particulate reservoir would have on vegetative concentrations. First, 25% of the volatilized vapor phase reservoir was transferred to the particulate reservoir, and then 50%was transferred. The results of this test are shown in Table 6-3.

Vapor phase transfers still explain more of the grass and fodder concentration than particle depositions, with 60-65% of the plant concentration explained by vapor transfers when 50% of the reservoir is transferred to the particle phase reservoir. The overall impact to grass and fodder concentrations, however, is small, with only an 21-26% reduction when 50% of the vapors are transferred. As has been discussed, particle depositions are important for vegetable/fruit concentrations. Therefore, increasing the particle phase reservoir while decreasing the vapor phase reservoir greatly increases the dominance of particle impacts to vegetations - 92% of vegetable/fruit concentrations are due to particle impacts when 50% of the vapor phase reservoir is transferred, and the vegetable/fruit concentrations increase by about a factor of 3 with this transfer.

It might be concluded from this test that modeling the sorption of volatilized vapor-phase dioxin-like contaminants would: 1) mostly impact the inhalation exposures, with vapor phase exposures being reduced equal to the amount modeled to move into the particulate reservoir and particulate inhalations increase by the additional amount added to the particulate reservoir - total vapor plus particle phase inhalation exposure would not change, 2) tend to increase the concentrations in vegetables/fruits with the subsequent impact on exposures from consumption of fruit and vegetables, and 3) apparently have little overall effect for grass and fodder concentrations, and hence little effect on beef and milk concentrations.

One important factor to note up front about below ground vegetable concentrations as compared to above ground vegetable concentrations (no underground fruits are assumed in this assessment) is that below ground vegetable concentrations are about two orders of magnitude higher than above ground vegetable concentrations for the soil contamination demonstration scenarios. Since no fruit is assumed to be grown underground, this difference does not affect fruit ingestion exposures. However, 28 g/day of a total of 106 g/day vegetable ingestion is assumed to be underground vegetables. Given the difference in concentration estimations, below ground vegetables explain 97%

Table 6-3. Results of the sensitivity test of modeling vapor/particle partitioning for volatilized residues (note: soil concentration equals 1 ppt in tests below).

Percent contribution of vegetative Percent

concentration due to: Vegetative Change from

Vegetation Vapor transfers Particle depositions Conc., ppt Baseline

I. Baseline

Veg/fruit 49 51 0.0000109

Grass 94 6 0.0038

Fodder 95 5 0.0019

II. 25% transfer of vapor phase to particulate reservoir

Veg/fruit 18 82 0.0000221 +103

Grass 79 21 0.0033 -13

Fodder 83 17 0.0016 -16

III. 50% transfer of vapor phase to particulate reservoir

Veg/fruit 8 92 0.0000333 +206

Grass 60 40 0.0030 -21

Fodder 65 35 0.0014 -26

of the total exposure via ingestion of impacted vegetables.  Sensitivity of underground vegetable concentrations to parameter changes for the soil contamination source category becomes important from this perspective.

On the other hand, the trend for the stack emission source category is exactly the opposite - above ground vegetable concentrations exceed below ground vegetable concentrations by two orders of magnitude. If air and soil concentrations were equal (or proportional) for the demonstration of both source categories, than the impact to vegetations would be equal (or proportional). Obviously, that was not the case. The soil concentration of 2,3,7,8-TCDD of 1 ppt (ng/kg) for the on-site soil source demonstration scenarios translated to a total airborne reservoir of 5E(-11) m g/m³ (vapor plus particle phase reservoirs). For the stack emission demonstration scenario 5, where the exposure site was 500 meters from the stack emission source, the vapor plus particle phase concentration of 2,3,7,8-TCDD was 1.4E(-12) m g/m³, which is reasonably similar to the air concentration of 5E(-11) m g/m³ modeled from a soil concentration of 1 ppt. However, the soil concentration for the stack emission demonstration scenario for the 20-cm mixing zone depth used for underground vegetation concentration was 0.00005 ppt, more than 4 orders of magnitude lower than the 1 ppt level for the soil contamination source categories. Therefore, below ground vegetables for the stack emission source category will be more than 4 orders of magnitude lower for the stack emission source category as compared to the soil source category.

This question now is whether this is a reasonable outcome. Should deposition from an airborne reservoir in the range of 10^-11 m g/m³ 2,3,7,8-TCDD result in a soil concentration closer to 1 ppt than 0.00005 ppt? Is the algorithm estimating soil impacts from depositions inherently underestimating soil concentrations? Likewise, should emissions from soils at 1 ppt result in air concentrations higher than 10^-11 m g/m³? Is the soil emission/dispersion algorithms inherently underestimating air concentrations? If air to soil impacts were being underestimated, and/or if soil to air impacts were being underestimated, than more correct model performance would lead to more equivalent outcomes with regard to vegetation impacts. As stated above, if soil and air concentrations for each demonstration scenario were equal (or proportional), then impacts to vegetation would be equal (or proportional). Air-to-soil and soil-to-air model performances are now examined briefly.

