ML20212K650
| ML20212K650 | |
| Person / Time | |
|---|---|
| Site: | Seabrook  |
| Issue date: | 01/23/1987 |
| From: | Sholly S MASSACHUSETTS, COMMONWEALTH OF, MHB TECHNICAL ASSOCIATES |
| To: | |
| Shared Package | |
| ML20212K635 | List: |
| References | |
| OL, NUDOCS 8701290141 | |
| Download: ML20212K650 (109) | |
Text
. _ . _ . ._ _ _ . . _ _ _ _ _ . . _ _ _ _ . . _ . _ . . _ . . _ _ , _ _ _ . _
-O
., 0-UNITED STATES OF AMERICA NUCLEAR REGULATORY COMMISSION Q- Before the ATOMIC SAFETY AND LICENSING BOARD
- O 3 In the Matter of )
t
)
PUBLIC SERVICE COMPANY OF ) Docket Nos. 50-443-OL NEW HAMPSHIRE, et al. ) 50-444-OL
) Of f-site Emergency 10 (Seabrook Station, Units 1 and 2 ) Planning Issues
- ) .
AFFIDAVIT OF STEVEN C. SHOLLY lO - -
1 Steven C. Sholly, being on oath deposes and says as follows:
- 1. I am an Associate Consultant with MHB Technical Associates,
- Q
- 1723 Hamilton Avenue, Suite K, San Jose, California, 95125. A statement of my professional qualifications is attached hereto and j marked Attachment A. I have more than five years experience in the O review, analysis, interpretation, and application of probabilistic risk assessment to the analysis of safety issues related to commercial nuclear power plants. I have served as a member of the peer review
O group for the NRC publication NUREG-1050 (Probabilistic Risk Assessment (PRA) Reference Document, September 1984), and have more recently served as a member of the Containment Performance Design i O Objective Workshop and the Panel on ACRS Effectiveness. I have j
- o .
i
. _ _ _ _ _ _ _ _ _ . _ _ . . , . _ _ . . _ _ _ _ _ , . . _ , _ _ _ . . . _ _ _ _ _ . . . . . _ . _ _ _ __ _ _ _ _ _ _ _ _ _ . _ ~ . , _ , - -
~
) 1 x
previously testified as an expert witness on probabilistic risk
) assessment matters in NRC proceedings on the Catawba and Indian Point nuclear plants, and also in the Public Inquiry regarding the proposed Sizewell-B nuclear plant in the United Kingdom.
- 2. MHB Technical Associates ("MHB") has ,been requested by the
)
Department of the Attorney General, The Commonwea'lth of Massachusetts, to perform a technical review of a recent filing by' Counsel for Public Service Company of New Hampshire ("PSNH") which petitions the commission for an exception or waiver of NRC regulations related to the size of the plume exposure pathway emergency planning zone (" plume EPZ"). Specifically, the petition seeks a reduction of the plu$e EPZ ,
)
from about 10 miles in radius to 1 mile in radius.
- 3. The PSNH filing consists of: (a) a letter dated ~18
. December 1986, from Thomas G. Dignan, Jr. (counselJ for'PSNH) , to the Secretary of the Commission; (b) -PSNH's petitilon for' waiver cc exception, dated 18 December 1986; (c) PSNH's memorandum in support of the petition, undated; (d) affidavits by Peter S. Littlefield,
) Shengdar Lee, John G. Robinson, and Messrs. Fleming, Torri, Woodard, Lutz, Henry, Budnitz, Aldrich, Hendrie, Rasmussen, Ritzman, Stratton, and Wilson; (e) three letters from Robert J. Sudnitz to J.
) DeVincentis, dated 9 November 1985, 17 January 1986, and 29 April 1986; (f) a report by Shengdar Lee and Peter S. Littlefield, Licensing Aspects of the Seabrook Emergency Planning Zone Study, YAEC-1502, .
} dated December 1985; (g) a Pickard, Lowe & Garrick report, Seabrook Station Risk Mangement and Emergency Planning Study, PLG-0432, dated December 1985; and (h) a second Pickard, Lowe & Garrick repont,
)
)
) Seabrook Station Emergency Planning Sensitivity Study, PLG-0465, dated April 1986.
- 4. In addition, closely related documentation has been reviewed. This documentation includes: (a) the original Seabrook PRA study, Seabrook Station Probabilistic Safety Assessment, Pickard, Lowe D
& Garrick, PLG-0300, dated December 1983; (b) the draft PRA review by Lawrence Livermore National Laboratory, A Review of the Seabrook Station Probabilistic Safety Assessment, dated 12 December 1984; (c) the review of the containment failure modes and source term analyses by Brookhaven National Laboratory, Review of Seabrook Station Probabilistic Safety Assessment: Containment Failure Modes and O Radiological Source Terms, NUREG/CR-4540, dated February 1986; (d) an NTS Engineering report, Seismic Fragilities of Structures and Comoonents at the Seabrook Generating Station, Units 1 and ~ 2, - Rev. 1,
[) NTS 15'89.01, June 1986; (e) a draft Brookhaven National Laboratory review of PLG-0465, Technical Evaluation of the EPZ Sensitivity Study for Seabrook, BNL Technical Report A-3852, December 5, 1986; and (f) g various submittals, formal and informal, concerning aspects of the PSNH-sponsored studies, prepared by consultants to PSNH, consultants to the NRC staff, and by members of the NRC staff.
, 5. In response to this request from the Attorney General's office, MHB has prepared a report documenting our review. This report is entitled, Review of PSNH EPZ Reduction Request for the Seabrook Station, and is dated January 23, 1987. A copy of this report O
attached hereto and is marked Attachment B.
9 O
)
- 1,
- 6. I performed the analysis and review documented in
) ,
Attachment B. The statements set forth in that report are true and correct to the best of my knowledge and belief.
d ,
a sh ',.
ll 5
h w .. . . I ,'
y' .
Ste'ren C. Sholly h l
23 January 1987 i
State of A / ,_ . On this the 3 day,of - - --
19.2.Z., before me, County of_ N -- , the undersigned Notary Pu , personalt ppeared
( -
wcA 9
- p i+ mgs, uno.n to me to he the ,e, son
. hose namer sehsc,,hed ;
. ?D-
'. 4578.' ::orp tAYP n.NA( !. EA*1P.Yq to the .ithin instrument and ackno.ledged that I# 1 ."
), ' 4j .
- tuc cAunt.tvA exa cuted the same fo'r the purposes therein contained. '
f kb sy c N e :: $. IC ,
4 IN WITNESS WHEREOF, I hereunto set my hand and official seal.
~
.w'f --
l
~
b [ .
GDetAAL ACEN9WLESC.90f0ff F0805 t
i i
D t
\
3 .
)
ATTACHMENT A PROFESSIONAL QUALIFICATIONS OF STEVEN C. SHOLLY
)
STEVEN C. SHOLLY MHB Technical Associates 1723 Hamilton Avenue Suite K San Jose, California 95125
) (408) 266-2716
. EXPERIENCE:
September 1985 - PRESENT
! Associate - MiB Technical Associates, San Jose, California Associate in energy consulting firm that specializes in technical and economic assessments of energy production facilities, especially nuclear, for local, state, and federal governments and orivate organizations. MHB
) is extensively involved in regulatory proceedings and the preparation of studies and reports. Conduct research, write reports, participate in
~
discovery process in regulatory proceedings, develop testimony and other documents for regulatory proceedings, and respond to client inquiries.
Clients have included: State of California, State of New York, State of Illinois.
) February 1981 - September .1985 l ,
Technical Research Associate and Risk Analyst - Union of Concerned Scien-tists, Washington, D.C.
) Research associate and risk analyst for public interest group based in Cambridge, Massachusetts, that specializes in examining the impact of ad-vanced technologies on society, principally in the areas of arms control and energy. Technical work focused on nuclear power plant safety, with emphasis on probabilistic risk assessment, radiological emergency planning and preparedness, and generic safety issues. Conducted
) research, prepared reports and studies, participated in administrative proceedings before the U.S. Nuclear Regulatory Commission, developed testimony, anlayzed NRC rule-making proposals and draft reports and prepared comments thereon, and responded to inquiries from sponsors, the general public, and the media. Participated as a member of the Panel on ACRS Effectiveness (1985), the Panel on Regulatory Uses of Probabilistic Risk Assessment (Peer Review of NUREG-1050; 1984), Invited Observer to
) NRC Peer Review meetings on the source term reassessment (BMI-2104; 1983-1984), and the Independent Advi-sory Comittee on Nuclear Risk for the Nuclear Risk Task Force of the National Association of Insurance Comissioners (1984).
)
)
January 1980 - January 1981
)- Project Director and Research Coordinator - Three Mile Island Public Interest Resource Center, Harrisburg, Pennsylvania Provided administrative direction and coordinated research projects for a public interest group based in Harrisburg, Pennsylvania, centered around issues related to the Three Mile Island Nuclear Power Plant. Prepared h fundraising proposals, tracked progress of U.S. Nuclear Regulatory Com-l mission, U.S. Department of Energy, and General Public Utilities activi-l ties concerning cleanup of Three Mile Island Unit 2 and preparation for restart of Three Mile Island Unit 1, and monitored developments related to emergency planning, the financial health of General Public Utilities, and NRC rulemaking actions related to Three Mile Island.
)
j July 1978 - January 1980 Chief Biological process Operator - Wastewater Treatment Plant, Derry Township Municipal Authority, Hersney, Pennsylvania
) Chief Biological Process Operator at a 2.5 million gallon per day ter-tiary, activated sludge, wastewater treatment plant. Responsible for bi-ological process monitoring and control, including analysis of physical, chemical, and biological test results, procees fluid and mass flow man-agement, micro-biological analysis of activiated sludge, and mintenance of detailed process logs for input into state and federal reports on treatment process and effluent quality. Received certification from the Commonwealth of Pennsylvania as a wastewater . treatment plant operator.
Member .of . Water Pollution Control Association of Pennsylvania, Central Section, 1980.
July 1977 - July 1978
) Wastewater Treatment Plant Operator - Borcuch of Lemoyne, Lemoyne, Penn-sylvania Wastewater treatment plant operator at 2.0 million gallon per day sec-ondary, activated sludge, wastewater treatment plant. Performed tasks as assigned by supervisors, including simple physical and chemical tests on -
) wastewater streams, maintenance and operation of plant equipment, and maintenance of the collection system.
September 1976 - June 1977 .
Science Teacher - West Shore School District, Camo Hill, Pennsylvania
) Taught Earth and Space Science at ninth grade level. Developed and im-plemented new course materials on plate tectonics, environmental geology, and space science. Served as Assistant Coach of the district gymnastics team.
)
)
September 1975 - June 1976 Science Teacher - Carlisle Area School District, Carlisle, Pennsylvania Taught Earth and Space Science and Environmental Science at ninth grade level. Developed and implemented new course materials on plate tecton-ics, environmental geology, noise pollution, water pollution, and energy.
Served as Advisor to the Science Projects Club.
)
1 EDUCATION:
B.S., Education, majors in Earth and Space Science and General Science, minor in Environmental Education, Shippensburg State College, Shippens-burg, Pennsylvania, 1975.
) Graduate coursework in Land Use Planning, Shippensburg State College, Shippensburg, Pennsylvania, 1977-1978.
PUBLICATIONS:
- 1. " Determining Mercalli Intensities from Newspaper Reports," Journal of Geological Education, Vol. 25, 1977.
- 2. A Critioue of: An Independent Assessment of Evacuation Times for Three Mile Island Nuclear Power Plant, Three Mile Island Public Interest
) Resource Center Harrisburg, Pennsylvania, January 1981.
- 3. A Brief Review and Critique of the Rockland County' Radiological Emergency Preparedness Plan, Union of Concerned Scientists, prepared for Rockland County Emergency Planning Personnel ;,nd the Chairman of the County Legis-lature, Washington, D.C., August 17, 1981.
) 4. The Necessity for a Prompt Public Alerting Capability in the Plume Expo-sure Pathway EPZ at Nuclear Power Plant Sites, Union of Concerned Scien-
, tists, Critical Mass Energy Project, Nuclear Information and Resource Service. Environmental Action, and New York Public Interest Research Group, Washington, D.C., August 27, 1981. *
) 5. " Union of Concerned Scientists, Inc., Comments on Notice of Proposed Rulemaking, Amendment to 10 CFR 50, Appendix E.Section IV.O.3 " Union of Concerned Scientists, Washington, D.C., October 21, 1981. *
- 6. "The Evolution of Emergency Planning Rules," in The Indian Point Book: A Briefinq on the Safety Investigation of the Indian Point Nuclear Power
} Plants, Anne Witte, editor, Union of Concerned Scientists (Washington, D.C.) and New York Public Interest Research Group (New York, NY), 1982.
- 7. "Unicn of Concerned Scientists Coments, Proposed Rule,10 CFR Part 50, Emergency Planning and Preparedness: Exercises, Clarification of Regula-tions, 46 F.R. 61134," Union of Concerned Scientists, Washington, D.C.,
January 15, 1982. *
)
)
)
- 8. Testimony of Robert D. Pollard and Steven C. Sholly before ' the Sub-
)
comittee on Energy and the Environment, Committee on Interior and Insular Affairs, U.S. House of Representatives, Middletown, Pennsylvania, March 29, 1982, available from the Union of Concerned Scientists.
, 9. " Union of Concerned Scientists Detailed Comments on Petition for Rulemak-l ing by Citizen's Task Force Emergency Planning,10 CFR Parts 50 and 70,
)- Docket No. PRM-50-31, 47 F.R.
Washington, D.C., May 24, 1982.
12639," Union of Concerned Scientists, L
- 10. Supplements to the Testimony of Ellyn R. Weiss, Esq., General Counsel, Union of Concerned Scientists, before the Subconraittee on Energy Conservation and Power, Comittee on Energy and Commerce, U.S. House of Representatives, Union of Concerned Scientists, Washington, D.C., August 16, 1982.
- 11. Testimony of Steven C. Sholly, Union of Concerned Scientists, Washington, D.C., on behalf of the New York Public Interest Research Group, Inc., be-fore the Special Committee on Nuclear Power Safety of the Assembly of the State of New York, hearings on Legislative Oversight of the Emergency Ra-
) diologic Preparedness Act, Chapter 708, Laws of 1981, September 2,1982.
- 12. " Comments on 'D. raft Supplement to Final Environmental Statement Related to Construction and Operation of Clinch River B. eeder Reactor Plant',"
Docket No. 50-537, Union of Concerned Scientists. Washington, D.C.,
September 13, 1982. *
- 13. " Union of Concerned Scientists Comments on 'Repart to the County Comis-stone'rs', by the Advisory Comittee on RadiolJgical Emergency Plan for Columoia County, Pennsylvania," Union of Concterned Scientists, Washing-ton, D.C., September 15, 1982.
) 14. " Radiological Emergency Planning for Nuclear Reactor Accidents," pre-sented to Kernenergie Ontmanteld - Congress, Rotterdam. The Netherlands, Union of Concerned Scientists, Washington, D.C., October 8,1982.
- 15. " Nuclear Reactor Accident Consequences: Implications for Radiological Emergency Planning," presented to the Citizen's Advisory Comittee to Re-
) view Rockland County's Own Nuclear Evacuation and Preparedness Plan and General Disaster Preparedness Plan, Union of Concerned Scientists, Wash-ington, D.C. , November 19, 1982.
- 16. Testimony of Steven C. Sholly before tre Subcomittee on Oversight and Investigations, Comittee on Interior ar.d Insular Affairs, U.S. House of
) Representatives, Washington, D.C., Union of Concerned Scientists. Decem-ber 13, 1982.
- 17. Testimony of Gordon R. Thompson and Staven C. Sholly on Comission Ques-tion Two. Contentions 2.1(a) and 2.1(d), Union of Concerned Scientists and New York Public Interest Research Group, before the U.S. Nuclear Reg-ulatory Comission Atomic Safety and Licensing Board, in the Matter of
.) Consolidated Edison Company of New York (Indian Point Unit 2) and the Power Authority of the State of New York (Indian Point Unit 3), Docket Nos. 50-247-SP and 50-286-SP, December 28, 1982.
- l c _
- 18. Testimony-of Steven C. Sholly on the Consequences of Accidents at Indian Point (Commission Question One and Board Question 1.1, Union of Concerned Scientists and New York Public Interest Research Group, before the U.S.
Nuclear Regulatory Comission Atomic Safety and Licensing Board, in the Matter of. Consolidated Edison Company of New York (Indian Point Unit 2) and the Power Authority of the State of New York (Indian Point Unit 3),
Docket Nos. 50-247-SP and 50-286-SP, February 7, 1983, as corrected February 16, 1983. *
( 19. Testimony of Steven C. Sholly on Commission Question Five. Union of Con-cerned Scientists and New York Public Interest Research Group, before the U.S. Nuclear Regulatory Commission Atomic Safety and Licensing Board, in
' the Matter of Consolidated Edison Company of New York (Indian Point Unit
- 2) and the Power Authority of the State of New York (Indian Point Unit 3), Docket Nos. 50-247-SP and 50-286-SP, March 22, 1983. *
- 20. " Nuclear Reactor Accidents and Accident Consequences: Planning for the Worst," Union of Concerned Scientists, Washington, D.C. , presented at Critical Mass '83, March 26,1983.
)
- 21. Testimony of Steven C. Sholly on Emergency Planning and Preparedness at Commercial Nuclear Power Plants, Union of Concerned Scientists Washing-ton, D.C., before the Subcomittee on Nuclear Regulation, Comittee on Environment and Public Works, U.S. Senate April 15,1983, (with " Union of Concerned Scientists' Response to Questions for the Record from Sena-tor Alan K. Simpson," Steven C. Sholly and Michael E. Faden).
)- 22. "PRA: What Can it Reall'y Tell Us About Public Risk from Nuclear Ac-cidents?," Union of Concerned Scientists, . Washington, D.C., presentation .
to the 14th Annual Meeting, Seacoast Anti-Pollution League, May 4,1983.
- 23. "Probabilistic Risk Assessment: The Impact of Uncertainties on Radt-ological Emergency Planning and Preparedness Considerations," Union of
) Concerned Scientists, Washington, D.C., June 28, 1983.
- 24. " Response to GAO Questions on NRC's Use of PRA," Union of Concerned Sci-entists, Washington, D.C., October 6,1983, attachment to letter dated October 6,1983, from Steven C. Sholly to John E. Bagnulo (GAO, Washing-y ton,D.C.). .
- 25. The Imoact of " External Events" on Radiological Emergency Response Plan-ning Considerations, Union of Concerned Scientists, Wasnington, D.C., De-cemoer 22, 1983, attachment to letter dated December 22, 1983, from Steven C. Sholly to NRC Comissioner James K. Asselstine.
) 26. Sizewell 'B' Public Inquiry, Proof of Evidence on: Safety and Waste Man-agement Imolications of the Sizewell PWR, Gordon Thompson, with supporting evidence by Steven Sholly, on benalf of the Town and Country Planning Association, February 1984, including Annex G. "A review of Probabilistic Risk Analysis and its Application to the Sizewell PWR,"
Steven Sholly and Gordon Thompson, (August 11, 1983), and Annex 0,
) " Emergency Planning in the UK and the US: A Comparison," Steven Sholly and Gordon Thompson (October 24,1983).
)
f l
- 27. Testimony of Steven C. Sholly on Emergency Planning Contention Number Eleven, Union of Concerned Scientists, Washington 0.C., on behalf of the Palmetto Alliance and the Carolina Environmental Study Group, before the j
U.S. Nuclear Regulatory Connission Atomic Safety and Licensing Board, in the Matter of Duke Power Company, et. al. (Catawba Nuclear Station, Units I and 2), Docket Nos. 50-413 and 50-414, April 16, 1984. *
- 28. " Risk Indicators Relevant to Assessing Nuclear Accident Liability Premi-ums," in Preliminary Report to the Independent Advisory Committee to the
_ NAIC Nuclear Risk Task Force, December 11, 1984, Steven C. Sholly. Union of Concerneo Scientists, Washington, D.C.
l 29. " Union of Concerned Scientists' and Nuclear Information and Resource Ser- .
vice's Joint Comments on NRC's Proposal to Bar from Licensing Proceedings the Consideration of Earthquake Effects on Emergency Planning," Union of Concerned Scientists and Nuclear Information and Resource Service, Wash-
) ington, D.C., Of ane Curran and Ellyn R. Weiss (with input from Steven C.
Sholly), February 28, 1985. *
- 30. " Severe Accident Source Terms: A Presentation to the Connissioners on the Status of a Review of the NRC's Source Term Reassessment Study by the i Union of Concerned Scientists," Union of Concerned Scie.ntists, Washing-
) ton, D.C. , April 3,1985. *
- 31. " Severe Accident Source Terms for Light Water Nuclear Power Plants: A Presentation to the Illinois Department of Nuclear Safety on the Status of a Review of the NRC's Source Term Reassessment Study (STRS) by the Union of Concerned Scientists," Union of Concerned Scientists,
) Washington, D.C., May 13, 1985.
- 32. The Source Term' Debate: A Review of the Current Basis for Predicting Se-vere Accident Source Terms with Special Empnasis on the NRC Source Term Reassessment Program (NUREG-0956), Union of Concerned Scientists, Cam-bridge, Massachusetts, Steven C. Sholly and Gordon Thompson, January y 1986.
- 33. Direct Testimony of Dale G. Bridenbaugh, Gregory C. Minor, Lynn K. Price, and Steven C. Sholly on behalf of State of Connecticut Department of Pub-lic Utility Control, Prosecutorial Division and Divi,sion of Consumer Counsel, regarding the prudence of expenditures on Millstone Unit III, February 18, 1986.
)
- 35. Review of Vermont Yankee Containment Safety Study and Analysis of
) Containment Venting Issues for the Vermont Yankee Nuclear Power Plant, prepared for New England Coalition on Nuclear Pollution, Inc., December 16, 1986.
Available from the U.S. Nuclear Regulatory Commission, Public Document Room, Lobby,1717 H Street, N.W., Washington, D.C.
)-
6
)
l.
Y .
~
I ATTACHMENT B REVIEW OF PSNH EPZ REDUCTION REQUEST FOR THE SEABROOK STATION b -
MHB TECHNICAL ASSOCIATES 23 January 1987 prepared for
) OFFICE OF THE ATTORNEY GENERAL l
COMMONWEALTH OF MASSACHUSETTS
) .I . INTRODUCTION On December 18, 1986, Public Service Company of New Hampshire (PSNH) filed a petition with the U.S. Nuclear Regulatory Commission (NRC) seeking a reduction in the size of the plume exposure pathway
)
emergency planni'ng zone '(" plume EPZ") from ten miles to one mile. The Department of the Attorney General, The Commonwealth of Massachusetts, requested MHB Technical Associates (MHB) to perform a focused technical review of the EPZ reduction petition and related documents.
)
The purpose of this report is to set forth the results of that review.
The remainder of this report is organized as follows. Section II describes the technical basis for the existing plume and ingestion EPZs. Section III addresses a number of specific technical issues
) that have arisen during our review of the Seabrook EPZ reduction petition and related documentation. Section IV sets forth our' conclusions and recommendations. Figures and Tables are found at the end of the report.
) This report was prepared under severe time constraints. The basis for PSNH's filing consists essentially of a full-scope probabilistic risk assessment (PRA), with two updates and two additional source term analyses, and various other submittals. Under
) normal conditions, it would be reasonable to expect that to perform a
)
r detailed review of such an assessment would require a team of reviewers at least six to twelve months. 1/ In addition, extensive calculations of accident sequence frequencies, containment event tree j split fractions, and sensitivity studies on these and other matters l would normally be performed in the conduct of such a review. Due to f
i the exigencies of NRC administrative law, the review period for this l
report was limited to a few weeks, and a full requantification was i simply impossible within this time frame.
