Temporal clustering of metals in a short sediment core of the Cascais Canyon ( Portuguese Margin )

1 Laboratório Nacional de Energia e Geologia, I.P., Unidade de Geologia Marinha, Estrada da Portela, Apartado 7586, 2721-866 Alfragide, Portugal. E-mail: mario.milhomens@ineti.pt 2 Geological Survey of Spain, Dept. of Geosciences Research and Prediction Global Change, c/ Ríos Rosas, 23, 28003-Madrid, Spain. 3 Instituto National de Recursos Biológicos, Instituto de Investigação das Pescas e do Mar (IPIMAR), Av. Brasília, 1449-006 Lisboa, Portugal. 4 Laboratório Nacional de Energia e Geologia, I.P., Laboratório de Análises Químicas, Azinhaga dos Lameiros, 1649-038 Lisboa, Portugal. 5 Royal Netherlands Institute for Sea Research, Department of Marine Chemistry and Geology, P.O. Box 59, 1790 AB Den Burg, Texel, The Netherlands.

Intense industrial activity around the Tagus estuary (the most urbanized area of Portugal) has accompanied population growth during the 20 th century, and is thus responsible for releasing large quantities of metals into the estuary (e.g.Figuères et al., 1985;Vale, 1990;Araújo et al., 1998;Canário et al., 2005;Vale et al., 2008).Part of these metals was exported from the estuary and deposited on the adjacent shelf and slope, particularly in fine-grained deposits of the Tagus prodelta (Paiva et al., 1997;Jouanneau et al., 1998;Queralt et al., 1999;Mil-Homens et al., 2009a;Mil-Homens et al., 2009b), and the Lisboa-Setúbal canyon system (Richter et al., 2009).The Cascais Canyon also emerges as an important target for assessing the transfer of contaminated shelf sediments to abyssal plains due to its proximity to these contaminated areas.The main goal of this study was to assess the temporal evolution and current level of contaminant enrichment for sediments of the middle Cascais Canyon.Cluster analysis (CA) of geochemical analytical data was performed to assess the history of deposition of major and selected trace elements (Cu, Cr, Hg, Li, Ni, Pb and Zn).This paper focuses only on one site within the Cascais Canyon, which appears to be representative of the transport of contaminated sediments through the canyon.A broader and promising study, in which a series of cores along the canyon are used to reconstruct the last 300 years of sediment transport and deposition, is in progress.

Sampling
Multi-core PE252-32 (8.36307ºN and 9.50690ºW; 2100 m water depth, 37 cm in length) was collected during cruise 64PE252 of the Dutch RV Pelagia.In the geomorphological context of the Cascais Canyon (Lastras et al., 2009), this multi-core was recovered at the thalweg of the main branch of the Cascais Canyon (Fig. 1), one of the shortest canyons on the western Portuguese Margin.It starts at around 175 m water depth in the shelf area adjacent to the Tagus prodelta, southwest of the Tagus estuary mouth, and extends down to the Tagus abyssal plain to water depths exceeding 4600 m.
We used one sub-core of multicore PE252-32 that was kept refrigerated at approximately 2ºC before being opened, described, photographed and x-rayed to characterize the sediment lithology and to identify its sedimentological and biological structures.This subcore was sampled using cut-off syringes or sliced at 0.5 or 1 cm intervals in the laboratory, and samples were freeze-dried.

Grain-size
Grain-size distributions were determined with a Coulter LS230 laser particle analyzer after sample disaggregation by ultra-sonication.These measurements were carried out for 12 sample intervals also used for 210 Pb determinations.relativamente baja, los resultados obtenidos ponen en evidencia la importancia del Cañón submarino de Cascais como vía de transporte de sedimentos contaminados depositados en el pro-delta del Tajo a las regiones profundas del margen de Portugal.

