ML20141E189
| ML20141E189 | |
| Person / Time | |
|---|---|
| Issue date: | 12/27/1985 |
| From: | Zurflueh E NRC OFFICE OF NUCLEAR REGULATORY RESEARCH (RES) |
| To: | Beratan L NRC OFFICE OF NUCLEAR REGULATORY RESEARCH (RES) |
| References | |
| NUDOCS 8601070631 | |
| Download: ML20141E189 (6) | |
Text
_.
~
- DEC 2 71985 MEMORANDUM FOR: Leon L. Beratan, Chief Earth Sciences Branch, DRPES
'THRU:
Andrew J. Murphy, Section Leader Seismology Section Earth Sciences Branch, DRPES FROM:
Ernst G. Zurflueh, Geophysicist Seismology 3ection Earth Sciences Branch, DRPES
SUBJECT:
EPRI POST-PEER REVIEW WORKSHOP 2
On October 30 and 31, 1985 I participated in a workshop in Denver to review post-peer review comments on the EPRI seismic hazard program. Participants at the meeting included EPRI personnel and consultants working on the seismic hazard program. Dr. Reiter of NRR and I were present as NRC observers.
The objective of the workshop was to review additional information and results of methodology. evaluations, present methodology improvements, and clarify residual concerns. Dr. Stepp presented an outline of the EPRI phase 2 program which is summarized in Table 1.
Dr. King discussed the reevaluation of the earthquake catalog and posed a series of questions for discussion as shown in Table 2.
Dr. Coppersmith together with Drs. Johnston and Arabasz then presented an evaluation of the maximum magnitude of earthquakes.
They stated that there are fundamental differences in source characteristics between the Eastern and Western U.S.
This is related to the question of how an intraplate earthquake is defined as compared to an interplate earthquake. A consistent view of this difference has, so far, not emerged. There are also divergent views _ of the importance of the historical record of seismicity.
In one sense, the historical record does not tell us much about what will happen in the future; in another. sense, the historical record is very important because without it an analysis of seimicity would not be possible.
An analysis of worldwide intraplate earthquakes was done using the following criteria to define such earthquakes:
7 1.
They occur within the continental crust.
- 2..
Mesozoic / Cenozoic orogenic belts are excluded.
3.-
Regions' of Neogene rifting or volcanism are excluded.
8601070631 851227 0FC:
-NAME:
DATE:
9 r "^ " 7 1995 Among the intraplate earthquakes, the Baffin Bay earthquakes may have been caused by normal faulting; otherwise there are very few examples of this type of movement. As an intraplate analog area, Australia may be one of the best examples. Central Asia has magnitude 8 earthquakes but will be excluded because the high magnitude may be associated with processes other than definite intraplate earthquakes.
If Central Asia is excluded, the New Madrid earthquake is a very unusual event. The Lisbon earthquake may be comparable, unless it is related to the Azores fracture zone. At present, the assumption used in these studies is that there is no gradation between intra-and interplate earthquakes.
In any case, the definition of what is an intraplate earthquake is seen as the key to interpreting the global data base.
Next, Dr. McGuire presented results of his parametric analyses. The uncertainties in models of seismicity are related to uncertainties in maximum magnitude (source) and in attenuation. Dr. McGuire computed four plots of uncertainty in terms of annual probability of exceedance vs. spectral velocity at 10Hz for each site:
Total uncertainty by team Uncertainty related to source zones Uncertainty related to maximum magnitude Uncertainty with smoothing option for seismicity parameters.
In analyzing the plots, a total range of variation of less than a factor of two is called low sensitivity, ranges higher that a~ factor of five high sensitivity, and intermediate ranges medium sensitivity.
For this analysis, all other factors and their possible combinations were averaged, not held constant. A criticism leveled at the analysis was that this is not a real sensitivity analysis because various things are lumped together.
In summary some teams showed high sensitivity on most factors, whereas other tended towards low sensitivity.. Among the factors analyzed for different sites and different teams, high sensitivity was found in 19 cases (35%) for seismicity, 13 cases (27%) for magnitude and I case (2%) for parameters.
Medium sensitivity was found in 7 cases (13%) for seismicity, *; cases (13%) for magnitude and 5 cases (9%) for parameters.
Comparisons of the EPRI study with that performed by LLNL for the NRC have shown that, in general, results are similar. The largest difference was found
.in the case of Shearon Harris, where LLNL results predicted a somewhat higher hazard. Different activity rates at magnitudes greater than five are responsible for the discrepancy. Among the various parameters, a and b values seem to be the most well defined, meaning they show-no difference between the 0FC:
NAME:
DATE:
-3_
DEC 2 71985 two studies. For further comparison, Dr. King suggested making up test cases with all starting parameters given and comparing analyses done by EPRI and LLNL.
