Origin, patterns and anthropogenic accumulation of potentially toxic elements (PTEs) in surface sediments of the Avilés estuary (Asturias, northern Spain)

Origin, patterns and anthropogenic accumulation of potentially toxic elements (PTEs) in surface sediments of the Avilés estuary (Asturias, northern Spain)

Marine Pollution Bulletin 86 (2014) 530–538 Contents lists available at ScienceDirect Marine Pollution Bulletin journal homepage: www.elsevier.com/l...

3MB Sizes 0 Downloads 73 Views

Marine Pollution Bulletin 86 (2014) 530–538

Contents lists available at ScienceDirect

Marine Pollution Bulletin journal homepage: www.elsevier.com/locate/marpolbul

Baseline

Origin, patterns and anthropogenic accumulation of potentially toxic elements (PTEs) in surface sediments of the Avilés estuary (Asturias, northern Spain) C. Sierra a, C. Boado b, A. Saavedra b, C. Ordóñez c, J.R. Gallego c,⇑ a b c

Facultad de Ciencias de la Tierra, ESPOL, Guayaquil, Ecuador Department of Statistics, University of Vigo, Spain Department of Mining Exploitation and Prospecting, Polytechnic School of Mieres, Spain

a r t i c l e

i n f o

Article history: Available online 10 August 2014 Keywords: Particulate emissions Sediment contamination Multivariate analysis Geostatistics Enrichment factors

a b s t r a c t Sediment quality has been assessed within the Avilés estuary, an important industrial area in the NW of Spain. The study started with a geochemical characterization of the superficial sediments that revealed some anomalous metal(oid)s concentrations in sensitive areas such as beaches or dunes. These data were studied by means of multivariate statistical techniques and enrichment factors calculation to evaluate the correlations and geochemical origin within the different elements. A novel approach using the combination of enrichment factors with a sequential application of factor analysis, clustering and kriging was essential to identify the possible sources of pollution. The collected information suggested that Cd (strongly correlated with Zn and Pb) was the potentially toxic element most widely distributed and problematic. Furthermore, particulate emissions from Zn metallurgy, as well as dust generated by the mineral loading and stockpile activities in the port were identified as the most important sources of pollution. Ó 2014 Elsevier Ltd. All rights reserved.

Marine estuary areas are crucial ecosystems for a variety of terrestrial aquatic species. Although new environmental policies have reduced anthropogenic alteration in the recent years, these areas are frequently altered by a number of human activities including among others (e.g. Kennish, 2002; Kennish, 2008; Fonseca et al., 2014; Wu et al., 2014): (a) industrial uses, such as land reclamation, atmospheric deposition and waste disposal; (b) navigation, with dike and ditch construction, impoundments, wells and dredge-and-fill operations and disposal; (c) agricultural, with land occupation deforestation, water extraction and diversion, introduction of non-native species, use of chemical fertilisers; (d) urban development, uncontrolled expansion in coastal watersheds, waste water discharges; and finally (e) recreational occupation. All of these activities have been affecting Avilés estuary with higher or lower intensity for decades and this implies an important deterioration of the environment of the area (López Peláez and Flor, 2008). The ways by which released potentially toxic elements (PTEs) (Hooda, 2010) interact with the sediment include precipitation, ion exchange, adsorption, and complexation (Tack, 2010). These processes are in turn strongly correlated with the organic matter content, the presence of clay minerals and hydrous Fe and Mn ⇑ Corresponding author. Tel.: +34 985 458 027; fax: +34 985 458 001. E-mail address: [email protected] (J.R. Gallego). http://dx.doi.org/10.1016/j.marpolbul.2014.06.052 0025-326X/Ó 2014 Elsevier Ltd. All rights reserved.

oxides (Tack, 2010). Thus, organic matter, which is usually present in some marine environments as is the case of estuaries, is able to physically trap the metals because it exhibits a negative surface charge, high surface area and high cation exchange capacity (Tack, 2010); meanwhile clays as well as hydrous Fe and Mn oxides collect and concentrate potentially toxic elements due to their large negative surface charge and moderate to high CEC (Sierra et al., 2010; Violante, 2013). Conversely, sands, which have larger grain size than clays and are composed mainly of SiO2, are less likely to accumulate pollution, although the existence of placers, tidal and atmospheric deposition can cause their enrichment in certain pollutants. In this regard pollution present in sands in locations such as beaches or dunes can cause a number of harming effects to human health (WHO, 2000). Consequently, monitoring sands in affected areas offers the possibility of not only preventing these sorts of problems, but also using them as a geochemical record to reflect the environmental conditions of the surrounding area (Tessier et al., 2011). The calculation of Enrichment Factors (EFs) over raw data is a usual strategy to maintain the hypothesis that trace metals are of anthropogenic origin (Sucharovà et al., 2012). However, regional geochemical surveys demonstrate that EFs can be high or low due to a number of reasons, of which pollution is only one. Consequently, the use of EFs to detect or prove human influence should