Core dating
The sediment core was dated through the vertical distribution of excess 210 Pb ( 210 Pb xs ), which is incorporated in the accumulating sediment from atmospheric fallout by decay of 226 Ra in the water column and indirectly by rivers (Cundy and Croudace, 1995;Boer et al., 2006). 210Pb was determined by alpha-spectrometry with Canberra Passivated Implanted Planar Silicon (PIPS) detectors using the grand-daughter isotope 210 Po (with a half-life of 138.4 days) after sediment samples were spiked with 209 Po and leached with HCl, following the procedure described by Boer et al. (2006). 210Pb has a half-life of 22.3 yrs, thus being very useful for reconstructing the last 100-150 years of sediment deposition, which is ca. 5 times the half-life of the isotope (Valette-Silver, 1993).Sediment age was determined on the basis of the downcore 210 Pb profile, fitted with the CF/CS (constant flux/constant sedimentation) model (Boer et al., 2006), in which a constant accumulation rate and negligible bioturbation are assumed.

Major and trace elements
Major elements (presented here as oxides SiO 2 , Al 2 O 3 , Fe 2 O 3 , MgO, CaO, Na 2 O, K 2 O, TiO 2 and MnO) were determined on fusion beads by WD-XRFS, using a sequential wavelength dispersive spectrometer (model AXIOS), according to an in-house LNEG-LAQ methodology.After being oven-dried for two hours at 105ºC, samples and standards were prepared as glass beads as follows: 1000 mg of sample were mixed with 9.0 g of lithium tetraborate flux (Li 2 B 4 O 7 ) and 20 mg of ammonium iodide (NH 4 I).This mixture was fused in Pt95-Au5 crucibles at 1150 °C for 20 minutes and then poured into Pt-Au dishes.The calibration curves were constructed using a set of 16 standards (National Research Centre for CRM's, Beijing, China) with a range of analytes and matrices similar to those of the samples.To correct for matrix interferences, theoretical a correction factors were applied.According to the laboratory quality control procedures, the calibrations were validated by daily analysis of certified reference material (CRM).
Total trace elements (Cu, Cr, Ni, Li, Pb and Zn) were measured by flame atomic absorption spectrometry (FAAS) after total acid digestion of sediment samples following the procedure described by Alves et al. (2009).An amount of 0.25 g of freeze-dried and ground sample was mixed with 2.5 ml of HNO 3 (HNO 3 Suprapur 65%), 2.5 ml of HClO 4 (HClO 4 Suprapur 70%) and 5 ml of HF (HF Suprapur 40%) on a Teflon beaker with lid.The mixture was heated for 15 minutes on an electric hotplate, followed by evaporation until dryness.After that, 5 ml of hot Milli-Q water (>18 mOhm), 5 ml of HNO 3 and 2.5 ml of HClO 4 were added to the solution and it was heated again on the electric hotplate and evaporated until dryness.The residue was dissolved with 5 ml of hot Milli-Q water (>18 mOhm) and 5 ml of HNO 3 for solubilization of salts.This procedure was repeated (usually twice) until the solutions were clear and transparent.After cooling down, the solution was transferred to a 50 ml volumetric flask and diluted to this volume with Milli-Q water.
Throughout the entire procedure blank reagent and CRM were regularly measured.Analytical precision expressed as relative standard deviation (RSD) of 8 replicates of MESS-3 (NRCC -Canada) was nominally lower than 10% (P<0.05).Accuracies determined by comparing results on CRM MESS-3 with certified values were within 7%.All FAAS measurements were performed on a SOLAAR 969 AA Thermo Elemental spectrometer equipped with a deuterium lamp background correction system, using hollow-cathode lamps from the Thermo Electron Corporation (Cambridge, UK) as radiation source at each element.
Total Hg was measured by atomic absorption spectrometry using a Leco AMA-254 silicon UV diode detector, after pyrolysis of each sample (approximately 0.05 g) in a combustion tube at 750°C under an oxygenrich atmosphere, and collection on a gold amalgamator (Costley et al., 2000).Precision of Hg measurements, expressed as RSD of 4 replicate samples of the CRM MESS-3, was less than 4% (P<0.05).Based on the mean values of the CRM, the results for Hg (4 replicates) indicate an accuracy of 106%.

