Probabilistic Ecological Risk Assessment of 1,2,4-Trichlorobenzene at

Mar 24, 2005 - Interuniversity Centre of Environmental Monitoring Research (CIMA), University of Genoa, Via Cadorna, 7, 17100 Savona, Italy, and Labor...
0 downloads 9 Views 526KB Size
Download PDF
Environ. Sci. Technol. 2005, 39, 2920-2926

Probabilistic Ecological Risk Assessment of 1,2,4-Trichlorobenzene at a Former Industrial Contaminated Site M A R C E L L O Z O L E Z Z I , * ,† CLAUDIA CATTANEO,† AND J O S EÄ V . T A R A Z O N A ‡ Interuniversity Centre of Environmental Monitoring Research (CIMA), University of Genoa, Via Cadorna, 7, 17100 Savona, Italy, and Laboratory for Ecotoxicology, Department of the Environment, Spanish National Institute for Agriculture and Food Research and Technology (INIA), Ctra de La Corun ˜ a km 7, 28040 Madrid, Spain

Measured concentrations of 1,2,4-trichlorobenzene (1,2,4-TCB) in soil and groundwater detected in an industrial contaminated site were used to test several probabilistic options for refining site-specific ecological risks assessment, ranging from comparison of single effects and exposure values through comparison of probabilistic distributions for exposure and effects to the use of distribution based quotients (DBQs) obtained through Monte Carlo simulations. The results of the deterministic approach, which suggest that risk exceeds a level of concern for soil organisms, were influenced mainly by the presence of hot spots reaching concentrations able to affect acutely a large proportion of species, while the large majority of the area presents 1,2,4TCB concentrations below those reported as toxic. Ground(pore)water concentrations were compared with aquatic ecotoxicity data in order to obtain an estimation of the potential risk for aquifers and streams in the adjacent area as well as for soil-dwelling organisms exposed via pore water. In this case, the risk is distributed over a large proportion of the site, while the local risk of hot spots was low, showing that risk characterization based exclusively on soil concentrations might be insufficient.

Introduction Risk assessment protocols are considered the best available tool for supporting, under scientific basis, decision-making processes on a wide range of areas, from economic development to environmental protection (1). In the past decade international institutions developed harmonized methodologies for the risk assessment of chemical pollution, and, at present, ecological risk assessment (ERA) can cover both generic and site-specific assessments. In the European Union, the Technical Guidance Document (TGD) provides a comprehensive method for assessing the generic risk of chemical substances to human health and the environment (2). Specific protocols for assessing certain groups such as pesticides, * Corresponding author phone: +39 019 23027220; fax: +39 019 862612; e-mail: [email protected]. † University of Genoa. ‡ INIA. 2920

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 9, 2005

biocides, or pharmaceuticals are also available (see the Scientific Steering Committee second report on the harmonization of risk assessment procedures for a comprehensive review of EU protocols) (1). Similar documents have been developed by the U.S. Environmental Protection Agency (3), which also provides site-specific risk assessment guidance, mostly under the Superfund activities. The estimations of risk used in these frameworks follow the principle of the “new risk assessment paradigm” (3) through comparisons of exposure (predicted estimations or measured values) levels with expected responses based on dose-response observations. The simplest comparison is a single-point estimate, such as the hazard quotient (HQ) (which is the measured or estimated environmental concentration divided by the toxicant reference value) (4, 5), which in the EU terminology is presented as the risk characterization ratio (RCR), defined as the quotient of the predicted environmental concentration (PEC) and the predicted no effect concentration (PNEC) (2, 6). If these ratios suggest that risk exceeds a level of concern, a risk refinement using higher tier approaches becomes necessary. A high level of conservatism is required to cover the extrapolation uncertainty in these initial point estimates (7), but it should be made clear that this is just a screening level approach, which can only be used to focus on the most important contaminants and emission sources. An HQ > 1 (or RCR > 1) does not necessarily indicate real risk, but that another more complex assessment must follow to reduce uncertainty to an acceptable level. Moreover, an HQ < 1 indicates that concentrations are well below those known to be toxic, but this approach cannot quantify the likelihood and magnitude of the risk, e.g., the probability for exceeding thresholds for long- and short-term effects (1, 3, 8). Many authors and scientific institutions have suggested the use of a tiered approach for exposure and effect determination and for risk characterization, and particularly the use of probabilistic approaches to risk assessment (5, 6, 8, 9). In the EU, the Scientific Steering Committee suggested five different levels for setting exposure assessment as well as five levels for the effect assessment, offering large possibilities for risk refinement (1). Typical refinements in risk assessment are the use of species sensitivity distributions (SSDs) combined with distributions of exposure concentrations to better describe the probability of exceedence of effect thresholds and thus the probability of adverse effects. Probabilistic approaches can cover both the uncertainty in the estimation and the true variability observed for most environmental variables (10, 11). Probabilistic risk estimations can be applied with use of different methodologies and for different purposes. The use of probabilistic approaches for generic risk assessments is included in regulatory-based tiered processes in several parts of the world (2, 12, 13). Similarly, probabilistic risk assessment methods can also be applied to site-specific risk assessments, e.g., to assess the risk of contaminated land. This paper presents an environmental risk assessment for an industrial polluted site in which high levels of 1,2,4trichlorobenzene (1,2,4-TCB) have been observed. The risk for soil organisms and ground(pore)water has been estimated using information on measured levels and reported toxicity values for 1,2,4-TCB. Several probabilistic options for refining the ecological risks in this site-specific assessment are presented, following a tiered approach, using this contaminated site as a case study for investigating the kind of information provided by each risk refinement option. 10.1021/es049214x CCC: $30.25

