The purpose of this chapter is twofold. First, it describes the algorithms used to determine exposure media concentrations of the dioxin-like compounds. Discussion of the algorithms are structured around four "source categories." These categories roughly translate to beginning points, or origins, of contamination. The source categories are also the basis for the example scenarios described in Chapter 5. Second, it provides information about all the model parameters and justification for the values selected for the demonstration of methodologies in Chapter 5. Parameter discussions appear immediately following descriptions of modeling methodologies.

Section 4.2 provides an introduction to the type of modeling used in this assessment. Section 4.3 describes the algorithms used for the first source category, on-site soil, where the contaminants occur in surface soils, and this contamination source and subsequent exposure occur at the same site. The second source category, described in Section 4.4, is termed off-site soil. The contaminated soil is remote from the exposure, such as in a landfill impacting a nearby residence. Section 4.5 describes algorithms to determine exposure media concentrations resulting from stack emissions, the third source category. Chapter 3 laid the groundwork for this section by describing the use of air dispersion/deposition models as applied to a point source to generate two key quantities: air-borne contaminant vapor phase concentrations at a site of exposure, and particulate phase deposition rates. Section 4.5 describes how modeled concentrations and depositions translate to soil, vegetative, and water concentrations. Section 4.6 concludes the chapter with a discussion of algorithms specific to the fourth source category, point-source effluent discharges into surface water bodies.

Algorithms are presented which estimate exposure media concentrations for: 1) surface soils, 2) surface water impacts: suspended and bottom sediment and dissolved phase concentrations, 3) air including the vapor phase and in particulate form, and 4) biota including beef, milk, fruit and vegetables, and fish.

Literally hundreds of fate and transport models have been published which differ widely in their technical sophistication, level of spatial or temporal resolution, need for site specific parameterization, and so on. This makes selection of the most appropriate one for any particular situation very difficult. EPA has published model selection criteria documents (EPA, 1987b; EPA, 1988d) and a software system (Integrated Model Evaluation System, IMES, Version 2.01, 1992, Office of Health and Environmental Assessment, Office of Research and Development, U.S EPA) to help assessors with model selection.

Relatively simple, screening level models are used to model fate, transport, and transfer of dioxin-like compounds from the source to the exposure media in this assessment. Simple assumptions are often made in order to arrive at the desired result, which is long-term average exposure media concentrations. Perhaps the most critical of the assumptions made is the assumption that the source strength remains constant throughout the period of exposure: the initial soil concentration of dioxin-like compound remains the same for that exposure period, and stack emissions and effluent discharges remain steady throughout this period. It is important to understand that EPA is not endorsing the algorithms of this assessment as the best ones for use in all dioxin assessments. They are suggested as reasonable starting points for site-specific or general assessments, and as will be discussed shortly, most multi-media exposure modeling has included similar screening level approaches. The assumptions behind models are described carefully throughout this chapter. If these assumptions do not apply to a particular situation, or where assessors require more spatial or temporal resolution, more complex models should be selected. References to other models are made in this and other sections throughout the chapter.

Finally, it cannot be overemphasized that measured concentrations are generally more reliable than modeled ones. Assessors should use measured concentrations if available and if such measurements can be considered spatially and temporally representative for the exposed populations.

The first examples of similar multimedia compartment modeling were probably the "fugacity" models proposed by Mackay (1979) and Mackay and Paterson (1981, 1982). Fugacity in this context is defined as the tendency for a chemical to escape from one environmental media compartment into another. The fugacity of a chemical present in an environmental media compartment is modeled using common fate and transport parameters such as octanol water partition coefficients, Henry's Constants, water solubilities, and so on. The fugacity concept is based on the fact that at equilibrium, equal fugacities are established in all compartments of a system. Examples of fugacity modeling include the transfer of nonionic organic chemicals between the atmosphere and surface water (Mackay, et al., 1986), between the atmosphere and plants (Riederer, 1990), and for food chain modeling (Travis and Hattemer-Frey, 1987). A definitive text on multimedia compartment modeling using the fugacity approach has recently been published (Mackay, 1991). One possible drawback for the fugacity approach applied to the types of source categories discussed in this assessment is that it does not consider spatial variability of concentrations within a compartment. For example, air concentrations vary depending on the distance from a source of air emissions, such as a stack or a site of soil contamination. The fugacity approach would typically treat air as a single compartment with a uniform concentration.

The transfer of contaminants between compartments and multimedia modeling approaches have been extensively studied at the National Center for Intermedia Transport at the University of California, Los Angeles. Their multimedia compartment model, MCM (Cohen and Ryan, 1985), provides several useful algorithms for intermedia transfer factors that would have application for dioxin-like compounds. More recently, this group has introduced the spatial multimedia compartment model (Cohen, et al., 1990), which allows for non-uniformity in some compartments. Such a model would be more suitable for the types of source categories of this assessment, since there is non-uniformity within a compartment as noted above in the air compartment example.

An early approach which merged simplistic multimedia modeling with human exposure was termed the exposure commitment method, developed by Bennett (1981). An exposure commitment is defined as a contaminant concentration in human tissue. Exposure commitments are calculated from transfer factors that are estimated as the ratios of the steady-state concentrations of a contaminant in adjoining compartments of an exposure pathway. An example of adjoining compartments is air to plants to livestock to diet. This method has been applied to both PCBs (Bennett, 1983) and 2,3,7,8-TCDD (Jones and Bennett, 1989). These applications have required measured concentrations of the contaminants in different compartments in order to estimate the transfer factors. The retrospective nature of this approach limits its usefulness for general applications.

One of the early multimedia models which also had human exposure as the endpoint, but did not require retrospective data, was the GEOTOX model (McKone and Layton, 1986). This model had air (vapor and particle phases), water (surface and ground water, including bottom sediments of surface water bodies), soil (soil gas, water, and solid subcompartments), and biomass (eggs, milk, meat, fish, and vegetation including food crops) compartments. The most recent evolution of this model can be found in McKone and Daniels (1991).

Multimedia modeling approaches have been extensively used to evaluate the exposure to dioxins. Paustenbach, et al. (1992) evaluated the exposure and risk to humans from residential and industrial soil contamination by 2,3,7,8-TCDD. Simple models were used to estimate the concentrations of 2,3,7,8-TCDD in air-borne suspended particulates and fish that reside in nearby streams impacted by the contaminated soil. Together with concentrations in contaminated soil, Paustenbach evaluated human exposures via soil ingestion, dermal contact, particulate inhalation, and fish consumption. They also used Monte Carlo techniques on exposure parameters (in contrast to fate and transport parameters) to determine a range of residential and industrial soil concentrations that would result in a specified risk level. The risk level chosen for their demonstration was 10^-5, which was determined by multiplication of the Lifetime Average Daily Doses (LADDs in mg/kg-day) and the cancer slope factor for 2,3,7,8-TCDD of 9700

(mg/kg-day)^-1 derived by Keenan, et al. (1991). Residential soil concentrations less than 20 ppb did not pose a lifetime cancer risk greater than 10^-5. For industrial sites, concentrations in soil that could pose a 10^-5 risk ranged between 131 and 582 ppb, depending on the amount of time the industrial worker spend outdoors under typical exposure conditions.

Travis and Hattemer-Frey (1991) evaluated human exposure to 2,3,7,8-TCDD from a broader perspective. The principal assumption of the Fugacity Food Chain model used for Travis' human exposure assessment is that atmospheric concentrations of 2,3,7,8-TCDD can be empirically linked to water, soil, and vegetative concentrations, which in turn are linked to agricultural produce, meat, milk, eggs, and fish concentrations. Simple models for atmospheric depositions onto plants, air-to-leaf transfers of vapor phase 2,3,7,8-TCDD onto plants, transfers to cattle beef and milk, and other models, are presented. They also compared their model predictions of exposure media concentrations to literature values, and concluded that their approaches resulted in concentrations comparable to those found in the literature. This effort by Travis and Hattemer-Frey is examined in more detail in Section 5.6 of Chapter 5.

Exposure to 2,3,7,8-TCDD using simplistic multimedia models has also been assessed for specific sources. Goeden and Smith (1989) evaluated the impact to fish and subsequent human exposure by consumption of fish to dioxins and furans emitted by a resource-recovery facility. Surface water sediment concentrations in a lake were estimated as a simple weighted average of concentrations on three kinds of particles entering the lake: soil via erosion whose concentration was estimated given contaminated particle depositions onto soil (and considering mixing and soil half-lives), deposition of background uncontaminated suspended particulates directly onto the lake, and direct deposition of contaminated particles onto the lake. Fries and Paustenbach (1990) also evaluated the impact of incinerator emissions of 2,3,7,8-TCDD, but they evaluated human exposure via consumption of food crops, meat, and milk. EPA (1990d) used a simple dilution model to evaluate the impact of pulp and paper mill effluent discharges of 2,3,7,8-TCDD and 2,3,7,8-TCDF into surface water bodies.

