Page 3 of 9 Data on fish concentrations in the literature

This assessment estimated fish concentrations of 2,3,7,8-TCDD for the various source categories to be:

1) on-site soil source category - 0.6 ppt,
2) off-site soil contamination - 3 ppt,
3) stack emission source category - 0.0004 ppt, and
4) effluent discharge source category - 0.4 ppt.
Data was not found to appropriately compare the stack emission source category results, and data on the effluent discharge source category is examined in the next section below. This section will examine some available data on fish concentrations in order to compare results from the first two categories with measured results.

The most appropriate study with which to make comparisons is the National Study of Chemical Residues in Fish (EPA, 1992b; hereafter abbreviated NSCRF, see Chapter 4, Section 4.5 of Volume II for more detail). Fish tissue data on a variety of species and contaminants of concern in aquatic environments and fish from around the country were developed. Most important for current purposes, the sites were carefully characterized in terms of potential sources of fish contamination.

There were 353 sites from which fish tissue data was available, of which 347 had data on 2,3,7,8-TCDD. Results from four site categories might be appropriate for comparison with concentrations estimated to occur from low, possibly background, soil concentrations of 2,3,7,8-TCDD. The four categories and number of sites per category were: the USGS water quality network NASQAN - 40 sites; Background (B) - 34 sites, Agricultural (A) - 17 sites, and Publicly Owned Treatment Works (POTW) - 8 sites. The average 2,3,7,8-TCDD concentrations measured for these four categories were: NASQAN - 1.02 ppt; B - 0.56 ppt; A - 0.75 ppt, and POTW - 0.90 ppt. The "on-site" source category was demonstrated in Chapter 5 using a soil concentration of 2,3,7,8-TCDD that might be typical of background conditions - 1 ng/kg (ppt).

The resulting fish tissue concentrations estimated for this soil concentration was 0.6 ppt. Four of the site categories of the NSCRF might be considered representative of sources characterized as land areas of high soil concentrations of 2,3,7,8-TCDD. These were: Industrial/Urban site (IND/URB) - 105 sites, Refinery/Other Industry (R/I) - 20 sites, Wood Preservers (WP) - 11 sites, and Superfund Sites (NPL) - 7 sites. Average fish tissue concentrations measured for these site categories were: IND/URB - 4.04 ppt, R/I - 4.38 ppt, WP - 1.40 ppt, and NPL - 30.02 ppt. The source category of this assessment most similarly characterized to these would the category of off-site soil contamination, where a bounded area of contaminated soil had 2,3,7,8-TCDD at 1 ppb.

As noted above, the fish tissue concentration estimated with this source category was 3 ppt. The two remaining site categories of the NSCRF were Paper Mills Using Chlorine (PPC), and Other Paper Mills (PPNC). These data served as the basis for the comparison discussed below in the effluent discharge source category.

In general, the range of fish tissue concentrations measured for (perhaps) background conditions, 0.56 - 1.02 ppt, were comparable to the fish tissue concentration estimated assuming the low (perhaps) background soil concentration of 1 ppt soil concentration, 0.6 ppt. It may also be appropriate to make the same observation for the source categories assuming higher soil concentrations as compared to measured concentrations noted above. In this case, the range of average measured concentrations - 1.4 - 30.02 ppt compares with the modeled 3 ppt.

This does not constitute a validation exercise, strictly speaking, since specific field data were not input and compared. However, a few key points can be made. One, the magnitude of concentrations appears to have been captured, and the approximate order of magnitude difference between background and higher source strength categories of the NSCRF also appears to have been duplicated. One data point from that study of interest is the 30.02 ppt concentration found for the NPL site. This is an order of magnitude higher than the 3.0 ppt estimated for the off-site soil source category.

No insights can be gained from this difference because information was unavailable on the seven sites which were characterized as Superfund sites and which were expected to have been the cause of the 30.02 ppt fish concentration. However, one exercise contained in Chapter 6 of this volume, the mass balance exercise (Section 6.4), does contain one interesting result. Assuming the 1 ppb concentration of 2,3,7,8-TCDD extended 15 cm (6 in) deep on the 40,000 m2 contaminated area, the mass of 2,3,7,8-TCDD that would be at this site would be 9 grams. This appears to be a small amount (and in fact the models of this assessment do not require depths of contamination - the fish and all other concentrations would be estimated if the contamination extended 15 cm or 15 m deep), and it would be interesting to know the source strengths of the 7 NPL sites, at least the soil surface concentrations.

