Policy Analysis Uncertainty in Octanol-Water Partition Coefficient: Implications for Risk Assessment and Remedial Costs IGOR LINKOV,* MICHAEL R. AMES, EDMUND A. C. CROUCH, AND F. KYLE SATTERSTROM Cambridge Environmental Inc., 58 Charles Street, Cambridge, Massachusetts 02141
The assessment of human health and ecological risks at chemically contaminated sites often includes the use of models to assess chemical transport, fate, and exposure/toxicity. These models require input data on a variety of physical and chemical properties for each compound of concern. Small changes in some of these parameters may result in significant differences in estimated human health or ecological risks and in the extent of required remediation efforts. The octanol-water partition coefficient (Kow) for hydrophobic organic compounds is one such parameter, particularly because it is often used to estimate additional partitioning and bioaccumulation parameters. Unfortunately, there is considerable variability among tabulated Kow values for many compounds of concern. This paper assesses the implications of using various values of Kow to calculate health-protective polychlorinated biphenyl (PCB) sediment quality objectives (SQOs) in a case study using a simplified food chain model and the range of Kow values available from or recommended by the U. S. EPA. For the site and Kow values considered in this study, which are a snapshot of values available in the spring of 2004, the SQOs differ by as much as a factor of 5. This range of SQOs is estimated to correspond to a difference in remediation costs of $48 million.
Introduction The octanol-water partition coefficient (Kow) describes the equilibrium ratio of the concentrations of a chemical substance in n-octanol and in water. This partition coefficient is widely used in risk assessments to approximate the distribution of chemicals between aqueous and organic media. Further, Kow is widely used to estimate other physical properties and toxicities. For example, in exposure modeling, it is used to estimate dermal permeabilities and absorption from the gastrointestinal tract and lung. Environmental models utilize Kow to estimate dissolved concentrations in water, bioconcentration coefficients between environmental media and living organisms, and soil and sediment adsorption coefficients. In principle, Kow is a well-defined and measurable property. In practice, the Kow values for many hydrophobic organic compounds are not well characterizedsa fact well* Corresponding author e-mail: linkov@CambridgeEnvironmental. com; phone: (617)225-0812; fax: (617)225-0813. 10.1021/es0485659 CCC: $30.25 Published on Web 08/11/2005
2005 American Chemical Society
known to both chemists and risk assessors (1). For example, a common problem in the preparation of mixtures for measuring Kow values above 106 is the presence of small quantities of emulsified octanol in the water phase, which presence results in erroneously high apparent water phase concentrations and erroneously low calculated Kow values (2). A complicating factor arises for mixtures of closely related compounds, for example PCBs or dioxins, that are often treated as a single “chemical” in risk assessments. For such mixtures, there is no single correct value of Kow that can be applied in all circumstances. It is possible for the uncertainty due to these two factors to be quantitatively characterized and incorporated in modeling. Advanced chemical analysis can be used to support selection of a site-specific Kow or to allow treatment of each specific chemical form separately. However, this is seldom practical and can rarely be carried through completely because ecotoxicological benchmarks are often available only for mixtures of such chemicals. Quantitative probabilistic analysis can be used to account for the uncertainty in Kow, but this method is also rarely applied, likely due to increased costs and the difficulty of obtaining regulatory approval. A more difficult situation to address arises when data of poor quality are recommended for use by reference documents due to the citation of inappropriate experimental results, inadequate documentation procedures, or simple errors in reporting. Pontolillo and Eganhouse (3) examined more than 700 publications reporting Kow values for DDT and DDE. They found variations of up to 4 orders of magnitude in reported Kow values for these compounds and little indication of a decline in the range of variation over the last five decades. Even the “recommended values” were found to range over more than 2 orders of magnitude. Eganhouse and Pontolillo (4) concluded that many values recommended in the literature tabulations are based on examination of erroneous and incomplete data compilations. The authors concluded that estimation of critical environmental parameters on the basis of Kow is inadvisable and could result in erroneous risk assessment results. This paper complements the study of Pontolillo and Eganhouse (3) by evaluating the variation of Kow for polychlorinated biphenyls (PCBs) present in databases which are either compiled or recommended by EPA, and which were available in the spring of 2004. Our overall goal is to assess the potential cost implications of the use of a range of these values for a site-specific remediation. The selection of PCBs for this evaluation is especially suitable because they are a primary risk driver for many contaminated sites, because there is considerable variation in Kow values for PCBs among available databases, and because PCBs, though actually a class of related compounds, are often treated as a single “chemical.” Possible explanations for the variation in reported log Kow values are considered, and we quantify Kow measurement uncertainty and variability for PCBs.