Statistical Analysis
We used the multivariate statistical method of cluster analysis (CA) following Ward's criterion (Davis, 2002) in order to relate analyses of variance.This allowed us to evaluate the similarity between the two variables (major and trace elements in a total of 16) and samples (corresponding to 19 depths), classifying them into relatively homogeneous meaningful clusters.CA was applied to identify both element associations and major changes in chemical composition through time (core depth).CA was performed using the SPSS software (version 13 for Windows) and was carried out on the standardized datasets (whose mean and standard deviation were set to zero and one, respectively -Z score) to minimize the effect by the difference in measured units, or variance, and to render the data dimensionless.
In addition we also used Spearman's correlations (r) in order to identify significant statistical correlations among the measured element variables (Davis, 2002).

Enrichment Factors (EF)
The impact of anthropogenic-derived metals was estimated based on the determination of enrichment factors (EF).EF were calculated as the ratio between the normalized sample and background (metal/conservative element) to reduce grain-size effects [EF = (metal/ conservative element) sample / (metal/ element conservative) background ].We considered the average concentrations from the three bottom samples of the core studied as local background values, because these samples comprise the local or regional variability.According to the 210 Pb chronology, these samples reflect pre-industrial deposition and thus represent the metal concentration free of contamination.

Grain-size and chemical composition
At 2 cm (core depth), core PE252-32 shows a thin (ca. 1 cm), dark-brown oxidized layer (10YR ¾, Munsell soil colour chart).Underneath this layer, a gradual change in colour is observed down-core from dark olive (5Y 4/3, Munsell soil colour chart) to olive-black (5Y 3.5/2, Munsell soil colour chart).No visible trace of bioturbation was observed.The sediments were mainly composed of the fine-grained size fraction (<63 mm), ranging between 89% and 99%.Fine contents are constant in the top 10 cm, with values of ca.42% of clay and 56% of silt, decreasing towards the bottom down to 32% and 53%, respectively (Fig. 2).Sand content is very low (<4%), except for the bottom two measured samples (depths 29-30 and 34-35 cm) with sand percentages of 8% and 14%, respectively.
Silica (SiO 2 ) is the most abundant oxide, ranging from 38% to 46%.Except for the top 10 cm and bottom 7 cm, SiO 2 and CaO show similar general trends in depth, decreasing gradually from bottom to top.This trend is op-posed to alumina (Al 2 O 3 ).Above 10 cm core depth, SiO 2 and Al 2 O 3 both show a slight decrease (n= 10, P=0.08; r=0.58) towards the surface, unlike CaO which increases slightly (Fig. 3).Thus, below 10 cm depth, the lithology is mainly siliceous-aluminosilicated, with increasing CaO towards the bottom.In the bottom 6 cm of the core, high contents of CaO, SiO 2 and sand (Figs. 2 and 3) suggest the presence of abundant quartz and carbonate sand.The higher silica content reflects dominant quartz grains in the bottom sample.Alumina, K 2 O, MgO, TiO 2 , Cr and Li profiles show similar down-core variations.Manganese oxide (MnO), Fe 2 O 3 , and Cu (Fig. 3) profiles exhibit a subsurface peak (at about 2 cm) coinciding with the oxidized layer.Below this level, MnO shows very homogeneous contents, while Fe 2 O 3 is comparable with Al 2 O 3 .Zinc, Cu, Ni, Pb, Hg and Na 2 O show a very similar down-core behaviour.This trend is marked by a continuous increase towards the present day, followed by slightly decreasing concentrations in the top 2 cm.Clearly, there is a significant positive relationship among these element concentrations (Table 1).This set of elements (Cu, Hg, Ni, Pb, Zn and Na 2 O) is also characterized by high and significant values of Spearman's correlation coefficients (r) with MnO and Fe 2 O 3 .However, considering only the top 10 samples, the Spearman's correlation coefficients (r) were not significant between MnO, Hg and Pb (n=10; P=0.07 60), suggesting a detrital origin.The absence of any relationship between Hg, Pb and Zn and the elements normally associated with aluminosilicates suggests that core samples above 10 cm depth have enrichments unrelated to the fine-grained sediment components (Fig. 4).Copper has a similar behaviour to Hg, Pb and Zn, showing a small variation between the two sets of samples (Fig. 4).In the case of Ni and Cr, this differentiation is not so clear suggesting that both elements have a detrital origin (Fig. 4).

