Page 6 of 9

7.2.4. Alternate Modeling Approaches for Estimating Environmental and Exposure Media Concentrations

This section examines alternate modeling approaches for estimating environmental and exposure media concentrations. This is by no means a comprehensive examination, nor is its purpose to justify the models selected. If the models examined can be shown to be similar or to arrive at similar results as the models of this assessment, perhaps some validity for modeling and/or the models selected for this assessment can be gained.

7.2.4.1. An alternate approach for estimating bottom sediment concentrations from watershed soil concentrations

The dilution of contaminated sediments entering a river system can be estimated using an alternate approach. The average runoff rate for the midwestern U.S. is about 15 inches/year (Linsley, et al., 1982), the value used in this assessment for determining the flow rate of the receiving water body. For a 10,000-acre watershed (4,000 hectares; the watershed size and effective drainage area for the example scenarios in Chapter 5), this yielded a stream flow of about 17.2 ft3/sec.

The sediment yield can be estimated from the stream flow as follows (Linsley, et al., 1982): Qs = aQn, where Qs = sediment flow rate (Eng. T/yr); Q = stream flow rate (ft3/sec); and a and n are empirical constants, reflecting the vegetative cover in the watershed. Linsley, et al., (1982) recommend using a=3,500 and n=0.82 for coniferous forest and tall grass, and a=19,000 and n=0.65 for scrub and short grass. Substituting these into the equation above (and Q = 17.2 ft3/sec) gives an annual sediment flow rate of 36,000 to 121,000 T/yr.

Annual sediment flows will be assumed to consist only of soils which have eroded during the year. As such, sediments will be comprised of contaminated as well as uncontaminated watershed soils. A "contaminant concentration ratio" can be calculated by estimating the sediment contributed by the contaminated areas and dividing by this sediment flow rate range; this assumes all other sediment contributions are uncontaminated.

The annual soil loss for Scenario 3 demonstrating the off-site scenario was 9.6 T/ac-yr. The contaminated site area for Scenario 3 was 10 acres (4 ha). The total soil erosion contributed by this site equals: the unit soil loss * area * soil delivery ratio; for Scenario 3, this equals 9.6*10*0.26, or 25 T/yr. The contaminant concentration ratio is (25 T/yr)/(36,000 to 121,000 T/yr), or a range of 0.0002-0.0007.

This can be compared to a contaminant ratio estimated using current methodologies. The example Scenario 3 which had soil concentrations at 1 ppb resulted in a bottom sediment concentration in the nearby water body of 0.0016 ppb, which leads to a contaminant concentration ratio of 0.0016. This is higher than the ratio range noted above. It does incorporate an "enrichment ratio", however, which is the ratio of contaminant concentration on soils eroding from a field to soils within the field. It is given a value of 3 for the demonstration scenario. The ratio range noted above did not consider enrichment; if it had, the range would instead by 0.0006-0.0021. Now the modeled 0.0016 and this range are comparable.

7.2.4.2. An alternate modeling approach for estimating water concentrations given a steady input load from overland sources

A study to evaluate the bioaccumulation of 2,3,7,8-TCDD in fish in Lake Ontario included an extensive modeling exercise (EPA, 1990a). The model used was WASP4 (Ambrose, et al., 1988). This is a substantially more complicated model than used in this assessment. The underlying principal for the WASP4 model is a conservation of mass. Contaminant source terms, described in mass/time units, enter what are termed control volumes, or segments.

The contaminant partitions between sorbed, bound, and dissolved phases; it is not required to specify whether the contaminant enters via soil erosion, water runoff, surface deposition, or otherwise. Contaminants are, however, assumed to enter via the surface or as part of inflows to the water body, in contrast to ground water recharge. The mass transported into a segment is either transported out of the segment, accumulates in the segment, or is transformed by chemical or biological reactions.

As noted, 2,3,7,8-TCDD input into the Lake Ontario application partitions within the water column into a sorbed compartment, a dissolved compartment, and a bound compartment. This bound compartment is further described as non-settling organic matter. Three analogous compartments receive 2,3,7,8-TCDD in the bottom sediment layer. Several exchanges between the six compartments and contaminant losses within each compartment are modeled. For example, losses from water column compartments include downstream transport, volatilization and photolysis; the loss mechanism from the bottom sediment layer is sedimentation. Exchanges between compartments consider partitioning, diffusion, and sediment settling and resuspension.