This is only a cursory summary of the wealth of multimedia modeling approaches that are available, and the application of such modeling approaches for evaluating human exposure to 2,3,7,8-TCDD. While there are many similarities and differences among the approaches, they all share one characteristic in common - they have all been described as "screening level models". Without attempting a definition of the qualifier, "screening level", such a qualifier for these models seems to imply the following types of common features: assumptions of equilibrium and/or steady state conditions between compartments, lack of substantial (if any) spatial or temporal resolution, the use of biotransfer or bioconcentration concepts which simply relate an environmental concentration (air or water concentration, e.g.) to a biomass concentration (plant or fish concentrations), and so on. A counterpoint to screening level models might be what are termed "site-specific" or "mechanistic" models. Such models are more theoretically sophisticated, contain more spatial and temporal resolution, attempt to simulate actual mechanisms of fate and transport rather than depend on empirical relationships developed from data, could involve complex food chain approaches to model biomass concentrations (to counter the simple biotransfer or bioconcentration approaches), and generally are highly parameterized requiring site-specific data that is often not readily available.

Because of the complexity of the multimedia environment, modeling of contaminant fate in such an environment has tended to remain simple. However, there are complex models which can be applied to smaller subsets of the multimedia environment, and which have been applied to assessments of dioxin-like compounds. One example is the COMPDEP model, which was used in this assessment to evaluate the impact of stack emissions of dioxin-like compounds. That model allows for complexities of terrain, varying weather patterns, vapor/particle partitioning, etc., to be considered. That model is further described in Chapter 3. Another example of more complex modeling was the use of the WASP4 model in a comprehensive evaluation of bioaccumulation of 2,3,7,8-TCDD in Lake Ontario (EPA, 1990b). That application required a substantial amount of site-specific parameterization.

With the exception of the COMPDEP model, the models used for this assessment are better described as screening level rather than mechanistic or site-specific. Many of the algorithms used are the same or very similar to the ones found in references above. Except for the effluent discharge source category, which uses a non-spatially resolved dilution model for surface water impacts, the algorithms do consider spatial differences between the source and site of impact or site of exposure. For example, the algorithm estimating surface water impacts from a site of soil contamination, while simple in its framework, does incorporate the following: the area of the site that is contaminated, the area of the watershed which drains into the water body, the erosion rates of the site of contamination as well as the rest of the watershed, the proximity of the site to the water body, the concentration of the contaminant at the site of contamination as well within the watershed other than the contaminated site, the lipid content of the fish, and the organic carbon fractions of the suspended and bottom sediments of the water body. Assignments for all these parameters impact water and fish concentrations, and it is certainly arguable that they are all site-specific parameters. From this perspective, it could be argued that most of the algorithms of this assessment are generally screening level in their theoretical sophistication, but site specific in their application.

Sections in other chapters of this volume address key issues relating to the use and credibility of the algorithms described in this chapter. Chapter 5, which demonstrates the methodology, makes observations concerning exposure media concentrations in Section 5.6.1. Chapter 6, on User Considerations for use of the models and algorithms of this assessment, discusses categorization of model parameters and conducts sensitivity analysis exercises on key fate, transport, and transfer algorithms. Chapter 7 on Uncertainty also has critical discussions including: when possible, comparison of exposure media concentration estimations of 2,3,7,8-TCDD made in the demonstration scenarios with literature values, comparison with alternate modeling approaches, and general discussions of parameter assignment uncertainty and algorithm uncertainties.

Figures 4-1 through 4-4 are flow diagrams showing interim compartment concentrations modeled, and principal processes modeled and assumptions made in the intermedia transfer. Sections 4.3. through 4.7 describe the algorithms for the four source categories considered in this assessment, and background and assignment of parameters for the demonstration scenarios of Chapter 5.

As earlier noted, the contamination and exposure occur at the same site for this source category. The contamination is assumed to originate at the soil surface. As such, the soil itself is the exposure media for the dermal contact and soil ingestion pathways. Sections 4.3.1 through 4.3.4 describe the algorithms for estimating concentrations of the dioxin-like compounds in: bottom sediment, suspended solids, and in the dissolved phase in the water column of surface water bodies (4.3.1), in the air in the vapor phase (4.3.2) and particulate phase (4.3.3), and in biota including fish (4.3.4.1), home-grown vegetables and fruit (4.3.4.2), and beef and milk (4.3.4.3).

The principal assumption in the algorithm estimating the impact to surface water and surface water sediments (suspended and bottom sediments) from an area of contaminated soil is that such an impact is correlated to surface soil concentrations at that site as well as surface soil concentrations within a larger area draining into the water body. This drainage area is commonly referred to as a watershed. Further, the impact to the water body is assumed to be uniform. This tends to be more realistic for smaller water

bodies as compared to large river systems. Other key assumptions in the surface water impact algorithm are:

• Soil erosion estimates, coupled with sediment delivery ratios, can be used to describe the impact of a contaminated site relative to other soils in the watershed which contribute sediments to the water body;

• The sorption of dioxin-like compounds onto surface soil, suspended solids and bottom sediments is principally a function of the contaminant's organic carbon partition coefficient, Koc, and the organic carbon content of soils and sediments;

• The concentration of contaminants in soil eroding from a site are initially higher than the concentrations at the site itself - it is "enriched" with contaminants. This enrichment occurs because some processes of transport, such as wind erosion or soil erosion, favor lighter soils (silts and clays) which have higher surface area to volume ratios (more binding sites) as well as higher organic matter contents on the average (which also favors more binding of organic chemicals). Other processes such as volatilization or degradation may counteract the enrichment noted at the edge of a site - concentrations on soil entering a water body may be less than those leaving the site;

• The concentration of contaminants in sediment suspended in the water column exceeds the concentration in bottom sediments. Similar reasoning as the above enrichment argument applies: particulates which remain in suspension tend to be lighter and more enriched with organic matter as compared to particulates which settle to the bottom of water bodies. It should be noted that suspended solids, in this algorithm, are simply a reservoir into which dioxin-like compounds sorb; more complex models consider sorption onto more than one reservoir of suspended materials including suspended particulates and dissolved organic matter;

• Suspended and bottom sediments originate principally as soil erosion; a mass balance is maintained such that a part of the soil reaching the water body through erosion remains as suspended particulates, and a part settles to bottom sediments.

• A steady state is achieved between concentrations in the dissolved phase in the water column, concentrations in the sorbed phase in the water column, and concentrations in bottom sediments;

• Volatilization out of the water body or degradation of residues in the water body are not modeled. Neglecting these dissipation processes has the net effect of overestimating water body impact. On the other hand, bottom sediment resuspension is not modeled. Not modeling resuspension would have a net effect of underestimating water column impacts; and

• Estimating the average impact to the water body, rather than a localized impact which may be the case if the contaminated soil is very near the water body, is suitable for purposes of this assessment procedure.

Concentrations in bottom sediment are desired because fish concentrations are estimated as a function of bottom sediment concentrations (see Section 4.3.4.1). Concentrations in suspended solids are desired because they are used to estimate bottom sediment concentrations, and dissolved phase concentrations are needed for estimating drinking water exposures.

The solution begins with the mass balance statement:

The mass of contaminant An amount which remains as dissolved in

entering the water body = the water column + An amount which remains

sorbed to suspended materials + An amount which

remains sorbed to particles settling to the bottom

This can be described mathematically as:

where:

C_swb = concentration on soil entering water body, mg/kg

ER_w = total watershed annual soil erosion, kg/yr

C_wat = dissolved-phase concentration in water column, mg/L

V_wat = water body annual volume, L/yr

C_ssed = concentration on suspended sediment, mg/kg

M_ssed = mass of suspended sediment introduced per year, kg/yr

C_sed = concentration on sediment settling to bottom, mg/kg

M_sed = mass of bottom sediment introduced per year, kg/yr

Other equations based on assumptions stated above and needed for this solution are:

1) mass balance of soil is maintained:

where:

f_s = fraction of annual erosion remaining as suspended materials, unitless

2) equilibrium between sorbed and dissolved phases is maintained; suspended sediments are enriched in comparison to bottom sediments:

where:

Kd_ssed = soil-water partition coefficient for contaminant in suspended sediment, L/kg

OC_ssed = fraction organic carbon in suspended sediment, unitless

OC_sed = fraction organic carbon in bottom sediment, unitless

Now, Equations (4-2) through (4-6) can be substituted into the right hand side of Equation (4-1) so that this side can be function a one concentration, C_ssed, and one erosion amount, ER_w. Factoring out C_ssed then gives:

The bracketed quantity in the right hand side of Equation (4-7) can be termed f , so that C_ssed can be solved as (C_swb ER_w)/f . Now, the numerator in this term can be expanded to describe contaminant contributions by a site of contamination and contaminant contributions by the rest of the watershed. Included in this solution is the assumption made above that soils eroding into water bodies are "enriched":

where:

C_swb = concentration on soil entering water body, mg/kg

ER_w = total watershed erosion, kg/yr

C_s = contaminated site soil concentration of dioxin-like compound, mg/kg

E = enrichment ratio, unitless

SL_s = unit soil loss from contaminated site area, kg/ha-yr

A_s = area of contaminated site, ha

SD_s = sediment delivery ratio for soil eroding from contaminated site to water body, unitless

C_w = average concentration of dioxin-like compound in effective area of watershed not including contaminated site, mg/kg

SL_w = average unit soil loss for land area within watershed not including contaminated site, kg/ha-yr

A_w = effective drainage area of watershed; the area contributing sediment which mixes with the sediment originating from A_s, ha

SD_w = sediment delivery ratio for watershed, unitless

Finally, the right hand side of Equation (4-8) can be termed, r , and the concentration in suspended sediment, C_ssed, is equal to r /f . All the terms in r /f are input parameters or can be solved as a function of input parameters. Other water body concentration terms, C_wat and C_sed, can now be solved using Equations (4-5) and (4-6). Note that this solution is most applicable to small water bodies to ponds and streams. The differences in the two water systems can be expressed in the parameters, effective watershed area, A_w, water body volume, V_wat, and organic carbon contents of suspended solids and bottom sediments, OC_ssed and OC_sed. Guidance on these terms and assignment of values for the demonstration scenarios in Chapter 5 is now given.