Case Study To ensure that this analysis of uncertain Kow values and contaminated sediment remediation costs represents a realistic scenario, we employ a case study approach. The case study is based on conditions at the Hylebos Waterways Superfund Site located in the southern basin of Puget Sound, near Tacoma, Washington. In 1983, EPA placed the Commencement Bay Nearshore/Tideflats Site on the National Priorities List of sites requiring investigation and cleanup VOL. 39, NO. 18, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
6917
under the Superfund Program. The nearshore/tideflats area was shown to be contaminated with a large number of hazardous substances at concentrations greatly exceeding those found in reference areas that were used for comparison (5). At Hylebos Waterway, one of the nine problem areas in Commencement Bay, PCBs were identified as the primary chemical of concern. In 1989, EPA issued a Record of Decision (ROD, 5) which established Sediment Quality Objectives (SQOs) at the site for a wide range of hazardous substances, as well as a tenyear remediation plan for meeting those objectives using a combination of sediment cleanup, source controls, and natural recovery. A site-specific biota-sediment accumulation factor (BSAF) was calculated based on available fish tissue and sediment data. This BSAF was used to establish the SQO for PCBs at 150 ppb, a value deemed to be health protective for local subsistence anglers. Using the same BSAF and revised input parameters for fish consumption rates and the toxicity factor for PCBs, EPA revised the SQO to 300 ppb in July 1997. The average PCB concentration in a reference area used for comparison was approximately 30 ppb. The Washington State Department of Ecology (WSDOE) used sediment data from Commencement Bay to demonstrate approaches for identifying PCB sediment contamination hotspots and for calculating urban bay, area-weighted, average PCB sediment concentrations (6). Using sediment sampling stations in Commencement Bay and the geographic information system ArcView (Environmental Systems Research Institute, Inc.) in conjunction with the Spatial Analyst extension, WSDOE mapped out areas in Commencement Bay associated with particular PCB concentrations. Although site-specific BSAFs were empirically determined for the area and were used by EPA and WSDOE for calculating SQOs at this site, such a process is often not practical. More frequently, SQOs are based on biotransfer parameters and modeling using tabulated data that are more easily obtained than site-specific information. We use this particular site for our convenience, since it provides readily available data allowing the assessment of the impact of employing a range of tabulated Kow values on derived SQO levels and remediation costs. The use of an actual Superfund site for the case study also provides the analysis with uncontrived contaminant distributions, modeling parameters, and remedial cost estimates. We modeled the partitioning and fate of PCBs at this site using the TrophicTrace risk model (7). Information required for the modeling (e.g., the spatial extent and level of PCB contamination, sediment characteristics, exposure parameters for local population, etc.) was taken from sitespecific measurements and estimates. Both WSDOE (8) and U.S. EPA (5) present the volumes and remedial costs associated with meeting a specified remedial action level. Required sediment remediation volumes and the approximate costs for various SQOs obtained in our case study were calculated by multiplying the ratio of the remediation costs to volumes from the WSDOE and EPA assessments by the measured volume of contamination above calculated SQOs.
Methods We used a combination of simple food chain and remediation cost models to analyze the effect of PCB Kow value uncertainty on the cost of a health-based cleanup of PCB contaminated sediments at a Superfund site. Details of each portion of the analysis are presented subsequently. Overall Model. The overall analysis proceeds as follows: (1) A PCB Kow value is selected for evaluation from the EPA chemical property databases. (2) The food chain model is run to estimate the transport of PCBs from sediments to game fish using an initial, nominal PCB sediment concentration (the cleanup level established by WSDOE) along with 6918
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 18, 2005
TABLE 1. Log KOw Values for Total PCBs and Aroclor 1254 Used in the Case Study database KowWin (calculated) KowWin (experimental) PhysProp ChemFate Water9 Chem9a SCDM (Superfund Chemical Data Matrix) SCDMWin Superfund PHEM (Public Health Evaluation Manual) STF (Soil Transport and Fate Database) HHRAP (Human Health Risk Assessment Protocol) PBT Profiler ATSDR
Log Kow Log Kow total Aroclor refer- date of PCBs 1254 ence download 6.34 6.29 7.10 3.90 7.31 7.31 6.70
9 9 10 11 12 13 14
4/12/2004 4/12/2004 4/12/2004 4/12/2004 3/25/2004 3/20/2004 3/30/2004
15 16
2/12/2004 NA
6.04
17
2/24/2004
6.21
18
NA
6.8 6.50
19 20
3/30/2004 NA
6.98 6.79 6.50 6.03 3.34
6.04 6.04 8.23
6.3
a The Chem9 database has been superseded by Water9 since the download date shown.
