draft revised response for nrc rai letter no. 037 (erai ... · approach to define the site-specific...

Proposed Turkey Point Units 6 and 7 Docket Nos. 52-040 and 52-041 FPL Draft Revised Response to NRC RAI No. 02.05.02-2 (eRAI 5896) of 73

NRC RAI Letter No. PTN-RAI-LTR-037

SRP Section: 02.05.02 - Vibratory Ground Motion

Question for Geosciences and Geotechnical Engineering Branch 1 (RGS1)

NRC RAI Number: 02.05.02-2 (eRAI 5896)

FSAR Subsection 2.5.2.4.5 describes new ground motion prediction equations (GMPEs) that the applicant developed for Caribbean region seismic sources. The basis of the new GMPEs is a scientific study conducted by Motazedian and Atkinson (2005) in the Puerto Rico area. In accordance with NUREG-0800, Standard Review Plan, Chapter 2.5.2, "Vibratory Ground Motion," and Regulatory Guide (RG) 1.208, "A Performance-Based Approach to Define the Site-Specific Earthquake Ground Motion":

a. Please explain why ground motion prediction models developed by Motazedian and Atkinson (2005) for the Puerto Rico region, which is primarily a subduction zone, provide an adequatebasis for the larger Caribbean region, especially for the region between Cuba and Florida.

b. Please provide the complete Senior Seismic Hazard Analysis Committee (SSHAC) documentation for the Level 2 study conducted to develop the Caribbean GMPEs for the staff to specifically evaluate the makeup of the Technical Integrator (TI) team, the peer review panel, how the experts’ opinions were integrated into the development of the final GMPE, whether any conflicting opinions among the experts were dealt with, and how the final GMPEs represent the consensus of the informed community.

c. Please provide copies of the following supporting calculations: Report #: 25409-000-K0C- 0000-00009, Report#: 25409-000-K0C-0000-00024, Report #: 25409-000-K0C-0000-00034 to enable the staff to evaluate the technical details of the final GMPEs.

d. In order for the staff to be able to compare the new Caribbean GMPEs with the 2004/2006 EPRI mid-continent GMPEs, please provide plots showing both ground motion models for earthquake magnitudes of 6.0, 7.0, and 8.0 in the distance range of 200 km to 1000 km at all seven frequencies defined in the EPRI 2004 and 2006 GMPEs.

e. Discuss evidence, if any, that seismic source scaling varies regionally and/or between source types in the Caribbean. For example, is there any evidence that stress parameter varies systematically between the northern Hispaniola sources, the Caribbean plate-boundary transform fault sources, and the Cuba sources? If so, what are the implications for the attenuation models and hazard calculated at TPNPP?

DRAFT


FPL RESPONSE:

Part a: Please explain why ground motion prediction models developed by Motazedian and Atkinson (2005) for the Puerto Rico region, which is primarily a subduction zone, provide an adequate basis for the larger Caribbean region, especially for the region between Cuba and Florida.

The Motazedian and Atkinson (FSAR Reference 287) ground motion prediction equations (GMPEs) are not used directly to model attenuation of strong ground motion between Cuba (and the northern Caribbean) and southern Florida. Rather, after performing a literature review of possibly applicable GMPEs, the regional attenuation and source parameters developed by Motazedian and Atkinson from a Puerto Rico earthquakes dataset were selected as reasonable starting points from which to develop simulation-based ground motions for the estimation of a regional suite of GMPEs for use in the Turkey Point PSHA.

As noted in Garcia et al. (FSAR Reference 255), there are currently no calibrated GMPEs for Cuba and the surrounding region. Strong motion instruments have only recently been installed in 1998 and the amount of recorded empirical ground motion data from these instruments, which is operated by the Cuban agency (FSAR Reference 255), is limited or unavailable. In the absence of a local regionally based empirical GMPE, Garcia et al. (FSAR Reference 255) employed the use of other GMPEs from similar regions. Specifically the rock versions of the Dahle et al. [1995] and Ambraseys et al. [1996] relations were used for the rock seismic hazard assessment; the stiff soil version of the Ambraseys et al. [1996], the soil version of the Dahle et al. [1995], and the Motazedian and Atkinson relation (FSAR Reference 287) GMPE, which was developed for Puerto Rico for soft rock site conditions, were used for the soil hazard computation.