 2005 American Chemical Society Published on Web 03/24/2005

FIGURE 1. Contaminated site map with location of sampling points.

Methods Samples were collected from a former industrial chemical site located in the North of Italy. Main products were (azo) dyes and color pigments and other industrial organic intermediates. These compounds and their intermediates contaminated the nearby river and the aquifer under the factory. Currently a waterproof underground side wall is under construction, and a series of wells located inside of it maintains the aquifer surface below the lowest river level to avoid further contamination. The analysis campaign was carried out between August 2000 and January 2001. The site was divided into nine main areas corresponding to different uses during industrial activity. Area A5 corresponded to the riverside, where side wall and wells are located. Samples were collected following regular grids of about 25 × 25 m (Figure 1). After thorough homogenization, the 2-mm soil fraction was used for analytical characterization. Several compounds were detected on the site; this paper covers exclusively 1,2,4TCB, identified as a major contaminant in the area. 1,2,4TCB determination was obtained by automated Soxhlet extraction (ASE) and gas chromatography/mass spectrometry (GC/MS) according to U.S. EPA methods 8270, 3541, and 3640 (14-16). A soil sample of 2-20 g was mixed with deuterated 4-dichlorobenzene as internal standard and azobenzene solution (process standard). ASE extraction was performed with dichloromethane, and the extract was dried and concentrated to 1-2 mL in rotovapor and finally

analyzed, after cleanup by gel permeation, by using GC/MS scan mode or selected ion monitoring (SIM) mode. Water samples were subjected to the same procedure. Method detection limits (MDLs) were the following: 0.1 mg/kg for soil and 1 µg/L for water concentrations. The results were used for a tiered risk assessment. For each sample point, the average of the concentrations measured in the top 2-m soil layer was used. Selection was based on information on waste management at the site and management options. At the same time, it allowed comparisons between soil and ground(pore)water related risks. These values were used as measured environmental concentrations (MECs) in the exposure assessment. Acute toxicity of 1,2,4-TCB to soil organisms was evaluated using validated data included in the European Union Risk Assessment Report (EU RAR) (17). In addition, one chronic toxicity value was provided by the EPA ECOTOX database: 14-days no-observed-effect concentration (NOEC) for Lactuca sativa was 10 mg/kg soil dry weight (18). Groundwater toxicity was assessed by using acute and chronic data on aquatic organisms adopted in the EU RAR (17). The geometric average was calculated when two or more data were available for the same species (2, 19). Toxicity data used for the SSD calculation are reported in the Supporting Information. The risk assessment process used in this work followed a tiered approach based on the levels of refinement proposed by the U.S. EPA Ecological Committee on FIFRA Risk Assessment (ECOFRAM) in its final Draft Report on Terrestrial VOL. 39, NO. 9, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2921

