One trial pit was opened near the Northwest corner of the site using a mechanical excavator and an additional bulk sample was collected at a depth of 40-50 cm.
ASSESSMENT OF HETEROGENEITY IN HEAVY METAL CONTAMINATION IN SOIL: AN EXAMPLE FROM A HISTORICAL MINING AND SMELTING WASTE DEPOSIT IN LAVRION, GREECE A.ARGYRAKI* AND V. ZOTIADIS** *Department of Economic Geology and Geochemistry, University of Athens, Panepistimiopolis Zografou, 157 84 Athens, Greece ** Edafomichaniki S.A., Em. Papadaki 19, N. Iraklio 141 21, Athens, Greece
SUMMARY: The aim of this study is to illustrate how routinely collected data in contaminated land site investigations may be utilized for characterizing the heterogeneity of contamination within a sampling target in three dimensions. To reach this aim, data from a geochemicalgeotechnical investigation at an urban site within the historical mining area of Lavrion, Greece were used as an example. The statistical technique of one-way analysis of variance (ANOVA) was applied to test the hypothesis that lateral variability of contamination is statistically significantly different of vertical variability with respect to heavy metal concentrations in soil, thus verifying that the used sampling protocol is fit for the purpose of delineating vertical and horizontal spatial distribution of contamination within the sampling target. Overall, the site heterogeneity with respect to potentially toxic elements concentration was found to vary for different elements not only as a result of their origin but also according to their different behavior under specific geochemical conditions. Vertical variability of elemental concentrations was found to be significantly higher than horizontal variability. This could be attributed to periodic deposition of waste material and downwards migration of the elements originating from waste. Vertical variability is influenced by buffering reactions, acting as a limiting factor of maximum depth of element leaching within the site.
1. INTRODUCTION Remediation and development of contaminated land demands thorough understanding of the nature and extent of soil contamination, i.e. determination of key-contaminants, their mobility and bioavailability and delineation of vertical and horizontal spatial distribution of contamination within the sampling target. Environmental site assessment involving a preliminary desk study to determine site history and on site sampling and performance of laboratory tests followed by data
interpretation are common practice in brownfields (US-EPA 2001). Decisions on sample number and sampling locations have an important role in determining the effectiveness of sampling and analysis as well as the level of uncertainty in the produced data. In this respect the assessment of the heterogeneity of contamination within the sampling target is of paramount importance. The aim of this study is to illustrate how routinely collected data in contaminated land site investigations may be utilized for characterizing the heterogeneity of contamination within a sampling target in three dimensions. To reach this aim, data from a site investigation at an urban site within the historical mining area of Lavrion, Greece are used as an example. The statistical technique of one-way analysis of variance (ANOVA) is applied to test the hypothesis that lateral variability of contamination is significantly statistically different of vertical variability with respect to heavy metal concentrations in soil, thus verifying that the used sampling protocol is fit for the purpose of delineating vertical and horizontal spatial distribution of contamination within the sampling target.
2. METHODOLOGY 2.1 Study area The study area is located in the town of Lavrion, Greece where waste material originating from sulfide ore mining, milling and metallurgical operations spanning over a period of at least 3000 years has been deposited at several spots (Fig. 1). The environmental site assessment was performed in parallel with a geotechnical study within an area of 18 acres that belongs to the local authorities and is located within an area of deposited gravity separation tailings known as “Telmata”. The topography is flat and the area lies at the mouth of the gentle sloping Lavrion mainland. The actual location of the site indicates that it has been receiving material from metal processing activity areas uphill, either on purpose or through the natural movement of the overburden due to gravity. The quantity and quality of the “Telmata” spoil has been previously described in detail (Xenidis et al. 2003). The predominant minerals in the spoil are quartz and calcite. Muscovite, chlorite, fluorite albite and dolomite are also present, while the most significant metal bearing phases are hydrous ferric oxide, phosgenite, smithsonite and mixed Pb-Mn, Pb-Fe and Pb-Zn-Mn oxides. The texture of the material is very fine, with over 75% of particles below 63 μm. Approximately 80% of the total Pb, Zn, Cd and As is contained in the fine fraction. The material was characterized as non-acid-generating but toxic because of the high solubility of the contained Pb and Cd. 2.2 Sampling and analysis A total of 28 overburden samples were collected from the site. Seven boreholes reaching a maximum depth of 25 m were drilled inside the field at locations shown in Figure 1. From each borehole, core samples were collected from 4 horizons (A, B, C, D) at depths of A = 0.50 m, B = 2 m, C = 3.5 m and D = 5 m by mixing the material from 10 cm above and below each horizon to produce the bulk sample. One trial pit was opened near the Northwest corner of the site using a mechanical excavator and an additional bulk sample was collected at a depth of 40-50 cm. Sampling depths were selected on the basis of soil material differentiation. According to geotechnical data up to the depth of 3.5 m the sampled material is a mixture of alluvial soil and historical mining and metallurgical waste (carbonate-rich tailings and slag).