The Motazedian and Atkinson (FSAR Reference 287) study uses a dataset of approximately 300 earthquakes of magnitude 3.0-5.5 located on and around the island of Puerto Rico to develop regional ground motion attenuation and source parameters. Motazedian and Atkinson acknowledge that their ground motion dataset consisted of recorded ground motions from both crustal and subduction zone earthquakes and that the separation of the earthquakes used in their dataset into crustal and subduction events was not possible due to the limited data precision for Puerto Rico events.

To address this possible concern about the influence of using both subduction and crustal events to estimate the regional attenuation and source parameters, Motazedian and Atkinson (FSAR Reference 287) provide a comparison of their GMPE with representative GMPEs for central and eastern United States (FSAR Reference 210), California (Atkinson and Silva, 2000), and an empirically-based GMPE based on the global ground motion dataset for subduction zones (Atkinson and Boore, 2003). Motazedian and Atkinson conclude that, overall the Puerto Rico relations agree well with the stochastic relations for California and eastern North America and are quite different from those calculated by the global subduction relations. This comparison and noted results provide a technical justification for using the source and attenuation parameters from the Puerto Rico ground motion dataset (FSAR Reference 287) as the starting point for the development of applicable GMPEs for the Caribbean seismic sources.

DRAFT


In developing the Caribbean GMPEs, a number of sensitivity studies were performed, including some that considered a range of anelastic attenuation parameters in an attempt to capture uncertainty in the attenuation of ground motion over propagation paths that, in part, run through the Gulf Coast region crust between the northern Caribbean and southern Florida. Details of the development and validation of the resulting Caribbean GMPEs are presented below in the response to Part b.

Part b: Please provide the complete Senior Seismic Hazard Analysis Committee (SSHAC) documentation for the Level 2 study conducted to develop the Caribbean GMPEs for the staff to specifically evaluate the makeup of the Technical Integrator (TI) team, the peer review panel, how the experts’ opinions were integrated into the development of the final GMPE, whether any conflicting opinions among the experts were dealt with, and how the final GMPEs represent the consensus of the informed community.

As noted above, only limited empirical strong motion attenuation information has been found for the Caribbean, and existing studies of earthquake hazard depend significantly on adoption of ground motion prediction equations (GMPEs) from other regions. No studies have been found modeling attenuation of strong ground motion between earthquakes in the Caribbean and sites in the CEUS so that no experts could be identified who could be characterized as a proponent, which NUREG/CR-6372 defines as “an expert who advocates a particular hypothesis or technical position” (Senior Seismic Hazard Analysis Committee [SSHAC], 1997). The absence of a proponent or proponents made it difficult to categorize the level of the study per the SSHAC guidelines. However, what has been done in this regard is otherwise fully consistent with industry practice in general and the SSHAC guidelines in particular as will be more fully discussed in the SSHAC Study Summary section of this response.

Initial development of a suite of Caribbean GMPEs was performed by, what in SSHAC nomenclature would be called, the Technical Integration team of two Bechtel seismologists: Drs. Nick Gregor and Behrooz Tavakoli, both with significant experience in generating and evaluating GMPEs. Peer review was provided by a Technical Advisory Group (TAG) at several points throughout the GMPE development process. A resource expert, Dr. Dariush Motazedian of Carleton University, Ottawa, Canada, was also consulted during this time.

This response first summarizes in more detail the initial development of the Caribbean GMPEs, which included, with the participation, guidance, and acceptance of the TAG, an examination of the sensitivity of these GMPEs to alternative source and anelastic attenuation parameters. This response then goes on to discuss the subsequent further confirmation of the reasonableness and conservatism of these Caribbean GMPEs as obtained through both the performance of additional sensitivity studies and the solicitation of comments from a panel of additional recognized strong ground motion attenuation experts.

DRAFT


Initial GMPE Development Summary

A suite of GMPEs was developed with guidance from a Technical Advisory Group (TAG) whose members were:

Dr. Robert Kennedy (RPK Structural Mechanics Consulting)

Dr. William McCann (Earth Scientific Consultants)

Mr. Donald Moore (Southern Nuclear Operating Company)

Dr. J. Carl Stepp (Earthquake Hazards Solutions)

Dr. Robert Youngs (Geomatrix Consultants, currently AMEC-Geomatrix)

Three TAG meetings and an additional TAG review after the third meeting provided a forum for the TAG members to review progress and to offer participatory peer review guidance on the development of the Caribbean GMPEs.