The next item of discussion was the analysis of seismicity via the tectonic framework.
For example, in the case of the Charleston seismicity, no specific structure has been associated with the earthquakes. With respect to earthqu.tke processes in general, it has not been possible to find a basis for predicting that an earthquake will occur at a given location.
Finally, the probabilit3 distribution is also uncertain; for instance, the likelihood of a similar earthquake occurring outside the Charleston area is still controversial.
General observations on tectonic framework analysis can be stated as follows:
Hazard analysis is not a prediction.
Ability to define causative structures post facto has been low.
Similar tectonic features may have very different probabilities.
Stress information has limited use; only the orientation of faults has been important.
There is skepticism regarding geologic / geodetic indications of activity.
Although difficult, the tectonic framework approach works.
~
The discussion then focused on the significance of P* and P**.
P* is defined as the probability that a specific source zone is active and can produce an earthquake of m greater than or equal to 5.
P** is the probability that a particular set of seismic sources is the " correct" set of active sources in a region, and it is derived from P* with additional tonsiderations including the relation: sum of all possible P**=1.
This leads to the question of background areas and their relationship whith tectonic features.
For instance, for the Gulf Coast region most teams except Law Engineering show a high probability of over.75 for earthquakes with m greater than or equal to 5.
A further question is whether a high P* implies high hazard. This would seem logical, but in some instances it is not the case. As a personal observation I would like to add that, with a limited number of source zones, a somewhat different choice of P*s (that is assigning different relative importance to different sources) can produce a very different outcome for the overall region, particularly with respect to seismicity assigned to background areas.
A subsequent discussion of the methodology used for this study again came back to the question of background. One reviewer's comment was that background is a
" cop-out", has no tectonic basis. A partial answer to the problem was that 0FC:
NAME:
DATE:
TABLE 1.
EPRI Seismic Hazard Program--Phase 2 t
1.
Development of Basic Information i
Earthquake Catalog Estimation of Maximum Earthquake Magnitude Alternative Seismicity Models I
2.
Methodology Review and Evaluation Cocparison with NRC/LLNL Seismic Hazard Methodology Parametric Analysis and Sensitivity Tests Peer Review 3..
. Methodology Improvements Estimation of Seismicity Parameters 4.
Revised Interpretations of Earthquakes Source Zones and Seismicity Parameters
. Site-Specific Interpretations for Nine Test Sites I
1 t
i y
TABLE-2.
Issues for Discussion 1.
What is the meaning of P*?
a)
Is it a probability of activity for.oiven tectonic conditions?
b)
Is time a consideration?
2.
What is trade-off between P* and rate of activity?
.3.
How does P* impact hazard estimates?
~
4.
Linkage of seismicity parameter estimation with standard approaches.
5.
Use of background zones.
i 6.
What are major contributors to differences in hazard results among tests?
7.
Are elements of the tectonic characteristics matrix independent?
8.
Are modifications to the methodology required for site-specific application and seismic hazard integration?
9.
How can interpretations be calibrated?
i
~.
DEC 2 ' M background areas are chosen by default, in other words are based on a lack of knowledge. Consequently background seismicity is one of the areas in which differences between different teams were greatest. As mentioned in the paragraph on P**, often an assignment of low ~ seismicity to source zones moves a high seismicity level into large background areas.
It was suggested that a team calibration be done based on a seismicity budget.
The question to be answered in this procedure would be: Do our predictions match historical observations?. A weighted average of all possible source combinations would be taken and compared with historical seismicity at each magnitude level.
A large number of model analyses and resulting graphs were presented by Dr.
Veneziano. For future analysis he suggested that the Eastern U.S. be divided into regions of incompleteness which are independent of tectonic zones. The-incompleteness of the data set would then be analyzed and earthquakes classified according to regions of incompleteness. By using only reliable portions of the historic record, the uncertainty in magnitude relationships could be quantified.
Finally the discussion centered around methods of aggregating the results.
It was stated that behavioral aggregation is more critical than mechanical aggregation, and each team was asked to state its assumptions explicitly.
Results could be aggregated at the final results level or at the components level. Aggregation at the components level would require consensus of the different teams. EPRI used the approach of aggregation at the final results level. This permits different interpretations by different teams and is considered a strong point of the EPRI approach.
Ernst G. Zurflueh, Geophysicist Earth Sciences Branch, DRPES
Enclosures:
1.
Table 1 2.
