ML16341F660
| ML16341F660 | |
| Person / Time | |
|---|---|
| Site: | Diablo Canyon  |
| Issue date: | 04/12/1990 |
| From: | Rood H Office of Nuclear Reactor Regulation |
| To: | Shiffer J PACIFIC GAS & ELECTRIC CO. |
| References | |
| TAC-55305, TAC-68049, NUDOCS 9004230077 | |
| Download: ML16341F660 (20) | |
Text
(
~
April 12, 1990 Docket Nos.
50=275 and 50-323 ter. J.
D. Shiffer, Vice Pres ident Nuclear Power Generation c/o Nuclear Power Generation, Licensing Pacific Gas and Electric Company 77 Beale Street, Room 1451 San Fr ancisco, California 94106
Dear Mr. Shiffer:
DISTRIBUTION QIKFTTB M RPi h
NRC 8
LPDRs GBagchi JZwolinski NChokshi PShea RRothman HRood CTrammell EJordan ACRS (10)
OGC (for information only)
PDV Plant File
SUBJECT:
TRANSMITTAL OF DOCUMENTS AND REQUEST FOR ADDITIONAL INFORhlATION RELATING TO NRC STAFF REVIELt OF DIABLO CANYON LONG TERt1 SEISMIC PROGRAM (LTSP)
(TAC NOS.
55305 AND 68049)
Enclosed are two consultant reports that relate to the NRC staff review of the Diablo Canyon Seismic Reevaluation Program.
The enclosed material is as follows:
Letter dated March 1, 1990 from Keiiti Aki of USC, to Jean Savy of LLNL; subject: Letter report on reviewing the PGIIE response to questions 1 through 19 (the questions were issued by the NRC by letter dated 2.
In order comments on April
- request, Letter dated March 12, 1990 from Ralph J. Archuleta of UCSB, to Jean Savy of LLNL; subject: Letter report on reviewing the PGImE response to questions 6, 8, 15, 16, 17, 18, and 19.
to maintain our review schedule, we request that you address the contained in the enclosures at the ground motion meeting to be held 30 and May 1, 1990.
If you have any questions regarding this please contact me.
Sincerely,
Enclosures:
as stated cc w/encl:
See next page R P/PD5
/(A)D:PD5 HRood ramme1 I 04/12/90 04/)+90 OFFICIAL RECORD COPY original signed by Harry Rood, Senior Project Manager Project Directorate V
Division of Reactor Projects - III, IV, V and Special Projects Office of Nuclear Reactor Regulation 9004230077 9004i2 PDR ADOCf( 05000275 P
PNU
l I
P l
(4gPg RfGIJ
~o ly A
~ I 0O
++*++
Docket Nos.
50-275 and 50-323 UNITEDSTATES NUCLEAR REGULATORY COMMISSION WASHINGTON, D. C. 20555 April 12, 1990 Mr. J.
D. Shiffer, Vice President Nuclear Power Generation c/o Nuclear Power Generation, Licensing Pacific Gas and Electric Company 77 Beale Street, Room 1451 San Francisco, California 94106
Dear Mr. Shiffer:
SUBJECT:
TRANSMITTAL OF DOCUMENTS AND RE(VEST FOR ADDITIONAL INFORMATION RELATING TO NRC STAFF REVIEW OF DIABLO CANYON LONG TERM SEISMIC PROGRAM (LTSP)
(TAC NOS.
55305 AND 68049)
Enclosed are two consultant reports that relate to the NRC staff review of the Diablo Canyon Seismic Reevaluation Program.
The enclosed material is as follows:
Letter dated March 1, 1990 from Keiiti Aki of USC, to Jean Savy of LLNL; subject: Letter report on reviewing the PG&E response to questions 1 through 19 (the questions were issued by the NRC by letter dated 2.
In order comments on April
- request, Letter dated March 12, 1990 from Ralph J. Archuleta of UCSB, to Jean Savy of LLNL; subject:
Letter report on reviewing the PG&E response to questions 6, 8, 15, 16, 17, 18, and 19.
to maintain our review schedule, we request that you address the contained in the enclosures at the ground motion meeting to be held 30 and May 1, 1990.
If you have any questions regarding this please contact me.
Sincerely,
Enclosures:
as stated cc w/encl:
See next page Harry Rood, Senior Project Manager Project Directorate V
Division of Reactor Projects - III, IV, V and Special Projects Office of Nuclear Reactor Regulation
1 l
Mr. J.
D. Shiffer Pacific Gas and Electric Company Diablo Canyon CC:
Richard F. Locke, Esq.
