PUBLIC HEALTH ASSESSMENT

Community Exposures to the 1965 and 1970 Accidental Tritium Releases

LAWRENCE LIVERMORE NATIONAL LABORATORY, MAIN SITE (USDOE)
LIVERMORE, ALAMEDA COUNTY, CALIFORNIA


APPENDIX 1: COMMENTS ON THE PUBLIC RELEASE VERSION OF PHA

ATSDR Responses follow each comment and indicate any changes that were made to the document as a result of the comment, or explains why changes were not made. The comments, which are presented verbatim, resulted in numerous changes from the public release version of the document (dated May 24, 2002).(18)


Reviewer 1 Comments

1. The assessment addresses all potential exposure pathways with one possible exception: the skin absorption pathway. I assume that skin absorption has been included in the inhalation pathway but no statement to this effect was made in the report. This point should be clarified when the report is revised.

ATSDR Response: A section on skin absorption has been added to Section 3 and a dermal absorption component has been added to the total tritium dose. The dermal absorption component is assumed to be equal to the HTO inhalation component.

2. The general model is appropriate. The level of sophistication and detail is consistent with our current understanding of the environmental transport of tritium following an HT release. Good use has been made of the available experimental data to augment the theory. The use of Monte Carlo analysis to estimate the uncertainty in the predicted doses is commendable, although I have reservations (described in the attachment) about some aspects of its application in this case. The model is sufficiently conservative that it is unlikely that the predicted doses will underestimate the true doses.

Having said this, several well-established environmental tritium codes (UFOTRI, ETMOD, STAR-H3, TRIF) are available for calculating doses from short-term HT releases. These codes are based on the available experimental data and in some cases have undergone extensive testing. Given the existence of these codes, use of the semi-empirical approach adopted in the assessment should be justified. The credibility of the assessment would be enhanced by comparing its predictions with those of one or more of the existing codes.

ATSDR Response: The purpose of this analysis is to determine if the exposures represent a potential health hazard. As the overall model "is sufficiently conservative that it is unlikely that the predicted doses will underestimate the true doses." comparison with the other referenced models is not necessary from a public health perspective.

3. The RASCAL model is a standard Gaussian dispersion model, the type of model normally used to calculate air concentrations following a short-term release. It gives results that are somewhat higher than those of Canadian standard N288.2, which is also a Gaussian model. However, the meteorological data used in RASCAL for both releases are suspect and consideration should be given to repeating the calculations with more appropriate parameter values (see the attachment for more details). The ISCST code is acceptable for calculating air concentrations from a ground level area source. When combined with experimental data on the loss rate of tritium from soil, the model produces time-integrated HTO concentrations in air that are about a factor of 2 higher with respect to experimental data. The intake of tritium via ingestion is determined from monitoring data, which is a good approach in theory but may underestimate the true intakes in this case. The OBT model is not realistic and results in large overestimates of OBT concentrations in foods. The models used to calculate doses from the environmental concentrations are state-of-the-art.

ATSDR Response: The meteorological data have been revised by use of an atmospheric stability category of "E" (which results in slightly higher concentrations than an "F" category). We have also modified temperature and plume rise factors based on newly available data (Peterson and others, 2002; Quarterman, 1965). The RASCAL results have also been compared with output from the HOTSPOT model, which is another Gaussian plume dispersion model. After the above meteorological adjustments, tritium air concentrations and deposition to soil are similar from the two models, although the RASCAL model does provide a slightly higher estimated concentration. Another adjustment to the model, as suggested in other comments, includes a broader range of HTO emissions from soil. These adjustments are described in the text and referenced as necessary. Tritium intake from ingestion will continue to be based on measured monitoring data. Although such measurements are subject to the uncertainties of sampling and analytical error, there is no a priori rationale for assuming modeled tritium concentrations will more accurately represent actual values.

4. Although I'm not an expert in the area of health effects and cannot comment authoritatively on this aspect of the work, I was able to understand the relationship between exposure and health effects as described in the assessment. A considerable amount of experimental data is presented, all of which is consistent in its implication that health effects are unlikely to occur at thelevel of the predicted doses.

ATSDR Response: Comment noted.

5. The accuracy of the predicted doses is in doubt because of problems with the meteorological data, the use of monitoring data to estimate ingestion doses and the underestimate of the uncertainty in the predicted air concentrations (see the attachment for more details). I believe the calculations of inhalation dose should be repeated using the correct values of the meteorological parameters. In these calculations, unnecessary conservatisms should be dropped and realistic standard deviations assigned to the air concentrations so that the output distributions are meaningful. The use of monitoring data to estimate ingestion doses should be justified or the doses should be calculated using a model to ensure they are not underpredicted. In either case, the OBT model should be modified to more closely reflect reality.

The information is presented clearly and is understandable, with a few exceptions:

  • The discussion of HT deposition, oxidation and re-emission is confusing at times. I have made suggestions for wording changes to improve the clarity.
  • Some parameter values (child breathing rates, dosimetric data for HT) and some intermediate outputs of the calculations (time-integrated HTO concentrations in air) are missing and should be included to allow readers to check the calculations.
  • The assumptions behind the OBT model are not spelled out.
  • The Crystal Ball reports of the output distributions are presented without explanation. It's possible to understand them but they are not user-friendly.

ATSDR Response: The predicted doses are recalculated as described above and in the document text. The discussion of HT deposition and OBT and HT dosimetry has been modified as suggested and parameter values for child breathing rates and HT dosimetric data added to document (in main document or appendices, as appropriate). Results of the Crystal Ball output that are included in the main document are discussed as suggested. Potential uncertainty in the estimated HT and HTO air concentrations has been addressed by conducting separate ISC model runs using mean and 95th percentile HTO emission rates. The resulting mean and 95th percentile HTO air concentrations were used to construct a lognormal probability distribution of HTO air concentrations at breathing height (1 m).

6. If the doses predicted in the assessment are accepted at face value, the conclusions and recommendations are appropriate. These doses are clearly below the level at which any health effects could occur. However, as noted above, there are problems with the accuracy of the dose estimates. Until these are revised, no formal conclusions can be drawn regarding health effects or the need for follow-up actions. Having said this, I fully expect that conclusions based on the revised estimates will be the same as those made in the assessment, namely that the exposures are not likely to produce adverse health effects, that doses are below levels of public health concern and that no specific recommendations are warranted.

ATSDR Response: Comment noted. The revised doses are similar to previously calculated doses such that the conclusions and recommendations are unchanged from the public release version of the PHA.

7. Community members near LLNL requested funds to hire an independent technical consultant to review the assessment. Although ATSDR does not have a process for providing funds to the community, it might consider involving the public in the selection of reviewers and sharing the results of the review with the community.

ATSDR Response: Community members were provided with an opportunity to recommend nominees for the peer review process. The person they recommended was solicited by ATSDR as one of the peer reviewers. However, this person, who was eminently qualified, indicated that he no longer accepts or conducts such peer reviews. The community members subsequently obtained an outside grant to pay for an independent expert review and ATSDR has extended the comment period to allow adequate time for those reviewers to complete their work. ATSDR has also recently presented the results of this public health assessment to the Livermore community (February 18, 2003) along with the major revisions prompted by the review process.

8. Incorrect meteorological data appear to have been used for both releases. Site-specific data for the 1965 release have recently come to light (Peterson et al. 2002). Use of these data would result in lower doses than those presented in the report, so it may not be necessary to re-do the calculations for this release. According to Myers et al. (1973), the stability during the 1970 release was class E or F rather than class B as assumed here. If this is the case, the predicted doses would increase substantially. Problems exist with the air concentrations even if the meteorological information used in the assessment is assumed to be correct (see items 26, 34 and 36 under Specific Comments), so it may be best to repeat all the dispersion calculations.

ATSDR Response: The PHA includes revised dispersion and dose calculations based on the newly available meteorological data for both the 1965 and 1970 releases.

9. There seems to be some confusion over the processes of HT deposition, oxidation and re-emission and over the various conversion and loss rates coming out of the 1987 HT experiment at Chalk River. HT is converted to HTO very soon after it diffuses into the soil. Any tritium remaining in the soil after the exposure is in the form of HTO; any HT that is not converted quickly diffuses back into the atmosphere. It therefore makes no sense to talk about HT concentrations in soil; the only tritium present in the soil is in the form of HTO. The HT deposition velocity accounts for both HT deposition and oxidation to HTO. Brown et al. (1988) discuss an effective oxidation rate of 1.5%/h. This value was calculated by dividing the ratio of HTO to HT air concentrations observed at the end of the release at a downwind distance of 50 m by the travel time from the source to 50 m. As such, it is a very specific parameter and can't be generalized to other situations. Ogram et al. (1988) measured the loss rate from soil and found values between 0.5 and 1%/h.

ATSDR Response: The discussion of HT deposition, etc., has been modified as suggested. A broader range of HTO loss rates have been utilized in the calculations based on additional studies, which are referenced in the document.