• C_s and C_w: These are concentrations of dioxin-like compounds in the contaminated site soil, C_s, and the average within the effective area of the watershed, C_w. The contaminated site concentrations drive the concentrations assumed for most exposures, and is a principal user input (for the on-site source category, the contaminated site is also the site of exposure). The simplest assumption for C_w is that it is 0.0. However, examination of soil data from around the world shows that, where researchers have measured concentration in what they described as "background" or "rural" settings, soil concentrations of PCDDs and PCDFs are in the non-detect to low ng/kg (ppt) range. Example Scenarios 1 and 2 in Chapter 5 demonstrate the on-site source category. For these example scenarios, C_w and C_s are both initialized at 10^-6 mg/kg (1 ppt) for all example compounds representing low concentrations that might be possible for basin-wide areas.

• E: Enrichment refers to the fact that erosion favors the lighter soil particles, which have higher surface area to volume ratios and are higher in organic matter content. Therefore, concentrations of organic contaminants, which are a function of organic carbon content of sorbing media, would be expected to be higher in eroded soil as compared to in-situ soil. While enrichment is best ascertained with sampling or site-specific expertise, generally it has been assigned values in the range of 1 to 5 for organic matter, phosphorous, and other soil-bound constituents of concern (EPA, 1977). The enrichment ratio would be expected to be higher in sandy soils as compared to silty or loamy soils because the finer silt particles which erode from a soil generally characterized as sandy are more a deviation from the norm compared to silt particles which erode from a soil generally characterized as silty or loamy. The example scenarios in Chapter 5 modeled mid-range agricultural loam soils (as modeled with organic carbon fractions, soil loss parameters as discussed below, etc.). The enrichment ratio will therefore be assigned a value of 3.0 in all circumstances.

• SL_s and SL_w: These are the unit soil loss, in kg/ha, from the exposure site and the average from the effective land area draining into the surface water body. In the simplest case, the unit losses can be considered equal. In the most complicated solution, the effective drainage area can be broken up into "source areas", where each source area can be unique in terms of the erosion potential, concentration of contaminant, and so on. The total contribution equals the sum of contributions from each source area, as: S C_j*SL_j*A_j*E_j*SD_j for the right hand side of Equation (4-8) for j number of source areas not including the exposure site. For direct input into Equation (4-8), the terms C_j, SL_j, A_j, E_j, and SD_j, should be determined and C_w, SL_w, E_w, and SD_w should be estimated as weighted averages over all source areas, A_j. The effective drainage area, A_w, would be the sum of all source areas, A_j.

For the example scenarios in Chapter 5 demonstrating the on-site source categories, SL_s and SL_w are assumed equal. This generally assumes that erosion parameters for the site of exposure mirror the averages for the drainage area. Also, the enrichment ratio, E, is assumed to be constant for all watershed soils. For the off-site soil source category, the site of contamination is assumed to have different erosion characteristics. The following is offered as general guidance and background for estimation of unit soil losses in this assessment.

The unit soil loss is commonly estimated using the Universal Soil Loss Equation. This empirical equation estimates the amount of soil eroding from the edge of a field (Wischmeier and Smith, 1965):

where:

SL = average annual soil loss, Eng. tons/acre-year

R = rainfall/runoff erosivity index, t-ft/ac-yr

K = soil erodibility factor, t/ac-(unit of RF)

LS = topographical factor, unitless

C = cover and management practice, unitless

P = supporting practices factor, unitless.

Several references are available to evaluate USLE factors for agricultural and non-agricultural settings (EPA, 1977; USDA, 1974; Wischmeier, 1972; Novotny and Chesters, 1981). For this assessment, values for these terms will based on assumptions about contaminated sites and rural soils. Justification and assumptions are given below. It should be noted that more sophisticated models are available for estimating erosion rates (i.e., CREAMS as described in Knisel, 1980), and should be considered in actual site-specific assessments.

• Rainfall/erosivity index, R: The R term represents the influence of precipitation on erosion, and is derived from data on the frequency and intensity of storms. This value is typically derived on a storm-by-storm basis, but it has been compiled regionally for the development of average annual values (EPA, 1977). Annual values range from < 50 for the arid western United States to > 300 for the Southeast. The value used in this assessment will be 160, which is typical of rainfall patterns seen in much of the midwestern United States.

• Soil erodibility, K: The soil erodibility factor reflects the influence of soil properties on erosion, with values ranging from <0.05 for non-erodible sandy soils to >0.50 for highly erodible silty soils. The value used in this assessment will be 0.30, which is typical of, for example, sandy or silty loam soils with 2% - 4% organic matter contents.

• Length-slope factor, LS: The topographic factor reflects the influence of slope steepness and length of the field in the direction of the erosion. Steeper slopes and longer lengths lead to higher LS values, with a range of 0.1 for slopes <1.0% and lengths <100 ft to >2.0 for slopes generally >10%. The two key considerations for its assignment, therefore, are the size of the field for which erosion estimates are being made, and the slope of that field. The example scenarios in Chapter 5 had field sizes of 0.4 ha (1 ac) for a rural residence, 4 ha (10 ac) for a small rural farm, and 10 ha (25 ac) for an off-site contamination site. Guidance for use of the Universal Soil Loss Equation stops short of defining appropriate sizes of field for which unit estimates are to be derived, except that the USLE was developed for agricultural "fields" where cover, slope, soil type, etc. are assumed to be uniform. For purposes of estimating erosion losses in this assessment, a field of 4 ha for estimating the LS factor will be used. In a rural watershed with agricultural and non-agricultural settings, this would be a reasonable average area of uniformity. If square shaped, a 4 ha area translates to a side length of 200 m. For purposes of assignment of the LS factor, it will be assumed that the contaminated site has a 2% slope. EPA (1977) (and other references as noted above) show nomagraphs giving the LS factor as a function of slope length and slope. With a 200 meter slope length and a 2% slope, the LS factor is approximately 0.20. This factor will be used for all soil loss estimates required in this estimates.

• Support practice factor, P: The support practice factor reflects the use of surface conditioning, dikes, or other methods to control runoff/erosion. P can be no greater than 1.0. However, values less than 1.0 should only be assigned when specific practices are employed which are designed to reduce erosion. For the example scenarios in Chapter 5, it will be assumed that no such practices are in place at the site of concern or throughout the watershed to control erosion. Therefore, a value of 1.0 will be assumed.

• Management practice factor, C: The final term in the USLE is the cover and management practice factor, C, which primarily reflects how vegetative cover and cropping practices, such as planting across slope rather than up and down slope, influences erosion. C values can be no greater than 1.0, with this value appropriate for bare soils. A C value of 1.0 is an appropriate choice for active landfills or sites of high soil contamination (like Superfund sites) mostly devoid of vegetation. For an inactive landfill with grass cover or any area with dense vegetative cover such as grass, a value of 0.1 or less is appropriate. Values greater than 0.1 but less than 0.7 are appropriate for agricultural row crops, which offer more protection than bare soil, but not as much protection as dense vegetation. Three erosion estimates are required for scenarios demonstrated in Chapter 5. One is for areas of high soil contamination, or the scenario demonstrating the "off-site" source category. It will be assumed that the off-site contaminated site is largely devoid of vegetation in this case, and a value of 1.0 will be assumed. A second erosion estimate is needed to characterize average unit soil loss throughout a watershed draining into a surface water body. The example scenarios are based on a rural setting which has agricultural and non-agricultural (i.e., rural residences) areas. The C value in this circumstance will be assumed to be 0.3. Finally, an soil erosion estimate is needed in the algorithm transporting contaminated soil from an area of high soil contamination to a nearby site of contamination, as part of the algorithms developed for the "off-site" source category. In this case, the land between a site of soil contamination and the nearby site of exposure will be assumed to be covered with dense vegetation, such as grass. In this case, the C value will be 0.1.

As just described, three unit soil loss estimates are required for this estimates and the difference between the three will be expressed in the C term. Multiplication of the five USLE terms gives unit soil loss estimates of 9.60 (with C = 1.0), 0.96 (with C = 0.1), and 2.88 (with C = 0.3) t/ac-yr. The value of SL_s and SL_w for the demonstration of the on-site scenario in Chapter 5 is 2.88 t/ac-yr. Since Equation (4-8) and other uses of unit soil loss estimates are needed in kg/ha-yr, these unit losses are easily converted to 21515, 2152, and 6455 kg/ha-yr.

• A_s and A_w: These are the area terms, including the area of the contaminated site, and the effective drainage area of the watershed, both in ha. The scenarios demonstrated in Chapter 5 have assumed 0.4 ha (1 acre roughly) for exposure sites described as rural residences, 4 ha (10 acres) for farms, and 10 ha (25 acres) for an off-site area of soil contamination. If the area of contamination is at the site of exposure, as in the "on-site" source category, then A_s should be assigned an area equalling the site of exposure (and the concentration term, C_s, should equal the average soil concentration over this site of exposure). If the area of contamination is away from the site of exposure, as in the "off-site" source category, A_s should equal the total area of contamination (and again C_s should equal the average soil concentration over this area).