II. TECHNICAL BASIS FOR EXISTING EPZ DISTANCES Before proceeding with a review of the PSNH pe tition and related documents, it is important for the reader to understand the technical basis for the existing plume and ingestion pathway EPZ distances. This section of the report discusses the origin and technical bases of these planning zones.
D -
II.1 NRC/ EPA Task Force on Emergency Planning In October 1975, the U.S. Nuclear Regulatory Commission (NRC)
D released the Reactor Safety Study (WASH-h.40 0) . WASH-1400 consisted in large part of a probabilistic risk assessment (PRA) of two nuclear power plants, namely the Surry PWR and the Peach Bottom BWR. 2_/
D
-1/ We note, for example, that Brookhaven National Laboratory required a review period of several months to produce a draft report addressing only limited aspects of the applicant's filings. BNL's report is not yet complete; moreover, it addresses selected issues raised by the PSNH/PLG filings.
3 2/ Surry is a two-unit nuclear power station employing a three-loop Westinghouse NSSS, U-tube steam generators, and a large, dry, subatmospheric containment. Peach Bottom is a three-unit ,
nuclear power station, two units of which employ a General Electric BWR/4 NSSS with a steel Mark I containment. The first J
) -A key feature of the WASH-1400 analysis was the use in the accident consequence model 3/ of an evacuation zone extending to five miles in all directions ' from the plant site, and extending to a l' distance of twenty-five miles in the downwind direction. 4/ At the time the WASH-1400 analysis was performed, and continuing until after
) the Three Mile Island accident in 1979, offsite emergency planning considerations for evacuation (and other population protective measures) were limited to the low population zone. The low population
) zone is defined in NRC's regulations at 10 CFR Part 100. For most plants, the low population zone is limited to a distance of no more than two miles from the plant. 5/
] The contrast between the regulatory requirements and the assumptions in the accident consequence model in WASH-1400 was noted by the Conference of State Radiation Control Program Directors. An ad g hoc Task Force of the Conference passed a resolution in 1976 requesting the NRC to "make a determination of the most severe Peach Bottom unit was a small gas-cooled reactor; it has been shut down and is in the process of being decommissioned.
p 3/ The accident consequence model used in WASH-1400 is documented in Appendix VI of that study. It is referred to as the CRAC code; CRAC is an acronyn for Calculation of Reactor Accident Consequences. An updated version of this code locame available In the early 1980s, and is referred to as CRAC2. A proprietary D version of tha CRAC code, called CRACIT, was used in the Seabrook PRA studies. Most recently, Sandia National Laboratories has developed the MELCOR Accident Consequences Code System (MACCS); thic is an updated version of CRAC2 with new health offacts models and enhanced meteorological modelling. MACCS is being used to calculate accident g consequences for the forthcoming NUREG-ll50 report.
4/ WASH-1400, Appendix VI, see page 2-5.
5_/ For example, the low population zone for the Seabrook site was set by the Atomic Safety and Licensing Appeal Board in 1977 to be 1.25 miles.
O
y .
accident basis for which radiological emergency response plans should
)
l be developed by offsite agencies." In addition, both the NRC and the U.S. Environmental Protection Agency (" EPA") received correspondence from state and local governments in this regard.
In response to these inquiries, the NRC and EPA organized a Task Force on Emergency Planning in November 1976._ The Task Force produced a report in December 1978 which is referred to by its report
) number, NUREG-0396. 6/ This report recommended the creation of two emergency planning zones ("EPZs"). EPZs were designated "as the areas for which planning is recommended to assure that prompt and ef fective
) actions can be taken to protect the public in the event of an accident. It was recommended that a plume exposure pathway EPZ with a radius of about ten miles and an ingestion exposure pathway EPZ with a 3 radius of about fifty miles be established. The EPZ concept set forth in NUREG-0396 is illustrated in Fiqure 1.
It is clear from NUREG-0396 as well as a related Sandia-Laboratories report 7/ that the EPZ distances resulted in large measure from accident consequence analysis carried out using the WASH-1400 PRA results for the Surry and Peach Bottom plants. In D
6/ Task Force on Emergency Planning, Planning Basis for the Development of State and Local Government Radiological
] Emergency Response Plans in Support of Light Water Nuclear Power Plants, U.S. Nuclear Regulatory Commission and U.S.
Environmental Protection Agency, NUREG-0396, EPA 520/1-78-016, December 1978.
7/
D.C. Aldrich, P.E. McGrath, and N.C. Rasmussen, Examination of g Offsite Radiological Emergency Measures for Nuclear Reactor Accidents Involving Core Melt, Sandia Laboratories, NUREG/CR-1131, SAND 78-0454, June 1978, reissued October 1979.
O
)
particular, the source terms 8_/ and relative frequencies of the source
) terms for Surry and Peach Bottom were used to generate dose / distance relationships which served as part of the basis for the EPZ distances.
) II.2 NRC/ FEMA Emergency Planning Guidance Following the accident at Three Mile Island in March 1979, the
} President designated the Federal Emergency Management Agency (FEMA) as the lead federal agency with respect to emergency response to nuclear reactor accidents. Together with the NRC, FEMA defined emergency planning standards to be used in the development and review of emergency plans for reactor accidents. This effort resulted in'the publication of NUREG-0654. 9/
The joint NRC and FEMA emergency planning guidance sets forth the the objective.of radiological emergency plans as follows: 10/ 0
)
-8/ A source term, in the reactor safety context, is a statement of the release magnitudes and chemical and physical
) characteristics of a radiological release from a nuclear power plant during a severe accident.
9_/ U.S. Nuclear Regulatory Commission and Federal Emergency Management Agency, Criteria for Preparation and Evaluation of Radiological Emergency Response Plans and Preparedness in Support of Nuclear Power Plants, NUREG-0654, FEMA-REP-1,
) January 1980. Revision 1 to this report was released in November 1980.
0 10_/ U.S. Nuclear Regulatory Commission and Federal Emergency Management Agency, Criteria for Preparation and Evaluation of Radiological Emergency Response Plans and Preparedness in
) Support of Nuclear Power Plants, NUREG-0654, FEMA-REP-1, Rev.
1, November 1980, page 6.
)
i' l.
)1 The overall objective of emergency response plans is to provide dose savings (and in some cases immediate life saving) for spectrum of accidents that could provide
)- offsite doses in excess of Protective Action Guides (PAGs). 11_/
Although a number of accident descriptions were used in the h development of this guidance (including the WASH-1400 core melt f'
accident releases), the NRC/ FEMA report states that the " selected planning basis is independent of specific accident sequences". M/ -
1 The FEMA /NRC guidance adopted the plume and ingestion EPZs proposed in NUREG-0396. In addition, these EPZs were incorporated into NRC regulations. The guidance defines the EPZs as"the area for
') which planning is needed to assure that prompt and ef fective actions can be taken to protect the public in the event of an accident." M/
FEMA /NRC guidance specifies that the plume EPZ is, concerned
) with exposures from whole-body ext'ernal exposure to gamma radiation ,
from the plume and deposited materials, and inhalation exposure from the passing plume. The duration of releases leading to such exposures was specified as ranging from "one-half hour to days." For the ingestion EPZ, the exposure pathways were specified as ingestion of contaminated water or foods (such as milk, fresh vegetables, or
)
M/ The EPA PAG doses are whole body doses from 1-5 Rom, and thyroid doses from 5-25 Rem. The PAG doses and the rationale behind them is fully discussed in,, U.S. Environmental
) Protection Agency, Manual of Protective Action Guides and Protective Actions for Nuclear Incidents, EPA-520/1-75-001, September 1975, Revised June 1980.
M/ Id., page 7.
M/ Id., page 10.
)
aquatic foodstuf fs) . The duration of potential exposure was specified
) as "from hours to months." M/
The size of the plume EPZ was based primarily on four considerations: M/
h
- a. projected doses from the traditional design basis accidents would not exceed Protective Action Guide levels outside the zone;
- b. projected doses from most core melt sequences would not exceed Protec tive Action Guide levels outside
) the zone;
- c. for the worst core melt sequences, immediate life threatening doses would generally not occur outside the zone;
) d. detailed planning within 10 miles would provide a substantial base for expansion of response efforts in the event that this proved necessary.
A similar set of considerations was used in selecting the size-
'of the ingestion EPZ: M/
- a. the downwind range within which contamination will generally not exceed the Protective Action Guides is
) limited to about 50 miles from a power plant because of wind shifts during the release and travel periods;
- b. there may be conversion of atmospheric iodine (i.e.,
iodine suspended in the atmosphere for long . time
) periods) to chemical forms which do not readily enter the ingestion pathway;
- c. much of any particulate material in a radioactive plume would have been deposited on the ground within about 50 miles from the facility; and M/ Id., pages 8-10.
5 1_5/ Id., page 12. ,
) M/ Id., page 13.
)
i
- d. the likelihood of e::ceeding ingestion pathway protective action guide levels at 50 miles is comparable to tha likelihood of exceeding plume
) exposure pathway protective action guide levels at 10 miles.
t III. SPECIFIC ISSUES RAISED BY PSNH FILING AND RELATED DOCUMENTATION This section of the report discusses a number of specific issues raised by the PSNH EPZ reduction petition and related documentation (i.e., the Seabrook PRA studies and reviews thereof) .
The issues are addressed in no particular order of significance.
) Due to the severe time constraints under which this report was prepared, it was not possible to always identify appropriate source terms for the changes to the applicant's and BNL's analyses which we propose. In addition, it was not possible to perform site-specific
) .
consequence analyses to il'lustrate in the most accurate manner the dose-response relationships that would resu'.t from the source terms which we do postulate herein.
) .
III.1 Accidents Caused by Sabotage / Terrorism
) PRAs to date have not explicitly considered accidents arising from acts of sabotage or terrorism. The principal reason given for this deliberate exclusion is the difficulty of establishing the
) frequency of sabotage attempts of varying severity and sophistication.
)
J
b
While the SSPSA constitutes a state-of-the-art probabilistic safety assessment, especially with respect to the issue of completeness, it does not include. a specific analysis of the risk associated with acts of war, terrorism, or sabotage. It simply was not considered appropriate nor purposeful to include such
} deliberate acts within an investigation of potential accidents. However, the analysis presented in this report includes the full spectrum of accident scenarios that are possible in the context of the level of damage that could result. In other words, the so-called
" worst-case" scenarios are included; i.e., those that involve postulated, albeit low likelihood, failure of all safety systems, worst case failure of the containment and most unfavorable me teorological conditions, and their interactions with emergency evacuation. Therefore, an act of war, sabotage, or terrorism could not cause a higher degree of damage to the public than that already considered within the scope 3 of the project. Hence, the exclusion of sabotage, terrorism, and acts .of war, is only with respect to the estimation of the frequency of occurrence.
J The " worst-case" accident for a pressurized water reactor would involve a gross failure of the containment boundary (above ground) at a time of maximum radioactive aerosol loading in the containment. In the context of the original PRA study, such a worst-case accident was 3
included, and the position regarding sabotage and terrorism accidents set forth above represented a reasonable position.
Among the source terms included in the SSPSA was Sl; S1 postulated the release of 94% of the noble gases, 75% of' the iodine
_1_7/ Karl N. Fleming, et 1., Seabrook Station Probabilistic Safety D Assessment, Pickard, Lowe & Garrick, Inc., PLG-0300, prepared for Public Service Company of New Hampshire and Yankee Atomic Electric Company, December 1983, page 1.2-1.
_ _ _ _ _ _ _ _ _ _ _ _ _ _ - - __---_____-______________j
)
and cesium gr'oups, 39% of the tellurium group, 9.3% of the barium
)- group, 46% of the ruthenium group, and one quarter of one percent of the lanthanide group. 18/ Similarly large source terms were identified in the 1975 WASH-1400 report and the 1982 "Sandia Siting Study" report. Table 1 compares the source term characteristics of
)
SSPSA source term S1, WASH-1400 source term PWR-1, and Sandia Siting Study source term SST-1. It would be difficult to postulate conditions which would give rise to a significantly larger release than postulated in the source terms listed in Table 1. Thus, it could be argued that WASH-1400 and the Sandia Siting Study included the largest feasible releases. Since the NUREG-0396 EPZ distances are
)
based on the WASH-1400 source terms, it could reasonably be argued that the existing EPZ distances implicitly account for the health effects' of accidents caused by sabotage and/or terrorism, although not
)-
for the probability of such accidents.
In the context of the more recent Seabrook studies (PLG-0432 and PLG-0465), this is no longer true due to reductions in source term
) magnitudes and increases in release times taken in the newer studies.
Even the newest source term computer models show very large release fractions when early containment failure is assumed. M/ Thus
)
M/ Id., page 2.1-8.
M/ The principal difference with the new source term models is that for transient sequences and very small LOCAs, the newer models take account of fission product deposition within the
) reactor coolant system, whereas WASH-1400-era models ignore this source term attenuation mechanism.* There are large uncertainties surrounding the RCS deposition modeling in that the NRC-sponsored codes do not account for the possibility of re-evolution of deposited fission products (due to heating, high gas flow rates, and/or chemical changes over time) . See, for example, Richard S. Denning, " Perspectives on Technical
) Issues Evolving from the Source Term Reassessment," Battelle
)
_____a
)- l l
I sabotage-initiated accident sequences cannot be dismissed out-of-hand.
2_0/ Moreover, it is meaningless in the context of a probabilistic analysis of risk to assert that the potential for risk from sabotage has been recognized, that methods to reduce this risk have been implemented, and that therefore such accidents need not be considered.
)
2J/ If this rationale were to be more broadly applied, one would never analyze any of the accidents traditionally considered in PRAs because those accidents could be excluded by an analogous argument --
) Columbus Laboratories, in NUREC/CP-0058, Vol. 2, Proceedings of the U.S. Nuclear Regulatory Commission Twelfth Water Reactor Safety Research Information Meeting, Gaithersburg, Maryland, 10/22-10/26/1984, published January 1985, 20/ An NRC staff reviewer also questioned the omission of sabotage.
"My final concern with the reported work, and the risk assessment approach, is that we are ignoring the potential risk of sabotage. This is just as real as an earthquake or any of the equipment failures which can initiate a severe accident, and it will impact risk. I do not believe that the rational (sic) of limiting ,the comparison to topics used by the
) formulators of the old emergency planning is sufficient when one is potentially considering one or two mile radius zones..
We are already violating this rational (sic) with the large number of potential accident paths considered in the Seabrook PRA and with seismic considerations. Of importance here, there are items that were negligible with the old, highly
) conservative assumptions that became significant with removal of conservatisms based upon recent knowledge. These may impact risk assessments when a reduced zone size is considered. Steam generator tube rupture due to overheating is one such item.
Sabotage is another." See, internal NRC memorandum dated Octcber 6, 1986, from Warren C. Lyon to Carl Berlinger, 3
Subject:
" Review of Seabrook Documents Pertaining to Change in Emergency Planning Lone Size," page 3.
21/ Such an argument war advanced in WASH-1400. See, WASH-1400, Main Report, Section 5.4.6, page 71. This line of reasoning fails to adequately account for insider sabotage. An " insider" saboteur is one with authorized access to and intimate knowlege
) of the nuclear power plant. Within a risk framework and considering insider saboteur capabilities, the key factors would seem to be the motivation and technical capabilities of the potential saboteur ( s) , the effectiveness of routine operational and security practices at detecting sabotage attempts, and the ef fectiveness of steps taken to mitigate an
) accident in progress.
) .
m==na.s--aei .
l l 1.e., that defense-in-depth, provision for multiple, independent, I
redundant, and diverse safety systems, and other reactor design practices would prevent accidents.
l On the basis of the historical record to date, one would not
) necessarily expect reactor accidents arising from sabotage or terrorism to be a dominant contributor to core melt frequency. While there have been some sabotage-initiated accident precursor events,
) they have been relatively few in number. 2 On the other hand, 22/
there have been a large number of precursors of accidents arising from causes other than sabotage, and the contribution of such precursors to '
) severe accident likelihood is well analyzed and documented. 2J/
This line of argument is only appealing to a point, however, and there are reasons for considering it to be misleading. First,
) there is no basis to assume that the existing historical record adequately indicates what future sabotage / terrorism attempts will
)
22,/
2 The incidents to date have involved sabotage directed against offsite power sources, or single on-site safety systems such as auxiliary feedwater, high pressure injec tion, or emergency AC power (diesel generators) .
23/
~~
See, J.W. Minarick & C.A. Kukielka, Precursors to Potential Severe Core Damage Accidents: 1969-1979, A Status Report, Oak Ridge National Laboratory, NUREG/CR-2497, ORNL/NSIC-182, Vols.
1 and 2, June 1982; and W.B. Cottrell, et al., Precursors to Potential Severe Core Damage Accidents: 1980-1981, A Status
) Report, Oak Ridge , National Laboratory, NUREG/CR-3591, ORNL/NSIC-217, Vols. I and 2, July 1984. For example for the 1980-1981 period, the " industry-average potential severe core damage frequency" was estimated at 1.6 x 10-4 per reactor-year for 58 precursor events. One of these was an insider sabotage attempt at Beaver Valley. The estimated frequency of severe core damage for this event was 2.4 x 10-6 per reactor year.
) This suggests, but does not prove, that sabotage may be a minor contributor to core melt frequency.
)
involve in terms of sophistication or the level of violence. 2J/
Second, while hypothetical accidents initiated by sabotage or terrorism may not be large contributors to core melt frequency, they may nevertheless be important contributors to overall risk by representing a leading cause of large releases. Third, the historical record is relatively short and may not be representative. 25/
A final consideration is that the totality of technica'l work done on Seabrook in specific and severe accidents at pressurized water reactors in general represents an important body of information for potential saboteurs or terrorists. This body of information sets forth in considerable detail the combinations of failures which lead to severe accidents, and indicates quite clearly what containment failure modes are needed to maxi:nize various types of consequences.
Prior to the release of the WASH-1400 analysis in 1975, and the
)
-24/ Indeed, recent history, involving the use of vehicle bombs in the Middle East, revealed a new mechanism for sabotage that has received attention in the context of nuclear plants in this country. Although much of the analytical studies involving the 3 threat posed to nuclear power plants by vehicle bombs are as Safeguards classified Information and thus publicly unavailable, it appears that such a sabotage / terrorism mode of attack poses the threat of substantial damage to a reactor even at relatively large setback distances.
g 25/ For example, in addition to there being only a limited number of sabotage precursors, there have been no severe accident precursors to date from seismic events. In contrast, PRA studies for a number of plants (Seabrook included) predict that seismically-initiated accidents are important contributors to core melt frequency and, in some cases, to various risk measures (such as early fatalities or cancer deaths) . Thus, 3 the historical. record may not accurately reflect all potential contributors to core melt frequency and various risk measures.
Indeed, it is precisely because the historical record is so limited and that accidents in general are relatively unlikely that one must resort to probabilistic risk analyses rather than actuarial techniques to estimate the level of risk arising from D nuclear power plant accidents.
O
release of the Seabrook PRA in 1983, the same body of technical information was not broadly available. Now such information exists on a detailed, plant-specific basis. 6 26f Thus, accidents caused by sabotage and/or acts of terrorism remain as a real but currently unquantified source of risk. There is currently no reasonable rationale --
either pr'obabilistic or deterministic --
for excluding accidents initiated by sabotage or
)- terrorism from emergency planning considerations. One is thus forced to consider the possibility that a large, early release could indeed occur. On this basis along, it seems unlikely that a 'small EPZ can be justified. Consideration of accidents initiated by sabotage and terrorism leads to a conclusion that a ten-mile plume EPZ and a fif ty-mile ingestion EPZ --
as set for th in NRC regulations --
should be
) retained.
III.2 Event V Modelinq
)'
Previous PRA analyses of " Event V" & have assumed that the result of the failure of the interfacing valves in the RHR piping is a
) -
26/ We do not mean to suggest that PRAs should be kept secret.
Rather, the public availability of PRA studies has resulted in a notable increase in available knowledge concerning the risks of nuclear power plant operation, and has resulted in a reduction of that risk with implementation of appropriate The usefulness of PRA informa tion to saboteurs is,
) backfits.
however, an unavoidable consequence of this process.
2y So-called in WASH-1400, " Event V" is a reference to a loss-of-coolant-accident created by the failure of valves separating the reactor coolant system from the residual heat removal system. These failures are postulated to result in the release
) of high pressure reactor coolant to the RHR system, which causes overpressurization of the RHR piping (which is designed for low pressure), resulting in a pipu rupture in the auxiliary
)
-1
( ,
pipe rupture in the auxiliary building. This assumption was first documented in the WASH-1400 study, and was carried through to the Seabrook PRA analys's, i published in December 1983.
In the Seabrook PRA updates represented by PLG-0432 and PLG-3 0465, PLG adopted an IDCOR model which asserts that the most likely failure in an Event V scenario is at the RHR pump seals. 28/ This results in a slower leak rate, a later core melt, and a reduced source 3 term.
In addition to this change to the Event V analysis, more recent assessments of the sequence have recognized. the potential for the RHR 3 leak point to be submerged at some point following accident initiation but before substantial fission product release. This provides an effect similar to " suppression pool scrubbing" as modeled in BWR
, severe accident analyses, resulting in a reduction of the source term (except for noble gases). ,
In the revised Event V analyses, PLG established a " fragility curve" representing the assessed conditional prcbability of RHR piping building. Reactor coolant is lost outside the containment, and is therefore unavailable for recirculation back to the reactor vessel as would normally be the case for a LOCA inside containment (in this case, coolant would collect in the sump-3 frcm which it can be pumped back into the reactor coolant system with the emergency core cooling system in the recirculation mode) . When the RWST inventory is depleted, the core uncovers and core melt ensues. Radioactivity is released outside the containment. Event V is also referred to as an
" interfacing LOCA", but it is not the only example of this g class of sequences (steam generator tube rupture is another example).
-28/ The IDCOR analysis of the Event V scenario resulting in RHR pump seal failure rather than pipe rupture was set forth in a 1984 report. See, Fauske and Associates, Inc., Evaluations of Containment Bypass and Failure to Isolate Sequences for the O IDCOR Reference Plants, Draft IDCOR Technical Report, FAI/84-9, July 1984, Section 4.
D
l
)
) rupture when overpressurized. The curve is essentially arbitrary, however. The failure likelihoods at yield and ultimate stress are assumed values based on the " engineering judgment" of the PLG analysts. These two assumed points are then used with other assumptions to construct the fragility curve. From this curve, a discrete probability distribution is created to mathema tically represent the curve. The validity of such a procedure in any context save the Bayesian analytical context adopted by PLG is certainly open to question. 29/ 30/
29/ In the Bayesian framework employed by PLG, all types of
} " evidence" are treated as having equal validity and significance. Thus, actual operating experience, results of experiments, and analysts' beliefs about reality are all accorded the status of data. Moreover, the Bayesian approach asserts that all persons, when presented with the same
" evidence", would select the same probability distribution to
) reflect their " state 'o f knowledge" -
in this case, a fragility curve representing the likelihood' of RHR pipe rupture ,
when overpressurized beyond design values. The degree of arbitrariness inherent in this approach boggles the mind. It is grounded in the unprovable assumption that all persons will evaluate the same evidence in precisely the same way and come 3 to the exact same conclusion. The the Indian Point ASLB criticized the use of Bayesian methodology in PRAs, sta ting, "While it would be justifiable to reject the Bayesian methodology on statistical grounds alone, perhaps the chief danger in using the Bayesian approach is that the seeming rigor of the algorithm and the ' engineering judgment' injected by the
] use of the priorg may lend a spurious air of reliability to the result. One may deceive oneself into believing that the calculated probability distributions are more realistic than they actually are." See, ASLB decision in the Indian Point Units 2 and 3 Safety Investigation, October 24, 1983, slip.
ol., page 44.