10. I did some calculations of my own to verify the accuracy of the results in the assessment. The HT concentrations in air predicted by RASCAL were about a factor 2 higher than concentrations from the Canadian Standard dispersion model N288.2, which is also a Gaussian plume model. The reason for this difference is not obvious but the RASCAL results are conservative. Results from the HT release experiments in France and Canada suggest that the ratio of the time-integrated HTO concentration in air to the time-integrated HT concentration is about 0.1. The ratio in the assessment is 0.2, indicating that the HTO concentrations are conservative by an additional factor of 2, perhaps because the initial concentration on day 1 is overestimated (see item 19 under Specific Comments). Thus the inhalation doses in the assessment appear to be the right order of magnitude, although conservative. However, I believe the ingestion doses, which were calculated from monitoring data in the assessment, could be underestimated. Predictions of a simple model suggest that the HTO ingestion dose could be as high as 20 mrem under assumptions similar to those made in calculating the inhalation dose. On the other hand, the contribution of OBT to the ingestion dose is severely overestimated in the assessment. Even with a much increased ingestion dose, the conclusion holds that the releases posed no health risk to members of the public.

ATSDR Response: The RASCAL results have been compared with results from HOTSPOT, which is another Gaussian plume dispersion model. Although the results are similar, RASCAL results are about 20% higher than HOTSPOT results, which is acceptable for the purpose of initial exposure assessment. The revised inhalation doses are based on maximum 12 hour average HTO concentrations rather than the maximum 1 hour average that was used in the previous estimate.

According to representatives of the California Agricultural Extension Service, very little vegetation is growing in the Livermore area during August. Cattle are being fed from stored forage and any garden crops must be irrigated to survive during this time period. Consequently, the measured tritium concentrations and estimated ingestion doses are probably an accurate estimate of exposure.

11. Although I support the use of stochastic assessments, I'm concerned over some of the distributions used in the analysis. These are by necessity subjective, and have to be carefully thought out and justified if the output distributions are to be meaningful. I believe the standard deviations for the HT and HTO air concentrations are much too small. Since these concentrations underlie the inhalation dose predictions, the uncertainties in the doses are likely much higher than indicated in the report. Moreover, I believe most of the distributions should be lognormal rather than normal, as most environmental parameters are (Hoffman 1979, Sheppard and Evenden 1988, Zach et al. 1989, Ott 1990, Blackwood 1992). Finally, the distributions should not be adjusted to conservative values, as recommended in the last paragraph on page 22. Conservative values should only be used in deterministic assessments. If they are used in stochastic assessments, the output distribution no longer has any meaning in terms of the probability of consequences. Obtaining a realistic estimate of this probability is the main reason for doing a stochastic assessment.

If the distributions are widened as much as I suggest, the 95th percentile doses may become larger than desirable. This is the cost of including conservatisms in a stochastic analysis; the solution is to reduce the conservatisms as much as possible. RASCAL appears to overestimate air concentrations but I'm not sure why. The time-integrated HTO concentration in air seems to be overestimated by an additional factor of 2, perhaps because the initial concentration on day 1 is overestimated, or because all the deposited HT is assumed to be re-emitted. The ingestion doses are not an issue if the monitoring data can be defended but doses predicted from model calculations could be reduced by making more realistic assumptions about the diet and location of the cow, the fraction of food that is contaminated and so on.

ATSDR Response: The revised dose calculations are based on a probability distribution of the HT and HTO air concentrations as described above and in the revised document. As indicated, the revised doses have a broader range of uncertainty, although mean doses are very similar. Output from a Gaussian dispersion model should be assumed to be normally distributed, as that is an underlying assumption of the analytical method. It should be further noted that many lognormal distributions of environmental parameters represent inappropriate sampling methods; the data are skewed because of poor spatial representation of the underlying population. A normal probability distribution should be used if the expected values are as likely to be greater than the mean as they are to be less than the mean. We agree that realistic values should be the basis of the probability assessment and have reviewed the bases for the required probability distributions.

12. Page (iii), paragraph 2, line 6 and elsewhere: The 1965 release occurred on January 20, not January 21.

ATSDR Response: The appropriate changes have been made.

13. Page (iii), paragraph 3, line 3: Replace "such as" with "known generically as".

ATSDR Response: This change has been made.

14. Page 8, paragraph 1, line 9: Replace "into HTO, or soil moisture" with "into HTO (tritiated water), which becomes associated with soil moisture".

ATSDR Response: This change has been made.

15. Page 8, paragraph 1, lines 9-10: Replace "Following transformation into HTO, a significant portion of the soil moisture is re-emitted" with "A significant portion of the tritiated soil water may be re-emitted".

ATSDR Response: This change has been made.

16. Page 8, paragraph 3: The references in lines 5 and 7 to the "soil HT concentration" should be changed to "soil HTO concentration". Also in line 5, replace "transformed to HTO" with "emitted to the air". In the same line, I'm not sure what "additional factors" refer to here. Either spell them out or delete the phrase in parentheses. Finally, uptake by plants and animals should be included specifically in the list.

ATSDR Response: This paragraph has been re-written as suggested.

17. Page 9, Figure 3: Replace the text in the middle right ("HTO dispersion") with "HTO emission from soil and dispersion in air". Also, replace the last sentence in the text to the right in the box with "The emission rate declines as HTO is lost from the soil".

ATSDR Response: These changes have been made.

18. Page 10, paragraph 1: It's stated here that the maximally exposed individuals are assumed to 0.5 miles from the source in 1965 and 1 mile in 1970. Elsewhere (page (iii), paragraph 2) the distance for the 1965 accident is given as "about 1 mile". This point should be clarified because the distances are invoked later in the assessment (Table 2) to justify the use of 1970 results alone, on the assumption that the 1970 doses were higher than those for 1965. This may not be the case if the MEI in 1965 was indeed at 0.5 miles.

ATSDR Response: This was a typographical error and has been corrected. The MEI for both releases was at the 1 mile location. Also, as a result of changing the atmospheric stability, the maximum HT concentration is also at the 1 to 1.5 mile location and has been so noted in the document.

19. Page 10, paragraph 2, line 4: Replace "the dispersion of HTO from the soil surface into the breathing zone" with "the emission of HTO from the soil and dispersion in the atmosphere".

ATSDR Response: The text has been changed as suggested.

20. Page 10, paragraph 4, lines 1-4: Replace the first two sentences with "The loss of HTO from soil to air results in a continually declining concentration of soil and air HTO".

ATSDR Response: These sentences were shortened and clarified.

21. Page 11, paragraph 2: Myers et al. (1973) note that a surface-based inversion probably prevailed over the Livermore Valley at the time of the 1970 release and that the atmosphere was very stable. This means the stability class was more likely F than B, which makes sense for that time of day. Predicted air concentrations for class F could be very different from those presented here for class B. The doses for the 1970 release should be recalculated using class F stability.

ATSDR Response: The previous assumption of a "B" classification was based on slight or moderate solar insolation. As sunrise tables indicate that the release occurred before sunrise, the release was recalculated using both "E" and "F" classifications. All text, tables, and figures have been appropriately modified.

22. Page 11, paragraph 3: It's artificial and possibly confusing to identify inhalation with the instantaneous concentration and deposition with the cumulative concentration. Both inhalation and deposition are calculated in the same way from the (assumed constant) air concentration during the passage of the plume and the release duration. This same remark applies to paragraph 4 on page 34 in Appendix 1.

ATSDR Response: Both inhalation and deposition must include a time component. In order to include a time component in the HT inhalation exposure the concentration must be based on the instantaneous air concentration in order to integrate the breathing rate over a 30 minute period. As the cumulative concentration is already integrated over time, it would be incorrect to integrate a cumulative air concentration over time. No changes have been made.

23. Page 11, paragraph 4: In line 1, replace "cumulative HT soil concentrations" with "cumulative HTO soil concentrations". In lines 2-3 replace the reference to Brown et al. with Ogram et al. (1988). Both Brown and Ogram presented results from the 1987 HT release at Chalk River but the deposition velocity results are in Ogram. In line 6, replace HT with HTO.

ATSDR Response: The HT-HTO changes have been made as suggested. We have also changed the reference and included additional references to HT deposition velocities.

24. Page 11, paragraph 5, line 1: Replace "deposited onto" with "diffused into".

ATSDR Response: This change has been made.

25. Page 12, lines 1- 2: The first 2 lines on page 12 should be changed from "by Brown et al. (1988), the transformation and re-emission " to " by Ogram et al. (1988), the re-emission".

ATSDR Response: This paragraph has been re-written, the above reference has been changed, and additional references added.

26. Page 12, paragraph 3, line 4: Replace 1.82E-03 Ci-sec/m3 with 1.82E-03 Ci/m3.

ATSDR Response: The value has been recalculated and the units changed.

27. Page 12, paragraph 5: In line 1, replace Brown with Ogram. In line 2, delete "transformation and". In lines 5, 6, 7, 8, 10 and 11, replace "transformation" with "re-emission".

ATSDR Response: This section has been completely re-written and includes these changes.