CRETE 2012
2
Figure 1. Left: Map of Lavrion urban area showing the study site and distribution of mining and metallurgical waste comprising carbonaceous tailings of flotation residuals, slag and sulphidic tailings. Modified from Xenidis et al. (2003). Right: Map of the project area showing boreholes (T8-T15) and trial pit (PIT 1) locations.
The deepest sample was collected at the depth of 5 m from a silty – clayey sand layer with gravels, representing weathered Quaternary deposits. Below this depth, Quaternary deposits of alternations of conglomerates and marls continue for 3-10 m until they meet the alpine geological basement of metamorphic schist. The collected soil samples were air dried at room temperature, disaggregated with a pestle and mortar, homogenized and sieved to a grain size of 0.425mm in the laboratory. Chemical analysis was performed for the elements Cu, Pb, Zn, Ag, Ni, Cr, Mn, Fe, As, Cd, Sb and Ba using ICPMS at the ACME Analytical Laboratories Ltd., Canada after digestion with a mixture of nitric and hydrochloric acids. Analytical precision was assessed by using analytical duplicates and was reported to be better than 3%. Analytical bias was estimated by analyzing a house reference material, i.e. a matrix matched soil with known elemental concentrations. No statistically significant bias (checked with a t-test at the 95% level) was observed on the analytical results of replicated measurements of this material. Soil pH was measured after mixing with distilled water in a ratio 1:2. One water sample was collected from the installed piezometer at borehole T8 using a weighted standard Teflon bailer of 1l capacity. This sample was analyzed for the elements Pb, Zn, As, Cu, Fe, Mn, Cd and Ni. Other parameters determined in the water sample included pH, anion concentrations of SO42-, Cl- and cation concentrations of Mg2+ and NH4+.
2.3 Statistical basis and application of ANOVA Analysis of variance is a broad class of statistical techniques for identifying and measuring the various sources of variation within a collection of data (Davies, 1986). It has been applied for the estimation of sampling uncertainty in contaminated land investigations (Ramsey and Ellison, 2007; Argyraki, 2010). ANOVA can be used for identifying relationships between criterion variables and predictor variables, whether those predictor variables are quantitative or qualitative in nature. Thus it is more general in scope than regression analysis and can be used in situations when the predictor variable is composed of values which differ in kind, rather than in quantity CRETE 2012
3
e.g. layers of waste material with different geotechnical and geochemical characteristics. In this study, in order to assess the heterogeneity of the contamination within the sampling target the horizontal variability of elemental concentrations was compared to the vertical variability. First, the elemental concentration means of the observations within each sampled horizon were calculated. The residuals represent the variation within each depth horizon. The means of each horizon were averaged to obtain a grand mean to understand the effect of sampling depth and finally the variation within the sampled horizons (horizontal variability) was compared to the variation across horizons (vertical variability).
3. RESULTS AND DISCUSSION 3.1 Total elemental concentrations The descriptive statistics of the total analyses of the core samples are presented in Table 1.The recorded concentrations of potentially toxic elements at the site signify that the whole field is heavily contaminated. Concentrations of Pb, Zn and As reaching maxima of 5.2%, 5.3% and 0.7% respectively in the analyzed samples, agree with concentrations reported by previous research on samples of carbonate-rich, oxidic tailings from the area (Xenidis et al. 2003). The high values of coefficients of variation for elemental concentration of As, Cd, Zn, Pb and Cu, ranging from 99% for As to 66% for Cu, indicate high variability of concentrations within the site for all elements associated with the metallurgical waste. This observation demonstrates the heterogeneous nature of sampled material but does not provide any information on the comparative degree of heterogeneity in the lateral and vertical dimensions. Principal component factor analysis with a varimax rotation was also applied to the data in order to create factors, each representing a cluster of interrelated variables within the data set. The rotated factor loadings, communalities and the proportions of the variance explained by 3 factors are presented in Table 2. The factor analysis explained 87.1% of the total variance in the data. Factor 1 comprises Pb, Zn, Ag and Cd which show high positive loadings with smaller contribution from Cu, Sb and Ba. Factor 2 contains Mn and Fe with similar high positive loadings. Arsenic also presents a contribution to this factor. Factor 3 is dominated by Ni and Cr. Table 1 – Summary statistical data of the drillcore samples (n =28) total chemical analysis Mean Median SDa CVb Min Max -1 -1 -1 -1 Element (μg g ) (μg g ) (μg g ) (%) (μg g ) (μg g-1) Cu 265 285 174 66 29 571 Pb 23880 26963 17160 72 830 52220 Zn 24680 28157 17890 73 980 53779 Ag 56 62 45 80 1 160 Ni 180 154.5 78 43 49 350 Cr 134 129.5 38 28 51 200 Mn 2740 2583 1230 45 930 5600 Fe 50000 49000 14000 28 22600 78400 As 2190 1660 2160 99 46 7200 Cd 116 130 88 76 4 279 Sb 351 410 250 71 9 680 Ba 222 240 101 46 54 380 a Standard Deviation b Coefficient of Variation CRETE 2012