the selected Kow value and various other chemical, ecological, and site-specific parameters. (3) The lifetime dose resulting from fish consumption by the critical angler population was then estimated. (4) The SQO was estimated by iterative selection of the initial PCB concentration in sediment that results in modeled dose equal to the EPA reference dose (RfD) for PCBs. The RfD is a lifetime dose that has been deemed unlikely to result in an appreciable risk of deleterious effects. (5) The cost for remediation of sediments with PCB concentrations above the SQO are estimated from sitespecific contamination profiles, remediation volumes, and averaged cost per sediment volume dredged. (6) The food chain and remediation cost models are then rerun using other Kow values from the databases, and the dependence of remediation costs on assumed Kow values is plotted. Kow Values for Total PCBs and Aroclor 1254. The term PCB does not identify a specific compound but instead refers to any mixture of 209 PCB congeners, each of which has its own Kow. Several databases and software packages reporting physical-chemical properties for PCBs are posted or referenced on EPA’s web site. In addition, several EPA guidance documents for performing human health and ecological risk assessments contain recommended Kow values for PCBs. These EPA chemical property databases (which term hereafter is used to include estimation programs often included with such databases) include values for PCBs identified either as an unspecified mixture of congeners or as mixtures specified by their commercial Aroclor number. Although many original publications and reference handbooks also report Kow values for PCBs (e.g., 2), only values obtained from databases and software packages available from or recommended by EPA have been used in the case study. Aroclor 1254 is the mixture that is tabulated most often in the databases, so Kow values for both unspecified PCB mixtures and Aroclor 1254 were used in the case study. Table 1 shows the log Kow values used in the case study (values of Kow are often tabulated as their base 10 logarithm; we follow this convention here). Bioaccumulation Model. The ecological food chain model for this analysis is implemented using the TrophicTrace software package (7). TrophicTrace is an Excel (Microsoft Corp.) add-in that provides a spreadsheet tool for calculating potential human health and ecological risks associated with bioaccumulation of contaminants in dredged sediments, and it has been used by the U.S. Army Corps of Engineers in developing dredging and remedial goals (21). TrophicTrace applies a steady-state bioaccumulation model based on the
TABLE 2. TrophicTrace Input Parameters parameter
mean value
reference
Human Health Exposure and Risk Model. The potential noncancer human health risk was estimated using the hazard quotient approach defined as
Sediment and Water
Kow initial sediment concentration (total PCBs, ng/g dry wt) water concentration (total PCBs, ng/L) TOC (%)
300
3.4%
lipid content (%)
Sandworm 1.2%
Generic Forage Fish body weight (g) 3 lipid content (%) 1% site use factor (%) 50% body weight (g) lipid content (%) site use factor (%)
Generic Game Fish 250 3.5% 50%
Human Ingestion body weight (kg) 70 reference dose 0.00002 (mg/kg-day) fish ingestion (g/day) 42 exposure duration 7300 (days) site use factor (%) 50%
assumed, variable cleanup level established by WSDOE (8) estimated based on equilibrium 8 25-27 assumed assumed assumed site-specific 8 assumed 8 28 8 assumed assumed
approach of Gobas et al. (22, 23). The model uses a common polychaete, Nereis virens (sandworm), to represent the prey base for a generic forage fish which is assumed to be the sole food source for a generic game fish. The human receptors are tribal anglers eating the generic game fish. The model input parameters include sediment concentrations for PCBs, weight and lipid content of aquatic organisms, food ingestion rate and body weight of fish, total organic carbon in sediment, and Kow. Water concentrations are calculated from sediment concentrations using equilibrium partitioning. This is a conservative assumption (i.e., likely to overestimate risk) because equilibrium is not likely to be reached in a marine ecosystem. Site-specific information and default TrophicTrace data used to define input parameters are given in Table 2. Site-specific information was used when available. Several sources provide equations for the rate constants used in the model (22-24). The model predicts PCB accumulation in fish through direct gill uptake of PCBs from water and dietary consumption of contaminated prey, and PCB loss or reduction due to elimination (proportional to concentration), metabolic loss, and fish growth (7)
Cf )
k1 × Cwd + kd × Cdiet k2 + ke + km + kg
where Cf is the concentration of PCBs in fish tissue (µg/kg); Cwd is the freely dissolved concentration of PCBs in water (µg/L); Cdiet is the concentration of PCBs in the diet (µg/kg); k1 is the gill uptake rate (L/kg-d); kd is the dietary uptake rate (1/d); k2 is the gill elimination rate (1/d); ke is the fecal elimination rate (1/d); km is the metabolic rate (1/d); and kg is the growth rate (1/d). The dependence of Cf on Kow enters the model both through the calculation of PCB concentrations in water and benthic invertebrates, and through the calculation of PCB uptake, and gill and fecal elimination rates as detailed by Gobas et al. (23).