C. Sierra et al. / Marine Pollution Bulletin 86 (2014) 530–538

be used carefully (Reimann and de Caritat, 2005). In this context, multivariable statistical methods can complement/substitute EFs approaches; in fact, they are commonly used in geochemical studies to distinguish between polluted and unpolluted areas or between natural and anthropogenic inputs, to obtain information beyond univariate or bivariate statistics, and to identify pollution sources as well as transportation (e.g. Yongming et al., 2006; Martínez-López et al., 2008; Yalcin et al., 2008). In this sense, cluster analysis is used to separate the samples into groups, and factor analysis makes it possible to simplify the explanation of a set of observations by reducing the dimensionality (Gallego et al., 2013). In addition, geostatistics, the application of probabilistic methods to regionalized variables (Matheron, 1965; Chilès and Delfiner, 2012), makes the characterization of the spatial patterns and can be successfully combined with multivariate statistics in environmental studies (Webster and Oliver, 2007). One of its main characteristics is that it allows the characterization of the phenomenon from a number of observations obtained from different available locations (Diggle and Ribeiro, 2007; Chilès and Delfiner, 2012), assuming that there is dependency between observations corresponding to proximate locations in space. This methodology includes two steps (1) the selection of a suitable variogram model which best describes the spatial distribution of the variable under consideration (2) the prediction of the additional values of the spatial variable by means of kriging systems. In this regard, kriging is a linear interpolation procedure that provides the best linear unbiased estimator for quantities that vary in space, and it allows the correlation between observations to predicted element concentrations at non-sampled locations (e.g. Alary and Demougeot-Renard, 2010; Saavedra et al., 2013). On the whole, the specific objectives pursued in this study were the following: (a) Assess the pollution levels in the Estuary of Avilés area establishing the main spatial correlations; (b) Identify the anthropogenic sources of pollution distinguishing them from the complex geochemical background of the zone, (c) To introduce a novel approach to calculate enrichment factors using a sequential application of factor analysis and clustering. Avilés is located in an important urban-industrial area (agglomeration higher than 100,000 habitants) including one of the main harbors in the North of Spain (Fig. 1). Northeast of the city lies the estuary of Avilés, geologically characterized for Post-Ordovicic soft sedimentary rocks such as Triassic siltstone, Jurassic limestone and dolomite from the Gijón Formation and Middle Jurassic siliceous conglomerate (Julivert et al., 1973).

531

Nowadays there are several sandy beaches in the area as well as some small beaches formed within the estuary margins in which clastic sediments are predominant. Moreover, sedimentary structures such as dunes, swash marks, wave and aeolian ripples, and also stratification profiles (reverse grading), are common throughout the study area where the dominant winds (annual maximum average frequency > 40%) along the Asturian coast are from the SW (Flor, 1981, 1986; Flor-Blanco et al., 2013). In this context, the locations studied were divided in three main areas avoiding docks, roads and paved sectors not available for sampling (Fig. 2): (a) Area A comprises El Espartal beach, and the largest eolian dune system of Asturias, located at the left side of the estuary. Both of them have been significantly deteriorated due to industrial and urban activities in the area. The geological substrate of this area is only formed by sand (Sierra et al., 2013); however, some dredged materials have been dumped in San Juan de Nieva area (Flor and Flor-Blanco, 2005) increasing its deterioration. The surrounding area includes two hydrometallurgical-electrolytic plants for Zn production, in addition to several other factories manufacturing chemical products, fertilizers and glass. (b) Area B includes the Llodero Cove, which is a zone under tidal influence occupied by intertidal sediments located at the mouth of the Vioño creek, and some other locations of small accessible beaches on the right side of the estuary (López Peláez and Flor, 2008). The substrate consists of sand which is gradually substituted by mud in which an important presence of organic matter is observed (Sierra et al., 2013). The zone suffers from significant wastewater discharges and land occupation (López Peláez and Flor 2008). Aluminium and steel factories are located at this side of the estuary, in addition to several spoil heaps and open-pit quarries. (c) Area C corresponds to Xagó Beach, it is located to the North of the estuary and isolated by San Juan Peninsula. It was selected to be used as a feasible background given its similar geological setting composed mainly of sands (Sierra et al., 2013), also because this peninsula acts as a barrier, making it likely that Xagó should be less affected by pollution than the other two areas. (d) Focusing on industrial activities, the whole district has experienced an important industrialization process within the last sixty years, starting with the installation of a steel factory in 1951. This plant was followed by the settlement of some of the abovementioned industries (Zn and Al production,

Fig. 1. Geographical setting of the Avilés estuary area in Asturias (North of Spain).

532

C. Sierra et al. / Marine Pollution Bulletin 86 (2014) 530–538

Fig. 2. Location of the study area. The dots represent the location of the sampling points.

chemical industries, etc.). These facilities caused not only morphological alterations such as the construction and expansion of the docks which altered marshes and dunes that surrounded the river, but also exponentially increased the atmospheric emissions converting Avilés into one of the world´s most polluted cities in the 1980s (Berciano et al., 1989), thus affecting soil and sediment quality (Gallego et al., 2002). Although pollution emissions have been moderated in recent years, they still remain remarkable in present (EEA, 2012) supported by secondary sources such as marine and road cargo traffic and the urban population of Avilés. In addition, regional effects such as those created by power plants and other heavy industries in the nearby densely populated urban and industrial areas of Gijón and Oviedo (less than 30 km far from Aviles) must be considered. Sampling campaign procedure followed a stratified scheme using a quasi-regular grid in each of the Areas above-defined. A total of 204 samples were taken (103 of them in the Area A, 82 in Area B and 19 in Area C). Since a geostatistical study was to be performed, samples were collected so that both the variability at the short and long distance would be taken into account in order to construct realistic variograms. Thus, from the upper 0–20 cm of sediment a couple of samples (approx. 100 g each) were collected with a modified Van Veen grab sampler, at a distance of up to 2 m, at each sampling point. The samples were first screened in situ in order to discard materials larger than 2 cm, then homogenised and stored in inert plastic bags. The pairs of sampling points were separated from each other with a minimum distance of 10 m. Samples were dried at room temperature in order to minimize loss of volatile elements, then quartered and ground using a vibratory disc mill (RS 100 Retsch) at 400 rpm for 40 s to obtain a grain