The total area impacting a river system has been termed a watershed. For purposes of this assessment, an "effective" drainage area will almost always be less than the total area of a watershed. A "watershed" includes all the land area which contributes water to a river system. For large river systems, this area is in the order of thousands of square miles and includes several tributaries and smaller streams feeding into the main branch of the river. Each stream and tributary has its own sub-basin, whose sediment originates from a land area much smaller than thousands of square miles. If the contaminated site lies within that sub-basin, that it would be appropriate to include only the area within that sub-basin as the effective drainage area. This is one circumstance where an "effective drainage area" would be less than a total watershed area. Another consideration for determining the effective drainage area is the positioning of the contaminated site with respect to the point where water is extracted for drinking and fish are caught for consumption. If these points are significantly upstream in the river system in relation to the contaminated site, there is no reason to conclude that sediments or water near where the water is extracted are impacted by the contaminated site. If these withdrawal points are downgradient of the contaminated site, then there is reason to believe that sediments and water are impacted. However, if they are downgradient from the contaminated site but not at the bottom of the watershed, then sediment and water quality further downgradient from the withdrawal points is not of concern and land draining into these downgradient portions would not be part of the "effective drainage area". One further possible consideration is how far upgradient in the watershed one should go when determining the size of the effective drainage area. While sediments introduced at the furthest points may eventually work their way down to the mouth of the watershed, this may take geologic time and not recent historic time. Therefore, sediment quality near a site of contamination need not consider these far reaches.

For a standing water body such as a lake or a pond substantially fed by ground water recharge, an assumption that probably should be made using the simple framework of this assessment is that all sediments within the lake/pond are completely mixed. Therefore, the effective area should equals all area around the lake/pond contributing sediment, and, as in the above discussion on river systems, a part of the land area contributing sediments to streams or rivers which may feed the standing water body. From this discussion, it is clear that determination of an effective drainage area depends on site specific considerations, but it will likely be less than the total watershed area. For purposes of demonstration, the effective drainage area, A_w, will be assumed to be 4,000 hectares (10,000 acres, 15.6 mi²). Furthermore, it will be assumed that the water body in question is part of a river system, which mainly impacts the assignment of the total suspended solids parameter, TSS (as discussed below). This assignment is not based on any specific sites that have been studied. It is only justified as being a reasonable size for dilution of contaminated soils which originate from contaminated site. Given the other area terms discussed above, 0.4, 4, and 10 ha, then the assignment of 4000 ha would appear to add a substantial amount of clean soil for mixing considerations.

A useful data source for this term and the suspended sediment term below, for specific sites in the United States, is Appendix F in Mills, et al. (1985). This appendix includes a compilation of data from river and reservoir sediment deposition surveys, including total drainage area, water body volumes, and rates of sediment deposition (mass/area-time). A caution in using this and similar data bases when evaluating specific sites is that, again, these total drainage areas are just that, total areas. Water bodies in this data base are located in the 48 conterminous states. An estimate of suspended sediment concentrations can be made using the water volume and the sediment deposition rates from this data, and an assumption on sediment deposition velocity. The specific weight of sediments in the water body, also supplied in this appendix, can be used to estimate sediment deposition velocity.

• SD_s and SD_w: These are the sediment delivery ratios applied to the exposure site and the watershed as a whole. Such a ratio is required because not all the soil which erodes from an area deposits into the receiving water body. The following delivery ratio was proposed for construction sites (EPA, 1977):

where:

SD_s = sediment delivery ratio from site of interest, unitless

DL = distance from site to receiving water body, m

3.28 = converts m to ft (empirical equation was developed for units of ft).

Note that the sediment delivery empirical equation simplifies all land features pertinent to erosion to a function only of length. The equation was developed to estimate sediment loads from construction sites to nearby surface water bodies, and from distances up to 250 m (800 ft, roughly). Without specific information on the sites from which it was developed, it is assumed that the land area between the construction sites and the receiving water body is "average" and this relationship can be used for applications other than construction sites.

As noted in previous bullets, the example scenarios demonstrating the on-site source category assumed C_s = C_w, and SL_s = SL_w. The impacted water body was assumed to be 150 meters away from the site of contamination, also the site of exposure for the on-site source category. This distance translates to a delivery ratio of 0.26. Site-specific conditions could result in a larger (steeper slope, e.g.) or smaller proportion of the eroded soil being delivered to the water body than would be estimated with this equation.

Figure 4-5 shows a watershed delivery ratio as a function of watershed size (figure from Vanoni, 1975). As seen, the ratio decreases as land area increases. The total watershed size assumed for the example scenarios in Chapter 5 was 4,000 hectares, or 40 km². From Figure 4-5, this translates to a watershed delivery ratio, SD_w, of 0.15.

• f_s: As soil erodes into the water body, it will settle onto the bottom to become bottom sediment. Part of the settled material will become resuspended because of turbulent flow. The finest materials in eroded soil may not settle for a long time, and essentially always be in suspension. One way to arrive at the fraction of annually eroding material which remains in suspension ("remains in suspension" for purposes of discussion - in reality, little, if any, will remain in suspension, but will rather deposit and resuspend) involves complex modeling. A wealth of such models exist, such as those described in Wang (1989). The approach used here is more simple than those in Wang (1989).

If an average level of suspended material in the water were specified, in units of mg/L, what would be known with otherwise required parameters is the total amount of erosion reaching the water body (as discussed above) as well as the annual water volume

(discussed below). A required parameter for this assessment will therefore be the level of

suspended solids in the water body, TSS. With this parameter and the annual water flow volume, V_wat, the total suspended load equals, TSS (mg/L) * V_wat (L/yr). The assignment of these two terms are 10 mg/L and 1.524*10¹⁰ L/yr, leading to a total suspended load of

1.524 * 10¹¹ mg/yr, or 1.524*10⁵ kg/yr. Total erosion into the water body, in similar units, equals, A_s * SL_s * SD_s + (A_w - A_s) * SL_w * SD_w. With parameter assignments as discussed above, the total annual erosion equals 3.87 x 10⁶ kg/yr. Therefore, the fraction of total load that is suspended is 0.04 (1.524*10⁴/3.87*10⁶).

Given this formulation, the f_s term is not a model input value, but is solved on the basis of the other parameters noted.

• TSS: This is the total suspended sediment in the water body. This value will be lower for standing water bodies such as ponds or lakes as compared to streams or rivers. The more turbulent flow in rivers will suspend sediments to a greater degree than a relatively calm lake. A complex modeling exercise evaluating the impact of 2,3,7,8-TCDD to Lake Ontario assumed a suspended sediment concentration of 1.2 mg/L (EPA, 1990b). For use in pond or lake settings, an assumption of a suspended sediment concentration of 1-2 mg/L is reasonable. All example scenarios in Chapter 5 assume that the 4,000 ha watershed drains into a river suitable for supporting fish for consumption and water for drinking purposes. General guidance offered for the potential for pollution problems in rivers and streams as a function of average suspended sediment concentration are: 10 mg/L or less - no problem, 100 mg/L or less - potential problem, and greater than 100 mg/L - probable problem. A cutoff concentration for protection of aquatic life is 80 mg/L (Mills, et al., 1985). The value assumed for TSS for all example scenarios in Chapter 9 is 10 mg/L, indicating no turbidity problems and a river supportive of fish for consumption.

• V_wat: The stream in the example scenarios will be assumed to derive its annual flow only from the effective drainage area, A_w. This would imply that the scenarios are best described as sub-basins ( see the discussion on effective drainage area, A_w, above). Given the area of drainage, one way to estimate annual flow volume is to multiply total drainage area (in length squared units) times a unit surface water contribution (in length per time). The Water Atlas of the United States (Geraghty et al., 1973) provides maps with isolines of annual average surface-water runoff, which they define as all flow contributions to surface water bodies, including direct runoff, shallow interflow, and ground-water recharge. The range of values shown include 5-15 in/yr throughout the Midwest cornbelt, 15-30 in/yr in the South and Northeast, 1-5 in/yr in the desert Southwest, and a wide range of 10-40 in/yr in the far West. For this assessment, an assumed 15 in/yr is used to estimate the annual flow volume. Over a 4,000 hectare drainage area, total flow volume equals 1.524 x 10¹⁰ L/yr (15 in/yr * 0.0254 m/in * 4,000 ha * 10,000 m²/ha * 1000 L/m³).

• Kd_ssed: This adsorption partition coefficient describes the partitioning between suspended sediment and the water column. For numerous applications for organic contaminants, particularly for estimating the partitioning between soil and soil water, this partition coefficient has been estimated as a function of the organic carbon partition coefficient and the fraction organic carbon in the partitioning media:

where:

Kd_ssed = partition coefficient between suspended sediment and water, L/kg

Koc = organic carbon partition coefficient for contaminants, L/kg or cm³/gm

OC_ssed = fraction organic carbon content of suspended sediment, unitless.