HQ )
IRf × Cf × ED BW × RfD × AT × 106
where HQ is the toxicity hazard quotient; IRf is the annual average fish ingestion rate (g/day); Cf is the concentration of PCBs in fish tissue (µg/kg); ED is the exposure duration (days); BW is the body weight (kg); RfD is the reference dose (mg/ kg-day); AT is the averaging time (days); and 106 is the unit conversion factor (103 µg/g × 103 g/kg). This equation was rearranged and used to calculate a concentration of PCBs in fish tissue that would result in an acceptable risk level based on the use of the following exposure parameters. Fish Ingestion Rate. For the purposes of this study, we used a mean fish ingestion rate of 42 g/day based on a survey of fish consumption among members of two Puget Sound tribes (8). Game fish from contaminated areas was assumed to comprise 50% of the anglers’ fish diet. Exposure Duration. An exposure duration of 7300 days (i.e., 365 d/yr for 20 yr) was used to characterize a long-term fish ingestion scenario. Body Weight. Body weight is set to 70 kg as is commonly used in EPA risk assessments (29). Reference Dose. The reference dose for Aroclor 1254 of 0.00002 mg/kg-d was selected for the assessment of both Aroclor 1254 and total PCBs following EPA guidance for food chain exposures (30). The value is from EPA’s Integrated Risk Information System (IRIS) database (28), and is specified as a point estimate also following EPA guidance (31). Averaging Time. An averaging time of 7300 days (i.e., 365 d/yr for 20 yr) was used to characterize long-term noncancer health risks. The choice of an acceptable risk level or HQ is a policy decision. Most regulatory programs in the United States set a risk criteria for noncarcinogens at an HQ of 1, or at a modeled dose equal to the Reference Dose. This risk criterion will be used for the purpose of deriving SQOs herein Remedial Cost Estimation. Remedial costs are usually estimated based on the volume of contaminated sediments requiring removal and site-specific dredging/disposal scenarios. In general, remedial cost estimation is done using site-specific information and is a very resource and timeconsuming process. For this case study the approximate remedial costs were estimated as the product of the area of sediments contaminated above the derived SQO, an assumed dredging depth, and an average dredging cost per unit volume. The spatial extent of contamination was taken from site study documents (6). The average dredging/disposal cost per unit volume was calculated using the ratio of overall remedial costs estimated by EPA (5) to the volume of sediment requiring removal estimated by the WSDOE (8). Note that this simplified approach relates to only one aspect of site contamination (its spatial extent) and does not take into account the large variability in sediment depth and costs expected in site-specific remedial projects.
Results Variability among Recommended Kow Values. A review of EPA databases (see Methods) reveals several sources reporting Kow values for total PCBs and/or Aroclor 1254 (Table 1). Reported log Kow values for total PCBs range from 3.90 to 8.23. The range of reported log Kow values for Aroclor 1254 is also quite wide (from 3.34 to 6.98). Some of the differences in Kow values are evident among different EPA offices. The STF model developed by the Office of Research and Development (ORD) recommends a log Kow VOL. 39, NO. 18, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
6919
value of 8.23 for total PCBs; the Office of Air Quality Planning and Standards (OAQPS) provides a value of 7.31 in their Water9 software; the Office of Pollution Prevention and Toxics (OPPT) gives a value of 6.3 in the PBT Profiler software; and the Office of Solid Waste and Emergency Response (OSWER) reports a value of 6.04 in the PHEM manual and in the SCDMWin database. Significant differences can also be found among the software/tabulation products generated within the same office. For example, both Water9 and Chem9 software were developed by OAQPS, but the Kow values for Aroclor 1254 that they generate differ by almost 3 orders of magnitude (log Kow values of 6.03 and 3.34, respectively). The log Kow value for total PCBs reported in the paper version of the OSWER SCDM database is 6.7, while its software implementation (SCDMWin) reports a value of 6.04. The PhysProp, KowWin, and ChemFate Software products were all developed by the Syracuse Research Corporation for EPA, but recommended Kow values for total PCBs from these databases also range widely (log Kows from 3.90 to 7.1 for total PCBs). It is possible that some of the differences are due to different intended purposes for the software/tabulationss different values of Kow may more accurately represent the mixture of PCBs likely to occur in different situations. However, no warnings are provided in any of the software/ tabulations on the applicability of the values obtained. Kow Variation due to Measurement Uncertainty and Variability. Variation in the Kow values discussed above could, conceivably, be attributable to such effects as the use of different measurement methods, variable congener distributions within an Aroclor, varying amounts of octanol and water used in the measurement, and different measurement temperatures. We have assessed the theoretical extent to which these factors would affect measurements of the octanol-water partition coefficient (see Supporting Information for