Another comprehensive data base of fish concentrations of 2,3,7,8-TCDD is from EPA's National Dioxin Study (EPA, 1987; abbreviated NDS), which actually provided the motivation for the NSCRF when significant residues of 2,3,7,8-TCDD were found in fish in the NDS. Fish concentrations from the NDS are also listed and discussed in Kuehl, et al. (1989). Travis and Hattemer-Frey (1991) summarized the fish data from the NDS.

Their summary is as follows. Data collected from 304 urban sites in the vicinity of population centers or areas with known commercial fishing activity, including the Great Lakes Region, showed concentrations to range from non-detected to 85 ng/kg (ppt). The geometric mean concentration was 0.3 ppt, and only 29% had detectable levels of 2,3,7,8-TCDD. The Great Lakes data had more contamination, with 80% detection rate and a geometric mean concentration of 3.8 ppt. Recall from above that the fish tissue concentrations estimated for 2,3,7,8-TCDD for the off-site soil contamination source category was 3.0 ppt.

The NSCRF also collected data on 2,3,4,7,8-PCDF, the second example compound demonstrated. Briefly, the range of average fish tissue concentrations noted for the site categories evaluated as background above is 0.42-0.78 ppt, very similar to the 2,3,7,8-TCDD range of 0.56-1.02 ppt. The modeled fish tissue concentration of 2,3,4,7,8-PCDF for background conditions was the same as that for 2,3,7,8-TCDD at 0.6 ppt. Actually, 2,3,5,7,8-PCDF concentrations were slightly higher due to slightly higher Koc values for 2,3,4,7,8-PCDF - which translates to less partitioning to water and more in sediments.

The key parameter leading to the similar result is the assignment of the bioaccumulation parameter, the biota to sediment accumulation factor (BSAF) the same value for both compounds. The range of 2,3,4,7,8-PCDF average fish concentrations for the sites of elevated soil concentration was 1.86-5.44 ppt, which compares to the modeled 2,3,4,7,8-PCDF concentrations of 3.0 for the off-site source category with initial soil concentrations of 1.0 ppt.

The NSCRF also collected data on PCB concentrations in fish, although the results were expressed in terms of total tetra-, hepta-, and so on. The data indicates concentrations well into the part per billion range for this breakout, and even higher considering total PCBs. The average concentration of total heptachlorobiphenyls over all study sites was 96.7 m g/kg (ppb). The average concentration of total PCBs over all sites was estimated as 1897.88 ppb, and the average concentration of total PCBs for background sites was 46.9 ppb. The modeled concentration of the example heptachlorobiphenyl, 2,3,3',4,4',5,5'-HPCB, for the background scenario (soil concentration = 1.0 ppt) was 10 ppt. The modeled concentration of 2,3,3',4,4',5,5'-HPCB for the off-site soil source scenario, where soil concentration were 1 ppb, was 80 ppt.

Data from the Great Lakes region indicate that PCB concentrations are significantly higher than PCDD/PCDF concentrations in this area. PCB concentrations from fish in Lake Ontario are in the tens to hundreds of ppb level (Niimi and Oliver, 1989), while 2,3,7,8-TCDD contamination in Lake Ontario was in the tens of ppt level (EPA, 1990a) - a three order of magnitude difference. Other data in Table B.10, Appendix B, Volume II, where concentrations were similarly in the tens to hundreds of ppb level were from Lake Michigan (Smith, et al. 1990) and Waukegan Harbor in Illinois (Huckins, et al., 1988). The single data point from that table for 2,3,3',4,4',5,5'-HPCB, the example PCB congener in Chapter 5, was for carp in Lake Michigan, and was 29 ppb (29,000 ppt).

While the modeled PCDD/PCDF fish concentrations seem reasonably in line with measured concentrations, this assessment may have underestimated concentrations of 2,3,3',4,4',5,5'-HPCB. As noted, concentrations for fish in the Great Lakes Region were in the tens to hundreds of ppb range, while this assessment derived estimates all under 1 ppb. It is inappropriate to make direct comparisons without also comparing source strengths. Concentrations of PCBs in bottom sediments ranged from the low ppb for the tri-PCBs, to the tens of ppb for the tetra through hexa-PCBs, back to the low ppb for the hepta and octa-PCBs, in Lake Ontario (Oliver and Niimi, 1988).

Another literature source showing fish concentrations in Waukegan Harbor, IL, in the hundreds of ppb range, had sediment concentrations of specific congeners as low as 5 ppb to as high as 131 ppm. The concentration of 2,3,3',4,4',5,5'-HPCB in bottom sediments was estimated to be only 40 ppt in the off-site soil scenario. Therefore, one reason PCB concentrations in fish estimated in this assessment are as much as three orders of magnitude lower than noted in the literature is because sediment concentrations estimated for the source categories in this assessment are three orders of magnitude lower. The BSAF for PCBs also was noted to be variable, with values below 1.0 to values over 20.0 (see Section, Chapter 4 of this volume). The BSAF for the example PCB congener in this assessment was 2.0. Higher BSAFs would also increase PCB concentrations estimated for fish.