30/
~~
The concern over the Bayesian approach is not academic. Robert Easterling, a classical statistician at Sandia National Laboratories, pointed out that Bayesian methods applied to the estimation of the initiating event frequency of a large LOCA for the Indian Point and Zion plants resulted in very different results. Despite being performed by the same analysts and with D no change in the in the world's state-of-knowledge, the large LOCA initiating event frequency for Zion resulted in the addition of 714 LOCA-free years (which had not yet occurred),
D
t The net result of these changes to the Event V analysis is that although the initiating event frequency is assessed at 7.6 x 10-6, the frequency of the more traditional Event V scenario is lowered to 3.4 x
-8 10 Even PLG acknowledges that if operator successes in mitigating b the scenario are deleted from the analysis, the frequency of the RHR
~
seal leakage Event V scenario (postulated by PLG) rises from 3.4 x 10 to 5.9 x 10 -7 H/ Thus, by altering the traditional Event V
) analysis, PLG lowers the frequency of Event V (and its associated source term) from 7.6 x 10 -6 32,/
2 to 3.4 x 10-8, a reduction by a factor of over 220 in li'kelihood . 3y The fraquency of the more 3' while the analysis for Indian Point added "only" 344 LOCA-free years. Easterling referred to this outcome as " Whimsy." In a separate analysis of auxiliary feedwater reliability, a Bayesian analysis resulted effectively in the elimination of failures which actually occurred and an increase in the number of challenges (both of which lowered the failure rate for the
] system). See, Robert G. Easterling, " Comments on: Plan to Evaluate the Commission's Safety Goal Policy Statement,"
enclosure to a letter dated 2 June 1983 from Robert G.
Easterling, Sandia National Laboratories, to Samuel J. Chilk, NRC.
H/ Karl N. Fleming, et al., Seabrook Station' Risk Management and
] Emergency Planning Study, Pickard, Lowe & Garrick, Inc., PLG-0432, prepared for New Hampshire Yankee Division, Public Service Company of New Hampshire, December 1985, pages 2-11 to 2-12, and 2-16.
3 32,/ Note that the initiat_gg event frequency for' Event V has been raiged from 1.8 x 10 in the original Seabrook PRA to 6.5 x 10 as a result of further analysis of operating data. A13g note that although the initiating event frequency of 1.8 x 10 was a mean value, it was greater than the 95th percentile value due to the skewed shape of the probability distribution for the intiating evegt. By this curve, an initiating event frequency O of 6.5 x 10- (as now assessed in the Seabrook PRA updates) would have been extremely unlikely. This is another example of the unreliability of the Bayesian me thodology. It must be clearly kept in mind at all times that what one is seeing in a Bayesian expression is the quantified degree of belief of the analysts.
9 The pipe rupture V sequence is assigned by PLG to an'even lower
)_3 /
frequency that noted above for the RHR seal failure V sequence.
O
) p >
l traditional pipe rupture Event V scenario (as in WASH-1400) is lowered
) still further to about 3 x 10 ~9 --
a reduction by a factor of over 2,000.
PLG asserts in PLG-0432 that a " full event tree analysis" of interfacing LOCA (i.e., " Event V") sequences was performed. M/ We believe that this characterization is misleading. A closer examination of the " Event V" event tree reveals that the tree consists largely of engineering aspect's of the Event V scenario.
Only three operator actions are considered in the PLG Event V sequence event tree. The first of these is misdiagnosis, which is assumed to guarantee failure of the second operator action. Operator misdiagnosis is given a conditional probability of 6.5 x 10 ~3 per event. H/ This is a very low pechability of failure given an initiating event of such low frequency. Considering that the frequency of'the interfacing LOCA initiating event is only 7.6 x 10-6
)
) PLG places the pipe rupture V sequence at a frequency of 3.9 x 10-9. See, PLG slides used at meeting with NRC in Bethesda, Maryland, August 6, 1986.
~
34/ Karl N. Fleming, et al. , Seabrook Station Risk Manacement and
} Emergency Planning Study, Pickard, Lowe & Garrick, Inc., PLG-0432, prepared for New Hampshire Yankee Division, Public Service Company of New Hampshire, December 1985, page 1-107.
M/ Karl N. Fleming, et al., Seabrook Station Risk Management and Emergency Planning Study, Pickard, Lowe & Garrick, Inc., PLG-0432, prepared for New Hampshire Yankee Division, Public
) Service Company of New Hampshire, December 1985, page 3-89.
)
y (or about.1 chance in 131,000 per reactor-year) , 36/ the conditional ',
probability of operator misdiagnosis seems to be rather low. 37/ J,8,/
The conditional probability of misdiagnosis of 6.5 x 10~3 cited above is equivalent to saying that one would expect the operators to correctly diagnose the Event V initiating event 99.35% of the time
) (i.e., one misdiagnosis in 154 attempts). Considering the low initiating event frequency, this seems quite unlikely. Rather, an-
" incredulity response" is possible, similar to that postulated for the
. occurrence of a large LOCA, which is a more likely initiating event than Event V. 3,9/ ,
) -36/ This is thq sum of the mean values for both injection and suction lines. Karl N. Fleming, et al., Seabrook Station Risk Management and Emergency Planning Study, Pickard, Lowe &
Garrick, Inc., PLG-0432, prepared for New Hampshire Yankee Division, Public Service Company ' of New Hampshire, December 1985, pages 3-20 and 3-22. ,
) J,7)
PLG asserts that " procedure enhancements" result in- the reduction in frequency from 6.5 x 10-6 to 5.9 x 10-7. 8,33, Karl N. Fleming, et al. , Seabrook Station Risk Management and Emergency Planning Study, Pickard, Lowe & Garrick, Inc., PLG-F 0432, prepared for New Hampshire Yankee Division, Public
) Service Company of New Hampshire, December 1985, page 2-12.
38/
~"~
We note here that Brookhaven National Laboratory's (BNL) draft review of the PLG human reliability analysis for the Event V scenario concluded that the analysis presented in PLG-0432 was
" superficial at best". See K. Bandyopadhyay, et al.,
Technical Evaluation of the S T, Sensitivity Study for Seabrook,
) Brookhaven National Laboratory, Technical Report A-3852, draft, 5 December 1986, page 2-15. We would agree with BNL that a detailed human reliability analysis should be performed. BNL did not perform such analysis, and we did not have the time or access to the plant, draft emergency procedures, etc., that BNL enjoyed in order to perform such an analysis ourselves. We
) have, therefore, presented an alternative assessment. 8,ee,,
below.
~39/ A.D. Swain & H.E. Guttmann, Handbook of Human Reliability Analysis with Emphasis on Nuclear Power Plant Apolicataons:
Final Report, Sandia National Laboratories, NUREG/CR-; . 27 8 ,
) SAND 80-0200, August 1983, pages 17-14 to 17-15.
)
o l
h - .
i
- y '
If the operators prematurely terminate infection; ;the benign i
) Event V scenario described in PLG-0432 then reverts ..to the high consequence event previously analyzed in other PRAs.. $ An alternative evaluation of the conditional probability of misdiagnosis f
l of the Event V initiator, based on consideration of the entire crew and considering the high stress level likely to be associated with such an event, is 0.04. & This is equivalent to assuining that the
- success rate is 964, or one misdiagnosis in 25 attempts. This is a
)
. factor of six increase in the HIP employed in th[e ' PLG analysis..
, Multiplying the alternative crew misdiagnosis HEP \ by[ the Event V initia ting event frequency yields a frequencf of 3,0 r 10 ~7 for an unmitigated Event V with a release point which is not sukHserged. g I
10/ By PLG's analysis, a large source te're is released to .the RNR
)' vault in even the RHR seal failure V sequence. The PLG-RNR
- vault release is estimated to include' 93% of the noble gases, 33% of the cesium-iodide, 20% of the cesium hydroxide, 204 of.
the tellurium, 1.5% of the strontium, and 2.9% of the ruthenium radionuclide groups. See PLG slides used at meeting with NBC in Be thesda, Maryland,"XTg,ust 6, 1906.
41 / This is based on a threat stress level of 0.25 (assumed not to abate because the situation will get worse over time), and a high level of dependence among the other three members of the operating crew (two operators, the shift supervisor, and the shif t technical advisor) . Thus, the joint HEP for the crew is 0.25 x 0.55 x 0.55 x 0.55 = .0.0416. A.D. Swain & H.E.
) Guttmann, Handbook of Human Reliab:,lity Analysis with Emohasis on Nuclear Power Plant ADDlicat.,ons: Finah Report, Bandia National Laboratories, NUREG/CR-1275, 5AND50-C200, August 1903, Chapter 17, generally. _
42/ In an unmitigated v sequence, it is expected that core melt -
)- would ensure relatively quickly. For example, calculaf.lons by Battelle Columbus Laboratories for a y seqence s.t surry j indicate that core melt begins in under 40 minutes, and is i complete at about one hour. See J.A. Gieseke, et al.,
( Radionuclide Release Under Speci @ , LWR Accident Conditions,
, Battelle Columbus Laboratories, BMI-2104, Vol. V, "PWR-Large, Dry Containment Design (Surry Plant Recalcalations)," July 1984, page 6-74.
i l
) -
This is a factor of ten greater than the value derived in PLG-0465, but is more in line with Event V frequencies for other similar plants
) (Indian Point, Zion, Millstone 3, etc.). 43/ 44/
In addition, an operator error in terminating injection should be assessed. The potential for such an error was identified in the 1 Lawrence Livermore National Laboratory review of the Millstone Unit 3 PRA study, but no quantification was attempted in that review. 45/
Even if the operators correctly diagnose Event V, there is a potential
) for the operators to termina te or significantly reduce injection to conserve RWST inventory in an attempt to increase the amount of time 7 BNL derived a much higher frequency, based on its analysis of 43/
l Event V scenarios for Seabrook -- i.e., 1.37 x 10-6.
Bandyopadhyay, et al., Technical Evaluation of the 375 ES , K.
l Sensitivity Study for Seabrook, Brookhaven National Laboratory, l Technical Report A-3852, draft, 5 December 1986, pages 2-17 and j 2-40. Compare ' the BNL result with the the PLG results, which total 4.57 x 10-8. See Karl W. Fleming, et al., See prook Station Risk ManagementTn,d Einergency Planning Study, Pictard, Lowe 6 Garrick, Inc., PLG-0432, prepared for New Hampshire Yankee Division, Public Service Company of New Hampshire, December 1985, page 3-101.
44/ We note that BNL has performed a preliminary review of the Event V sequence. BNL has also derived a much larger frequency of occurrence for Event V than did PLG. The total frequency of Event V damage states involving containment bypass was estimated by BNL to be 1.2 x 10-6 per reactor-year. See, K.
Bandyopadhyay, et al., Technical Evaluation of the EP3
) Sensitivity Study for Seabrook, Brookhaven National Laboratory, Technical Report A-3852, draf t, 5 December 1986, pages 2-17 and 2-40. By contrast, PLG calculated a frequency of about 4.0 x 10-8 per reactor year. See, Karl N. Fleming, et al., Seabrook Station Risk Management and Emergency Planning Study, Pickard, Lowe 6 Garrick, Inc., prepared for New Hampshire Yankee Division, Public Service Company of New Hampshire, December
}
, 1985, page 3-101. BNL's value is a factor of about 30 greater than PLG's.
45/
~
A.A. Carcia, et al., A Review of the Millstone 3 Probabilistic Safety Study, Lawrence Livermore National Laboratory, NUREG/CR-4142, UCID-20330, April 1986, page 3-133.
)
)
f
),
available to take mitigating actions. 46,/ No assessment is presented
) here due to insufficient review time, but this crew error should be examined an,d quantified.
4 i
The scarce term for an unmitigated (i.e. , no " pool scrubbing" -
/
) 4
- the refease point is not covered by water) Event V is addressed in PLG-0465. This source term can be compared with source terms for an W
analogous sequence at the Surry plant (calculated using state-of-the-
) art, NRC-sponsored " Source Term Code Package") --
see, Figure 2. A 1 '
" WASH-1400 method" source term calculation for Seabrook is also j
i presented. 47/
Upon examining the PLG-0465 discussion of the Event V sequence phenomenology and source term assessment, it is not evident that the 3 ,
analysis considered the fate of hydrogen released into the auxiliary 33 building. There would seem to be a considerable potential for this .
hydrogen t'o burn or detonate in the auxiliary building. It is likely that such an occurrence would severely damage the auxiliary building.
Thus, we consider it unlikely that the fission product ; attenuation
) assumed in the PLG analysis would in factionsue. In addition, the PLG analysis appears to be based on an assumption that the RNR failure is a pump seal failure (PLG assessed the conditional probability of pipe rupture at 6 x 10~3). 48/ As discussed above, we find this analysis
~~46/ In an Event V sequence, the onset of core melt is directly related to the amount of time that injection can be maintained by drawing suction on the RWST. The longer the RWST inventory
). , lasts, the later in time core melt might occur. Moreover, if s RWST inventory can be replenished, core melt could theoretically be avoided entirely.
E/ This is from the original Seabrook PRA.
) 48/ Karl N. Fleming, et al., Seabrook Station Risk Management and Emergency Planning Study, Pickard, Lowe & Garrick, Inc., PLG-4
)
is poorly justified, and do not consider the probability estimate
)
credible. 49/ In the absence of a convincing analysis, we see little reason for making an assumption other than that pipe rupture will not occur with a.very high likelihood.
) Thus, we believe that the NUREG-0956 calculation of the Event V source term for Surry, being based on pipe rupture and minimal attenuation in the safeguards building, is more representative of the
) release that could occur at Seabrook in an unmitigated Event V sequence, and recommend its use in the absense of a credible plant-specific analysis. Certainly, as a minimum, such a source term (or
) one calculated specifically for Seabrook) should be considered as part of the uncertainty analysis for the V sequence. We also believe that the frequency for the V sequence is significantly higher than
) estimated by PLG. An alternative estima te of the frequency of 'an 0432, prepared for New Hampshire Yankee Division,. Public Service Company of New Hampshire, December 1985, page 3-30.
-49/ PLG references a draft NRC case study by the Office for the Analysis and Evaluation of Operational Data (AEOD) of BWR
). overpressurization incidents. The final version of that study became available in September 1985. See, _ Peter Lam, "Overpressurization of Emergency Core Cooling Systems in Boiling Water Reactors," U.S. NRC, AEOD Case Study Report AEOD/C502, September 1985. This study, unlike the draf t, did consider the dynamic effects of the overpressurization of BWR
) ECCS systems. The final version of the report assessed a pipe rupture probability of 0.01 to 0.001 for piping pressurized to twice its design pressure. The study also made reference to a separate CRC analysis which assessed a probability of 0.1 for the same conditions. See, AEOD/C502, pages 24-26. It should be noted that in the PWR case, the piping will be pressurized
) to more than thrice its design pressure. In addition, an Oak Ridge National Laboratory study indicates that under Event V conditions, water hammer and pipe whip rupture may be more significant sources of pipe rupture than simple overpressuriza tion, however this study did not develop a pipe rupture probability estimate. See, J.D. Harris & J.W.
Minarick, An Evaluation of BWR Over-Pressure Incidents in Low
} Pressure Systems, Oak Ridge National Laboratory, preliminary draft, May 1985, Section 7.
~
)'
~
unmitigated V sequence is 3.0 x 10 . An uncertainty assessment for the V sequence should consider the possibility that the sequence frequency is equal to the intiating event frequency -- 7.6 x 10-6, y, recognize that this is a conservative assumption, but it is in line
)
with the assumption made in many other PRA studies. 50/ 51/ '
III.3 Steam Generator Tube Rupture Scenarios
)
The rupture of one or more steam generator tubes constitutes a loss-of-coolant-accident. More specifically, since coolant is lost
) from the reactor coolant system to the secondary side of the plant in such a scenario, steam generator tube rupture (SGTR) swquences represent a loss of coolant outside containment. 52_/ The coolant
~
) .
5_Of one final aspect of the Event V scenario needs to be mentioned -
here. It has been suggested that it may be possible to
- initia te an Event V sequence as the result of a fire, an electrical short circuit, or testing involving jumpers. See, internal NRC memorandum dated October 6, 1986, from Warren Lyon to Carl Berlinger,
Subject:
" Review of Seabrook Documents
) Pertaining to Change in Emergency Planning Zone Size,"
Enclosure 1, " Review Comments on 'Seabrook Station Risk Management and Emergency Planning Study' (Pickard, Lowe and Garrick, Inc., PLG-0432, December 1985)," page 5. Such initiating modes for interfacing LOCAs have not generally been considered in previous PRAs. This is likely due to the fact
) that such failure modes would be less likely than the normal modes considered in V sequence analyses.
Sl/ A cable-spreading room fire initiating event is given a frequency of about 5.2 x 10-7 in the Seabrook PRA. See, Karl N. Fleming, et 1., Seabrook Station Probabilistic Safety
) Assessment, Pickard, Lowe & Garrick, Inc., PLG-0300, prepared for Public Service Company of New Hampshire and Yankee Atomic Electric Company, December 1983, page 5.1-2. This is one example of a fire which should be carefully examined to ascertain if it is possible to cause interfacing valves to open.
) There have been five historical SGTR incidents. These occurred
-52/
at Point Beach Unit 1 (2/26/75), Surry Unit 2 (9/15/76), Doel
) lost to the secondary side of the steam generator can pressurize the steam side of the plant to a pressure sufficient to lift the main steam safety valves or atmospheric relief valves. Since these valves relieve pressure directly to the atmosphere, a containment bypass pathway is created.
) The Seabrook PRA displays a matrix of initiating events and release category (i.e., source term) frequencies. This matrix suggests that steam generator tube rupture (SGTR) sequences have been 3 misclassified in the Seabrook PRA. The matrix shows essentially a zero frequency for containment bypass resulting from SGTR sequences.
53/ Another table shows a combined frequency of less than 1x 10 -8
} for SGTR sequences with nonisolated steam generators (i.e., with a steam leak to the environment) . 54/ 55/
Compared with review of Millstone 3 PRA by Lawrence Livermore g National Laboratory . (LLNL) , this result is so low as to be incredib,le.
In contrast with the Seabrook results, LLNL estimates a frequency of 2
~
J Unit 2 (Belgium, 6/25/79), Prairie Island Unit 1 (10/2/79), and R.E. Ginna (1/25/82). See, J.E. Newell, CEGB Proof of Evidence On: Steam Generator Tube Integrity, Sizewell 'B' Public Inquiry, CEGB P15, November 1982, page 9.
J. 53/ Karl N. Fleming, et 1., Seabrook Station Probabilistic Safety Assessment, Pickard, Lowe & Garrick, Inc., PLG-0300, prepared for Public Service Company of New Hampshire and Yankee Atomic Electric Company, December 1983, page 2.3-10.
54/ Id., page 5.1-8.
J An NRC staf f reviewer noted that the PRA (incorrectly) did not
--55/
recognize SGTRs without a stuck-open main steam relief valve as a bypass pathway. See, internal NRC memorandum dated October 8, 1986, from Warren Lyon to Carl Berlinger,
Subject:
" Review of Seabrook Documents Pertaining to Change in Emergency Planning Zone Size," page 2.
J
)
-6 x 10 for SGTR sequences with steam leak for Millstone 3. 56/ 57/
) Considering that Seabrook and Millstone are relatively similar, it is very unlikely that Seabrook SGTL sequences with steam leak to the atmosphere are a factor of 100 lower in likelihood at Seabrook.
) The Seabrook PRA estimated that there is a contribution to core melt frequency for SGTR-initiatted sequences of 1.7 x 10 -6 & These sequences will cause releases :o the environment via periodic lifting
) of the main steam safety valvos. The resultant source term should be similar to that derived by Battelle for the induced tube rupture scenario analyzed for NUREG-1150. (See,Section III.4, below) .
)
56/ A.A. Garcia, ec al., A Review of the Millstone 3 Probabilistic Safety Study, Lawrence Livermore National Laboratory, NUREG/CR-4142, UCID-20330, April 1986, page 5-3.
~
57/ Results similar to the Millstone 3 results were obtained in the Indian Point. PRA, the NRC staff review of the Indian Point PRA, and an NRC staff generic study of SGTR sequences. The NRC staff rev!.ew of the Indian Point PRA suggasted that the frequency of core melt from SGTR sequences was about 4 x 10-6.
) See, Testimony of Frank Rowsome and Gary Holahan, Indian Point Special Proceeding. The Indian Point PRA estimated the frequency of SGTR core melts at 2 x 10-6. See, PAGNY and Consolidated Edison, Indian Point Probabilistic Safety Study, March 1382, revised 1983. The NRC generic SGTR study estimated the core melt frequency for single SGTR sequences at about 5x See, A. Akstulewitz, et al., NRC Integrated Program for
) 10-6.
the Resolution of Unresolved Safety Issues A-3, A-4, and A-5 Regarding Steam Generator Tube Integrity, U.S. NRC, NUREG-0844, draft for comment, April 1985.
58/ Karl N. Fleming, et 1., Seabrook S ta tion Probabilistic Safety Asr.e s sme n t , Pickard, Lowe & Garrick, Inc., PLG-0300, prepared
) fer Public Service Company of New Hampshire and Yankee Atomic Electric Company, December 1983, page 2.3-10.
)
1 1
) 1 III.4 Induced Steam Generator Tube Ruptures
)
Induced steam generator tube rupture refers to an SGTR that develops during an accident without SGTR being the initiating event.
In other word's, the SGTR in the induced tube rupture sequences is a
" consequential" failure. This failure could result from a variety of factors, including pre-existing tube defects, pressure transients in the reactor coolant system while the secondary side of the tubes is
) depressurized, fission product deposition and heating of the SG tubes, etc.
The Seabrook studies and the BNL review give induced SGTRs a very low likelihcod. In contrast, recent Sandia National Laboratories containment event tree studies in support of the forthcoming NUREG-1150 study indicate that for some sequences induced tube ruptures may
) be relatively likely. The S'andia study ,rovides p containment event tree branch probabilities for three passes through the containment event tree (CET) , referred to as optimistic, central, and
) pessismistic. 59/
-59/ Sandia explains the derivation of its results as follows: "We have attempted to define the three sets of values in such a way tha the central estimate represents the median o fhte reactor 3 safety community on any particular issue. That is, if a large number of experts were to be polled with regard to their views, a substantial fraction would respond that the actual situation would be better than that indicated by the central estimata, and an approximately equal number would respond that it is worse. Of those who responded that the situation is actually better than that represented by the central estimate, we would expect that a substantial number would view our optimistic estimate as an appropriate representation of the issue; likewise, those who viewed the situation as being worse thatn the central estimate implies would agree that reality is or could be as unfavorable as that reflected by our pessimistic estimate." Sandia also stated, "In this report we do not -
.) propose weighting factors or distributions for these estimates, J
)
For some sequences, the central and pessimistic CET
) probabilities for induced tube ruptures are non-negligible. For example, for a station blackout sequence (including loss of auxiliary
-3 f eedwa te r) , the central estimate for Surry is that there is a 1 x 10
) probability of an induced tube rupture, while the pessimistic estimate is that the probability is 1 x 10 -1 60/ 61/
A recent source term analysis for a 4-loop Westinghouse PWR
) NSSS for an induced SGTR sequence (designated TMLU-SGTR, resulting from a loss of feedwater transient) using state-of-the-art NRC source term codes (i.e., the Source Term Code Package) calculated a source
) term of 38% of the noble gases, 14.5% of the iodine, 14.3% of the cesium, 9.2% of the tellurium, and 0.2% of the barium groups. This study stated that these results should be applicable to other PWRs of
}- .
nor purpose that one is more reasonable than another. We present these estimates as a reflection of the information that is available." See, A.S. Benjamin, et al., Containment Event
) Analysis for Postulated Severe Accidents: Surry Power Station, Unit 1, Sandia National Laboratories, NUREG/CR-4700, SAND 86-1135, Vol. 1, Draft report for Comment, November 1986, pages 7 and 8.
