PAGE 8 OF 9

4.4.3.3. Sites studied in the National Dioxin Study

The National Dioxin Study (EPA, 1987a) focused on sites of known or suspected contamination of soil by 2,3,7,8-TCDD. There were 7 "Tiers" of investigation, with roughly decreasing expectations of finding 2,3,7,8-TCDD. Tiers 1 and 2 included 2,4,5-TCP production and associated disposal sites (Tier 1) and sites where 2,4,5-TCP was used as a precursor in the manufacture of pesticidal products and associated disposal sites (Tier 2). These tiers had the highest expectation for finding 2,3,7,8-TCDD. There were originally thought to be 450 sites that would fall in Tiers 1 and 2, but after investigation, only 100 sites were included for study. Some were downgraded into Tier 3. Of the 100 sites studied, 20 were on or were proposed for inclusion in the Superfund National Priorities List. Tiers 3 and 5 were associated with 2,4,5-TCP formulation (Tier 3) and use (Tier 5). Tier 6 were organic chemical or pesticide manufacturing facilities were 2,3,7,8-TCDD was suspected of being present. Tier 4 included combustion sources and are not discussed further in this section. Tier 7, basically an examination of background areas, are also not discussed here.

Issues that are identified as important in fate and transport modeling for this subcategory of off-site sources include concentrations, the possibility of ground water contamination, and site-specific characterization.

These are discussed in turn.

. Concentrations:
Only 11 of the 100 Tier 1 and Tier 2 sites were eventually classified as requiring "no further action" because 2,3,7,8-TCDD soil concentrations were very low, < 1 ppb, or not detected (with detection limits generally at 1.00 ppb). Where it was detected, a general trend was to find very high concentrations where 2,4,5-TCP production wastes were stored or disposed of, with much lower concentrations at soils near these particular areas. At hot spots, concentrations were as high as 2,000 parts per million, but generally soil concentrations were in the parts per billion.

It was this parts per billion generalization that led to the assignment of a 1 ppb soil concentration for the demonstration of the off-site source category in Chapter 5. There were findings in the low ppb range for Tiers 3, 5, and 6, but at much lower frequency and no findings higher than the tens of ppb range. For exposure assessments, the characterization of soil concentrations in a site containing hot spots has to be carefully considered.

For site evaluations and proposed options for remediation, an areally weighted average might be considered, although this could dilute loss estimates depending on the area chosen - choosing a large area corresponding to property lines might, for example, lead to an "average" concentration orders of magnitude lower than concentrations found in hot spots. One approach which should be considered is a "hot spot" impact compared to an areally averaged impact. It should also be remembered that removal of highly contaminated soils is a common practice and another option for evaluation would be a concentration assuming hot spots are removed.

. Potential for Ground Water Contamination:
PCBs have been found in ground water in sites associated with dielectric fluids of transformers. Oils can migrate through soils as a separate immiscible phase and reach ground water, which has been the common explanation for PCB impacts to ground water. Ground water contamination by 2,3,7,8-TCDD has very rarely been found in ground water, although it has been released to the environment in an oil matrix. The Times Beach area of Missouri is the principal example of this release, where waste oils containing 2,3,7,8-TCDD were used for dust control.

Ground water sampling did occur in many of the Tier 1 and 2 National Dioxin Study sites, but the results were mostly non-detects. One occurrence at 0.18 ppt was noted for the Hyde Park site of Hooker Chemical in Niagara, NY, and a high of 1.8 ppb was found in an on-site monitoring well at National Industrial Environmental Services in Furley, KS. There were, however, numerous high occurrences in sub-soil samples in hot spot areas, in bottom sediments of evaporation lagoons, and so on, in the hundreds of ppb range.

There have been some limited experimentation showing different patterns of 2,3,7,8-TCDD migration in soils in the presence of solvents or in an oily matrix. Palusky, et al (1986) studied the mobility of 2,3,7,8-TCDD in soils associated with each of 6 solvents. Migration was found to be higher with aromatic solvents and chloroform in comparison to saturated hydrocarbons and methanol.