Risk Assessment (12). The methodology deviates from the U.S. EPA approach for hazardous waste site assessment, but it is in line with the refinement process suggested in the EU guidance document (2). The following tiered approach was adopted: (1) Level 1 involves deterministic quotients that are simply ratios of single values of exposure divided by toxicity. HQs were calculated using the lowest chronic toxicity value and the threshold value (PNEC) suggested in the EU RAR. Exposure values used for the comparison were 1,2,4-TCB average (calculated by both excluding and replacing the values below the MDL with half of the detection limit) and maximum area concentrations. (2) Level 2 represents the comparison of exposure distributions with fixed values for effects or, conversely, the comparison of fixed exposure values with effect distributions. These provide the probability that exposure levels exceeded preestablished effect levels or vice versa. As stated by Travis and Land (20), the assumption of log normality for environmental data is fairly universal, but ElShaarawi and Esterby (21) suggest that such an assumption should not be accepted automatically and its suitability should be checked. The suggested approach is the use of the quantile-quantile (Q-Q) plot. Hence, the (ln)transformed concentrations above the MDL were plotted versus their normal scores, resulting in a straight line whose slope and intercept provided an estimate for the average and the standard deviation of the distribution. Alternative approaches for distribution fitting in the presence of censored values were evaluated using the whole soil exposure data set (see the Supporting Information). Although exposure data could reasonably be described by a log-normal distribution and in such a case many authors recommend the use of maximum probability estimation (MLE) methods (22-24), the robust regress method was chosen because estimates of extrapolated values can be directly retransformed and summary statistic computed in the original units, thereby avoiding transformation bias (25). Moreover, according to Gilliom and Helsel (26), it represents the most robust estimation method of minimizing errors in estimates of the average, standard deviation, median, and interquartile range of censored data. Following this procedure, the fitted distribution of concentrations above the MDL was used to extrapolate a collection of less-thans, and these data were used collectively to estimate the summary statistic by using the MLE routine of STATGRAPHIC Plus 4.1 (Statistical Graphic Corp.). The species sensitivity distribution approach was used for setting effect distributions, assuming the same distribution for the single test species as for all community species (27). Distribution fittings were calculated by using STATGRAPHIC Plus 4.1. The parameters of the exposure distributions and the SSDs (average, standard deviation, percentiles) are reported in the Supporting Information. The calculated cumulative distribution function (CDF) of exposure values was compared with the PNEC, the lowest chronic NOEC, and acute EC50 values. The intercept of the environmental concentrations CDF with toxicity values was used to calculate the probability that these concentrations would be exceeded (level 2.1). When possible, the same procedure was applied to single areas by using distributions of measured values within each area. In parallel, SSD was compared to different levels of exposure (average and maximum concentration for each area), and the percentage of affected species was calculated (level 2.2). (3) Level 3 utilizes both exposure and effect distributions. The margin of safety (MOS10) (9) was calculated as the quotient between the 10th percentile for SSD and the 90th percentile for the exposure distribution. In addition, the 2922

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 9, 2005

reverse cumulative distribution of exposure (or exceedence probability function, EXF, calculated as 100 - CDF) and the SSD were used to generate a joint probability curve (JPC), which describes the probability of exceeding the concentration associated with a particular degree of effect (5, 12). JPCs were calculated for different levels of protection. They were obtained by dividing the toxicity values by different factors (10 and 100) (i.e., to generate a set of “surrogate NOECs”), then by calculating the SSD, and finally by combining the SSD with the EXF of exposure. (4) Level 4 involves the so-called distribution-based quotients (DBQs) (12). Risk was expressed as the probability of exceeding preselected hazard quotients (exposure/toxicity ratios). The Monte Carlo technique was used to sample randomly values from distributions of exposure and toxicity (10 000 simulations) and to generate a distribution expressing the probability of the quotients to be exceeded. The Monte Carlo analysis was performed by using Crystal Ball 2000 v.5.2 (Decisioneering, Inc.). The decision to distinguish the last two levels was taken as regulations in the European Union frequently establish regulatory-binding quotients. Monte Carlo simulation can easily handle a large variety of data distributions and data sets even when data do not fit specific distributions.

Results and Discussion Soil Toxicity. Level 1. The worst-case assessment, obtained through the comparison of the maximum concentration and the PNEC (calculated by applying a factor of 1000 to the lowest EC50 mg/kgsoil, not normalized for the soil organic carbon content (2)), suggested a potential for adverse effects to terrestrial organisms with all quotients higher than 1 except for site A1b and reaching values over 1000 for four sites. The use of the worst-case average (only values above MDL) or “real” average (all values using MDL/2 for samples below the MDL) reduced the actual HQ values but did not change the pattern of areas above the 1 trigger. A more realistic refinement, based on comparisons of maximum and mean concentrations with the lowest NOEC, suggested values higher than 1 only for the maximum concentrations. For the real averages the highest HQ was 0.39, and it did not exceed 1 also for the upper 95% confidence value. These results offer additional information on the potential risk identified by using the PNEC value. The PNEC includes an application factor for covering species more sensitive than those tested. The comparison of the graphs presented in Figure 2 indicates that the potential risk is mostly related to this application factor. However, some individual values present concentrations exceeding those observed as toxic for some organisms, so confirming the possibility of effects at least in hot spots. Point estimate quotients are the most commonly used tool for risk assessment of contaminated sites (28-32). Simplicity, transparency, and low data requirements are their main advantages. However, these quotients do not offer quantitative information on the probability and magnitude of the effects. The sampling efforts and the availability of ecotoxicity data allowed refinement of the estimated risk by using additional risk characterization tools. The initial refinement was performed by using probability distributions for the exposure assessment. Level 2.1. Comparisons of exposure distribution and effect data indicated probabilities of exceedence of the PNECsoil and lowest NOEC of 36.45 and 2.71%, respectively. The probability of exceeding acute EC50s was close to zero (Figure 3). Probability distributions were also obtained for each single area (areas A1, A1b, and A3 were excluded due to the lack of detectable concentrations). Areas A4b and A6b showed the highest values, with probabilities of 61.2 and 80.9% of exceeding PNEC value (all other areas remained below 30%)

FIGURE 2. HQs calculated by dividing maximum concentrations and averages respectively by PNECsoil (left) and lowest NOEC (right).