Table 2 I FES Filc9
! subjcet File 50-Distribution /R-2811 Circ /Chron RMinogue EConti j
i GlCS/PDR Dross LBeratan i
ESB Sbj/Rd KGoller AMurphy EZurflueh a ra. _
- " *_ n,
~ ' ~ ~ ~
0FC: RES;ESB RES:ESB : ?R S ESB :
NAME: EZurflueh
- AMurphy
- LBe tan *
._ m l
" "~":'
~~
DATE
DEC 3 I 1985 MEMORANDUM FOR: Leon L. Beratan, Chief Earth Sciences Branch, DRPES THRU:
Andrew J. Murphy, Section Leader Seismology Section Earth Sciences Branch, DRPES FROM:
Ernst G. Zurflueh, Geophysicist Seismology Section Earth Sciences Branch, DRPES
SUBJECT:
EPRI POST-PEER REVIEW WORKSHOP On October 30 and 31, 1985 I participated in a workshop in Denver to review post-peer review comments on the EPRI seismic hazard program. Participants at the meeting included EPRI personnel and consultants working on the seismic hazard program. Dr. Reiter of NRR and I were present as NRC observers.
The objective of the workshop was to review additional infonnation and results of methodology evaluations, present methodology improvements, and clarify residual concerns. Dr. Stepp presented an outline of the EPRI phase 2 program which is summarized in Table 1.
Dr. King discussed the reevaluation of the earthquake catalog and posed a series of questions for discussion as shown in Table 2.
Dr. Coppersmith together with Drs. Johnston and Arabasz then presented an evaluation of the maximum magnitude of earthquakes. They stated that there are fundamental differences in source characteristics between the Eastern and Western U.S.
This is related to the question of how an intraplate earthquake is defined as compared to an interplate earthquake. A consistent view of this difference has, so far, not emerged. There are also divergent views of the importance of the historical record of seismicity.
In one sense, the historical record does not tell us much about what will happen in the future; in another sense, the historical record is very important because
~
without it an analysis of seimicity would not be possible.
An analysis of worldwide intraplate earthquakes was done using the following criteria to define such earthquakes:
1.
They occur within the continental crust.
2.
Mesozoic / Cenozoic orogenic belts are excluded.
3.
Regions of Neogene rifting or volcanism are excluded.
_. /1A L H L rw*
DATE:
DEC 2 71985 2-Among the intraplate earthquakes, the Baffin Bay earthquakes may have been caused by normal faulting; otherwise there are very few examples of this type of movement. As an intraplate analog area, Australia may be one of the best examples. Central Asia has magnitude 8 earthquakes but will be excluded because the high magnitude may be associated with processes other than definite intraplate earthquakes.
If Central Asia is excluded, the New Madrid earthquake is a very unusual event. The Lisbon earthquake may be comparable, unless it is related to the Azores fracture zone. At present, the assumption used in these studies is that there is no gradation between intra-and interplate earthquakes.
In any case, the definition of what is an intraplate earthquake is seen as the key to interpreting the global data base.
Next Dr. McGuire presented results of his parametric analyses. The uncertainties in models of seismicity are related to uncertainties in maximum magnitude (source) and in attenuation. Dr. McGuire computed four plots of uncertainty in terms of annual probability of exceedance vs. spectral velocity at 10Hz for each site:
Total uncertainty by team Uncertainty related to source zones Uncertainty related to maximum magnitude Uncertainty with smoothing option for seismicity parameters.
In analyzing the plots, a total range of variation of less than a factor of two is called low sensitivity, ranges higher that a factor of five high sensitivity, and intermediate ranges medium sensitivity.
For this analysis, all _ other factors and their possible combinations were averaged, not held constant. A criticism leveled at the analysis was that this is not a real sensitivity analysis because various things are lumped together.
In summary some teams showed high sensitivity on most factors, whereas other tended towards low sensitivity. Among the factors analyzed for different sites and different teams, high sensitivity was found in 19 cases (35%) for seismicity, 13 cases (27%) for magnitude and I case (2%) for parameters.
Medium sensitivity was found in 7 cases (13%) for seismicity, 7 cases (13%) for magnitude and 5 cases (9%) for parameters.
Comparisons of the EPRI study with that performed by LLNL for the NRC have shown that,.in general, results are similar. The largest difference was found in the case of Shearon Harris, where LLNL results predicted a somewhat higher hazard.
Different activity rates at magnitudes greater than five are responsible for the discrepancy. Among the various parameters, a and b values seem to be the most well defined, meaning they show no difference between the OFC:
NAME:
DATE:
i DEC 2 7 E95
- two studies. For further comparison, Dr. King suggested making up test cases with all starting parameters given and comparing analyses done by EPRI and LLNL.