They speculated that the extent of migration related to the solubility of 2,3,7,8-TCDD in the solvent. Puri, et al. (1989) studied the migration potential of 2,3,7,8-TCDD in soil, water, and waste oil mixtures. Over time, they observed a reversible sorption pattern of TCDD, and concluded that a carrier medium with a significant amount of waste oil would play a dominant role in the movement of TCDD through soils.

. Site-specific Characterization:
In the case of landfills or sludge land application sites, the assignment of a soil concentration and an area can be made with some reasonableness. Such is not the case with the industrial contamination sites such as those studied in the National Dioxin Study, as briefly discussed above in the concentration bullet. Most of the sites studied in the National Dioxin Study were in the order of tens of hectares to below ten hectares.

On the other hand, the Dow Chemical site in Midland, Michigan is described as a site 607 ha in size (Nestrick, et al, 1986). That area corresponds to the size of the property, and the many soil sampling sites within that area were described as "background". Several of the pesticide formulator sites studied in Tier 3 were 2 hectares or less in size. Many of the them were extensively or partially paved with buildings, which complicate fate and transport modeling. Some of the Tier 5 sites of 2,4,5-TCP use were agricultural fields, which are less complicated to describe.

However, two sites were described as 2500 acres in size, which again is very large and makes assignment of an average soil concentration non-trivial. Other use sites were described as railyards and railroad rights of way. While estimates of loss into air could be made in complicated sites such as these, use of soil erosion modeling becomes very complicated if not undoable with paved areas, buildings, drainage ditches, roads, and the like.

4.5. ALGORITHMS FOR THE STACK EMISSION SOURCE CATEGORY

Contaminants emitted from incinerator stacks are transported in air and deposit on the exposure site, water bodies that may be used for drinking or fishing purposes, and on surrounding land. Chapter 3 describes the application of the COMPDEP (reference) model to obtain vapor-phase air concentrations and deposition rates of particles at a specified distance from an example stack emission source. These quantities are assumed to be given for purposes of discussion in this section; further discussion of the air transport modeling is given in Chapter 3.

Estimating soil concentrations based on particulate depositions follows a similar approach as estimating exposure site soil concentrations resulting from erosion of contaminated soil from off-site areas of contamination. Section 4.5.1. describes how soil concentrations are estimated given total (wet plus dry) deposition rates. Surface water impacts are assumed to result from direct deposition onto surface water bodies as well as erosion from the impacted effective drainage area. This solution is an extension of the solution given in Section 4.3.1. for the on-site source category, and is given in Section 4.5.2. Following now are bullet summaries for similarities and small refinements to algorithms previously discussed:

. Air impacts:
The atmospheric transport modeling described in Chapter 3 was comprised of two computer simulations: one which considered that emissions were in a vapor form and were transported as such, and one which considered that emissions were in particle form and likewise were transported as such. The result of the vapor-phase runs was a unitized ambient air concentration at various distances up to 50 km in all directions from the stack. Only the results in the predominant wind direction were used in this demonstration.

The result of the particle-phase runs were an ambient reservoir of air-borne contaminants sorbed to particulates (used only for inhalation exposures), and wet and dry deposition unit rates also at various distances up to 50 km. By "unitized", what is meant is that emissions for the vapor or particle runs can be thought of as "1" mass/time (g/sec) emissions. Results for all distances are linear with respect to this emission rate; that is, if the rate of vapor contaminant determined to be emitted is "5", than ambient air concentrations at any location are 5 times what they are when "1" is assumed to be emitted. The same holds true for emissions in the particle phase. Chapter 3 developed a framework for assigning a vapor and a particle fraction for specific dioxin congeners.

For example, 2,3,7,8-TCDD was assumed to have a vapor fraction of 0.55 (55% was in vapor form) and a particle fraction of 0.45. The final model results for air concentrations, and dry and wet deposition rates for all congeners, starting from these unit model runs and then incorporating congener-specific emission rates and vapor/particle splits, are given in Tables 3-12 to 3-17. The vapor-phase air concentrations were used to model vapor phase transfers in the vegetative bioconcentration algorithms. They were also used, summed with the simulated reservoir of particle-bound contaminants, to estimate the total reservoir of contaminant available for inhalation exposures.