Pacific Gas 5 Electric Company Post Office Box 7442 San Francisco, California 94120 Ms. Sandra A. Silver 660 Granite Creek Road Santa Cruz, California 95065 Mr. Peter H.
Kaufman Deputy Attorney General State of California 110 West A Street, Suite 700 San Diego, California 92101 Managing Editor The Count Tele ram Tribune 13 o nson venue P. 0.
Box 112 San Luis Obispo, California 93406 Ms.
Nancy Culver 192 Luneta Street San Luis Obispo, California 93401 Regional Administrator, Region V
U.S. Nuclear Regulatory Commission 1450 Maria Lane, Suite 210 Walnut Creek, California 94596 Mr. John Hickman Senior Health Physicist Environmental Radioactive Mgmt. Unit Environmental Management Branch State Department of Health Services 714 P Street, Room 616 Sacramento, California 95814 NRC Resident Inspector Diablo Canyon Nuclear Power Plant c/o U.S. Nuclear Regulatory Commission P. 0.
Box 369 Avila Beach, California 93424 Bruce Norton, Esq.
c/o Richard F. Locke, Esq.
Pacific Gas and Electric Company Post Office Box 7442 San Francisco, California 94120 Dr. R. B. Ferguson Sierra Club - Santa Lucia Chapter Rocky Canyon Star Route Creston, California 93432 Chairman San Luis Obispo County Board of Supervisors Room 270 County Government Center San Luis Obispo, California 93408 Michael M. Strumwasser, Esq.
Special Assistant Attorney General State of California Department of Justice 3580 Wilshire Boulevard, Room 800 Los Angeles, California 90010
I
Hr. J.
D. Shiffer Pacific Gas and Electric Company
~
Diab lo Canyon Long Term Seismic Program CC:
Dr. Keiiti Aki Department of Geological Sciences University Park University of Southern California Los Angeles, California 90089-0741 Dr. Ralph J. Archuleta Department of Geological Sciences University of California Santa Barbara Santa Barbara, California 93106 Dr. Steven H. Day Department of Geological Science San Diego State University San Diego, California 92182 Dr. George Gazetas Dept. of Civil Engineering 212 Ketter Hall SUNY-Buffalo Buffalo, New York 14260 Dr. Robert D. Brown, Jr.
U.S. Geological Survey Mail Stop 977 345 Hiddlefield Road Henlo Park, California 94025 Dr. David B. Slemmons Center for Neotectonic Studies MacKay School of Hines University of Nevada-Reno
- Reno, Nevada 89557-0047 Dr. Robert Fitzpatrick Building 130 Brookhaven National Laboratory
Box 808 Livermore, California 94550 Dr. Anestis S. Veletsos 5211 Paisley Avenue
- Houston, Texas 77096 Dr.
Ken Campbell U.S. Geological Survey P.O.
Box 25046, Mail Stop 966 Denver Federal Center
- Denver, Colorado 80225 Dr.
C. J. Costantino Building 129 Brookhaven National Laboratory
- Upton, New York 11973 Dr.
M.
K. Ravindra E(E 3150 Bristol Street, Suite 350 Costa Mesa, California 92626 Dr. Michael Bohn Sandia Lab. - Organization 6412 Post Office Box 5800 Albuquerque, New Mexico 87185 Dr. J.
Johnson EgE 595 Market Street - 18th Floor San Fran ci s co, Ca 1 iforni a 94105
'I
DEPARTMENTOF GEOLOGICALSCIENCES TELEFHONB (213) 743-2717
~OVtNg@
o~
March 1, 1990 Dr. Jean Savy MSI 196 LLNL P. O. Box 808 Livermore, CA 94550 Dear Jean Mis is my letter report on reviewing the P.G, &E. response to questions 1 through 19.
Let me followthe list ofissues I raised in my letter to you dated Match 9, 1989.
The first one was the issue ofthe selection rule, especially, the effect ofincluding the Parkfield and Morgan Hillearthquake data. This issue was addressed in thee questions, namely, 2, 13, and 14. The P.G. &,E response to these questions reveal that the effect of the selection rule is small but significant. The effect is greater for frequencies lower than 5 Hz, and the expanded data base gives the result very similar to the site specific spectrum, and we can no longer say that the latter envelops the former. The peak value ofthe 84th-percentile spectra indeed exceeds that ofthe site specific spectrum by about 15% ifsoil site records were excluded from the data base. This confirms the diluting effect ofsoil site records, although the reduced sample size may also be responsible at least in part as claimed by P.G. &E.
My second issue was actually a non-issue, because I agreed with P.G. &E. about the surprising magnitude dependence ofdata dispersion.