-60/ A.S. Benjamin, et al., Containment Event Analysis for
] Postulated Severe Accidents: Surry Power Station, Unit 1, Sandia National Laboratories, NUREG/CR-4700, SAND 86-1135, Vol.
1, Draft report for Comment, November 1986, page 45. Similar results are presented in a published conference paper; see, A.S. Benjamin, et al., " Evaluation of Uncertainties in Risk Emanating from a Number of Severe Accident Issues," Sandia National Laboratories, SAND 86-1225C, presented at the 3rd Containment Integrity Workshop.
-61/ BNL reports similar NRC staff estimates, ranging 0.01 to 0.3.
BNL used a value of 0.2 given failure to depressurize the
. reactor coolant system. See, K. Bandyopadhyay, et al.,
Technical Evaluation of the EPZ Sensitivity Study for Seabrook,
) Brookhaven National Laboratory, Technical Report A-3852, draft, 5 December 1986, page 2-24.
J i
J 3 similar primary system design.- 62/ It would appear that the Seabrook PRA would " bin" this sequence into either a no containment failure or a late containment failure source term. This would be incorrect, and would result in a misleading dose vs. distance relationship for severe accidents at Seabrook since these two containment failure modes would be expected to be associated with relatively small source terms in
, contrast with the SGTR source term described above.
J There are a large number of Seabrook PRA accident sequences which turn out to be station blackout or station blackout analogues in the context of an assessment of the likelihood of induced tube 3
ruptures. For example, five sequences in the top twenty that are essentially station blackout sequences. 62 / In addition, there are other sequences which are analogous (no reactor coolant system heat 3 removal or injection capability). 6_4/
There are two types of station blackout analogues. In the first type, auxiliary feedwater fails early, leading to early core J melt due to boil-off of the RCS inventory. In the second type, auxiliary feedwater fails late (due to DC battery depletion) , leading to a much later core melt. The sum of the frequency of Seabrook O
--62/ R.S. Denning, et al., Radicnuclide Release Calcula tions for Selected Severe Accident Scenarios: PWR, Ice Condenser Design, Battelle's Columbus Division, NUREG/CR-4624, BMI-2139, Vol. 2, July 1986, pages 5-42 and 6-2. The sequence analyzed was a complate loss of feedwater, failure of high pressure injection, g and a consequential steam generator tube rupture occurring at the time of core slump.
-63/ These are sequences 1, 2, 8, 9, and 18, with a cumula'tive frequency of 5.33 x 10-5 per reactor year.
, 64f These are_gequences 5 and 13, with an additional frequency of 1.15 x 10 per reactor year. Sequence #5 in the Seabrook PRA also involves scram failure -
i.e., it is an ATWS sequence involving conditions equivalent to station blackout.
D
) .
accident sequences which are early melt station blackout analogues is
) 2. 7 . x 10-5, Qf This represents about 12% of overall core melt frequency. 66/ The frequency of Seabrook accident sequences which are late melt station blackout analogues is 8.0 x 10-5 . This is about
] 35% of overall core melt frequency. The principal difference in the early and late melt sequences should be the timing of the release and some reduction in source term due to radioactive decay.
3- Taken together, station blackout analogues (i.e., sequences with injection failure and lack of secondary side heat r emoval) ,
including early and late melt variations, have a frequency of about
~ 1.1 x 10-4, or about 47% of overall core. melt frequency. Of this,
-)
about 9% is associated with large seismic events with implications for impaired offsite emergency response.
I f. ten percent of these station. blackout analogue sequences involve an induced tube, rupture (as postulated in the Sandia pessimistic containment event tree quantification for NUREG-ll50),
then the frequency of sequences which should be binned into an induced O
- tube rupture source term category is about 1.1 x 10-5, or about 5% of Q/ BNL reports that PSNH estimated the frequency of "high pressure sequences" at 4 x 10-5. See, K. Bandyopadhyay, et al.,
g Technical Evaluation of the EP U ensitivity Study for Seabrook, Brookhaven National Laboratory, Technical Report A-3852, draft, 5 December 1986, page 2-24.
66/ It should be noted that about 33% of the 2.7 x 10 -5 frequency of early melt station blackout analogues cited a ve is 5
g associated with seismic events (20% of the 2.7 x 10 frequency is associated with seismic initiating events with ground accelerations of 0.7g to 1. 0g) . These seismic sequences will frequently be associated with loss of offsite power, meaning that the siren alerting system will not function. Furthe r, it is likely that emergency communications will be disrupted due to earthquake-associated damage., *These and other offsite O impacts due to seismic events pose unusual emergency response problems (see,Section III.5, below) .
Q y~-.- -,-,_e-, .
.,,,,%.,,. ,,__,,,y..
h )
overall core melt frequency. Another way of interpreting this result
) !
is that about one core melt accident in twenty will be associated with an induced tube rupture source term. g/
It should be observed that all of the core melt frequency associated with station blackout analogues involves high pressure core melt sequences. Thus, for these sequences, as well as others, high pressure melt ejection and direct containment heating are a likely
)
result. (See, Section III.6, below, for a discussion of these issues.)
Even for sequences with induced tube ruptures, it is plausible
) that RCS pressure at the time of vessel failure will be sufficiently high as to present the risk of high pressure melt ejection and direct ,
containment heating. 8 6_8/ Should high pressure melt ejection and I
67/ 'This presu'mes that the " late melt" scenarios identified above are subject to induced tube ruptures. Sandia National Laboratories' recent containment event. tree analysis for Surry ,
binned the earlye melt and late melt station blackout sequences into the same plant damage state. See, A.S. Benjamin, et al.,
) Containment Event Analysis for Postulated Severe Accidents:
Surry Power Station, Unit 1, Sandia National Laboratories, NUREG/CR-4700, SAND 96-ll35, Vol. 1, draft report for comment, November 1986, page 16. It is plausible that this is not correct, but we believe it is. Assuming that the " late melt" blackout analogues do not apply, the frequency of induced SGTRs
) would be 2.7 x 10-6, rather than 1.1 x 10-5 as identified above. This would represent about 1.2% of the overall core melt frequency, or about one accident in 85, rather than one in twenty. The alternative estimate of 2.7 x 10-6 calculated in this footnote is very close to BNL's upper range estimate.
See, K. Bandyopadhyay, et al., Technical Evaluation of the EPZ
) Sensitivity Study for Seabrook, Brookhaven National Laboratory, Technical Report A-3852, draft, 5 December 1986, page 2-27.
6_8/ The Battelle Columbus Laboratories analysis of the induced tube rupture sequence estimated the RCS pressure at the time of bottom head failure to be 1900 psia. The RCS pressure throughout the sequence is estimated to remain above 1500 psia.
,) These results were obtained for an assumed rupture of five steam generator tubes. See, R.S. Denning, et al. , Radionuclide Release Calculations fo r Selected Severe Accident Scenarios:
)
-en--- ,--n ,,-nwwan- -,e-av- - , - - - - - - _ , - , - , , - - - - v----. - - e~n- --. --~,,,----,-_a - - - - - - - - - - - - - - - - - - - - - ,
' ^
I
. \
)
direct containment heating result in containment failure for induced
) tube rupture sequences, the source term associated with containment failure would be additive to the induced tube rupture source term (of course, one would have to account for the release which has already occurred through the tube rupture when estimating the HPME/DCH source term). The potential would exist under these circumstances for a very much larger source term than considered by PLG for those accidents
) which PLG would otherwise consider the containment to either fail very late or not fail at all.
Seismic Sequences
) III.5 The total contribution to core melt frequency from accidents initiated' by seismic events is estimated to be 2.9 x 10-5, Qj og
) '
this contribution, 1.4 x 10 -5 of the core melt frequency arises from very large ground accelerations of 0.7g to 1.0g. 3/ Thus, about 6%
of overall core melt frequency and about 48% of all seismically-initiated core melt accidents will be associated with large ground accelerations.
Such large ground accelerations ~have the potential to
)
substantially interfere with offsite emergency response. First, at PWR, Ice Condenser Design, Battelle's Columbus Division, NUREG/CR-4624, BMI-2139, Vol. 2, pages 4-6 and 4-62.
69/ Karl N. Fleming, et al., Seabrook Station Probabilistic Safety Assessment, Pickard, Lowe & Garrick, Inc., PLG-0300, prepared for Public Service Company of New Hampshire and Yankee Atomic Electric Company, December 1983, page 2.3-10.
70/ Karl N. Fleming, et al., Seabrook Station Probabilistic Safety Assessment, Pickard, Lowe & Garrick, Inc., PLG-0300, prepared
) for Public Service Company of New Hampshire and Yankee Atomic Electric Company, December 1983, page 9.2-24.
)
) such large ground acceleration values, area-wide loss of AC power is very likely. This will result in the failure of the siren alerting 3 system installed around Seabrook; the system is intended to be used to alert the population offsite to the need to listen to radio or television to receive emergency information in the event of an
) accident at Seabrook.
Second, communications are likely to 'be disrupted. Other potential impacts include significant reduction in sheltering
) capabilities due to structural damage to homes and public buildings and possible physical damage to evacuation routes. H/ It should also be recognized that these ef fects will have a significant impact
) on the potential for expanding emergency response efforts outside the
'EPZ.
In addition to impacting emergency response, seismic accident sequences associated with large ground accelerations might'be expdcted
)
to be accompanied by a relatively large aftershock within a~few hours after the initial seismic event. The combined loading of pressure from a seismically-initiated core melt accident in progress and a seismic loading from an aftershock should be investigated as a possible source of early containment failure. H/ The potential for
)
71/ In past assessments of risk,
) -
practice of assuming very the NRC staff has adopted a poor emergency response characteristics for accidents initiated by large earthquakes (referred to by the NRC staff as " area-wide disasters").
72/ Such an investigation could not be carried .out in the limited time avairable for this report.
)
3
1 1
)-
this type of failure mode has been raised by other studies, H/ but
) has not been included in any published PRA to date. H/
Finally, seismic PRA studies usually do not make allowances for degraded operator performance in' accident sequences initiated by
) earthquakes. Preliminary research carried out for NRC indicates that 73/
See, for example: letter dated 2 April 1982 from J.W. Reed (Jack R. Benjamin & Associates, Inc.) to W.A. von Riesemann (Sandia National Laboratories,
Subject:
" Proposed Research Program to Improve the Seismic Analysis Procedures for
) Probabilistic Risk Assessment of Nuclear Power Plants"; R.D.
Campbell & D.A. Wesley, Potential Seismic Structural Failure Modes Associated with the Zion Nuclear Power Plant: Seismic Safety Margins Research Program, Engineering Decision Analysis Company, Inc., prepared for Lawrence Livermore National Laboratory, NUREG/CR-1704, UCRL-15140, March 1981; D.A.. Wesley,
) et al., Conditional Probabilities of Seismic ' Induced Failures for Structures and Components for the Zion Nuclear Generating Station, Structural Mechanics Associates, prepared for Pickard, Lowe & Garrick, Inc., SMA-12901.02, October 1980; J.B. Rivard, et al., Identification of Severe Accident Uncertainties, Sandia National Laboratories, NUREG/CR-3440, SAND 83-1689, September
) 1984. It should be noted that J.W. Reed (Jack R. Benjamin &
Associates, Inc.) has been an NRC staff consultant on the review of seismic risk assessments, and participated in the review of 'the seismic risk assessments for Seabrook, Indian Point, and Millstone 3. Structural Mechanics Associates have performed seismic fragility analyses for many nuclear power plant PRA studies (under contract to Pickard, Lowe & Garrick,
) Inc., including Seabrook, Indian Point, Millstone 3, and Midland.
H/ A study for Maine Yankee for design basis accident pressure loadings and seismic events did report a considerable reduction in seismic fragility. In the absence of any accident-related
)- pressure loading, the fragility of the Maine Yankee containment was estimated to be 2.lg. With the design-basis accident pressure of 55 psi present, the fragility was reduced to 1.2g.
The reported results were characterized as representing high confidence in low probability of failure, and are considered by the authors of the report to be conservative values. See, P.S.
) Hashimoto, et al. , Conservative Seismic Capacities of the Maine Yankee Roactor Containment Including and Excluding Design Incident Pressure, Structural Mechanics Associates, Inc.,
prepared for Maine Yankee Atomic Power Company, NTS/SMA-36001.01, December 1984. The calculated reduction in fragility appears to be quite substantial. Unfortunately, no results
) were reported for the higher pressures which would be typical of core melt accidents in progress.
)
O performance under severe stress brought on by earthquakes results in high error rates, difficulty in revising an initial perception of an O
event, part task fixation (" tunnel vision"), and a tendency to repeat habitual responses. It was concluded further that these general findings to not provide a basis for assuming that reactor operators
.O are different in this regard. M/ The implications of these variations in human performance may not have been reflected in the Seabrook PRA seismic analysis. Detailed review, which was not O possible in the short time frame available for preparation of this report, would be required to ascertain the impact of these human performance factors on seismic risk.
40 III.6 High Pressure Core Melt Sequence Modeling O The Seabrook PRA studies model high pressure core melt sequence phenomenology in what can only be characterized as an optimistic fashion. Recent PRAs, including Seabrook, have indicated that rather
- O than failing the entire bottom head, molten core material is more likely to cause a localized failure. In particular, the in-core instrument tubes which penetrate the bottom head of the reactor vessel have only a small circumferential weld just inside the vessel wall which is postulated to fail under molten core thermal loadings.
- j. This failure mode opens up a rela tively small opening in the l bottom head. Sandia National Laboratories experiments have indicated
!O .
~
7s/ James P. Jenkins & Karl R. Goller, " Reactor Operator Performance Under the Stress of a Seismic Event," U.S. NRC, presenta tion to the Seismic Risk and Heavy Industrial O Facilities Conference, San Francisco, California, May 12, 1983.
O
r- - - -
)'
that for sequences in which the reactor coolant system pressure is more than a few hundred psia, melt will be forcibly ejected from the reactor vessel, a mechanism referred to as high pressure melt ejection (HPME).
) The depiction of HMPE, as modeled in PLG PRAs, is described by Sandia National Laboratories as follows: ". (T) he discharged core material is assumed to flow out o fhte cavity and spread evenly over j- the floor of the contaiament building, forming a shallow debris bed that can be easily quenched by existing spray systems. The relocation of the debris from the cavity is predicted to occur over a wide pressure range, typically 1.4 to 17.1 MPa. Because a protracted core
]
debris-concrete interaction is avoided, it is predicted that the accident promptly terminates without significant gas or aerosol
) generation or added pressurization of the containment building." H/
PLG PRAs, including Seabrook, credit this mechanism in their containment analyses, and this is a key reason why much of the accident sequence frequency in the Seabrook PRA -and other PWR PRAs winds up getting binned to a "no containment failure" or " late containment failure" plant damage state. Such plant damage states are typically associated with low source terms -- in effect, rather benign J accidents from a public exposure standpoint. E/
~
76/ William W. Tarbell, Marty Pilch, & John E. Brockmann, " Melt Expulsion and Direct Containment Heating in Realistic Plant Geometries," Sandia National Laboratories, SAND 85-1631C, in
]
- Proceedings of the International ANS/ ENS Topical Meeting on Thermal Reactor Safety, American Nuclear Society, San Diego, California, 2/2-6/86, Vol. 1, page II.6-1.
H/ It is clear from a recent NRC meeting transcript that PLG analysts (specifically, Fauske & Associates) are persuaded by a
.) series of HPME experiments involving Wood's metal that HPME is a non-issue. Wood's metal is a mix of 12.5% tin, 12.5% '
D
B In contrast, Sandia researchers have demonstrated through B
analysis and experiments that the PLG analyses may be incorrect.
Although the Sandia experiments confirm that the core debris is expelled from the reactor vessel and dispersed out of the reactor D cavity region, the expulsion process is not a benign " film-like" flow of material, but is rather a "very energetic discharge of melt particles distributed in a high velocity gas." 78/ A number of 3 significant weaknesses in the PLG analysis of high pressure melt ejection have been noted by researchers at Sandia National Laboratories. In particular, the PLG analyses fail to consider the 3 presence of gases in solution in the melt and the production of aarosols by tha malt ajection process. y Sandia research has cadmium, 25% lead, and 50% bismuth. Wood's metal has a melting point of about 73 degrees C (150-160 degrees F). Regardless of the ejection pressures used, regardless of the scale of the
') test, and regardless of the geometry of the cavity into which it is ejected, Wood's metal will promptly solidify on any surface whi.ch it touches unless that surface is hot. In our '
view, the results of the Wood's metal HPME tests - i.e. , that large fractions of the ejecta are retained in the cavity area -
are no more than a reflection of the basic properties of 3 Wood's metal, and have little or nothing to do with a realistic simulation of reactor accident conditions. These test results are therefore meaningless, and prove nothing whatsoever concerning HPME phenomena.
78/
-~
William W. Tarbell, Marty Pilch, & John E. Brockmann, '" Melt
] Expulsion and Direct Containment Heating in Realistic Plant Geometries," Sandia National Laboratories, SAND 85-1631C, in Proceedings of the International ANS/ ENS Topical Meeting on -
Thermal Reactor Safety, American Nuclear Society, San Diego, California, 2/2-6/86, Vol. 1, page II.6-1.
3
- 79/
~~
Indeed, based on the results of an early series of HPME tests (referred to as the System Pressure Ejection Tests or SPIT),
Sandia concluded, "The results from the Phase I SPIT tests ...
have shown the behavior of a molten stream ejected at high pressure to be uncharacteristic of that hypothesized in the ZPSS [ Zion Probabilistic Safety Study). ... The tests have demonstrated an unexpected aerosol source term accompanying the O jet propogation, assumed to be caused by three or more separate mechanisms." W. Tarbell, J. Brockmann, & M. Pilch, High-
D indicated conclusively that the benign termination postulated by PLG D analyses does not occur; ra ther , " highly fragmented debris may be ejected into the containment a tmosph'ere where thermal and chemical energy is liberated rapidly from the debris. " 8_0/ HPME experiments have been found by numerous experiments at Sandia to create a dense aerosol cloud which is capable of being forced into the upper containment through a variety of pathways. The hot aerosol particles oxidize, serve as distributed ignition sources for hydrogen in the containment, and result in a rapid pressurization, temperature increase, and production of prodigious quantities of radioactive aerosols. The pressurization of the containment results from what is D referred to as direct containment heating (DCH).
HPME/DCH has been identified in the NUREG-ll50 analyses as a potential containment failure mode for the Surry and Zion . nuclear power plant containments. Since the Seabrook co,ntainment is slightly smaller than Zion, and since Seabrook has a larger amount of metal in the core available to participate in HPME/DCH (by virtue of its higher 3 power level), HPME/DCH should also be an issue for Seabrook. 8J/
Moreover, Seabrook's containment is particularly susceptible to DCH Pressure Melt Streaming (HIPS) Program Plan, Sandia National
] Laboratories, NUREG/CR-3025, SAND-82-2477, August 1984, page 60.
8_0_/ M. Pilch & W.W. Tarbell, High Pressure Ejection of Melt From a Reactor Pressure Vessel: The Discharge Phase, Sandia National Laboratories, NUREG/CR-4383, SAND 85-0012, September 1985, page g 1.
8J/ We recognize that PLG and PSNH would not agree due to their assessment of the failure pressure of the Seabrook containment.
However, if BNL's assessment is used (and we believe BNL's assessment is more consistent with the s tate-o f-the-ar t) , the failure pressure of Seabrook is not markedly different from O Zion (i.e., 150 psia for Zion and 172 psia for Seabrook).
4
- p. due to the fact that the intermediate floors in the Seabrook containment are largely grated rather than concrete in construction.
I This will allow good mixing paths for the containment for all regions of the structure. 82/
The Seabrook PRA defines high pressure sequences as transients B and small LOCAs. 83/ Employing this ?imited definition and ,
considering only the twenty most likely sequences, high pressure sequences in the top twenty are sequences numbers 1, 2, 4, 5, and 7
]) through 20. These sequences account for a core melt frequency of 1.04
~4 x 10 just among the top twenty sequences, and represent 45.2% of the overall core melt frequency. Thus, at least 45.2% of the core melt frequency may be associated with HPME/DCH conditions.
3 Alternatively, the Seabrook PRA identifies the plant damage states with high vessel pressure at the time of vessel failure. 84/
These plant damage states represent 98% of the core melt frequency.
85/ Thus, HPME/DCH will be very important to calculated source terms unless the conditional failure probability is very low.
3 82/ Karl N. Fleming, et al., Seabrook Station Probabilistic Safety Assessment, Pickard, Lowe & Garrick, Inc., PLG-0300, prepared for Public Service Company of New Hampshire and Yankee Atomic Electric Company, December 1983, pages 3.2-1 to 3.2-2.
83/
~~
Karl N. Fleming, et al., Seabrook Stat' ion Probabilistic Safety
] Assessment, Pickard, Lowe & Garrick, Inc., PLG-0300, prepared for Public. Service Company of New Hampshire and Yankee Atomic Electric Company, December 1983, page 11.4-4.
84/ Karl N. Fleming, et al., Seabrook Station Probabilistic Safety Assessment, Pickard, Lowe & Garrick, Inc., PLG-0300, prepared for Public Service Company of New Hampshire and Yankee Atomic Electric Company, December 1983, page 2.3-7.
85/ Karl N. Fleming, et al., Seabrook Station Probabilistic Safety Assessment, Pickard, Lowe & Garrick, Inc., PLG-0300, prepared for Public Service Company of New Hampshire and Yankee Atomic Electric Company, December 1983, pages 2.3-7 and 5.1-8.
O
D Sandia calculations with a preliminary model 86/ involving the Zion plant suggest that " direct heating may be a significant, and potentially dominant, contributer (sic) to containment pressurization." g In view of the general similarities between D
Zion and Seabrook (e.g., containment free volume, mass of core materials, reactor cavity geometry, and failure pressure) , it can be concluded that HPME/DCH may also present the potential for early D
containment failure for Seabrook.
Containment pressures generated by HPME/DCH can be quite high.
For example, the NRC-sponsored Containment Loads Working Group
)' estimated a pressure of 176 psig and temperature of 1340 degrees F.
for whole-core dispersal and 100% oxidation of cladding (but not steel or uranium oxide, nor considering metal steam reactions in the 3 containment) at the Surry plant. 88/ If oxidation of steel and uranium dioxide and metal / steam reactions in the containment are considered, higher pressures can be calculated.
] Similar results are shown by other analyses. Sandia calculations with the CONTAIN and DHEAT codes show pressure increments for Zion for direct heating as high as 85-88 psia for 100% direct M/ Among other things, the model does not consider hydrogen recombination from localized reactions akin to a distributed ignition. source. The aerosol particles generated by HPME can act as distributed hydrogen ignition sources and cause recombina tion or burning of hydrogen in an otherwise " steam-inerted" atmosphere.