28. Page 12, paragraph 5, line 4: My own unpublished analysis of the short-term HT release carried out at Chalk River in 1987 suggests that the time-integrated HTO concentration in air arising from re-emission from soil and plants was about 4% of the HT concentration in air integrated over the release period (30 minutes) at a downwind distance of 400 m. The corresponding value from the 1986 French release (2 minute release, 800 m downwind) was 8%. In both experiments the data on HTO concentrations in air covered a period of only 4 days after the release, so the HTO/HT ratios integrated over all time would be larger and a value of 10% is reasonable. The ratio of the concentrations in the present report is about 20%, indicating that the estimate of the time-integrated HTO concentration in air is conservative by about a factor 2.

ATSDR Response: Although we have not recalculated the HTO/HT concentration ratios, the revised dose calculations are based on 12 hour average concentrations rather than 1 hour values. This change should effectively address the over-estimation of the HTO air concentrations.

29. Page 12, last paragraph: Replace "HT" with "HTO" in lines 1, 3 and 4.

ATSDR Response: These changes have been made.

30. Page 13, paragraph 1: It's not clear from the discussion in this paragraph how the highest one-hour tritium concentration from 1991 or 1993 can substitute for hourly meteorological data over a 12-day period. This is explained better in the second paragraph on page 41 and that explanation should be given here. Having said this, I don't understand why it's necessary to go through this process. It may be true that hourly meteorological data are not available for the release periods but the ISCST code is used only in the first hour, for which meteorological conditions are known. Basing the time-integrated HTO concentrations on the meteorological conditions in effect at the time of the release, rather than on the conditions that lead to the maximum concentration, could reduce the conservatism of the HTO inhalation dose considerably.

ATSDR Response: The available site-specific meteorological data for 1970 (or 1965) are not sufficient to run the ISC model. Because the HTO is emitted from the entire footprint of the HT plume, we felt it was necessary to use ISC or a similar model that could integrate breathing zone concentrations from an areally distributed source. The worst case conditions during a five year period are generally considered to be a reliable proxy for the site and time-specific meteorological data. This paragraph has been re-written to include the paragraph from page 41.

31. Page 13, paragraph 3: In line 1, replace "The HT concentration" with "The HTO concentration", and "Brown et al. (1988)" with "Ogram et al. (1988)". In lines 2-3, replace "The HTO concentration" with "The HTO air concentration". In lines 3-4, delete "of HT and HTO".

ATSDR Response: This paragraph has been deleted and figure 4 replaced with 2 figures showing the estimated HTO emission rates and the probability distribution of the HTO air concentration.

32. Page 13, paragraph 4, line 1: What exactly does "HTO concentrations in plant moisture are in equilibrium with soil moisture" mean? That concentrations in plants and soil are the same or just proportional? Spencer et al. (1988) show that the vegetation/soil (0-2 cm) ratio is about 1/3 after the lag of 12-24 hours.

ATSDR Response: The paragraph has been changed to clarify that the concentrations are proportional.

33. Page 13, paragraph 4, line 6: Replace "underestimates the decline of the HT soil source" with "overestimates the decline of the HTO soil source".

ATSDR Response: This sentence has been rewritten as suggested.

34. Page 14, Figure 4: In the heading, in the key, in the axis label on the left and in the caption, replace "soil HT" with "soil HTO".

ATSDR Response: This figure has been replaced as described above (comment 20).

35. Page 14, paragraphs 1 and 2: In a review of the 1965 release, Peterson et al. (2002) state that the wind was from the southwest (230o - 250o) at 3.6 m/s at the time of the release, and that the stability was class B. This information could be used to generate new predictions for this release, although this is perhaps not absolutely necessary. The present analysis is conservative since the lower wind speed and greater stability will result in higher concentrations than would be obtained using the real data.

ATSDR Response: The newly available information has been used to recalculate the 1965 doses.

36. Page 15, paragraph 2: In line 4, replace "cumulative soil HT loading" with "cumulative air HT concentration". In line 5, replace 3.0e-4 with 4.0e-4. In lines 5-6, replace "the HT to HTO conversion rate" with "the HTO loss rate from soil". The same comments apply to the last sentence in the caption to Table 1.

ATSDR Response: These sentences have been rewritten as suggested.

37. Page 15, paragraph 3: Tables A-2 and A-3 show that the predicted time-integrated HT concentration in air was lower in 1965 than in 1970 close to the source but higher at 0.7 km and beyond. This is to be expected for an elevated release and the release rates, wind speeds and stabilities (class B in 1970 and class C in 1965) assumed in the assessment. Since the HTO emission rates from soil are proportional to the integrated HT concentrations, I expected to find the same pattern in Table 1. But this was not the case, as the table shows higher emission rates for 1970 than for 1965 at all downwind distances. I was able to reproduce the values for 1970 and believe these to be correct. But I wasn't able replicate the 1965 values and suspect an error here.

ATSDR Response: There was an error in entering the 1965 cumulative HT (Table A-3) values. The plume centerline (30o) cumulative HT concentrations should have been 5.6, 17.8, 14.0, 7.3, 4.3, 2.4, 1.2, and 0.7 (Ci-sec/m3) for distances of 0.1, 0.2, 0.3, 0.5, 0.7, 1.0, 1.5, and 2.0 miles (respectively). This error led to another error in calculation of the HTO emission rates in Table 1. Using the correct values, relative to the 1970 values, the 1965 cumulative HT concentrations and HTO emission rates were higher at on-site locations, but lower at off-site locations (1 mile or more from stack). All of these values have been recalculated and the tables re-written using the actual weather data.

38. Page 15, paragraph 3, line 3: Since the atmosphere was more stable (class C) in 1965 than 1970 (class B), there was decreased atmospheric dispersion in 1965, not increased dispersion as stated. The plume would have covered a smaller area of ground in 1965 than in 1970 and a more accurate emission rate would have been obtained with smaller rectangles rather than larger ones.

ATSDR Response: As stated above, the stability classes for the releases have changed and all concentrations, etc., have been recalculated. The text has been re-written to reflect these changes. Also, because of greater certainty regarding meteorological conditions in 1965 and because the cumulative HT concentrations and HTO emission rates are lower for the 1965 release than for the 1970 release, we have not re-calculated the HTO dispersion using ISC for the 1965 release. For the previous calculations, we used the larger footprint areas because of greater uncertainty in the wind directions.

39. Page 16, paragraph 1: Give an idea of the fraction of the time-integrated concentration that is missed by stopping at 12 days rather than at infinity. This will justify stopping at 12 days, which otherwise seems arbitrary.

ATSDR Response: A justification for the 12 day exposure duration is given on page 17 (paragraph 3) and in Figures 4 and 5 (public release version). The 12 day period accounts for >95% of the soil HTO and based on HTO emission rates more than 4 environmental half lives. Inhalation doses for day 13 are less than of the daily background rate and we believe are accounted for in the chronic doses which are added to the acute doses estimated in this evaluation. A footnote with this justification has been added to this section of the revised PHA.

40. Page 16, paragraph 3: Change the first sentence to read "The 30-minute HT inhalation dose is calculated from the maximum instantaneous HT concentration and the release duration". In lines 3-4, replace "as the HT deposited on the soil surface is transformed and dispersed through the environment" with "as the HTO in the soil is dispersed through the environment".

ATSDR Response: These changes have been made.

41. Page 16, paragraph 4: In item 6, replace "Table 1" with "Table A-1". In item 7, replace "Table 2" with "Tables A-2 and A-3". Replace item 9 with "HTO loss rates from soil on the basis of measured data (Ogram et al. 1988)".

ATSDR Response: These changes have been made.

42. Page 17, paragraph 3, line 9. The loss of HTO from the soil follows first order kinetics. The 1% loss rate applies to the HTO that remains in the soil at a given time, not to the initial concentration at the end of the exposure. After 24 hours, the amount left in the soil is 0.9924 = 0.786 times the initial concentration, not 0.76. This is a small difference after one day but it becomes more important at longer times. For example, if the loss were applied linearly, there would be no HTO left in the soil after 100 hours, when in fact the concentration is still about one-third of the initial concentration.

ATSDR Response: This section has been re-written and no longer includes these sentences. We have also included a specific statement describing the HTO loss rate as an exponential decay parameter.

43. Page 17, Table 2: I tried to reproduce the numbers in this table but found I didn't have enough information to do so. In a report of this kind, it is essential to provide enough information to allow readers to check the calculations if they have a mind to. Dosimetric data for HT should be added to the report, plus breathing rates for children and predicted time-integrated HTO concentrations in air from the ISCST code.

ATSDR Response: HT dosimetric data, child breathing rates and time-integrated HTO air concentrations have been added to the appendices as suggested.

44. Further to Table 2: The HT dose rates at 1 and 2 miles should be in the same ratio as the HT concentrations in air at these distances. The concentration ratio from Table A-2 is 3.27/0.962 = 3.4 whereas the dose ratios range from 1.67 for children to 2.0 for adults. There seems to be something wrong here. Moreover, I couldn't reproduce the HTO adult dose at 1 mile. Taking the initial HTO concentration in air as 2.1 x 10-6 Ci/m3 from Appendix 3, and assuming a 1%/hour decrease, I calculate the integrated HTO air concentration over 12 days to be 0.71 Ci s/m3, or 2.64 x 1010 Bq s/m3. With a breathing rate of 1.41 x 10-4 m3/s, the total activity taken into the body due to inhalation over the 12-day period is 3.71 x 106 Bq. Adding 50% to this to account for skin absorption gives a total intake of 5.57 x 106 Bq. Using this as the "concentration" in the equation on page 40 yields 4.88 x 10-5 Sv or 4.88 mrem. The value given in Table 2 (19 mrem) is 3.9 times higher than this and so is conservative but the difference needs to be accounted for.