Initially a literature review was performed with the goal of retrieving acceptable GMPEs for the Cuba and Caribbean region. However, this literature review only retrieved one GMPE developed recently by Motazedian and Atkinson (FSAR Reference 287). In addition to the interaction with the TAG members, correspondence was conducted with Professor Motazedian during the initial development of the Caribbean GMPEs. These initial technical discussions were for the possible application of the published Motazedian and Atkinson (FSAR Reference 287) GMPE for the PSHA study. However, based on the limitations of this GMPE (e.g., incomplete suite of necessary spectral frequencies, limited application for distances greater than 500 km, and site-specific ground conditions of soft rock), it was determined that this published GMPE was not directly acceptable for use in the PSHA.

For the development of the GMPEs for the Cuba and Caribbean region, the regionally determined source and attenuation parameters from Motazedian and Atkinson (FSAR Reference 287) were used with a point-source stochastic ground motion methodology to develop an applicable large-distance GMPE for the magnitude and distance range needed for the PSHA.

At the suggestion of TAG members over the course of the three TAG meetings, alternative attenuation relationships were considered to examine the sensitivity of the initial GMPEs to alternative source and anelastic attenuation properties. Specifically, the effect on GMPE results of adopting a double corner (2C) seismic source model was examined, recognizing that the active northern Caribbean plate margin might be expected to act like a more active tectonic environment such as the western United States (WUS) rather than the less active tectonic environment of the central and eastern United States (CEUS). In addition, a Gulf Coast region anelastic attenuation factor (Q) model rather than the Puerto Rico region-specific Q model from Motazedian and Atkinson (FSAR Reference 287) was analyzed, recognizing that much of the propagation path from the Caribbean sources to the Turkey Point Units 6 & 7 site is through the Gulf Coast crust.

It was found that adoption of these alternatives (i.e., different suite of regional attenuation and seismic source parameter values) led to ground motion values that were equal to (at large distances based on the anelastic attenuation rates) or lower than (based on the different magnitude scaling from the WUS- based double corner model) the suite of original nine new ground motion attenuation models adopted for the Cuba and Caribbean region. A

DRAFT


comparison of the weighted combination of the original nine GMPE models and the inclusion of these additional sensitivity models resulted in a slightly lower weighted mean ground motion attenuation curve over the magnitude and distance range needed for the PSHA. Therefore, their incorporation into the final PSHA results would have slightly lowered the already low hazards, and thus, the use of the original nine GMPE models was accepted by the TAG members because the inclusion of these additional GMPE models would be expected to lead to lower, and less conservative, ground motion results.

In a review of these results after the third and final TAG meeting, and in a comparison of initial GMPE models with the accepted EPRI 2004 eastern US ground motion models, the TAG concluded that the results satisfactorily responded to their recommendations.

With satisfaction of all TAG recommendations, the original suite of nine GMPE models was concluded to be acceptable and was used for the Turkey Point Units 6 & 7 FSAR. In response to this RAI, supplemental sensitivity studies were performed and solicitation of additional expert comment has been made. The results of this additional work are presented below in the Supplemental Sensitivity Studies and SSHAC Study Summary sections, respectively.

Supplemental Sensitivity Studies

After the suite of GMPEs was developed, two additional sensitivity analyses were performed to further validate the technical assessment of the stochastic Caribbean GMPE models developed for Turkey Point Units 6 & 7. These analyses sought additional empirical data with which to compare the Caribbean GMPEs and examined the effect of alternate suites of GMPEs on the PSHA results at the site. The results of these two additional sensitivity studies are presented here.

The first supplemental sensitivity analysis compared the suite of GMPEs used in the Turkey Point Units 6 & 7 FSAR PSHA with empirical regional ground motion data. To assist in the technical evaluation of the current Caribbean GMPE models developed for the Turkey Point Units 6 & 7 FSAR a comparison of empirical ground motions from five regional earthquakes occurring since 2004 in the Gulf of Mexico and northern Caribbean region were analyzed. These earthquakes were selected from among those available as being most representative of earthquakes whose tectonic settings and locations might be expected to be like those of future earthquakes contributing to the overall PSHA at the site under the source model adopted. Specifically, these were shallow crustal earthquakes in the northern Caribbean or Gulf of Mexico.

Regional broadband empirical data from selected Incorporated Research Institutions for Seismology (IRIS) stations were obtained, processed and compared to the suite of Caribbean GMPE models and the suite of EPRI (FSAR References 242 and 203) GMPE models for both the Mid-continent and Gulf Coast regional models. Based on these comparisons, a technical assessment can be made on the applicability of using the Caribbean GMPE for the Turkey Point Units 6 & 7 PSHA.

