrogeologic complex, indicating that more detailed hydrogeochemical surveys are

SC

required for assigning the threshold value moving from the regional scale to the local scale.

Acknowledgements

Research funded by the Regione Autonoma della Sardegna (Cooperation Agreement between DSCG-

NU

UNICA and RAS-ADIS-STGRI, December 2013). The Editor and two anonymous reviewers are

MA

acknowledged for their useful suggestions.

References

AAVV., 1986. Progetto Sardegna: studi geologico-strutturali ed idrogeologici; prospezione idrogeochimica

D

per la valutazione del potenziale geotermico. Commissione delle Comunità Europee, Rapporto Finale -

PT E

Contratto EG-A2-O52-I, pp. 167, Bruxelles.

Appelo, C.A.J., Postma, D., 2005. Geochemistry, Groundwater and Pollution, second edition. Balkema, pp. 536, Rotterdam.

Ardau F., Barbieri G., 2000. Aquifer configuration and possible causes of salination in the Muravera plain

CE

(SE Sardinia), Italy. In Proceedings of 16th Salt Water Intrusion Meeting, pp. 11-18, Wolin Island, Poland. Aru, A., Tomasi, D., Vacca, A., 2006. Aspects of environmental degradation in Sardinia. In: Watershed

AC

Management: Water Resources for the Future (L. Tennyson and P.C. Zingari, Eds) 16, 163-178. Barbieri, G., Barroccu, G., Muscas, L., Pedditzi, E., Raes, H., Uras, G., Vernier A., 1991. La salinizzazione degli acquiferi del Campidano sud-orientale. In: Convegno sulla ricerca e protezione delle risorse idriche sotterranee in aree montuose. Vol. II, pp. 311-323, Brescia, Italy. Barroccu, G., 2007. Hydrogeology of granite rocks in Sardinia. Groundwater in Fractured Rocks: IAH Selected Paper Series (J. Krásný and J.M. Sharp, Eds) Volume 9, pp. 33-44, CRC Press. Belkhiri, L., Boudoukha, A., Mouni, L., Baouz, T., 2010. Multivariate statistical characterization of groundwater quality in Ain Azel plain, Algeria. African Journal of Environmental Science and Technology, 4, 526-534. Biddau, R., Cidu, R. 2005. Hydrogeochemical baseline studies prior to gold mining: a case study in Sardinia (Italy). Journal of Geochemical Exploration, 86, 61-85.

ACCEPTED MANUSCRIPT Biddau, R., Cidu, R., Frau, F., 2002. Rare earth elements in waters from the albitite-bearing granodiorites of central Sardinia, Italy. Chemical geology, 182, 1-14. Biddau, R., Cidu, R., Ghiglieri, G., Da Pelo, S., Carletti, A., Pittalis, D., 2017. Nitrate occurrence in groundwater hosted in hardrock aquifers: estimating background values at a regional scale. Italian Journal of Geosciences, 136, 1, 109-120, doi: 10.3301/IJG.2016.03. Caboi, R., 2001. Hydrogeochemistry in the Flumendosa river basin (Sardinia, Italy). Proceeding of the 10 th International Symposium on Water Rock Interaction - WRI-10, Villasimius, Italy, June 10-15, 2001 (R.

PT

Cidu, Ed). Vol. 1, pp. 465-468, A.A. Balkema, Rotterdam. Caboi, R., Noto, P., 1982. Dati isotopici sulle acque termali e fredde dell'area campidanese. Estratto da

RI

Ricerche geotermiche in Sardegna con particolare riferimento al Graben del Campidano. CNR-PFE-SPEGRF-10, pp. 124-132, Pisa, Italy.

SC

Cerar, S., Mali, N., 2016. Assessment of presence, origin and seasonal variations of persistent organic pollutants in groundwater by means of passive sampling and multivariate statistical analysis. Journal of Geochemical Exploration. 170, 78-93.

NU

Chessa, P.A., Delitala, A., 1997. Il Clima della Sardegna. Servizio Agrometeorologico Regionale (SAR) per la Sardegna, nota tecnica n°2 http://www.sar.sardegna.it/pubblicazioni/notetecniche/nota2/index.asp. Last

MA

accessed in June 2016.

