ML20099J512

Applicant Exhibit A-149,consisting of NUREG/CR-2723, Estimates of Financial Consequences of Nuclear Power Reactor Accidents, Dtd Sept 1982

ML20099J512
Person / Time
Site:	Limerick
Issue date:	05/22/1984
From:	Strip D SANDIA NATIONAL LABORATORIES
To:	NRC OFFICE OF NUCLEAR REGULATORY RESEARCH (RES)
References
CON-FIN-A-1334 NUREG-CR-2723, OL-A-149, NUDOCS 8411290093
Download: ML20099J512 (22)

Text

NUREG/CR-2723 SAND 82-1110 Printed September 1982 ('hClf,g

, o 0, 6

[ j, [7 v ^ ,

, c3 , ,_ >

%y2x "@ uE <  :

Estimates of the Financial 4 Consequences of Nuclear Power *DWO Reactor Accidents David R. Strip Y A i t m ionee Atumergue, New Memco 87185 and Uvermore Califome 94550 for me 4.hted States Departnent of Energy under Contract DE-AC04-760P00780 NUCLEAR REGULATORY COMMISSION Docket No,b?>D 3@ 0 f; cia! Esh. No ' *k in the m.mer et _(MQ\ fig \un_ w sw, _.

wwoJ -

Applicant _ .' ' in Ir:cy :ncr _ _ _ _ _._ _ _ C: ct 3 _ _ _ _ _ _,

Cant'a .r '- __ ._. _ _ _ _ .  ;

Ccw .1a ___ .__ _ ___ _._ _

• ~

fj 21.

Other . ue

~

Reporter h __.

S%ATo88Po!888l!!

O Prepared for U. S. NUCLEAR REGULATORY COMMISSION

t ,

l O! j l

NOTICE '

This swpor. was prepared as an account of work sponsored by an agency of the Uniend States Gosernment. Neither the United '

States Government nor any agency thereof,or an of their employ-or answnes any >

ees, makes any warranty ^, empressed or im

- '"- for an party's use, or the legal resultsliabdity or . ,of any of such use, apparatua product j

process disclosed in this report, or represents that its use by such ,

1 third party would not infringe privately owned rights.

r.

' a

+

Available from i GPO Sales '

Division of T Information and Document Control U.S. Nuclear Co==W '

Washington, D.

and National Technical Information Service g4 Springfield, Virginia 22161 1

^

O a

4 I

, e F

.g l

l I

i

)-

NUREG/CR-2723 SAND 82-1110 Printed Septcmber,1982

. Estimates of the Financial Consequences of 1._

5 ^ Nucient Power Reactor. Accidents David R. Strip Safety and Environmental Studies Division 9415 Sandia National Laboratories Albequerque, New Mexico 87185 o

0 ABSTRACT This report develops prehminary technique for estimating the Snancial conse-m quences of potential nuclear power reactor accidents. Ossite cost estimates are based on CRAC2 calculations. Costs are assigned to health efects as well as property damage. Onsite costs are estimated for worker health eSects, replacement power, and cleanup costs. Several classes of costs are not included, such as indirect costs, socio-ecommic costs, and health care costs. Present value discounting is explained

,C and then used to calculate the life cycle cost of the risks of potential reactor acci-dents. Results of the Rnancial consequence estimates for 156 reactor-site combina-

'( tions are summarised, and detailed estimates are provided in an appendix. The results indicate that,in general, onsite costs dominate the consequences of potential accidents.

p PreppN for

' Division of n.;k Analysis U.S. Nuclear Regulatory Commission Washington, D.C. 20555 Under Memorandum of Understanding DOE 40-550-75 NRC Fin No. A1334 a

5 s'

- . . . - - - , - - . . . , _ , . _ . _ . . . _ , _ _ . - - , - - - - . . _ - - . . ~ . _ . . - . , - . - - - . - , - , , - - - - - - . . . . . - . - - . -

- - 1. Introduction

[

\ . A number of recent developments have created an increased awareness of the need for the ability to evaluate the Anancial consequences of reactor accidents. Included in these is the Nuclear Regulatory Commission's current eforts to develop a set of safety goals which include an ALARA (as low as

. reasonably achievable) criterion, which dennes the nature of cost-eSective improvements, and therefore requires the ability to measure the degree of improvement. Even before such a criterion is adopted, Anancial risk analysis tools' are valuable additions to the decision making process for evaluating cost-i- benett tradeofs for proposed design requirements or backSt proposals. In addition, recent experience at -

- Three Mile Island has focused attention on the potential for tremendous Anancial consequences in even minor accidents (at least when minor le deined in terms of health eSects). In a recent paper, Starr and

-Whipple [1] explore the possibility of using the utilities' Anancial self-interest as a basis for increased

. - cooperation with the NRC and a less adversarial relationship with the regulatory body. .

In order to further explore the potential role of Saancial risks in the regulatory process, it is necessary to have a better understanding of what these risks are. In this report we will examine the Anancial consequences of potential accidents at existing nuclear power plants. The estimates of health

. consequences and oshite Anancial consequences will be based on CRAC2 (Calculation of Reactor Accident

, Consequences, Version 2) predictions which were made to support a recent project which evaluated the

, impact of alternative siting criteria [2]. Costs for replacement power are based on studies ongoing now at Argonne National Laboratories.

The methods developed in tais project, and the results presented in this report have a number of potential applications. Value/ impact analyses are playing an increasingly visible role in decision making in the regulatory process, and there is a specine, immediate need for simple value/ impact analysis tools for support of the decision making processes la the Severe Accident Rulemaking. In addition, a means of value/ impact analysis is critical to the practical implementation of an ALARA criterion based on cost, such as the one proposed in'the ACRS (Advisory Cosamittee on Reactor Safeguards) safety goals [3].

l The results presented in this report can be used to gain a better understanding of the reistive importance 7

- of the contributors to Anancial consequences. In addition, the information on the range of consequences l can be usefulin a reevaluation of the liability limits of the Price-Anderson Act.