The organic carbon partition coefficient, Koc, can be a measured value or it can be estimated. Schroy, et al. (1985) listed an organic solids/water partition coefficient of 468,000 for 2,3,7,8-TCDD. Information in Jackson, et al. (1986), imply that this is a very low partition coefficient for 2,3,7,8-TCDD. They obtained soil samples contaminated with 2,3,7,8-TCDD from 8 sites in the Times Beach area of Missouri, and 2 from industrial sites in New Jersey. These contaminated soils had 2,3,7,8-TCDD concentrations ranging from 8 to 26,000 m g/kg (ppb), and organic carbon contents ranging from 0.015 to 0.08. They determined soil water partition coefficients, Kd_s, for these soil samples, and using the organic carbon fraction data, estimated Kocs for 2,3,7,8-TCDD. The mean Koc from these ten samples was roughly 24,500,000. EPA (1990b) evaluated the Koc for sorption of 2,3,7,8-TCDD onto Lake Ontario sediments. They concluded that log Koc was greater than 6.3 (Koc = 2,000,000), but less than 7.3 (Koc = 20,000,000).

In the absence of measured values, the Koc can be estimated from a chemical's octanol water partition coefficient, Kow. Empirical equations relating Kow to Koc are listed in Lyman, et al. (1982). Of six different equations listed in that reference, the following derived by Karickhoff, et al. (1979) is used to estimate the Koc for the example compounds in Chapter 5:

where:

Koc = organic carbon partition coefficient, L/kg

Kow = octanol water partition coefficient, unitless

This equation was empirically developed from a limited number of hydrophobic contaminants (n=10, R² = 1.00). It implies that Koc is very similar to Kow for strongly sorbed compounds such as the dioxin-like compounds. Using the log Kow of 6.64 given in this assessment for 2,3,7,8-TCDD in Karickhoff's relationship estimates a Koc of roughly 2,700,000.

• OC_sed, OC_ssed: The organic carbon content of solids and sediments of water bodies are generally higher than organic carbon contents of the surrounding lands. Furthermore, organic carbon contents of suspended organic materials and solids are typically greater than those of bottom sediments. A significant sink for strongly hydrophobic contaminants such as the dioxin-like compounds is thought to be suspended, or non-settling, organic material. In modeling 2,3,7,8-TCDD in Lake Ontario (EPA, 1990b) using the WASP4 model, a compartment separate from suspended solids termed "non-settling organic matter" served as a permanent sink. For purposes of this assessment, a single reservoir of suspended materials onto which incoming dioxin-like compounds sorb is principally characterized by OC_ssed, and the values selected for OC_sed and OC_ssed should reflect the relative partitioning behavior of suspended and bottom materials. As noted above, these water body carbon contents are also related to the organic carbon contents of surrounding soils. The model parameter, OC_sl, is the soil organic carbon fraction and is required for modeling of soil contamination by dioxin-like compounds. Foth (1978) lists the organic nitrogen content of several soil types ranging from sand and sandy loam to clay. The range from that list is 0.0002 - 0.0024 on a fractional basis. Assuming a carbon to nitrogen ratio of 10 (Brady, 1984; who notes that C:N ratios vary from 8 to 15, with the typical range of 10 to 12), organic carbon ranges from 0.002 to 0.024. A soil organic carbon fraction, OC_sl, is assumed to be 0.01 for all example settings in Chapter 9, which is in the middle of this range. The organic carbon content of bottom sediments, OC_sed will be higher at 0.03. Bottom sediments originate as erosion from surrounding land, but also include decay of organic materials within water bodies. The organic carbon content of suspended materials can approach 0.20, but OC_ssed will be assumed to be 0.05 for the example settings in Chapter 5.

The algorithms for estimating vapor-phase concentrations of contaminants were presented and derived in Hwang, et al. (1986). These procedures were developed for soil surface and subsurface contamination with polychlorinated biphenyls, PCBs. The models are based on the assumptions that: 1) PCBs move through the soil primarily by vapor phase diffusion, i.e., leaching is not considered, 2) PCB vapor in the soil matrix reaches a local equilibrium with pore air, 3) degradation processes for PCBs were not considered , and 4) the PCB contamination occurs at the surface and extends down infinitely. These assumptions are similar to the general types of assumptions that have been made for all the algorithms estimating exposure media concentrations in this assessment. The procedures in that PCB assessment were also used for this assessment. Details of the derivation are presented in Hwang, et al. (1986).

The average flux rate over an exposure duration of ED can be estimated as:

where:

FLUX = average volatilization flux rate of contaminant from soil, g/cm²-s

E_slp = soil pore porosity, unitless

D_ea = effective diffusivity of contaminant in air, cm²/s

C_s = contaminant concentration in soil, ppm or mg/kg

H = Henry's Constant of contaminant, atm m³/mol

Kd_s = soil/water partition coefficient, cm³/g

ED = exposure duration, s

I = interim undefined term for calculation, cm²/s

= D_ea E_slp

E_slp + P_soil (1-E_slp) [Kd_s/(41 H)]

P_soil = particle bulk density of soil, g/cm³.

The effective diffusivity, D_ea, is solved as a function of contaminant diffusivity in air, and soil pore porosity:

where:

D_ea = effective diffusivity of contaminant in air, cm²/s

D_c = molecular diffusivity of contaminant in air, cm²/s

E_slp = soil pore porosity, unitless.

The soil adsorption partition coefficient, Kd_s, is given as:

where:

Koc = contaminant organic partition coefficient, L/kg

OC_sl = fraction organic carbon in soil, unitless.

It is noted in Hwang, et al. (1986) that this procedure would tend to overestimate emissions and resulting exposures in situations involving small spills which would not involve deep contamination. It is also noted that the average flux rate is inversely proportional to the square root of the duration of exposure - the longer the duration of exposure, the lower will be the average flux rate. Whereas this solution assumes an unlimited reservoir of contaminant, it is an unsteady state solution (unlike other solution strategies) and is essentially an average flux rate over an amount of time defined by the exposure duration. Inherent in the solution was the consideration that residues dissipate by volatilization at the surface layers, resulting in contaminants diffusing upwards from deeper soil layers over time. With this longer path of diffusion, volatilized amounts decrease, and hence the average flux over time also decreases.

Vapor-phase concentrations along the center (y=0.0) of an area source can be estimated from (Hwang, 1987):

where:

C_va = vapor-phase concentration of contaminant in air, m g/m³

FLUX = Average volatilization flux rate of contaminant from soil, g/cm²-s

a = side length parallel to the wind direction, m

U_m = mean annual wind speed, m/s

S_z = vertical dispersion coefficient in air, m

z = height of the exposed individual, m

erf = error function

e = error function term, unitless

= b/(2*(2*S_y)^.5)

b = side length perpendicular to the wind direction, m

S_y = horizontal dispersion coefficient in air, m

10¹⁰ = converts g/cm²-m to m g/m³.

This was the model used to estimate on-site vapor-phase concentrations. The dispersion terms, S_z and S_y can be estimated using site-specific wind rose data. In the absence of data, these terms can be estimated assuming the most common stability class, D, as:

where:

S_y,z = horizontal and vertical dispersion coefficient, m

X = distance upwind of the contaminated site, m.

Background on Koc and OC_sl were given in Section 4.3.1. above. Guidance for other terms in this algorithm now follow.

• E_slp: Porosity is defined as the pore space in soils occupied by air and water, and for sandy surface soils show a range of 0.35-0.50. Medium to fine-textured soils (loams, clays, etc.) show a higher range of 0.40-0.60 (Brady, 1984). Soil porosities in the example settings were 0.50.

• H: Henry's Constants were discussed in Volume 2, Chapter 2. The values of H used for the three example compounds were: for 2,3,7,8-TCDD - 1.65*10^-5 atm m³/mole; for 2,3,4,7,8-TCDF - 4.99*10^-6 atm m³/mole; and for 2,3,3',4,4',5,5'-HPCB - 1.00*10³ atm m³/mole.

• P_soil: Particle bulk density is defined as the mass of a volume of soil solids. This contrasts the more common parameter, bulk density, which is the mass of a unit of dry soil, which includes both pores and solids. Particle bulk density, P_soil, has a narrow range of 2.60 to 2.75, and for general calculation purposes, Brady (1984) recommends a value of 2.65 for average mineral surface soils, the value used for the example settings.

• ED: The exposure duration is simply the amount of time individuals are exposed. Two exposure durations were used in the demonstration scenarios, 9 years for "central" and 20 years for "high end" exposures. Used in this algorithm, and as discussed earlier, longer exposure durations translate to lower average volatilization fluxes. This presumes a soil concentration assumed to be uniform over depth starting at time zero, and to become depleted over time. The selected exposure durations of 9 (2.83*10⁸ sec) and 20 years (6.31*10⁸ sec) was used.

• D_c: Molecular diffusivities in air of the example compounds could not be found in the literature. However, diffusivities of one compound can be estimated from another with the following (Thibodeaux, 1979):

where:

D_a,b = Molecular diffusivities of compounds a and b, cm²/s

MW_a,b = Molecular weights of compounds a and b, g/mole

Thibodeaux (1979) lists the molecular diffusivity of diphenyl at 25 C at 0.068. Given the molecular weight of diphenyl of 154 g/mole, the diffusivities of the example compounds are: 2,3,7,8-TCDD (MW = 322) = 0.047; 2,3,4,7,8-TCDF (MW = 340) = 0.046; and 2,3,3',4,4',5,5'-HPCB (MW = 396) = 0.043.

• U_m: Mean annual windspeeds vary from between 2.8 and 6.3 m/s (EPA, 1985b). An assumption of 4.0 m/s in the absence of site-specific average wind speeds was made for the example scenarios of this assessment.