. Vegetative impacts:
The rates of wet and dry deposition modeled by COMPDEL were used to determine vegetative impacts. The model for particle deposition impacts to vegetations is described in Section 4.3.4.2 above. Of course, this above section solves for dry deposition as a reservoir times a dry deposition velocity (for dry deposition), and as a reservoir times rainfall and a washout factor (for wet deposition); such a solution is not required for the stack emission source category since the deposition totals are estimated by the COMPDEP model. Other parameters for the vegetative model - the Bvpa (air-to-leaf vapor transfer factor), the Rw (fraction of wet deposition retained on vegetation surfaces), crop yields and interceptions, and the vegetative washout factor, kw, are used for the stack emission source category.

. Biota concentrations:
The algorithm estimating concentration in fish tissue based on bottom sediment concentrations is the same as in previous source categories. Modeled rates of contaminant deposition on particles onto the exposure site are used to estimate a "tilled" and an "untilled" soil concentration, as described below in Section 4.5.1. Underground vegetable concentrations are a function of tilled soil concentrations. The soil concentration used for cattle soil ingestion is untilled. Beef and milk concentrations are again a function of vegetative and soil concentrations, diet fractions, and bioconcentration and bioavailability factors as described in Section 4.3.4.3.

4.5.1. Steady-State Soil Concentrations

Chapter 3 describes the use of the COMPDEP Model to estimate the particulate phase deposition rates at the exposure site. This total deposition rate, F, includes both dry and wet deposition, and is used to estimate the steady state soil concentrations. The deposition of contaminated particulates from the air is assumed to be somewhat analogous to the process of eroding contaminated soil from an off-site source depositing on an exposure site. Specifically, the following assumptions are also made:

1) only a thin layer of soil becomes contaminated,

2) this layer is either "untilled" or "tilled", depending on surface activities, and

3) surface residues are assumed to dissipate with a half-life of 10 years corresponding to a first order decay rate of 0.0693 yr-1.

Considerations of upgradient erosion and exposure site soil removal are not made. Depositions occur over the exposure site and surrounding land area on an on-going basis. It might be said that upgradient soil concentrations are similar to exposure site concentrations at all times. Like the soil source categories, a tilled mixing depth of 20 cm is assumed.

However, an untilled mixing depth of 1 cm is assumed for this source category, in contrast to the 5 cm assumed for the off-site soil source category. It is felt that the process of erosion assumed to transport contaminated soil in the off-site soil source category to a site of exposure is a more turbulent process. It assumes that contaminated soils mix with "clean" soils that are between the site of contamination and the site of exposure.

In contrast, ongoing airborne deposition of particles is felt to be a less turbulent process impacting all watershed soils simultaneously; hence the assumption of a 1-cm mixing depth. The qualitative mass balance statement (similar to the one made above in Section 4.4.1, with _C equalling change in exposure site soil concentrations over time) can now be made as:

Diagram V3 4-4

This is mathematically stated as:

Equation V3 4-48

The solution to this equation is:

Equation V3 4-49

which computes C as function of time, t. Similar to the assumption made above in Section 4.4.1., the steady state solution for C is simply F/kM. The deposition rates supplied by the COMPDEP model are in units of g/m2-yr, so a conversion to mg/yr requires a multiplication by the land area of the exposure site and a multiplication of 1000 mg/g. Procedures to estimate M are given above in Section 4.4.1.

4.5.2 Surface Water Impacts

The solution for stack emission impacts to surface water bodies is an extension of the solution for soil contamination described in Section 4.3.1. Stack emissions deposit onto soils within the effective drainage area to result in an average basin-wide soil concentration. Soil erosion then delivers contaminants to surface waters as in Section 4.3.1. Stack emissions also directly deposit onto and impact the surface water body as well. All the assumptions laid out at the beginning of Section 4.3.1 apply here as well. New quantities needed for this solution include: a rate of contaminant deposition onto soils of the effective drainage area used to estimate average soil concentrations (such concentrations are estimated using the approach given in Section 4.5.1. above), a rate of contaminant deposition onto the water body, and a rate of particulate matter deposition onto the water body.