This model requires substantial parameterization. Once values were selected for the Lake Ontario application, an evaluation was made on the impact of different levels of 2,3,7,8-TCDD input. Dynamic and steady state results were discussed. Principally examined for the steady state results were the concentrations of bottom sediment sorbed 2,3,7,8-TCDD and water column dissolved (soluble) phase 2,3,7,8-TCDD. A given level of steady 2,3,7,8-TCDD input, in kg/yr, resulted in a steady state concentration sorbed to bottom sediment and dissolved in the water column.

The premise in both the Lake Ontario steady state application of WASP4 and the water concentration algorithms in this assessment is that contaminants continue to enter water bodies over time unabated. Ground water entry of contaminants is not considered in either approach. Although a direct modeling comparison cannot be done, it is possible to slightly adjust the algorithms of this assessment to evaluate how results from a simple partitioning approach would compare with results from the complex fate and transport approach of the WASP4 steady state application.

Assume a surface water body is initially free of contaminant and at time t equals 1 day, a strongly hydrophobic contaminant, such as the dioxin-like compounds of this assessment, begins to enter a lake. Assuming the contaminant enters via soil particles, as in the approach of this assessment, it will then partition between those soil particles and surrounding water. The soil particles will slowly move toward the bottom of the lake at a rate described by a particle settling velocity. A settling velocity of 1 m/day is assumed in the Lake Ontario simulations.

The amount of time it takes to settle to the bottom once entering from the surface equals the lake depth divided by this settling time. The Lake Ontario depth was 86 m. Therefore, it might take 86 days to settle. This, of course, neglects resuspension of settled particulates. With this simplistic framework, a steady state amount coming into the lake after 86 days is matched by an amount depositing onto the lake bottom; the amount of contaminant within the water column has reached steady state. Water concentrations can then be estimated assuming equilibrium partitioning.

Results of sediment and water column steady state concentrations are described for any loading of 2,3,7,8-TCDD in the WASP4 steady state application; those loadings are described in kg/yr. Loadings in kg/yr are easily correlated to a steady state water column amount, given the above analysis. For example, a loading of 1.0 kg/yr could translate to a within water column steady state amount of 0.24 kg (1.0 kg/yr * (86 d)/(365 d/yr)).

This steady water column amount partitions between suspended sediment and surrounding water. First, the total concentration (sorbed + soluble) simply equals:

Equation V3 7-1

The dissolved phase portion of total is given by:

Equation V3 7-2

Parameters in this equation for the Lake Ontario WASP4 application include VOL, Koc, OCssed, and TSS. Lake Ontario volume was given as 1.68 x 1012 m3, Koc was estimated for the WASP4 application as 3,162,000, OCssed was estimated at 0.03, and TSS was estimated 1.2 mg/L. For a steady load of 1 kg/yr and a resulting LD of 0.24 kg, the steady state water column 2,3,7,8-TCDD concentration, using the simplistic approach described above, is estimated as 0.13 pg/L (ppq). The steady state water column concentration estimated by WASP4 given the same parameters and a load of 1 kg/yr is roughly 0.20 pg/L. An uncertainty analysis done with these WASP4 results concluded that 95% confidence limits around this prediction are 0.03 and 0.40 pg/L.

This would seem to imply that the simple partitioning approach used in this assessment compares favorably with the more complex fate and transport modeling assessment using WASP4, for Lake Ontario.

7.2.4.3. Estimating fish tissue concentrations based on water column concentrations rather than bottom sediment concentrations

EPA has prepared a document titled, "Interim Report on Data and Methods for Assessment of 2,3,7,8-Tetrachlorodibenzo-p-Dioxin Risks to Aquatic Life and Associated Wildlife" (EPA, 1993). That document provides details on the two key bioaccumulation parameters used for the methodologies of this document, the Biota Sediment Accumulation Factor, BSAF, used for the soil and stack emission source categories, and the Biota Suspended Solids Accumulation Factor, BSSAF, used for the effluent discharge source category. That document also discussed several water column based bioaccumulation factors, which are the focus of this section.