The fish concentration of 2,3,7,8-TCDD estimated for the stack emission source category was lowest at 0.00007 ppt. Data was unavailable to place this in any comparative framework. It was noted that the watershed soil concentration estimated was 0.0001 ppt 2,3,7,8-TCDD. The fish-to-soil ratio of 0.7 compares favorably to the fish-to-soil ratio of 0.6 (0.6 ppt/1.0 ppt) for the on-site soil demonstration which assumed basin-wide 1.0 ppt soil concentrations. Impact of pulp and paper mill effluent discharges on fish tissue concentrations

a. Description of Exercise and Model Parameters

This section describes a validation exercise of the effluent discharge algorithm. Validation is a cautious description of the exercise, since much of the data used is of uncertain quality. Discharge rates of 2,3,7,8-TCDD (mass/time units) into surface water bodies from a subset of 104 pulp and paper mills, which were sampled on a one-time basis in 1988 for such discharges and other parameters (EPA, 1990d; hereafter referred to as the 104-mill study), represent the key observed source term for this exercise. Fish concentrations of 2,3,7,8-TCDD for fish sampled downstream of these sources as part of the National Study of Chemical Residues in Fish (EPA, 1992b; abbreviated NSCRF hereafter) represent the key predicted model result for this exercise.

The National Council of the Paper Industry for Air and Stream Improvement (abbreviated NCASI hereafter) has already performed this exercise, and a brief description of their efforts and results can be found in Sherman, et al. (1992). NCASI carefully matched NSCRF data to appropriate mills of the 104-mill study. In many cases, they found more than one fish sample to correspond to a given discharge. Also, they considered circumstances where more than one mill effluent discharge can be considered to have impacted the environment where fish were sampled. In these cases, discharge rates from the contributing mills were fed into the model as source terms.

In NCASI's careful examination of the available data, they only considered 47 of the 104 mills as appropriate for this type of model testing. From these 47 mills, 95 fish samples with detectable residues of 2,3,7,8-TCDD were identified. Some mills had only one fish sample corresponding to it while others had up to four fish samples. The following explains why 57 of the remaining mills were not considered for this exercise:

1. Downstream of 10 pulp and paper mills was an estuary. NCASI considered the model appropriate for riverine situations only and did not calculate fish concentrations for estuarine settings.

2. The measurement for 2,3,7,8-TCDD in the effluent was listed as non-detect, and no further data examination and modeling occurred. There were 13 mills in this category.

3. NCASI could not identify appropriate fish measurements in the NCSRF downstream of the mill, and did not model further. Seven mills were in this category.

4. Some of the mills in NCASI's exercise were only considered "proximate" mills adding to the source term associated with another mill and one or more fish concentrations. Five mills were described in this manner.

5. For the remaining 22 mills, no explanation was provided for their lack of inclusion in the validation exercise.

Details of the NCASI modeling assumptions were supplied to EPA by NCASI (personal communication, Steven Hinton, PhD., P.E., NCASI, Inc.; Department of Civil Engineering, Tufts University, Medford, MA, 02155) and adopted for this exercise. Several other source materials were used to develop the parameters for this exercise. First, Figure 7-1 shows the effluent discharge model and all the numerical quantities required, including the source term and the observed fish concentration, and model parameters associated with the mill discharge and the aquatic environment. Further description of the effluent discharge model can be found in Chapter 4, Section 4.6. The model parameters and their source materials are now listed.

1) Mill parameters including the 2,3,7,8-TCDD discharge rate, the effluent flow rate, the suspended solids content of the effluent flow, and the organic carbon content of the suspended solids in the effluent flow: The 104-mill pulp and paper mill study (EPA, 1990d), a cooperative study between EPA and the paper industry, measured mass releases of 2,3,7,8-TCDD (actually effluent flow and concentrations, from which mass releases can be estimated),...

table Figure 7-1 Schematic of effluent discharge model showing all parameter inputs and observed fish concentrations.

... effluent flow, and total suspended solids content of the effluent flow (and other information such as releases of 2,3,7,8-TCDF, which were not needed for this exercise).