• a, b, z, and x: Simple assumptions can be made to assign values to the length terms above: a, b, z, and x. Assuming a square-shaped contaminated site, a equals b which equals the square root of the area of the site. A common assumption for z, the height of the exposed individual, is 2 m. The x term can be assumed equal to a side length (a or b), or can equal the side length plus the distance to the exposed individual if the contamination is not on-site and dispersion is modeled as "near field." For the residence and farm setting examples in Chapter 5, where the contamination was on-site, the x term was equal to a side length.

The method for determining the flux of soil particles due to wind erosion for on-site conditions was developed in EPA (1985b) based on Gillette's (1981) field measurements of highly erodible soils. A key assumption for this solution is that the soil surface is assumed to be exposed to the wind, uncrusted, and to consist of finely divided particles. This creates a condition defined by EPA (1985b) as an "unlimited reservoir" and results in maximum dust emissions due to wind only. This wind erosion flux is given as (EPA, 1985b):

where:

E_e = total dust flux of <10 m m particle due to wind erosion, g/m²-hr

V = fraction of vegetation cover, unitless

U_m = mean annual wind speed, m/s

U_t = threshold wind speed, m/s

F(x) = a function specific to this model.

EPA (1985b) provides details allowing for the application of this equation under a variety of circumstances. The following is offered as guidance specific to on-site conditions:

• V: For a "residence" or "farm", grass or crops are likely to substantially cover the soil, and the fraction of vegetative cover can range from 0.5 (minimal coverage) to 0.9 (more lush coverage). For the residence example settings, V was set at 0.9 which assumes a continual grass cover over the contaminated soil. The V for the farm settings was instead 0.5. The area of contamination for the example farm settings was larger than the residence setting, 10 acres to 1 acre. The land where crops were grown was also contaminated; the 0.5 value for V assumes that the cropland is totally or partially bare at some times - perhaps during spring land preparation and fall harvest.

• U_m, U_t: As given above in Section 4.3.2. on vapor phase air concentrations, the mean annual wind speed, U_m, assumed in the example scenarios was 4.0 m/s. The threshold wind velocity, U_t, is the wind velocity at a height of 7 m above the ground needed to initiate soil erosion. It depends on nature of surface crust, moisture content, size distribution of particles, and presence of non-erodible elements. It can be estimated on the basis of the following procedure (EPA, 1985b):

This is the wind speed measured at the surface needed to initiate soil erosion. EPA (1985b) shows how it can be determined as a function of soil aggregate size distribution. However, for the "unlimited reservoir" approach for which Equation (4-19) was developed, soil particles are assumed to be fine at 1.5 mm or less as an average. This translates to a threshold friction velocity of 75 cm/s and less. A value of 50 cm/s might be reasonably assumed to be representative of these types of surfaces, and was assumed for this assessment.

EPA (1985b) graphically shows the roughness height for a range of possible conditions. Included in this range are a roughness height of 0.1 cm for natural snow, 1.0 cm for a plowed field, 2.0-4.0 cm for grassland, 4.0 cm for a wheat field or for suburban residential dwellings, and up to 1000 cm for high rise buildings. The assumption made for the residence and farm example settings was 4.0 cm, following the information given for a wheat field or a suburban residence.

A chart provided by EPA (1985b) shows this ratio as a function of roughness height. For a roughness height of 4.0 cm, this ratio is seen to be 13.

This is simply the product of the ratio given in step 3 above and the friction velocity. Using values given above, 50 cm/sec * 13 = 6.5 m/sec.

• F(x): The model-specific function, F(x), is determined by first calculating the dimensionless ratio x, where x = 0.886 U_t/U_m, and finding F(x) from a chart of F(x) versus x, as provided in EPA (1985b). For U_t = 6.5 and U_m = 4.0, x = 1.44 and F(x) = 1.05.

The unit dust flux is easily converted to a total contaminant flux by multiplying by soil concentration and area:

where:

WE = contaminant wind erosion emission rate, g/s

E_e = total dust flux of <10 m m particle due to wind erosion, g/m²-hr

C_s = contaminant concentration in soil, ppb or ng/g

A_sc = area of contaminated site, m²

2.8*10^-13 = converts ng/hr to g/sec.

The next step in estimate particulate-phase contaminant concentration is to estimate the dispersion term. The model that is used is the same as the one used to estimate on-site vapor phase dispersion given in Equation (4-16) above. The following two changes obtain the correct "FLUX" term for use in Equation (4-16):

1) Instead of WE, a total flux term presented here in units (g/sec) consistent with other particulate flux terms discussed in this chapter, an appropriate "FLUX" for Equation (4-16) is a unit flux term: C_s*E_e (ng/m²-hr). Since the algorithm for E_e was developed for 10 m m size particles, the multiplication of E_e by C_s assumes that the concentration of contaminant on 10 m m size particulates is the same as that for the soil overall.

2) C_s*E_e is still not in the right units for Equation (4-16). The conversion term of Equation (4-16), 10¹⁰, should instead be, .00028.

Substituting C_s*E_e for FLUX, and .00028 for 10¹⁰ in Equation (4-16) will allow for the estimate of C_pa, the particulate phase concentration of contaminant in air, in units of m g/m³.

This section summarizes the algorithms to estimate contaminant concentrations in fish, vegetation (including vegetables for human consumption and pasture grass or fodder grown on contaminated soil for beef cattle consumption), beef, and milk. As will be shown, all algorithms are simple empirical equations which relate an environmental media concentration to a biota concentration, using a "biotransfer" or "bioaccumulation" factor.

The procedure and supportive data for the algorithm to estimate fish tissue concentrations can be found in Cook, et al. (1991), and more recently in an assessment of risk of 2,3,7,8-TCDD to aquatic life and associated wildlife (EPA, 1993) which EPA is conducting as part of its reassessment of dioxin-like compounds. The information in those reference focuses on 2,3,7,8-TCDD, although there is discussion on the related compounds covered in this assessment including other PCDDs, PCDFs, and PCBs. These compounds share a high degree of hydrophobicity that increases as the degree of chlorination increases. Cook, et al. (1991) note that this corresponds in general to an increase in lipophilicity and an increase in ability to bind to organic carbon in sediments and to dissolved organic matter in water. However, these tendencies are not paralleled by an increase in bioaccumulation. Only the 2,3,7,8-chlorine-substituted congeners are substantially bioaccumulated by fish, although large quantities of other PCDD and PCDF congeners are found in sediments. This pattern of bioaccumulation results because of higher rates of metabolism of PCDDs and PCDFs in fish as compared to the 2,3,7,8-chlorine-substituted congeners (EPA, 1992; Cook, et al., 1991, with references to Muir et al., 1986; Gobas, 1990). While the highly chlorinated 2,3,7,8-substituted congeners are very slowly accumulated, they have very slow elimination rates.

2,3,7,8-TCDD and other planar polyhalogenated aromatic hydrocarbons often have not been detected in water from aquatic ecosystems even when biota are highly contaminated. Because surface layers of bottom sediments are a good indicator of the relative amount of chemical in the system over the time scale involved for bioaccumulation of super-hydrophobic chemicals, a term known as the Biota to Sediment Accumulation Factor, or BSAF, has been offered as a measure of site-specific bioaccumulation potential. This term was recently proposed to replace equivalent terms which were known as the Bioavailability Index, or BI (Kuehl, et al., 1987; Cook, et al., 1991; EPA, 1990b), the Accumulation Factor, AF (Lake, et al., 1990) and the Biota to Sediment Factor, or BSF (Parkerton, et al., 1993; Parkerton, 1991; Thomann, et al., 1992). BSAF is defined as:

where:

BSAF = biota to sediment accumulation factor, unitless

C_lipid = concentration of contaminant in lipid of fish, mg/kg,

C_oc = concentration of contaminant in bottom sediment organic carbon, mg/kg

The surface water algorithms estimate concentration of contaminant in bottom sediments (see Section 4.3.1 above). This concentration, C_sed, can be converted to an organic carbon basis as a function of OC_sed:

where:

C_oc = concentration of contaminant in bottom sediment organic carbon, mg/kg;

C_sed = concentration of contaminant in bottom sediment, mg/kg;

OC_sed = fraction organic carbon in bottom sediment, unitless

The organic carbon content of bottom sediments was assumed to 0.03; see Section 4.3.1. for the derivation of C_sed.

Since the accumulation of contaminant is assumed to occur only in fish lipid, a correction term to estimate the whole fish tissue concentrations is needed since fish consumption in g/day refers to whole fish consumption. The correction term is simply f_lipid, and so whole fish concentrations are simply C_lipid * f_lipid.