FIGURE 3. Probability of exceeding soil toxicity values. and 5 and 7% of exceeding NOEC value, respectively. The probability of exceeding the EC50 was below 1.5%. Level 2.2. A parallel assessment was performed on the toxicity of 1,2,4-TCB. Only acute end points on enough soil species were available for SSD estimations. Data on invertebrates, plants, and soil respiration were combined in a single SSD. The inclusion of all taxa in a single curve was justified by the lack of differences in the sensitivity ranges. Van Beelen et al. (33) have reported an SSD for 1,2,4-TCB slightly higher than the one obtained here, but the number of species considered was lower and the data were normalized to a standard soil with 10% organic matter. The same authors reported small differences for 1,2,4-TCB between the SSD obtained for soil organisms and the SSD estimated from toxicity data on aquatic organisms and extrapolated to soil using the equilibrium partitioning method. The comparison of exposure values with the SSD showed a negligible percentage of species with expected EC50 below the average of the measured values. The higher average concentration (3.33 mg/kg, area A6b) exceeded the EC50 for only 0.1% of the species. The situation was different for percentages calculated according to maximum area concentrations. About 70% of species are expected to be acutely affected in area A4, while the percentage for area A6b reached almost 50%. Percentages in other areas were lower, and only areas A2 and A5 exceed 5% of affected species. These two approaches for level 2 should be considered complementary. They present different dimensions of the magnitude of the risk. Exposure distributions inform on the relative extent of the risk within the study area, while effect distributions bring up the intensity of the adverse effects. The conceptual difference between the two approaches may result in different rankings, e.g. between areas A4b and A4. Each approach offers complementary information for setting priorities for the risk management and risk reduction decision-making process. Level 3. The following level of refinement was the combination of exposure and effect distributions (shown in

FIGURE 4. Comparison of exposure and effect data as cumulative distribution functions. Figure 4) to calculate MOS10. The obtained value was 49.13 [18.28; 132.07] (MOS10 indicates the factor between the 90th percentile of exposure and the 10th percentile of effects; the higher the risk, the lower is the MOS10). It should be considered that the exposure distribution is based on real measurements and therefore represents the true variability for the distribution of the contaminant in the soil, not the uncertainty in the exposure estimation. As the effect distribution is based on acute effects, the MOS10 indicated the probability of acute and severe effects. The calculated value, around 50 with a low confidence limit of about 20, shows that acute severe effects are not expected over the whole area. Following ERA protocols (1, 2), the longterm assessment might be covered by the application of a factor to the EC50 values. The MOS10 was calculated also for areas that previous levels indicated as the most contaminated ones by applying a factor of 100 to the EC50 distribution. It resulted in 1.06, 0.11, and 0.09 respectively for A4, A4b, and A6b, so providing a classification based on risk potential as A6b > A4b > A4. JPCs resulting from direct comparison of EXF and SSD, shown in Figure 5, offer a better representation of the overall risk. The use of JPC helps to obtain a visual indication of the risk. As underlined by Solomon et al. (5), the closer the JPC is to the axes, the smaller the probability of adverse effects. In other words, the x-axis represents the intensity of effects (i.e., “how bad”), while the y-axis represents their probability (“how often”). Application factors between 10 and 100 for covering acute-to-chronic differences as described above are employed in several regulatory programs when chronic information is not available (2, 34). The JPCs corresponding to these values are also shown in Figure 5. The probability of exceeding the EC50 for 5 and 10% of the species ranged from 0.73 to 0.89%, while using a safety factor of 100, the probability increased to 14.23-15.88%. Looking at the JPCs of Figure 5, the site could be classified as relatively safe (5, 11, 12) with respect to 1,2,4-TCB. VOL. 39, NO. 9, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2923

FIGURE 5. Joint probability curves for soil toxicity.

FIGURE 6. Joint probability curves for groundwater toxicity.