Ciappeddu, L., Dettori, B., Filigheddu, M.R., Marchi, M., 1981 Contributo alla valutazione dei caratteri idrologici e idrogeochimici delle acque sorgentizie della Sardegna. Atti Istituto Mineralogia e Geologia, 1,

D

89-150.

PT E

Cidu, R., Biddau, R., Fanfani, L., 2007. Impact of past mining activity on the quality of groundwater in SW Sardinia (Italy). Journal of Geochemical Exploration, 100, 125-132. Cidu, R., Caboi, R., Biddau, R., Petrini, R., Slejko, F., Flora, O., Stenni, B., Aiuppa, A., Parello, F., Valenza, M., 2008. Caratterizzazione idrogeochimica ed isotopica e valutazione della qualità delle acque superficiali

CE

e sotterranee campionate nel foglio 549 Muravera. In: GEOBASI - Il foglio IGMI N° 549 Muravera - (G. Ottonello, Ed), pp. 149-183. Pacini Editore, Pisa, Italy.

AC

Cidu, R., Caboi, R., Fanfani, L., Frau, F., 1997. Acid drainage from sulphides hosting gold mineralization (Furtei, Sardinia). Environmental Geology, 30, 321-237. Cidu, R., Dadea, C., Desogus, P., Fanfani, L., Manca, P.P., Orrù, G. 2012. Assessment of environmental hazards at abandoned mining sites: a case study in Sardinia, Italy. Applied Geochemistry, 27, 1795-1806. Cidu, R., Da Pelo, S., Frau, F., 2013. Legacy of Cyanide and ARD at a Low-Scale Gold Mine (Furtei, Italy). Mine Water and the Environment, 32, 74-83. Cidu, R., Mulas A.D., 2003. Geochemical features of thermal waters at Benetutti (Sardinia). Rendiconti Seminario Facoltà Scienze Università di Cagliari, 73, 1, 39-53. Cidu, R., Pala, A., Serra, S., 1991. Valutazione delle riserve idriche del bacino del Rio Sos Molinos di Gavoi (Sardegna centrale). Geologia Tecnica, 2, 14-29.

ACCEPTED MANUSCRIPT Cloutier, V., Lefebvre, R., Therrien, R., Savard, M.M., 2008. Multivariate statistical analysis of geochemical data as indicative of the hydrogeochemical evolution of groundwater in a sedimentary rock aquifer system. Journal of Hydrology, 353, 294-313. Cruz, J.V., Andrade, C., 2015. Natural background groundwater composition in the Azores archipelago (Portugal): A hydrogeochemical study and threshold value determination. Science of the Total Environment, 520, 127-135. Da Pelo S., 1993. Studi di interazione acqua-roccia nel Montiferro sud-occidentale (Seneghe-Narbolia-

PT

Cuglieri). Tesi di Laurea Inedita, Dipartimento di Scienze della Terra, Università degli Studi di Cagliari, Italy.

RI

Da Pelo, S., Musu, E., Cidu, R., Frau, F., Lattanzi, P., 2009. Release of toxic elements from rocks and mine wastes at the Furtei gold mine (Sardinia, Italy). Journal of Geochemical Exploration, 100, 142-152.

SC

Daughney, C. J., Moreau-Fournier, M., Morgenstern U., Raaij, R., 2012. Use of hierarchical cluster analysis to assess the representativeness of a baseline groundwater quality monitoring network: comparison of New Zealand’s national and regional groundwater monitoring programs. Hydrogeology Journal, 20, 185-200.

NU

Daughney, C.J., Reeves, R.R. 2005. Definition of hydrochemical facies in the New Zealand National Groundwater Monitoring Programme. Journal of Hydrology, New Zealand, 44(2): 105-130.

MA

De Vivo, B., Boni, M., Costabile, S., 1998. Baseline Formational anomalies versus mining pollution: geochemical risk maps of Sardinia, Italy. Journal of Geochemical Exploration, 64, 321-337. De Vivo, B., Boni, M., Marcello, A., Di Bonito, M., Russo, A., 1997. Baseline geochemical mapping of

D

Sardinia, Italy. Journal of Geochemical Exploration, 60, 77-90.