Equations (4-1) through (4-8) are now displayed again with these additions.

Equation V3 4-50

Mass balance and equilibrium equations continue:

Equation V3 4-51, 4-52, 4-53, 4-54, 4-55.

Substituting again as in Equation (4-7):

Equation V3 4-56

As before, the bracketed quantity in the right hand side of Equation (4-56) can be termed f , so that Cssed can be solved as (Cswb ERw + DEPc)/f . The numerator in this term can be expanded to describe contaminant contributions by the effective drainage area which has received depositions, the first quantity in the numerator, and to describe direct depositions, the second quantity:

Equation V3 4-57

Again as before, the right hand side of Equation (4-57) can be termed, r , and the concentration in suspended sediment, Cssed, is equal to r /f . Other water body concentration terms, Cwat and Csed, can now be solved using Equations (4-54) and (4-55).

Guidance on these terms and assignment of values for the demonstration scenarios in Chapter 5 is now given.

. Cswb and ERw:

Equation (4-57) shows all the terms necessary to arrive at an estimate of the annual contaminant entry into the water body via erosion, the Cswb * ERw term. Section 4.5.1 describes the algorithm to estimate soil concentrations given a deposition rate of contaminant. One deposition rate will be chosen to represent average deposition rates over the effective drainage area of the watershed (the effective drainage area is termed Aw). This rate will be the rate given in COMPDEP modeling at 0.5 kilometers, which implies that the water body and the effective drainage area into the water body are near the stack. Tables 3-15 and 3-16 (Chapter 3) display wet and dry deposition rates for this distance.

These rates are added to arrive at total deposition, shown in Table 3-17. Second, a representative mixing depth to characterize average watershed soil concentrations needs to be selected. Previous algorithms used a mixing depth of 20 cm for tillage activities, specifically home gardening, and 1 and 5 cm for non-tilled soil concentrations (1 cm for the stack emission and 5 cm for the off-site soil source category). For the sake of demonstration, it will be assumed that a representative watershed depth will equal 10 cm, which might be interpreted as an average of tilled and untilled lands within the effective drainage area. The values for SLw (6455 kg/ha-yr), Aw (4000 ha), ER (3), and SDw (0.15) were all given and discussed in Section 4.3.1. and will not be repeated here.

. DEPc:

The second quantity of Equation (4-57) describes the annual input to the surface water body that comes from direct deposition. This term is RDEPc * Awat * 1000, where RDEPc is the rate of contaminant deposition onto the water body, Awat is the area of the water body, and 1000 converts g to mg. The rate of contaminant deposition at 0.5 km will also be used to describe direct deposition impact to the surface water body, since for demonstration purposes, there is no justification for saying this distance is further or nearer the point of stack emission. The area of the water body has not been required for any other reason, and one will now be given. First, the effective drainage area of 4000 ha is relatively small and will result in a relatively small stream, at 1.524 x 107 m3/yr flow volume.

This volume is also equal to the average cross sectional area of the stream (m2) times stream velocity (m/yr). Assuming a stream velocity of 4.73 * 106 m/yr (15 cm/sec; 1/2 ft/sec), which is reasonable for a small stream, the cross sectional area is solved as 3.22 m2. An average 1 meter depth and 3.22 meter width appear reasonable. This width times the stream length would give stream surface area, Awat. Assuming a rectangular shaped watershed, dimensions of 40 ha wide by 100 ha long (to arrive at the 4000 ha effective drainage) seem reasonable. This length of 100 ha translates to 10000 meters, and the full surface area of the stream is 32200 m2. This will be the value assumed for Awat.