Age model
The sediment accumulation rate of 210 Pb, in g cm -2 yr -1 (Fig. 5) was calculated by transforming the depth scale into cumulative dry mass (CMD) using the dry bulk density in order to correct compactation (Robbins, 1978; Appleby and Oldfield, 1992).The 210 Pb profile showed a regular exponential decrease towards the bottom.Two samples from the nearest levels, where a fragment of synthetic rock (sinter probably dumped from a ship) was found, were not considered for defining sediment accumulation rates.Nevertheless, the sinter fragment in the core may actually support the age model, because it seems to correspond to a period (first half of the 20 th century) when steam ships may have passed over the core site.The obtained sediment accumulation was 0.061 g cm -2 yr -1 , equivalent to an average linear sedimentation rate of 0.11 cm yr -1 .

Statistical analysis
The cluster analysis (hierarchical clustering, Ward's method, squared Euclidean distance; Davis, 2002) of 16 variables is illustrated by the dendrogram in Figure 6.Distance on the horizontal axis represents the level of association between groups of variables (Davis, 2002).Low distance values indicate greatest similarity between variables (Davis, 2002).The variables fall into four main clusters (A, B, C and D) that seem to have a geochemical implication in terms of sediment composition and element sources.The high proximity between clusters A and B suggests that their components represent the fine fraction of sediments.While cluster B contains elements (Al, Ti and Li) commonly associated with the detrital fine fraction of sediments, cluster A also includes elements typically associated with natural elements connected with diagenesis (Mn, Fe) and other elements (Pb, Zn, Hg, Cu, Ni) that may also have a natural origin, or can be derived from diagenetic processes or from anthropogenic sources.This association suggests that Fe and Mn oxide-hydroxides are probably controlling (totally or partially) their distributions in sediments.This cluster can be separated into two subgroups (A 1 and A 2 ).Subgroup A 2 is formed only by Mn, which indicates that this element has a  In this subgroup, samples show a slight increase in Hg, Pb and Zn contents, concomitant with increasing concentrations of Al and Li, elements associated with the fine-grained sediment components.

Natural vs. anthropogenic variability
To compensate for grain-size variations, total metal contents are commonly normalized using Al or Li as a proxy for the fine-grained fraction (Loring and Rantala, 1992).Both elements are present in the lattice of fine-grained aluminosilicate minerals and are not substantially influenced by anthropogenic activities (Windom et al., 1989;Loring and Rantala, 1992).Prior to 1900 AD (underneath the 10 cm core depth), despite the similar trends of Li and Al and the fine fraction (Figs. 2 and 3) and metals (Cu, Cr, Ni, Hg, Pb and Zn), Al shows higher regression coefficients (r 2 ) with metals than Li (Fig. 4).Thus, Al seems to be a better geochemical normalizer for explaining the natural variability in metal concentrations.The regular down-core normalized profile of Cr suggests a detrital origin for this element (Fig. 8), agreeing with the results of cluster analysis (Fig. 6).The normalized profiles of Cu, Zn, Pb and Hg exhibit approximately constant values prior to AD 1900, suggesting that slight changes in metal concentrations (Fig. 3) are grain-size dependent.After this period, normalized profiles of Hg, Pb and Zn show a gradual increase up to AD 1980 (2 cm core depth), followed by a slight decrease towards the present-day (surface sample) (Fig. 8).These variations in normalized metal contents, unrelated to grain-size variation, probably reflect predominantly anthropogenic contributions derived from the intense industrial and urban activities in the Tagus estuary.
Although the normalized profile of Cu for the last 100 years (top 10 centimetres) displays an increasing trend, a marked peak stands out at AD 1981 (3.5 cm), below or coincident with the peaks of Mn (1.5 cm; Fig. 3) and Fe (2.5 -3.5 cm; Fig. 3).The existence of metal peaks such as that of Cu at, or near, the redox boundary, and to a lesser extent that of Ni, suggests an association with other mineral phases (such as oxide/ hydroxides of Fe and Mn) and hence a connection with diagenetic processes and metal redistribution (Thomson et al., 1996(Thomson et al., , 1998)).This is also indicated by the Cu and Ni clustering with Fe (Fig. 6, subgroup A 1.2 ) and with Mn (Fig. 6, subgroup A 2 ).