Note that the current PSHA results were developed using the suite of EPRI (FSAR References 242 and 203) Mid-continent GMPE models for other non-Caribbean sources. In this comparison all of the GMPE attenuation curves are for the assumed CEUS hard rock

DRAFT


site conditions with Vs= 2.83 km/sec, whereas the empirical IRIS data are for the individual unknown site conditions of each station which is expected to be less than Vs= 2.83 km/sec. Based on simplified site response analysis it would be expected that an adjustment for the station specific site conditions to the more stable CEUS hard rock site conditions would require a reduction factor in the observed empirical ground motion values.

The current deaggregation of seismic hazard at the Turkey Point Units 6 & 7 site from all sources, Cuban, Caribbean, and southeast United States, is shown in Figure 1 (FSAR Figure 2.5.2-226 and Figure 2.5.2-228) for longer period motions (for which the relative contribution of the larger, more distant Cuban and Caribbean sources would be expected to have their greatest relative contribution) and for the 10-4 and 10-5 mean annual frequency of occurrence probabilities used to develop design ground motions under current NRC regulatory guidance. The relative contribution for the higher frequency cases of 5 and 10 Hz (see FSAR Figures 2.5.2-227 and 2.5.2-229) from the more distant Caribbean sources is significantly reduced relative to the closer local seismic sources. For this sensitivity comparison, the empirical ground motions for 1 and 2.5 Hz are presented for each of the five regional earthquakes and GMPEs.

Figure 1. Magnitude and Distance Deaggregation for 1 and 2.5 Hz at 10-4 and 10-5 Annual Frequencies of Exceedance.

DRAFT


The five regional earthquakes that have been analyzed are:

12/14/2004 – Caribbean Sea Region (Mw6.8, 19.05 N, -81.52 W, hypocenter depth 12.0 km)

09/10/2006 – Gulf of Mexico (Mw5.9, 26.32 N, -86.84 W, hypocenter depth 29.6 km) 02/04/2007 – Cuba Region (Mw6.2, 19.49 N, -78.34 W, hypocenter depth 12.0 km) 05/28/2009 – North of Honduras (Mw7.3, 16.50 N, -87.17 W, hypocenter depth 12.0 km) 01/12/2010 – Haiti (Mw7.0, 18.61 N, -72.62 W, hypocenter depth 12.0 km)

Note that based on the hypocentral location and geographical location relative to tectonic plate boundary, all five of these earthquakes are considered to be shallow crustal events and are not associated with any regional subduction zones. For each event, a standard time history processing methodology was applied with the final results being a dataset of the acceleration response spectra for a spectral damping of 5 percent at each station. The geometric mean of the two horizontal components was computed and comparison plots for the 1 and 2.5 Hz spectral frequencies were generated showing the empirical data and the suite of GMPE models (i.e., both of the EPRI, 2004 GMPE models and the Caribbean GMPE models). Both the individual attenuation curves and the weighted mean attenuation curve for a given set are shown in the comparison plots for each spectral frequency and earthquake.

Caribbean Sea Region (Mw6.8), 12/14/2004:

This event occurred in the Caribbean Sea region and its epicenter is shown in Figure 2a along with the Turkey Point Units 6 & 7 site location, and the location of the IRIS stations that recorded this earthquake. Based on the observed station distribution, the IRIS station DWPF located in central Florida has the most applicable source to site path for this comparison. The attenuation curves and empirical data are shown in Figures 2b and 2c for the 1 and 2.5 Hz spectral frequencies. The data point from the DWPF station is highlighted as a solid blue circle. Overall the empirical data falls below the median Caribbean attenuation curve (heavy red line) and has values which are in the lower distribution range of Caribbean attenuation curves. In addition, the observation from the DWPF station is lower than the entire range of the Caribbean attenuation curves. DRAFT


Figure 2a. Map showing the earthquake location for the 12/14/2004 Caribbean Sea Region earthquake (Mw6.8), the IRIS station locations used in the analysis (shown as black triangles in this and similar subsequent figures), and Turkey Point Units 6 & 7 site location. DRAFT


0.000001

0.00001

0.0001

0.001

0.01

0.1

1

100 1000 10000

Sp

ec

tra

l Ac

ce

lera

tio

n (g

)

Hypocentral Distance (km)

Attenuation Models: 1Hz, 12/14/2004, M6.8

Mean 1CC-QBase-SDAll 1CC-QHigh-SDAll 1CC-QLow-SDAll 1CV-QBase-SDAll

1CV-QHigh-SDAll 1CV-QLow-SDAll 2C-QBase 2C-QHigh 2C-QLow

EPRI(Mean) MidCont EPRI(1) MidCont EPRI(2) MidCont EPRI(3) MidCont EPRI(4) MidCont