ATSDR Response: We believe the differences in the dose ratios may be an artifact of rounding error. Rounding off of these small numbers can provide significant changes in the ratios. As we have completely recalculated the doses and concentrations, if there was an editing or transcription error in the previous table, it has been corrected by insertion of the revised values. With regard to the overall magnitude of the inhalation doses, we believe the high values are largely a result of using 1 hour maxima (for each day). As stated in the response to comment 3, the revised dose estimates are based on use of the 12 hour average values and are more similar to those estimated in this comment.

45. Two final points regarding Table 2. (i) If doses from the 1965 release are indeed higher than those from 1970 at 1 and 2 miles (or if the MEI is 0.5 miles from the source rather than 1 mile), then the numbers in the table may have to be based on the 1965 data. At the very least, the phrase "due to increased dispersion" should be removed from line 5 of the caption since dispersion was less in 1965 than in 1970. (ii) Do the doses include the contribution from skin absorption? If not, they should. If they do, this should be mentioned so it's clear that an important exposure pathway hasn't been missed.

ATSDR Response: The MEI for both releases was located 1 mile from the tritium facility stack. The newly available 1965 meteorological data indicates that the centerline of the 1965 plume was ~60o from the tritium facility stack. Based on this revised wind direction, we have recalculated the number of people and distance to the potentially exposed population for the 1965 release. As with the previous calculations, potential doses from the 1970 release are higher and more people were potentially exposed. Consequently, we are continuing to base our maximum exposure estimations on the 1970 release. The reason that the 1965 doses are lower than the 1970 doses is because of decreased dispersion of the 1965 release. The caption has been modified to clarify that the lower dispersion is for the 1965 release. A skin absorption factor (100% of HTO inhalation) has been added to the total doses.

46. Page 19, paragraph 1: Although I support the use of monitoring data to calculate doses in general, the data must be reliable. More information is needed before you can convince me that this is the case here. Exactly where were the plant and milk samples taken? If they're going to be used to estimate doses, they must be taken on the plume centerline (as determined by measurements and not by wind direction) at the downwind distance of the maximally exposed individual. When were the samples taken? They must be taken when the plant concentrations peak. The fact that the milk concentrations were so much lower than the vegetation concentrations makes me even more suspicious. As a check on the values given in the report, I estimated ingestion dose using a simple model that's described in an Appendix to these comments. I found doses of about 20 mrem for ingestion of fruits and vegetables, two orders of magnitude higher than the values based on the monitoring data. Two conclusions are possible from this: either the models are overpredicting or the monitoring data is underpredicting. I suspect it's a combination of the two, but some recognition should be given in the report to the possibility that the ingestion doses could be higher than indicated from the monitoring data.

ATSDR Response: As stated in the response to comment 3, the actual measurements of vegetation and milk appear to be representative of growing conditions at the time of the 1970 release. Winter conditions, as during the 1965 release, are much wetter such that cattle may be consuming pasture grasses during that time. However, there are a limited number of cold weather garden crops that will be harvested and consumed during that timeframe such that the overall ingestion dose from the 1965 release is unlikely to be significantly higher than that estimated for the 1970 release. Also, newly available 1965 weather data indicates that it rained shortly after the 1965 release such that soil HTO concentrations, and water taken up by plants, would be greatly diluted.

The specific locations and timeframes of the measured samples are uncertain. Although we used those locations with the highest measured values, it is possible that higher values may have been present in the environment. We have acknowledged that uncertainty by assuming that the highest measured value represents the 90th percentile value. This distribution assumes that 10% of the samples will have higher values. On the basis of all of these factors and the human urinanalysis of the people living in the plume area that did not indicate any detectable tritium exposures, we are very confident that the models over-estimate the total tritium doses.

47. Ingestion doses are not calculated separately for the 1965 release on the assumption that air concentrations were higher in 1970. But, as pointed out above, the 1965 concentrations were higher at and beyond 0.7 miles, and the 1965 concentrations at 0.5 miles were higher than the 1970 concentrations at 1 mile. Thus the monitoring data from 1970 would underestimate the ingestion doses in 1965. Separate estimates should be carried out for 1965.

ATSDR Response: As indicated above, the 1965 concentrations were higher at on-site locations, but lower at all areas of potential off-site exposure. Although there may be some difference in the ingestion doses for the two releases, overall the 1970 release would have resulted in higher total tritium doses. As those doses are not likely to result in adverse health effects, there is no need for separate estimates of the 1965 ingestion doses.

48. Page 19, paragraph 2: The monitoring data likely underestimate the ingestion dose from milk as well as from fruits and vegetables. It would be desirable to model this dose, but reliable concentrations in milk are difficult to calculate because of the uncertainty in the amount of tritiated water taken in by the cow with its feed. Also, milk concentrations would not follow the air concentration. They would build up initially as the air concentrations drop off and decrease more slowly than the air concentrations once they reach their peak. But the time-integrated concentration calculated assuming a 1%/h decrease would probably give a result that wasn't far off reality. Using the vegetation concentrations from the Appendix, an ingestion rate of 40 L/d vegetation water by the cow, a fraction of daily intake appearing in 1 kg of animal produce of 0.014 d/L and a milk ingestion rate of 1 L/d, the predicted ingestion dose is about 37 mrem, or about the same as the modeled dose from fruits and vegetables. The possibility that ingestion doses from milk could have been substantially higher than those based on monitoring data should be acknowledged.

ATSDR Response: Language regarding uncertainty of the ingestion doses has been added to the text.

49. Page 19, paragraph 4: From the description of the OBT model given here, it appears that the OBT concentration in plants is assumed equal to the HTO concentration and that the total activity of tritium ingested in the form of OBT equals the activity ingested as HTO. Neither of these assumptions is defensible. OBT concentrations will behave very differently than HTO concentrations, starting out very low and showing a steady increase over the 12-day assessment period rather than a decrease. The intake of tritium in the form of OBT will be much less than the intake of HTO because most plants are primarily water. Using data from the 1987 experiment at Chalk River (Spencer et al. 1988), I calculated the dose from ingestion of OBT in fruits and vegetables as 0.7 mrem over the first 12 days. This is trivial compared to the modeled ingestion dose from HTO. The OBT concentration and the corresponding ingestion dose will stay relatively high over the next month or two but would still contribute relatively little (~ 20%) to the total time-integrated ingestion dose.

ATSDR Response: We have relied on the findings of an expert panel to determine the relative dose contribution from OBT (ATSDR 2002). The revised calculations and dose estimates assume an average increase of 32% of the HTO ingestion dose due to the contribution of OBT. Consequently, about 2/3 of the resulting ingestion doses are due to HTO ingestion and about 1/3 due to OBT ingestion.

50. Page 20, Figure 6: In the second line of the caption, replace "most likely" with "average". The most likely dose occurs at the peak of the distribution.

ATSDR Response: The caption has been changed.

51. Page 21, Table 3: The ratio of mean to 95th percentile doses is the same for adults and children for the inhalation doses but not for the ingestion dose. Why is the 95th percentile so close to the mean for the adult ingestion dose?

ATSDR Response: The ratios of mean to 95th percentile values for adult and child ingestion doses are similar for the revised dose calculations. An exception to this which provides an example of rounding error is in Tables 2 and 3 of the revised document. The 1-mile adult 30 minute HT inhalation dose shows a mean value of 0.1 mrem and 95th percentile value of 0.2 mrem. The ratio of 95th% to mean value is 2, however, the mean value was rounded up from 0.05 and the ratio using unrounded values is 4. The rationale for using rounded numbers is our underlying belief that due to inherent uncertainties, none of the calculated numbers have any significance beyond one, or possibly two, significant digits.

52. Tables 2 and 3: Make sure the numbers in these tables are consistent. For example, the 95th percentile 12-day HTO inhalation dose to a child is given in Table 2 as 130 mrem but in Table 3 as 134. Also, the total doses in mSv in Table 3 are not exactly 0.01 times the doses in mrem.

ATSDR Response: These tables and their values have been corrected for consistency.

53. Page 21, paragraph 3: Why are Myers' estimates of dose from milk ingestion so much higher than the present estimates? His approach and results should be critiqued or readers may wonder if he is right and the present estimates wrong.

ATSDR Response: Myers et al. (1971) estimate of dose from milk was based on a hypothetical cow consuming pasture grasses contaminated at the maximum measured HTO concentration and an assumed moisture content of 4 L/m2. The Myers report states that "a) no other locations reaching this level of contamination were found and b) it is unlikely that any such locations were missed because of the extensiveness of the survey." Our dose estimates are based on milk measurements from the real dairy cows present in the plume area. Statements outlining these differences in approach and the resulting doses have been added to the text.