PT E

Dettori, B., 1978. Nuove osservazioni sul chimismo e sul termalismo delle acque minerali della Sardegna settentrionale. Annali della Facoltà di Agraria dell'Università di Sassari, 26, 375-388. Ducci, D., Condesso de Melo, M.T., Preziosi, E., Sellerino, M., Parrone, D., Ribeiro, L. 2016. Combining natural background levels (NBLs) assessment with indicator kriging analysis to improve groundwater

CE

quality data interpretation and management. Science of Total Environment, 569–570, 569–584. Edmunds, W.M., Shand, P., 2008. Natural Groundwater Quality (W.M. Edmunds and P. Shand, Eds).

AC

Blackwell Publishing Ltd. ISBN: 978-14051-5675-2. Edmunds, W.M., Shand, P., Hart, P., Ward, R.S., 2003. The natural (baseline) quality of groundwater: a UK pilot study. Science of Total Environment, 310, 25-35. ESRI, 2013. Environmental Systems Research Institute, Esri, 380 New York Street, Redlands, California 92373-8100, USA. EU, 1991. Directive 91/676/EC of the European Parliament and the Council of 12 December 1991. Concerning the protection of waters against pollution caused by nitrates from agricultural sources. Strasbourg. EU, 2000. Water framework directive, 2000, Directive 2000/60/EC of the European parliament and the council of 23 October 2000. Establishing a framework for community action in the field of water policy.

ACCEPTED MANUSCRIPT Everitt, B.S., Landau, S., Leese, M., Stahl, D., 2011. Cluster analysis. 5th edition, pp. 330, John Wiley & Sons, Ltd. West Sussex, PO19 8SQ, United Kingdom. Ghiglieri, G., Oggiano, G., Fidelibus, D., Barbieri, G., Vernier, A., Tamiru, A., 2009. Hydrogeology of the Nurra

Region,

Sardinia

(Italy):

basement-cover

influences

on

groundwater

occurrence

and

hydrogeochemistry. Hydrogeology Journal, 17, 447-466. Golzar, H., Selim Reza, A.H.M., Lutfun-Nessa, M., Ahmed, S.S., 2013. Factor and Cluster Analysis of Water Quality Data of the Groundwater Wells of Kushtia, Bangladesh: Implication for Arsenic Enrichment

PT

and Mobilization. Journal Geological Society of India, 81, 377-384. Güler, C., Thyne, G., McCray, J.E., Turner, K.A., 2002. Evaluation and graphical and multivariate statistical

RI

methods for classification of water chemistry data. Hydrogeology Journal, 10, 455-474. Helsel, D.R., Hirsch, R.M., 2002. Statistical methods in water resources. In: USGS (Ed.), Techniques of

SC

Water-resources Investigations of the United States Geological Survey. Book 4: Hydrologic Analysis and Interpretation, Chapter 3. Publication available at: http://water.usgs.gov/pubs/twri/twri4a3/. Last access: May 2016.

NU

Hernández-Crespo, C., Martín, M., 2015 Determination of background levels and pollution assessment for seven metals (Cd, Cu, Ni, Pb, Zn, Fe, Mn) in sediments of a Mediterranean coastal lagoon. Catena, 133,

MA

206-214.

Ibba, M.F., 2011. Studio per la valutazione del potenziale geotermico dell’Anglona. Tesi di Laurea inedita, Dipartimento di Scienze della Terra, Università degli Studi di Cagliari, Italy. (Istituto

Nazionale

di

Statistica)

2011.

Censimento

della

popolazione

2011.

D

ISTAT

PT E

http://www.istat.it/it/sardegna. Last accessed in June 2016. Jiang, Y., Guo, H., Jia, Y., Cao, Y., Hu, C., 2015. Principal component analysis and hierarchical cluster analyses ofarsenic groundwater geochemistry in the Hetao basin, Inner Mongolia. Chemie der Erde, 75, 197-205.