EPRI(5) MidCont EPRI(6) MidCont EPRI(7) MidCont EPRI(8) MidCont EPRI(9) MidCont

EPRI(Mean) Gulf EPRI(1) Gulf EPRI(2) Gulf EPRI(3) Gulf EPRI(4) Gulf

EPRI(5) Gulf EPRI(6) Gulf EPRI(7) Gulf EPRI(8) Gulf EPRI(9) Gulf

IRIS Data

Figure 2b. Comparison of Caribbean GMPE (red lines), EPRI Mid-continent (blue lines) and Gulf Coast region (green lines) GMPE and empirical IRIS processed data for a spectral frequency of 1 Hz. DRAFT


0.000001

0.00001

0.0001

0.001

0.01

0.1

1

100 1000 10000

Sp

ec

tra

l Ac

ce

lera

tio

n (g

)


Attenuation Models: 2.5Hz, 12/14/2004, M6.8Mean 1CC-QBase-SDAll 1CC-QHigh-SDAll 1CC-QLow-SDAll 1CV-QBase-SDAll1CV-QHigh-SDAll 1CV-QLow-SDAll 2C-QBase 2C-QHigh 2C-QLowEPRI(Mean) MidCont EPRI(1) MidCont EPRI(2) MidCont EPRI(3) MidCont EPRI(4) MidContEPRI(5) MidCont EPRI(6) MidCont EPRI(7) MidCont EPRI(8) MidCont EPRI(9) MidContEPRI(Mean) Gulf EPRI(1) Gulf EPRI(2) Gulf EPRI(3) Gulf EPRI(4) GulfEPRI(5) Gulf EPRI(6) Gulf EPRI(7) Gulf EPRI(8) Gulf EPRI(9) GulfIRIS Data

Figure 2c. Comparison of Caribbean GMPE (red lines), EPRI Mid-continent (blue lines)

and Gulf Coast region (green lines) GMPE and empirical IRIS processed data for a spectral frequency of 2.5 Hz.

Gulf of Mexico (Mw5.9), 09/10/2006:

This event occurred in the Gulf of Mexico and its epicenter is shown in Figure 3a along with the Turkey Point Units 6 & 7 site location, and the location of the IRIS stations that recorded this earthquake. Based on the observed station distribution, the IRIS station DWPF located in central Florida has the most applicable source to site path for this comparison. The distribution of stations in the central United States have a less applicable travel path azimuth and may show different attenuation properties based on these different tectonic travel paths. The attenuation curves and empirical data are shown in Figures 3b and 3c for the 1 and 2.5 Hz spectral frequencies. The data point from the DWPF station is highlighted as a solid blue circle. In general the empirical observations fall within the range of the Caribbean attenuation curves with the single exception of station LRAL (i.e., at a distance of about 750 km) for 2.5 Hz in which the empirical data exceeds the highest Caribbean attenuation curve. As noted earlier in this response, the empirical observation has not been adjusted to a common site condition of Vs= 2.83 km/sec and it could be expected that this empirical data point could possibly be reduced if a site amplification adjustment were to be applied. It can also be concluded from the comparison plots in Figures 3b and 3c, that the distribution of the current Caribbean attenuation curves

DRAFT


adequately captures the range of the empirical data. In addition, the observation from the DWPF station is in the lower range of the Caribbean attenuation curves.

Figure 3a. Map showing the earthquake location for the 09/10/2006 Gulf of Mexico

earthquake (Mw5.9), the IRIS station locations used in the analysis and Turkey Point Units 6 & 7 site location.

DRAFT


0.000001

0.00001

0.0001

0.001

0.01

0.1

1

100 1000 10000

Sp

ec

tra

l Ac

ce

lera

tio

n (g

)


Attenuation Models: 1Hz, 09/10/2006, M5.9Mean 1CC-QBase-SDAll 1CC-QHigh-SDAll 1CC-QLow-SDAll 1CV-QBase-SDAll1CV-QHigh-SDAll 1CV-QLow-SDAll 2C-QBase 2C-QHigh 2C-QLowEPRI(Mean) MidCont EPRI(1) MidCont EPRI(2) MidCont EPRI(3) MidCont EPRI(4) MidContEPRI(5) MidCont EPRI(6) MidCont EPRI(7) MidCont EPRI(8) MidCont EPRI(9) MidContEPRI(Mean) Gulf EPRI(1) Gulf EPRI(2) Gulf EPRI(3) Gulf EPRI(4) GulfEPRI(5) Gulf EPRI(6) Gulf EPRI(7) Gulf EPRI(8) Gulf EPRI(9) GulfIRIS Data