size of < 125 lm. The milled samples were shipped to the accredited (ISO 9002) ACME laboratories (Vancouver, Canada). There, 1-g representative sub-samples of the ground product were leached by means of an ‘Aqua regia’ digestion (HCl + HNO3), and the concentrations of 18 major, minor and trace elements (Ca, Mg, K, Na, Al, Fe, S, Cu, Pb, Zn, Cd, Ni, Mn, As, Sr, Sb, La, Cr) were analysed for the digested material by inductively coupled plasma optical emission spectrometry (ICP-OES). In this work we have combined factor analysis, cluster analysis and discriminant analysis techniques to perform a multivariate analysis with the concentrations of the chemical elements. All calculations were performed with the R freeware program (R Development Core Team, 2012). Further information on the procedures used can be found in the Supplementary material SM2. Kriging comprises a set of point interpolators which require a point map as input and return a raster map with predictions and prediction variances. The predictions are weighted averaged input point values where weights are determined such that the prediction error in each output pixel is minimized. These steps were employed to map the spatial distribution of the factors calculated by factor analyses. Additional details are given is the Supplementary material SM2. An initial univariate statistical characterization was performed on the raw data of a selected group of elements; the statistical parameters obtained are depicted in Table 1. Possible contaminants such as Zn and Pb exhibited the largest standard deviations, and differences between minimum and maximum concentrations, whereas much lesser variations corresponded to major elements such as K, Na and Al. In addition, according to the variations in the coefficients of skewness (also confirmed by means of the Kolmogórov-Smirnov test), the distribution of each variable is far from Gaussian (Lattin et al. 2003). Justification to this result could be that wind and water currents

C. Sierra et al. / Marine Pollution Bulletin 86 (2014) 530–538 Table 1 Descriptive statistics (minimum and maximum concentrations, mean, standard deviation, skewness) for the analysis of the sediment samples (mg kg 1). Element

Mean

SD

Min

Max.

Skewness

Ag Al As Ca Cd Co Cr Cu Fe K La Mg Mn Na Ni Pb S Sr V Zn

0.53 3600.00 18.81 17,700.00 5.96 2.92 7.37 34.92 11,600.00 600 4.51 4100.00 207.23 900.00 6.99 176.06 800.00 76.60 10.63 1437.29

0.63 4000.00 27.72 21,300.00 10.80 3.85 12.44 107.07 15,000.00 900 2.42 6200.00 310.88 1500.00 12.69 668.35 800.00 89.28 7.57 1895.98

0.1 1100.0 4.0 200.0 0.5 1.0 2.0 1.0 4400.0 200.0 2.0 700.0 39.0 100.0 1.0 3.0 500.0 3.0 4.0 37.0

5.0 29,700.0 228.0 101,300.0 87.2 49.0 151.0 1272.0 195,500.0 8600.0 24.0 42,500.0 2226.0 10,300.0 167.0 8710.0 5500.0 411.0 67.0 10,000.0

4.1 4.4 5.5 1.8 4.2 9.9 9.2 9.2 10.8 5.9 4.4 3.8 4.4 3.4 11.6 11.9 3.6 1.9 4.8 2.8

continuously transport sediments in the area leading to their sensitivity to skewness factors such as allochthonous natural or anthropogenic inputs. The specific values for each one of the studied areas are provided in Supplementary materials SM1. The Enrichment Factors (EFs) approach can be applied to the univariate data in order to estimate potentially toxic elements (PTEs) contamination according to a comparison between nonpolluted and polluted areas with a conservative lithogenic element as a reference (Sucharovà et al., 2012). The non-polluted location selected was Area C (Xagó beach) since it has the same underlying geology as the other two; it is close to the estuary and topographically sheltered from the industrial effluents and emissions by the San Juan Peninsula. The normalizing element was Al mainly because no specific anthropogenic enrichment was suspected for this metal in this area, but also because it behaves conservatively for most marine environments (Herut and Sandler, 2006). Therefore the EFs (Table 2) were calculated as follows: EF = (M/N)Areas A and B/(M/N)Area c, wherein M is the concentration of the element of interest, and N the concentration of Al (Herut and Sandler, 2006). Table 2 also includes some regulatory levels for elements of environmental concern in marine sediments. The EFs values obtained advocate an abnormal behavior (positive anomalies) for Cd, Cu, Pb and Zn thereby suggesting their probable anthropogenic origin. In addition, environmental standards revealed clear anomalies for Cd, Zn and Pb. Both procedures, when interpreted in conjunction, substantiate a significant enrichment in Cd, Zn and Pb in the study site (Areas A and B). Following the preceding considerations, spatial distribution of the individual samples exceeding the Spanish environmental standards (CEDEX, 2007) for Cd, Pb and Zn are schematized in Fig. 3. Table 2 Enrichment factors (EFs), mean (mg kg 1), and reference values of heavy metals and metalloids of concern according to the Spanish ‘‘Centro de Estudios y de Experimentación de Obras Técnicas’’ (CEDEX, 2007), Canada´s ISQG (threshold effect level), and PEL (Predicted Effect Level) regulations (CCME, 2001)). Element