As in all studies of this nature, it is important to point out the large uncertainties in all results. The

[T uncertainties in the CRAC2 code have been discussed at length in other places [2,4]. In addition, source l-p terms, wi.ich play a signi8 cant ro'e in determining the consequences predicted with the CRAC2 model, are the subject of considerable discussion [5]. Studies examining the' sensitivity of results to changes *

- of the magnitude proposed by Rahn and Levenson [2] indneste that oskite consequence predictions could change considerably if the current source terms are discarded in favor of the new, smaller ones proposed. In additio't to the uncertainties in the CRAC2 predictions, additional uncertainty is added in the estimates of onsite efects, such as cleanup and replacement power costs. These uncertainties will be .

quantlSed to the extent possible, and sensitivity studies will h: lacinded to help judge the impact of the uncertainty, We begin in the next section with an overview of the CRAC2 code and the prediction of otsite consequences. The following section will discuss the methods used for converting the CRAC2 results

!- 'I into the form appropriate for use in this study. Section 4 explains the use of present value discounting and presents the formulse for discounting used in this study. The next section discusses the assignm_ent -

of dollar values to health efects. In section 6 we discuss the estimation of onsite consequences. Section 7

~'

is a summary of the results. In section 8 we examine the sensitivity of the results to assumptions made in the course of the calculations. This section is followed by an appendix containing estimates of the Raancial consequences at all sites which are currently operating or hold construction permits.

2. Offsite Consequences j The otsite consequence analyses presented in this report were calculated using the CRAC2 [4]

code, an improved version of the code origins!ly developed for the United States Reactor Safety Study O

• 1 T

~&+---r9---s-1 e n- T-pup gee av.4.--S'W=p+my--w. gar - - m. gr p- .3% emgg.y.gp-.-m-am"TW'_m*d&e-'N- m w+ W -'h= M -y 'enw-w-W='rw=' vff e i'r T-e'r *W9'wwT=-'eeew'we*M'u' v 9 W *Nw & WT 4"t"1--*de"T"Wum"-tr

s (RSS). CRAC2 employs a straightline Gaussian plume model to represent the transport and dispersion of radionuclides refased in reactor accidents. The model allows for changes in weather (based on hourly observations) during the transport of the plume; however, it assumes that wind direction remains constant (hence, straightline). Radiation dose to the public is based on both external exposure from airborne and deposited radionuclides and internal exposure from Inhaled and ingested radionuclides. Duration of external dose is determined by the evacuation scenario defined by the user, which allows specification ,

of both evacuation and sheltering zones, as well as unprotected areas. Internal dose duration is the remainder of the life of the exposed Individual.The manner in which health effects are calculated from dose will be described in detail in a later section.

In addition to calculating health effects, the CRAC2 model also provides estimates of offsite economic consequences. The data on which the economic consequence estimates are b& sed are generally detailed I only down to the state level. Thus, particular industries or areas of economic activity in the vicinity -

of a plant are not taken into consideration. All costs are expressed in 1980 dollars [2, Appendix A].

h Economic consequences not included in the CRAC2 code are all onsite costs (capital loss, replacement k power, cleanup), economic costs of health effects, costs of litigation, and indire t costs (such as shut down of adjacent or other reactors, loss of industrial capacity and jobs, etc).

CRAC2 results are based on the simulation of a number of weather sequences sampled according to a statistical model. (A comparison of the'efhetiveness of the CRAC2 sampling procedure with the sampling procedure used in the RSS version of CRAC can be found in [6]). The results from these sequences are combined to generate a distribution of results, which is frequently presented as a complementary cumulative distribution function (CCDF). In this report we will present all results as means (averages) of the distributions estimated by CRAC2.

The CRAC2 results presented in this report are from calculations performed as part of a project ,

to support NRC activity on reactor siting criteria [2]. Calculations were performed for 91 sites in the United States which had reactors with operating licenses or construction permits. Many of these sites .

have more than one reactor, and therefore the results presented in the appendix to this report show 156 reactor-site :ombinations. The CRAC2 runs used the actual population distributions surrounding the sites. based on the 1970 census, and a wind rose recorded at the site over a one-year period. All persons within 10 miles of the reactor were assumed to evacuate to 15 miles (at which point they are assumed to receive no additional exposure) at a speed of 10 miles per hour after a delay of either 1, 3, or 5 hours5.787037e-5 days <br />0.00139 hours <br />8.267196e-6 weeks <br />1.9025e-6 months <br /> (with weights on the delay times of 30Fo,40Fe, and 30Fo respectively). Because of the 1

difBculty in obtaining consistently high quality hourly weather observations for all the selected reactor sites, a surrogate meteorological record was derived from data collected by the National Weather Service at a station with meteorological conditions similar to those at the plant [7]. Since the objective of the

' project for which these CRAC2 runs were performed was to provide guidance in siting, all CRAC2 runs assumed an 1120 MWe PWR reactor, not the reactor existing at the site. Thus, ths results calculated in these CRAC2 runs are not directly representative of the actual potential consequences of an accident at the site. In a following section we discuss how we scaled these results to derive an approximrtion of potential consequences for the actual site reactor.

[ An accident spectrum consisting of five accident groups, ranging from a gap activity release to a

large scale fuel melt with large atmospheric release, was developed by the NRC to represent the range of fission product release in potentall accidents (8). These groups are

e Group 1-Severe core damage. Essentially involves loss of all installed safety features. Severe direct breach of containment. (similar to PWR2) f e Group 2---Severe core damage. Containment fails te isolate. Fission product release mitigating l

systems (e.g., sprays, supression pool, fan coolers) operate to reduce release. (similar to PWRb) g e Group 3-Severe core damage. Containment fails by basemat melt-through. All other release l

mitigation systems have funtione.1 as designed. (similar to PWR6) e Group 4-Limited to moderate core dr. mage. Containment systems operate but in a somewhat degraded mode. (similar to PWR9) 2 I

e Group 5-Limited core damage. No failures of engineered safety features beyond those postulated

' by the various design basis accidents are assumed. The most severe accident in this group includes substantial core melt, but containment functions as designed. (an order of magnitude smaller than

~(

PWR9) ~

For the purpose of decision making such as in siting and emergency response, the NRC has denned Ave releases, denoted SST1-SST5 (Siting Source Term), to represent the Sie accident groups [2]. By assigning appropriate probabilities to the Ave source terms, this set of releases can be used to represent the risk from any current LWR design.