CE

KNOW, 2012. Sistemi innovativi ed integrati per la definizione dei processi di degrado delle risorse idriche sotterranee, in particolare da inquinamento da nitrati in bacini pilota della Sardegna (KNOW: implementing

AC

the Knowledge of NitrOgen in groundWater, R. Cidu Coordinator). Project funded by LR N. 7, 7 August 2007, "Promozione della Ricerca Scientifica e dell'innovazione Tecnologica in Sardegna", Cagliari, Italy. Lee, L., Helsel, D., 2005a. Statistical analysis of water-quality data containing multiple detection limits: Slanguage software for regression on order statistics. Computers & Geosciences 31, 1241-1248. Lee, L., Helsel D., 2005b. Baseline models of trace elements in major aquifers of the United States. Applied Geochemistry, 20, 1560-1570. Li, J., Wang, Y., Xie, X., Su, C., 2012 Hierarchical cluster analysis of arsenic and fluoride enrichments in groundwater from the Datong basin, Northern China. Journal of Geochemical Exploration, 118, 77-89. Lorrai, M., 2004. Geochimica dell'uranio nelle acque sotterranee della zona di Sarrala (Sardegna). Tesi di Dottorato inedita, Dipartimento di Scienze della Terra, Università degli Studi di Cagliari, Italy.

ACCEPTED MANUSCRIPT Lorrai, M., Fanfani, L., Lattanzi, P., 2004. Processes controlling groundwater chemistry of a coastal area in SE Sardinia (Italy). Proceeding of 11th International Symposium on Water Rock Interaction (R.B. Wanty and R.R. Seal II, Eds) , 439-443. A.A. Balkema Rotterdam. Matschullat, J., Ottenstein, R., Reimann, C., 2000. Geochemical background - can we calculate it? Environmental Geology, 39, 990-1000. McLachlan, G.J., Peel, D., 2000. Finite mixture models. Applied Probability and Statistics Section, Wiley Series in Probability and Statistics, pp. 419, Wiley, New York.

PT

Melis, F., 2010. Valutazione della qualità di acque di sorgente in Gallura. Tesi di Laurea inedita, Dipartimento di Scienze della Terra, Università degli Studi di Cagliari, Italy.

RI

Monjerezi, M., Vogt, R.D., Aagaard, P., Saka, J.D.K., 2011. Hydro-geochemical processes in an area with saline groundwater in lower Shire River valley, Malawi: an integrated application of hierarchical cluster

SC

and principal component analysis. Applied Geochemistry, 26, 1399-1413.

Morgenstern, U.; Daughney, C.J.; Leonard, G.S.; Gordon, D.; Donath, F.M.; Reeves, R.R. 2015. Using groundwater age and hydrochemistry to understand sources and dynamics of nutrient contamination

NU

through the catchment into Lake Rotorua, New Zealand. Hydrology and Earth System Sciences, 19, 803822.

MA

Musio, A., 1990. Studio delle acque sorgive del territorio di Orune. Tesi di Laurea inedita, Dipartimento di Scienze della Terra, Università degli Studi di Cagliari, Italy. Oggiano, G., Di Pisa, A., 2001. Introduction to the Sardinia geologic evolution. Rendiconti Seminario

D

Facoltà Scienze Università di Cagliari, 71, 2, 5-11.

PT E

Panno, S.V., Kelly, W.R., Martinsek, A.T., Hackley, K.C., 2006. Estimating background and threshold nitrate concentrations using probability graphs. Ground Water, 44, 697–709. Parkhurst, D.L., Appelo, C.A.J., 1999. User’s Guide to Phreeqc (version 2) – A Computer Program for Speciation, Batch-reaction, One Dimensional Transport and Inverse Geochemical Calculation. USGS

CE

Water-Resources Investigation Report 99-4259, U.S. Geological Survey, Denver, USA. Pichiri, F., 2007. Valutazione della qualità delle acque nell'area di Nurri-Orroli. Tesi di Laurea inedita,

AC

Dipartimento di Scienze della Terra, Università degli Studi di Cagliari, Italy. Piper, A.M. (1953) A graphic procedure in geochemical interpretation of water analysis. USGS Groundwater note, 12, 63, USA.

Preziosi, E., Parrone, D., Del Bon, A., Ghergo, S., 2014. Natural background level assessment in groundwaters: probability plot versus pre-selection method. Journal of Geochemical Exploration, 143, 4353. R Development Core Team, 2013. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. http://www.R-project.org/ R.A.S. 2005. Deliberazione della Giunta Regionale n. 1/12 del 18/01/2005 con cui la Regione Sardegna ha designato, quale zona vulnerabile da nitrati di origine agricola (ZVN), una porzione del territorio del Comune di Arborea. Regione Autonoma della Sardegna, Cagliari (in Italian).