87/ M. Pilch & W.W. Tarbell, Preliminary Calculations on Direct Heating of a Containment Atmosphere by Airborne Core Debris, Sandia National Laboratories, NUREG/CR-4455, SAND 85-2439, July 1986, page 19.
D g/ Containment Loads Working Group, Estimates of Early Containment Loads from Core Melt Accidents, U.S. NRC, draft report for comment, NUREG-1079, December 1985, page xxvii.
O
y . .
heating, but no oxidation of metal. With 90% participation in
) , oxidation of metal, pressure increments of 158-168 psia were calculated for Zion. 89/ To achieve conditions for Zion similar to the analysis .in NUREG/CR-4700 for Surry, 90% participation in
) oxidation gave an additional increment of 73-80 psia beyond simple ,
direct heating; thus 25% participation in oxidation would give a pressure increment of about 18-20 psia, for a total increment of about
) 103-108 psia above the pressure existing at - the time just before vessel failure. With even modest pressures existing just prior to vessel failure, a real threat to containment - for Zion would result.
} The calculations with DHEAT and CONTAIN assumed an initial pressure of 43 psia; this would yield total pressures of 146-151 psia. 90/ We believe that similar, but somewhat higher results, would be obtained
) for Seabrook (due to slightly smal'ler containment free volume, greater .
, thermal power level, larger mass of metal in the core, and a greater RCS water and steam volumes) . These additional increments would
)
By K.D. Bergeron & D.C. Williams, "CONTAIN Calculations of Containment Loading of Dry PWRs," Sandia National Laboratories, in NUREG/CP-0056, SAND 84-1514, Proceedings of the Second Workshop on Containment Integrity, August 1984, T. Molina & R.
Cochrell, editors, pages 115-128.
}.
90/ The CLWG standard problem for Zion provided a containment pressure for station blackout just prior to vessel failure of 0.4 MPa, or 58 psia. See, K.D. Bergeron, D.C. Williams, & P.E.
Rexroth, " Applications of the CONTAIN 1.0. Computer Code to - the Analysis of Containment Laoding Under Severe Accident
} Conditions," Sandia National Laboratories, in NUREG/CP-0058, Proceedings of the U.S. Nuclear Regulatory Commission Twelfth Water Reactor Safety Re search Information Meeting, Gaithersburg , Maryland, 10-22-26/84, Vol. 3, published January 1985, page 426. If this is realistic, direct heating (particularly considering oxidation of metal, oxidation of uranium dioxide, and metal water reactions in the containment
) a tmosphere) would appear to present a substantial threat of containment failure for Zion.
J produce pressures in the range required to cause failure of the 3; Seabrook containment.
Other Sandia analyses for Zion indicate that pressures of 58 to 134 psia- 'are possible due to coincident steam spike, RCS depressurization, and hydrogen combustion. These pressures -are obtained without direct heating; slightly higher pressures should prevail for Seabrook given similar conditions, as noted above.
Considering direct heating, if both thermal and chemical energy are transferred directly tc the containment atmosphere, involvement of "only modest fractions of tne core debris (10 to 35%) could generate severe threats to containment integrity even for Zion." g/
PLG analyses to date simply have not adequately responded to the analytical and experimental work performed by Sandia National Laboratories. As a result, PLG analyses are repetitive, and, 3 '
therefore, consistently optimistic ' in. asserting so ' confidently that containment f ailure will not' occur as a result of HPME/DCH. 91 / The 91/
F.E. Haskin, V.L. Behr, & L.N. Smith, " Combustion-Induced Loads 3 in Large Dry PWR Containments," Sandia National Laboratories, in NUREG/CP-0056, SAND 84-1514, Proceedings of the Second Workshop on Containment Integrity, August 1984, T. Molina & R.
Cochrell, editors, pages 137 and 139.
92/
The latest' argument raised by PLG in defense of its position
] that HPME/DCH presents no substantial threat of containment failure is that structures in the path of the ejected material would have a " substantial effect" on removing core melt debris .
from the gas flow resulting from HPME. See, Karl N. Fleming, '
et al., Seabrook Station Risk Management and Emergency Planning Study, Pickard, Lowe & Garrick, Inc., PLG-0432, prepared for g New Hampshire Yankee Division, Public Service Company of New Hampshire, December 1985, page 4-15. In contrast, recent Sandia experiments have demonstrated that this argument is incorrect. Three experiments in the HIPS series (HIPS-3C, HIPS-7C, and HIPS-8C) indicated that simple structures may not greatly mitigate the dispersal of debris into rhe containment atmosphere. Indeed, in HIPS-7C, over 95% of the debris was O found outside the simila ted reactor cavity (whereas PLG had O
J uncertainties are such that the potential for early failure of the Seabrook containment due to HPME/DCH cannot confidently be precluded.
J In PLG-0432, PLG presented a three-step model for DCH. PLG held that three steps were necessary: (1) debris must be displaced from the reactor cavity region, (2) the debris must be particulated into fine aerosol, and (3) the debris must be distributed throughout the containment volume. PLG acknowledges that a " substantial impact would result if 30% or more of the core inventory is involved" in DCH.
O g7 Based on the Sandia experiments, the conditional likelihoods of items 1 and 3 in the PLG assessment must be very high 1.e., about O 0.95) . The overall result would be driven by estimates of the amount of core participating in the ejection and the amount par ticularized i'nto fine aerosol, according to PLG's model O We disagree, however, that the aerosol must be. finely particularized to be effective in participating in direct containment heating. Sandia HIPS experiments indicate that nearly all of the core g debris released from the reactor vessel under HPME conditions will be ejected into the containment atmosphere. The mean particle size for this material ranges from 0.1 to 1 mm in diameter. Essentially all of v,
predicted the debris would be located within the cavity) . See, William W. Tarbell, Marty Pilch, & John E. Brockmann, " Melt Expulsion and Direct Containment Heating in Realistic Plant Geometries," Sandia National Laboratories, SAND 85-1631C, in Proceedings of the International ANS/ ENS Tooical Meetino on O
Thermal Reactor Safety, American Nuclear Society, San Diego, California, February 2-6, 1986, Vol. 1, pages II.6-1 through 11.6-8.
M/ Karl N. Fleming, et al., Seabrook Station Risk Management and Emergency P la nni-ng Study, Pickard, Lowe & Garrick, Inc., PLG-0432, prepared for New Hampshire Yankee Division, Public Q Service Company of New Hampshire, December 1985, page 4-13.
the ejected material will support direct heating effects to some extent. Sandia's results also indicate that 0.3% to 5.2% of the 3
ejected material will be in the form of aerosol particles which will remain airborne for long periods of time and will be available for release if containment failure occurs. 94/ 95/
J-This may not be the limit of the releases, however, since there is the potential for enhanced release of some species (due to conversion to volatile oxide forms). There may also be some later releases as core debris deposited on containment surfaces oxidizes after containment failure. These two mechanisms have not been extensively studied, but some results for the former mechanism may be O available within the year. 96/
-94/ Telephone conversation with John Brockmann. See, also, Marty Pilch, William W. Tarbell, & John E. Brockmann,' Melt Expulsion-
') and Direct Containment Heating in Realistic Plant Geometries,"
Sandia National Laboratories, SAND 85-1631C, in , Proceedings of
- the International ANS/ ENS Topical Meeting on Thermal Reactor Safety, American Nuclear Society, San Diego, California, February 2-6, 1986, Vol. 1, pages II.6-1 through II.6-8.
3 9_5,/ The recent DCH-1 test in the SURTSEY facility produced results which indicate that from 5% to 25% of the debris was aerosolized. See, William W. Tarbell, John E. Brockmann, &
Marty Pilch, "'D6H-1 D : The First Direct Containment Heating Experiment in the SSURTSEY Test Facility," Sandia National Laboratories, SAND 86-2220C, in NUREG/CP-0081, Transactions of 3' the Fourteenth Water Reactor Safety Information Meeting, Gaithersburg, Maryland, 10/27-31/86, page 16-2.
~
96/ Sandia has conducted some HPME experiments, including the recent DCH-1 test in the SURTSEY experimental facility, in which fission product mocks have been used. Analysis of these test data is in progress. The DCH-1 test and the SURTSEY 3 facility are discussed in a recent conference paper. See, William W. Tarbell, John E. Brockmann, & Marty Pilch, "DCH-1:
The First Direct Containment Heating Experiment in the SSURTSEY Test Facility," Sandia National Laboratories, SAND 86-2220C, in NUREG/CP-0081, Transactions of the Four teen th Water Reactor Safety In fo rma tion Meeting, Gaithersburg, Maryland, 10/27-O 31/86, pages 16-1 to 16-2.
3
b Sandia's " pessimistic" estimate in NUREG/CR-4700 assumes that 100% of the core is ejected, while the " central" estimate assumes that 75% of the core participates. Since essentially all of this material will participate in direct heating, a significant threat to containment integrity for Zion would result in either case. By analogy, due to similarities between Zion and Seabrook, we believe that Seabrook's containment would be similarly threatened.
Sandia National Laboratories, in work supporting the
) forthcoming NUREG-ll50 report, has also created a model for HPME/DCH in the context of quantifying the containment event tree for the Surry plant. For a plant damage state' designated as TNNN (defined as intact
) RCS, no RWST inventory discharged to containment, containment sprays failed -- i.e., essentially station blackout) , Sandia predicts early containment f ailure to occur 0.4% of the time in the " central"
) estimate and 51% of the time' for the " pessimistic" estimate. 97f For .
scenarios at high pressure with RWST inventcry injected into, the containment, and sprays and containment heat removal available (plant
] damage state TYYB), Sandia still predicts HPME/DCH failure 0.1% of the time in the " ce ntra l" estimate and 83% of the time in the
" pessimistic" estimate (sprays fail 46% of the time, and remain 3- operating 37% of the time) for Surry. 98/
97/ This represents the sum of the conditional probabilities for
" containment release modes" 7 and 8. We believe that this is correct for plant damage state TNNN since the sprays are always
) failed in this damage state.
Containment Event Tree See, A.S. Benjamin, et al.,
Analysis for Postulated Severe Accidents: Surry Power Station, Unit 1, Sandia National Laboratories, NUREG/CR-4700, SAND 86-1135, Vol. 1, draft report for comment, pages 50-51.
-98/ A.S. Benjamin, et al., Containment Event Tree Analysis for
) Postulated Severe Accidents: Surry Power Station, Unit 1,
~
b.
of course, it is true that Seabrook has, in comparison to Surry, a larger containment free volume and a higher estimated failure pressure. On the other hand, Seabrook also has a larger thermal power level, a larger mass of metal in the core, and larger RCS water and J steam volumes. Thus, although the Surry probabilities cannot be directly applied to Seabrook, we nonetheless believe that although there is substantial uncertainty, HPME/DCH-caused containment failure
) for Seabrook cannot be dismissed. ,99/ 100/ 101/ Early containment Sandia National Laboratories, NUREG/CR-4700, SAND 86-ll35, Vol.
1, draft report for comment, pages 50-51.
) g/ BNL's earJy draft assigns a low conditional probability to DCH failure for Zion - probabilities range from 2.2 x 10-4 to 3.3 x 10-4. See, M. Khatib-Rahbar, et al., -Rebaselining of Risk for Zion, Brookhaven National Laboratory Technical Report A-3293, first draft, 4/30/86, page 3-19. We understand (personal communication with W.T. Pratt, BNL) that this first draft was
] not based on any specific calculations of DCH pressure increments, but rather represented what is essentially
" engineering judgment" of the likelihood of containment failure due to DCH. Dr. Pratt informs us that more recent versions of the Zion analysis, which are structured in a similar fashion to Sandia's Surry calculations, come up with results very similar 3 to Surry, although slightly better (properly reflecting Zion's stronger containment) . Considering the results and discussion presented above, we do not believe that the Zion result predicted by BNL represents a best estimate. Rather, it may be characterized as representing approximately Sandia's
" optimistic" estimate group.
) 100/ We no te , in addition, that PLG has recently presented their assessment of what BNL would calculate as the conditional probability of early containment failure for Seabrook due to HPME/DCH. PLG's estimate was 1.5 x 10-4 for the conditional containment failure probability for DCH. See, U.S. NRC, Transcript of meeting da ted 1/14/87, " Meeting of NRC with
) Brookhaven National Lab RE: Seabrook EPZ," 113-122, and last attachment. We are informed (personal communication, W.T.
Pratt, BNL) that the basis for the PLG calculation is incorrect; moreover, a final draft of BNL's assessment for Zion is due out at the end of this month, and it will contain a more complete assessment of DCH containment failure likelihoods for g Zion. BNL intends to perform such an assessment for Seabrook in its final report; that report is not due to be completed for some time.
O
}
failure due to HPME/DCH must be considered to ,be within the uncertainties involved in severe accident analyses of Seabrook. Thus, we disagree with the PLG-0432 and PLG-0465 " peer review group" comments that the results of those studies are not affacted by uncertainties. We further disagree with PLG and BNL that HPME/DCH either cannot cause the Seabrook containment to fail or cause it to fail at a very low conditional probability.
)
III.7 Containment Failure Pressure Analysis
) The containment failure pressure analysis in the original Seabrook PRA resulted in the creation of a probability distribution which was asserted to express the analysts' degree of belief in the
) failure pressure. The original Seabrook PRA stated that there was a
" negligibly sma l'. probability" that the containment would survive
' pressures in excess of 210 psia. 102/ ~Now, after further analysis,
)
101/ Sandia performed an assessment of the conditional probability of the failure of Zion's containment from HPME/DCH in a 1985 draft report. For station blackout (with early failure of auxiliary feedwater) , Sandia's central estimate was a
) conditional probability of 2 x 10-3; the pessimistic estimate was 43%. Sandia presented results for other sequences as well.
For a reactor coolant pump seal LOCA with injection failure (sprays succeed), the HPME/DCH conditional failure probability was 2% for the central estimate and 77% for the pessimistic estimate. For a small LOCA with injection failure (sprays succeed), the results were 3 x 10-3 for the central estimate
) and 60% for the pessimistic es tima te . In each case, the optimistic analysis result was a conditional probability of essentially zero. See, A.S. Benjamin, et al., Containment Event Analysis and Estimation of Source Term Frequencies, Sandia National Laboratories, draft for review, 22 March 1985, pages 77-79.
102/ Karl N. Fleming, et al., Seabrook Station Probabilistic Safety Assessment, Pickard, Lowe & Garrick, Inc., PLG-0300, prepared
)
D we are asked to believe that the median failure pressure (i.e., 50th percentile failure pressure) is 213 poia. 103/
Table 3 provides a comparison cf the containments for Seabrook, Millstone 3, and Zion Unit 1. All of these reactors have been the 3
subject of full-scope PRAs.
Brookhaven National Laboratory reviewed the containment failure analysis in the original Seabrock PRA (PLG-0300, December 1983) . BNL b noted that, as analyzed in th e PRA , the median pressure which causes yield of both the containment liner steel and the containment rebar is 157 psig. 104/ This pressure corresponds well to the definition of O containment failure used in a number of previous PRA studies.
Nonetheless, the Seabrook PRA went beyond this pressure to estimate the ultimate failure pressure, which the PRA placed at 216 psig. 105/
9 BNL found that the containment penetrations would leak at lower pressures. The fuel transfer tube was found to result in a 3 in 2 leak area at 172 psig; the main feedwater penetrations were found to result 3 in a total leak area of 200 in 2 at a pressure of 180 psig; and the safety injection lines were found to result in a leak area of 2 in 2 at for Public Service Company of New Hampshire and Yankee Atomic Electric Company, December 1983, page 4.4-13.
D 103/ This indicates the unreliability of " engineering judgment" as expressed in terms of a probability distribution. There is reason to question whether a person's " state of knowledge" can be reliably be expressed in such a manner. Of course, upon further reflection or additional analysis one can change one's mind --
the concern here is whether this can be reliably O reflected in a probability distribution. We emphatically do not believe it can.
104/ M. Khatib-Rahbar, et al., A Review of the Seabrook Station Probabilistic Safety Assessment: Containment Failure Modes and Radiological Source Terms, Brookhaven National Laboratory,
- D NUREG/CR-4540, BNL-NUREG-51961, February 1986, page 24.
105/ Ibid.
S
h a pressure of 166 psig. BNL classified the main feedwater penetration failure at 180 psig as a Type C failure per the definition in the SSPSA study. 106/ The Seabrook PRA assessed different shell failure pressures for wet and dry sequences; for wet sequences, the failure pressure was assessed at 211 psia, for dry sequences, the failure
)
pressure was assessed at 187 psia. 107/
It is clear from this comparison, as well as additional data in NUREG/CR-4540, that for HPME scenarios, Zion and Seabrook should
) achieve very similar pressures. In addition, Seabrook has slightly larger masses of water, steam, uranium dioxide, and zirconium (and slightly less steel) . These results would also tend Seabrook toward
) slightly higher pressures.
BNL assumed that' for all sequences where there is full containment spray operation, no containment failure would result.
) 108/ This may not necessarily be true considering HPME and hydrogen burn. 109/
106/ NEED REFERENCE
) 107/ NEED REFERENCE 108/ M. Khatib-Rahbar, et al., A Review of the Seabrook Station Probabilistic Safety Assessment: Containment Failure Modes and Radiological Source Terms, Brookhaven National Laboratory, NUREG/CR-4540, BNL-NUREG-51961, February 1986, page 43 109/ For example, see the conditional containment failure probabilities due to HPME/DCH for Zion as estimated by Sandia.
For a reactor coolant pump seal LOCA with injection failure (sprays succeed), the HPME/DCH conditional failure probability was 2% for the central estimate and 77% for the pessimistic
) estimate. For a small LOCA with injection failure (sprays succeed), the results were 3 x 10-3 for the central estimate and 60% for the possimistic estimate. In each case, the optimistic analysis result was a ccaditional probability of essentially zero. See, A.S. Benjamin, et al., Containment Event Analysis and Es tima tion of Source Term Frequencies,
) Sandia National Laboratories, draft for review, 22 March 1985,
)
i I
) ,
The Seabrook PRA also appears to assume that if th'ere is a wet cavity, no " vaporization" release occurs. This may be true for the WASH-1400 definition of vaporization release, but considering HMPE/DCH sequences and steam explosions (both in-vessel and ex-vessel) , it may D
not be true. Both HPME/DCH' and steam explosions present the opportunity enhanced releases when compared to the WASH-1400 definition of " vaporization" release. ,
D I III.8 Steam Explosions
) Most recent PRAs have assigned a probability of 1 x 10-4 or less to the conditional probability that an in-vessel steam explosion-could result in the production of a missile capable of penetrating the S containment. The Seabrook PRA , study assigned a conditional probability of 5 x 10 -4 to this failure mode. 110/
There is a considerable range of opinion among reactor safety 3 analysts as to the likelihood of containment failure arising from in-vessel steam explosions. Some analysts believe that such a failure mode is physically impossible. Others believe that the probability is g either indeterminate or possibly rather high.
pages 77-78. In bY ~ 7 ) these sequences, sprays and fan coolers operate. Ic snv., J be noted that although containment failure is predicted to occur, the source term should still be S relatively low due to spray operation. Whether sprays and fan ,
coolers would continue to operate after an HPME/DCH/ containment failure event is open to debate, but appears questionable in vur view.
110/ M. Khatib-Rahbar, et al., A Review of the Seabrook Station 3 Probabilistic Safety Assessment: Containment Failure Modes and Radiological Source Terms, Brookhaven National Laboratory, NUREG/CR-4540, BNL-NUREG-51961, February 1986, page 42.
O
)J A 1983 study carried out for the UKAEA by a researcher at the j Winfrith Laboratory p'rovided rela tively large containment failure probabilities for in-vessel steam explosions. For low pressure sequences, the analysis predicted an upper limit conditional probability of 1.3 x 10-2; for high pressure sequences, an upper limit
)
conditional probability of 8 x 10 -3 was predicted. 111/ If these results are applied to Seabrook, one obtains an " upper limit" (by Briggs' definition) probability of containment failure by in-vessel steam explosion of about 1.9 x 10 -6 This represents about 0.8% of the overall core melt frequency. 112/
Sandia National Laboratories performed an uncertainty study of
) PWR steam explosions for the Zion plant in 1984. The results of this study indicated broad undertainty over the probability of containment failure due to in-vessel steam, explosions -- the so-called " alpha-mode
) failure". In essence, Sandia found that the uncertainties spanned the range from a conditional probability of alpha-mode failure of zero to 100%. 113/
)
111/ A.J. Briggs, The Probability of Containment Failure by Steam Explosion in a PWR, Safety & Engineering Science Division, AEE Winfrith, AEEW-R1692, December 1983, page 20.
) 112/ A model calculation based on a Monte Carlo sampling method estimated a conditional containment failure from in-vessel steam explosions of 2.67% to 4.48% (i.e., 2.67 x 10-2 to 4.48 x 10-2). See,. M. Berman, J.M. McGlaun, & M.L. Corradini, " Core Melt / Coolant Interactions: Modelling," Sandia National Laboratories, SAND 83-1852C, in Proceedings of the U.S. Nuclear Regulatory Commission Eleventh Water Reactor Safety Research
} Information Meeting, NUREG/CP-0048, Vol. 3, Gaithersburg, Maryland, October 24-28, 1983, pages 226-248. These results are largely consistent with the Briggs analysis.
113/ M. Berman, D.V. Swenson, & A.J. Wickett,, An Uncertainty Study of PWR Steam Explosions, Sandia National Laboratories,
) NUREG/CR-3369, SAND 83-1438, May 1984, page 49.
)
).
The results of this uncertainty study created an immediate controversy since many previous studies had taken the position that
) the conditional probability of alpha-mode failure was either extremely low (i.e., of the order of 1 x 10-4) or zero. The NRC convened a panel called the " Steam Explosion Review Group" (SERG) to review the
) ma tte r. The SERG report concluded that: 114/
(T)he consensus of the SERG is that the occurrence of a steam explosion of ' sufficient energetics which could lead to alpha-mode containment failure has a low probability. This conclusion is reached despite the
) expression of differing opinions on modeling of basic steam explosion sequence phenomenology. An opinion supported by most members of the group is that the probability of containment failure is reduced due to the expectation of limited melt mass involvement in the '
explosion and/or low thermal-to-mechanical energy
) conversion. Other members placed more emphasis on reduction by other physical means.
. It should be noted, however, that the SERG considered the
) Indian Po' int Unit 2' plant, not Zion. 115/ In addition, neither SERG nor Sandia consider heatup of the vessel head bolts. Such heating could occur in severe accidents; this would result in failure of the
) head bolts at a lower level of impact energy, thus leaving more of the energy generated by the steam explosion available for impacting the containment. This factor may increase the likelihood of containment
)
114/ Steam Explosion Review Group, A Review of the Current Understanding of the Potential for Containment Failure From In-Vessel Steam Explosions, U.S. Nuclear Regulatory Commission, NUREG-lll6, June 1985, page iv.
115/ S team Explosion Review Group, A Review of the Current Understanding of the Potential for Containment Failure From In-Vessel Steam Explosions, U.S. Nuclear Regulatory Commission, NUREG-lll6, June 1985, page 7. It is no t immediately clear whether this difference is significant.
)
) failure from steam explosions, but has not been considered in any published analysis to date.
)-
It should also be recognized that individual SERG . members '
alpha-mode f ailure probability estimates for low pressure sequences varied considerably, ranging from essentially zero up to "less than 10-2", and included specific best estimates as high_as 4 x 10 -3 and upper limit estimates of 0.1. 116/ On the basis of _ this large range of estimates, we do not consider the SERG's conclusion to be credible.