CRAC2 calculates three mWor classes of public health eSects. These are: early fatalities, early 4 injuries, and latent cancer fatalities. In addition, CRAC2 also calculates thyroid and genetic efects, which will not be discussed in this report. Early fatalities are estimated on the basis of exposure to

& bone marrow, lung, and gastrointestinal tract, and are observed within one year of exposure. Bone

-marrow damage is the major contributor. The risk of early fatalities for a given dose is determined by a dose-response curve. The dose-response curve used for the calculations AM here is taken from the RSS, and is called Curve B, which assumes support!ve treatment of exposed ladividuals. (Supportive treatment indicates procedures such as reverse isointion, large doses of antibiotics, and transfusions of whole-blo'o d packed cells or platelets.) Dose-response curves are deined for both greater and lesser degrees of treatment. However, it is believed that adequate facilities do not exist to supply the greater

' level of treatment required to justify the " heroic esort" response curve, while the minimal eBort response curve assumes thst only standard hospitilisation techniques are used to trent the exposed population.

Curve B has an LDaefoe (dose to bone marrow at which 50% of & exposed papulation is expected to

- die within 60 days) of 510 reds. All persons exposed to greater than 615 rads are assumed to have a

- 100% mortality rate, and no deaths are assumed for persons receiving less than 320 reds.

Persons receiving large doses who do not die are subject to early injuries, which are deined as non-

- fatal radiation induced illnesses requiring medical attention or hospitalisulon, and include prodromal vomiting, skin illnesses, and immunological system impairment. These early health essets are estimated

- on the basis of early dose to b whole body,lang, and gastrointestinal tract. Whole body dose dominates the efect. Rate of essets is determined from dose by.means of dose-response curves, as were early fatalities. The dose-response curves for early injuries were also drawn from the RSS.

n The Anal health eseet we are concerned with is latent cancer fatalities. Latent cancers are based on early and chronic dose, and are assumed to have a ten year latency period followed by a pened at risk for the remainder of the ladividual's life (except leukemia which has a 30 year plateau). The dose-response

- curve for latent cancer fatalities is taken from the BEIR I report, and is linear in dose. Dose efectiveness factors are used to reduce the eSect of low doses in a manner similst to the linear-quadratic model of dose response.

Two other CRAC2 results will be presented for each reactor site combination in & tables accom-panying this report. These are person-rem and property damage. Person-rem is the total population dose commitment, expressed in rem, from b postulated accident. Property damage, which is measured in 1980 dollars,is a measure of the economic consequences of an accident. Economie esects taken into account include lost wages, relocation expenses of the evacuated population, decont==i== tion costs, lost

' public and private property, and laterdicted land and farm erop costs, all calculated on the basis of statewide laaduse and land value data, and & population distribution surrounding the speciSc site.

Further details on the treatment of economie consequences can be found in references [2,9]. Economic consequences which are not included are the cost of providing health care to & aSected population, all onsite costs, litigation costs, and indirect costs. In addition, no dollar value is assigned to health efects.

The manner in which these factors are treated in this report will be detailed in the following sections.

Table 1 below lists the means of the selected efects for the Ave releases at the Indiss Point site, (which is located in the Hudson Valley approximately 40 miles north of New York City), conditional on the stated release and an 1120 MWe PWR. This table shows that even at one of the most densely populated sites is the United States, SST4 and SST5 lead to essentially no ossite consequences. Because of this lack of olkite consequences, results from SST4 and SST5 will not be presented in the tables for 3

I J

+y-.t- wa.% my-,.wq, ,.p74_9.,,wwg.gy9.,,y ,,g.%y., yg <.,.,y9, y,.g,, ,,g,p

, , ,,w"-PMPhww w-ywu n"'s'"'"emwhm""'TW'C"W'9T"WN"m"WN#CNm'PWMe-mw'we*'""'*M8-PW

each of the 156 reactor-site combinations. Treatment of onsite consequences for SST4 and SST5 will be

~

discussed in the section on onsite consequences.

Indian Point '

CRAC2 RESULTS: Mean Effects Conditional on Release RELEASE EARLY EARLY LATENT CANCER PERSON PROPERTY CATEGORY FATALITIES INJURIES FATALITIES REM DAMAGE SST1 831.0 3640.0 8110.0 1.25E+08 1.18E+10 SST2 0.1 18.0 587.0 1.10E+07 1.46E+08 SST3 0.0 0.0 1.8 3.23E+04 1.95E+07 '

SST4 0.0 0.0 0.04 7.70E+02 0.0 .

SST5 0.0 0.0 0.003 7.70E+01 0.0 Table 1

3. Scaling to Power Level All the CRAC2 calculations performed for the siting study, and which will be used for this project, assumed an 1120 MWe PWR at the site, rather than the actual reactor type and size. The type of reactor is not critical since the source terms were derived to represent any LWR design by appropriate selection of weights for the five releases. The power level of the reactor, and hence the core radionuclide inventory, plays a significant role in determining the magnitude of consequences. Sensitivity analyses (2]

indicate that inventory scales fairly linearly with power level, and that consequences scale approximately linearly with inventory in a range surrounding 1120 MWe that includes most of the reactor sizes covered in this study. The differences between linearly scaled consequences and those calculated exactly can err in either direction,(that is, there is no systematic blas), and are within the range of values representing uncertainties due to other factors. Therefore, all results presented in this report as being representative of the true power level at the plant are derived by taking *the results from CRAC2 analyses assuming an 1120 MWe power level and then scaling the result by "'[*"7 '"". Table 2 below shows the predicted consequences for Indian Point Unit 2, which has power level'of 873 MWe. These mean values are the same as those in Table 1 scaled by 6 = .86.

Indian Point Unit 2 SCALED RESULTS: Mean Effects Conditional on Release RELEASE EARLY EARLY LATENT CANCER PERSON PROPERTY CATEGORY FATALITIES INJURIES r'ATALITIES REM DAMAGE SSTI 647.7 2837.3 6321.5 9.74E+07 9.20E+09 SST2 0.1 14.0 457.5 8.57E+06 1.14 E+08 SST3 0.0 0.0 1.4 2.52E+04 1.52E+07 Table 2

4. Discounting In evaluating the economic consequences of a potential accident that can occur anytime in the life of a plant, we must sum terms for costs or risks occurring over a period of several years. This creates a problem since a dollar expended today and a dollar expended ten years in the future do not necessarily

=

have the same value today to the recipient. This arises from the fact that the dollar received today could be put in a bank and interest collected for ten years, in addition to the flexibility of having the dollar 4

[ availab'e for use before the ten years are over. One way to compare dollars that arrive at diferent points

' f in time is to And the amount of money which must be placed in a bank today to have the same amount f

X of money at the time the other income is scheduled to arrive. The sum of money which must be put in the bank today to achieve a speci8ed sum at a point in the future is called the discounted present value of the later sum, and the laterest rate is called the discount rate. Thus, the discounted present value of

$1.00 received one year from today, assuming a 10% discount rate, is $.91 ($.91 x 0.1 + 0.91 1.00).