Enrichment factor

Concentration mean

Reference value CEDEX

ISQG

PEL

As Cd Cr Cu Pb Zn

0.7 4.5 1.22 4.79 2.35 2.66

18.81 5.96 7.37 34.92 176.06 1437.29

80 0.5 200 50 60 250

7.24 0.7 52.3 18.7 30.2 124

41.6 4.2 160 108 112 271

533

Comparisons between these maps indicate remarkable similarities in the distribution patterns of the three elements, in addition to the following deductions: (a) no abnormal levels were detected in Area C (Xagó beach); (b) notable concentrations of the three elements were observed along Area A (El Espartal beach and dunes), specially in the subarea closer to the left side of the estuary mouth; (c) high concentrations of Cd, and lower but also significant concentrations of Zn and Pb, were detected in Area B (Llodero Cove). Multivariable statistical analysis was used to ascertain the pollution sources. Prior to this study, and since the original data do not follow a normal distribution, the original measurements were logtransformed to reduce skewness. Subsequently, in order to avoid the effect of the different units of measurement, the log-transformed variables were standardized. Then, factor analysis using maximum likelihood extraction followed by varimax rotation with the original variables transformed was performed. The results summarized the 18 variables over 204 samples in four groups of inter-related variables (factors). Table 3 shows the factor loadings constituting the matrix K as well as other associated parameters. These parameters include, among others, the cumulative variance indicating that the factors represent 78.9% of the variance in the data, as well as the proportion of the variance of the variables that is both error free and common to other variables (communalities), which, despite the exceptions of Ag, As and Mn and S, was high enough for most of the elements. Factor 1 (F1) explains 23.9% of the total variance and it is strongly correlated with chalcophiles such as Zn, Cd, Pb, ad Cu. A moderate loading factor is attributed to As and Ag, although as mentioned before, these two presented low communalities. The very well-known geochemical association of Zn, Cd and Pb (Burton et al., 2005) corresponds with ores used in Zn production (Sphalerite –ZnS–, usually containing Cd, and Galena –PbS– usually containing Ag). If, as we stated before, we consider that the concentrations of these elements in Areas A and B are high, the origin of these exogenous metals could be attributed to the dust generated by the mineral loading and stockpile activities performed in the port and particulate emissions of Zn metallurgy (see Li et al., 2006; Mattielli et al., 2009 for similar cases). The presence of Zn particles has been corroborated by means of Scanning electron microscopy with energy dispersive X-ray spectroscopy (SEM/ EDX) as can be observed in Supplementary material SM3. Factor 2 (F2), accounting for an additional 21% of the variance, is mainly linked with Fe, Ni, and Co, with moderate factor loadings for V and Cr (see Table 3). This factor could be attributed to the presence of natural iron oxy-hydroxides within the sampled material. However, the strong industrial and urban activities in the area have to be considered; in fact V-Ni associations are usually related to fossil fuel combustion in heating systems, vehicles and industrial plants (e.g. Celo et al., 2012). As a consequence, a mixed origin including both natural and urban-industrial sources is proposed. Factor 3 (F3) explicates another 17.4% of the total variance and is strongly correlated with lithogenic elements such us K, Al, and also La as a representative of Rare Earths. This is clearly a natural factor which reflects the siliciclastic materials present in the local geology (López Peláez and Flor, 2008). Factor 4 (F4), which describes another 16.6% of the total variance, comprises elements associated to the alteration of the carbonates, particularly of biogenic origin (for instance Ca and Mg from shells). The relatively high load found for Na also suggests the marine influence in this factor. The negative load of Cd and Zn is remarkable thereby suggesting a negative correlation between Zn–Cd–Pb contamination and sea-influenced samples. Once the main four factors explaining approximately 80% of the variability in the study area were obtained, the factor score matrix was calculated and a cluster analysis was performed. Thus, after the application of Ward’s clustering algorithm, the dendrogram

534

C. Sierra et al. / Marine Pollution Bulletin 86 (2014) 530–538

Fig. 3. Areas exceeding environmental standards indicated in Table 2 for Cd, Pb and Zn. Note that the size of the dot is proportional to the level of contamination (classification into five groups according to their lowest -smaller dots- and the highest -bigger dots-values).

Table 3 Factor loadings, communalities, sum of squares of the factor loadings, percentage of variance explained, and cumulative percentage of variance explained by the varimax rotated factors (extracted by maximum likelihood). Factor loadings higher than 0.6 are indicated in bold and those close or equal to zero are omitted.

Ln Cu LnPb LnZn LnAg LnNi LnCo LnMn LnFe LnAs LnSr LnCd LnV LnCa LnLa LnCr LnMg LnAl LnNa LnK LnS Sum of squares of loadings Proportion variance Cumulative variance

Factor1

Factor2

Factor3

0.741 0.789 0.867 0.651 0.385 0.422 0.525 0.255 0.616

0.416 0.472 0.370

0.304

0.828 0.267 0.397 0.277 0.446 0.343 0.252 4.781 0.239 0.239

0.220 0.157 0.162 0.282 0.146 0.221 0.217

0.825 0.726 0.162 0.876 0.402 0.103 0.142 0.756

0.294 0.495

0.226 0.700 0.104 0.327 0.165 0.274 0.355 4.200 0.210 0.449

0.679 0.544 0.445 0.754 0.498 0.910 0.468 3.486 0.174 0.623

was cut at a height 40, obtaining a classification in four clusters (groups of samples) spatially distributed in each one of the studied zones (shown in Fig. 4). In order to explain the four clusters in terms of their relationship with the previously described factors, Fig. 5 shows box-plots for each factor corresponding to each of the four clusters (note that ordinate axis are in logarithmic scale therefore small distances indicate high variations): – Group 2 is the most clearly related with anthropogenic sources (particularly with the Zn–Cd–Pb association). The majority of samples belonging to this group is located on the left side of the estuary and concentrated in the surroundings of the industrial facilities, especially in the Espartal dune area. – Group 3 is correlated with carbonates and Na as it is highly represented by factor 4. This is consistent with the spatial distribution showed in Fig. 4 given that samples from the beaches (Espartal and Xagó) and others very close to the estuary mouth (right side) are included in this group. In fact, the unquestionable inclusion of Xagó samples in this group indicates that they were correctly proposed as a background, and that this background can be complemented with additional data from samples different to those of Xagó which are also included in this group. This does not imply that it is a completely natural