Before discussing these factors, it is noted that food chain modeling is a well developed alternate approach for estimating fish tissue concentrations of bioaccumulating contaminants (Thomann, 1989), which has also been applied to 2,3,7,8-TCDD (Parkerton, 1991). This approach is significantly more complex than the bioaccumulation/biotransfer approach of this methodology. It involves detailed site-specific characterizations, specifically the identification and transfer modeling between trophic levels of a food chain in a water body. Food chain modeling is a mechanistic approach, while the transfer approaches of this methodology are empirical. No judgement is rendered as to the relative merit of food chain models versus use of bioaccumulation coefficients. If detailed site-specific data is available, and given time and resources, assessors should consider food chain modeling for estimating fish tissue concentrations.

One water column measure which has been classically used is termed the Bioconcentration Factor, or BCF. Bioconcentration refers to the net accumulation of a chemical from exposure via water only, and BCFs are most often obtained in laboratory conditions. BCFs are defined as the ratio of the chemical concentration in organism (mass of chemical divided by wet weight of organism tissue) to that in water.

Another water column measure of the potential for a contaminant to accumulate in fish tissue is termed the Bioaccumulation Factor, or BAF. Bioaccumulation refers to the net accumulation of a chemical from exposure via food and sediments as well as water. Similar to the BCF, BAFs are defined as the ratio of the chemical concentration in the organism to that in the water.

For chemicals that are not strongly hydrophobic (unlike the dioxin-like compounds), the distinction between bioconcentration and bioaccumulation is small. Whereas food intake is generally a few percent of body weight per day, water passing over gills will equal hundreds to thousands times the organism weight per day, depending on species, activity, temperature, and other factors. Given this, the concentration of chemical in food must be 3 or more orders of magnitude greater than that in water before food can substantially contribute to uptake. EPA (1993) estimates that food intake becomes a critical contributor to the accumulation of contaminants in fish tissue for contaminants with log Kow of 5 and greater.

Since the dioxin-like compounds fall into this category, the remainder of this section will focus on the Bioaccumulation Factor. EPA (1993) defines steady-state lipid-based BAFs for total chemical in water and freely dissolved chemical in water (i.e., chemical which is truly in a dissolved phase and not bound to dissolved or suspended particulate organic materials) as:

Equation V3 7-3a 7-3b

EPA (1993) then develops relationships between ssBAFld and ssBAFlt, based on dissolved and particulate organic carbon reservoirs in the water column, and partition coefficients for these reservoirs. This is meaningful in complex modeling where these two reservoirs of organic carbon can be accounted for, such as in the WASP4 model. Alternately, EPA (1993) defines the TBFoc, a total binding factor to organic carbon, which empirically considers the reservoir of dissolved organic material (i.e., increases total binding and reduces truly dissolved phase concentrations) when such a reservoir is not explicitly modeled. The modeling frameworks in this assessment have only one compartment of suspended material to which contaminants sorb, with one associated organic carbon content. A second reservoir to which contaminants bind, the reservoir of dissolved organic material, is not modeled.

EPA (1993) developed a ssBAFlt and a ssBAFld for lake trout, 2,3,7,8-TCDD, and for Lake Ontario 1987 contamination conditions. The WASP4 model was used to model three hypothetical loading conditions that might have resulted in fish tissue concentrations observed in 1987: steady state loading, a steady state loading followed by a 90% reduction in annual loads for 20 years (i.e., 1968-1987), a steady state loading followed by a 100% reduction (i.e., no loading) for 20 years. The BSAF for lake trout estimated for 1987 data is given in EPA (1990a) as 0.07.

The BSAF is determined from measured bottom sediment concentrations and fish tissue concentrations; an assumption of historical loading is not necessary for BSAF development. Details of the Lake Ontario study, including initial modeling efforts with the WASP4 model can be found in EPA (1990a). Slight refinements to the WASP4 runs were later made (cited in EPA, 1993 as an unpublished report: Endicott, D.D., W.L. Richardson, T.F. Parkerton, and D.M. DiToro. 1990. A steady-state mass balance and bioaccumulation model for toxic chemicals in Lake Ontario: Report to the Lake Ontario Fate of Toxics Committee. U.S. EPA, Environmental Research Laboratory, Duluth, MN: 121 pp). The BAFs determined in these later runs will be tested using the models of this assessment.