This type of calculation is called present value discounting. An excellent introduction to this . et -

. . enn be found in an essay by Kenneth Arrow in [10] or most introductory economics texts.

In this report we have used discounting formalse based on continuous discounting. The $rst forinula

. presented here is used in calculations of the present worth of early health efects and ofsite property -

damage. The present value of a cost Co which occurs with a frequency / is given by:

't g-rt. _ g-sty e Cof dt = Cof r (1) .

s.

where r - efective discount rate f - frequency of accident costing Co

!. tg - time of onset of risk of accident

' i f- time of end of risk of accident.

For an operating plant like Indian Point Unit 2, t, is 0. We will assume a 40 year plant life, so tf - 32 for Indian Point Unit 2, which began operation in 1974 and therefore.has only 32 years of remaining operational life. We will use r .04 which is typical of the true discount rate experienced over the.

4 past several years. (The true, or real, discount rate is approximately the diference between the rate of

. Inaation and the rate of interest on debt, such as the banking ladustry's prime rate.) In the discussion section we will examine the impact of this assumption. For these values, the multiplier of Cof for formula (1) is 18.05.

' ' [' -

The next formula is for calculating the present value of an expense which recurs for a number of l

years, such as the 10 year cleanup expense discussed in Section 6. The present value of an expense Ce j recurring for M years is given by:  ;

• • (2) f Cee - dt' dt = 1 - e-'I't-'*hl - e-' A')

~

Cleanup expense has t4,i f , and r as above, and M - 10. The multiplier of Cof is therefore 148.76.

The last formula we need is for calculating the present value of an expense that will recur until a Axed date, rather than for a Axed number of years. This is the formula that applies to the replacement power costs which are charged for the remaining life of the plant (see section 6).

.' tr *r 'e-* - e-'*r -e -d'(i -

-d' di' di - Cef

~

Ce e f ig) (3) l V ,/ r r For Indiar. Point Unit 2 the multiplier of Cof is 228.79.

4

- (For the reader laterested la see details, the use of the freguesey f la the above formule impiteltly allows repeat accidents at a resetor. The formuis esa be corrected to prohibit this situatloa. However, the correction would considerably cosapticate the formals, and for the accident frequeneles under considerstlos the diference la results would be estremely .

small, appearing la the4 bird or fourth decimal place of the answer.)

The imi,act of the start date on the multipliers arises from two mechanisms. As the start date moves farther into the past, the multiplier shrinks since the remaining life of the plant is reduced, and therefore t, remains 0 while ifgets smaller. As the start date mores farther into the future, the multiplier shrinks because there are more years of discounting intervening between the present and the onset of the risk.

v 5

j. .

Thus, the multipliers achieve their maxima for start dates in the current year,1982. The table below shows values of the three multipliers for several start dates to give an indication of the impact of ' the start date on the discounted present value.

Comparison of Multipliers for Diferent Start Dates START DATE MULTIPLER (1) MULTIPLER (2) MULTIPLIER (s)

M-10 1967 15.80 130.25 165.15 1974 18.05 148.76 228.79 1982 19.95 164.45 296.92 1987 16.34 134.64 243.10 1992 13.37 110.23 199.03 Table 3 The selection of the actual discount rate will have a signiacant impact on the predicted consequences, and to some degree on their relative magnitudes. Sensitivity to discount rate will be examined in Section 8.

5. Health Effects Costs The assignment of a dollar value to various health efects is certain to be one of the most dillicult and potentially controversial aspects of the estimation of the Anancial consequences of an accident. There does not appear to be any universaq accepted method for assigning a value to a life, let alone a single price ascribed to the value of a life. One method that has been used is to impute a social perception of the worth of a life from the expenditures that society is willing to make in order to prevent a death.

The work of Cohen [11] is an example of this technique. This approach leads to widely varying values on human lifes, from the low tens of thousands of dollars for some cancer prevention tests and highway maintenaace, to hundreds of thousands of dollars per life for some auto safety features, to millions of dollars per life for some mine safety and radiological standards. A second complicating feature of assessing the value of a life is that studies have indicated that people place a diferent value on a lost life depending on the circumstances under which the death occurred. For example, a death la an involuntary, novel situation is perceived as being worse, or more costly, than a death from a familiar activity such as automobile driving [12]. Another confounding factor, especially for the reactor accident case, is the assessment of the diference in value for an immediate death compared to a delayed death, like a cancer fatality. One approach to dealing with this aspect is to use a concept of life shortening rather than a value for a life. Life shortening measures the number of lost years, and therefore may assign a lower value to the death of an older person than for younger person, or for a delayed death versus an immediate death.

A recent report by a subcommittee of the ACRS [3] proposes values for early deaths and delayed '

deaths. The values are $5 million and $1 militon respectively. (The value for early deaths was proposed in combination with a risk aversion criterion, and therefore is actually expressed in ' equivalent deaths".

Equivalent deaths are calculated for a given accident by raising the consequences of the accident to the 1.2 power, and therefore for accidents with large consequences the efective valuation on life will be larger than $5 million.) The values were proposed in the context of an ALARA (as low as reasonably achievable) criterion to require improvements to a reactor facility. The Nuclear Regulatory Commission has proposed, in NUREG-0880 [13), the use of a value of $1000 per man rem averted. This extrapolates to approximately $10 million per latent cancer fatality. In neither of these cases has a rationale for selecting these figures been given, and therefore there is no basis for discussing their merit in this report.