Level 4. The probabilities of exceeding preestablished levels of risk were obtained directly from probabilistic quotients distribution generated by Monte Carlo simulations. The exceedence probability was 0.39% for the DBQ of MEC/ EC50 g 1 and 0.65% for the DBQ of MEC/EC50 g 0.5, while the application of a factor of 100 (corresponding to the DBQ of MEC/EC50 g 0.01) increased the probability to 9.69%. The exceedence probability resulted in about 26.61% considering the same factor used for the deterministic assessment (MEC/ EC50 g 0.001). The availability of several acute toxicity data allowed use of the probabilistic approach as a refinement for the more conservative lower tier approach based on a deterministic PNEC. The initial estimations, showing high HQ values, suggested that a generalized potential risk for the whole area could not be excluded. The refinement methods allowed the estimation of the actual risk, associated with the presence of hot spots with concentrations able to affect acutely a large proportion of soil species, while the large majority of the area presents 1,2,4-TCB concentrations below toxic levels. The effects covered in this risk assessment are exclusively based on direct toxic effects using standard end points. Other effects, such as avoidance behavior, can be more sensitive (35) and have been suggested as screening methods (36). In addition, phenomena such as acclimatization, bioavailability, and aging, etc., require additional tools, such as direct toxicity testing with the contaminated soil (37) and are not covered in this assessment. Ground(pore)water Toxicity. Level 1. Maximum concentration compared with PNECwater (calculated by dividing the lowest NOEC by 10, according to the EU protocol (2)) resulted in HQs higher than 20 for most of the area and about 320 for the most polluted area (A4b). The HQs based on the comparison between the “real” averages of the measured values and the lowest NOEC resulted in HQs lower than 1 for three areas, A1, A3, and A6, and comprised between 1.3 and 22 for the others. Level 2.1. The probability of exceeding fixed threshold values of toxicity was calculated by using the exposure distribution of the whole site (as the monitoring effort for groundwater did not allow independent assessment for each area). The probability of exceeding the PNECwater, and the lowest NOEC and LC50 values was 72.39, 37.42, and 9.69%, respectively. Level 2.2. For effects assessment, considering data availability, SSDs were estimated for both acute LC50s and chronic NOECs. In addition, the SSD for aquatic invertebrates (five crustacean species) was estimated. The resulted curve (average, 1968.86 µg/L; standard deviation, 2172.41 µg/L) was within the confidence interval for the overall SSD and therefore the overall curve (13 species) was used for the assessment. SSDs for acute and chronic toxicity were compared with the maximum and the average (calculated by replacing the censored values with half of the detection

limit) concentrations detected in each area (Figure 6). As far as maximum concentrations were concerned, the most polluted zone was area A4b, with a measured concentration expected to affect 99.5% of the species considering chronic toxicity and about 33.20% considering acute effects. Area A5 exceeded 97% of affected species for chronic toxicity (15.8% for acute), while all other areas (excluding areas A3 and A6) exceeded 30% of affected species (chronic SSD). Level 3. Comparing exposure and effect distributions, MOS10 resulted in 1.30 for acute toxicity, and it was 1 order of magnitude lower than 1 for chronic toxicity already without applying any safety factor. Applying a safety factor of 10 to the NOEC, the resulting JPC (Figure 6) allowed classification of the site as relatively unsafe (5, 12, 13). Exceedence probabilities for 10 and 5% of the species were 8.30 and 10.56% for the acute LC50 and 34.94 and 39.62% for the chronic NOEC values, respectively. The TGD applies a factor of 1000 to the LC50 values and of 10 to the NOECs if long-tern tests for three species of three trophic levels are available. Other protocols, including several European protocols for risk assessment of particular groups of chemicals, apply a factor of 100, instead of 1000, to the LC50 (1). Figure 6 shows that the assessment factor of 1000 can be regarded as conservative in this case. In fact, the exceedence probability curve observed for the application of a safety factor of 10 to the chronic SSD was quite similar to that obtained for the application of a safety factor of 100 to the acute SSD. Level 4. The results of Monte Carlo simulations for chronic toxicity showed an exceedence probability of 21.59% for the DBQ of MEC/NOEC g 1, while a safety factor of 10 (corresponding to the DBQ of MEC/NOEC g 0.1) increased the probability to 52.96%. A factor of 10 is usually employed to cover interspecies extrapolation when a full chronic data set is available (2). The exceedence probability of acute toxicity ranged from 3.92% (DBQ for MEC/EC50 g 1 to 48.36% (DBQ for MEC/EC50 g 0.01). Comparison of Soil and Ground(pore)water Results. The comparison of ground(pore)water concentrations with aquatic ecotoxicity data offers an estimation of the potential risk of the contaminated site for aquifers and streams in the adjacent area. Ground(pore)water concentrations are used as subrogate for concentration in draining waters and leachates moving from the site to adjacent water bodies. The comparison of the observed risk and the river flow can be used for estimating the required efficacy of the adopted riskmanagement measure by calculating the maximum amount of drainage/leaching water that could flow into the river without creating unacceptable risk levels for aquatic communities. The variability in the dilution factor associated with river flow and the magnitude of rain events can also be covered by probabilistic estimations (38). In addition, measurements represent pore water concentrations, which, through comparison with aquatic ecotoxicity data, offer a