ACCEPTED MANUSCRIPT R.A.S. 2006. Piano di Tutela Acque (Regional plan for water protection) approvato con Deliberazione della Giunta Regionale n. 14/16 del 4 aprile 2006. Regione Autonoma della Sardegna, Cagliari (in Italian). http://www.regione.sardegna.it/j/v/25?s=26251&v=2&c=1260&t=1. Last accessed in June 2016. R.A.S. 2011. Caratterizzazione, obiettivi e monitoraggio dei corpi idrici sotterranei della Sardegna approvato con Deliberazione della Giunta Regionale n. 1/16 del 14/01/2011, allegato al Piano di gestione del Distretto Idrografico.

Regione

Autonoma

della

Sardegna,

Cagliari

(in

Italian).

https://www.regione.sardegna.it/documenti/1_328_20130906131252.zip. Last accessed in June 2016.

PT

R.A.S. 2013a. Sardinia Digital Elevation Model. Regione Autonoma della Sardegna, Cagliari. http://www.sardegnageoportale.it/webgis/sardegnamappe/mappa.html?mapname=DTM. Last accessed in

RI

June 2016.

R.A.S. 2013b. Carta dell'Uso del Suolo in scala 1:25.000. Regione Autonoma della Sardegna, Cagliari.

SC

R.A.S. 2013c. Carta geologica di base della Sardegna in scala 1:25000. Regione Autonoma della Sardegna, Cagliari

http://www.sardegnageoportale.it/argomenti/cartageologica.html. Last accessed in June 2016.

NU

R.A.S. 2014. Progetto di Aggiornamento del Piano di Gestione del Distretto Idrografico - Caratterizzazione, obiettivi e monitoraggio dei corpi idrici sotterranei. Allegato Delibera Comitato Istituzionale n.4 del

MA

18.12.2014. Regione Autonoma della Sardegna, Cagliari

http://www.regione.sardegna.it/documenti/1_470_20141222141305.pdf. Rayner, J., Manis, D., 2001. Gold in Sardinia: recent developments in exploration and exploitation. In:

D

Proceedings of the 10th International Symposium on Water-Rock Interaction. (R. Cidu, Ed), 741-744. A.A.

PT E

Balkema Rotterdam.

Reimann, C., Filzmoser, P., 1999. Normal and lognormal data distribution in geochemistry: death of a myth. Consequences for the statistical treatment of geochemical and environmental data. Environmental Geology, 39, 1001-1014.

CE

Reimann, C., Filzmoser, P., Garrett, R.G., 2005. Background and threshold: critical comparison of methods of determination. Science of the Total Environment, 346, 1,16.

AC

Reimann, C. Garrett, R.G., 2005. Geochemical background - concept and reality. Science of the Total Environment, 350, 12- 27. Reimann, C., Filzmoser, P., Garrett, R.G., Dutter, R., 2008. Statistical Data Analysis Explained: Applied Environmental Statistics (R. Wiley Ed), pp. 362. ISBN: 978-0-470-98581-6. Reimann, C., Matschullat, J., Birke, M., Salminen, R., 2009. Arsenic distribution in the environment: the effect of scale. Applied geochemistry, 24, 1147-1167. Rothwell, K.A., Cooke, M.P., 2015. Comparison of methods used to calculate normal background concentrations of potentially toxic elements for urban soil. Science of the Total Environment, 532, 625634.

ACCEPTED MANUSCRIPT Rotiroti, M., Di Mauro, B., Fumagalli, L., Bonomi, T., 2015. COMPSEC, a new tool to derive natural background levels by the component separation approach: application in two different hydrogeological contexts in northern Italy. Journal of Geochemical Exploration. 158, 44-54. Singh, K.P., Malik, A., Mohan, D., Sinha, S., 2004. Multivariate statistical techniques for the evaluation of spatial and temporal variation in water quality of Gomti River (India)—a study. Water Research, 38, 39893992. Singh, K.P., Malik, A., Sinha, S., 2005. Water quality assessment and apportionment of pollution sources of

PT

Gomti River (India) using multivariate statistical technique, a case study. Analytica Chimica Acta, 538, 355-374.