In order to do this exercise, all critical model parameters used to develop the BAFs for this WASP4 modeling exercise will be used in the model framework of this assessment. The most critical parameter is the organic carbon partition coefficients, Koc, assumed for 2,3,7,8-TCDD. BAFs were determined assuming Koc of 107 and 108. Since the models of this assessment assume steady loading into water bodies, only the BAFs developed under "steady state" loading conditions will be used. As noted, the WASP4 model considers binding to more than one suspended compartment.

The increased binding can be modeled using a TBFoc, which was assumed to be 1.5 for Lake Ontario by Cook. For the models of this assessment, this factor will be applied to Koc - it effectively increases Koc by 50%. The concentration of suspended solids in Lake Ontario and used in the WASP4 modeling exercise was 1.2 mg/L. The other critical parameters are the fraction organic carbon contents of the suspended solids and the bottom sediments, OCssed and OCsed, respectively. Assigned values to these parameters, based on Lake Ontario data, in the WASP4 exercise and in this exercise were 0.15 (15%) and 0.03 (3%), respectively.

Since the purpose of this exercise is to evaluate how the modeling approaches of this document perform using the BSAF or the alternate BAF approach, duplicating the source strength terms used in the WASP4 modeling exercise is not necessary. The pertinent question is, with a given source strength, how would both approaches predict fish tissue concentrations. For simplicity, the on-site source category as demonstrated in Chapter 5 will be used. In this scenario, the soil within the watershed is assumed to uniformly be 1.0 ppt, and the loadings are via soil erosion.

In summary, the parameters for this exercise including the steady state BAFs are:

diagram 7-2

The 1.5 in the Kocs was the TBFoc noted above. The BAFs specific to each Koc were the ones developed also specific to those Koc in the WASP4 modeling exercises. For both tests: soil concentration of 2,3,7,8-TCDD = 1.0 ng/kg (ppt), total suspended solids (TSS) = 1.2 mg/L, the organic carbon content of suspended sediments (OCssed) = 0.15, and the organic carbon content of bottom sediments (OCsed) = 0.03. Whole fish tissue concentrations are estimated as Clipid * flipid, where flipid is 0.07.The whole fish tissue concentration for the BSAF approach in Test 1 was estimated to be 0.61 ppt. Using the ssBAFlt and ssBAFld, the whole fish tissue concentrations were estimated very nearly to be the same at 0.867 ppt for ssBAFlt and 0.863 ppt for ssBAFld. The test results did not change substantially for Test 2. The BSAF approach led to a fish tissue concentration of 0.62 ppt, and the concentration was identical for BAFs at 0.869 ppt.

While it appears that the water column based approaches estimate fish tissue concentrations identical to each other and very close to estimates made based on bottom sediment concentrations, in fact the performance of the models differ when parameters are changed in these tests. More incoming 2,3,7,8-TCDD can be modeled to remain in the water column with an increase in the reservoir of total suspended solids, the TSS parameter initialized in above tests at 1.2 mg/L. Continuing with Test 1 parameters above, increasing TSS from 1.2 mg/L to 10 mg/L has the following changes to fish tissue concentrations: 0.54 ppt for the BSAF test, 4.85 ppt for the ssBAFlt test and 0.76 ppt for the ssBAFld test.

Decreasing the organic carbon content of the suspended solids will have the effect of reducing the amount of incoming 2,3,7,8-TCDD simulated to remain in the water column, while increasing the amount modeled to reside in bottom sediments (because a mass balance of 2,3,7,8-TCDD is maintained), and also increases the dissolved phase concentration. Changing the TSS back to 1.2 mg/L and reducing the organic carbon content of suspended solids from 0.15 to 0.05 results in the following changes to fish concentrations: 0.62 ppt for the BSAF test, 0.45 ppt for the ssBAFlt test and 0.88 ppt for the ssBAFld test.

These two tests have demonstrated the variability in fish tissue concentrations when key water column parameters are altered. Fish concentrations would also differ if the key bottom sediment parameter, the organic carbon content of bottom sediments, was different. Returning to original Test 1 parameters and reducing the organic carbon content of bottom sediments from 0.03 to 0.01 results in the following changes to fish concentrations: 1.73 ppt for the BSAF test, 2.45 ppt for the ssBAFlt test and 2.44 ppt for the ssBAFld test.