6 umm mi a mi

4 1- Rather than enter the controversy by attempting to place a value on human life, we will treat

. the issue by using empirical values of society's willingn-ss to expend resources to avert a death. This

./ approach is useful in the context of this report since we are trying to develop techniques which can 3

be used to develop a value/ impact approach for potential regulatory application. Therefore, the use of empirical societal values from life-saving technological " Axes" in comparable circumstances is an appropriate approach in this context. We have chosen a value of $1 million for early fatalities and

$100,000 for early injuries and latent cancers. The choice of these values is only a starting point for discussion; a later section on sensitivities will examine the impact of the selection of these particular

' values, and provide insight into the results if other values are selected. The lower Agure is in the range (although slightly larger) of imputed life values based on various medical treatments or screening techniques, mostly related to cancers, which are comparable to the delayed deaths caused by radiological accidents. The higher Agure is larger than most values for traBe safety programs or equipment, which are used to prevent prompt deaths, comparable to the early death as deAned for our purposes. In addition,..

it is in the range of the imputed life values based on other considerations such as aircraft safety. (These comparisons are based on the imputed values in [11).) For the purposes of accounting for cost, it is assumed that eSects are charged for when the accident occurs. While these particular values may cause controversy, the discussion of sensitivities in Section 8 shows that these assumptions have relatively little

- impact on the overall conclusions. Thus, while the actual values used are important in many areas of .

application, the conclusions of this report are relatively unatected by values within the range of general' discussion. Note that these costs are used to represent the expenditure that society is willing to make to avert a loss of life, and do not incidde the co,ts for medical care. Table 4 below shows the mean

'(average) costs for the oHhite consequences conditional on the stated release.

Indian Point Unit 2 OFFSITE COSTS: Conditional on Release RELEASE EARLY EARLY LATENT CANCER fROPERTY

'- CATEGORY FATALITIES ($'s) INJURIES ($'s) FATALITIES ($'s) DAMAGE ($'s)

SST1 6.47E+08 2.84E+08 6.32E+08 9.20E+09 3ST2 1.00E+05 1.40E+06 4.58E+07 1.14E+08

/'

(. SST3 0.0 0.0 1.40E+05 1.52E+07 Talsle 4 i

Table 5 below shows the discounted present value of the costs in Table 4 summed over the life of Indian Point Unit 2,~ assuming frequency f4 (expresssed per year) for release 1. Thus, early fatality, early lajury, latent cancer fatality, and property damage values are discounted to the present using discounting formula (1). The values in the tables have units of dollars and represent the discounted present value of the risks summed over the plant life; they are not the cost of may single accident. Thus, if one were to assign a valse of 10-s to fa, (this value is selected for illustrative purposes only), the expected mean lifetime risk due to property dunage would be $1.66 x 10 8. This does not imply, however, that accident.t with larger consequences could not occur.

I ,

Indian Point Unit 2

$ DISCOUNTED PRESENT VALUE OF OFFSITE COSTS (times frequency /4)

L RELEAPE EARLY EARLY LATENT CANCER PROPERTY CATEGORY FATALITIES ($'s) INJURIES ($'s) FATALITIES ($'s) DAMAGE ($'s)

SST1 1.17E+10x f 5.12E+09x fi 1.14E+10x f 1.66E+11x f SST2 J,08E+06x /s 2.53E+07 x /s 8.26E+08 x /s 2.05E+09x /s SST3 i,.bOE+00x /s 0.00E+00 x /s 2.53E+06 x /s 2.74E+08 x /s Table 5 4

7 i

+7=-yrt- , - y .mg ww-9  % p--,9,qv9, _ym.yg.g.g-g ngengs-,.gra r -em w oww,.9, g w. yop -eee swWW N Te w* w u ww 7- M-S e 9M e N-P--FNh- o'"'me-9'hM*

6. Onsite Consequences The CRAC2 code was designed to provide estimates of the offsite consequences of reactor accidents and therefore does not provide estimates of the onsite consequences of an accident, either in terms of health efibets or Anancial efects. In order to carry out the task of estimating the consequences of an accident, we have divided the consequences into four areas; plant personnel health efects, replacement power costs, cleanup costs, and capital costs.

An examination of NRC regulations concerning reactor operating procedures during emergencies, as well as procedures of the utilities, indicated that during a major emergency in which a significant '

release is imminent there would be approximately 40 persons on the site in either the control room or the Mchnical support center, both of which are required to provide a degree of protection from a radiological release (14]. Using an assumption of essentially a worst possible case, we assume that an SST1 release '

results in 10 early fatalities and 30 early injuries onsite. The signiacantly lower levels of hazard far the other releases are assumed to cause no early efects to reactor personnel. In the discussion and sensitivity analysis section we w examine the impact of this assumption on the overall conclusions, and how to estimate the consequences under diferent assumptions.

Estimates of replacement power costs are based on preliminary results from an ongoing research project at Argonne National Laboratory (15]. In this method replacement power costs are estim.2ed on the basis of the cost of replacement fuels and power availability for each National Electric Reliability Council (NERC) region. The dominant factor in determining these costs is the relative proportion of oil Ared backup plants versus economical alternative sources (for example coal or hydro). In (15] the cost of replacement power is estimated to be

~

Co = (0.286 x R + 0.086)10' 4 per M W year there R is the fraction of replacement energy by oil-fired or noneconomy power purchases. The values of R by NERC region (defined in Figure 1) are:

MARCA 20 %

NPCC 95 %

MAAC 50 %

MAIN 15 %

ERCOT 50 %

SPP 40 %

WSCC California 95 %

not California 25 %

SERC 15 %

ECAR 5%

Table 6 This cost formula is based on the assumption that the reactor had a 65% availability prior to the accident.

The value derived by this method is then multiplied by the power level of the plant, and summed over a tumber of years corresponding to the remaining life of the plant, with discounting and inflation in fuel price taken into account.

Using the replacement cost of electricity for the full remaining life of the plant imputes a value for the lost capital cost of the plant, since, by purchasing pewer, the income stream of the plant is maintained, although the net income may become negative if rates do not increase corresponding to the replacement power costs. Thus, this replacement power cost estimation procedure using actual cost for the remaining life of the plant eilminates the need to include a seperate term for the lost capital expense.

It should be noted that for cases with remaining plant life greater than ten years, it may be possible to build a new facility, either coal or nuclear, that could replace the lost capacity at a lower cost than that obtained using the calculation above. However, estimating the effect of a replacement plant is very complex since it involves changes to the capacity expansion plan of the utility, interaction with other 8

I 4

d_l

\

C l

i ,

y -

\

p -

),

~'

g .

lC n- Wr scan l i f (  %

,s .

j.fs Encor y e

\_

\

4 l

Figure 1. NERC Regions O 9 l

utilities la the region, as well as other factors relating to the overall economy. Means of incorporating -

these factors are currently being researched at Argonne. Estimates of the impact of including these ellbets will be included in the discussion section at the end of this report.