The BSAF was developed as a measure of bioaccumulation potential rather than as a predictor, as it is being used here. It is uncertain as to whether measured BSAFs are generally applicable to other water bodies. Efforts are underway to evaluate the general applicability of BSAFs (P. Cook, Duluth Environmental Research Laboratory, US EPA, 6201 Congdon Boulevard, Duluth, MN 55804, personal communication). Using the BSAF approach as a predictive tool greatly underplays the complexity of the processes transferring contaminants from aquatic ecosystems to aquatic organisms. EPA (1993) provides a comprehensive discussion on aquatic impacts and processes for 2,3,7,8-TCDD and related compounds. Following are some of the key issues to consider:

1) Resident vs. Migratory Species: Parkerton (1991) applied a bioenergetics-based bioaccumulation model in an attempt to duplicate BSAFs for 2,3,7,8-TCDD found for carp and blue crabs in the Passaic River, New Jersey. He showed nearly a ten-fold difference in 2,3,7,8-TCDD BSAF calculated from data for resident species as compared to migratory species in the Passaic River. This would be expected for fish which also reside part of the time in relatively clean water bodies; migration would enable depuration of residues from fish. The possibility that migration patterns might explain some of the results for fish concentrations of 2,3,7,8-TCDD in the Lake Ontario bioaccumulation study was also raised (EPA, 1990b). That assessment also discussed a related issue of concern - to consider lakewide average sediment concentrations or concentrations near where sampled fish were captured in calculating the BSAF. Even within a large lake, more sedentary populations of fish may be impacted by localized contamination.

2) Past history of contamination: If contamination of surface water bodies with hydrophobic compounds like the dioxin-like compounds has occurred principally in the past, then it can be expected that most of the contamination occurs in or near the bottom sediment layer and not within the water column. Furthermore, if inputs to water bodies are declining or low in comparison to past loadings, then sediments would be undergoing depuration - residue levels would be declining, and the system may not be equilibrium. EPA (1990b) noted that very low BAF*s (defined as a fish to sediment ratio not including the sediment organic carbon and fish lipid considerations of BSAFs) and BSAFs for 2,3,7,8-TCDD in Lake Ontario contrasts higher BAF*s for other hydrophobic compounds such as DDE or PCBs. An explanation offered is that loadings to the Lake may be declining, such that there is a substantial disequilibrium between sediments, water, fish, and their prey. One parameter required in the bioenergetics model Parkerton (1991) used (referred to in the above bullet) was a ratio of contaminant concentration in bottom sediment to that in suspended sediment, rs/rw. In modeling exercises on the Passaic River, he found closer agreement between measured and predicted BSAFs with this ratio equal to 10 in contrast to 1, the only two values tested; a ratio of ten means that the concentration of contaminant in bottom sediment is ten times higher than it is in the suspended sediment. BSAFs predicted by the model were developed as the ratios in modeled fish lipid concentrations divided by modeled bottom sediment organic carbon normalized concentrations. Measured BSAFs used actual Passaic River fish lipid and bottom sediment concentrations of 2,3,7,8-TCDD. BSAFs predicted with this ratio equal to 1 were roughly 4 times as high as measured BSAFs, and BSAFs found with rs/rw equal to 10 were twice as high as measured. A related result of his modeling exercise was that, at best fit between modeled and measured BSAFs where the rs/rw was 10, dietary exposures explained over 50% of the BSAFs for carp and 85% in blue crabs, in contrast to water column exposures. He speculates that prey organisms consist of benthic animals which ingest contaminated bottom sediment. If the food chain begins near bottom sediments, and if food chain exposures are a principal explanation for fish tissue dioxin concentrations, than it follows that a model would perform better when bottom sediment concentrations drive fish tissue concentrations rather than water column concentrations, or equivalently, when rs/rw = 10. Finally, he notes that 2,3,7,8-TCDD contamination in Passaic river largely occurred as a result of historical loadings. The picture that emerges from Parkerton's modeling is as follows: sediments are serving as an internal source of contaminants due to past historical loadings, and the water column is in disequilibrium with bottom sediments and driven only by depuration of bottom sediment concentrations. The bioaccumulation of these compounds in carp and blue crabs appears to be mediated by trophic transfer via the benthic foodweb. In both the Lake Ontario and Passaic River studies, concentrations of 2,3,7,8-TCDD were higher in deeper bottom sediments as compared to surficial bottom sediments - this implies historical loadings and possibly depuration of surficial residues.

This issue is non-trivial for the methodology of this assessment, since an assumption for deriving suspended and bottom sediment concentrations is that the contamination is ongoing, and that the hypothetical water body may be closer to a state of equilibrium as compared to situations where contamination was principally in the past. The BSAF assumed for 2,3,7,8-TCDD in the demonstration scenarios of 0.09 is more in line with data from EPA (1990b) on Lake Ontario and from Parkerton (1991) from data in Passaic River, then with other data (presented later) where historical loadings are not as clear a principal source of bottom sediment contamination. The issue of ongoing versus historical contamination should be considered when assigning site-specific BSAFs.

3) Variations among fish species: Feeding habits, age, migratory patterns, and lipid contents (including lipid content of edible vs. inedible fish tissues) are just a few of the interacting factors which determine a site-specific BSAF as a function of fish species. The demonstration of this approach in Chapter 5 assigns a single BSAF to each of the three example contaminants. Although not unlike other simplifications of this assessment, such approaches are recognized as oversimplifications.

4) Study designs to obtain BSAFs: Although there is some evidence that BSAFs specific to a contaminant may be applicable to other aquatic settings, data to evaluate such a hypothesis is still sparse. Even data sets that do exist need to be carefully evaluated before deriving BSAFs. Such an evaluation should consider sediment as well as fish species data. Critical factors for sediment sampling include location, number, depth of sampling, variability, availability of organic carbon fraction information, and so on. Similar issues are pertinent for fish sampling and analysis.

Following now are guidance for the terms required for estimating fish tissue concentrations.

• BSAF: Table 4-1 summarizes literature from which biota sediment accumulation factors for dioxin and furan congeners could be developed. Only five sets of data were found in the literature. The data from the Wisconsin River (Kuehl, et al. 1987) and that from 1 lake in Sweden (Kjeller, et al. 1990) both show decreasing BSAF with increasing chlorination. The BSAF of 2.94 for 2,3,7,8-TCDD determined from a lake in Sweden should be questioned since it is more than an order of magnitude different than any of the other data. Causes for this discrepancy could be manifold. Some observations from Kjeller, et al. (1990) might shed some light on this result. Although sediment data was from three water bodies, 8 of the 9 Pike samples (pike samples were composites of 2-5 fish from one location in the water body) were from one of the water bodies. This is why only data from the one water body was summarized in Table 4-1. This water body, Lake Vanern, was clearly the most contaminated of the three water bodies studied. A paper mill was located at the northern part of this lake and the authors concluded that discharges from this mill impacted the lake. The average of 2,3,7,8-TCDD and 2,3,7,8-TCDF organic carbon normalized concentrations for five sediment samples from this lake was 297 pg/g; the analogous average concentration for 10 samples taken from another lake, Lake Vattern (6 samples), and a river, Dala (4 samples), was 65 pg/g. A similar disparity between Lake Vanern and the other water bodies is found with the penta-CDD/CDF concentrations: 205 pg/g vs. 108 pg/g, with similar comparisons for the hexa-, hepta, and octa-CDD/CDF. The sediment and corresponding pike sample nearest this mill had the highest concentrations reported - pike samples were given as 3000 and 833 pg/g lipid normalized 2,3,7,8-TCDF and 2,3,7,8-TCDF (a composite from 5 pike taken at this sampling station), respectively, and sediment was 1800 and 244 pg/g organic carbon normalized for 2,3,7,8-TCDF and 2,3,7,8-TCDD. Note the BSAF for 2,3,7,8-TCDD implied from this data point is 3.41. Another consideration for high BSAFs might be the source of contamination. Speculation from the Lake Ontario and Passaic River field data was that contamination principally occurred in the past, whereas in the Swedish data, contamination appears to have been ongoing at the time of sampling. This might be one indication that BSAFs for aquatic systems where contamination is ongoing might be greater than from systems where the contamination is primarily historical.

Table 4-1. Available Biota to Sediment Accumulation Factors, BSAF, for dioxin-like compounds.

Fish Water # Sed. samples

Reference/Congener Species Body # Fish samples BSAF Comments

Kuehl, et al., 1987 Carp Wisconsin 1/1 Laboratory flow through experiment using Wisconsin

River River sediment and Lake Superior water; BSAFs

determined from one "representative" sediment sample

2,3,7,8-TCDD 0.27 and one "composited" fish sample; sediment organic

2,3,7,8-TCDF 0.06 carbon and fish lipid contents given in article; no

1,2,3,7,8-PeCDD 0.06 other details provided.

1,2,3,6,7,8- &

1,2,3,4,7,8-HxCdd 0.035

1,2,3,6,7,8-HxCDF 0.037

1,2,3,4,6,7,8-HpCDD 0.0048

1,2,3,4,6,7,8-HpCDF 0.0033

US EPA, 1990b Lake Comprehensive field study on bioaccumulation of

Ontario 2,3,7,8-TCDD in Lake Ontario; BSAFs are estimated

2,3,7,8-TCDD Brown Trout 55/81 0.03 given 55 sediment samples and specific number of fish

Lake Trout 55/81 0.07 samples as noted; report evaluates matching fish with Smallmouth Bass 55/14 0.05 with sediment data from sites where fish were caught.

White Perch 55/38 0.20

Yellow Perch 55/77 0.03

Kjeller, et al., 1990

Pike Lake 4/6 Results presented at right derived from data in

2,3,7,8-TCDD Vanern 2.94 Kjeller, et. al (1990); data includes sediment

1,2,3,7,8-PeCDD in Sweden 1.03 samples from four sites in Lake Vanern and 6

1,2,3,4,7,8-HxCDD 0.17 composited (2-5 fish in composite) pike associated

1,2,3,6,7,8-HxCDD 0.086 with the four sites; pike concentrations reported

1,2,3,7,8,9-HxCDD 0.018 in article on a lipid basis; Lake Vanern is near a

OCDD 0.002 paper mill.