The predictions for all tests might be considered reasonably close, given the uncertainties in the bioaccumulation and water modeling parameters. The one test described above where the BSAF and BAF approaches led to the most differences was the one which increased suspended material contents from 1.2 mg/L to 10 mg/L. In that case, nearly a ten-fold difference was noted in fish concentrations with the ssBAFlt as compared to the BSAF or the ssBAFld.

An important consideration in using the water column based approaches is that the BAFs developed by Cook (or that could be developed otherwise) are based on modeled rather than measured water column concentrations, and measured lake trout tissue concentrations. In that sense, the BAFs were calibrated for Lake Ontario conditions and specific to the WASP4 modeling exercise. Therefore, using these BAFs in the modeling framework of this assessment is, strictly speaking, invalid. Further, the values of the BAFs varied depending on the assumptions on historical loadings into Lake Ontario.

As noted above, three loading conditions were tested. The steady state BAFs were given above.

For the 20 year - 90% reduction tests, the following BAFs were determined: BAFld was 3.03x106 for Koc = 107 and 2.86x107 for Koc = 108, and BAFlt was 8.26x105 for Koc = 107 and 1.02x106 for Koc = 108.

For the 20 year - 100% reduction tests, the following BAFs were determined: BAFld was 3.86x106 for Koc = 107 and 3.40x107 for Koc = 108, and BAFlt was 1.05x106 for Koc = 107 and 1.21x106 for Koc = 108.

The BSAF developed for lake trout for Lake Ontario was developed using measurements of both fish tissue and bottom sediment concentrations. Both the BSAF and BAF are most appropriately developed using site specific data (coupled with a modeling exercise for BAF). Inasmuch as that can be impractical or difficult for many sites, efforts are underway to determine the general applicability of BSAFs and BAFs determined for one site to other sites. EPA (1993) proposes that BAFls for different congeners can be roughly estimated as the BAFl for 2,3,7,8-TCDD multiplied by the ratio of the BSAF for the congener and the BSAF for 2,3,7,8-TCDD. Such an estimate will incorporate differences in uptake, metabolism and chemical partitioning but not differences caused by chemical loss processes such as volatilization and photolysis. This approach for estimating BAFls for other congeners does allow for some generality since sediment and fish tissue data for other congeners and water bodies is available.

Another bioaccumulation term discussed in one literature article for dioxin is termed the Regulatory Bioaccumulation Multiplier, or RBM (Sherman, et al., 1992). Multiplication of this term and a "nominal water concentration" estimates a 3% lipid fish concentration. A nominal water concentration equals an amount of a contaminant, 2,3,7,8-TCDD in this application, added or entering a water body over time, divided by a flow volume over that same time. Assuming a fish lipid content of 3%, an RBM of 5000 was recommended based on examination of laboratory flow through data, simulated field data, and actual field data (EPA's Lake Ontario study and data downstream of pulp and paper mills). Dividing the 5000 by 0.03 gives 1.67*105, and this number is now analogous to the ssBAFlt developed by EPA (1993) described above, and in the same range as the 5.2-6.8*105 range for ssBAFlt.

7.2.4.4. Other modeling approaches and considerations for air concentrations resulting from soil volatilization

Volatilization flux was modeled using an approach given in Hwang, et al. (1986), developed for PCB flux from soils. Principal assumptions for their derivation were that contamination extended indefinitely, biodegradation or other degradation processes were not considered, residues were in equilibrium between soil and soil air, and vertical movement was through vapor phase diffusion. Their analytical solution was integrated over time and a solution was presented which gave average unit flux as a function of time during which volatilization occurs. PCBs and other dioxin-like compounds resist degradation, although there is evidence of photodegradation, which may influence surficial residues. These compounds sorb tightly to soil, so that an assumption of vertical movement primarily through vapor phase diffusion (rather than in a soluble phase with leaching, runoff, or evaporating water) is a tenable one. Also, presentation of an average flux rate solution made Hwang's approach amenable to spreadsheet analysis, the computer software tool used in this assessment.

An alternate model for estimating volatilization flux was presented in Jury, et al. (1983). It is a generalized analytical solution which assumes equilibrium between the sorbed, soluble, and vapor phases. It incorporates considerations of steady state water fluxes and degradation mechanisms. A depth over which contamination occurs is specified. A computer code of this model was obtained from the author (William A. Jury, Professor and Chair, Department of Soil and Environmental Sciences, University of California, Riverside, 92521-0424). Tests were run holding all pertinent quantities the same with both models including initial concentrations, organic carbon partition coefficients, Henry's Constant, molecular diffusivity, fraction organic carbon in soil, soil bulk density, porosity, and an assumption of contaminant non-degradation.