Cleanup costs for reactor accidents are diflicult to estimate due to a lack of experience and data.

For the initial stages of this project we will use a value of $100 million dollars per year for ten years which represents the cost of early decommisioning or cleanup and repair, depending upon the severity of the accident. This figure is comparable to current estimates of the cleanup costs for the Three Mile Island accident [16].

Table 7 below shows the estimates of these contributors for the Indian Point Unit 2. Note that the health edects costs only apply to SST1, and the other costs to SST1,2, and 3. SST4 and 5 are not expected to produce any onsite health eithets, and may leave the plant in a repairable state. Thus, the ,

replacement power may not have to be purchased for the remaining life of the plant, but rather for a shorter period. The discounted value of that cost could be estimated using discounting formula (2). At the moment, we have no estimate for the cleanup (or repair) costs of SST4 or 5.

Indian Point 2 Onsite Health Efibets: $ 1.30E+07 (conditional on SST1)

$ 2.35E+08 x ft (discounted over remaining plant life)

Replacement Power: $ 3.12E+08 (per year, conditional on SST1,2, or 3)

$ 7.14E+10 x f4 (discounted over remaining plant life)

Cleanup: $ 1.00E+08 (per year, conditional on SST1,2, or 3)

$ 1.49E+10 x f6 (discounted over remaining plant life)

Table 7 Table 8 below shows the mean (average) costs of an accident at Indian Point in 1982. The health ellhets costs are magnitude times the dollar values discussed earlier, the of! bite property damage costs are the scaled estimates of means from CRAC2, conditional on the release, and the onsite costs are the per year conditional costs from Table 7 above, summed over the period for which they are paid (10 years for cleanup costs and 32 years (remaining plant life) for replacement power), and discounted assuming a 4% real discount rate. Thus all the entries ' are the present net value in 1982 of an accident occurring in 1982.

Indian Point Unit 2 MEAN TOTAL COSTS: Conditional on Release RELEASE OFFSITE OFFSITE ONSITE TOTAL CATECORY HEATLH COSTS PROPERTY COSTS COSTS COSTS SST1 1.56E+09 9.20E+09 6.43E+09 1.72E+10 SST2 4.73E+07 1.14E+08 6.43E+09 6.59E+09 SST3 1.40E+05 1.52E+07 6.43E+09 6.45E+09 .

Table 8 The following table shows the discounted present value of the life cycle risk from Indian Point Unit

2. This risk represents the sum over all future years of operation, and therefore is not conditional on an accident. Rather, the expressions contain a term for the accide'it frequency f, which must be multiplied out to arrive at a discounted present value for the life cycle risk.

10

- w a .. , i a u uu

Indian Point Unit 2 DISCOUNTED PRESENT VALUE OF TOTAL COSTS '

OFFSITE OFFSITE ONSITE TOTAL sy -

RELEASE CATaGORY MEATLH COSTS PROPERTY COSTS COSTS COSTS SSTI 2.82E+10x f 1.66E+11 x f 8.66E+10 x fi 2.81E+11x f SST2 8.52E+08x f 2.05E+09 x f2 8.63E+10x /2 8.92E+10x fa SST3 2.53E+06x fa 2.74E+08x fa 8.63E+10x /s 8.66E+10x fa Table 9 i

7. Summary of Results

. i This section summarises the detailed results presented la Appendix A. It is important to again caution h reader and potential user of these data that h results are derived from CRAC2 analyses assuming sa 1120 MWe PWR at'ench site. These results were them scaled linearly by power level. Thus, the data do not represent the conclusions of PRAs performed for each site and reactor, but rahr are approximations. As la all applications of this nature, h data are subject to uncertalaty. However, we believe that the nature of the uncertalaties is such that comparisons among plaats are fairly accurate la relative rankings. _,

. Early fatalities show & most variability of the various health essets. Figure 2 is a CCDF (complementary cumulative distribution function) of the mean number of early fatalities at each of the 156 reactor-site combinations, conditional on an SST1 release. The mean of this CCDF is 69, which is exceeded almost 25% of k time. As can be seen la the CCDF, the mesa number of fatalities for b various reactor site combinations spaa more than four orders of magnitude.

Figure 3 is the CCDF of mena early lajuries. Mesa early lajuries for the reactor site combinations span about two and one. half orders of magnitude, compared to the four orders for mena early fatalities.'

f The mesa of this CCDF is 345, and like early fatalities, & mesais exceeded spyr- :- " $y 30% of &

\

time. The CCDF of latest cancer fatalities is shown la Figure 4. With & exception of one outlier, &

data fall within two orders of magnitude. The mean of & distribution is 1450, sad is exceeded about 30% of the time. Thus, & shape of h three health eSects CCDFs are very similar to each other la

that each has a mesa exceeded 25-30% of the time, and each has an 's'-shape (ladicative of a lognormal distribution).

Figures 5,6,and 7 are the CCDFs of mean person-rem for SST1,2, and 3 respectively. The three curves are almost identical in shape, each is just the next larger release shifted to the left. As in the CCDFs presented earlier, the mesa of each of these distributions is exceeded 30-40% of the time.

Figures 8,9, and 10 tilastrate the CCDFs for various economic consequences. Figere 8 is a plot of

& mesa oSbite property damage costs, conditional on an SST1 release. There are about two orders of

! magnitude spread, although 90% of the values fall withis a one order of magnitude range. Figure 9 is a CCDF of mesa total health essets costs conditional on sa SST1 release. These costs show about two i: a.

orders of magnitude of variation, which is much less tham & four orders observed la early fatalities. The narrowing of the spread indicates the importance of & costs of latent enacer fatalities is determining j the overall health essets costs. Figure 10 is the CCDF of mean total Ramacial costs for an SST1 release.