4
Table 2 - Varimax component loadings of three factors, communality and percentage of variance explained for twelve soil variables. Variable Factor1 Factor2 Factor3 Communality Cu 0.726 0.570 -0.246 0.913 Pb 0.974 0.160 -0.043 0.976 Zn 0.969 0.189 -0.079 0.981 Ag 0.943 0.238 -0.045 0.947 Ni -0.166 -0.094 0.923 0.888 Cr -0.008 0.123 0.924 0.869 Mn 0.120 0.850 0.185 0.771 Fe 0.356 0.851 -0.021 0.852 As 0.336 0.684 -0.527 0.858 Cd 0.929 0.180 -0.134 0.914 Sb 0.795 0.505 -0.213 0.934 Ba 0.582 0.459 -0.054 0.553 % Variance 0.452 0.240 0.179 0.871 The variability of elemental concentrations in the first factor, accounting for 42% of the total variance, appears to be controlled by the primary sulphide mineralization, and thus the composition of the processed ore, as all elements present in this factor are characteristic of the chemistry of ore smelted at Lavrion. The second factor, accounting for 24% of the total variance relates to the less important Fe-Mn mineralization of the area and the oxide phases of the tailing deposits. Arsenic is also grouped in this factor, showing the affinity of this element with Fe and Mn oxides. Finally, the third factor explaining 18% of the total variance contains elements which are enriched in the basic and ultrabasic rocks within the metamorphic schist system of the local geology. 3.2 Vertical and horizontal contaminant distribution One-way ANOVA was applied to the total data for testing the hypothesis that variability of contamination within each sampled horizon is significantly statistically different of variability between horizons with respect to heavy metal concentrations in soil. The results of this statistical technique showed that vertical variability is significantly higher than horizontal variability within each sampled horizon at 95% confidence level the eight elements. The ANOVA results are given in Table 3 and are presented graphically on side by side box-plots for the four sampled horizons (Fig. 2). In this respect the used sampling protocol was judged fit for the purpose of delineating contamination at the site in three dimensions. In terms of variability of contamination within each sampled horizon, horizon B not only is the most contaminated but also relatively homogeneous in respect with Pb concentration (Fig. 2(a)). The same pattern is observed for the elements Cu, Zn and Cd. This observation coupled by visual inspection of the drillcore material signified the existence of deposited metallurgical waste at least until the depth of 2 m. Horizon C is also demonstrating a high mean value of Pb, Zn and Cd but lower than horizon B and also displays the highest range of contaminant concentrations within the 7 sampled locations. The large heterogeneity of this horizon can be attributed to two distinct processes: (a) direct elemental enrichment of the material due to the existence of localized slag horizons, e.g. in boreholes T13 and T14, giving rise to high concentrations and (b) downwards leaching of the elements originating from the surface tailings deposits. Long term downwards migration of elements in polluted sites is affected by several physiochemical factors (Maskall et al. 1996) with most significant being the pH and Eh. Soil pH measurements at this
CRETE 2012
5
Table 3- ANOVA results for eight elements, showing statistically significant differences (p< 0.05) among population means at different depth horizons. (DF = degrees of freedom, SS= sum of squares, MS= mean square). Variable Source DF SS MS F p 6 6 Cu Factor 3 4.056×10 1.352×10 7.89 0.001 Error 24 4.110×106 0.171×106 Total 27 8.167×106 Pb Factor 3 3.379×109 1.126×109 5.91 0.004 9 9 Error 24 4.572×10 0.191×10 Total 27 7.951×109 Zn Factor 3 3.586×109 1.195×109 5.67 0.004 Error 24 5.058×109 2.107×109 Total 27 8.644×109 Ni Factor 3 88658 29553 9.25 0.000 Error 24 76714 3196 Total 27 165372 Cr Factor 3 11066 3689 3.22 0.041 Error 24 27491 1145 Total 27 38556 As Factor 3 6.594×107 2.198×107 8.78 0.000 7 7 Error 24 6.008×10 0.250×10 Total 27 12.603×107 Cd Factor 3 78189 26063 4.74 0.010 Error 24 132063 5503 Total 27 210252 Fe Factor 3 16.81 5.60 3.50 0.031 Error 24 38.45 1.60 Total 27 55.26 site were in the range of 5.8 – 6.1 for all the samples and are explained by the mineralogical composition