2924

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 9, 2005

TABLE 1. Information Obtained for Each Level of the Refinement Process level

type of information

1 2.1

no information on probability and magnitude of effects quantitative information on one dimension of the magnitude of the risk: the extent (percentage of samples/ site surface under a determinate effect level) quantitative information on one dimension of the magnitude of the risk: the intensity (percentage of species with effect values at or below the established exposure level) information on probability could be obtained by comparing results based on average, Xth percentiles, and maximum concentrations MOS provides general information about the level of risk information on probability based on the combination of distributions; all dimensions for expressing the magnitude of the risk are combined in a single number joint probability curves offer visual indication of risk, allowing distinctions between the extension and the intensity of the expected effects possibility to evaluate a wide range of possible levels of protection/risk combined quantitative information on the probability and magnitude of the risk information on the probability of both types of effect; again, possibility to obtain the probability for different levels of protection

2.2

3

4

risk estimation for soil-dwelling organisms exposed via pore water. This approach is frequently used (2) as toxicity data are more often found for aquatic organisms than for soildwelling organisms. Van Beelen et al. (33) have recently published a comparison for 24 chemicals including 1,2,4TCB, showing an acceptable agreement for this particular chemical and for the equilibrium partitioning theory in general. The combined approach is particularly suitable to cover the uncertainty related to other soil taxonomic groups and the required acute to chronic extrapolation, as chronic SSD for soil organisms cannot be derived. A recent ECETOC report (39) presents acute-to-chronic ratios for 1,2,4-TCB of 23 for invertebrates and 34 for fishes. The comparison of aquatic SSD for acute and chronic data agrees well with these estimations, suggesting that a factor slightly higher than 10 should be applied. Therefore, it can be assumed that the JPC for 1/10 and 1/100 of the EC50, shown in Figure 6, represents a proper interval for the predicted JPC for chronic effects on soil organisms. Overall, the risk assessment, based on soil data, indicates a low general risk. However the presence of a small percentage of hot spots with levels above those capable of affecting a large number of species has been detected. A detailed evaluation of the expected effects can be obtained from the location of taxonomic groups within the SSD. The most sensitive effects reported for soil organisms are acute effects on soil respiration and effects on sensitive plants. The recovery of microbial soil population from the initial chemical stress is a well-reported mechanism, which has been included in the new OECD guidelines (40). In fact, for 1,2,4-TCB the adaptation phenomenon has been observed by Schroll et al. (41) even for the mineralization of the substance. These authors observed a rapid mineralization of 1,2,4-TCB by an adapted population obtained from a contaminated site with respiration rates between 0.1 and 0.6 mg of CO2 production/ gsoil for soils contaminated at initial concentration of 5 mg of TCB/kgsoil, 10 times lower than the reported 24-h EC50. Regarding plant species, the acute effects were observed within a wide range, of about 2 orders of magnitude, indicating different species sensitivities. The probabilistic soil risk assessment indicates that, even in hot spots, the effects are expected to be associated with changes in the microbial community and plant biodiversity. However, they allow the maintenance of basic soil functions and plant coverage. The assessment of pore water risk leads to a different set of conclusions. The potential risk is distributed on a large proportion of the site, while the local risk of hot spots is lower, as observed through the comparisons of the HQ for

the PNECs and of the percentage of species with acute EC50s below the maximum measured concentration. The highest risk for pore water cannot be explained by a higher sensitivity of aquatic organisms. The opposite has been reported by van Beelen et al. (33). Therefore, these results indicate that the larger risk of 1,2,4-TCB in the studied site is related to pore water concentrations. The comparison of the results obtained for the different levels offers a proper estimation of the information obtained for each method, which has been summarized in Table 1. It should be considered that the higher level methods offer information on the overall risk of the site but require additional information in order to set the relevance and magnitude of effects in hot spots.

Acknowledgments M.Z. and C.C. thank the Italian Delegate Commissioner for the remediation of the ex-ACNA site and the Valbormida for the research grant awarded to CIMA, which allowed the realization of the present work.

Supporting Information Available Toxicity data of 1,2,4-TCB to soil and water organisms, statistical treatment of data, and descriptive statistic of the exposure and the toxicity data. This material is available free of charge via the Internet at http://pubs.acs.org.