All of these parameters, the contaminant as well as the physical parameters, were the ones assumed for 2,3,7,8-TCDD and the surface soils of this assessment. In applying Jury's model, the depth of contamination was assumed to be 10 cm. Also, Jury's model allowed for a selection of water flux to be 0.5 cm/day (heavy leaching), -0.5 cm/day (heavy evapotranspiration), or 0.0 cm/day (no water flux). The latter selection of no water flux was chosen. This model comparison test showed that the Hwang model predicted an average flux over 10 years roughly three times higher than the average flux predicted by the Jury model over the same time period. Running both models over 50 years showed similar results. The average flux over that time dropped by about 50% for both models and there was still a three-fold difference in predicted volatilization fluxes.

The exact reason for this three-fold difference was not investigated, and could lie in differences in assumed boundary conditions (Hwang, et al. (1986) discusses differences in boundary conditions between his and Jury's models). In any case, it is judged that both models predict comparable volatilization fluxes. The Hwang model might be considered conservative in that it predicts 3 times higher volatilization flux (with 2,3,7,8-TCDD parameters, etc.).

The Jury model also provides other informative results. It provides a mass balance which, for the 50-year test, showed that only 2.6% percent of the original mass within the 10-cm layer had volatilized. By implication, the Hwang model predicts a 7.3% loss by volatilization over that time period. With the other parameters and assumptions - no degradation and tight sorption to soil - the Jury model showed that 97.4% remained in the profile and that only a minute quantity diffused below 10 cm. Also, the Jury model gives a concentration profile over time. After 50 years, it showed that all volatilization loss was contained within the upper 2 cm of soil profile. This implies that the boundary condition assumption for the Hwang model, that contamination extends indefinitely, is not consequential for the dioxin-like compounds.

A near-field dispersion model is used to estimate air concentrations resulting from soil volatilization, for the on-site source category (where contamination and exposure occur at the same site). An alternate approach to estimating on-site dispersion given a volatilization flux is the "box-model" approach. This simple approach can be visualized as follows: air above soil is contained within a structure which has two walls, say a north and south wall, and a ceiling - wind blows through the building in an east-west direction mixing the volatilized flux.
This is expressed mathematically as:

Equation V3 7-4

Before testing the box-model equation, results for the approach used in this assessment are summarized. The key factors impacting air concentration calculations in Scenarios 1 and 2 is the duration of exposure and area over which contamination occurs. In the central scenario, Scenario 1, the area was 4,000 m2 (1 acre) and in the high end scenario, Scenario 3, the area was 40,000 m2 (10 acres). The exposure duration was 9 years in Scenario 1 and 20 years in Scenario 2. The volatilization flux was different for both scenarios, but not because of area considerations, but because of exposure duration assumptions; the average flux of 2,3,7,8-TCDD for the high end scenario was 1.1x10-21 g/cm2-sec, whereas the average flux for the central scenario was 1.7x10-21 g/cm2-sec. The air concentration estimated for both the central and high end scenario was the same at 4.4x10-11 m g/m3. Larger areas tend to increase air concentration prediction; the larger area of the high end scenario countered the effect of having a lower average volatilization flux; hence similar air concentrations were predicted for the central and high end scenarios.

The values used to evaluate the box model approach were the fluxes, as given above, the mixing zone wind speed, 2 m/sec, which is half the average wind speed assumed in this assessment, the areas noted above, the side length, estimated as the square root of the area, and a mixing zone height estimated initially at 2 m. The box-model air concentration for the central scenario with these parameters is 2.7x10-10 m g/m3. This is 6 times higher than the concentrations predicted in this assessment. The box-model concentration estimation for the high end scenario, given slightly lower flux as noted above and the larger land area, was 5.5x10-10 m g/m3, which is over an order of magnitude higher than the concentration estimated for this assessment.