The response to Question 6 gives further convincing argument that the observed magnitude dependence is not due to the artifact ofdata collection and processing.
The explanation offered by Steve Day and accepted by P. G. &E. that the reduction of dispersion for large earthquakes may be due to the fact that th'e contributions from many segments ofa fault plane tend to average and smooth the effect from each segment is also acceptable to me, The third issue Iraised in my March 9 letter was with regard to the use of the word "empirical source function". This is confusing and misleading because it does not represent the source effect ofthe target earthquake, but is simply the empirical Gtcen's function corrected for the propagation effect in a very crude manner. Furthermore, the word "empirical source function" wrongly suggests that it may not include the effect ofthe recording site, I repeat this objection because P.G. &E. still use the word "empirical source function" in the responses to Questions 4 and 7.
A more fundamental issue is about the physical model offaults used for the numerical simulation. In the response to Question 18, P.G. &E. introduces three stress drops, namely, global static stress drop, global rupture duration stress drop, and local stiess drop. I can followthe definition and description of the first two stress drops given by P.G. &E., but cannot accept those of "local stress drop".
UNIVERSITYOF SOUTHERN CAUFORNIA, UNIVERSITYPARK, LOS ANGELES, CAUFORNIA 90089-0740
7
First, global rupture duration stress drop as defined by Somerville and others (1987) is very close to the so called "Brune's stress drop" ifthe corner frequency is replaced by the reciprocal ofthe rupture duration. The physical significance ofthis stress drop as used in strong motion simulation is unclear. For example, Boore and Atkinson (1987) stated "this parameter is known by several names; we prefer to refer to itsimply as the stress parameter and thereby not attach any physical significance in terms offault models".
On the other hand, the local stxess drop defined for the specific barrier model of Papageorgiou and Aki(1983) referred in the P.G. &E. response to Question 18 has a clear physical meaning. Their model is composed ofcircular cracks ofthe same size fillinga rectangular fault plane, and the local stzess drop is the static stress drop occurring for each circular crack.
Since the local stress drop is assumed to be the same for all cracks in the specific barrier model, itwillbe different &om the global rupture duration stress drop, unless one crack occupies the whole fault plane.
The final model ofP. G. &E. for the Hosgri fault (as i interpreted in my March 9 letter) is a fault plane filledwith 4x22 subevents, each occupying 3x4 km area, with the slip velocity of 50 cm/s and local stress drop of50 bar. This is very similar to the specific barrier model with local stress drop of 50 bar except for the introduction of asperities by allowing a variation of stress drop &om a subevent to another.
I am very much confused now because earlier Iunderstood that the stress drop now called "global rupture duration stress drop" by P. G. &E was the local stress drop of subevent similar to the circular crack ofthe specific barrier model. Now, by definition given in the response to Question 18, it is defined as the Brune stress drop with the corner frequency replaced by the teciprocal ofduration.
I wonder ifthis serious inconsistency is something existing only in my mind because ofmy misunderstanding of the physical model ofP.G. &E., or is due to the nebulous nature oftheir physical model as teflected in the statement of Boore and Atkins mentioned earlier. In any case, since the model ofP.G. &E. should have a clear physical meaning, it must avoid the use of a parameter regarded as something to which any physical significance in terms offault models should not be attached.
In response to Question 18, the local stress drop estimated by Papageorgiou and Aki(1983) is quoted as 200 to 370 bars. This estimate was made without the consideration oflocal site effect on the strong motion accelerograms.
Recently, Akiand Papageorgiou (Proc. of 9th World Conf. Earthq. Eng., Aug. 2-9, 1988, Tokyo-Kyoto, Japan, Vol. VIII, P. 163-169) revised the stn:ss drop using the frequency dependent site effect estimated by Phillips and Aki(1986). The revised value now ranges from about 100 to 200 bars.
The fifthissue raised in my March 9 letter is satisfactorily answered by responses to Questions 9 and 11. The reliable frequency range ofthe numerical simulation is above 2 Hz.
The sixth issue is the step-by-step explanation oferrors in the numerical simulation, and was addressed in detail by P.G. &E. in the response to Question 7. This response gives a clear documentation of the quantitative measure of the goodness offitfor strong motion simulation used for estimating the uncertainty of simulated ground motion. The
l l
separation ofuncertainty into three parts, namely, modeling, random and parametric, helps to clarify the sources ofuncertainty in strong motion simulation.
It is clear that the important sources of uncertainties are (1) smallness ofsample size presently available for validation, and (2) some arbitrariness in the choice ofparameters for future earthquakes. Iwonder ifitis possible to include the Lorna Prieta earthquake data for further validation.