. The CCDF shows very little variability. This is due to b domlanace of cleanup costs and replacement power costs la determining the overall cost. (The Indian Point site used la & example tables la this ,

report is the exception to this general comelusion.)

h

8. Sensitivity and Discussion I

In a number of areas discussed above bre is considerable uncertalaty regarding the values selected.

f In this section we will examine the sensitivity of results and conclusions to the particular values select d.

l

(

11 1

r ~ .e m .- - ne.,-n,-,-n . -

.--.,n <.-~ ~-v.- -.m--- -. - . - - - - - - . , - - - --.A,--

The area which may contain the greatest uncertainty is the selection of values for loss of life, since this value cannot be determined in a technical manner; it is really a social or political question. Using the values for health eSects given in Section 5, health effects costs account for approximately 12% of the total ofthite discounted costs for the SST1 release. (This is the average over all sites and reactors.) The mininum value was 5% of total otsite cost, while the maximum was less than 25%. Figure 11 is a CCDF of the ratio of health efects costs to total otsite effects costs (health efects costs plus property damage), conditional on SST1. It is apparent that the 12% mean is exceeded approximately 40% of the time. Thus, in general an order of magnitude or greater increase in the value assigned to loss of life is necessary to increase the cost of health efects to a level making it a significant contributor to the offsite cost of an accident. Figure 12 is a CCDF of the ratio of discounted health effects costs to total accident costs including replacement power and cleanup costs, conditional on SST1. At most, health effects costs contribute just over 10% of the total ecst, and more typically contribute 5% or less, further reducing the impact of changes in values assigned to loss of life. The portion of. total offsite costs represented by '

health effects costs for SST2 is 16% minimum,30% average, and 48% maximum, while SST3 is 0.5%

minimum,3% average, and only 11% maximum. Thus, SST2 has a greater sensitivity to change in life values, although a factor of five change is still necessary to bring health effects cocts to the level of offsite property damage. It should be noted, however, that for SST2 and 3 the total offsite costs are more than an order of magnitude smaller than the onsite costs, and therefore an order of magnitude change in the health effects costs would have negligible effect la determining the overall costs for these accidents.

Some participants in the debate on the valuing of human lives argue that it is improper to discount the dollars associated with loss of life, since this is implicitly discounting the value of a life lost in the future. While we do not agree with this argument since it can lead to inefficient utilization of social resources, we have examined the impact on the conclusions presented in this paper if health effects costs are not discounted. Figure 13 is a CCDF of the fraction of total offsite costs which represent

'non-discounted health effects, conditional on SST1. In comparing this figure to Figure 11, one can see

-that the maximum fraction has increased from about 20% of the total costs to 50%. However, this large fraction occurs with very low frequency, and 80% of the time health effects, even without discounting, account for at mos> 25% of the offsite costs, which in turn represent only a fraction of the total costs.

Recently, some attention has been focused on the use of $1000 per person rem averted as the basis for an AI. ARA criterion on reduction of publl: risk from reactor accidents [17]. For SST1 accidents, the discounted present value (over the life of the plant) of person tem times $1000 ranges from a minimum of 79% of the total (onsite and offsite) cost of an accident to more than 16 times the total cost, with average discounted nresent value cost equal to 9 times the total cost using the formulation described in this paper. SST2 nas a minimum of 6%, maximum of almost 1000%, and a mean of 200%. For the SST3 release, the maximum value represented by using the $1000 per person rem value is only 3% of the total accident cost. The use of the $1000 per person rem as a proxy for only the health effects costs is also subject to considerable variability. For the the SST1 release, the person tem estimates exceed the modeled value by factors ranging from 50 to over 450, with an average of approximately 240. SST2 and 3 both'have overestimates by factors averaging 450. Thus, even if the values assigned to loss of life were to be increased by an order of magnitude, the person rem proxy value would still overestimate the value of health effects by a large amount.

Several uncertainties are associated with the onsite cost estimates. The first is the magnitude of the onsite health effects. Onsite health effects account for less than 1% of the total onsite costs for SST1, and therefore a significant increase in the values assigned to loss of life would be required to have any impact. (It seems implausible that more injuries would occur, since the 40 casualties used assumes a

, 100% casualty rate, with 25% of the casualties being fatal.) Even assuming this level of casualties would have little or no effect on the other releases since the cleanup costs and replacement power costs remain the same, and completely dominate the onsite costs.

The second area of question in onsite costs is the assumption that replacement power is ' tchased for the remaining life of the plant. Because of discounting, approximately 60% of the presen ralue for replacement power is due to expenditures in the first 10 years which is probably the minimum amount of time necessary to build a new plant. In addition, even the remaining years would still have a cost, which could be no less than one-fourth the estimated replacement power cost, and therefore the total 12

+

- replacement power costs would change by no more than 30%. Replacement power costs account for 50%

of the total costs (average over all sites) for SST1, and therefore the approalmation based on purchasing replacement power for the remaining life of the plant could change & total costs by no more than 15%

. O3

^

DiSerent discount schedules (with identical discount rates) are used for various costs and therefore changes la the discount factor asset diferent costs la diSerent ways. For the health effects and oEhite property damage which are discounted at & time of occurrence, and cleanup, which has a cost recurring for ten years, the present value increases by about 40% when the discount rate is dropped from .04 to

.02. Eliminating discounting altogether (0.0 discount rate) causes present values to double for plaats with start dates near or before 1982. As the start date moves farther lato the future, the disprence increases to almost a factor of four for start dates well lato b 1990s. The value drops by about 40%

! when time discount rate is doubled to .08. The present value of the replacement power changes only-

i slightly faster than these Srst two terms, i A series of sensitivi*y calculations was performed as part of b siting study [2] to evaluate the

) -lapact of uncertalaties la the source term, especially la line with reductions suggested la (5). The i

table below shows the relative reductions la mean ogbite consegmences at the Indian Point site, with all  :

==1==- scaled to 100 for the original SST1 source term. In general, it appears that & relative contribution to odhite costs of its constituent factors will renais unchanged with reductions la h source term (although the importance of early fatalities does decrease somewhat). The contribution of oEbite

!. costs to total cost is reduced however, since the oasite costs are fairly ladependent of the sise of h  ;

source term.

Sensitivity of Mesa Consequences to Reductions la SSTI l'

RELEASE EARLY EARLY LATENT CANCER PARSON PROPERTY

{ aIEa FATALITm3 INJURmS FATALITIES ram DAMAGE J SST1 100 100 100 100 100 SST1x .5* 30 35 74 82 62 L

i SST1x.1 1 4 32 39 11 SST1x.05 0.2 2 19 24 5  :

SST1x.01 0.03 6.7 0.71

[ 1 5

• Resense treetione redaeed ror allisotopes eseert noble sness l.

Table 10 A number of costs for sa accident were not lacluded la b calculatione presented in this report.