Factor4

0.168 0.449 0.170 0.140 0.946 0.218 0.143 0.980 0.128 0.110 0.770 0.134 0.523 0.162 0.294 3.312 0.166 0.789

Communality 0.815 0.898 0.918 0.468 0.908 0.755 0.552 0.908 0.568 0.911 0.839 0.908 0.979 0.686 0.875 0.803 0.893 0.667 0.993 0.435

background; in fact, we could consider it as a general background that takes into account a regional pollution in coherence with the slightly significant weight of Factor 2, including even the positive outliers of the other two factors with negative weight in the group. – Groups 1 and 4 are mostly influenced by Factor 3, i.e. these clusters are quite similar, and they have probably been formed based on the content of lithogenic elements whose concentration is conditioned by grain size parameters, thus suggesting a higher influence of fines in the formation of the cluster. In fact, grain-fine sediments (even soils) in the left side of the estuary, and muds belonging to the Llodero zone (right side of the estuary) are included; all of them belong to sedimentary environments not strongly affected by wind. – Group 4 has higher correlation values with Factor 3 than with the rest of the factors, thus indicating its higher influence in the formation of the group. – It must be pointed out that Factor 2 revealed similar and neutral correlation values with all the groups, thus fostering their previously suggested mixed origin that plays the role of ‘‘anthropo-geochemical’’ background. On the other hand, given that clustering is a limited approach, some groups are not well-distinguished (1 and 4); in fact, in these

C. Sierra et al. / Marine Pollution Bulletin 86 (2014) 530–538

535

Fig. 4. Spatial distribution of the clusters obtained by Ward’s algorithm. The numbered dots indicate each one the different clusters found.

Fig. 5. Box-and-whisker plots of the varimax rotated factors for each of the four clusters determined by Ward’s algorithm.

sorts of methods equal importance is given to each one of the studied elements, while some of them pose a higher environmental concern. However, it was possible to acquire more useful information by obtaining EFs referred to the abovementioned background Group 3, that is to say, EF = (Metal/Al)group 1,2,4 /(Metal/Al)group 3 as showed in Table 4. Data in Table 4 indicate that only Cd shows significant EFs for Groups 1, 2 and 4 whereas Zn and Ag also have high EF values in

Group 2. This suggests a higher mobility (or possibly more than one source) for Cd than for Zn. Therefore, Zn, and especially Cd, present anomalies visibly above the background and this surely reveals a powerful source of pollution (Zn metallurgy as suggested above) for these elements being predominant over minor sources. Additionally, it appears an Ag anomaly that fosters this hypothesis, i.e., Ag is a minor contaminant usually not considered in environmental studies, nevertheless in our scenario the only familiar

536

C. Sierra et al. / Marine Pollution Bulletin 86 (2014) 530–538

Table 4 Average concentration (mg kg than 1.5 in italic). Element

1

), ratio element to aluminum, in addition to EFs, which have been calculated for each group and element, with respect to Group 3 (values higher

Average concentration

Cu Pb Zn Ag Ni Co Mn Fe As Sr Cd V Ca La Cr Mg Al Na K S

Ratio element/Al

Enrichment factor

Group 1

Group 2

Group 3

Group 4

Group 1

Group 2

Group 3

Group 4

Group 1

Group 2

Group 4

6.7 29.3 587.6 0.3 4.4 1.7 73.1 81,100.0 10.6 11.9 2.1 7.9 23,600.0 3.8 4.7 14,200.0 22,400.0 8500.0 3900.0 6000.0

43.2 194.9 2,658.7 0.7 6.2 3.1 199.0 91,300.0 16.5 25.4 13.3 8.7 51,800.0 3.7 4.7 15,300.0 24700.0 1300.0 3300.0 5700.0

36.4 242.3 812.9 0.4 8.4 3.4 213.9 149,100.0 17.0 148.2 1.4 10.4 349,200.0 3.9 7.9 63,500.0 23,200.0 6600.0 3700.0 7400.0

65.6 262.4 2,396.8 0.9 9.4 3.8 408.3 142,400.0 37.0 109.9 11.3 17.4 257,200.0 7.6 13.6 72,000.0 89,900.0 20,200.0 17,100.0 15800.0

0.00030 0.00131 0.02623 0.00001 0.00020 0.00007 0.00326 3.62054 0.00047 0.00053 0.00009 0.00035 1.05357 0.00017 0.00021 0.63393 1.00000 0.37946 0.17411 0.26786

0.00175 0.00789 0.10764 0.00003 0.00025 0.00013 0.00806 3.69636 0.00067 0.00103 0.00054 0.00035 2.09717 0.00015 0.00019 0.61943 1.00000 0.05263 0.13360 0.23077