The next item of discussion was the analysis of seismicity via the tectonic framework. For example, in the case of the Charleston seismicity, no specific structure has been associated with the earthquakes. With respect to earthquake processes in general, it has not been possible to find a basis for predicting that an earthquake will occur at a given location.
Finally, the probability distribution is also uncertain; for instance, the likelihood of a similar earthquake occurring outside the Charleston area is still controversial.
General observations on tectonic framework analysis can be stated as follows:
Hazard analysis is not a prediction.
Ability to define causative structures post facto has been low.
Similar tectonic features may have very different probabilities.
Stress information has limited use; only the orientation of faults has been important.
There is skepticism regarding geologic / geodetic indications of activity.
-Although difficult, the tectonic framework approach works.
The discussion then focused on the significance of P* and P**.
P* is defined as the probability that a specific source zone is active and can produce an earthquake of m greater than or equal to 5.
P** is the probability that a particular set of seismic sources is the " correct" set of active sources in a region, and it is derived from P* with additional considerations including the relation: sum of all possible P**=1.
This leads to the question of background areas and their relationship whith tectonic features.
For instance, for the Gulf Coast region most teams except Law Engineering show a high probability of over.75 for earthquakes with m greater than or equal to 5.
A further cuestion is whether a high P* implies high hazard. This would seem logical, but in some instances it is not the case. As a personal observation I would like to add that, with a limited number of source zones, a somewhat different choice of P*s (that is assigning different relative importance to different sources) can produce a very different outcome for the overall region, particularly with respect to seismicity assigned to background areas.
A subsequent discussion of the methodology used for this study again came back to the question of background. One reviewer's comment was that background is a
" cop-out", has no tectonic basis. A partial answer to the problem was that 0FC:
NAME:
DATE:
i
?
TABLE 1.
EPRI Seismic Hazard Program--Phase 2 1.
Development of Basic Information Earthquake Catalog Estimation of Maximum Earthquake Magnitude Alternative Seismicity Models 2.
Methodology Review and Evaluation Comparison with NRC/LLNL Seismic Hazard Methodology Parametric Analysis and Sensitivity Tests Peer Review 3.
Methodology Improvements Estimation of Seismicity Parameters 4
Revised Interpretations of Earthquakes Source Zones and Seismicity Parameters Site-Specific Interpretations for Nine Test Sites l
1,
~.
TABLE 2.
Issues for Discussion i
I..
What is the meaning of P*?
I a)
Is it a probability of activity for given tectonic conditions?
I b)
Is time a consideration?
2.
What is trade-off between P* and rate of activity?
]
3.
.How does P* impact hazard estimates?
4.
Linkage of seismicity parameter estimation with standard approaches.
5.
Use of background zones.
6.
What are major contributors to differences in hazard results among tests?
[
l j
7.
Are~ elements of the tectonic characteristics matrix independent?
8.
Are modifications to the methodology required for site-specific application and seismic hazard integration?
9.
How can interpretations be calibrated?
i i
i i
3 4
i I
4
~
n;pt ;
On background areas are chosen by default, in other words are based on a lack of knowledge. Consequently background seismicity is one of the areas in which differences between different teams were greatest. As mentioned in the paragraph on P**, often an assignment of low seismicity to source zones moves a high seismicity level into large background areas.
It was suggested that a team calibration be done based on a seism.icity budget.
The question'to be answered in this procedure would be: Do our predictions match historical observations? A weighted average of all possible source combinations would be taken and compared with historical seismicity at each magnitude level.
A large number of model analyses and resulting graphs were presented by Dr.
Veneziano. For future analysis he suggested that the Eastern U.S. be divided into regions of incompleteness which are independent of tectonic zones. The incompleteness of the data set would then be analyzed and earthquakes classified according to regions of incompleteness. By using only reliable portions of the historic record, the uncertainty in magnitude relationships could be quantified.
Finally the discussion centered around methods of aggregating the results.
It was stated that behavioral aggregation is more critical than mechanical aggregation, and each team was asked to state its assumptions explicitly.
Results could be aggregated at the final results level or at the components level. Aggregation at the components level would require consensus of the different teams. EPRI used the approach of aggregation at the final results level. This permits different interpretations by different teams and is considered a strono point of the EPRI approach.
I Ernst G. Zurflueh, Geophysicist Earth Sciences Branch, DRPES l
Enclosures:
1.
Table 1 2.
Table 2 Distribution /R-2811 Circ /Chron RMinogue EConti DCS/EDR Dross LBeratan ESB Sbj/Rd KGoller AMurphy EZurflueh 0FC: RES;ESB RES:ESB:
R S ESB :
NAME: EZurflueh
- AMurphy
- LBe tan
- DATE:lh26-85 ps 9 i
J
-