Figure 3b. Comparison of Caribbean GMPE (red lines), EPRI Mid-continent (blue lines)

and Gulf Coast region (green lines) GMPE and empirical IRIS processed data for a spectral frequency of 1 Hz. DRAFT


0.000001

0.00001

0.0001

0.001

0.01

0.1

1

100 1000 10000

Sp

ec

tra

l Ac

ce

lera

tio

n (g

)


Attenuation Models: 2.5Hz, 09/10/2006, M5.9







IRIS Data



Cuba Region (Mw6.2), 02/04/2007:

This event occurred south of the island of Cuba and its epicenter is shown in Figure 4a along with the Turkey Point Units 6 & 7 site location, and the location of the IRIS stations that recorded this earthquake. Only three IRIS stations were analyzed from this earthquake and their station locations are shown in Figure 4a. Based on the observed station distribution, the IRIS station DWPF located in central Florida has the most applicable source to site path for this comparison. The attenuation curves and empirical data are shown in Figures 4b and 4c for the 1 and 2.5 Hz spectral frequencies. The data point from the DWPF station is highlighted as a solid blue circle. Overall the empirical data falls below the median Caribbean attenuation curve (heavy red line) and has values which are in the lower distribution range of Caribbean attenuation curves. In addition, the observation from the DWPF station is similar to the lowest Caribbean attenuation curve or lower.

DRAFT


Figure 4a. Map showing the earthquake location for the 02/04/2007 Cuba Region earthquake

(Mw6.2), the IRIS station locations used in the analysis and Turkey Point Units 6 & 7 site location. DRAFT


0.000001

0.00001

0.0001

0.001

0.01

0.1

1

100 1000 10000

Sp

ec

tra

l Ac

ce

lera

tio

n (g

)


Caribbean Attenuation Models: 1Hz, 02/04/2007, M6.2

Mean 1CC-QBase-SDAll 1CC-QHigh-SDAll 1CC-QLow-SDAll 1CV-QBase-SDAll1CV-QHigh-SDAll 1CV-QLow-SDAll 2C-QBase 2C-QHigh 2C-QLowEPRI(Mean) MidCont EPRI(1) MidCont EPRI(2) MidCont EPRI(3) MidCont EPRI(4) MidContEPRI(5) MidCont EPRI(6) MidCont EPRI(7) MidCont EPRI(8) MidCont EPRI(9) MidContEPRI(Mean) Gulf EPRI(1) Gulf EPRI(2) Gulf EPRI(3) Gulf EPRI(4) GulfEPRI(5) Gulf EPRI(6) Gulf EPRI(7) Gulf EPRI(8) Gulf EPRI(9) GulfIRIS Data

Figure 4b. Comparison of Caribbean GMPE (red lines), EPRI Mid-continent (blue lines) and

Gulf Coast region (green lines) GMPE and empirical IRIS processed data for a spectral frequency of 1 Hz. DRAFT


0.000001

0.00001

0.0001

0.001

0.01

0.1

1

100 1000 10000

Sp

ec

tra

l Ac

ce

lera

tio

n (g

)


Caribbean Attenuation Models: 2.5Hz, 02/04/2007, M6.2


Figure 4c. Comparison of Caribbean GMPE (red lines), EPRI Mid-continent (blue lines) and

Gulf Coast region (green lines) GMPE and empirical IRIS processed data for a spectral frequency of 2.5 Hz.

North of Honduras (Mw7.3), 05/28/2009:

This event occurred north of Honduras in Central America and is the largest earthquake in the suite of five events analyzed in this sensitivity study. The location of its epicenter is shown in Figure 5a along with the Turkey Point Units 6 & 7 site location, and the location of the three IRIS stations that were analyzed. Based on the azimuths from the earthquake to the three IRIS stations, none of the associated seismic ray travel paths are ideal for this comparison study. The attenuation curves and empirical data are shown in Figures 5b and 5c for the 1 and 2.5 Hz spectral frequencies. The comparisons provided in the figures indicate that the empirical data from this earthquake are lower than any of the Caribbean attenuation curves.