0.00157 0.01044 0.03504 0.00002 0.00036 0.00015 0.00922 6.42672 0.00073 0.00639 0.00006 0.00045 1.505172 0.00017 0.00034 2.73707 1.00000 0.28448 0.15948 0.31897

0.00073 0.00292 0.02666 0.00001 0.00010 0.00004 0.00454 1.58398 0.00041 0.00122 0.00013 0.00019 2.86096 0.00008 0.00015 0.80089 1.00000 0.22469 0.19021 0.17575

0.19 0.13 0.75 0.80 0.54 0.51 0.35 0.56 0.65 0.08 1.58 0.78 0.07 1.01 0.62 0.23 1.00 1.33 1.09 0.84

1.11 0.76 3.07 1.76 0.69 0.87 0.87 0.58 0.91 0.16 8.98 0.78 0.14 0.89 0.56 0.23 1.00 0.19 0.84 0.72

0.46 0.28 0.76 0.60 0.29 0.29 0.49 0.25 0.56 0.19 2.09 0.43 0.19 0.50 0.44 0.29 1.00 0.79 1.19 0.55

source for Ag is also Zn metallurgy given that the geochemical association between Zn–Pb–Cd and Ag in Zn and Pb ores is recognized. Conversely, for Pb, no high EF was found; however this does not mean that there is no Pb pollution linked to the Zn metallurgy, in fact there must be as shown in Fig. 4. As a consequence, the low EFs found mean that Pb anomalous concentrations have a mixed origin (several industries, dense traffic, etc.) that has been assimilated in the background designed in this work. To further support the results, a geostatistical approach by ordinary kriging was applied. The study was conducted using the factor scores. In order to avoid interpolation between areas with different underlying geology, as well as landforms, kriging was performed independently in the three study Areas A, B, and C with the only exception of not including the isolated beaches of the estuary mouth. In order to model the spatial distribution of the elements, omnidirectional variograms were built from measures of each one of the factors, and then this data adjusted to the possible theoretical variograms (linear, spherical and exponential) according to Matheron’s estimator or Cressie and Hawkins’ robust estimator in the three different zones (see Supplementary material SM3). Those variograms in which the median value of h(x) was closest

to 0.455 were selected to the confection of the maps, and are shown in Table 5. According to this premise, better fitting variograms correspond to exponential and spherical approximations over the linear (indicators of non-stationary variables), however fitting some data. The measured nugget value for the different fitting models for each factor in every zone ranged from 0 to 0.7547. The value of the sill varied from 0.0273 to 893.3321. The range, which indicates the distance over which the samples are spatially dependent, varied widely from 0 m (pure nugget, thus having no spatial dependence) to over 394,000 m. The nugget ratio was used as an indicator of the degree of spatial dependence of the factors. Accordingly, nugget ratios smaller than 0.25 were said to indicate strong spatial dependence; between 0.25 and 0.75 moderate spatial dependence; and values higher than 0.75 a weak spatial dependence (Jabro et al., 2006). These values suggest that stronger spatial dependences predominate in fluvial environments (zone B) over aerial environments (zones A and C). To identify the distribution patterns of the factors, contour maps were represented. The unknown values of the factors were predicted using interpolation by kriging, which is carried out taking into account the dependency structure determined in

Table 5 Parameters and the cross-validation results of the best fitted omnidirectional variograms (Matheron and Cressie and Hawkins) for each one of the factors in the three studied zones. Co (nugget), C (partial sill), C + C0 (sill), C0/C + C0 (nugget ratio). RSS 500 stands for residual sum of squares at 500 m, and mean and median are referred to h(x). Zones correspond with the three areas indicated in Fig. 3. Estimator

Model

C0

RSS 500

MSE

Mean

Median

Area 1

F1 F2 F3 F4

Matheron Cressie–Hawkins Cressie–Hawkins Cressie–Hawkins

Exp. Exp. Exp. Exp.

0.7547 0.0923 0.3581 0.0000

C 0.3141 0.3900 0.0654 1.4679

C + C0 1.0688 0.4823 0.4235 1.4679

C0/C + C0 0.7061 0.1914 0.8455 0.0000

Range 1,297.7944 0.0000 1297.8000 383.0363

62.7448 15.5251 13.7424 28.4215

7.9212 3.9402 3.7071 5.3312

0.7294 1.9257 1.0037 1.6595

0.3972 0.3662 0.4587 0.4392

Area 2

F1 F2 F3 F4

Cressie–Hawkins Matheron Matheron Cressie–Hawkins

Sph. Exp. Sph. Exp.

0.1982 0.2264 0.4528 0.2142

0.3242 893.1057 152.2429 51.5312

0.5224 893.3321 152.6957 51.7453

0.3794 0.0003 0.0030 0.0041

113.8793 394,092.8100 93,228.6894 155,788.4250

26.4901 61.0984 62.5637 9.6599

5.1469 7.8165 7.9097 3.1080

0.6725 1.4750 0.8052 1.0622

0.3415 0.5227 0.3044 0.3248

Area 3

F1 F2 F3 F4

Matheron Matheron Cressie–Hawkins Cressie–Hawkins

Sph. Sph. Sph. Exp.