Literature Cited (1) SSC. The Second Report on the Harmonisation of Risk Assessment Procedures. EC Health & Consumer Protection Directorate-General, Directorate C-Scientific Opinions. C1-Followup and Dissemination of Scientific Opinions, April 2003. http:// europa.eu.int/comm/food/fs/sc/ssc/outcome_en.html. (2) EU. Technical Guidance Document on Risk Assessment in Support of Commission Directive 93/67/EEC on Risk Assessment for New Notified Substances, Commission Regulation (EC) No. 488/94 on Risk Assessment for Existing Substances Directive 98/8/EC of the European Parliament and of the Council Concerning the Placing of Biocidal Products on the Market, 2003; EUR 20418 EN/1. (3) U.S. EPA. Guidelines for Ecological Risk Assessment. Risk Assessment Forum; Environmental Protection Agency: Washington, D.C., 1998; EPA/630/R-95/002F. (4) U.S. EPA. Ecological Risk Assessment in SuperfundsGlossary of Terms. http://www.epa.gov/region5/superfund/ecology/ html/glossary.html#hazard. (5) Solomon, K.; Giesy, J.; Jones, P. Probabilistic Risk Assessment of Agrochemicals in the Environment. Crop Prot. 2000, 19, 649655. (6) Jager, T.; Vermeire, T. G.; Rikken, M. G. J.; van der Poel, P. Opportunities for a Probabilistic Risk Assessment of Chemicals in the European Union. Chemosphere 2001, 43, 257-264. VOL. 39, NO. 9, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2925

(7) Tarazona, J. V. The Identification of Thresholds of Acceptability and Danger: The Biological Route. Arch. Toxicol., Suppl. 1997, 19, 137-146. (8) Klaine, S. J.; Cobb, G. P.; Dickerson, L.; Dixon, K. R.; Kendall, R. J.; Smith, E. E.; Solomon, K. R. Ecological Risk Assessment for the Use of the Biocide, Dibromonitrilopropionamide (DBNPA), in Industrial Cooling Systems. Environ. Toxicol. Chem. 1996, 15 (1), 21-30. (9) Solomon, K. R.; Baker, D. B.; Richards, R. P.; Dixon, K. R.; Klaine, S. J.; La Point, T. W.; Kendall, R. J.; Weisskopf, C. P.; Giddings, J. M.; Giesy, J. P. Ecological risk assessment of atrazine in North American surface waters. Environ. Toxicol. Chem. 1996, 15 (1), 31-76. (10) Aldenbreg, T.; Jaworska, J. S. Uncertainty of the Hazardous Concentration and Fraction Affected for Normal Species Sensitivity Distributions. Ecotox. Environ. Saf. 2000, 46, 1-18. (11) Chapman, P.; Reed, M. Revised Preliminary Paper on Methods of Uncertainty Analysis. EUFRAM Work Package 4, 2004. http:// www.eufram.com/ outputs.cfm. (12) ECOFRAM. Terrestrial Draft Report. U.S. EPA. May 1999. http:// www.epa.gov/ oppefed1/ecorisk/index.htm. (13) SETAC. Guidance Document on Higher Tier Aquatic Risk Assessment for Pesticides (HARAP); Campbell, P. J., Arnold, D. J. S., Brock, T. C. M., Grandy, N. J., Heger, W., Heimbach, F., Maund, S. J., Streloke, M., Eds.; 1999. ISBN 90-5607-011-8. (14) U.S. EPA Method 8270C. Semivolatile Organic Compounds by Gas Chromatography/Mass Spectrometry (GC/MS). http:// www.epa.gov/SW-846/8_series.htm. (15) U.S. EPA Method 3541. Automated Soxhlet Extraction. http:// www.epa.gov/SW-846/3_series.htm. (16) U.S. EPA Method 3640. Gel Permeation Cleanup (GPC). http:// www.epa.gov/SW-846/3_series.htm. (17) ECB. 1,2,4-Trichlorobenzene European Union Risk Assessment Report, European Commission Joint Research Centre., 2003; EUR 20540 EN. (18) U.S. EPA. ECOTOX: Ecological Toxicity Database, MidContinent Ecology Division, National Health and Environmental Effects Research Laboratory, 2001 (19) United Nations. UN Globally Harmonized System of Classification and Labeling of Chemicals (GHS), New York and Geneva, 2003. http://www.unece.org/trans/danger/publi/ghs/ officialtext.html. (20) Travis, C. C.; Land, M. L. Estimating the Mean of Data Sets with Nondetectable Values. Environ. Sci. Technol. 1990, 24, 961962. (21) El-Shaarawi, A. H.; Esterby, S. R. Replacement of Censored Observation by a Constant: Anevaluation. Water Res. 1992, 6, 835-844. (22) Haas, C. N.; Sheff, P. A. Estimation of Averages in Truncated Samples. Environ. Sci. Technol. 1990, 24, 912-919. (23) El-Shaarawi, A. H. Inference about the Mean from Censored Water Quality Data. Water Resour. Res. 1989, 25 (4), 685-690. (24) Singh, A.; Nocerino, J. Robust Estimation of Mean and Variance Using Environmental Data Sets with below Detection Limit Observations. Chemomet. Intell. Lab. 2002, 60, 69-86. (25) Helsel, D. R. Less than Obvious. Statistical Treatment of Data below the Detection Limit. Environ. Sci. Technol. 1990, 24, 17661774. (26) Gilliom, R. J.; Helsel, D. R. Estimation of Distributional Parameters for Censored Trace Level Water Quality Data. 1. Estimation Techniques. Water Resour. Res. 1986, 22 (2), 135146.