) All the SERG did was confirm that the uncertainty was very large, differing from the NUREG/CR-3369 Sandia analysis in the cutoff for the upper end of the uncertainty (i.e., SERG put the upper limit at 0.1,
) Sandia put the upper limit at 1) . 117/ In addition, the SERG report shows amply that there is little agreement on how to model steam explosions -
i.e., virtually every SERG member who employed a model
] ,
in deriving their estimate used a model that was different from every other model.
)' 116/ Steam Explosion Review Group, A Review of the Current Understanding of the Potential for Containment Failure From In-Vessel Steam Explosions, U.S. Nuclear Regulatory Commission, NUREG-lll6, June 1985, page 12. -
117/ Marshall Berman of Sandia National Laboratories, in a recent
) -
paper (written af ter the SERG report) , concluded, "Every stage leading to a possible alpha-mode failure is credible, and some stages are currently believed by some experts to be highly likely. There are no experiments or models which are capable of demonstrating low probabilities for any of the stages or for the process as a whole." See, M. Berman, "An Evaluation of the Bases for' Estimating Alpha-Mode Failure Probabilities," Sandia
)' National Laboratories, Paper XI.7-1, in Proceedings of the International ANS/ ENS Topical Meeting on Thermal Reactor Safety, San Diego, California, February 2-6, 1986. See, also, Marshall Berman, " Molten-Core Coolant Interactions Program,"
Sandia National Laboratories, presented at the U.S. NRC Twelfth Water Reactor Safety Research Information Meeting, October 24, 1984.
}
3
i Sandia ran a number of calculations on a Monte Carlo sampling
} program for its steam explosion model. In the " full width" calculation, which took note of the full range of uncertainty in the input parameter, Sandia calculated an alpha-mode failure probability
} of 2.67% for a large missile with a velocity greater than 90 meters /second. 118/ Although the Sandia analysts do not accept this as their "best estimate", we believe that it represents a plausible
) estima te of the alpha-mode failure probability for low pressure sequences (no specific estima te was provided for high pressure sequences). If this estimate is applied to the Seabrook results, about 2% of the core melt frequency consists of low pressure sequences
). -- i.e., a frequency of 4.8 x 10 -6 for low pressure sequences.
Multiplying this by the 2.67% conditional probability of alpha-mode failure yields a. total frequency of alpha-mode failure of 1.3 x 10 ~7 D
Alternatively, Sandia's more recent-analyses for NUREG/CR-4700 included alpha-mode failure es tima tes . The " central" estima te for both low and high pressure sequences was zero, while the " pessimistic" D
estimate was 10%. 119/ Applying the pessimistic value to Seabrook, this would result in an alpha-mode failure frequency of ten percent of the core melt frequency, or 2.4 x 10-5, O We have provided three alternative calculations of what the overall frequency of alpha-mode releases might be. These estimates, 9 118/ M. Berman, D.V. Swenson, & A.J. Wickett, An Uncertainty Study of PWR Steam Explosions, Sandia National Laboratories, NUREG/CR-3 369, SAND 83-14 38, May 1984, page 35.
119/ A.S. Benjamin, et al., Containment Event Analysis for Postulated Severe Accident: Surry Power Station, Unit 1, Sandia g National Laboratories, NUREG/CR-4700, SAND 86-1135, Vol. 1, draft report for comment, November 1986, pages 44-46.
O
E
) I based on different models, span the range as high as 2.4 x 10 -5 and as low as 1.3 x 10 ~7 Using the Seabrook PRA's value of 5 x 10 ~4 for the
)
conditional probability of alpha-mode failure, one would obtain an es tima te o f 1. 2 x 10 ~7 All of these results are plausible within the current state of
) knowledge; they clearly indicate that there are large uncertainties involved in estimating alpha-mode failure likelihood. Such a large-range of uncertainties indicates that alpha-mode failure scenarios
) cannot be ignored in the context of emergency planning zone distances.
III.9 Volatile Iodine Forms
)
PLG-0 465 and PLG-043 2, as well as the draf t BNL review of these documents, ignore the potential for releases of iodine in forms other
) than cesium iodide. In contrast, NRC-sponsored research indicates tha t a non-negligible fraction of the iodine release could be in the form of volatile iodine, specifically organic iod'ide. A letter report
) documenting this work concludes that 0.5% (i.e., 5 x 10-3) of the iodine released to containment would be in the form of organic iodide.
120/
The availability of volatile iodine at this level sets a floor for the lower limit of iodine releases. The key factors to be considered are whether there are mechanisms for reducing the volatile iodine concentration (i.e., are containment sprays ef fec tive? ) , and
) '
120/ Arlin K. Postma, Discussion of Organic Iodide Formation Under Severe Accident Conditions, Benton City Technology, BCT-LR 1, page 6, attachment to a letter dated March 4, 1985, from Arlin K. Postma, Benton City Technology, to Jocelyn Mitchell, NRC.
)
J
) .
the time delay between accident initiation and containment failure.
Iodine releases . would be unlikely to be as low as postulated by PLG when. volatile forms are considered. 121/
In addition, experiments at the VGES facility at Sandia
)
National Laboratories indicate that airborne cesium iodide (the assumed chemical form of iodine in the PLG studies and the draft BNL review) can be converted to molecular iodine (i.e., I) 2 by hydrogen
) burns. The experiments indicated that 20% of the airborne CsI was converted to I 2 122/ There are also conflicting experimental results which appear to indicate the possibility that cesium iodide
) (CsI), which is the assumed form of iodine in the PLG studies, may be dissociated by exposure to intense gamma radiation. 123/ Other forms of iodine would then be dominant in severe accidents, such as HI, HOI,
) or I 2
These vola tile forms would behave very differently from CsI.
121/ It should be noted that, even in the TMI-2 accident, with a release pathway- through water, the iodine measured in the
) containment atmosphere represen4ed 0.03% of the core inventory.
See, C.A. Pelletier, et al., " Preliminary Results of the TMI-2 Radiological-Iodine Mass Balance Study," in Transactions of the 1982 Winter Meeting, American Nuclear Society, Vol. 43, 1982, page 17.,
122/ " Chemical Modification of CsI/Containing Aerosols During aH Burn," NRC Weekly Information Report, Week Ending December 14,2 1984, Enclosure E to a memorandum dated December 20, 1984, from T.A. Rehm, Office of the Executive Director for Operations, NRC, to the NRC Commissioners.
123/ R.M. Elrick & D.A. Powers, " Effects of Ionizing Radiation on
) the Transport Chemistry of Cesium Iodide," Sandia National Laboratories, 1985. See, also, A.R. Taig, et al.,
" Interpretation of Experimental Results on the Interactions of Fission Product Vapours With Reactor Materials," Sandia Na tional Laboratories, presented at the ANS Meeting, Snowbird, Utah, 7/15-19/85; and NRC memorandum dated January 10, 1986,
) - from Christopher Ryder and Lisa Chan to Mel Silberberg,
Subject:
" Summary of the Meeting on the Effect of Ionizing Radiation on the Stability of Cesium Iodide," with attachments.
)
)
124/. Therefore, the PLG analyses and assumptions regarding iodine
)
source terms may be substantially in error. Significant errors in estimating iodine source terms are important to early dose calculations since iodine is of ten a key contributor to early doses.
III.10 Fission Product Reevolution
)
NRC-sponsored source term calculations (as provided in NUREG-0956, BMI-2104, and NUREG/CR-4624) do not account for phenomena which might serve to increase source terms. A key issue is whether deposited fission product aerosols could undergo chemical reactions and/or physical disturbances which would re-evolve these materials as aerosols available for release to the environment following
) containment failure.
This is an area of considerable uncertainty at the current, time. We mention it here briefly to point out that the source term
) estimates made to date with the new source term codes may be underestimates to the extent that re-evolution is not addressed.
) III.11 Accident at Under Shutdown Conditions During the review of the PLG-0465 and PLG-0432 submittals, Brookhaven Na tional Laboratory and NRC staff reviewers raised the
)
issue of the risk arising from accidents occurring during shutdown 124/ The potential for photolysis of CsI was recognized as early as 1984. See, letter dated June 29, 1984, from Richard S.
,) Denning, Battelle Columbus Laboratories, to Prof. Richard Wilson, Harvard University, attachment, " Responses to APS Questions of June 6."
)
}
conditions. This is an area that is usually neglected in PRAs due to the assumed greater level of risk from accidents occurring at power,
)
and the Seabrook PRA was no exception in this regard.
Both BNL, and PLG in response to BNL, rely in part on a PLG study of shutdown accidents for the Zion nuclear plant, referred to as NSAC-84. 125/ NSAC-84 represents one of the first attempts at assessing risk from accidents during shutdown conditions, and is represented as an extension of the Zion Probabilistic Safety Study
) (ZPSS) which PLG performed. 126/
NSAC-84 concluded that the mean frequency of core damage from cold shutdown ' accidents was 1.8 x 10 -5 per year for Zion; the median value was estimated at 2.6 x 10 -6 These results represent about 27%
of the frequency of core damage from power operations for mean results and about 5% for median results. This indicates, as noted in NSAC-84,
) that there are greater uncertainties.about the cold shutdown accident frequency estimate. 127/
Although the ZPSS included accidcnt sequences initiated by
) " external events" (such as firea, floods, earthquakes, etc.), the analysis in NSAC-84 was limited to " internal events". In our view, the exclusion of external events, and, in particular, seismic
)
125/ D.C. Bley, et al., Zion Nuclear Plant Residual Heat Removal PRA, Pickard, Lowe & Garrick, Inc., prepared for the Nuclear Safety Analysis Center, NSAC-84, July 1985.
j 126/ Pickard, Lowe & Garrick, Inc., Fauske & Associates, Inc., and Westinghouse Electric Corporation, Zion Probablistic Safety
, prepared for Commonwealth Edison Company, September 127/ D.C. Bley, et al., Zion Nuclear Plant Residual Heat Removal PRA, Pickard, Lowe & Garrick, Inc., prepared for the Nuclear
) Safety Analysis Center, NSAC-84, July 1985, page 6-2.
).
initiators, from NSAC-84 represents a significant limitation on the usefulness of that study. By analogy, any "rebaselining" of the NSAC-84 to represent Seabrook will also be affected by this limitation. To i date, no results for shutdown accidents initiated by external events '
).
has been presented by the NRC staff, BNL, or PLG.
There is at least one previous assessment of cold shutdown accidents which did include consideration of seismic events. This is a 1982 analysis by Science Applications, Inc., of the Sequoyah plant
("SAI Study"). 128/ The SAI Study was limited to earthquakes of the size of the Safe Shutdown Earthquake (SSE). It is well-recognize'd
) that most of the ' seismic risk . for nuclear power plants comes, from seismic events which are several multiples of the ground acceleration represented by the SSE. 129/
) Despite these limitations, the results of the SAI Study are of l .
interest. .The SAI Study performed a conservative (" upper limit")-
l analysis of core melt frequency from cold shutdown events from j operator errors and seismic events up to the severity of the SSE. A l
total of 20 LOCA cases were examined in the SAI Study. With no l maintenance underway at the time of the LOCA, the core melt frequency
. was estimated to be in a range from 3.96 x 10 -5 to 1.14 x 10 ~7 per l
l 128/ Peter Lobner, William Horton, & Bruce Kirstein, A Preliminary
( Assessment of Core Melt Probability in Cold Shutdown Following l a Postulated LOCA at the Sequoyah Nuclear Plant, Science
(/ Applications, Inc., prepared for the U.S. Nuclear Regulatory Commission, SAIOl382-147LJ, Rev. 1, July 16, 1982.
129/ A case in point is the Seabrook PRA, which found that almost half of the seismic core melt frequency came from seismic events with a ground acceleration of 0.7g to l'.0g; the Seabrook SSE is 0.25g. Thus, about half of the seismic core melt
) frequency comes from earthquakes with ground accelera tions approximately three to four times larger than the SSE for Seabrook.
) .
reactor-year. If maintenance (affecting one electrical or cooling water support system train) was in progress, the core melt frequency
} increased to a range of 7.53 x 10 -5 to 8.46 x 10 -6 130/ A significant conservatism in the study is that in the event of an SSE seismic event, a pipe break occurs with a , probability of 1.0. 131/
) The study also assumed, however, that plant conditions in cold shutdown do not significantly affect equipment failure probabilities.
132/ It is not clear that the later assumption is correct.
The SAI Study reported that there were no Technical
}
Specification requirements to maintain containment l' solation during cold shutdown. A preliminary survey of practices at several PWRs it was common practice
) indicated that for containment equipment hatches to be open and both airlock doors open simultaneously during cold shutdown to support the movement of personnel and equipment into
) '
130/ Peter Lobner, William Horton, & Bruce Kirsteld, A Preliminary Assessment of Core Melt Probability in Cold Shutdown Following a Postulated LOCA at the Sequoyah Nuclear Plant, Science Applications, Inc., prepared for the U.S. Nuclear Regulatory
} Commission, SAIOl382-147LJ, Rev. 1, July 16, 1982, page 1.
131/ The initiating event frequency for the SSE seismic event was assumed to be 2.0 x 10-4. Peter Lobner, William Horton, &
Bruce Kirstein, A Preliminary Assessment of Core Melt Probability in Cold Shutdown Following a Postulated LOCA at the Sequoyah Nuclear Plant, Science Applications, Inc., prepared
} for the U.S. Nuclear Regulatory Commission, SAIOl382-147LJ, Rev. 1, July 16, 1982, page 75. It should be noted that this is not much higher in frequency than some rather substantial seismic accelera tions for the Seabroook plant. For example, the initiating event frequency for a 0.79 transient for
) Seabrook is 1.97 x 10-5.
1.00 x 10-6.
For a 0.7g LOCA, the frequency is See, Karl N. Fleming, et al., Seabrook Station Probabilistic Safety Assessment, Pickard, Lowe & Garrick, Inc.,
PLG-0300, prepared for Public Service Company of New Hampshire and Yankee Atomic Electric Company, December 1983, page 5.1-2.
132/ Id., page 71.
J
) .
l and out of the containment. 133/ Depending upon the specific practices employed, it mught be difficult to re-establish containment
) integrity in the short term. For instance, the time required to close the equipment hatch was found to be comparable to the minimum time available for operator protective actions for cold shutdown accidents
) '
occurring during the interval from eight to thirty-five hours after shutdown. 134/ 135/
In the SAI Study, seismic events were postulated to be
) important contributors to core melt frequency from cold shutdown accidents. The SAI Study examined the issue parametrically.
Nonetheless, the results suggest that seismic contributors to cold
) shutdown accidents should not be ignored. We therefore recommend that the analysis of cold shutdown accidents be expanded to include external initiating events, particularly seismic. It is plausible D that bounding arguments can be made to limit the consideration , of 1 '
other external events, but it is unlikely that such an argument can be l' made for seismic events. Absent this step, we conclude that the cold j shutdown accident risk as presented by BNL and PLG may misrepresent the actual risk. Full consideration of seismic initiating events, as well as consideration of altered equipment and containment isolation states, is required in order to complete the cold shutdown risk analysis.
t i
O 133/ Id., page 14.
l 134/ Id., page 15.
135/ We have not had the time to investigate these matters for
! Seabrook specifically; such an examination would be a priority j item for a longer-term study.
o
- f. -
III.12 De-Inerting Burns Leading to Containment Failure
)
The PLG analyses in PLG-0465 and PLG-0432 assume that if AC power and/or containment sprays, as appropriate, are recovered.during an accident sequence, the sequence is benignly te rminated. This analysis appears to credit operator actions to recover AC power and/or containment sprays without considering the possibility that adverse consequences could result from such actions.
When sprays are recovered several hours af ter vessel failure, the containment atmosphere --
which was previously steam inerted --
begins to experience condensation of, steam. This leaves behind very high concentrations of oxygen and hydrogen. A "de-inerting" burn or I detonation can then occur, which might raise pressure sufficiently to
) threaten the structural integrity of the containment. This potential has not been addressed by PLG.
An analysis of de-inerting burns for Zion indicates that under
) some circumstances, hydrogen burn pressures could be generated that are marginally threatening to containment integrity. 136/ A " Severe Accident Sequence Analysis" (SASA) program study for Zion calculated a
} pressure of greater than 170 psia for a deinerting burn at 8.5 hours5.787037e-5 days <br />0.00139 hours <br />8.267196e-6 weeks <br />1.9025e-6 months <br /> into a station blackout accident. 137/ Similar results, employing 136/ F.E. Haskin, V.L. Behr, & L.N. Smith, " Combustion-Induced Loads in Large-Dry PWR Containments," Sandia National Laboratories, in NUREG/CP-0056, SAND 84-1514, Proceedings of the Second Workshop on Containment Integrity, Augus t 1984, T. Molina & R.
Cochrell, editors, Figure 3, page 141.
137/ F.E. Haskin, et al., Analysis of a Hypothetical Core Meltdown
) Accident Initiated by a Loss of Offsite Power for the Zion 1 Pressurized Water Reactor, Sandia National Laboratories, NUREG/CR-1988, SAND 81-0503, November 1981.
)
l ~
l I ' state-of-the-art severe accident model's, were recently obtained for j Zion, however, more time was required to de-inert. 138/ This would allow for greater fission product attenuation due to spray effects.
The point here is not that de-inerting burns would necessarily result in a 1&rge release of fission products, although noble gases
) would certainly be released 139/ (along with aerosols generated by re-evolution of deposited fission products, radioactive decay, 140/ and other processes), following containment failure due to the de-inerting
) burn. The point here is that, in contrast with the PLG assessment, containment integrity may not be maintained indefinitely. Rather than maintaining containment integrity for more than a day, de-inerting burns may result in containment failure in a matter of hours if AC power is recovered in station blackout sequences. Since station blackout analogues make up a significant fraction of Seabrook's "
estimated core melt frequency, de-inerting burns may be an important
)
138/ F.E. Haskin, V.L. Behr, & L.N. Smith, " Combustion-Induced Loads in Large-Dry PWR Containments," Sandia National Laboratories, in NUREG/CP-0056, SAND 84-1514, Proceedings of the Second Workshop on Containment Integrity, August 1984, T. Molina & R.
) Cochrell, editors, Figure 3, page 141.
139/ Noble gases are unaffected by fission product attenuation mechanisms other than decay. These gases are chemically inert, and evolve from water very rapidly.
) 140/ Sandia Na tional Laboratories work has shown that decay of tellurium-132 into iodine-132 can result in significant amounts of iodine-132 being available for release if containment failure is delayed a few hours. See, D.C. Williams, " Effects of Radioactive Decay Chains: Tellurium and Iodine," Sandia National Laboratories, Appendix D in R.J. Lipinski, et al.,
) Uncertainty in Radionuclide Release Under Specific LWR Accident Conditions, Sandia National Laboratories, SAND 84-0410, draft, Volume II, "TMLB' Analyses", February 1985. The chemical form of this iodine will be important, as will the rate at which it is generated following a de-inerting burn which causes containment failure. This issue should be studied in the context of Seabrook.
l l
E factor in the conditional frequency of exceeding the EPA PAG doses of 1 and S rem whole-body. Noble gas releases alone --
even after 24 hour2.777778e-4 days <br />0.00667 hours <br />3.968254e-5 weeks <br />9.132e-6 months <br /> decay, can exceed the PAG doses 5 miles downwind if no protective actions are taken. 141/ (See, Figure 2, reproduced from the APS review of draft NUREG-0956.) With only an 8-hour decay, these doses
)
will be somewhat higher.
III.13 Secondary Containment Behavior Under Severe Accident g Conditions The Seabrook containment enclosure (" secondary containment")
has a low pressure capacity. The walls range in thickness from 15-36 inches, but are unlined. The PRA estimated the pressure capacity of the secondary containment at no more than 10 psi differential. 142/
.This differential pressure should easily be exceeded for most accidents. 'Thus, the secondary containment should not be expected to provide for much source term attenuation.
There are other reasons for concluding that the role of the D secondary containment in mitigation of severe accidents will be minimal. First, it is not as substantial a structure as the primary containment; thus, when the primary fails, the secondary will also e
141/ APS Study Group, Report to the American Physical Society of the Study Group on Radionuclide Release from Severe Accidents at Nuclear Power Plants, February 1985, draft, page 29 and Figure 11.B.1.
G 142/ Structural Mechanics Associates, Internal Pressure Capacity for the Containment and Enclosure Building, prepared for Pickard, Lowe & Garrick, Inc., SMA 12811.02, November 1983, Appendix H.1 in Karl N. Fleming, et al., Seabrook S ta tion Probabilistic Sa fe ty Assessment, Pickard, Lowe, & Garrick, Inc., PLG-0300, prepared for Public Service Company of New Hampshire and Yankee
$ Atomic Electric Company, December 1983, page H.1-87.
4 '
fail. 143/ Second, the secondary containment is an unlined structure, such that when concrete cracking occ ur s , the containment
) function is lost. This is not true with respect to the primary containment since the steel liner will maintain its integrity and prevent leakage beyond the pressure at which concrete cracking is
) predicted to occur, up to the failure pressure of the containment.
Third, hydrogen burns in the secondary containment (resulting from primary containment leakage or failure) will likely either cause
) structural failure of the secondary containment or failure of the filtration units. 144/ At most, therefore, we conclude that the secondary containment may offer a slightly greater chance of an s
elevated release point than would be the case if the secondary containment were not present; this presumes that the structure t
survives the severe accident loads, which we believe is rather unlikely except in the case of very slow leakage 'of the primary containment.
l 143/ The original Seabrook PRA concluded that the secondary containment would fail when the primary containment pressure reaches 180 psia due to cracking of the secondary containment concrete. See, Karl N. Fleming, et al., Seabrook Station Probabilistic Safety Assessment, Pickard, Lowe & Garrick, Inc.,
PLG-0300, prepared for Public Service Company of New Hampshire and Yankee Atomic Electric Company, December 1983, page 11.3-24.
144/ The volume of the secondary containment is only 524,344 cubic l feet. See, Karl N. Fleming, et al., Seabrook Station Probabilistic Safety Assessment, Pickard, Lowe & Garrick, Inc.,
) PLG-0300, prepared for Public Service Company of New Hampshire and Yankee Atomic Electric Company, December 1983, page 11.2-l 16. Because the secondary containment volume is so small and because the structure has a low pressure capacity, burning relatively small quantities of hydrogen will cause structural failure of the secondary containment.
)
v--v-. r--- ,- -,,-,,~---._n.. , , , . ~ , - , - ..n.,-,- -,---.nnn-m.-_ _-_-m,.,--_-_ ,,-__m.,__m- , nun --.rn __me.,,--
III.14 Adequacy of " Peer Review" of PLG-0431, PLG-0432, and PLG-0465 Peer review in the context of a PRA study -- particularly one so detailed and complex as the Seabrook PRA and its recent supplements i
(P LG-0 30.0, PLG-0431, PLG-0432, and PLG-0465) is not accomplished in a few days by a panel of experts, no matter what the qualifications of the experts nor their standing among their peers. It is simply not
! possible to perform a meaningful peer review in a few days of meetings.
The PLG-0432 " peer review" group first met on October 30 and l
[ 31, 1985. 145/ The final report pf the " peer review" group - was
. transmitted to the New Hampshire Yankee Division of PSNH on November less than two weeks la te r.