a detailed description of our methodology). The selection of appropriate measurement methods was found to be the most crucial contributor to Kow variability. In the 1970s and 1980s, the shake-flask method was the most common way to measure octanol-water partition coefficients. This method involves vigorously shaking a container to mix the measured substance with the octanol and water phases. However, it may also introduce microemulsions of octanol into the water phase of the experiment. As shown in the Supporting Information, for substances with a large Kow, these microemulsions introduce nonnegligible amounts of experimental error. For example, for a substance with an actual log Kow of 6, if 1 of every 100 000 parts of octanol in the experiment exists in the water phase as part of a microemulsion, the experimenter will measure an apparent log Kow of 5, or a full order of magnitude low. As a result, the shake-flask method is not reliable for use when a substance’s log Kow is larger than 4 or 5. Nevertheless, if an appropriate method is used, the variability between Kow measurements is low; for example, recently Tolls et al. (32) coordinated a ring test involving interlaboratory comparison of Kow values measured using the slow-stirring method. The entire range of values reported by 15 participating laboratories spanned less than one log unit for each chemical. Additionally, we calculated the effects of variable congener distributions within an Aroclor, varying the amounts of octanol and water used in the measurement, and different measurement temperatures. Varying the congener distribution between those found in different historical batches of Aroclor 1254 contributes less than 0.2 log units to the Kow spread, while varying the amounts of octanol and water used in measurements has essentially no effect for any practical laboratory setup and imparts a spread of about 0.4 log units between all theoretical Aroclor 1254 Kow measurements. The effects of temperature are available for 21 PCB congeners, 6920
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 18, 2005
FIGURE 1. Sediment Quality Objectives calculated for different log Kow values. Arrows indicate potential costs of remediation to selected SQO levels. and examples presented in the Supporting Information show that a temperature difference of 25 °C would, on average, make a difference of approximately 0.3 log units in the measurement. Financial Implications. Figure 1 presents the Sediment Quality Objectives for PCBs calculated using log Kow values from 6.0 to 7.2. This range does not include extreme values recommended in some databases (e.g., the low values of 3.34 and 3.90, or the high value of 8.23). Selection of the remaining Kow values from different databases can result in SQO levels in the case study that differ by more than a factor of 7. The case study costs associated with remediation requirements to achieve these health-based SQO levels range from approximately $7.5 million for a log Kow value of 6.0 to $55 million for a log Kow value of 7.2.
Discussion Our review of Kow values available in EPA databases reveals a range of values that covers more than 4 orders of magnitude for total PCBs and more than 3 orders of magnitude for Aroclor 1254. These ranges are even wider than the 2 orders of magnitude range found for recommended Kow values for DDT and DDE by Pontolillo and Eganhouse (3). These are also wider than the range of log Kow values (6.0 to 6.8, with one cited value of 4.08) given by Mackay et al. (2), the most extensive handbook on chemical properties of organic contaminants. Additionally, these large ranges cannot be explained through simple measurement variability and uncertainty. As discussed above, we evaluated (see Supporting Information) the theoretical range of variation for Kow values for PCBs in well-controlled experiments (that are not confounded by experimental difficulties such as micro-emulsification of phases) due to such effects as variable congener distributions within an Aroclor, varying amounts of octanol and water used in the measurement, and different temperatures between measurements. We found that these variables are relatively insignificant (less than one log unit). Variations in other experimental conditions and procedures must contribute the bulk of the variability among the reported measurements. One notable example of poor quality data in an EPA database is the ChemFate log Kow value of 3.90 for total PCBs. This value can be traced back to a documentsTulp and Hutzinger (33)swhich specifically cites it as erroneous, but the value made its way into reference literature through a series of citations made without adequate examination of the original source (see Supporting Information for greater detail). Of course, such problems are found in other collections of Kow values as well; for example, Mackay et al. (2) cite multiple secondary or irrelevant papers in support of their Kow values for Aroclor 1254 (Supporting Information). It is clear that incomplete examination or misinterpretation