0.0039 0.0531 0.0398 0.0000

0.0504 0.2240 0.0273 0.0627

0.0543 0.2771 0.0672 0.0627

0.0715 0.1916 0.5931 0.0000

103.0225 0.0000 34.3407 26.0657

0.0595 0.5443 0.1036 0.1142

0.2439 0.7377 0.3219 0.3379

0.8882 0.9894 1.1227 0.8461

0.5643 0.4290 0.4626 0.5031

C. Sierra et al. / Marine Pollution Bulletin 86 (2014) 530–538

537

Fig. 6. Kriging spatial distribution maps of the four geochemical factors.

the variograms. The kriged maps of the four factors are depicted in Fig 6. The factors were classified into five quantiles corresponding to the lowest (light areas) and the highest values (dark areas). The kriging contour map of Factor 1 (Fig. 6a) reveals the highest influence of this factor in the area to the north of El Espartal– Salinas site as a consequence of the dominant winds which are from the SW and the vicinity of the Zn industry (represented by the Zn–Cd–Pb–Ag previously established connection), as well as a moderate influence over the Llodero Cove area, and very low in the Xagó area. The spatial distribution of Factors 2 and 3 (Fig. 6b and c), illustrates a predominance of high values for these factors in Llodero Cove area, moderate in El Espartal–Salinas, and low values for Xagó. This is in accordance with the smaller influence of the Zn–Pb–Cd correlation, and the lower levels of pollution in the cove as a consequence of dilution and sediment transport. Especially for Factor 3, values are decreasing with growing distances to the sea.

Finally, as regards Factor 4 (Fig. 6d), it has the highest values over the Xagó Beach area, once again indicating the lower anthropic influence on this zone. The marine influence is also clearly represented by this factor, as it can be observed in the plot for the Espartal–Salinas area where darker colors increased offshore. The joint interpretation of Figs. 5 and 6 leads to a congruent conclusion in those groups that had been previously associated with some factors. That is to say, group 2 of samples, characterized by factor 1, is more or less located in the dune area. Similarly, group 3, which embodies the initial suggested background and is represented by Factor 4, agrees with Xagó area, which also exhibits in kriging the highest values for this Factor 4. Finally groups 1 and 4, mostly represented by Factor 3, present in Fig. 5 a spatial distribution approximately coincident with kriging of Factor 3 in Fig. 6 (abundant it the Llodero Cove area). The information gathered provided a basis for delimitating the polluted zones and their sources, thus facilitating the development of specific air and soil monitoring activities, urban planning and environmental policies.

538

C. Sierra et al. / Marine Pollution Bulletin 86 (2014) 530–538

Acknowledgements Carlos Sierra thanks the Prometeo Project (SENESCYT-Ecuador) for financial support provided during the redaction of this research. The authors are also thankful to Dr. Teresa Albuquerque, at the Polytechnic Institute of Castelo Branco for assistance provided in this research. Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.marpolbul.2014. 06.052. References Alary, C., Demougeot-Renard, H., 2010. Factorial Kriging Analysis as a tool for explaining the complex spatial distribution of metals in sediments. Environ. Sci. Technol. 44, 593–599. Berciano, F.A., Dominguez, J., Alvarez, F.V., 1989. Influence of air pollution on extrinsic childhood asthma. Ann. Allergy Asthma Immunol. 62, 135–141. Burton, E.D., Phillips, I.R., Hawker, D.W., 2005. Geochemical partitioning of copper, lead, and zinc in benthic, estuarine sediment profiles. J. Environ. Qual. 34, 263– 273. CCME (Canadian Council of Ministers of the Environment), 2001. Canadian environmental quality guidelines. Chapter 6: Canadian Sediment Quality Guidelines for the Protection of Aquatic Life, Winnipeg, Canada. CEDEX, 2007. Ficha técnica. Clave 6.1. Materiales de dragado, Madrid, Spain. Celo, V., Dabek-Zlotorzynska, E., Zhao, J., Bowman, D., 2012. Concentration and source origin of lanthanoids in the Canadian atmospheric particulate matter: a case study Atmospheric. Pollut. Res. 3, 270–278. Chilès, J.P., Delfiner, P., 2012. Geostatistics: Modeling Spatial Uncertainty. Wiley, New York, USA. Diggle, P., Ribeiro, P.J., 2007. Model-based Geostatistics. Springer, New York, USA. EEA, 2012. Costs of air pollution from industrial facilities in Europe. http:// daviz.eionet.europa.eu/data/local-files/costs-of-air-pollution-from-industrialfacilities-in-europe/view (accessed 29.06.2014). Flor, G., 1981. Las dunas eólicas costeras de la playa de Xagó (Asturias). Trabajos de Geologia 11, 61–71. Flor, G., 1986. Sedimentología de una duna lingüiforme en la playa de Xagó (Asturias). Actas del IX Congreso Nacional de Sedimentología, Universidad de Salamanca 1, 317–328. Flor, G., Flor-Blanco, G., 2005. An introduction to the erosion and sedimentation problems in the coastal regions of Asturias and Cantabria (NW Spain) and its implications on environmental management. J. Coastal Res. 49, 58–63. Flor-Blanco, G., Flor, G., Pando, L., 2013. Evolution of the Salinas-El Espartal and Xagó beach/dune systems in north-western Spain over recent decades: evidence for responses to natural processes and anthropogenic interventions. Geo-Mar. Lett. 33 (2–3), 143–157. Fonseca, E.M., Baptista Neto, J.A., Pereira, M.P.S., Silva, C.G., Arantes Junior, J.D., 2014. Study of pollutant distribution in the Guaxindiba Estuarine System – SE Brazil. Mar. Pollut. Bull. 82, 45. Gallego, J.R., Ordóñez, A., Loredo, J., 2002. Investigation of trace element sources from an industrialized area (Avilés, northern Spain) using multivariate statistical methods. Environ. Int. 27, 589–596. Gallego, J.R., Ortiz, J.E., Sierra, C., Torres, T., Llamas, J.F., 2013. Multivariate study of trace element distribution in the geological record of Roñanzas Peat Bog (Asturias, N. Spain). Palaeoenvironmental evolution and human activities over the last 8000 cal yr BP. Sci. Total Environ. 454–455, 16–29. Herut, B., Sandler, A., 2006. Normalization methods for pollutants in marine sediments: review and recommendations for the Mediterranean, Report H18/ 2006. Israel Oceanogr. Limnol. Res..