9, 1985 --
The support provided by the
" peer rev-iew" group' for the results of the PLG-0432 is so qualified as
) to be.of little importance.- In particular, the following passage from the " peer review" group's letter is instructive 146/
The Peer Review Group brought to the review a collective
)
broad experience in the many technical issues covered by the RMEPS, especially in the areas of source term, risk assessment, and consequences analysis. However, our review effort by its nature did not enable us to delve into the numerous details of the many calculations in
- the study -- rather, our approach was to review the
} methodology and judgments used by the study team, 145/ According to the meeting notice sent to the peer review group participants, the group obtained the PLG-0432 report on about October 8, 1985. The two-day meeting consisted of a series discussions of the original PRA and PLG-0432 on October 30, 1985, and a mee ting of the peer review group on October 31, 1985 to draft their findings. See, letter to peer review group members from John DeVincentis, October 1, 1985. This letter is provided as Appendix 1 to this report.
146/ Letter dated 9 November 1985 from Robert J. Budnitz to J.
) DeVincentis, page 1.
) '
. - - , . - ~ r - - . . , - , - ,.,w,-,n-.~,.-, ,-__,n ..-.___,,,,,-,,,-,,_,,,,n_,. , , . , , _ , - - - , . - ,--,_-,,,,n .-
).
( including the models and data. When we state that we
> . concur with the study team's approach and conclusions, we mean that the technical approach taken, the engineering judgments made, and the data used seem to us appropriate. Nevertheless, we are not in a position to verify all of the numerous results in detail, al' chough <
we did concern ourselves with enough of the details to
) become convinced of their reasonableness.
We do not dispute the qualifications of the members of the PLG-
) 0465 " peer review" group. Indeed, we are acquainted with the members of the group either personally or by reputation. We do question, however, whether a meaningful peer review can take place in less than
) two weeks, and principally on the basis of two days of joint meetings.
Moreover, the characterization of the " peer review" as an
" independent" peer review must be placed into context.
Dr. Rasmussen was on the Technical Review Board for PLG-0300. Drs. Stratton and Wilson have argued in public forms for reduction of emergency planning zone sizes for some time (Dr. Wilson qu'estioned the rationale for the .
original EPZs when they were adopted) . The exclusion of persons from
)
the peer review panel who are even mildly critical of PRA methodology, or who hold views which are different from those of IDCOR contractors, raises questions about the " independence" of the review. We do not mean here to criticize the motives of the participants on the peer review panel --
rather, it is well recognized that if one wants critical review and comment, one does not seek out a panel of persons
) who are generally of similar viewpoints to the authors of the work being reviewed. The fact that we were able to find so many potentially significant omissions and optimistic assumptions in the
} PLG analyses in such a limited period o f . time is indicative of the nature of the " peer review" performed.
)
i
).'
! - i
- . The " peer review" letters by Budnitz, et al., do not represent peer review in the sense that phrase is normally used in PRA circles.
Peer review implies a searching and detailed review of the l
assumptions, data base, modeling, etc., used in the PRA. A fe#
meetings and -days of deliberations .do not constitute a peer review in any technical sense. Moreover, as the Sandia National Laboratories work underlying NUREG-ll50 demonstrates, there is a'much broader range
[ of views on safety issues than is represented on the PSNH/NHY peer review team.
It is also important to note the makeup of the PSNK peer review group. Although containment strength is a key aspect of the PLG l studies which the peer review group examined (this is acknowledged by the " peer review" group), not one of the seven members of the peer
- j. review group is a structural engineer.
l III.15 Assessment of PLG Dose / Distance Curves for 1- and 5-Rem Doses JEPA PAG Doses)
)
i In this subsection, we review some of the dose / distance curves
- generated by PLG and draw conclusions related to the bases for plume l
) EPZ distances, namely the likelihood of exceeding the 1-Rem and 5-Rem I EPA Protective Action Guide doses. The results of this review are set
- forth in tabular form in Table 4 and Table 5, which show approximate dose / distance relationships for PLG source terms in PLG-0432 and PLG-l
) .
.0465. 147/ These results are' estimated from the PLG dose versus distance curves. 148/
The results in Tables 4 and 5 clearly show that regardless .of whether the "best estimate" or " conservative" or " WASH-1400-type" 3
source terms defined by PLG are used, with the exception of S3W, there is a very substantial probability of exceeding both the 1-Ren and .5-Rem EPA PAG doses at distances from one to ten miles, and, in some 3 cases, to distances of 20. miles or more. Mo reove r, it must be kept in -
mind that these results were calculated by PLG's CRACIT code, which
.models variable plume trajectory, multi-puff releases, and terrain 3 effects. Had these results been calculated by CRAC (as in NUREG-0396) or CRAC2 (as in NUREG/CR-2239) , the values would have been similar to significantly higher in some cases, because these consequence codes 3 model single-puff releases' and invariant plume trajectory, and ignore terrain effects. (See,Section III.17, below.)
It should also be noted that the PLG reports and the PSNH 3 papers filed in support of its plume EP2 reduction request do not 147/ Karl N. Fleming, et al., Saabrook Station Risk Management and Emergency Planning Study, Pickard, Lowe & Garrick, Inc.,
prepared for New Hampshire Yankee Division, Public Service 3 Company of New Hampshire, PLG-0432, December 1385, Appendix A; and Karl N. Fleming, et-'al., Seabrook Station Emergency Planning Sensitivity Study, Pickard, Lowe & Garrick, Inc.,
prepared for New Hampshire Yankee Division, Public Service Company of New Hampshire, April 1986, 5-8 through B-8..
O 148/ These likelihoods are conditional on the occurrence of the indicated source terms. The values in the tables are estimated from the PLG dose / distance curves. The precise values are not important, only their rough magnitudes. Ideally, access to the data on which the PLG curves are drawn would have permitted more precise figures to be reported here, but the results are 9 sufficient for the purposes at hand. Given sufficent time for comparison, we would have utilized CRAC2' to perform a comparison.
O-
)
~
~ address at all the matter of the size of the ingestion EPZ. Thus, the
)
size of the ingestion EPZ must remain at 50 miles.
III.16 Decay Chains and Their Impact on Late Conta'inment Failure
) Source Terms As noted above in Section III.12, decay of tellurium-132 to
- iodine-132 may be an important source of radio-iodine for source terms
)
resulting from very late containment failure. Tellurium-132 has a half-life of 78 hours9.027778e-4 days <br />0.0217 hours <br />1.289683e-4 weeks <br />2.9679e-5 months <br />, and it decays to iodine-132 which has a half-life of 2.3 hours3.472222e-5 days <br />8.333333e-4 hours <br />4.960317e-6 weeks <br />1.1415e-6 months <br />.
)
The inventory of tellurium-132 for a 3200-MWt Westinghouse PWR at shutdown is about 120 million curies. 149/ Thus, the amount oi!
iodine-132 which could evolve from decay of tellurium-132 may be .
) rather large. -
This source,of radio-iodine has not been considered in the NRC source term reasessment work (except the QU3ST study cited to in this section), in the IDCOR source term work, or in the various Seabrook studies by PLG. The tellurium-132 to iodine-132 decay chain thus merits close attention in the emergency planning context since iodine-
) 132 is a potent source of gamma radiation for short-term (i.e. , early) exposure.
l l
149/ U.S. Nuclear Regulatory Commission, Reactor Safety Study, WASH-
) 1400, Appendix VI, " Calculation of Reactor Accident Consequences," page 8-3, October 1975.
)
v- -% . < - - . , , , - _ _ . .
. __.-..-,.-,,-v-. - - . . . - , - . . . - _ - - . . -,_--,m.,, -
..--..---.e.-,-,,_-.4 . . . . - - , _ _ _ _ , - . - - - - , . - . - _ - , - . - , , - - - - - - - - -
>~
III.17 Differences in Modeling Between CRACIT, CRAC, and CRAC2
)
CRACIT is a PLG proprietary modification. of the CRAC code. CRAC
(" Calculation of Reactor Accident Consequences") was first used in the 1975 Reactor Safety Study (WASH-1400), and has been used in a number of applications since that time. These applications include the dose and consequence calculations in NUREG-0396.
) It must be recognized that CRAC does not directly yield dose / distance curves. For NUREG-0396, the dose / distance curves (Figure 3) were obtained by interpolation of values from the WASH-1400
)
source terms and mathematically summed to produce the curve.
For comparison, BNL has performed calculations using CRAC2 (which, as a model output option, does produce dose / distance curves)
) for the WASH-1400 PWR source terms, and then produced a summary curve.
These results are shown in Figure 4.
The CRACIT code employs several innovations which affect dose / distance relationships and consequence calculations in comparison
)
with similar calculations performed by CRAC2 and CRAC. CRACIT has models for terrain effects on the plume, variable plume trajectory, and multi-puff releases. CRAC is constrained to model single-puff, relatively short-time period releases; CRAC2 can model longer ' duration releases, but not of multiple source terms. 150/ Both CRAC and CRAC2
)
150/ CRAC2 employs a sir.ple correction equation to vary the horizontal dispersion parameter. Releases longer than 10 hours1.157407e-4 days <br />0.00278 hours <br />1.653439e-5 weeks <br />3.805e-6 months <br /> in duration are treated as 10-hour releases. See, L.T.
Ritchie, et al., CRAC2 Model Description, Sandia National
) Laboratories, NUREG/CR-2552, SAND 82-0342, March 1984, page 3-11.
)
i
)- y are constrained to employ invariant plume trajectory, and neither has f
the capability to model terrain effects.
The above factors are relevant when no emergency response is assumed to occur. When emergertcy response is considered, there are I ~
additional differences between CRACIT and CRAC/CRAC2. CRACIT employs evacuation vectors (considering both changes in direction associated with the evacuation network and changes in speed) derived from an
). analysis of evacuation time estimates. CRAC (in a revised form) and CRAC2 employ a model which evacuates the population radially away from the reactor at a constant speed. These model dif forences can result
) in substantial differences in consequences.
As a result of these modeling difference ~s, significant differences can result in calculated consequences (whether the
)
consquence of interest is a dose / distance relationsh'ip, early fatalities, latent fatalities, or land contamination area) solely as a result of moving from one consequence model to another. This means one must be very cautions when making comparisons from one report to another when different consequence models were used.
Based on our familiarity with the results of model comparisons, we conclude that some portion of the reduction in risk --
and the
)
reduction in dose / distance relationships upon which PLG and PSNH builds its case for a 2-mile plume EPZ -- comes solely from a change from the CRAC model (as used in NUREG-0396) to the CRACIT model (as
)~ employed by PLG in PLG-0432 and PLG-0465). This difference is ephemeral, and has absolutely nothing whatsoever to do which differences between Surry (upon which the NUREG-0396 curves are based)
) and Seabrook (upon which the PLG curves are based) . The magnitude of
)
).
this effect cannot be accurately quantified absent a comparison calculation using CRACIT. Since this code is proprietary, we are powerless to perform such a calculation ourselves.
Nonetheless, we believe that some appreciation of the differences
)
between CRAC and CRACIT calculations can be obtaine'd from two sources.
First, BNL has run sample calculations with the MACCS code (MELCOR Accident Consequence Code System) which, like CRACIT, employs modeling
) to handle terrain effects, variable plume trajectory, evacuation vectors, and multi-puff releases. Second, the Committee on the Safety of Nuclear Installations of the Organization for Economic Co-Operation
) and Development sponsored a code comparison study in which CRACIT, CRAC, and CP.AC2 were compared with other codes.
The BNL calculations were for PLG release cateogry S2W. These
) calculations are depicted in Fiqure 5. Assume for the moment that the curves depicted there represent the summary curve for 200-Rem dose as ,
in NUREG-0396. One might be tempted, on the basis of the curves shown
) in Figure 5, to set the plume EPZ distance at either 2 miles, 4 miles, or 7 miles, depending upon whether one wanted to use the MACCS, 200 Rem, 3-Puff curve, the MACCS, 200 Rom, 1-Puff curve, or the MACCS, 1-Puff, 0.5 hour5.787037e-5 days <br />0.00139 hours <br />8.267196e-6 weeks <br />1.9025e-6 months <br /> duration curve. Note that there is no dif ference in
)
any of these three curves as to the total quantity of fission products released. The only difference is in the modeling of the release by the consequence model.
The results of the CSNI/OECD code comparison study were published in 1983. 151/ Detailed results from the comparison study are set
) 151/ Committee on the Safety of Nuclear Installations, International Comparison Study on Reactor Accident Consequence Modeling,
)
l
)
forth in a notebook, of . results transmitted to study participants in
) July 1983. 152/ The CSNI/OECD study considered hypothetical releases at hypothetical sites, thus the results are not directly applicable to Seabrook. The results are, nonetheless, instructive as an indication
) of the impact on calculated consequences of modeling differences between the CRAC, CRAC2, and CRACIT accident consequence models.
The source terms employed in these comparison calculations are
) identical for BMR-1, BMR-2, and BMR-3, insof ar as release magnitudes; the source terms differ in timing of release and thermal energy release rates. The release fractions are 100% for noble gases, 30%
for iodine, cesium, and tellurium, 3% for barium and ruthenium, and 0.3% for lanthanide groups. .
153/ Benchmark Problem 7 provides a useful comparison among CRAC, CRAC2, and CRACIT.
Fiqure 6 shows the consequence curves for early deaths for
)
release cateogry BMR-1 with a uniform populat: ion density and emergency response modeling as normally used in the codes. Examining Figure 6, CRACIT results in no early deaths until a conditional probability of
)
Nuclear Energy Agency, Organisation for Economic Co-Operation and Development, Summary Report, September 1983.
(52/ Committee on the Safety of Nuclear Installations, Comparison of Reactor Accident Consequence Models: The Results of the NEA Committee on the Safety of Nuclear Installations' International
) Comparison Study on Reactor Accident Consequence Modeling, Nuclear Energy Agency, Organisa tion for Economic Co-Operation and Development, June 1983, transmitted to study participants by letter dated July 14, 1983, from Daniel J. Alpert, Sandia National Laboratories.
) 153/ Committee on the Safety of Nuclear Installations, Comparison of Reactor Accident Consequence Models: The Results of th e NEA Committee on the Safety of Nuclear Installations' International Comparison Study on Reactor Accident Consequence Modeling, Nuclear Energy Agency, Organisation for Economic Co-Operation and Development, June 1983, transmitted to study participants -
) by letter dated July 14, 1983, from Daniel J. Alpert, Sandia National Laboratories, " Introduction," page 14.
P
)
, about 2%; for comparison, CRAC records at least one early death at a
) conditional probability of about 40% and CRAC2 records this result at a conditional probability of about 80%. Figure 7 shows similar results for a non-uniform population distribution. The conditional
- probability of at least one early death ranges from about 2% for CRACIT to 40% for CRAC and about 80% for CRAC2.
Fiqure 8 shows results for early injuries for release BMR-1 and a
) uniform population. The conditional probability of at least one early injury ranges from about 10% for CRACIT to about 80% for CRAC and 100%
for CRAC2. For a non-uniform population, Figure 9 shows a conditional
) probability of at least one early injury ranges from about 20% for
'CRACIT to 60% for CRAC and 100% for CRAC2.
The results depicted in the paragraphs above and on Figures 6
) through 9 indicate tha t for identical releases, differences in modeling yield significant differences in the probability of causing early fatalities or early injuries. This is further indicated by the following comparison. Examining Figure 6, at a conditional probaility
)
before the CRACIT code shows any early f atalities at all, CRAC2 has already peaked at several hundred fatalities. Similar results are shown in Figure 7, and similar results for early injuries are shown in Figures 8 and 9.
Based on these comparisons, it is evident that at least some of the dif ference tha t PSNH shows in its comparison with NUREG-0396 --
) for which PSNH and PLG claim is related to the strength of the containment --
is nothing more than a difference in consequence modeling. We believe that if the dose / distance relationships for the PLG source terms were re-calculated using CRAC (and interpolating) or
)
CRAC2 the dif ference between the NUREG-0396 results and the Seabrook results calcula ted by PLG would narrow significantly. Whether the
) CRACIT models are more realistic is not the issue here -- what is the issue is the reasonableness and fairness of the comparison being made.
! It must be recognized that the difference between the PLG
) dose / distance curves and the dose / distances curves from NUREG-0396 does not arise solely due to dif forences between Surry and Seabrook --
- l. some of the difference (which we cannot now quantify, but believe may be significant) arises solely from a difference in accident consequence models.
) IV. CONCLUSIONS AND RECOMMENDATIONS This section of the report sets forth our conclusions and recommendations.
) The original EPIs set for the Seabrook site were defined in terms of the NRC's regulatory requirement for a plume EP5 of about ten miles, and an ingestion EPS of about 50 miles. These EPS dis ta nces , first set forth in NUREG-0396, were dervied from a number l of factors, perhaps the most important of which were the risk assessment results from WASH-1400.
l The ' uncertainties in the WASH-1400 risk assessment results, l while discussed briefly in NUREG-0396, were not explicitly addressed on the dose-versus-distance curves in NUREG-0396. Moreover, it is now well-recognized that the risk assessment presented in WASH-1400 is incomplete in that it failed to adequately consider, for example,
[ reactor coolant pump seal LOCAs (a s both initia ting events and consequential failures) , external events (such as earthquakes) , steam
)
l
)
77
) generator tube rupture accidents, and others. It is also well-
, recognized that some severe accident phenomena were either handled poorly or were not recognized at the time the analysis was done (such as high pressure melt ejection and direct containment heating, for
)
example). The calculational tools ' employed in WASH-1400 are, in comparison with the current state-of-the-art, primitive. In addition, it is now well-recognized that some aspects of the WASH-1400 analysis are conservative (i.e., overestimates consequences, frequency, or both) and some are optimistic (i.e., underestimates of consequences, frequency, or both).
) All of this is not to denigrate the achievements of WASH-1400.
Clearly, it was a pioneering analysis, and the application of the methodology first used in WASH-1400 has contributed to a new and
) greater understanding of the nature of reactor accident risks.
. Nonetheless, it must be recognized that the existing EPE .
distances were set based on an incomplete understanding of risk.
) Although this understanding has improved, we are still quite far from a definitive statement on reactor accident risk. This is clearly evidenced by the differences in the assessmentr of various severe
) accident phenomena between, for example, IDCOR and the NRC and its contractors.
The EPZ distances set in NUREG-0396 and adopted by the j Commission represented a considered judgment by the authors of NUREG-0396 of what constituted adequate protection for the public health and safety. All of the bases for those judgments were not fully explained at the time (or since), and those which were were heavily contested by I
i L
)
parties on all sides of the nuclear safety issue as being either too
)
conservative or too optimistic.
PSNH has now requested a substantial reduction in the plume EPZ distance from ten to two miles. In order to assess the reasonableness of this request, it is necessary to go beyond a simple , comparison of
.the dose curves presented in NUREG-0396 and in the applicant's various l reports. One must, in making such a judgment, consider the totality of available evidence, and one must also consider the degree of i
uncertainty in that evidence.
In examining the available evidence, it must be recognized that
) there was insufficient time available to perform a detailed review and requantification of the risk posed by Seabrook in the light of the available evidence. Sufficient time to perform such a task. simply was
) not available. In addition, there is a considerable- body of-information which will become available in the near future in the form of the NUREG-ll50 report and the results of the NRC-IDCOR discussions.
) These reports and results will need to be reviewed in detail, by the Commission, by the industry, and by their critics, before final reports on these ma tters are rendered. The decision paths laid out
, j' i
) for these interactions are by now well-established. The applicant seeks to intervene late in the game and essentially try an "end run" around this well-defined process. The NRC stafT quite properly y rejected an earlier attempt for reduce the plume EPZ for the Calvert Cliffs plant because the process described above was then incomplete.
Although progress has been made, the process is far from complete.
The progress made to date consists largely of a much better definition of what issues are " risk drivers" --
that is, the issues s
l l
l l
l
) about which th ere is great uncertainty, and about which there are sharp differences in results between industry, 'he t NRC staff, and their critics. These are the same issues which have the potential to significantly affect risk estimates toward either much lower or much
)
higher risk.
These issues are rather well-formulated by now. Our assessment in Section III above identifies many of the more potentially signficant issues, such as high pressure melt ejection, alpha-rgode containment failure, and others. The current state of knowledge about i
these issues is such that there is great uncertainty. Different teams of analysts, relying in many cases upon the same data base, can clearly come up with quite different answers to specific risk-related questions.
) -
The, P.ickard, Lowe & Garrick analyses which underly the applicant's plume EPZ reduction request represe'nt, in our view, one end of a spectrum of possible results for a severe accident analysis
) of Seabrook. Considering the uncertainties involved in severe accident analysis at the current time, their results are neither surprising nor unique. There are, however, a range of possible
) results. Not all of these results support a plume EPZ reduction for Seabrook (or any other site) .
Considering this state o' affairs, we strongly recommend that
) the full range of severe accident knowledge and understanding be considered in making a decision on the validity of the applicant's request. We believe that when such an assessment is made, the result
) at this time must be that the uncertainties involved are of such a magnitude that a reduction of the plume EPZ distance is not justified.
)
i .
) In the short review time available for this report, we have identified a number of faults with the PSNH/PLG analyses. In almost every case, these faults are found to potentially underestimate risk
) or, at the very least, to understate the magnitude of the uncertainties that exist in the estimates of risk. The faults we,have l
found are the following:
l b 1. The analyses fail to consider the risk posed by accidents due to sabotage. With WASII-1400-era analyses, it would be possible to argue that sabotage risks are implicitly included due to the relatively high accident sequence frequencies and large source terms posited in these analyses. In the PLG and BNL analyses, the "conserva tisms" in the WASH-1400-era analyses are stripped
) away, leaving a residue which no longer can be said to inplicitly include sabotage risks. Indeed, the very process which is undertaken to quantify risks from other types of accidents provides (properly interpreted) a virtual roadmap for potential saboteurs to follow. This paradox of PRAs is well-recognized, but the fact that it is well-recognized does not
) mean that the issue goes away. It is doubtful that the risk posed by sabotage can be accurately qu'antified'. Thus, consideration of the issue boils down to .two key questions: (a)
Is sabotage a. credible source of severe accidents?, and (b) Can ,
saboteurs cause a large, early release of radioactivity? We believe that the answer to both of these questions is yes, and that this argues for retaining the existing EPZ distances.
- 2. There are large uncertainties concerning the modeling of a key accident sequence, namely Event V (the interf acing LOCA) . PLG has attempted a "first-of-a-kind" analysis which results, on its own terms, in a reduction in frequency by a factor of over
) 2,000 in the likelihood of a large containment bypass accident of this rype. While the PLG analysis may adequately represent a lower-bound estimate, the uncertainties in both the frequency of the accident scenario and the magnitude of the source term resulting therefrom are such that the PLG analysis cannot be considered a best estimate. This is especially significant since it is quite clear that the PLG 200-Rem dose / dis tance
) curve (upon which PLG places great emphasis) is largely driven by the results of their Event V analysis. There are alternative assumptions and models which result in significantly higher sequence frequencies and larger source terms than employed by PLG in their analyses.
} 3, PLG has incorrectly modeled steam generator tube rupture (SGTR) l accidents. Unless a stuck-open steam relief valve is present (to which PLG assigns a very low conditional probability), PLG l
models SGTR sequences as involving an intact containment, when
)
O
)
l in fact they still represent containment bypass accidents.