of sources and citation of second-hand literature is a general problem in the field, leading to incorporation of poor-quality data into reference documentation. Data quality is of the utmost importance - we have demonstrated the implication of selecting different Kow values for developing sediment quality objectives (SQOs) at a Superfund site using a standard approach applied through a peer-reviewed software package that was developed and recommended for such use by the Army Corps of Engineers. There are, of course, many other parameters where selection of differing values can lead to widely ranging risk estimates in a chemical transport, fate, and toxicity model. The sensitivity of risk estimates to these modeling uncertainties are often acknowledged and discussed at length in a risk assessment report. For example, Linkov et al. (34) quantitatively evaluates sources of uncertainty and variability in estimating risk to an ecological receptor (osprey) from trophic transfer of PCBs in sediments from the New York-New Jersey (NY-NJ) Harbor. The implemented food chain model is similar to the one presented in this study. The log Kow was varied in that study in a much narrower range (5.24 through 7.36). The sensitivity analysis showed that the Log Kow is the second most important parameter influencing the osprey toxicity quotients (after the osprey Toxic Reference Values, TRVs), followed by the sediment and water concentrations. These results are consistent with both Burkhard (24, 35) and Ianuzzi et al. (36) who concluded that the lipid content of the exposed organisms and the Kow of the contaminant influence estimates of tissue concentrations more than other parameters. A critical point of this case study exercise is that variation of the Kow value can have a large and usually unexamined cost implication for the site remediation. Ignoring other uncertainties in the modeling, variation of the Kow value for PCBs over the range available from the EPA databases examined produced extreme differences in cleanup criteria and costs (a range of 7 to 1 in the latter, even without the use of the most extreme values). Even though these quantitative estimates are specific to this case study and the simplified remedial cost estimation procedures that we used, our basic assumption of remedial cost proportionality to the spatial extent of contamination is valid for many sites, and our conclusions are generally valid. In modeling PCB transport at a specific site, it is often impossible to determine the appropriate Kow to use without assessing the concentrations of each PCB congener and developing a site-specific Kow. If this is not done, and a Kow value is simply selected from an available database, one cannot readily judge how well the Kow value describes conditions at the particular site. If such a Kow selection is made in the modeling of a Superfund site, it may be presumed that citation of an EPA source (such as those listed in Table 1) would provide adequate justification for any of the values in the databases (since EPA has jurisdiction over such sites). However, since it is possible to select a value from EPA sources that gives such a wide range of outcomes (at least in our case study), the results have to be considered substantially arbitrary even assuming no arbitrariness in other parts of the process. Moreover, without standardization of the sources used for parameter values (i.e., selection of a single value for a particular chemical), the results at otherwise identical sites may also differ arbitrarily for the same reason. In a risk assessment, parameter uncertainties such as those described here can be dealt with in many ways. Probabilistic methods can be used to explicitly characterize uncertainty in values resulting from different measurement techniques or an unknown mixture of components at a site (35). Sitespecific calibrations can be used to narrow uncertainty distributions (37); however, these methods assume that the empirical inputs, including those from databases, are valid.
Measurement and modeling of specific congeners may reduce the uncertainties associated with poorly defined environmental mixtures of PCBs, but Kow values cited for individual congeners may still display a rather large degree of variability (2). Because certain PCB mixtures may be more likely to occur under certain environmental conditions, an appropriate Kow value might be selected from those available based on site-specific factors. However, little direction exists among the databases and regulatory guidance documents for performing and reviewing such a selection. While it may be inappropriate to use a single Kow value for modeling all PCBs in all environmental scenarios, there is a need to reduce the uncertainty and arbitrariness of selecting a Kow value from among those databases described above. Eganhouse and Pontolillo (4) concluded that, to a large extent, the lack of data quality procedures and the proliferation of erroneous data and references may be responsible for the wide range of Kow values for DDT and DDE reported in the literature and recommended by agencies. Our study highlights the significance of their conclusion. Rigorous data quality and peer review procedures are required to ensure a consistent use of values for Kow and other parameters in risk assessments and remedial action planning.
Acknowledgments We are grateful to Bruce Hope and Laura Green for comments and discussions. Discussions with John Wakeman helped in developing the case study presented in this paper. This work was supported by the U.S. Chamber of Commerce.
Supporting Information Available Additional database information, calculation methods, and variables. This material is available free of charge via the Internet at http://pubs.acs.org.