In Section 7.2.3.1 in Chapter 7, the capability of the deposition algorithm to estimate soil concentrations is examined. The hypothesis examined is that air concentrations of 2,3,7,8-TCDD in a rural setting should correlate to untilled soil concentrations in a rural setting, and that hypothesis can be evaluated using the deposition algorithms of the stack emission source category. The conclusion from that section was that it would appear that the deposition algorithm may, in fact, be underestimating soil concentrations. The amount of underestimation was speculated to be about one order of magnitude. Uncertain model parameters identified in that section were the untilled mixing zone depth, the velocity of deposition, and the dissipation half-life for depositing residues. It is also noted that the algorithms of this assessment do not consider detritus production as an input to the soil reservoir. If the deposition algorithm is indeed underestimating soil concentrations, than in fact an airborne reservoir on the order of 10^-11 m g/m³ might translate to a concentration higher than 0.00005 ppt, although a one order of magnitude increase to 0.0005 ppt is still much less than 1 ppt.

One could also even question the use of a contaminant dissipation rate which is applied to depositing residues which are assumed to be tilled in a home garden or an agricultural field. Routes of dissipation for dioxin-like compounds are physical transport, such as wind or soil erosion, or volatilization, or chemical, such as photolysis, which is the only degradation route shown to be relevant to these compounds. These are phenomena relevant to surface residues, not buried residues. The use of a 10-year dissipation rate for both tilled (a mixing zone depth of 20 cm) and untilled (a mixing zone depth of 1 cm) settings could be a fundamental flaw in the approach - perhaps a dissipation rate should only be applied to untilled soil impacts. In addition to the argument presented above, this is another reason suggesting tilled soil concentrations resulting from stack emissions should be higher than are estimated in this assessment, and hence underground vegetable concentrations should be higher than are estimated in this assessment.

The other hypothesis is that the soil to air algorithms of the soil source categories are underestimating air concentrations. In fact, evidence developed in other parts of this document suggest that air concentrations resulting from soil concentrations may be underestimated. One piece of evidence discussed in different sections of this Volume is that air concentrations modeled to result from background soil concentrations are lower than air concentrations measured in pristine settings. In one literature article measuring concentrations in an area described as a "remote countryside" in Sweden (Broman, et al. 1991), air concentrations of 2,3,7,8-TCDD were measured at 2*10^-10 m g/m³. The air concentration of 2,3,7,8-TCDD modeled in this assessment from a 1 ppt background soil concentration of 2,3,7,8-TCDD is nearly an order of magnitude lower than that at 4*10^-11 m g/m³. This suggests that the models of this assessment underestimate air concentrations resulting from releases from soils. Another piece of evidence is developed in Section 7.2.3.8, which compares plant:soil ratios as determined by the model with those developed in experimental and field conditions. The soil contamination models of this assessment appear to be leading to above ground plant:soil ratios that are 1-2 orders of magnitude lower (i.e., plant concentrations may be 1-2 orders of magnitude underestimated) than analogous ratios for experiments where soil can be surmised to be the only source of dioxin. One possible explanation offered is that the air concentrations are being underestimated.

If air concentrations resulting from soil concentration would be modeled to be higher than they are currently, then the dichotomy identified above would be narrowed. If soil concentrations resulting from air depositions are modeled to be higher than they are currently, then the dichotomy identified would be even further narrowed. Evidence summarized above suggests that both are plausible. Clearly, more evaluation of the soil to air algorithms of the soil contamination source categories and the air to soil algorithms of the stack emission source category, is warranted.

Given a soil concentration, in any case, the impacts of parameter changes for the algorithm predicting concentrations in underground vegetables are shown in Figure 6-10. The two orders of magnitude range for the root concentration factor, RCF, translates to a two order of magnitude range of concentration estimation. The same is true for the empirical correction factor applied to below ground vegetables, VG_bg, and the organic carbon partition coefficient, Koc. A smaller impact is noted for the organic carbon fraction of soil, OC_sl. Koc and OC_sl are required for this algorithm because vegetable concentrations are a function of soluble phase concentrations, not soil concentrations. Increasing Koc and/or increasing OC_sl results in decreasing the water concentrations, explaining why the high values for these parameters reduce vegetable concentrations.

One final note is that the dry to fresh weight ratio, FDW, is not on this figure, while it does appear on Figure 6-8. This is because the RCF was developed on fresh weight basis already, so no conversion to a fresh weight is required.

6.3.3.10. Beef fat concentration estimation The impacts of parameter changes to beef fat concentration estimation for the soil source category is shown in Figures 6-11.

First, it is noted that changes in soil concentration result in linear changes in beef concentrations - a ten-fold increase in soil concentration results in the same ten-fold increase in beef concentration. This is because changes to soil concentration alone result in the same proportional change in the concentrations estimated to be in grass and cattle feed. Changes to grass and feed concentrations with no change in soil concentrations (which could result from different parameter selections in estimating grass and feed concentrations) do not result in as substantial a change - a tenfold increase in vegetation

concentrations only increases beef concentration by a factor of two; a tenfold decrease in vegetation concentration only reduces beef fat concentrations by about 20%.

This is a key and insightful result. If the models and their parameterization are valid, it indicates that the bulk of impact to beef and milk fat is from ingestion of soil for the soil source categories. With a closer look at the results for the demonstration of the on-site soil category, demonstration scenario #2, it is found that soil ingestion explains 90% of the beef fat concentration, despite being only 4% of their diet. Grass (48% of

their diet) and feed (48%) of their diet explain 7 and 3% of beef fat concentrations. The story is the same for milk fat concentrations. Soil explains 87% of the milk fat concentrations, despite being only 2% of the dairy cattle's diet. Feed (90% of their diet)