54. Page 22, paragraph 3: Replace the third sentence with "Plants are a primary conduit for transferring tritium from soil to air, but the air concentration includes the tritium that is actually in the plants."

ATSDR Response: The sentence was changed as follows: Transpiration from plants and evaporation from soil are the primary conduits for transferring tritium from soil to air.

55. Page 22, paragraph 4, lines 9-15: No credit can be claimed for dilution of plant concentrations by uncontaminated water when ingestion doses are calculated from observed plant concentrations. But this argument can be used to maintain (legitimately) that modeled ingestion doses are too high.

ATSDR Response: This is true if the vegetation measurements had included garden or food crops. The Myers report indicates that "The vegetation samples were mostly varieties which remain green during the dry season". Our statement indicates that garden crops are unlikely to remain green during the dry season. No changes to this sentence have been made.

56. Page 29, paragraph 2, lines 7-8: The sentence beginning "Due to increased dispersion conditions " needs to be changed to reflect whatever resolution is reached regarding the relative concentrations of the two releases.

ATSDR Response: The phrase "increased dispersion" has been changed to "meteorological conditions".

57. Page 34, paragraph 3, line 7: Delete "decreasing atmospheric dispersion and", since dispersion is independent of release height (in the model at least). Also, the effect of release height for these stabilities would extend only a few hundred meters downwind, not the 3 miles claimed.

ATSDR Response: This section has been rewritten. Because of newly available information on stack parameters, plume rise has been included in the revised RASCAL calculations. However, total dispersion of a plume is highly dependent on release height, which significantly effects the distribution of ground level contaminant concentrations.

58. Page 34, paragraph 4, line 6: Replace "Q (Ci-sec)" with "Q (Ci)". Also, in the last line, replace Brown with Ogram.

ATSDR Response: These changes have been made.

59. Page 36, Table A-4: The values of wind speed, direction, stability class etc. are missing from the bottom of the table.

ATSDR Response: The revised table includes the complete input summary of the RASCAL evaluation.

60. Page 38, paragraph 2, line 3: A southwest wind was assumed for the 1965 release, so the rectangles should cover directions 10º to 50º (as indicated on page 15) rather than 180º to 220º.

ATSDR Response: Wind directions for the 1965 release have been revised on the basis of newly available data. This paragraph has been changed accordingly.

61. Page 38, paragraph 2: Replace lines 6-10 with "HTO emission rates for each rectangular area were derived from the time-integrated HT concentrations in air, the HT deposition velocity and the HTO loss rate from soil, assumed to be 1%/h (Ogram et al. 1988)."

ATSDR Response: This paragraph has been changed as suggested.

62. Page 40, the equation for calculating dose: I'm not an expert in dosimetry so the following comments may not be relevant. But if I can't understand this equation, others may not be able to either. The first term in the equation is defined as a concentration but has units of Bq rather than Bq/m3 or Bq/L. Should this be total intake rather than concentration? The units of the tritium decay energy should be MeV/disintegration to make the equation dimensionally correct. The parameters in the equation other than concentration appear to act like a dose coefficient, but when multiplied out have a value of 8.76 x 10-12 Sv/Bq, a factor 2 lower than the commonly accepted value of 1.8 x 10-11. Has the dose coefficient been revised down recently? Finally, it's not clear how to apply the equation for HT.

ATSDR Response: The equation referenced shows the amount of energy deposited in a defined mass from a defined amount of tritium. The equation is used for the total intake of tritium regardless of the concentration. If the concentration is known, extra coefficients are used to estimate the total intake. These could include liters consumed per day, amount of air inspired during a day based on activity levels, or the amount of food consumed. As for the math, we used the following values and approximated the ICRP value of 1.8E-11 Sv/Bq:

27 picocuries = 1 Bq
1.6 E-13 J/Mev
tritium decay energy (average) = 6
seconds per year = 31.5 E 6

multiplying these numbers out gives 8.2e-10 J/Bq in one year

a Sievert is defined as 1 J/kg and a standard "man" mass is assumed to be 70 kg.

If you divide 8.2 e-10 by 70, then one obtains 1.8E-11 Sv/Bq.

63. Page 41, paragraph 2: Replace the last 3 sentences with "The declining air concentrations are due to the loss of HTO from the soil, which is the source of activity for the air".

ATSDR Response: This paragraph has been changed as suggested.

64. Page 41, paragraph 3, line 2: Replace "dispersion from soil" with "the loss rate from soil".

ATSDR Response: This paragraph has been changed as suggested.

65. Page 41, paragraph 6: It's likely that the uncertainties in environmental HT and HTO concentrations and in the assumptions about the ingestion pathways are much larger than the uncertainties in the dosimetric parameters and would dominate the uncertainty in the dose if they were taken into account. This should be pointed out here.

ATSDR Response: The revised sensitivities show that, as suggested, the environmental parameters account for most of the variance in the dose estimates. This paragraph has been changed accordingly.

66. Page 43, bottom distribution: This should be the "initial soil HTO concentration", not the "initial soil HT concentration".

ATSDR Response: This forecast label has been changed as suggested.

67. Page 44: Add units to the distribution at the top of the page. Also, the description of the lower distribution sounds very much like that of the lower distribution on page 42, but the numerical values are different. What is the difference between these two distributions?

ATSDR Response: This forecast label has been changed as suggested. The cumulative HT concentration forecasts are redundant, however, the "cumulative air act (HT)" forecast is not used in the calculations and has been removed from the report.

68. Page 45, upper distribution: The mean value of the distribution is given as 0.0238 mrem but the average 30-min HT dose for a child in Table 2 is 0.05 mrem. What is the reason for the difference?

ATSDR Response: The Crystal Ball simulations were run iteratively for adult and child intake rates and body weights, and for concentrations at the 1 and 2 mile locations. The page 45 forecast is for an adult exposure. The revised Crystal Ball report has been labeled to clearly identify the specific parameters it is simulating.

69. Pages 46-49: I accept the distributions for the dosimetric parameters since those were set by a panel of experts (but units should be added to the distribution for tritium decay energy.

ATSDR Response: The decay energy units of MeV have been added to the figure.

70. Cumulative HT air concentration/Q: The shape and standard deviation of the distribution must be justified. I believe the standard deviation is much too small. The method of expert elicitation was recently used to estimate as rigorously as possible the uncertainty in the predictions of Gaussian dispersion models (NUREG 1994). Results of this study indicate that the ratio of the 95th to 5th percentiles of the centreline concentration is about a factor of 10 for downwind distances of 500 to 1000 m. A large uncertainty would be expected in the present case from the uncertainty in stability class alone. For the 1970 release, estimates of stability ranged from B to F, which can generate differences of more than a factor of 10 depending on the downwind distance. As a secondary point, the units should be indicated for the distribution.

ATSDR Response: As the revised calculations are based on observed weather conditions, we feel there is considerable justification for use of the E (or F) and B (or C) atmospheric stabilities. In both cases we have used the stabilities that lead to the highest estimated concentrations in the areas of potential maximum exposure. The shape (distribution) of the HT concentrations must be assumed as normal because such normality is the underlying basis for the Gaussian dispersion (concentration distributions are normally distributed within the plume). While the standard deviation of the assumed HT distribution may be small, because it is a normal distribution, the simulation results are based principally on the mean value.

We have compared the results of the RASCAL dispersion with another Gaussian model (HOTSPOT) and found that while the results are similar, RASCAL results are about 20% higher than HOTSPOT results. A validation study of the RASCAL model also found that predicted concentrations over-estimated measured concentrations (Ramsdell, personal communication). Consequently, we are confident that the RASCAL results do not under-predict actual concentrations and that the actual doses will be less than the predicted doses. Unit labels will be added to the Crystal Ball report as suggested.

71. HTO air concentration: The shape and standard deviation of the distribution must be justified. I believe the standard deviation is much too small. The HTO concentrations reflect not only the uncertainty in the HT air concentrations but also the uncertainties in the deposition velocity, the re-emission rate and the dispersion model in ISCST, all of which will be large.

ATSDR Response: The derivation of the HTO concentration distribution has been changed. As you suggest, the revised HTO concentration distribution is based on cumulative HT concentrations (normal distribution), HT deposition velocity (assumed normal distribution), and the HTO re-emission rate (exponential distribution; the HTO re-emission rate has been termed the HTO loss rate in order to differentiate it from the HTO soil emission rate used in the ISC model). Multiplication of these factors produces an apparently lognormally distributed HTO soil emission rate (Ci/sec-m2). The mean (50th%) and 95th% values of that emission rate distribution were used in separate ISC model runs to produce mean and 95th% HTO concentrations. The resulting mean and 95th% concentrations were then used to describe a lognormal HTO concentration distribution which was used to estimate doses. This method has been described in the revised document.

72. HT deposition rate: The literature contains many values of the deposition velocity apart from those reported by Ogram et al. (1988). These should be considered in setting the mean and the standard deviation. Also, I believe the distribution should be lognormal rather than normal.