Hooda, P.S. (Ed.), 2010. Assessing bioavailability of soil trace elements. Trace Elements in Soils. Springer, Berlin. Jabro, J.D., Stevens, W.B., Evans, R.G., 2006. Spatial relationships among soil physical properties in a grass-alfalfa hay field. Soil Sci. 171, 719–727. Julivert, M., Truyols, J., Marcos, A., Arboleya, M.L., 1973. MAGNA 50 (2ª Serie). Hoja 13 –Avilés. Instituto Geológico y Minero de España. Madrid. Kennish, M.J., 2002. Environmental threats and environmental future of estuaries. Environ. Conserv. 29, 78–107. Kennish, M.J., 2008. Environmental future of estuaries. In: Polunin, B. (Ed.), Aquatic Ecosystems: Trends and Global Prospects. Cambridge University Press, Cambridge, UK. Lattin, J., Carroll, D., Green, P., 2003. Analyzing Multivariate Data. Duxbury Press, New York. Li, Y., Wang, Y., Gou, X., Su, Y., Wang, G., 2006. Risk assessment of heavy metals in soils and vegetables around non-ferrous metals mining and smelting sites, Baiyin, China. J. Environ. Sci. 6, 1124–1134. López Peláez, J., Flor, G., 2008. Evolución Ambiental del estuario de Avilés (1833– 2006). Trabajos de Geología 28, 119–135. Martínez-López, J., Llamas, J.F., De Miguel, E., Rey, J., Hidalgo, M.C., Sáez, A.J., 2008. Multivariate análisis of contamination in the mining district of Linares (Jaén, Spain). Appl. Geochem. 23, 2324–2336. Matheron, G., 1965. Les variables régionalisées et leur estimation: une application de la théorie des fonctions aléatoires aux sciences de lanature. Masson, Paris, France. Mattielli, N., Petit, J.C.J., Deboudt, K., Flament, P., Perdrix, E., Taillez, A., RimetzPlanchon, J., Weis, D., 2009. Zn isotope study of atmospheric emissions and dry depositions within a 5 km radius of a Pb–Zn refinery. Atmos. Environ. 43, 1265– 1272. R Development Core Team, 2012. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. http:// www.R-project.org/ (accessed 29.06.2014). Reimann, C., de Caritat, P., 2005. Distinguishing between natural and anthropogenic sources for elements in the environment: regional geochemical surveys versus enrichment factors. Environ. Sci. Technol. 337, 91–107. Saavedra, A., Ordóñez, C., Taboada, J., Giráldez, E., Sierra, C., 2013. Grade control in a quartz deposit using universal fuzzy kriging. Dyna 189, 61–69. Sierra, C., Gallego, J.R., Afif, E., Menéndez-Aguado, J.M., González-Coto, F., 2010. Analysis of soil washing effectiveness to remediate a brownfield polluted with pyrite ashes. J. Hazard. Mater. 180, 602–608. Sierra, C., Ordóñez, C., Gallego, J.R., 2013. Functional outlier detection in grain size distribution curves of detrital sediments: a tool for geochemical, sedimentological and coastal research studies. Sed. Geol. 297, 31–37. Sucharovà, J., Suchara, I., Hola, M., Marikova, S., Reimann, C., Boyd, R., Filzmoser, P., Englmaier, P., 2012. Top-/bottom-soil ratios and enrichment factors: what do they really show? Appl. Geochem. 27, 138–145. Tack, F.M.G., 2010. Trace Elements: General Soil Chemistry, Principles and Processes. In: Hooda, P.S. (Ed.), Trace Elements in Soils. John Wiley & Sons Ltd., Chichester, UK. Tessier, E., Garnier, C., Mullot, J.U., Lenoble, V., Arnaud, M., Raynaud, M., Mounier, S., 2011. Study of the spatial and historical distribution of sediment inorganic contamination in the Toulon bay (France). Mar. Pollut. Bull. 62, 2075–2086. Violante, A., 2013. Chapter three – elucidating mechanisms of competitive sorption at the mineral/water interface. In: Sparks, Donald L. (Ed.), Advances in Agronomy. Academic Press, pp. 111–176. Webster, R., Oliver, M.A., 2007. Geostatistics for Environmental Scientists, 2nd Ed. Wiley, Chichester. WHO, 2000. Draft guidelines for safe recreational water environments: coastal and freshwater, draft for consultation (EOS/DRAF1798.14), Geneva. Wu, G., Shang, J., Pan, L., Wang, Z., 2014. Heavy metals in surface sediments from nine estuaries along the coast of Bohai Bay, Northern China. Mar. Pollut. Bull. 82, 194–200. Yalcin, M.G., Narin, I., Soylak, M., 2008. Multivariate analysis of heavy metal contents of sediments from Gumusler creek, Nigde, Turkey. Environ. Geol. 54, 1155–1163. Yongming, H., Peixuan, D., Junji, C., Posmentier, E.S., 2006. Multivariate analysis of heavy metal contamination in urban dusts of Xi’an, Central China. Sci. Total Environ. 355, 176–186.