. DEPp:

The rate of particulate deposition onto the lake is required to achieve a mass balance of all annual soil erosion + particle deposition contributions to water body solids. The rate of particulate matter emitting from the stack and arriving at downwind locations was not supplied in Chapter 3. Instead, a literature value of 0.03 g/m2-yr developed by Goeden and Smith (1989) will be used. This value was based on modeling emissions from a resource recovery facility. Total particulate emissions from the stack were projected to be 4.63 g/s, and the deposition rate onto a nearby lake was modeled to be 0.03 g/m2-yr. No further information was supplied. Now, with the surface area as solved for above at 32200 m2, the total particle deposition, DEPp in kg/yr, is 966 g/yr.

. fs and fsd:

These are the fractions of total erosion and depositing particles remaining as suspended materials within a year. As discussed in the solution for the "on-site source category" in Section 4.3.1, fs was solved for as:

a value for total suspended solid, TSS of 10 mg/L, multiplied by a total flow volume Vwat of 1.524 x 1010 L/yr, divided by the total erosion into the water body, 3.87 x 1012 mg/yr. This resulted in an fs of 0.039. Note that this implies a total suspended load of 152,400 kg/yr. It could be assumed that the minuscule 1 kg/yr of particles directly depositing onto the stream remain in suspension during the year, on the basis of being smaller in size than eroded soil. This assumption will, in fact, be made, but it will be supported as follows.

In a quiescent water body, settling occurs through gravity and can be expressed in terms of Stokes Law:

Equation V3 4-58

For purposes of this discussion, a reasonable assignment of particle density of is 2.5 g/cm3 for depositing particles or eroding soil. Therefore, making substitutions, the right hand side of Equation (4-58) reduces to 918 d2.

Now, assumptions for the particle sizes of eroding soil and depositing particles can be made to arrive at a ratio of settling velocities, Vssoil/Vspart. The basis for assigning an enrichment ratio for delivery of contaminants via soil erosion was that fine-sized particles were the ones eventually reaching the water body via erosion. Lick (1982) states that a major fraction of the sediments (suspended and bottom) in the Great Lakes are fine grained, silts and clays, and that data from Lake Erie indicates that 90% of the sediments are of this category.

Brady (1984) shows USDA's classification of soils according to particle size, and gives a range of 0.0002 to 0.005 cm for silt sized particles and less than 0.0002 for clay size particles. The following assumptions are made to arrive at a representative diameter for particles in eroded soil: eroded soil is comprised of a 50/50 split of these two sized particles, silt-sized particles are, on the average 0.0026 cm in diameter, and clay size particles are 0.0001 cm in diameter. With these assumptions, the average particle size for eroding soil is 0.0014 cm.

The settling velocity for a 0.0014 cm particle is 1.8 x 10-3 cm/sec. In Section 3.4.3, Chapter 3, the argument was developed that 87.5% of the total emission rate of dioxin-like congeners would be associated with particles less than 2 m m. The basis of this argument was a surface area to volume ratio, with smaller particle sizes having significantly larger ratios. This does not mean that 87.5% of the 1 kg/yr depositing particles are of this size. However, for this discussion, the size of depositing particles will be assumed to be 2 m m (2 x 10-4 cm), since these size particles deliver most of the dioxin-like compounds to the water body (and the ultimate purpose of this exercise is to determine a value for the fraction of depositing particles which remain suspended and impact suspended sediment concentrations). The settling velocity, Vspart, is estimated as 3.7 x 10-5 cm/sec.

The ratio Vssoil/Vspart is about 50. Said another way and with all the assumptions and simplifications made above, depositing particles will remain in suspension 50 times longer than eroding soil in a quiescent water body.

Given this high a difference in settling velocities, it seems reasonable to assume fsb equals 1.0. The fraction of soil erosion remaining in suspension, fs, will be estimated given TSS, Vwat, etc., as before (see Section 4.3.1), only DEPp (the total amount of depositing particles, in kg/yr) will comprise a given increment of suspended materials when solving for fs.