Finally, there are several questions raised with regard to the topographic effect on strong ground motion. In none of the responses to these questions, the P. G. k, E. paid attention to the anomalous effect expected forthe SV incidence at the critical angle as discussed by Kawase and Aki(BSSA, fg, No. 1, 1990) and Nava ~ g. (Seismological Research Letters, 59, No. 6, 1989). The anomalous effect described by Kawase and Aki for the WhittierNarrows earthquake indicates that the Diablo Canyon site topography effect due to SV waves from the sources in the ocean side may be "deamplification" rather than "amplification". 'Ms is relevant to Question 12.
The anomalous ef'feet at the critical incidence is also relevant in the inteqaetation of the Nahanni records at sites 1 and 2 in terms of a souse coupled site effect.
Sincerely yours, KeiitiAki KA:st
~ UNIVERSITYOF CALIFORNIA, SANTA BARBARA BERKELEY
~
DAVIS
~
IRVINE
~
LOS ANGELES
~ RIVERIE
~
SAN DIEGO
~
SAN FRANCISCO SANIA BARBARA
~
SANIA C RUE DEPARTMENT OF GEOLOGICAL SCLENCES SARI% BARBARA, CALIFORNIA93106 Fhx (SOS'61.231'arch 12, 1990 Dr. Jean Savy Mail Stop I 196 Lawience Livermore Laboratory P. O. Box 808 Livermore, CA 94550
Dear Jean,
Having read PG8cE's responses to questions 6, 8, 9, 15, 17, 18 and 19 there are a few comments and some questions ofmy own. Before addressing each of these responses, there is a fundamental dilemma presented by the PG8cE methodology for numerical simulations. Namely, ifthe radiation coefficient that modulates the seismic radiation is homogenized (made practically isotropic), what then is the difference between a thrust faulting mechanism and a strike-slip faulting mechanism?
As far as I can determine, there isn't any. Myreasoning goes as follows. Ifslip occurs on a fault plane prescribed by its strike and dip, the angle which determines the style offaulting is the rake.
Because the dip and strike angles are fixed by the geometry, the only angle left to be homogenized is the rake. Thus homogenization of the radiation is equivalent to randomizing the rake which in turn is equivalent to randomizing the sense ofslip, i.e., randomizing the style offaulting. Ifthis is true, all differences between thrust, strike-slip and oblique faulting are due to geometry, not style offaulting. This is particularly bothersome given that the numerical simulations are validated against data, e.g. Question 8, comparing styles offaulting and Question 17 comparing empirical data with numerical simulations.
Question 6.
This analysis seems complete. Physically itmakes sense with what Steve Day once said.
Namely, one might expect less dispersion with the larger magnitude events because the larger the fault plane the more opportunity there is fordifferent rays to interfere with each other. Again one has to look at what is being considered, peak acceleration, This is a single number. A magnitude 6.5 earthquake might generated 10-15 seconds ofstrong acceleration.
Ten seconds sampled at 100 or 200 samples/sec leaves one with 1000 to 2000 data points from which to pick one number.
However, a magnitude 5 earthquake might generate only 3-5 seconds ofstrong motion, Now, one picks a single value out of 300-500 samples.
As mentioned in the response, "no differentiation was made between soil-and rock-site data in the analysis..." The figures Q6-3 would not support such differentiation even ifthe assumed regression function allowed forit. Those figures are telling. There is one datum within 10 hn for a M 6.8 - 7.4 earthquake showing peak acceleration about 0.45 g; all other data (3 of them) are larger than 0,6 g.
Question 8.
\\
IfIunderstand this correctly, this answer says that peak acceleration and average response spectral level (3 - 8.5 Hz), scale in the same way for different styles offaulting. As noted above, what is the meaning ofstrike-slip or thrust ifthe radiation coefficient is homogenized?
Even ifthat were not the case, why would one expect the response spectral level to scale in the same way as the peak acceleration unless the peak acceleration was primarily due to the 3-8.5 Hz radiation. I don' consider the difference between 16% and 19% to be significant given the analysis shown in Question 6.
Question 9.
In PG&E's response to this question there are statements made that may or may not be true. The purpose of this comparison was to validate ifpossible the truth of such statements.
In particular, on Page 2, third paragraph:
"However, at the frequencies of importance to the plant site, strong ground motions are characterized by incoherent source and wave propagation phenomena, and these theoretical seismograms methods do not provide adequate representations of the high-frequency ground motions." "these theoretical seismogram methods" refers to frequency-wavenumber. I take exception to this statement.