Literature Cited (1) Renner, R. The Kow Controversy. Environ. Sci. Technol. 2002, 36, 411A-413A. (2) Illustrated Handbook of Physical-Chemical Properties and Environmental Fate for Organic Chemicals, Volume 1; Mackay, D., Shiu, W. Y., Ma, K. C., Eds.; Lewis Publishers: Chelsea, MI, 1992. (3) Pontolillo, J.; Eganhouse, R. P. The Search for Reliable Aqueous Solubility (Sw) and Octanol-Water Partition Coefficient (Kow) Data for Hydrophobic Organic Compounds: DDT and DDE as a Case Study. U. S. Geological Survey Water-Resources Investigations Report 01-4201; USGS: Reston, VA, 2001. (4) Eganhouse, R. P.; Pontolillo, J. Assessing the Reliability of Physico-Chemical Property Data (Kow, Sw) for Hydrophobic Organic Compounds: DDT and DDE as a Case Study. SETAC Globe 2002, July-August. (5) EPA Superfund Record of Decision: Commencement Bay, Near Shore/Tideflats; EPA/ROD/R10-89/020; U.S. Environmental Protection Agency EPA ID: WAD980726368 OU 01, 05; Pierce County, WA, 1989. (6) PCBs in Sediments at Selected Sites in Puget Sound; Washington State Department of Ecology, Publication No. 02-03-003; WSDOE: Olympia, WA, 2002. (7) TrophicTrace: A Tool for Assessing Risks from Trophic Transfer of Sediment-Associated Contaminants. U.S. Army Corps of Engineers, Prepared by Menzie-Cura & Associates, Inc.: Chelmsford, MA, 2004. (8) Developing Health-Based Sediment Quality Criteria for Cleanup Sites: A Case Study Report; Publication No. 97-114; Washington State Department of Ecology: Olympia, WA, 1997. (9) Estimation Program Interface (EPI) Suite, Version 3.11; U.S. Environmental Protection Agency: Washington, DC, 2000. Available at http://www.epa.gov/opptintr/exposure/docs/ episuitedl.htm. (10) Interactive PhysProp Database; Syracuse Research Corporation: Syracuse, NY, 2004. Available at http://www.syrres.com/ esc/physdemo.htm. (11) CHEMFATE Chemical Search; Syracuse Research Corporation: Syracuse, NY, 2004. Available at http://www.syrres.com/esc/ chemfate.htm. VOL. 39, NO. 18, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
6921
(12) Technology Transfer Network Clearinghouse for Inventories & Emission Factors: Water9; U.S. Environmental Protection Agency Office of Air Quality Planning and Standards: Washington, DC, 2004. Available at http://www.epa.gov/ttn/chief/ software/water/index.html. (13) Technology Transfer Network Clearinghouse for Inventories & Emission Factors: Chem9. U.S. Environmental Protection Agency Office of Air Quality Planning and Standards: Washington, DC, 2004. Available at http://www.epa.gov/ttnchie1/software/ chem9/. (14) Superfund Chemical Data Matrix Methodology Appendix A; U. S. Environmental Protection Agency: Washington, DC, January 28, 2004. Available at http://www.epa.gov/superfund/sites/npl/ hrsres/tools/scdm.htm. (15) Superfund Chemical Data Matrix - User Version, Windows Version 1.0, (SCDMWin); U.S. Environmental Protection Agency Office of Emergency and Remedial Response: Washington, DC, September 1997. (16) Superfund Public Health Evaluation Manual (Superfund PHEM); EPA/540/1-86/060; U. S. Environmental Protection Agency Office of Emergency and Remedial Response: Washington, DC, October 1986. (17) Soil Transport And Fate Database 2.0 and Model Management System; U. S. Environmental Protection Agency Office of Research and Development: Washington, DC, 1991. Available at http://www.epa.gov/ada/csmos/models/stf.html. (18) Human Health Risk Assessment Protocol for Hazardous Waste Combustion Facilities; EPA/530/D-98/001A; U.S. Environmental Protection Agency Office of Solid Waste and Emergency Response: Washington, DC, 1998. (19) The PBT Profiler. U.S. EPA Office of Air Quality Planning and Standards: Washington, DC, 2004. Available at http://www. epa.gov/oppt/pbtprofiler/. (20) ATSDR. Toxicological Profile for Polychlorinated Biphenyls (PCBs); U.S. Department of Health and Human Services, Agency for Toxic Substances and Disease Registry: Washington, DC, 2000. (21) Comparative Risk Assessment of Dredged Material Management Alternatives, Moss Landing Harbor, Moss Landing, California, Peer Review Draft; U.S. Army Corps of Engineers, San Francisco District: CA, 2003. (22) Gobas, F. A. P. C. A Model for Predicting the Bioaccumulation of Hydrophobic Organic Chemicals in Aquatic Food-Webs: Application to Lake Ontario. Ecol. Modell. 1993, 69, 1-17. (23) Gobas, F. A. P. C.; Z’Graggen, M. N.; Zhang, X. Time Response of the Lake Ontario Ecosystem to Virtual Elimination of PCBs. Environ. Sci. Technol. 1995, 29, 2038-2046. (24) Burkhard, L. P. Comparison of Two Models for Predicting Bioaccumulation of Hydrophobic Organic Chemicals in a Great Lakes Food Web. Environ. Toxicol. Chem. 1998, 17, 383-393. (25) Briggs, D. E. G.; Kear, A. J. Decay and Preservation of Polychaetes: Taxonomic Thresholds in Softbodied Organisms. Paleobiology 1993, 19, 107-135. (26) Lemieux, H.; Blier, P. U.; Dufresne, F.; Desrosiers, G. Metabolism and Habitat Competition in the Polychaete Nereis virens. Marine Ecol. Prog. Ser. 1997, 156, 151-156.