Estimation of anthropogenic enrichments
Anthropogenic enrichments were only estimated for Hg, Pb and Zn following the output grouping obtained by the cluster analysis.They were estimated through the determination of EF.
In this study, background values were obtained from the average contents of the three deeper sam- ples that are considered as representing the local pre-industrial values.Based on the 210 Pb chronology, and using the calculated sediment accumulation rate of 0.061 g cm -2 yr -1 , the three bottom samples of the core fall within the interval between AD 1728 and AD 1674.As the underlying assumption of constant sediment accumulation rates throughout the record cannot be tested, we only show EF values of Hg, Pb and Zn for samples younger than AD 1800 (Figure 9).The EF for the three elements shows quite constant values until the beginning of the 20 th century.
Thereafter we observe an increasing trend until the middle of the 1980s followed by a slight decrease in metal enrichments (Fig. 9).Mercury reaches the highest level of enrichment (EF Hg(max) =5), followed by Pb and Zn.Enrichments are lower than in finegrained deposit of the Tagus prodelta (EF Hg(max) =20; EF Pb(max) =7; EF Zn(max) =4) (Mil-Homens et al., 2009a), reflecting both the greater distance to the source of contaminated sediments (the Tagus estuary) and dilution with marine uncontaminated sediment.On the other hand, the same order of enrichment was found in a depocentre located to the north of the Nazaré Canyon (Mil-Homens et al., 2008).

CONCLUSIONS
The temporal evolution of anthropogenic metal enrichment trends between the 1900s and the 1980s is comparable between the Tagus prodelta and site PE252-32 in the Cascais Canyon.Though sediment accumulation rate in the canyon is approximately 3 times lower than on the Tagus prodelta, the similarity of anthropogenic enrichment trends reveals an efficient transfer of contaminated sediments from the shelf to 2100 m water depth within the Cascais Canyon.The results prove that this sector of the canyon acts as a sink of contaminated sediments, extending the negative effects of contamination to deep regions of the Portuguese Margin.

Fig. 1 .
Fig. 1. -Location of multi-core site in the Cascais Canyon.Shading denotes the Tagus prodelta, characterized by sediment with more than 90% mud (silt + clay) and less than 30% of CaCO 3 (data extracted from the Chart of Surface Sediments of the Portuguese shelf [sheet number 5], published by the Portuguese Hydrographic Institute [2005]).

Fig. 4 .
Fig. 4. -Relationship between metals (Hg, Pb, Zn, Cu, Ni and Cr) and Al 2 O 3 , and Li in down-core samples of core PE252-32.Sediment samples from the top 10 cm are represented by filled symbols.

Fig. 5 .
Fig. 5. -Vertical distribution of 210 Pb in core PE252-32 plotted against cumulative mass depth (CMD).Crosses represent 210 Pb values not considered for determination of sediment accumulation rates.

Fig. 7 .
Fig. 7. -Hierarchical dendrogram (Ward's method of linkage, squared Euclidean distance) for 19 down-core samples using as clustering variables the subgroup composed of Hg, Pb and Zn, interpreted as reflecting anthropogenic contributions.

Fig. 8 .
Fig. 8. -Vertical profiles of metal (Cr, Cu, Hg, Ni, Pb and Zn) to Al ratios in multi-core PE252-32.Chronology is derived from 210 Pb.The extrapolation of 210 Pb chronology to ages of 1650 AD is done assuming constant sediment accumulation rates.

Fig. 9 .
Fig. 9. -Historical trends in normalized enrichment factors for Hg, Pb and Zn.EFs are calculated by dividing the Al-normalized metal contents by the Al-normalized metal background for each core.Samples older than 1800 AD are not represented.