2,3,7,8-TCDF 1.40

12348/12378-PeCDF 0.25

2,3,4,7,8-HxCDF 0.71

(continued on next page)

Table 4-1. (cont'd)

Fish Water # Sed. samples

Reference/Congener Species Body # Fish samples BSAF Comments

Kjeller, et al. (1990) (cont'd)

123479/123478-HxCDF 0.036

1,2,3,6,7,8-HxCDF 0.065

1,2,3,7,8,9-HxCDF 0.27

2,3,4,6,7,8-HxCDF 0.047

1,2,3,4,6,7,8-HpCDF 0.0009

1,2,3,4,7,8,9-HpCDF 0.023

1,2,3,4,6,7,8-HpCDF 0.006

OCDD 0.0001

Parkerton, T.F. 1991 Passaic 7 "resident" fish species were best represented by

River carp; "migratory" species were eel and striped bass;

2,3,7,8-TCDD Resident fish 61/11 0.081 BSAFs at left are for 2,3,7,8-TCDD, given 61 bottom

Migratory fish 61/15 0.009 sediment samples and specific number of fish samples

Blue Crab 61/14 0.055 as noted; TCDD contamination attributed to historical

industrial input, particularly a 2,4,5-T plant operation

1940s to 60s.

Connecticut Department of 21 different Data supplied by CDEP (1992); complete data Environmental Protection (CDEP, 1992) water bodies description, study design, and interpretation in

Chapter 7, Section 7.2.3.2. CDEP established a

2,3,7,8-TCDD Carp 346/521 0.86 monitoring program to evaluate the impact of newly

2,3,7,8-TCDF Channel catfish 346/521 0.25 operating resource recovery facilities to soil,

2,3,4,7,8-PCDF White catfish 346/521 0.47 sediment, and fish. BSAFs at left are for the four

Total TEQ White sucker 346/521 0.24 congeners; they are derived from the average of 346 Brown bullhead sediment samples and 521 total fish samples covering

Yellow perch the six species noted at left.

(continued on next page)

Table 4-1. (cont'd)

Fish Water # Sed. samples

Reference/Congener Species Body # Fish samples BSAF Comments

US EPA, 1993 different Results compiled by EPA (1993) for 2,3,7,8-TCDD;

water bodies details can be found in each study:

2,3,7,8-TCDD Smelt 0.04 Batterman, et al. (1989)

Sculpin 0.12 same

Herring Gull 0.43 EPA (1990b) and Braune and Norstrom (1989)

Bullhead 0.05 Cook (unpublished) as listed in EPA (1993)

Sandworm 0.48 Rubinstein, et al. (1983)

Clam 0.93 same

Shrimp 0.73 same

The Swedish data also illustrates some of the complexities of interpreting literature data. First, the sediment data was expressed concentrations normalized to "sediment contents of organic material" (sic). This was interpreted as organic matter normalized, not organic carbon normalized. Parkerton (1991) assumed that organic carbon was 45% of organic matter to derive BSAFs when organic carbon data was unavailable; following this lead, organic matter normalized concentrations in Kjeller, et al. (1990) were divided by 0.45 to arrive at organic carbon normalized concentrations. Also, there was not an exact match in "sites" between sediment samples and fish samples; these sites were physical locations within the large lake where samples were taken. There were five sites where sediment samples were taken, and five sites where composited pike samples were taken in Lake Vanern. However, one of the sediment and one of the pike samples were from unique sites; only four sites had both sediment and pike samples. The results in Table 4-1 were derived using average sediment and pike concentrations from only these four sites. Another way to have derived BSAFs would be to average all lake sediment and pike concentrations; since there may be some relationship between sediment and pike concentrations based on lake location, it was decided to include only the four sites with both fish and sediment samples. Finally, there were two sets of results listed for 1,2,3,4,6,7,8-HpCDF as though there were two unique sets of analyses for the same congener; this is why there are two entries for this congener in Table 4-1.

A complete discussion of the data generated by the Connecticut Department of Environmental Protection (CDEP, 1992) is included in Chapter 7, Section 7.2.3.2. Generally, water bodies tested were mostly in rural/suburban settings rather than urban settings. Concentrations of 2,3,7,8-TCDD in surface soils and bottom sediments were in the low ppt level, indicating background impacts. BSAFs generated with that data were 0.24 to 0.85 for TEQs, 2,3,7,8-TCDD, 2,3,7,8-TCDF, and 2,3,4,7,8-PCDF.

Excluding the Swedish data, there are 26 reported BSAFs for dioxin-like congeners in Table 4-1. These range from 0.009 to 0.93, with lower BSAFs associated with higher chlorinated congeners. A BSAF of 0.09 will be assumed for 2,3,7,8-TCDD in the demonstration scenarios in Chapter 5. Although there is indications of declining BSAFs with increasing chlorination, there is probably not sufficient grounds to assign a BSAF for the second example compound, 2,3,4,7,8-PCDF, significantly different from that of 2,3,7,8-TCDD. The BSAF for this example furan will also be 0.09. In demonstrating the suite of dioxin-like congeners for the stack emission scenario, a profile of BSAF values is crafted generally reflecting the trend of lower BSAF for higher chlorinated congeners, but this profile cannot be rigorously defended, for obvious reasons.

It should be noted that all bioconcentration or biotransfer parameters, such as the BSAF, are qualified as second order defaults for purposes of general use. Section 6.2. of Chapter 6 discusses the use of parameter values selected for the demonstration scenarios, including a categorization of parameters. Second order defaults are defined there as parameters which are theoretical and not site specific, but whose values are uncertain in the published literature. The parameter values in this category should be considered carefully by users of the methodology.

EPA (1990b) estimates BSAFs for PCBs and other selected chemicals (DDE, HCB, etc.) for Lake Ontario from several data sets. Parkerton, et al. (1993) summarizes BSAFs for PCBs and other compounds from other water bodies using other data sets. A selected summary by water body taken from these two sources for PCBs is given in Table 4-2.

Two trends are apparent. First, the BSAFs for PCBs appear to exceed those of the dioxin and furan congeners by an order of magnitude and more. Second, and from limited data, it would appear that BSAFs increase from dichloro- through hexa- or perhaps hepta-chloro PCBs, and then decrease thereafter. An assignment of a BSAF for 2,3,3',4,4',5,5'-HPCB is not apparent from the data summary below. The data point from Siskiwit for the single heptachloro-PCB, which was 2,2',3,4',5,5',6-HPCB, was estimated by Parkerton (1991) as 12.5. The BSAF for flounder from New Bedford Harbor estimated by Parkerton (1991) was 0.84, with BSAFs for lobster and crab as 1.29 and 2.74, respectively. A value of 2.00 is assigned to 2,3,3',4,4',5,5'-HPCB for the example scenarios in Chapter 5.

Finally, it should be noted that these assignments are based on data on vertebrate rather than invertebrate aquatic species. It is generally recognized that invertebrates do not possess the enzymatic capability to metabolize hydrophobic compounds as effectively as higher chordates. As a result, invertebrate species such as mussels, clams, oysters, shrimp, crabs and lobsters may have BSAF values much higher than those observed for fish. Parkerton (1991) and Parkerton, et al. (1993) reviewed the literature to estimate BSAFs of 1 to 5 for species including grass shrimp, sandworms, deposit feeding clams, and blue mussel for PCDD/PCDFs and PCBs.

Table 4-2. Available Biota to Sediment Accumulation Factors, BSAF, for PCBs.

Fish Water

Congener Species Body BSAF Comments

PCB trout, salmon, Lake Ontario 1.40, 0.77 Compiled in EPA (1990b) from several data sources, years of study, and fish species.

perch, bass 0.52, 0.86, Summary in this table includes all uniquely derived BSAFs for PCBs for species noted.

3.35, 1.42, PCBs not further identified except BSAF value of 0.58 specific to Aroclor 1254.

trichloro-PCB lake trout, Siskiwit Lake 0.45-2.6 Compiled by Parkerton, et al. (1993) from data in Swackhammer, et al. (1988)

tetrachloro-PCB whitefish 0.71-1.3 and Swackhammer and Hites (1988); Parkerton presents data for individual pentachloro-PCB 3.4-9.4 congeners - summary at left aggregates by chlorination and includes both fish hexachloro-PCB 2.9-20.8 species; only one data point presented for heptachloro-PCB.

heptachloro-PCB 12.5

octachloro-PCB 2.2-12.7

Total PCB Three species Rio de La Plata 4.40 Determined by Parkerton et al. (1993) from Columbo, et al. (1990) on total PCBs;

of marine fish Argentina Columbo reference also has data on PCB IUPAC congeners 5-8, 14, 19, 28-31, 52,

101, 110, 138, 153, 180.

dichloro-PCB New Bedford 0.11-0.59 Compiled by Parkerton, et al. (1993) from data in BOS (1990); summary at left

trichloro-PCB Harbor 0.26-0.65 is the range of values specific to the PCB congener grouping, and averaged across

tetrachloro-PCB flounder, 0.65-1.02 noted species.

pentachloro-PCB lobster, crab 1.05-2.08

hexachloro-PCB 1.29-4.00

heptachloro-PCB 0.84-2.74

octachloro-PCB 0.23-1.17

nonachloro-PCB 0.02-0.38