DRAFT


Figure 5a. Map showing the earthquake location for the 05/28/2009 North of Honduras

earthquake (Mw7.3), the IRIS station locations used in the analysis and Turkey Point Units 6 & 7 site location. DRAFT


0.000001

0.00001

0.0001

0.001

0.01

0.1

1

100 1000 10000

Sp

ec

tra

l Ac

ce

lera

tio

n (g

)


Caribbean Attenuation Models: 1Hz, 05/28/2009, M7.3Mean 1CC-QBase-SDAll 1CC-QHigh-SDAll 1CC-QLow-SDAll 1CV-QBase-SDAll 1CV-QHigh-SDAll1CV-QLow-SDAll 2C-QBase 2C-QHigh 2C-QLow EPRI(Mean) MidCont EPRI(1) MidContEPRI(2) MidCont EPRI(3) MidCont EPRI(4) MidCont EPRI(5) MidCont EPRI(6) MidCont EPRI(7) MidContEPRI(8) MidCont EPRI(9) MidCont EPRI(Mean) Gulf EPRI(1) Gulf EPRI(2) Gulf EPRI(3) GulfEPRI(4) Gulf EPRI(5) Gulf EPRI(6) Gulf EPRI(7) Gulf EPRI(8) Gulf EPRI(9) GulfIRIS Data

Figure 5b. Comparison of Caribbean GMPE (red lines), EPRI Mid-continent (blue lines)

and Gulf Coast region (green lines) GMPE and empirical IRIS processed data for a spectral frequency of 1 Hz.

DRAFT


0.000001

0.00001

0.0001

0.001

0.01

0.1

1

100 1000 10000

Sp

ec

tra

l Ac

ce

lera

tio

n (g

)



Mean 1CC-QBase-SDAll 1CC-QHigh-SDAll 1CC-QLow-SDAll 1CV-QBase-SDAll 1CV-QHigh-SDAll1CV-QLow-SDAll 2C-QBase 2C-QHigh 2C-QLow EPRI(Mean) MidCont EPRI(1) MidContEPRI(2) MidCont EPRI(3) MidCont EPRI(4) MidCont EPRI(5) MidCont EPRI(6) MidCont EPRI(7) MidContEPRI(8) MidCont EPRI(9) MidCont EPRI(Mean) Gulf EPRI(1) Gulf EPRI(2) Gulf EPRI(3) GulfEPRI(4) Gulf EPRI(5) Gulf EPRI(6) Gulf EPRI(7) Gulf EPRI(8) Gulf EPRI(9) GulfIRIS Data


and Gulf Coast region (green lines) GMPE and empirical IRIS processed data for a spectral frequency of 2.5 Hz

Haiti (Mw7.0), 01/12/2010:

The Haiti earthquake is the most recent event in this suite of five earthquakes analyzed. The epicentral location and the suite of IRIS stations analyzed are shown in Figure 6a. Based on the observed station distribution, the IRIS station OTAV located in between the south part of Florida and the island of Cuba has the most applicable source to site path for this comparison. Note that the data from the DWPF was not available for this earthquake. The attenuation curves and empirical data are shown in Figures 6b and 6c for the 1 and 2.5 Hz spectral frequencies. The data point from the OTAV station is highlighted as a solid blue circle. Overall the empirical data falls in the lower range or lower than the Caribbean attenuation curves. In addition, the observation from the OTAV station is significantly lower than the entire range of the Caribbean attenuation curves.

DRAFT


Figure 6a. Map showing the earthquake location for the 01/12/2010 Haiti earthquake (Mw7.0),

the IRIS station locations used in the analysis and Turkey Point Units 6 & 7 site location. DRAFT


0.00001

0.0001

0.001

0.01

0.1

1

100 1000 10000

Sp

ec

tra

l Ac

ce

lera

tio

n (g

)


01/12/2010 Earthquake: 1Hz, 01/12/2010, M7

Mean

1CC-QBase-SDAll

1CC-QHigh-SDAll

1CC-QLow-SDAll

1CV-QBase-SDAll

1CV-QHigh-SDAll

1CV-QLow-SDAll

2C-QBase

2C-QHigh

2C-QLow

EPRI(Mean) MidCont

EPRI(1) MidCont

EPRI(2) MidCont

EPRI(3) MidCont

EPRI(4) MidCont

EPRI(5) MidCont

EPRI(6) MidCont

EPRI(7) MidCont

EPRI(8) MidCont

EPRI(9) MidCont

EPRI(Mean) Gulf

EPRI(1) Gulf

EPRI(2) Gulf

EPRI(3) Gulf

EPRI(4) Gulf

EPRI(5) Gulf

EPRI(6) Gulf

EPRI(7) Gulf

EPRI(8) Gulf

EPRI(9) Gulf

IRIS Data

Figure 6b. Comparison of Caribbean GMPE (red lines), EPRI Mid-continent (blue lines) and

