ML19208A430
| ML19208A430 | |
| Person / Time | |
|---|---|
| Issue date: | 08/14/1979 |
| From: | Hendrie J NRC COMMISSION (OCM) |
| To: | Nordhaus W YALE UNIV., NEW HAVEN, CT |
| Shared Package | |
| ML19208A431 | List: |
| References | |
| NUDOCS 7909130571 | |
| Download: ML19208A430 (7) | |
Text
0A y%
g&
UNITED STATES
-l 9
NUCLE AR REGULATOR Y CO'.U.11SS!ON Y 2[c J W;h[
W /d m N G T ON. D.C. N255 v [V)
% +....f
.u August 14, 1979 g{
omcc or Tur cn un wa Professor William D. Nordhaus John Musser Professor of Economics Department of Economics Yale University Box 2125, Yale Station New Haven, Connecticut 06520
Dear Professor yo'rdhaus:
[f Thank you for sending me a copy of your thoughtful analysis; I quite agree with your main point.
It is now common knowledge that the Reactor Safety Study (RSS) has broader uncertainties than stated and that its overall risk predictions should not be used in public policy decisions.
This was the message of the Lewis Conmittee, the Conmission policy state-ment, and is fully consistent with many statements by the RSS authors.
The essence of your macroanalysis rests on the comparison between the HEW estimate of the casualties to be expected from Three Mile Island (TMI) and the RSS predictions.
This comparison really suffers from the apples and oranges problem.
The HEW estimate should be expected to exceed an RSS prediction by a large factor because it was made on a conservative oasis while the RSS predictions attempted to be more realistic.
The magnitude of the difference is difficult to estimate precisely, but it should be over a factor of 10.
This difference in values lies in both the dose-response relationship and the estimated population dose. The dose response model employed in the HEW calculation predicts roughly ten times more latent cancer fatalities for a given low level population dose than does the model used in the RSS.
Also, the estimate of the population dose employed in the HEW calculation contains a number of deliberately conservative assumptions.
Although the NRC was involved with HEU and others in the preparation of that early estimate, our staff is currently reexamining its accuracy.
It appcars, as of now, that the actual population dose could be smaller but we have not yet done enough work to estimate the reduction.
The differences between the HEW and RSS estimates are described in more detail in th9 attachment hereto.
SM50N 7909130h
Professor William D. Nordhaus 2
August 14, 1979 There are additional insights, associated with the RSS microanalysis that could affect your macroanalysis.
Accidents that stop short of full core melt were given very cursory treatment in the RSS.
Such accidents would affect the low end of the predicted consequence curves and, consequently, were not given full treatment.
The major interest of the study was in accidents that could have high public consequences of the type that could only be associated with large amounts of melted fuel which would cause loss of containment integrity.
Thus comparisons of the probabilities of small consequence accidents with the RSS predictions would necessarily have large uncertainties.
We have also found that the extrapolation from the two reference plants analyzed in the RSS to 100 plants was more presumptuous than we previously thought.
Differences among plants can result in differences in predicted risks, and we are finding, with continued studies, that there are differences that can have significant risk implications.
In particular, the TMI plant is not well represented by the reference pressurized water reactor (PWR) studied in the RSS since it does not fit the ensemble of RSS accident sequences.
The basic difference is that, in TMI, a loss of the main feedwater system surely caused the pressurizer relief valve to open in a few seconds, independently of the operation of the auxiliary feedwater system.
For the reference PWR in the RSS, and for most PWRs, the loss of both main and auxiliary feedwater systems are required for about thirty minutes before the relief valve would lift.
I am happy to be able to say that the plant changes we have recently required the utilities to implement have corrected this matter on all operating plants similar to TMI so that they are now a much closer fit to the RSS.
Th9 main point to be made, however, is that the question raised about the validity of the RSS results by your macroanalysis are qualitatively confirmed by RSS microanalysis of the TMI reactor.
However, I don't believe that we can confirm your quantitative analysis.
If we, for instance, apply the reduction factor of 10 to the HEW estimate, it becomes 0.1 to 1 latent cancer fatalities.
Further, as indicated in the attachment, it appears that you may have misread our curves since the lowest value shown is 30. As I have already said, the RSS did not do extensive calculations at low levels of consequences.
Thus any attempt to compare the probabilities of predicted accidents in the RSS with the known probability of the TMI accident cannot be made on the basis of the information presented in the RSS.
We are all convinced that improvements of some of the RSS methods are needed and we are now doing more work in this area. We have to improve methodology arJ data, as well as to gain a better knowledge of plant differences.
The value of the RSS methods to our licensing process lies mainly in the framework it provides for microanalysis.
This is turning out to be of great assistance to us in differentiating between important and unimportant contributions to safety of various plant features and in identifying features in plants that require improvement.
g,.,
,,o ej]&Ya A AS 35.&
Q
Professor 1lilliam D. flordhaus 3
Augus t 14, 1979 If you would like more detailed information, please feel free to call Bob Budnitz, whom I believe you kno.i, or Saul Levine who was one of the principals in the RSS.
Thank you again for your thoughts. (
Sincerely, i
- \\fy-
/
\\
Joseph M' Hendrie Chairman
Enclosure:
As stated An d a,. m n
s>
Attachment Discussion of HEW and RSS Estimates of Latent Cancer Fatalities This attachment discusses the differences in comparing the projections of the latent cancer fatalities from the Three Mile Island-2 accident made by HEW with the projections made by the Reactor Safety Study (RSS).
The basic assumption that the observations of accident consequences at TMI-2 is a given, and can be directly compared with the Reactor Safety Study is not proper.