These box-model estimations are higher than the ones made for this assessment. An uncertain parameter for both modeling approaches is the area of soil contamination. The mixing zone height for the box model is also a parameter of uncertainty. Users of the box model approach have often assumed a conservative 2 m height approximating the height of exposed individuals. However, others have claimed this is far too low a mixing height, suggesting 10 meters or even an atmospheric height closer to 100 meters. Higher mixing zone heights would have brought the box model estimations more in line with estimations made in this assessment. The closest analogous parameter in the dispersion model to the mixing zone height is the height of exposed individual, which is more unambiguously the breathing zone height of 2 m.

One key assumption concerning the exposure site air concentrations resulting from an off-site area of soil contamination should be questioned. The current approach assumes that air-borne contaminates originate at the site of contamination and are transported to the site of exposure. On the other hand, this assessment also assumes that exposure site soil becomes contaminated over time due to erosion. Also, some of the example scenarios have tested the impact of very low, perhaps "background", levels of dioxin-like compounds, which would occur surrounding a site of exposure. It is at least plausible that volatilization from soils other than the area of elevated contamination would contribute to air-borne contamination, and concentrations to which individuals are exposed to at sites of exposure near sites of contamination.

This was tested by using the on-site algorithms and developing soil concentrations for these algorithms based on soil concentrations predicted to occur in the off-site scenario. Specifically, the off-site demonstration scenarios included a 10 ha field at 1 ppb 150 m from the exposure site, also at 10 ha. The soil concentrations estimated to occur at the exposure site were 0.28 ppb for a 5-cm no-till depth and 0.08 ppb for a 20 cm tilled depth. The on-site algorithms for volatilization and dispersion were run starting with these concentrations, and resulting concentrations were compared with those estimated to occur only from volatilization from the contaminated site and transport to the exposure site. The air concentration estimated to occur from untilled soil is 2.5 times higher than that estimated to occur from the off-site area and transported; the air concentration estimated to occur from tilled soil is 25% less than estimated to occur from volatilization and transport.

This might imply that exposure site air concentrations are being underestimated if air concentrations at the site of exposure are assumed to only originate at the site of contamination, and not also at the site of exposure, or even from other areas. Lower estimated air concentrations also would result in lower estimates of impact to above ground vegetations, including fruits and vegetables for consumption, and grass and cattle feed, whose concentrations partially determine beef and milk concentrations. This exercise implies that the underestimation might be less than a factor of 5.0. Of course, this conclusion is contingent on the off-site impact algorithms which have estimated that a 0.28 or a 0.08 ppb soil concentration will result 150 meters from an area whose concentration is 1.00 ppb.

7.2.4.5. Alternate models for estimating plant concentrations from soil concentrations

The models of this assessment separate above and below ground vegetations for estimating concentrations. Root concentrations, which in this assessment translates to below ground vegetations, are a function of soil water concentrations and a Root Concentration Factor, RCF. Above ground vegetations, which in this assessment include above ground fruits and vegetables, pasture grass, and cattle feed, are modeled as a function of vapor phase transfers and wet plus dry particle depositions. This section examines one alternate approach for above ground vegetations; alternate approaches for below ground vegetations could not be found.

One approach to modeling plant concentrations would be with passive uptake via evapotranspiration. The assumption here is that soluble phase contaminants move passively with transpiring water. This approach has been applied for contaminants which are soluble in water. However, nearly all the evidence suggests that this would not be appropriate for the dioxin-like compounds. Specifically, the evidence suggests that residues do not translocate to within portions of either above or below ground vegetations. Such would be case for soluble contaminants moving passively with transpiring water. This conventional wisdom was, however, challenged with a recent experiment by Hulster and Marschner (1993b) on vegetations of the cucumber family. Their results were most striking for zucchini, which showed uniform plant concentrations from inner to outer portions of the zucchini fruit, and the highest whole fruit concentrations they had ever measured, despite careful experimental conditions which physically isolated the fruit from the soil. Pumpkins also showed high plant contamination, with more expected plant concentrations measured for the cucumber. Assuming the vegetations of this assessment - fruit/vegetables for human consumption and vegetations of the beef/dairy food chain - do not behave as in Hulster and Marschner's (1993b) experiment, than translocation to inner plant parts is not expected.