Radionuclide release will still occur due to periodic lifting of the main steam relief valves; these valves relieve pressure to the environment. gLG estimates the frequency of SGTR i accidents at 1.7 x 10 per reactor-year. We have not been l able to perform a reassessment of this likelihood. The source term for such an accident may involve the rapid release of 10-15% of the core inventory of iodine, cesium, and tellurium, h along with about half the noble gas inventory and minor amounts
! of other species. The offsite doses from such a release are not included in PLG results. Moreover, we have found that it is possible for SGTR sequences to be associated with high RCS pressure at the time of reactor vessel failure. This presents the potential for high pressure melt ejection and direct
) containment heating, which we believe may cause containment failure. Thus, this containment failure source term may be added to the SGTR/ main steam relief valve source term for at least some SGTR accidents.
l 4. The PLG studies did not include the potential for induced tube h ruptures for sequences in which RCS pressure remains high during core meltdown while the secondary side of the plant is depressurized. PLG presented arguments to show that such induced tube ruptures cannot occur. We have cited other sources (Sandia National Laboratories) which indicate that such failures are not only possible, but relatively likely. The frequency of accidents in which induced tube rupture would be
)
of concern is estimated to be about 47% of the overall core-melt frequency. It is estimated that 10% of these accidents woul'd be associated with induced tube ruptures, for an overalf frequency of induced tube rupture accidents of about 1.1 x 10 per reactor-year, or about one accident in twenty. The source term and potential for high pressure melt ejection and direct
) containment heating are analogous to that described above for SGTR sequences.
- 5. Seismic accident sequences are estimated by PLG to contribute about 13% of overall core melt frequency, and about half of this comes from earthquakes with ground accelerations in the
) range from 0.7g to 1.0g or higher. Such accident sequences will be associated with significant offsite damage which will impair emergency response. In addition, we have identified the potential for aftershocks to occur during severe accidents in progress. An unanalyzed issue, but one which is acknowledged by a number of analysts to be potentially important, is whether
)
the combination of the existing internal pressure lead on the containment (from the accident in progress) and a seismic load from an aftershock wculd be capable of causing containment failure. It is known that the seismic fragility of the containment will be lowered by increased internal pressure, but the precise relationship has not yet been determined for Seabrook. Such a containment failure mode could be an
)~
important cause of early containment failures if the fragility
__ a
D of the' containment is sufficiently lowered by increased D' internal pressure and if af tershocks are relatively likely.
- 6. High pressure ejection of molten core debris from the reactor vessel during a severe accident can lead to direct containment heating and present the threat of containment failure at a time of significant radioactive aerosol loading in the containment.
3 We have concluded tha t PLG's modeling of high pressure melt ejection and direct containment heating is incorrect, and may result in a significant underestimate of risk, in part because their analysis is based on fundamentally flawed experiments.
Other more credible experiments indicate that high pressure melt ejection and direct containment heating present a severe threat to containment integrity. High pressure core melt
) sequences are estimated to account for about 98% of the overall Seabrook core melt frequency. Thus, unless the likelihood of containment failure for these conditions is very low, high pressure melt ejection and direct containment heating may be the most important cause of early containment failure, and may completely dominate the dose / distance curve for Seabrook.
- 7. The containment failure pressure estimated by PLG for Seabrook is very high compared with other plants. Other analyses of the Seabrook containment suggest that the failure pressure estima ted by PLG is too high. If Seabrook is examined using methods similar to other plants, the failure pressure is very
] close to that of Zion.
- 8. The secondary containment for S'eabrook is expected to play little or no role in mitigating severe accident doses.
- 9. The probability of a steam explosion causing containment 3 failure (the so-called " alpha-mode failure") is es timated by PLG to be very low. We have found that there is a wide variety of opinion as to the probability of alpha mode failure.
Various estimates of this probability range from impossible to a 10% probability given a core melt accident. There is also dispute over whether the same probability should apply for high pressure sequences as for low pressure sequences. Based on
] three different models, we have estimated alternative alpha-mode failure probabilities which, together with Seabrook accident likelihoods, result. in alpha-mode failure release frequencies ranging from 1.2 x 10-7 to 2. 4 x 10-5 per reactor-year. At the high end of this range, alpha-mode failures may dominate the risks posed by Seabrook. Even at the low end of D this range, they may affect the risk curves calculated by PLG.
- 10. PLG's source term estimates have not accounted for vola tile forms of iodine which may be created during an accident. The failure to consider volatile iodine forms is most significant for those accidents in which containment failure is estimated D to occur late. PLG has also failed to account for the results of experiments which indicate that cesium iodide (the assumed form of iodine in the PLG studies) is changed by hydrogen burns e
( .
in the containment. atmosphere. About 20%' of the cesium iodide i may be converted to the very volatile molecular iodine. There are also uncertainties about whether cesium iodide will be the dominant form of iodine. Some recent experiments indicate that cesium iodide dissociates under gamma radiation exposure to form volatile iodine species.
t
- 11. 'The PLG analyses fail to account for phenomena which can cause l re-evolution of deposited fission products a late times in an accident sequence.
These phenomena are currently under study, but it is well-recognized that source term estimates made to date may have to be revised upward to account for these phenomena.
)
- 12. Both PLG and BNL have performed preliminary analyses of the risk posed by accidents during plant shutdown conditions.
Neither of these analyses, however, have considered accidents during shutdown conditions caused by earthquakes or other external events. Since Seabrook may reasonably be expected to
) operate about 70% of the time (or less), and since seismic events occur at random, the failure of PLG and BNL to consider seismically-initiated accidents during shutdown conditions may be significant.
- 13. PLG assumes that if AC power is recovered, containment sprays will successfully depressurize the containment and terminate a
) severe accident.- In contrast, other studies show tha.t under these _ conditions a very large hydrogen burn can occur --
referred to as a "de-inerting burn'. Such hydrogen burns may be capable of causing containment failure, but were not-analyzed by PLG or BNL.
I 14. The " Peer Review" of PLG-0432 which was sponsored by PSNH does not represent peer review as that term is normally used within the PRA community. A true peer review consists of a lengthy and detailed assessment of the models, data, assumptions, and -
methods used in a PRA, along with alternative calculations, confirmatory calculations, and sensitivity and uncertainty
) studies. The " Peer Review" of PLG-0 43 2 does not conform to this definition of peer review. Moreover, although the asserted containment strength of Seabrook was relied upon by the " Peer Review Group" for its conclusions, none of the " Peer Review Group" members were structural engineers capable of performing a technical review of the design of the containment.
- 15. Careful examination of the dose / distance curves presented by PLG indicates that when the Environmental Protection Agency
" Protective Action Guide" (PAG) doses are considered, there is no basis for reducing the size of the plume EPZ.
- 16. PLG's analyses have ignored the decay of tellurium-132 to
) iodine-132 in their assessment of source term magnitudes.
Although their accident consequence code considers this decay after release from containment, the containment analytical
)
)
-84 _
tools used by PLG do not. There is a large amount of tellurium-132 released from the core in an accident, and decay of this material to iodine-132 can provide a large airborne I
iodine source very late in an accident sequence.
- 17. PSNH has not requested or justified a change in the size of the ingestion EPZ. This zone must remain at 50 miles in radius,
- 18. Using PLG's calculated dose / distance relationships (from the CRACIT cod e) , it is clear that for nearly l all release categories postulated by PLG (whether " realistic",
"conserva tive", or " WASH-1400-type") , the EPZ Protective Action Guide doses for whole-body exposure are exceeded with a D significant probability out to distances of ten miles or more (with only a single exception) .
- 19. Moreover, there are significant modeling differences between the CRAC code (used to generate the dose / distance curves in NUREG-0396) and the CRACIT code used by PLG to generate . the 3 Seabrook dose / distance curves. We conclude that some portion of the difference between the NUREG-0396 and PLG curves is due solely to accident'conseq6ence code modeling dif ferences, and
~
has no thing whatsoever to do with real differences between Seabrook and NUREG-0396.
D D
D D
e 9
3-
-85 _.
TABLE 1
) COMPARISON OF SOURCE TERM CHARACTERISTICS FOR PWR ACCIDENT RELEASES REPRESENTING EARLY CONTAINMENT FAILURE Release h Characteristic SSPSA, S1 WASH-1-4 0 0 , PWR-1 Sandia, SST-1 l Release Time (hrs) 1.4 2.5 1.5 Release Duration (hrs) 0.5 0.5 2.0 Warning Time for 7 Evacuation (hrs) 0.3 1.0 0.5 Release Elevation (m) 10 25 10 RELEASE FRACTIONS:
Xe-Kr Group .94 0.9 1.0 Iodine Group .75 .70 ,
.45 Cesium Group .75
.40 .67 .
Tellurium Group .39 .40 .64 Barium Group .093 .05 .07 Ruthenium Group .46 .40 .05
-4 -3 Lanthanide Group 2.8x10 3.0x10 9.0x10 -3 SOURCES: SSPSA, Table 11.6-3; WASH-1400, Main Report, Table 5-1; Sandia Siting Study, NUREG/CR-2239, 2.3.1-2.
)
)
)
)
)
- -__ _ _____________ __ ____ 1
l
)~ ,
TABLE 2
)
COMPARISON OF EVENT V SOURCE TERMS Fission Product PLG-0465 NUREG-0956 SSPSA, S6V Xe-Kr Group .93 1.0 0.97
) , ,
Iodine Group .094 0.3- 0.43 Cesium Group .092 0.3 0.43 l Tellurium Group .083 0.06 0.40 Barium Group ~4 2.3x10 0.004 0.048
) Ruthenium Group 3.8x10
~4 2.0x10 ~7 0.033
~4 Lanthanide Group 3.8x10 3.0x10 ~4 0.0053 f Cesium iodide
) **
Cesium hydroxide , ...
Assumed based on other NRC-calculated source terms SOURCES: PLG-0465, page 4-82; NUREG-0956, page 4-53 (BNL results);
PLG-0300, page 11.6-33.
)
)
)
)
).
) TABLE 3 -- CONTAINMENT DESIGN COMPARISON l
j Design Parameter Seabrook Zidn 1 Millstone 3 Reactor Thermal Power (MWt) 3650 3250 ,
3411 Containment Type RC Prestresse,d RC/Sub Containment Free Volume (cubic feet) 2.7x10 0 2.73x10 6 2.3x10 6 Estimated Con-(J tainment Failure Pressure (psia) 211/187 150 142-146 SOURCES: NUREG/CR-4540,.Eebruary 1986, page 5; Millstone 3
) PRA, page 9-2.
)
)
)
i
, l Table 4 PLG Source Term Dose / Distance Relationships (With No Protective Action)
). .
Conditional Probability of Exceeding'l Rem Dose at x Miles PLG Source Term Des'ignation 1 Mile 2 Miles 5 Miles 10 Miles 20 Miles j S2-BM 100% 100% 70% 30% 7%
S3-BM 100% 100% 30% 9% 0.8%
S6-BM 100% 100% 80% 50% 2%
S7-BM 100% 100% 60% 40% 5%
Sl-CM 100% 1,00% , , ,. ,
80%. 70% 50%
S2-CM 100% 100% 80% 60% 9%
S3-CM 100% 100% 50% 40% 10 % '
S6-CM 100% 100% 90% 80% 50%
S7-CM 100% 100% 70% 60% 50%
S2W 100% 100% 100% 90% 60%
)
SlW 100% 100% 80% 80% 70% 1 S6W 100% 100% 100% 100% 80%
S7W 100% 100% 50% 30% 3%
b S3W 4% 4% Not Rep. Not Rep. Not Rep.
)
)
l 1
) .
1 Table 5 l
PLG Source Term Dose / Distance Relationships (With No Protective Action)
) Conditional Probability of Exceeding 5 Rem Dose at x Miles PLG Source Term Designation 1 Mile 2 Miles 5 Miles 10 Miles 20 Miles S2-BM 100% 90% 3% 0.3% Not Rep.
S3-BM 70% 50% 4% 3% Not Rep.
S6-BM 100% 90% 10% 2% Not Rep.
S7-BM 90% 90% 20% 7% Not Rep.
Sl-CM 100% 100% 70% 60% 20%
S2-CM 100% 100% 20% 3% Not Rep.
S3-CM 90% 80% 50% 5% Not Rep.
D
,S6-CM 100% 100% 70% 40% 3%
. S7-CM 100% 90%~ 60% 50% 4%
S2W 100% 100% 70% 40% 8%
SlW 90% 90% 70% 70% 60%
S6W 100% 100% 100% 90% 50%
S7W 70% 60% 10% 5% Not Rep. -
S3W Not Rep. Not Rep. Not Rep. Not Rep. Not Rep.
)
) .
J
%e e- ev -w,v-2'= m*'"w w **
v n*
yE t v H E
T N
- E C
o,G r,N N aoI S O
E S E
T "oS fE E C
/ L P N S M E
",C
\ M PO A U R O AS E XER L
P
^% PR ERA g k c a
U m i
\
I d/
/
s e
I n
\ o U
\w / N4 Z g
n in
~
d a
n
~ l P
y c
n e
' g U N E r
e m
N \
1
!3
/
/
f t
o p
e c
n g
\ C o
U , 1 N }a h
,o er u
g N i A F N B R
U g
s g
.u U
~
~ N
' t p
(
E E U N N E N
l O O O Z Z Z M I
G N G 1 N A NU N E ,
/E O 4 A I
N l
a t
M l l
v N S .t N u A R A L AhM L t 0 t
p g
g A N
U P
6 y g g
'* _ O W 4 EoV _
I S
f U E g L G C g X E
M(
U
)
f FIGURE 2
< iOm . . . . . .
) _
~ ~
r.aso +
RATIO OF DOSE FROM INDIVIDUAL h GROUPS TO DOSE FROM NOBLE GASES -
\ NORMALIZED TO I AT 5 MILES N
N 3 joo _ 2 HR DECAY -
1.0
) ' nuclear plant
'y "
~% 2
- ]'- -
%*x"S*"[,, _ 24 HR DECAY life saving b - -
, activities , ,k ,
? M \
l < _- N ~
C
\ 2HR DECAY c
$ \ 4 8I \ -
0.1
) z T N 95th PERCENTILE DOSE -
- \Q00% NOBLE GASES RELEASE-N(\
rEPA Protective 'Y - _
Action Guideline -
, limit for pro- U _
N g 24 HR DECAY _
) }jecteddosestoj Q the general public N _
emergency planning 1 zone distance (; }
j l I l l l l 0 2 4 6 8 10 12
) DISTANCE (MILES)
FIC. II.B.1. The spatial dependence of the ratio of dose from individual groups to dose from noble gases inormalized to 1 at 5 miles) (CRAC2 Calculations).
)
)
) .
) FIGURE 3 I e 8 4 4 558g 4 5 5 5 5 4ILg 4 1 5 5 3 4 5 8
) .
~
! O " ~
1 REM
)
"8 c
0.1 -. 5 REM --
Q "
"i 9 e8 - -
g< . -
) o[
z m, 52 . -
U$ 50 REM uo .
UU m<
c2
) D $ 0.01 30 E . t 200 REM -
~
)
~
.r.-
' ' ' I ' ' ' ' ' ' ' ' ' ' ' ' ' '
0.001 1 10 -
100 1000 DISTANCE (MILES)
Figure 111. Conditional Probability of Exceeding Whole Body Dose Versus Distance. Probabilities are Conditional on a Core Melt Accident (5 x 10-5).
) Whole body dose calculated includes: external dose to the whole body due to the passing cloud, exposure to radionuclides on ground, and the dose to the whole body from inhaled radionuclides.
Dose calculations assumed no protective actions taken, and straight line plume trajectory.
) .
)
l FIGURE 4
)
I ?
! 4, 5
, 8, .8891 .
P .
S. 8 .#885 I.C ~, , , , , , ,
a . . . ,
ji x . , .
s ,m . , , . . . , . , . u . m . , ,, .
/ . ,a_..,...\..a ., u. r - .- .
' r u . s.. .
g 8
"<7: s:.( ..qar , ; .g!1@ - . 6 .u :.18 : r w # 1 A i e .* i et ' g-pt= Nh'i_*_4 :g:hg .::;$ r. .i.=:.4.:y ;;j 4,l.,.p.-pq O 8.\, .
, _xT., . .\ .ii . . . .
- 4 .6 . v. i . ;, . N e . -, +- .i s r rr8.76 -
, f~4\' y -
A -\ 1 .iit - se \ :-f --t .: t t \ . . . . i, - is s p A il 8 4 a' :
E == 'W4*\:* W :'R paie ,1. Mik'
) en 8
==Wv=.n=.!
. . . ..\ q.;-kii .:'.31. - . ..l y\ 1 -- .it. _p; -.l4. _.:s. 4.;
3 _. _._ _A s;_3 .t.. u 3
-- - i g.. . . .=n" 7-'f.j. l ;
g g.. _ . . , .;;
. . , h,..v.;--
o
._. 4 . __ . .e e, % o, . . A. . .
an.
4 1: 1 7 i e e 3 j__ s 1
,.,v . t .
a 1
8,- r. .hs -sn 3. . w. -g , .
1 . . i . . !
E '
) ' a =>>m x i n i m a m. ..& . .-r- i . - i i i .,
I I, \, 2-. =J=la -=-4N.fi) .iel.4, - *.t + : L 1-1-4
. . , .\t- , 1 4.. s_s . o . ._ . . .
Q l . \ .{Q s t -, r.--i : L - 9 _9 {'. J .6 l 4 a,a as * : rs- ! ' e6 4 3 ' '
0 1. ) .-d!Ei .'. i Al- hs fAl i 1 .. -
u M ~_..' :Fi-=%MNIIIHit: "t*i-I-+-T - =- kd}
h N.hh 'kNb! E---N.W;;{$$?
~
o
) =4 r, --~: ;.-9 .x= q +-~w . '
v -
g:~ % -
, .ib-v - .
rx
,n e,. -
+
- i. , s . , i11 , s. , . >
! .i 1 % i ' L 1 P g I 4
1 -l e o
,,..t-g ., , ., . , . . . ,-6 . ;1 v nL
....-# s e .
8
-.,,.c...,..e .. ig . ..w g . , . . . ,
WMM fa[b "l :
-4 *M:' T.4." O Il
% M$15 \" 3- I '+ b i s.s 5
,A,g. g _. ---t. m :. -.
g, g .u u p .' L :.: a-- . 4 r-8
.-.g.. . ..p..s .s _> , -
- i,Xg a wl 1.. .. 4.,.
.a,s Hi . g. m%.:'- tT 1 s-'.* \\ ; ;--A. ' 8 0 ' t, i,-
-- M -
M'si-4 LM2- - Ela:!!!~~ C M T. a.R M 8N I C 5: A@+
I[' k(~?&NYN T=$E-l~ ~ - ~~b Y-15 U N fr." NU -
-dl--,_-.
, _gg - , - .- --, e - t tu .. u=) b,3 ,,,JA .
- r- g4_],,d; d,, ,.f*._..
f -
-~ , ,
) .. Q.y.g-.' --.. -.
y
.o on I so too
, M lt.E s
) Figure 6.1 Components of NUREG-0396 curv'e as computed by BNL using CRAC2. The sumary curve is normalized to 6x10-5 core melt probability.
The result differs from NUREG-0396
)
)
)
FIGURE 5
) ,. i .' .
r i :
' N '~
. j a , '
)
- is ..
-l
_ y ite
\ fuFp
- g. - te
-t
= ;
..s w -
p h o \
==
<l.
M II l
- -2 I "
a 18 gg
-f
. 3 I i P
e : \ . : :
0
- d N
.a N, b
$,,-5 : .
m" ,o :, .
J. .~ # d M % 2,o A em v.te
-f Ag
-t a m206.88 REM *.fvFP l
. +56.48 REN '
S 5.88 NEN
- as ,1.08 REN
) is .- /" .
,, 4 se-t .'ne i i
st isa DISTANCE CMILE53
)
Figure 6.6 Dose versus distance curve for release category S2W from Seabrook for no frmediate protective action with BNL results using MACCS superimposed.
) -
) .
FIGURE 6 1.00E*00b -
- .~-
a w
)
i '
i 1
. c FIN o KIK (
. NOR l
1
, PLG
\ RSS 1.0CE-01 7 x
. . , SAN
). -
.: =
'3 2
5 g (
3
+ _- --
l . 00E -
.r-
)
P
)
EARLY DEATHS (-
1.00E-03 -
! I <
l.00E-00 1.00E*01 1.00E+02 1.00E*03 1.00E-04
. Fia. 10.5.1 PROBL,EM 7.9MRI UNIFORM POPULATION
)
) -
FIGURE 7
)
1.00E-00 __
r-A i
\
1.0CE-01 7 b FIN l ,
o KFK a NOR
, PLC
) x RSS
.* = _ _
- }
1.00E-02 ; \
C 3 ,
E
) E
- 1. 00E-03 e-o-
)
EARLT DERTHS
) 1.00E-04 '
1.00E+00 1.0dE*01 1.0dE+02 1. 0 d.E-03 1.' ode *04 1.00E-05 Fia. 10.5.2 PROBLEM 7.SMR1 NON-UNIFORM POPULATION
)
FIGURE 8 1.00E+00 ,
- x -
)
i t
- 1. 00E -
- O (
5 .:
2 e
8 m
c FIN
- 1. 0CE -
, PLG x RSS
. SRN ,%
(
- 1. 00E-03 --
)
EARLT INJURIES
)
1.00E-04 ' '
- 1.00E+00 1.00E*01 1.00E+02 1.0dE+03 1.00E-04 1.00E-05
(
Fla. 10.5.7 PROSLEM 7.8MRI UNIFORM PO.*ULATION
)
)
r
)
I
) FIGURE 9
) 1.00E+00 E i
x_ .
I h '
i
- 1. 00E-01 -- *
) -
O
) -- r .
! . 00E-02 -- : "'N* l
{
, PLG x RSS '
g , SAN l ,E d *
.r.- ..
m e
c Q.
1.00E-03 1
- p. ~:
^
ERRLT INJURIES 1.00E-04 ' '
1.00E-00 1.00E-01 1.00E*02 1.00E+03 1.0CE-04 1.00E-05 Fla. 10.5.8 PR SLEM 7.5MRI NCN-UNIFORta P C.:UL A T
- 0'.
- ' ~ L * =." a ~J :
Af$9)D[K j
'O
?_*E
"* SEABROCK STAMON 1
Engineering Office t
i::, .' . J 3y gg October 1,1985
_g mWWM6 53_19933 steusHempeNre Yankee DMeien T.F. V2.2.4 1
No Response Requested 1
Peer Review Group (see Attachment B) ,-
EPZ Optimization Study / Peer Review Group
Dear Sir:
i O This letter confirms verbal requests for your participation in a Tcchnical Peer Review Group regarding the subject study for the Seabrook Station. As previously discussed with you, tne review and foMow-on meeting win occur in October. We are now confident that the fonowing dates will not change: ,
.O October 8,1985 Peer Review Group will have Technical Report October 30 and 31, 1985 Peer Review Group Meeting (Attachment A) i The review meeting proposed agenda is provided in Attachment, A. The agenda is subject to change as determined by the Peer Review Group. The proposed agenda anows for eachnical presentations to the Peer Review Group on O the first day. The Peer Review Group would question the technical authors in data 11 both days as required. The Peer Review Gr'oup is expected to draft their findings the second day and a Draf t Report is scheduled for November 8,
( 1985.
For your convenience, Attachment 5 contains a list of individuals whom O you may want to contact during your review and preparation for the meeting.
If you have any questions, pluse feel free to can Jim Moody at (617) 872-8100.
l i In addition, please send a copy of your resume and a fee schedule to Mr. R. W. Romer (see Attachment B).
1 0 ~
-Very truly your ,
.+
. 8-1 #
ohn DeVincentis, Director lQ ,
Engineering and Licensing cc K. Fleming F. Torri .
K. Woodard B. Lutz B. Henry J. Moody K. Kiper R. Romer 10 1
l P O. Box 300 Sectrock.NH C3674 . Telechone (603) 474 9521 1
O
. __ . - . _ _ _ _ _ - - - - __ ._____ __ _