RI

Stiff, H.A.Jr. 1951. The interpretation of chemical water analysis by means of patterns. Journal of Petroleum Technology, 3, 1-16.

SC

Stumm, W., Morgan, J.J., 1996. Aquatic Chemistry – Chemical Equilibria and Rates in Natural Waters. John Wiley and Sons, 2nd edition, New York.

Templ, M., Filzmoser, P., Reimann, C., 2008. Cluster analysis applied to regional geochemical data:

NU

Problems and possibilities. Applied Geochemistry, 23, 2198-2213.

Urresti-Estala, B., Carrasco-Cantos, F., Vadillo-Pérez, I., Jiménez-Gavilán, P., 2013. Determination of

MA

background levels on water quality of groundwater bodies: A methodological proposal applied to a Mediterranean River basin (Guadalhorce River, Málaga, southern Spain). Journal of Environmental Management, 117, 121-130.

D

Venturelli, G., Caboi, R., Frau, F., Toscani, L., 1999. Geochemistry of spring and well waters and

PT E

mineralogy of the interacting rocks in the Logudoro region (NW Sardinia, Italy). Quaderni di Geologia Applicata, 6, 11-23.

Voutsis, N., Kelepertzis, E., Tziritis, E., Kelepertsis, A., 2015. Assessing the hydrogeochemistry of groundwaters in ophiolite areas of Euboea Island, Greece, using multivariate statistical methods. Journal of

CE

Geochemical Exploration, 159, 79-92.

Wendland, F., Berthold, G., Blum, A., Elsass, P., Fritsche, J.-G., Kunkel, R., Wolter, R. 2008. Derivation of

AC

natural background levels and threshold values for groundwater bodies in the Upper Rhine Valley (France, Switzerland and Germany). Desalination, 226, 160–168. Yidana, S.M., 2010. Groundwater classification using multivariate statistical methods: Southern Ghana. Journal of African Earth Sciences, 57, 455-469.

ACCEPTED MANUSCRIPT Captions of figures Figure 1. Location of the Sardinia island; a) digital elevation model (R.A.S., 2013a); b) simplified soil use map (modified after R.A.S., 2013b); c) simplified lithological map (modified after R.A.S., 2013c) with locations of the groundwater sampling sites. Figure 2. Box plots showing the electrical conductivity values in groundwater samples grouped according to the predominant hydrogeologic complex (All data).

PT

Figure 3. Piper diagrams showing major components in groundwater (GW) distinguished according to the predominant hydrogeologic complex (All data).

RI

Figure 4. Plots showing sulfate versus chloride concentrations in groundwater (GW) distinguished according to the predominant hydrogeologic complex (All data). The straight line represents the dilution line of

SC

seawater (SW).

Figure 5. Map showing the distribution of nitrate concentrations in Sardinian groundwater (All data).

NU

Figure 6. Dendrogram resulting from hierarchical cluster analysis performed on All data (a) and the Selected data (b).

MA

Figure 7. Stiff diagrams based on median concentrations reported in Table 3; (a) groundwater chemistry in clusters identified using All data, and (b) those identified using the Selected data. Figure 8. Regional distribution of threshold values of chloride (a), sulfate (b) and fluoride (c) calculated on

D

the Selected data. Gray lines in the maps indicate the boundaries of groundwater bodies (R.A.S., 2011).

PT E

Figure 9. Plot showing the fluoride activity versus the calcite saturation index (SI) in groundwater belonging

AC

CE

to clusters G4Sel, G8Sel and G9Sel.