For purposes of this validation exercise, actually only the total suspended solids content of effluent discharges was used from the primary reference of this study (EPA, 1990d).
Data from the 104-mill study was also used in a modeling study, described more fully below, and in that reference, it was more conveniently organized and compiled. As such, effluent flow and 2,3,7,8-TCDD discharge rates came from a secondary reference.

expand table Figure V3 7-1

The organic carbon content of the solids in the effluent was assumed to be 0.36. This was the value used in the example scenario of Chapter 5, and was based on the fact that effluent solids are principally biosolids).

2) Receiving water body parameters including flow rate, suspended solids content, and organic carbon content of suspended solids. A modeling study conducted by EPA (EPA, 1990e) used a simple dilution and the EXAMS model to evaluate the impact from discharges of 2,3,7,8-TCDD and 2,3,7,8-TCDF from chlorine bleaching mills. Mills from the 104-mill study were the ones evaluated in this report. This study developed key receiving water parameters for these mills which are pertinent to the dilution model of this assessment, including harmonic mean flows at the point of effluent discharges, which were based on the nearest STORET sampling point, and suspended solids concentration of the receiving water body at this point. Details on how these key quantities were developed are included in that report and will not be discussed here. The organic carbon content of the suspended solids was assumed to be 0.05, which was also the content assumed for the example scenarios in Chapter 5.

3) Parameters associated with 2,3,7,8-TCDD, including the organic carbon partition coefficient, Koc, and the biota suspended sediment accumulation factor (abbreviated BSSAF). The Koc for 2,3,7,8-TCDD was the same 2.69x106 otherwise assumed in this assessment, and the BSSAF value was assumed to be 0.09, which is the same value as the BSAF, Biota (bottom) Sediment Accumulation Factor. Sections in Chapter 4 further discuss the Koc, BSAF, and BSSAF.

4) Fish data including the fraction lipid and the observed fish concentrations: The core reference for this information is the National Study of Chemical Residues in Fish (EPA, 1992b), as noted above. NCASI compiled the fish concentrations and associated lipid content of the samples as part of their modeling exercise, and these were used here as well.

Table 7-5 lists the parameters used for each identified mill and receiving water body, as well as the modeled and observed fish concentrations. Not included in this table are the parameters assumed for all model runs, including the organic carbon contents of the suspended solids terms, and the 2,3,7,-TCDD Koc and BSSAF.

b. Results and Discussion

One important point to discuss up front is that 38 of the 47 eligible mills discharged into surface water bodies that were characterized as "low", while the remaining 9 mills discharged into "high" receiving water bodies. This characterization refers to the flow rates of the receiving water bodies. The average flow rate of the 38 low water bodies was 5.4 * 108 L/hr, with a range of 107 to 109 L/hr, while for the other nine, the average flow was 2.8 * 1010 L/hr, with a narrow range of 1 to 4 * 1010 L/hr.

This distinction appears to be non-trivial for a few reasons. One, model predictions appear to more closely match observations for the smaller water bodies. The average of 38 mills and 74 fish for modeled and observed fish concentrations is 7 ppt and 15 ppt, respectively. The average of 7 mills and 21 fish associated with large receiving water bodies for modeled and observed fish concentrations is 0.7 and 5.3 ppt, respectively. As evaluated by NCASI, another important feature of the larger receiving water bodies is that they were the ones principally considered to have multiple discharges.

table Table 7-5 Model parameters and results for effluent discharge model validation testing.
A final observation concerning the large receiving water bodies is that the suspended solids data is also significantly different than the low receiving water bodies. For the 38 water bodies associated with the small water bodies, the receiving water body solids content averaged 9 mg/L, while for the nine high receiving water bodies, the suspended solids content averaged 78 mg/L.

This importance of the suspended solids content is principally seen for mills 39-42. The solids content of these water bodies ranged from 107 to 221 mg/L. The average modeled fish concentration for these mills was 0.04 ppt, while the average observed fish concentrations was 3.0 ppt.
expand table Table V3 7-5

The impact is one of "dilution": discharged 2,3,7,8-TCDD mixes into a larger reservoir of suspended particles, leading to a low 2,3,7,8-TCDD concentration on suspended solids concentration and lower predicted fish tissue concentrations. This dilution effect may also be real, as the average observed fish concentrations for these circumstances of 3.0 ppt may indicate a significant difference with the average 15.0 ppt observed for the smaller receiving water bodies. Nonetheless, these high suspended solids data must be considered suspect.

Considering all 47 mills and 95 fish observations, it was found that 73 and 87% of predictions within a factor of 10 and 20 of observed concentrations, respectively. The predicted and observed results of this exercise for these 47 mills is shown graphically in Figure 7-2. As seen here, a best-fit simple linear regression (using the least squares method) has a slope of 0.86 and an intercept of 8.06. The positive intercept indicates that the model, with all current parameter assignments, has a tendency to underpredict fish concentrations. Also of note and perhaps not ironically, the highest observed fish concentration of 143.3 ppt is matched by the highest predicted fish concentration of 89.2 ppt.