For consistency, the same assumptions must' be used to obtain health effects.
Substantive differences exist between the methods that were used to make these estimates.
In particular, the population dose and dose response calculation techniques are not compatible.
Additionally, Prof. fiordhaus appears to have misinterpreted the number of latent cancer fatalities given in the RSS by a factor of 30.
Population Dose (in person-remi TMI-2 population dose estimates have a significant number of conservative assumptions (over estimates of both dose and population) which would not have been used in an RSS type analysis.
The following factors are given in the report of the Ad Hoc Interagency Dose Assessment Grgup or have been mentioned by the authors of that report in discussions 1.
Dose measurements at TMI were overestimated due to biases in the detection devices (TLDs).
2.
Background radiation at TMI was not adequately subtracted from the dosa measurements.
3.
Population was assumed to remain in the area near TMI throughout the course of the accident with no evacuation.
4.
Population near TMI is assumed to remain outdoors throughout the course of the accident with no shielding or sheltering factors of the buildings taken into account.
The combination of these factors contributes to significant overestimates of the total population dose. As of this time, we have not attempted to make a more realistic estimate of the total person-rem or to quantify the total amount of conservatism in the TMI-2 person-rem estimates.
Dose Response OEM$.
Dose res?onse models convert given radiation exposures into impacts on health (total man-rem into total latent cancer fatalities).
Because of the sparcity of useable data, there are significant variations in the models that have been developed to predict dose response.
The uncertainties are greatest for low doses of low-LET radiation *.
The radiation that was emitted from radionuclides released during the accident at TMI-2, and
- tiote : X-rays, ganma-rays and beta particles are low-LET radiations.
2 considered in the RSS is low-LET radiation (beta and gamma emitters).
l All authoritative reports.2.3,4 on the subject indicate that "for exposures to low-LET radiation at low doses, most cancer risk estimates based on the linear (no threshold) hypothesis regarded as more than uoper limits of risk." 5It was for this reason that the RSS, in attempting to make a realistic estimate of latent cancer fatalities, did not follow the traditional linear hypothesis.
HEW Approach to Dose Response HEW's estimate of or2 fatality fcr TMI was calculated based on $dels for dose response given in the 1972, Biological Effects of Ionizing Radiation (BEIR) Report.
The BEIR models are predicated upon the "lir. ear, no-threshold hypothesis" of dose versus effect.
In addii. ion, BEIR proposed " relative" risk and " absolute" risk models for the health '
effects calculation.
HEW's estimate used a geometric mean between the
" relative" and " absolute" risk estimates rounded ul to one latent cancer fatality.
HEW's estimate of 10 fatalities represents the extreme opinion among the scientific comunity and should not be regarded as even a reasonable upper bound (95th percentile).
There are no accepted scientific data to support a codel which~would predict 10 latent carcer fatalities from even the most conservative projections of person-rem from TMI-2.
RSS Acoroach to Dose Response RSS projeg;tiens of latent cancer fatalities were based on the " central estimate"u of dose response.
The central estimate assumes a reduced incidence of latent cancer fatalities for low doses and rates but still uses a no-threshold model.
RSS also proposed a lower bound model for latent cancer fatalities which assumed a zero incidence of latent cancer fatalities below 10 rem.
For an upper bound dose response, RSS used the absolute risk model based on BEIR 1972.
Differences Between HEW and RSS For the low-LET radiation emitted and the very low doses projected at TMI-Z, the RSS model would predict about a factor of 10 less latent cane.er fatalities than would be predicted by HEW.,This factor of ab 10 is derived from a factor of five reduction due.to very low doses,out(less than one ren) and a factor of about 2 from HEW's use of both the "aMolute" and " relative" risk models.
Therefore, given the same projection of person-rem, RSS dose response model would predict about one tenth of a latent
- cancer fatality.
This point would not have shown on the curves given in the RSS because they would be off scale.
34hU2.2
3 Latent Cancer Fatalities Misinterpretation Figure 5-5, page 90 of the Main Report of the Reactor Safety Study shows theprobabilitydistributionsoflatentcancerfgtalitiesperyear,given an accident, in the text explaining this figure it is stated that this curve reflects the per year rate of latent cancer fatalities that could be expected over a 30 yeir period (10 to 40 years after the accident).
The minimum number of total latent cancer fatalities given in RSS is thus 30 instead of I as assumej by Prof. Nordhaus.
Conclusions It is not our intent to question the techniques and methods that were used in either the HEW observation or the Reactor Safety Study, but only to point out that there are significant differences between the two.
It is noted that the RSS fatality curves do not predict the probability of any consequence smaller than 30 total latent fatalities.
Further, if the HEW observation of one latent fatality is made in a manner consistent with the RSS technique, the prediction would be at least one order of magnitude smaller.
e 9,
,5
%d3
- References 1.
BEIR,1972, The Effects an Population of Exposure to Low Levels of Ionizing Radiation.
I i
2.
UN5CEAR, United Nations Scientific Committee on Atomic Radiation,1972, Ionizing Radiation: Levels and Effects.
ICRP, Intern'ational Commission on Radiological Protection, Reports 3.
9,10,14 and 26.
4.
NCRP, National Council an Radiation Prptection, Reports 39 and 43.
5.
BEIR Report,1979, BEIR-III Report page 3 6.
WASH-1400,1975, Appendix VI page 9-25 Population Dose and Health Impact of the Accident at TMI Ruclear Station 7.
Ad-Hoc Interagency Dose Assessment Group, NUREG-0558, May,1979.
8.
WASH-1400, 1975, Main Report, page 74.
e e
.(