ATSDR Response: As suggested, we have broadened the references on which the HT deposition velocity is based. However, these additional references, including Sweet and Murphy (1981), Dunstall et al. (1985), Spencer and Vereecken-Sheehan (1994), and a review by Brown et al. (1996) confirm that our assumed distribution of rates is consistent with a variety of measured or modeled rates. Additionally, there does not appear to be any a priori basis for assuming a non-normal distribution of those rates.

73. Vegetation activity: More detail should be given on the number, location and timing of the measurements that form the basis for this distribution, so the reader can judge whether the standard deviation is reasonable or not. I believe the distribution for this parameter should be lognormal rather than normal.

ATSDR Response: To the extent possible, based on the locational descriptions of the samples, we have used those sample results most likely to represent the plume footprint. As previously stated, because of growing conditions at the time of the 1970 release, we believe these measured samples are a reasonable indicator of plant tritium concentrations. More complete information on sample locations and times are not available. If we had chosen to include the many lower concentration and non-detected results, the resulting distribution would appear lognormal. However, as the HT plume must be considered normally distributed, it is reasonable to assume that plant moisture concentrations in the plume footprint are also normally distributed.

74. Milk concentration: A distribution based on just 3 points has no credibility. Since the cows are eating the local vegetation, the distribution for milk should be at least as wide as the distribution for vegetation and should be lognormal rather than normal.

ATSDR Response: There were many more milk analyses. There were, however, only three detectable tritium measurements. As previously stated, these results are based on the cows from dairy farms in this area. As cows would not be consuming pasture grasses during this time period, inclusion of the dairy component is conservative.

75. Adult and child body mass: I have no expertise here but I would have thought body mass would be better described by a lognormal than a normal distribution. Also, the units should be given for these distributions.

ATSDR Response: The revised simulations and dose estimations assume lognormal distributions of body masses and breathing rates.

76. Ingestion Dose Estimates from a Simple Model

Data from Spencer et al. (1988) collected during the short-term HT release experiment at Chalk River in 1987 show that, after a lag time of a few hours, HTO concentrations in vegetation were consistently a factor of about 3 lower than concentrations in the 0 - 2 cm soil layer. This provides a means of predicting concentrations in vegetation, from which ingestion doses can be calculated.

The initial HTO content of the soil is 2.5 x 10-3 Ci/m2 = 9.3 x 107 Bq/m2 (Figure 4 in the assessment). Following an HT release, HTO concentrations in soil drop off exponentially with depth with a scale length of 2.3 cm (Taeschner et al. 1995), so the HTO content of the top 2 cm of soil is 58% of the total, or 5.4 x 107 Bq/m2. Assuming a water content in midsummer of 10% by volume, the total amount of water in the top 2 cm is 2 L/m2. The initial HTO concentration in soil water is therefore 2.7 x 107 Bq/L. Assuming a loss rate of 1%/h, the integrated concentration in soil water over 12 days is 1.1 x 108 Bq d/L. The time-integrated HTO concentration in plant water is a factor 3 lower than this, or 3.5 x 107 Bq d/L. With an intake rate of 0.3 L/d vegetation water, the total tritium activity taken in through ingestion of fruit and vegetables is 1.1 x 107 Bq. Using a dose coefficient of 1.8 x 10-11 Sv/Bq, the ingestion dose becomes 2.0 x 10-4 Sv or 20 mrem. This is substantially larger than the dose estimated from the monitoring data, and comparable to the inhalation dose. It's unlikely that these doses would have been achieved in reality but they were obtained with a model similar to that used for the inhalation dose.

ATSDR Response: We agree that the above calculated doses would be unlikely. Further, we believe that basing the ingestion dose estimates on measured milk and vegetation tritium concentrations provides a more realistic assessment of potential doses. Considering that human biological measurements failed to detect tritium exposure from the 1970 release, it is more important to communicate the conservatism of the inhalation doses rather than to equally over-estimate the ingestion doses.


Reviewer 2 Comments

77. Determining the affected population by estimating the total (55 and 52 persons) and then using census data to estimate numbers of women, children, elderly, etc. doesn't seem reasonable given the small numbers of people that were exposed. Maybe, since the assessment only differentiates child and adult doses, these statistics aren't necessary?

ATSDR Response: Part of ATSDR's task in conducting a public health assessment is to determine how many people may be exposed to the site-related environmental contamination and if there may be some portion of that population that is particularly susceptible to those contaminants. As there are no population data available for the limited areas of these air plumes, the identification of buildings and use of average per household occupancy is a reasonable estimate of the maximum exposed population within the plume areas.

78. There is no mention as to why HT is the only form of tritium released from the stack at LLNL. A paragraph describing why this is the case would be informative.

ATSDR Response: All of the available accident reports and descriptions state that the releases were comprised of HT. As this was the form supposedly contained within the vessels that leaked, there is no basis for assuming that any other form of tritium comprised a significant proportion of the releases. The above sentences have been added to the document.

79. The assessment does not include the person(s) that lived in the area during both releases.

ATSDR Response: The doses are summed to provide annual dose estimates and compared with annual dose limits. As the biological half life of tritium is less than one year, there is no long term dose contribution and there would be no cumulative effect from the different releases. Additionally, the revised document includes newly available weather data that indicates that the two plumes did not affect the same area so it is unlikely that any individual was exposed to both plumes (via a residential exposure scenario).

80. Page 15 (end of 3rd full paragraph) - it is stated that the air concentrations from the two releases are modeled as equal, but this finding isn't stated anywhere else in the document, except in that sentence (it sounds as if the reader should already have known this to be the case).

ATSDR Response: The revised dispersion calculations indicate that doses for the 1970 release were likely to have been greater than those from the 1965 release. The document has been revised to indicate that all dose estimates and evaluation of potential adverse health effects are based on the larger 1970 dose estimates.

81. Page 16 - mention is made of breathing rate and how it was estimated, but there is no differentiation between adult and child breathing rates. This needs explanation.

ATSDR Response: The phrase "adult and child breathing rates" has been added to the text. The assumptions concerning child breathing rates have also been added to the Crystal Ball report in Appendix 4.

82. Page 18 - why didn't the author's provide uncertainty for the fruit consumption variables? (several other variables are not included in the uncertainty analysis).

ATSDR Response: The food ingestion rates (fruit, milk, and vegetables) are based on recommended values from the EPA Exposure Factors Handbook (EPA 2002). We have re-run the Crystal Ball simulations using these values as the mean of a normal distribution and there is no difference in the ingestion dose. Consequently we have not changed the point value ingestion rates into probability distributions.

83. Page 19 - what data supports the use of normal distributions for tritium concentrations in vegetables and milk? Environmental concentrations are most often distributed lognormally.

ATSDR Response: The distributions of tritium concentrations in vegetation and milk are based on use of samples from the plume footprint using the lowest measured values as the 10th percentile and the highest value as a 90th percentile of a normal distribution. A normal distribution was used because there is no a priori reason for assuming that the mean will be skewed between these upper and lower values. Additionally, the initial dispersion of the HT plume is based on a Gaussian (or normal) model which assumes that concentrations in the plume are normally distributed. Consequently, HT distribution is assumed normal. Also, although some environmental and biometric parameters are lognormally distributed, many apparent lognormal distributions are an artifact of inadequate sampling.

84. Page 24 (bottom) - children may have increased breathing rates, but their volumes are probably smaller - what is the minute volume of adults vs children?

ATSDR Response: The minute volumes of adults vs. children are 0.011 m3/minute and 0.008 m3/minute, or 15.3 m3/day and 11.4 m3/day, respectively. The change in relative doses is based on the ratio of the breathing rates and body weights.

85. A normal distribution is used for HT deposition rate, but typically uniform or loguniform distributions are used for these type variables. The use of normal is highly biasing, unless there are data supporting such a distribution; the supporting data should be presented or referenced.

ATSDR Response: Additional references to measured or modeled HT deposition velocities have been added to this section. The distribution we have assumed uses the range of rates included in these references. The use of a uniform distribution requires knowledge of the upper and lower boundary values. Deposition velocities have been found to vary with soil type, soil moisture, vegetation cover, and season. Considering the range of parameters that affect deposition rates, it is unlikely that the measured velocities capture the entire range of variation. Likewise, when integrating the range of velocities over a plume footprint, is seems reasonable to assume that the velocities will approach an average value. The normal distribution assumes that values higher than the mean are just as likely as values lower than the mean.

86. Page 12 (end of 3rd full paragraph) - this is a repeat of what appeared on the previous page.

ATSDR Response: This section has been re-written.

87. Page 12 (4th full paragraph, next to last sentence) - what is "a uniform distribution with a value of 1%/hr"?

ATSDR Response: This uniform distribution is essentially a single point value. On the basis of additional empirical values (as referenced in revised document) the HTO emission rate has been changed to an exponential distribution ranging from 0 to 8%/hour with a mean value of 1%/ hour.

88. 1% per hour does not translate into 24% per day - the 1% refers to the loss rate of an amount of material present at a given time - that value shouldn't be multiplied by 24 hours in a day to get %/day. Carrying this same logic to a weekly value gives us 168%/wk?