These methods have been used by a number of authors to model strong motion records. The upper limiton frequency is more a matter of computer time than appropriateness, PG&E has certainly not shown any waveform comparisons to indicate that the generalized ray method works. The comparisons an: always in the form of a PGA or response spectrum The PGA comparison is almost useless.
Again one generates 2000-3000 acceleration points. The maximum value is selected and compared with the maximum number from some recorded accelerogram that also contains 2000-3000 points. Given that the stress drops are approximately correct as they willbe since one doesn't find 10 m ofslip on areas of 100 m, and allowing for conservation of energy, i.e., 1/R attenuation which is supported by the data, why shouldn't one ofthe 2000-3000 numbers agree with one datum? The PGA from the numerical simulations doesn't even have to be within one second of the time for the PGA in the recorded accelerogram.
There is no attempt at waveform modeling. Likewise, response spectrum comparisons ignores the phase information in the accelerOgra.
This statement by PG&E is based on other authors'ttempts to model waveform data. To say that these other methods are inferior contradicts the statement made by PG&E on Page 6, third paragraph: "Atfrequencies above 2 Hz, there is generally good agreement between the response spectra generated using the two methods, and there is excellent agreement in peak acceleration."
Comparing the computed accelerograms using the two methods, Figures Q9-1, Q9-2, it seems clear that the generalized ray method is giving a limited view ofthe ground response.
To say that the rhultiples are coherent, and thus irrelevant, presupposes the answer.
One has to wonder what an accelerogram would look likeifthe fullGreen's functions (frequency-wavenumber) were used and not a simple set ofgeneralized rays. Although the comparison might be more exact between the tangential components for reasons given in the response, what do the radial and vertical components look like for the frequency wavenumber computations? I would like to see those plots at the next meeting.
The averaged response spectrum is presumably the average response of the 22 traces shown in either Figure Q9-1 or Figure Q9-2. Which Figure, i.e., which depth.
It's clear that the frequency-wavenumber method produces a significant peak in the response spectrum in precisely the 3-8.5 Hz range. Is this important to the engineering analysis that follows?
Question 15.
The numerical simulations were not used to assist in determining the relations for magnitude and distance.
Why? I thought a primary concern was the lack of data within 10-20 km.
Thus the numerical simulations were meant to fillin that data gap. The curve in Figure Q15-1 is
constrained by only seven data points. The curves are extrapolations Gum distances greater than 20 km (See Figure Q6-3). The regression analysis used by PG&E has a pseudo distance factor that dominates at the 4.5 km epicentral distance ofthe site. The pseudo distance is the factor 0.616 exp (0.524 M) which is equal to 27 km for a M7.2 earthquake and 19 km for a 6.5. This pseudo distance is added to the epicentral distance (4.5 km) and used for the regression analysis. In short, the distance of4.5 km makes a 25% difference for a 6.5 and about a 20% difference for a 7.2..I have always been confused why the peak acceleration should have a distance dependence like 1/R2. (1/R Rom geometrical spreading and 1/R Gom intrinsic attenuation?)
Imcall PG&E (Sadigh) once showing that the footwall accelerations were no different Gom the hanging wall. That is contradicted in the statements concerning the style offaulting. How can the average radiation coefficient for a thrust event be larger than a strike-slip event. For example, consider a vertical fault. In one case the style is strike-slip, in the other reverse. Would two such faults produce equal amplitudes within one fault depth?
Question 17.
The style offaulting is the one number consistently less than 1.0 in converting Rom thrust event recordings to strike-slip for the site. Again the issue of style offaulting indicates that thrust events produce higher accelerations.
Yet, in the response to Question 18 PG&E argues that stress drops are not higher for thrust compared to strike-slip.
Question 18.
The average stress drop of50 bars and the local stress drop of500 bars seems to encompass a wide range ofpossibilities. The numerical simulations may be overestimating the strike-slip ground motion ifa real difference exists between thrust and strike-slip faulting.
Question 19.
The issue ofdircctivityis contmversial. UC Berkeley willget a different Ml.from Caltech and seismologists attribute this to directivity. MLthough is based on 1.0 Hz, a fairlyhigh frequency. Differences for the 1922, 1934 and 1966 Parkfield earthquakes are noted for stations south and north of the epicentral region. Is this directivity? Iquestion whether ~vr~ae horizontal peak accelerations willever show directivity. Why average the results Aom four test cases?
Why not show each one separately and let us see what the results look like? I do like the plots in the figures for this question; I would have preferred to see each of the four runs separately plotted in this manner.
Sincere y, Ralph J. Archuleta
~
i1