of the material which predominantly contains quartz and calcite (Xenidis et al. 2003). Slightly acidic conditions may contribute to the mobilization of Pb and other metals depending on the buffering capacity of the material (Gee et al. 2001). However, the carbonaceous composition of the waste in the upper horizons creates buffer conditions and acts as a limiting factor that controls the mobility and downwards migration of the elements. Thus, the lower concentration of leached elements appears in horizon C. Evidence supporting the argument that the lowest elemental concentrations at this depth are influenced by downwards leaching is also provided by the large drop of As content in horizon C (Fig. 2 (b)) and a significant rise in Cr and Ni concentration (Fig.2 (c)). Arsenic, which is contained in the metallurgical waste that lies on the surface, has a strong affinity to Fe, Mn oxides that prevent its leaching in deeper layers under oxidizing conditions (Smedley and Kinniburgh 2002). Concentration of Cr, which originates from the local rock basement, rises reflecting the composition of the underlying metamorphic schist that hosts basic and ultrabasic rock. Elemental concentrations of samples at the depth of 5 m (Horizon D) approach the natural geochemical background of the area. Samples from this horizon are enriched in Ni and Cr but concentrations of most potentially toxic elements are falling dramatically compared to the upper horizons. However, there is still evidence of contaminants leaching through the vertical profile demonstrated by relatively high elemental concentrations of borehole T9 at this horizon.
CRETE 2012
6
Figure 3. Boxplots comparing concentrations of selected elements at the four sampled horizons. Solid circles indicate mean values and asterisks indicate outliers.
4. CONCLUSIONS One-way analysis of variance provided a useful statistical tool for demonstrating that within the tailings deposit, vertical variability of elemental concentrations is significantly higher than horizontal variability. This could be attributed to periodic deposition of waste material of varying quality during the operation of the metal processing plants. Vertical stratification has also occurred as a result of downwards leaching of elements through time. Lead, Zn, Cd and As show decreasing concentrations with increasing depth while Ni and Cr appear with maximum concentrations in the deepest layer reflecting their geogenic origin. Downwards migration of the elements originating from waste is influenced by buffering reactions, acting as a limiting factor of maximum depth of element leaching within the site. Overall, site heterogeneity with respect to potentially toxic elements concentration has been shown to vary for different elements not only as a result of their origin but also according to their different behavior under specific geochemical conditions. In light of this information pole foundation with minimum excavation and overburden disturbance was proposed for construction of a planned building on the site. This work was performed according to routine site assessment procedures and its findings illustrate how routinely collected data may be utilized for characterizing the heterogeneity of contamination within a sampling target in three dimensions. The used example demonstrated the use of one-way ANOVA for judging the fitness of the used sampling protocol for the purpose of delineating the contamination horizontally as well as vertically, enabling informed decision making for planning the development of the site.
CRETE 2012
7
REFERENCES Argyraki, A. (2010) Analysis of Variance to estimate sampling uncertainty. In: Approaches in Material Sampling, B. Geelhoed (Ed.), IOS Press, Amsterdam, pp. 61-88. Davis, J.C. (1986) Statistics and Data Analysis in Geology, J Wiley and Sons, New York. Gee, C., Ramsey, M.H. and Thornton, I. (2001) Buffering from secondary minerals as a migration limiting factor in lead polluted soils at historical smelting sites. Applied Geochemistry, 16, 1193-1199. Maskall, J., Whitehead, K., Gee, C., Thornton, I. (1996) Long-term migration of metals at historical smelting sites, Applied Geochemistry, 11, 43-51. Ramsey, M. H. and Elisson, S.L.R. (eds.) (2007) Eurachem/EUROLAB/CITAC/Nordtest/AMC Guide: Measurement uncertainty arising from sampling: a guide to methods and approaches, Eurachem. Smedley, P.L., Kinniburgh , D.G. (2002) A review of the source and distribution of arsenic in natural waters. Applied Geochemistry, 17, 517-568. US-EPA (2001) Environmental Investigations Standard Operating Procedures and Quality Assurance Manual. Region 4, Science and Ecosystem Support Division. Athens, GA. Xenidis, A., Papassiopi, N., Komnitsas, K. (2003) Carbonate-rich mining tailings in Lavrion: risk assessment and proposed rehabilitation schemes. Advances in Environmental Research 7, 479-494.
CRETE 2012
8