The specific issue of uptake and translocation via transpiration was investigated using soybean and corn plants grown hydroponically in carefully constructed growth chambers (McCrady, et al., 1990). Roots and the hydroponic growth solution were separated from the shoots and leaves of these plants using two separate chambers, one inverted over the other. Separate air-flow systems for each chamber included traps for volatile organics. Mass balance on the tritiated TCDD experiments was able to recover 98% in the soybean experiment and 86% for the corn experiment. Most of the recovered material was found in the roots; 75% for soybeans and 67% for corn, with the second highest recovery was on the inside surface of the root chamber, around 15% for both experiments. Recovered TCDD was also found, in order of decreasing percentage, in the growth solution, root chamber air, shoot chamber air, and shoots. The recovery from the shoots was negligible at 0.004% and 0.001% of the total TCDD for the soybean and corn, respectively. McCrady, et al. (1990) concluded that transpiration stream transport of 2,3,7,8-TCDD to plant shoots is an insignificant mechanism of plant contamination, and that volatilization of TCDD is an important transport mechanism that can result in significant quantities of airborne TCDD being absorbed by plant shoots.

Briggs, et al. (1982) provide another way to evaluate the translocation of contaminants from roots to above ground vegetation. Experiments with barley roots in growth solution led to the development of an empirical parameter describing the efficiency of transport of organic chemicals to plant shoots from root uptake. This parameter is called the Transpiration Stream Concentration Factor (TSCF) and is defined as (concentration in transpiration stream)/(concentration in external solution).
The empirical formula presented for this factor is:

Equation V3 7-5

Given a log Kow for 2,3,7,8-TCDD of 6.64, TSCF is solved for as roughly 5 * 10-5. Assuming that the concentration of external solution concentration for the experimental conditions of Briggs' experiments is equivalent to the concentration in soil water in a field situation, then the TSCF for 2,3,7,8-TCDD implies that the transpiration stream water of a plant is over 5 orders of magnitude lower than the soil water concentration. Like McCrady's experiments, this also shows the insignificance of translocation of residues from roots to shoots.

The one approach that was found that might have been used in the place of the algorithms for above ground vegetation, is simpler and more general in nature. It was developed from field data on above ground vegetation concentrations correlated to soil concentrations of contaminants and the octanol water partition coefficient (Travis and Arms, 1988). This correlation led to an empirical bioconcentration factor for vegetation, Bv, regressed against the contaminant log Kow, and defined by the authors as the concentration in above ground plant parts divided by the concentration in soil:

Equation V3 7-6

With 2,3,7,8-TCDD log Kow equal to 6.64, the Bv translates to a value of 0.0056. Note that this Bv is defined identically to the plant:soil contaminant concentration ratios that were discussed in Section 7.2.3.8 which compared the model's estimations of these ratios with those found under experimental. As discussed in that section, contaminant concentration ratios were estimated for the two scenarios demonstrating the on-site source category in Chapter 5, Scenarios 1 and 2: above ground vegetables/fruit - 7*10-5, grass - 6*10-3, and feed - 3*10-3.

It is not clear how to compare the Bv of 0.0056 to these ratios without retrieving the studies which Travis and Arms (1988) used, although this value is clearly higher than the fruit/vegetable ratio and consistent with the grass/feed ratio estimated for Scenarios 1 and 2. The studies used by Travis and Arms were not retrieved. An examination of the chemicals used by Travis and Arms show that 25 of 29 used are pesticides, which suggests that plant concentrations may be those of agricultural crops.

If so, a comparison of the above-ground 1*10-5 ratio with this 0.0056 ratio would be appropriate. An examination of the chemicals also reveals that 10 of the 29 are moderately to very soluble (log Kow less than 4.00), while others are similarly insoluble as the dioxin-like compounds (including DDT, TCDD, Aroclor 1254, and others; 15 with log Kow greater than 5.0). Developing such an empirical relationship which mixes chemicals whose mode of action is passively with water (which would be the case with aldicarb and simazine, among others on the list) with those whose mode is through vapor transfers or particle depositions (TCDD, and so on) does not appear to be technically valid.

Nonetheless, the fact that the Travis and Arms Bv is much higher than the plant:soil ratio generated for the on-site soil contamination source category demonstration is noteworthy. First, it was noted in Section 7.2.3.8 that the plant:soil ratios generated by the models were lower than had been measured in the literature, and this is an additional piece of evidence in that direction. Second, other evidence in this assessment suggests that the air concentrations resulting from soil contamination may be underestimated by over an order of magnitude. This was discussed in Section 7.2.3.7 above.