Figure 1

AC

CE

PT E

D

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

AC

CE

PT E

D

MA

Figure 2

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

AC

CE

PT E

D

Figure 3

PT E

D

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

AC

CE

Figure 4

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

AC

CE

PT E

D

Figure 5

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

AC

CE

PT E

D

MA

Figure 6

PT E

D

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

AC

CE

Figure 7

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

AC

CE

PT E

D

Figure 8

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

AC

CE

PT E

D

Figure 9

ACCEPTED MANUSCRIPT Table 1. Summary statistics of All data: n = number of samples, n

Min

Max MAD

10th

20th

25th

Median 75th 80th 90th 95th 99th

1414

0

0

4.2

9.3

0.4

6.5

6.8

6.9

7.2

7.5

7.6

7.9

8.1

8.5

V

1414

0

0

-0.14

0.86

0.07

0.18

0.26

0.31

0.37

0.40 0.41 0.46 0.48 0.51

DO

mg/L

1112

0

0

0.20

11

2.2

2.0

3.0

3.3

4.9

6.3

6.6

7.6

8.3

9.1

TDS

g/L

1414

0

0

0.04

8.6

0.43

0.23

0.33

0.39

0.63

1.0

1.1

1.5

1.9

3.0

EC

mS/cm 1414

0

0

0.07

10.4

0.65

0.36

0.50

0.58

0.94

1.5

1.7

2.3

3.0

5.0

Ca

mg/L

1414

0

0

1.7

559

48

15

22

27

55

95

105

134

165

248

Mg

mg/L

1414

0

0

0.80

414

17

6.9

10

12

21

37

41

57

74

133

Na

mg/L

1414

0

0

6.1

1750

79

29

41

47

92

180

209

319

421

737

K

mg/L

1414

16

1.1

<0.5

120

4.0

1.2

1.8

2.1

4.4

8.9

11

16

23

53

HCO3 mg/L

1414

0

0

0.10

1890

145

60

91

113

212

309

336

391

443

679

Cl

mg/L

1414

0

0

8.0

2200

113

42

60

69

133

256

299

468

647 1207

SO4

mg/L

1414

0

0

1.0

2930

40

14

21

24

46

90

108

169

251

591

F

mg/L

1414

51

3.6

<0.01

13

0.15

0.08

0.10

0.10

0.20

0.40 0.40 0.70

1.0

2.9

NO3

mg/L

1414

83

5.9

<0.02

384

17

0.80

2.0

3.0

13

103

209

NO2

mg/L

289

259

90

<0.02

17

nc

<0.02 <0.02 <0.02

<0.02

0.10 0.10 0.10 0.11

1.3

NH4

mg/L

1149

938

82

<0.01

13

nc

<0.01 <0.01 <0.01

<0.01

0.05 0.09 0.21 0.92

1.8

PO4

mg/L

1136

751

66

<0.1

5.0

nc

<0.1

<0.1

0.2

2.0

MA

NU

SC

RI

Eh

PT

pH

n

<0.1

40

67

0.5

1.0

CE

PT E

D

<0.1

33

Table 2. Number of samples in clusters per hydrogeologic complexes (All data).

G1

138

G2

207

G3

AC

Tertiary Mesozoic Paleozoic Quaternary Quaternary Tertiary Granitic Metamorphic volcanic carbonatic carbonatic sediments basalts sediments rocks rocks rocks rocks rocks

Cluster

8

32

4

22

2

0

42

28

102

14

12

26

0

7

36

10

120

15

0

10

9

25

13

19

29

G4

208

43

24

43

35

13

13

24

13

G5

299

157

11

40

50

18

8

13

2

G6

169

49

14

59

13

18

13

2

1

G7

94

16

3

28

24

1

6

4

12

G8

28

4

1

17

5

0

0

1

0

G9

151

66

4

30

28

14

2

2

5

Total

1414

460

103

243

212

91

62

143

100

1.0

ACCEPTED MANUSCRIPT

Cluster

n

TDS

Ca Mg Na

K

HCO3

Cl

SO4

F

NO3

PT

Table 3. Median values of TDS and dissolved components in clusters identified using All data as compared to those derived from the Selected data (in italics).