ATSDR Response: You are correct in that a 1%/ hour HTO loss rate is not equal to a loss rate of 24% /day. The percentage loss depends on the magnitude of the time step. Using an hourly time step, the percentage loss for the first day would be ~21%. As we are calculating daily doses, we made the simplifying assumption that 1%/ hour is equal to 24%/day for the first day (the percentage loss for the second day is ~18% of initial total and so on). The revised document includes a figure showing the percentage loss over 12 days (>95% of initial total) and an explanation of this assumption. As explained in the document, the absolute magnitude of the HTO loss rate does not have a significant effect on the total dose. The use of an hourly time step would increase the number of days required to deplete the HTO soil source by 95%, but would also cause a commensurate decrease in the HTO air concentration such that the total dose would be unchanged.

89. It is stated several times that the "most probable" doses are determined by a Monte Carlo simulation technique. "Most probable" doesn't mean average of the distribution, but it occurs where the distribution peaks. Also, the Monte Carlo is done such that the authors are estimating uncertainty (in a limited way) of the maximum exposure (and they go on to say that the doses are probably greater than anyone actually received). So, these are not "most probable" doses. Likewise, when 95th percentiles (and distributions) are given, those are the 95th percentile estimates of the maximum dose, not 95th percentile estimates of the expected dose. This could be quite confusing to the public.

ATSDR Response: The revised document uses the term "average (or mean) dose" to describe the estimated doses. We have further included a chart of the probability distribution of the estimated doses to show that the mean doses over-estimate the "most probable" doses.

90. Page 21 (top) - only the laws applicable to LLNL should be stated regarding dose limits.

ATSDR Response: The Livermore community is very knowledgeable with regard to regulations and laws pertaining to radiological releases and dose limits. In this case, it is necessary to explain why specific laws are not applicable, as well as those that are applicable.

91. Biological half-life of tritium varies between 1 and 40, with a triangular distribution? This parameter is expected to be lognormally distributed over a population.

ATSDR Response: A triangular distribution was used for the Monte Carlo simulations for the biological half-life of tritium. The conditions for our usage of this type of distribution are as follows. Limited data exist on the biological half-life; however, we have a reported minimum and a reported maximum value. Furthermore, the most likely value is somewhere between these values. The software we used to run the simulations, recommends this type of distribution for these limited data. A lognormal distribution is typically used when the data are skewed toward one end of the distribution. In this case, the assumption of a most likely value of 10 days skews the triangular distribution in a similar manner as a lognormal distribution.

92. Why is the dose and dose rate reduction factor used here? This parameter is normally used to convert high-dose-rate risk to low-dose-rate risk. Is that the intent here?

ATSDR Response: The dose and dose rate effectiveness factor (DDREF) is used to extrapolate from high doses (derived from A-bomb studies) to low dose and dose rates. Typically, the DDREF is applied when the doses are less than 20 rads as recommended by the UNSCEAR 1993 report.

93. Table 3 - "mean" and "95th" doses are reported - but these are the mean and 95th percentile of the maximally exposed person (for which the author's say is an overestimate of dose) does "mean" and "95th" really describe what is intended? The caption states, "The potential average tritium doses (to the maximally exposed individuals)" - seems odd to be talking this way. Is the assessment meant to determine the actual mean or most probable dose? or the uncertainty of estimating a mean dose?

ATSDR Response: Several sentences have been added to the introduction to clarify that this evaluation focuses on estimated exposures and the potential for those exposures to produce adverse health effects to those individuals with the highest potential doses. References to the mean and 95th percentile doses have been clarified by indicating that those parameters refer the probability distribution of the estimated doses and that the most frequent doses from the Monte Carlo simulations are less than the mean values.

94. Page 21/22 - how soon after the 1970 release did Meyers take his measurements of urine in exposed individuals? Timing might be the reason for not seeing tritium in the samples.

ATSDR Response: Most urine samples were taken from off-site individuals on the day following the release (one each were taken 2,3, and 4 days following release). About one half of the on-site employee samples were collected on both the day of and the day following the release. Others were collected up to six days following the release. We agree that the timing of the sampling could reduce the potential for detecting any tritium exposures. However, based on our estimated doses, even exposure for 1 day should produce a dose well above the stated detection limit (5,000 pCi/L; or a 0.025 mrem dose).

95. Page 17 (top line) - states that the 95th percentile dose is 0.14 mrem, but then the corresponding dose (or so it seems) in Table 2 shows 0.2 mrem.

ATSDR Response: This was probably due to multiple runs of the Crystal Ball simulations and rounding error in the results. The revised tables and text have been checked and corrected for consistency.

96. Page 19 (bottom) - again, the "most probable" of the distribuiton is NOT the arithmetic average of those distributions - the most probable or "most likely" is at the peak.

ATSDR Response: References to the mean and 95th percentile doses have been clarified by indicating that those parameters refer the probability distribution of the estimated doses and that the most frequent doses are less than the mean values.

97. Page 40 - it is stated that in order to adequately represent the range of all parameters, thousands of iterations are needed that's not true Latin hypercube methods solve this problem and have been shown to require only a hundred or so samplings to adequately represent input distributions and describe an output distribution.

ATSDR Response: The phrase "in a Monte Carlo analysis" has been added to that statement.

98. There are errors in the calculation of dose, if the DDREF variable was used. Also, the "weighting factor" is mandated by 10 CFR 20 as unity (1) for beta emitters. That parameter probably shouldn't be varied in an assessment such as this.

ATSDR Response: See response to comment 16 regarding DDREF. The radiation weighting factor of 1 is recommended by the Nuclear Regulatory Commission as well as the ICRP and the NCRP. The expert panel convened by ATSDR recommended a higher value for tritium and that is the value used by ATSDR. As referenced in the Expert Panel Report (ATSDR 2002), there are many references in the literature reporting a tritium radiation weighting factor in excess of 1.0.

99. The technical form of the document is still quite high in places.

ATSDR Response: We agree that this document presents information in a technical form. However, the Livermore community is quite sophisticated with regard to their concerns about radiological exposures and has requested that we do not omit information from our analyses. In addition to distributing the PHA documents, we have made several presentations of this information to the community and have made every effort to address and answer any questions the community may have about this material.

100. Page 13 (1st paragraph, next-to-last sentence) - the phrase "from weather years 1991 and 1993" would seem to belong in parentheses.

ATSDR Response: This section (and sentence) have been re-written in response to newly available weather information.

101. Page 14 (2nd sentence) - according to other statements in the document, the time of the accident was 6:14 am, not 6:30 am.

ATSDR Response: Different reports and news articles have reported slightly different times for the release. Although the specific time of the release may seem unimportant, dispersion of the HT plume is highly dependent on the atmospheric stability classification, which in turn, is dependent on whether the release occurred before or after sunrise. A newly available accident report unequivocally states that the release occurred at 5:49 AM. All references to the time of release have been so corrected.

102. Page 15 (top line) - what is a "flow vector"? It's meaning probably should be explained.

ATSDR Response: This section has been re-written in accordance with newly available 1965 weather data.

103. Page 20 (bottom) - "360 mrem" should be "360 mrem/yr".

ATSDR Response: This change has been made.

104. Page 40 (equation) - the tritium concentration parameter is shown to have units of Bq, but the dimensions of concentration should be something like Bq/L or Bq/m3.

ATSDR Response: See the response to comment 51 (page 54).

105. Page 40 - is the weighting factor "wt" the same as a radiation weight factor (quality factor) for beta emitters? That should be made more clear.

ATSDR Response: Yes, the term "radiation weight factor" has been inserted in the bulleted description. The assumption label is an abbreviation.

106. Page 40 - again, the DDREF is used incorrectly. It should not be in this assessment.

ATSDR Response: See the response to comment 16 (page 62).

107. Page 45 (sensitivity chart) - what parameters do cells D18 and K14 represent?

ATSDR Response: Cell D18 represents an inhalation dose from a light work breathing rate and K14 represents the strenuous work breathing rate. The Crystal Ball report and sensitivity chart have been updated and all of the significant assumptions and forecasts labeled.

108. Many of the Crystal Ball assumptions are overly biased by making them normal when there may not be sufficient data to support such assumptions.

ATSDR Response: As previously stated, the assumption of any probability distribution presents some bias into the evaluation. However, an assumption of normality is the least biased of the probability distributions. For example, a lognormal distribution assumes that most occurrences of a variable will be significantly lower than the mean value. Conversely, a normal distribution assumes those values are equally likely to be lower or higher than the mean. It is our belief that most parameter distributions should be assumed normal, unless there is adequate information to support a non-normal distribution. There is sufficient data to indicate that body masses and breathing rates should be lognormally distributed. These parameters and the associated text and figures have been so changed.

109. Page 47 - vegetable and milk concentrations given in different units. This is confusing.

ATSDR Response: Milk and vegetable concentrations units have been revised to similar units (Ci/L).

110. Page 48 - the inhalation rate distributions are given for adults or children? Or both?

ATSDR Response: The revised Crystal Ball report (Appendix 4) contains both child and adult breathing rates.

111. Page 49 - normal distributions for body mass? It's been shown that lognormal is a better distribution to use for body mass (and other physiological and environmental parameters).