6922
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 18, 2005
(27) Schrock, M. E.; Barrows E. S.; Rosman, L. B. Biota to Sediment Accumulation Factors for TCDD and TCDF in Worms from 28Day Bioaccumulation Tests. Chemosphere 1997, 34, 1333-1339. (28) Integrated Risk Information System. Records for Polychlorinated Biphenyls, Aroclor 1254; U.S. Environmental Protection Agency: Washington, DC, 2004. Available at http://www.epa.gov/ iris/subst/0294.htm (PCBs), and 0389.htm (Aroclor 1254). (29) Exposure Factors Handbook, Volume I: General Factors; EPA/ 600/P-95/002Fa. U.S. Environmental Protection Agency, Office of Research and Development: Washington, DC, 1997. (30) PCBs: Cancer Dose-Response Assessment and Application to Environmental Mixtures; EPA/600/P-96/001F; U.S. Environmental Protection Agency, National Center for Environmental Assessment, Office of Research and Development: Washington, DC, September 1996. (31) Risk Assessment Guidance for Superfund, Volume 1 - Human Health Evaluation Manual, Part A, Interim Final; EPA/540/ 1-89/0002; U. S. Environmental Protection Agency, Office of Emergency and Remedial Response: Washington, DC, 1989. (32) Tolls, J.; Bodo, K.; De Felip, E.; Dujardin, R.; Kim, Y. H.; MoellerJensen, L.; Mullee, D.; Nakajima, A.; Paschke, A.; Pawliczek, J. B.; Schneider, J.; Tadeo, J. L.; Tognucci, A. C.; Webb, J.; Zwijzen, A. C. Slow-stirring method for determining the n-octanol/water partition coefficient (Pow) for highly hydrophobic chemicals: Performance evaluation in a ring test. Environ. Toxicol. Chem. 2003, 22, 1051-1057. (33) Tulp, M. T. M.; Hutzinger, O. Some thoughts on aqueous solubilities and partition coefficients of PCB, and the mathematical correlation between bioaccumulation and physicochemical properties. Chemosphere 1978, 7, 849-860. (34) Linkov, I.; von Stackelberg, K.; Burmistrov, D.; Bridges, T. S. Ecological Risk Assessment: Uncertainty and Variability from Trophic Transfer in Management of Contaminated Dredged Sediments. Sci. Tot. Environ. 2001, 274, 255-269. (35) Burkhard, L. P. Factors influencing the design of bioaccumulation factor and biota-sediment accumulation factor field studies. Environ. Toxicol. Chem. 2003, 22 (2), 351-360. (36) Iannuzzi, T. J.; Harrington, N. W.; Shear, N. M.; Curry, C. L.; Carlson-Lynch, H.; Henning, M. H.; Su, S. H.; Rabbe, D. E. Distributions of key exposure factors controlling the uptake of xenobiotic chemicals in an estuarine food web. Environ. Toxicol. Chem. 1996, 15 (11), 1979-1992. (37) Baseline Ecological Risk Assessment for the Hudson River Remedial Investigation/Feasibility Study; U. S. Environmental Protection Agency, Prepared by Menzie-Cura & Associates and TAMS Consultants, Inc. July, 1999.
Received for review September 14, 2004. Revised manuscript received May 2, 2005. Accepted July 12, 2005. ES0485659