2926

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 9, 2005

(27) Wagner, C.; Lokke, H. Estimation of Ecotoxicological Protection Levels from NOEC Data. Wat. Res. 1991, 25, 1237-1242. (28) U.S. EPA. EPA Superfund Record of Decision: West Site/Hows Corners Plymouth, ME. EPA ID: MED985466168. EPA/ROD/ R01-02/021. 2002. http://www.epa.gov/superfund/sites/rods/ fulltext/r0102021.pdf. (29) U.S. EPA. Record of Decision Summary: Central Wood Preserving Superfund Site. East Feliciana Parish, Louisiana. EPA ID: LAD008187940. 2002. http://www.epa.gov/earth1r6/6sf/ pdffiles/atsf_final_rod.pdf. (30) Riccardi, C.; Berardi, S.; Di Basilio, M.; Gariazzo, C.; Giardi, P.; Villarini, M. Environmental Assessment of a Site Contaminated by Organic Compounds. J. Environ. Sci. Health, Part A: Toxic/ Hazard. Subst. Environ. Eng. 2001, 36 (6), 957-970. (31) Aslibekian, O.; Moles, R. Environmental Risk Assessment of Metals Contaminated Soils at Silvermines Abandoned Mine Site, Co Tipperary, Ireland. Environ. Geochem. Health 2003, 25 (2), 247-266. (32) Lu, H.; Axe, L.; Tyson, T. A. Development and Application of Computer Simulation Tools for Ecological Risk Assessment. Environ. Model. Assess. 2003, 8, 311-322. (33) van Beelen, P.; Verbruggen, E. M.J.; Peijnenburg, W. J. G. M. The Evaluation of the Equilibrium Partitioning Method Using Sensitivity Distributions of Species in Water and Soil. Chemosphere 2003, 52, 1153-1162. (34) Swartjes, F. A. Risk-Based Assessment of Soil and Groundwater Quality in The Netherlands: Standards and Remediation Urgency. Risk Anal. 1999, 19 (6), 1235-1249. (35) Schaefer, M. Assessing 2,4,6-Trinitrotoluene (TNT)-Contaminated Soil Using Three Different Earthworm Test Methods. Ecotoxicol. Environ. Saf. 2004, 57 (1), 74-80. (36) da Luz, T. N.; Ribeiro, R.; Sousa, J. P. Avoidance Tests with Collembola and Earthworms as Early Screening Tools for SiteSpecific Assessment of Polluted Soils. Environ. Toxicol. Chem. 2004, 23 (9), 2188-2193. (37) Ronnpagel, K.; Janssen, E.; Ahlf W. Asking for the Indicator Function of Bioassays Evaluating Soil Contamination: Are Bioassay Results Reasonable Surrogates of Effects on Soil Microflora? Chemosphere 1998, 36 (6), 1291-1304. (38) Fernandez C.; Carbonell G.; Tarazona J. V. Probabilisitc Approximation to Risk Assessment of Basins by Ecotoxicological Evaluation. In Integrative Modelling of Biophysical, Social and Economic Systems for Resource Management Solutions; Post, D. A., Ed.; The University of Western Australia UNIPRINT: Perth, 2003; pp 637-641. (39) ECETOC. ECETOC Aquatic Hazard Assessment II. Technical Report No 91, Brussels, 2003. ISSN-0773-8072-91. (40) OECD/OCDE. OECD Guidelines for the Testing of Chemicals. 217. Soil Microorganisms: Carbon Mineralization Test. 2000. http://www.oecd.org/dataoecd/17/45/1948325.pdf. (41) Schroll, R.; Brahushi, F.; Dorfler, U.; Kuhn, S.; Fekete, J.; Munch, J. C. Biomineralisation of 1,2,4-Trichlorobenzene in Soils by an Adapted Microbial Population. Environ. Pollut. 2004, 127 (3), 395-401.

Received for review May 27, 2004. Revised manuscript received February 4, 2005. Accepted February 15, 2005. ES049214X