. Vwat, OCssed, OCsed, and Kdssed:

These have all been discussed in Section 4.3.1.
The values for these parameters in the demonstration scenarios in Chapter 5 are:
Vwat = 1.524 x 1010 L/yr, OCssed = 0.05, OCsed = 0.03, and Kdssed = OCsed * Koc,

where Koc is the organic partition coefficient of the contaminant.

4.6. ALGORITHMS FOR THE EFFLUENT DISCHARGE SOURCE CATEGORY

As discussed in Volume II, Chapter 3, dioxin-like compounds can be released to waterways via various types of effluent discharges such as discharges from municipal waste water treatment facilities and pulp and paper mills using chlorine bleaching. Also discussed is the fact that these emissions have declined substantially in recent years, especially from pulp and paper mills. Since the procedures for considering point source discharges to waterways are somewhat different than those associated with the nonpoint source procedures for soil contamination and stack emissions, they are covered separately in this section. This source category is also different from others in that effluent discharges into surface water bodies are assumed only to impact fish and water.

The approach used in this report is an extension of the "simple dilution" model described in the Superfund Exposure Assessment Manual (EPA, 1988c). Other models are available which offer more spatial and temporal resolution than the model described here. One such model is the EXposure Analysis Modeling System, or EXAMS (Burns, et al., 1982, and Burns and Cline, 1985). The EXAMS and a simple dilution model were both applied in an assessment of effluent discharges from pulp and paper mills (EPA, 1990d). In this assessment, 98 of the 104 pulp and paper mills were modeled with both models using site-specific information (water body flow rates from STORET for all but 6 of the mills, effluent flow rates and contaminant discharges, etc.).

Three key quantities - one model result and two model parameters - led to a range of exposure conditions for humans consuming fish impacted by discharges from these pulp and paper mills: a water column concentration, a bioconcentration factor (BCF) applied to the water column concentration to get fish tissue concentration, and a fish ingestion rate.

The simple dilution model was used to estimate total water concentrations - i.e., mg TCDD total/L water. The EXAMS model was used to estimate dissolved phase water column concentration - i.e., mg TCDD dissolved in water column/L water. Then, with each set of water concentrations, two sets of exposure estimates (a low and a high estimate, in one sense) were generated - one with a BCF of 5,000 and a fish ingestion rate of 6.5 g/day, and one with a BCF of 50,000 and a fish ingestion rate of 30 g/day. Note that in deriving the range of results in that exercise, the BCF was applied to both a total and a dissolved phase water concentrations.

EPA (1993) discusses several bioconcentration/bioaccumulation empirical parameters for 2,3,7,8-TCDD, and makes the clear distinction for those which are to be applied to a total water concentration versus those applied to a concentration in the dissolved phase. The dilution and EXAMS model study indicated that the simple dilution model generally estimated higher water column contaminant concentrations compared to the EXAMS model, although this trend was not consistent among all water bodies modeled. The results from both models were comparable when the receiving water body had relatively low suspended solids concentration.

One key limitation of the EXAMS and the simple dilution model for use with dioxin-like compounds in aquatic systems is that they do not account for sediment transport processes. The EXAMS model was designed to determine the fate of transport of contaminants in the dissolved phase. Another spatially and temporally resolved model for this source category is the Water Analysis Simulation Package, the most up-to-date version termed WASP4 (Ambrose, et al., 1988). This model does include sediment processes and has been applied in a comprehensive evaluation of 2,3,7,8-TCDD bioaccumulation in Lake Ontario (EPA, 1990b). It requires extensive site-specific parameterization, but should be considered for more detailed site-specific evaluations of strongly hydrophobic and bioaccumulating contaminants such as the dioxin-like compounds.

The dilution model described below will be demonstrated in Chapter 5 with a set of data developed using site-specific data from the 104 pulp and paper mills of the 104-mill study. As will be discussed below, a hypothetical effluent discharge will have characteristics developed as the average of key characteristics from the 104 mill study. These key data include: flow rates of the receiving water bodies, suspended solids concentration in these receiving water bodies, effluent discharge flow rates, suspended solids in the effluent discharges, organic carbon content of solids in the effluent stream, and discharges of 2,3,7,8-TCDD.