Although the variance, r2, of 0.27 is not compelling evidence that the model is valid, one must consider the usefulness of this measure. It would certainly be more meaningful if several discharge measurements per mill were made and several fish measurements were made downstream of the discharge. In fact, only one discharge measurement was made per mill and a very limited number of fish samples were taken per mill. To be more rigorous, the several measurements of discharge would have to have been made over time to best reflect an average discharge.

table Figure 7-2 Comparison of predicted and observed fish tissue concentrations for validation of the effluent discharge model.
Likewise, other mill-specific parameters are uncertain, such as receiving water body flow, suspended solids in the water body, and so on.

Finally, and perhaps most importantly, the assumption of this exercise is that the mill discharges of 2,3,7,8-TCDD represent the only sources impacting the fish. This is most unlikely to be the case for the large receiving water bodies, which may be receiving other industrial point discharges or non-point sources (runoff, atmospheric deposition).

Given the factor of 2 difference in average predicted and observed fish concentrations

expand table Figure V3 7-2

for the low receiving water bodies, and the factor of 10 difference between predicted and observed for 73% of the fish results over all mills, one might cautiously conclude that the effluent discharge model of this assessment is generally valid for purposes of general discharge assessment.

Given this last cautious statement, one can continue this exercise by attempting a calibration on an appropriate parameter(s) so that predictions better match observations. The appropriate parameter for calibration is the BSSAF. The choice of 0.09 for the 2,3,7,8-TCDD BSSAF was based on data from Lake Ontario (EPA, 1990d). Specifically, 0.09 was the BSAF - lake bottom sediment to fish lipid accumulation factor - for measured fish and bottom sediments of Lake Ontario. As this is a lake and not a riverine situation, and inasmuch as 2,3,7,8-TCDD contamination of Lake Ontario sediments have been attributed to historical impacts and not ongoing causes, the 0.09 may be inappropriate.

As well, a range of BSAFs for 2,3,7,8-TCDD were noted in the literature in Chapter 4 of this assessment, ranging from less than 0.05 to greater than 1.00. EPA (1993) suggests that data collection methods limit the usefulness of some of the available literature, particularly those showing very high BSAF, and in a similar examination of BSAF data, suggests a range of 2,3,7,8-TCDD BSAF from 0.03 to 0.3. In any case, this suggests that the BSSAF is a reasonable candidate for calibration in this exercise.

A different selection for BSSAF significantly improves model performance. If the BSSAF is increased to 0.20 (up from 0.09), the average predicted fish tissue concentration for the 38 mills discharging into the smaller water bodies increases as expected from 7.0 to 15.6 ppt, comparing better now to the average observed concentration of 15.0 ppt. Also, the factor of 10 and 20 test including all mills now improves to 84% of predictions within a factor of 10 of observations, and 87% of predictions within a factor of 20.

Conclusions from this exercise include:

1. For at least smaller receiving water bodies, those with harmonic mean flows on the order of 107 to 109 L/hr, the effluent discharge model is appropriate for assessing effluent discharge impacts to fish for 2,3,7,8-TCDD and perhaps other dioxin-like compounds.

2. There appears to be a distinction in model performance for the large and small receiving water bodies. The high suspended solids concentrations generated in an earlier modeling exercise for the larger water bodies is one cause for model underprediction; these solids concentrations should be further reviewed. Also, these water bodies were evaluated by NCASI as ones with multiple sources. Other sources not identified by NCASI could also have been the cause for higher measured fish concentrations as compared to model predictions.

The NSCRF report (EPA, 1992b) contains an appendix giving a matrix indicating point source categories of discharges which may have affected fish

concentration results. Pulp and paper mills with and without chlorine were listed as point sources for 125 episodes (an episode is a fish sampling site). In 37 of these episodes, other point sources were identified, including one or more of the following: refinery (refinery using the catalytic reforming process), NPL site (a Superfund site), or other industry (an industrial discharge other than a paper mill or refinery). Given other sources, it is in fact a benefit to the exercise that predictions were lower than observations.

3. The model more closely predicts fish concentrations for the smaller receiving water bodies when the BSSAF is calibrated from 0.09 to 0.20. Considering that 0.09 was a value for 2,3,7,8-TCDD developed with data from Lake Ontario, a standing water body with principally historic and not ongoing 2,3,7,8-TCDD impacts, this setting is probably an inappropriate surrogate for ongoing discharges to a riverine situation. This would argue that a calibration is warranted.