12

7.4

31

1.5

50

45

14

0.10

9.9

G1Sel

104 0.18

13

6.7

32

1.6

50

51

15

0.10

3.4

G3

120 0.37

47

14

30

1.4

172

53

25

0.10

1.0

G3Sel

113 0.39

50

18

28

1.4

187

46

26

0.10

0.6

G2

207 0.40

24

14

72

4.0

98

101

30

0.21

11

G4

208 0.50

54

16

51

2.3

214

79

27

0.10

13

G4Sel

95

0.35

37

14

45

2.5

134

68

29

0.30

5.0

G5

299 1.02

72

32 181 8.9

262

260

79

0.30

25

G5Sel

73

0.61

24

15 154 6.8

147

190

35

0.30

3.0

G6

169 0.71

83

23

89

5.9

268

135

48

0.12

22

G6Sel

102 0.74

79

32

91

4.4

274

146

56

0.10

4.5

G7

94

0.81

72

37 123 4.8

255

166 108 0.30

1.0

G7Sel

62

0.74

63

34 122 5.4

239

G8

28

0.86

13

2.8 279 3.8

239

G8Sel

25

0.86

13

2.9 271 3.8

G9 G9Sel

D

MA 87

0.30

0.7

201

79

0.92

2.0

229

201

85

1.00

1.0

151 1.85 139 65 373 15

370

588 197 0.50

36

67

309

495 134 0.60

5.0

50 368 12

AC

CE

96

PT E

171

1.67

SC

138 0.18

NU

G1

RI

mg/L

ACCEPTED MANUSCRIPT Table 4. Background ranges calculated by the Median±2MAD (backtransformed data) of sulfate, chloride and fluoride in groundwater belonging to each cluster identified using All data and Selected data in italics. Sulfate (mg/L)

Fluoride (mg/L)

ln Median ln MAD Median±2MAD

ln Median ln MAD Median±2MAD

ln Median ln MAD Median±2MAD

0.40

20

98

2.62

0.48

5.2

36

-2.30

0.53

0.03

0.29

G1Sel

3.93

0.49

19

135

2.69

0.64

4.1

53

-2.30

0.69

0.03

0.39

G3

3.97

0.66

14

199

3.23

0.78

5.3

121

-2.30

0.39

0.05

0.22

G3Sel

3.83

0.65

13

168

3.26

0.68

7

102

-2.30

0.50

0.04

0.27

G2

4.62

0.63

29

357

3.40

0.45

12

73

-1.56

0.80

0.04

1.04

G4

4.37

0.38

37

168

3.30

0.60

8.1

91

-2.30

0.87

0.02

0.57

G4Sel

4.22

0.43

29

160

3.36

0.63

8

101

-1.22

0.86

0.05

1.67

G5

5.56

0.48

99

681

4.37

0.48

30

205

-1.20

0.60

0.09

1.00

G5Sel

5.25

0.44

78

460

3.56

0.64

10

127

-1.20

0.60

0.09

1.00

G6

4.90

0.35

66

273

3.87

0.40

21

107

-2.12

0.60

0.04

0.40

G6Sel

4.98

0.49

54

391

4.03

0.69

14

223

-2.30

0.95

0.01

0.67

G7

5.11

0.44

69

401

4.68

0.76

23

494

-1.20

0.68

0.08

1.17

G7Sel

5.14

0.68

44

670

4.47

0.80

18

428

-1.20

0.60

0.09

1.00

G8

5.30

0.41

89

453

4.37

0.93

12

505

-0.08

0.80

0.19

4.56

G8Sel

5.30

0.31

108

375

4.44

0.88

15

492

0.01

0.70

0.25

4.07

G9

6.38

0.58

184

1884

5.28

0.57

63

612

-0.69

0.70

0.12

2.02

G9Sel

6.21

0.54

170

1444

4.90

0.98

19

962

-0.51

0.74

0.14

2.65

D

PT E

SC

RI

PT

3.80

MA

G1

NU

Chloride (mg/L)

CE

Table 5. Threshold values in clusters identified using the Selected data. Values in bold are above limits. Chloride Sulfate Fluoride

AC

(mg/L)

G1Sel

140

53

0.4

G3Sel

170

100

0.3

G4Sel

160

100

1.7

G5Sel

460

130

1.0

G6Sel

390

220

0.7

G7Sel

670

430

1.0

G8Sel

380

490

4.1

G9Sel

1450

960

2.7

ACCEPTED MANUSCRIPT Highlights Geochemical, statistical and GIS methods were used to distinguish hydrogeochemical groups

PT

The estimator median±2MAD was appropriate to calculate background ranges in groundwater

AC

CE

PT E

D

MA

NU

SC

RI

Threshold values for chloride, sulfate and fluoride in groundwater were proposed at the regional scale