ATSDR Response: The revised dose calculations use lognormal distributions for body weights and breathing rates.

112. In several places it is stated that there is "no public health concern" regarding tritium doses. But, hasn't concern been expressed by the public? Maybe the wording should be altered to state that "ATSDR has no public health concern". Also, the second bullet on page 27 and at the end of the first paragraph on page 28, the author's state that "incidence was elevated" but it was "not statistically significant"? It would seem more reasonable to conclude simply that "there was no statistical increase in incidence", and leave it at that.

ATSDR Response: The phrase "no public health concern" has been replaced with the phrase "below levels of public health concern". Further, "levels of public health concern" are clearly defined by addition of a subsection on "Tritium Doses of Public Health Concern" to the Public Health Implications Section. This subsection clearly identifies the health comparison values and standards that ATSR uses in making its health determinations. The sentences describing the expected or elevated incidences of disease frequency accurately convey the information that those rates were higher than expected, but that the increase was not significant. Some members of the community might find it misleading if we did not state that there was a higher than expected incidence rate.


Reviewer 3 Comments

113. The public health assessment provides a complete description of the potential pathways of human exposure. The figures are helpful in presenting these pathways.

An additional statement on the significance, or lack of significance, of the HTO inhalation pathway would be helpful. It is stated (page 12) that it has been found that "there is little direct atmospheric conversion of HT to HTO (compared with the conversion rate from microbes)". This is a comparison of atmospheric conversion of HT with the conversion of HT deposited onto the soil, and only a fraction of the atmospheric HT is deposited onto the soil. Thus, this comparison does not provide complete information on the size of the HTO inhalation dose relative to the doses from other pathways considered. It would be helpful to add a statement on the DOSE from HTO inhalation relative to the DOSES from other pathways considered.

ATSDR Response: The RASCAL model estimates a maximum dose of 103 mrem (at 1 mile) based on the assumption that 100% of the tritium plume was present as HTO. Noguchi et al. (1989) and Noguchi (1995) conclude that the conversion of HT to HTO in air will be less than 0.5% over the time period of a few hours (according to the RASCAL model the HT plume has completely dispersed in 30 minutes). Using the RASCAL dose estimates, if 1% of the HT plume was present in an HTO form, it would add about 1 mrem to the total tritium dose at the point of maximum exposure. A paragraph indicating this source of uncertainty will be added to the section on "Total Tritium Doses and Uncertainties in Dose Modeling Procedures."

114. The pathways and models for tritium fate and transport are appropriate.

115. The RASCAL computer model developed by the Nuclear Regulatory Commission and the ISC model developed for the Environmental Protection Agency are very appropriate for the exposure analyses. Also, considering uncertainties in model parameters, it was appropriate to compute probability distributions, and the Crystal Ball model used is a well-established model.

116. The literature on the relationship between human exposure to ionizing radiation (all sources) and the potential health effects associated with that exposure is vast. The public health assessment presents data from only a small fraction of these references, but the information presented is adequate to allow conclusions and recommendations to be formulated. Including material from the report of the United Nations Scientific Committee on the Effects of Atomic Radiation (UNSCEAR) was particularly appropriate since the work reported was based on a review of many studies. Similar reviews of many studies have been conducted by the National Academy of Sciences' Committee on the Biological Effects of Ionizing Radiation (BEIR). The report of this work would also be an appropriate reference.

117. The public health assessment accurately and clearly communicates the absence of a health threat posed by the accidental tritium releases from this site. I think that the Summary of the report is excellent and very clearly presents the results of this assessment.

118. The conclusions and recommendations are appropriate in view of the potential doses as described in the public health assessment.

Many conservative assumptions were made in this assessment, which will result in an overestimation of the calculated doses. Even with these conservative assumptions, the calculated doses are small. Furthermore, it is noted (page 21) that following the 1970 accident, analyses were made of urine samples from potentially exposed people, including both LLNL workers and potentially affected community workers. The detection limit for these analyses corresponded to a very low dose of 0.025 mrem, and no elevated tritium body burdens were detected. This is quite significant, and it is appropriately noted in the report that these data indicate that the modeled exposures over-estimate the actual human exposures.

The modeled exposures and the actual data collected following the 1970 accident provide strong support for the conclusion that the accidental HT releases are "no apparent public health hazard."

ATSDR Response: The above comments are noted, no responses are necessary.

119. In the first paragraph on page 10 it is stated that the hypothetical maximally exposed individual for the 1965 accident is assumed to be 0.5 miles from the LLNL tritium facility. In the Summary it is stated that this distance is about 1 mile. Also, in Appendix A it is stated those distances less than 0.7 miles are on site. Thus, the stated distance on page 10 needs to be clarified.

ATSDR Response: The revised dispersion calculations and dose estimates use newly available weather data for the 1965 release. Dispersion calculations for both the 1965 and 1970 releases indicate that maximum HT and HTO air concentrations occurred at a distance of 1 mile from the release source. All references to the locations of maximum exposure have been revised.

120. In the fourth paragraph on page 11 it is stated that soil HT concentrations are represented by probability distributions. However, in Table 1 on page 15 soil HTO emission rates are presented as point values. The deposition rate used in the computation of these values of the soil HTO emission rates should be explained.

ATSDR Response: The ISC model can only use point values for HTO emission rates. The revised calculations accommodate this limitation by running the ISC model iteratively using the mean and 95th% HTO emission rates. The resulting mean and 95th% HTO air concentrations are used to develop a lognormal HTO air concentration probability distribution which is used in subsequent dose calculations. Descriptions of this procedure (and the HT deposition velocity assumptions) have been added to the document.

121. In the second paragraph on page 15 there is reference to computing soil HTO emission rates using "the cumulative soil HT loading values from Table A-2". I think that this should refer to the "cumulative HT air concentrations" rather than to the "cumulative soil HT loading values". This issue occurs in other places within the report also, including the caption of Table 1 on page 15.

ATSDR Response: These suggested changes have been made in the revised document.

122. Many conservative assumptions are made in this assessment, which will result in an overestimation of the calculated doses. This fact is appropriately noted in the report. Assumptions in the ingestion exposure pathway are especially conservative. These include the assumption that all fruits and vegetables eaten were grown at the point of maximum exposure. The probability distribution for tritium in vegetation, which was based on actual measurements along the centerline of the 1970 plume, was censored so that no values less than the detection limit were used. I am not aware of a technical justification for censoring the data. Tritium concentrations in milk were also based on actual measurements. Only three out of thirteen milk samples had detectable tritium concentrations, and only these three measurements were used for the probability distribution. This method of analysis produces a positive bias. Even with these overly conservative assumptions, the doses from the ingestion pathway are negligible. Thus, these assumptions did not effect the conclusions in this assessment. It is recommended that this degree of conservatism not be a precedent for all assessments.

ATSDR Response: The revised vegetation tritium concentration parameter is not censored at the detection limit. Also, the section on "Total Tritium Doses and Uncertainties in Dose Modeling Procedures" includes a paragraph on uncertainties associated with the ingestion dose component.

123. In the fourth paragraph on page 23, one sentence states that "the dose-response relationship appears to have a threshold at about 25 rads" and another sentence in this paragraph states that " these data suggest a linear dose-response curve without a threshold". The same reference is given for both statements. Are these different conclusions from different aspects of the same study? It would be helpful to add some statement clarifying these statements or just acknowledging the differences.

ATSDR Response: Although both health effects relate to central nervous system development, the referenced study concludes that the collective data may support either a threshold or no threshold dose response. More important, is that, overall, these health effects occurred at doses that are 500 to a 1000 times higher than the doses estimated from the LLNL tritium releases. A statement clarifying the uncertainty of the dose response has been added.

124. Table A-4 appears to be incomplete - at least, it does not contain all of the data presented in Table A-5.

ATSDR Response: Table A-4 (re-numbered as A-1) has been modified to include all of the output contained in Table A-5 (A-2).

125. In Appendix 3 on page 40 an equation used in the Monte Carlo simulation is presented. The equation contains a term for the "tritium concentration in curies". The location of the tritium concentration should be stated. Is it in the body? Also, concentrations have units of activity per unit volume or mass, which does not seem to work in this equation. Is the term actually an "activity" rather than a "concentration"? This should be clarified.

ATSDR Response: The term "tritium concentration" refers to the total amount of tritium in the body. The equation as it is expressed is calculating the amount of energy deposited inside the body. This is shown in the equation by the "wt factor/body mass" term. In the equation, the tritium concentration is dependent on the total amount of tritium in the body and the water content of the body to yield the concentration in picocuries/liter. However, we are calculating to dose to the entire body (hence the body mass term) so only the total amount of tritium is important. For the purpose of calculating the radiation dose to the entire body, the total amount of tritium in the body is used. Typically, this equation is used after one has calculated the intake of a substance using concentration factors, intake amounts, and time frame of intake to give the total intake.

126. The questions asked of peer reviewers are appropriate ones, and the opportunity is provided to submit additional comments, which is good.

18 Note that all references for the appendices are listed in the preceding References section.

Next Section     Table of Contents


Error processing SSI file