These costs imelade medical care, litigatloa, and ladirect costs. It is discult to estimate these costs, but some very erude estimates indicate ht some may be of importance. Medical costs esa be divided among treatment of early fatalities, early injuries, and latent caneers, both fataJ and mondatal. The costs

) assigned to b early eSects will probably be small la comparison to the $1 million and $100,000 assigned to these eSects. The costs of usedical treatment of & latest caaeers may be larger than the $100,000 associated with latest cancer fatalities, especially when non fatal ceaeers are taken lato account. The

)

4 ladirect costs could be very large if they are to take lato account mellosal economic repercussions such

• - as closing of other nuclear power plaats, loss of industrial espacity and jobs, essets on Anancial markets,-

soelo-econoanic impacts, etc. Estimates of these essets [13,19l ladicate that they may be larger than all

the essets discussed la this report. However, many of these ecsts are widely dispersed and would be diScult to llak to the pecident, at least la terms of establishlag legal liability.

While the range of variation due to uncertalaty may have signiacant impact when comparlag .

estimated values to absolute Agures, such as the cost of a desiga moditentloa, they do not signi8cantly

1. change the relative contributions of h various factors to & overall Associal consequences of an accident.

In additloa, except for possible ladirect costs which were not evaluated la this paper, & magnitude of these changes is generally withia & range of uncertainties due to & CRAC2 code, the selection of  !

source term, the estimation of accident probabilities, and oasite cost estimates.

13

( _ ____ -.

O ri I I

ri -

rlt -

g .

se

- u.

E

$jli P

~IA Y r3 ri l .

y , . . .

1 3 3 3 3 3 i 3 i Isugeooss sysegg p sopseg O

l'  !'

r,E

- :n i [3 m -

.)

a f ri

~

. H l  :

g -

3 3

g i i i

_ , , 3 y pm g p sopsey I 8'W'**** "d # **U**M O

ll

-- ._4 l

l v-M

g I I i

, M"

.E h M

R r1 1 *

- g i i i i i i i y 3sypeness svesig p sopasq X Sugposess W M MW O

[

th Mk '

!l } '} }

'a ,sj

.a

~

i i i i g assegg p X 881peesse mid M W O 15- i

l O

yg n-i 3 3 .

g 3

3

I 3 g

.g E If h

. 3 N 3 3

- - R- -

i i i i -8 9 '

i i i i I3mipsesse syngd p sottaeq I3miposan syssid M sot **d N O

M "I

i Q

Iv f - 3, E

.s 3, .

a  :

m

1 A

g Y "

h 3

3 3 3 3 3 3 3 I3mipewn egnld p nottmd ISu1pesan sinta p nottaeq O

16-

a REFERENCES

(C

'\ 1. Starr, C , and Whipple, C, "Caping with Nuclear Power Risks: The Electric Utility Incentives",

Nuclear Safety, Vol. 23, No.1, January-February,1982, ppl-7,

2. Aldrich, D., Sprung, J., et. al., Technical Guidance for Siting Criteria Development, NUREG/CR-2239, SAND 81 1549, Sandia National Laboratories,1982.

3. ACRS, An Approach to Quantitative Safety Goals for Nuclear Power Plants, NUREG-0739, October, 1980.

4. Ritchie, L., Johnson, J., and Blond, R., Calculations of Reactor Accident Consequences, VerA= 2: User's G=ua NUREG/CR-2326, SAND 81-1994, Sandia National Labcratories,1982.

' 5. Levenson, M., and Rahn, F., " Realistic Estimates of the Consequences of Nuclear Accidents,"

Nuclear Technology, Vol. 53, May,1981, pp99-110.

6. Ritchie, L., Aldrich, D., and Blond, R., " Weather Sequence Sampling for Risk Calculations,"

Transactions of the American Nuclear Society, June,1981.

7. Alpert, D., Aldrich, D., and Ostmeyer, R., " Meteorological Data for Reactor Accident Risk Calculations,"

Transactions of the American Nuclear Society, November,1981.

8. USNRC, Regulatory Impact of Nuclear Reactor Source Term Assumptions, NUREG-0771, June,1981.

9. USNRC, Reactor Safety Study Appendix VI: Calculation of Reactor Accident Consequences, WASH-1400, NUREG 75/014, October,1975.

10. Arrow, K., Energy and the Environment,Ashley, H., Rudman, R., and Whipple, C., eds., Pergamon Press, New York,1976. pp*113-140.

n 11. Cohen, B., " Society's Valuation of Life Saving in Radiation Protection and Other Contexts,"

Health Physics, Vol. 38, January,1981, pp. 33-51.

12. Slovic, P., " Images of Disaster: Perception and Acceptance of Risks from Nuclear Power,"

' Proc. FIAeenth Annual Meeting of the NCRPM, March,1979, pp.34-36.

i 13. USNRC, Safete Goals for Nuclear Power Plants: A Discussion Paper, NUREG-0880 Febr'ary,1982.

14. Burke, R., and Rasmussen, N., personal communication, March,1982.

15. Buehring, W., personal communication, February 17,1982.

16. USGAO, Three Mlle Island: The Financial Fallout. EMD-80 89, July,1980.

. 17. Strip.D., An Analysis of a Proposed $1000 per man-rem ALARA Criterion,SANT-1870, Sandia National Labortories,1982.

d

18. Stucker, J., Batten, C., Solomon,K.,andHirsch,W., Costs of Closing Indian Point Nuclear Power Plant,

' Rand Corporation, R 2857-NYO, November,1981.

19. Cartwright, J., Beemiller, R., Trott, E., and Younger, J., Industrial Impacts of Hypothetical Accidents at the Catawha Nuclear Reactor. Bureau of Economic Analysis, U.S. Department of Commerce, January,1982.

I 17 l .

l ACKNOWLEDGEMENTS The author would like to thank Richard Burke whose work in related areas has contributed greatly to this report, and Jay Johnson for performing many of the CRAC2 calculations necessary for this project. .

I 6

e O

l i

l l

l l

O l

\

18

' ' ' - - - ------n.

9 8, / (',

/ ~',

, x/ ny  :,

2-f '7

\~

,  :! W;lN!? J':j>

' -!  :!>U:)

c. ,C1 ~E

\..1

Jjit -

LJ

[ \ th

/

[

/ / e

/p, / ,,#

f'%e7.' v, .yr;- - ,e

\,s ,1 W ~Y

'Wy .// ,,/l / !-]

6. 4 / ,,//

lls h/,/ l l / .

e

/

*(('h'h*F

• s

//

'/

I l

I O

. . _ _