Gulf Coast region (green lines) GMPE and empirical IRIS processed data for a spectral frequency of 1 Hz. DRAFT


0.00001

0.0001

0.001

0.01

0.1

1

100 1000 10000

Sp

ec

tra

l Ac

ce

lera

tio

n (g

)


01/12/2010 Earthquake: 2.5Hz, 01/12/2010, M7

Mean

1CC-QBase-SDAll

1CC-QHigh-SDAll

1CC-QLow-SDAll

1CV-QBase-SDAll

1CV-QHigh-SDAll

1CV-QLow-SDAll

2C-QBase

2C-QHigh

2C-QLow

EPRI(Mean) MidCont

EPRI(1) MidCont

EPRI(2) MidCont

EPRI(3) MidCont

EPRI(4) MidCont

EPRI(5) MidCont

EPRI(6) MidCont

EPRI(7) MidCont

EPRI(8) MidCont

EPRI(9) MidCont

EPRI(Mean) Gulf

EPRI(1) Gulf

EPRI(2) Gulf

EPRI(3) Gulf

EPRI(4) Gulf

EPRI(5) Gulf

EPRI(6) Gulf

EPRI(7) Gulf

EPRI(8) Gulf

EPRI(9) Gulf

IRIS Data



Based on this first supplemental sensitivity analysis, it is concluded that the suite of Caribbean GMPE models used in the current PSHA predicts larger ground motions on average than the observed empirical data from these five earthquakes for the spectral frequencies of 1 and 2.5 Hz. In addition, given the subset of data from just those stations that have a more appropriate source to seismic ray path, this conclusion can be extended to state that the use of the Caribbean GMPE models should provide conservative (i.e., higher) ground motion predictions compared to the observed empirical data from the region.

The second supplemental sensitivity determined the sensitivity of the ground motion response spectrum (GMRS) at the Turkey Point Units 6 & 7 site to the GMPEs used for the Caribbean seismic sources, which include the Cuba areal source plus nine fault sources. Five sets of seismic hazard calculations for five different GMPE suites were used to develop the corresponding GMRS. The GMPEs used for Caribbean seismic sources are: those of the base case which is (1) the Caribbean GMPE models developed for Turkey Point Units 6 & 7 (FSAR Subsection 2.5.2.4.5.2), (2) EPRI Mid-continent region (FSAR References 242 and 203) equations, (3) EPRI Mid-continent region “mod1” (FSAR References 242 and 203) equations, (4) EPRI Gulf region (FSAR References 242 and 203)

DRAFT


equations, and (5) EPRI Gulf region “mod 1” (FSAR References 242 and 203) equations. The modification of the EPRI GMPEs for both the Mid-continent and Gulf Coast cases was to exclude the GMPE which predicts significantly higher ground motions for large distance, such as those from the contributing Cuba and Caribbean sources. All seismic hazard calculations were made for the hard rock site conditions and include the other non-Caribbean sources in the PSHA. Thus the observed differences are based solely on the use of different GMPE models for the Caribbean seismic sources. For each of the five cases, ground motion values were estimated from the mean hazard curves for the seven standard spectral frequencies. Site specific horizontal GMRS spectra are plotted in Figure 7 and are developed using the site amplification factors from the FSAR. In general, the GMRS results using the EPRI Gulf Coast GMPEs are equal to or lower than the results using the Caribbean GMPE (i.e., indicated as Base Case in Figure 7). For the EPRI Mid-continent GMPE, the opposite result is concluded in which the GMRS values exceed the GMRS values using the Caribbean GMPE models. However, based on the previous additional sensitivity, the EPRI Mid-continent models predict higher ground motions than the empirical ground motions and as such may not be applicable for the modeling of ground motion attenuation for seismic sources in this Caribbean region and the Turkey Point Units 6 & 7 site location in southern Florida. In addition the tectonic structure and potential attenuation of ground motions for Caribbean sources might be more consistent with the EPRI Gulf Coast GMPE models based on the southern region of Florida being located in the Gulf Coast tectonic region of the eastern United States and based on the results shown in Figure 7, the use of the EPRI Gulf Coast GMPE for the PSHA gives similar or lower GMRS values.

DRAFT


0.01

0.1

1

0.1 1 10 100

Horizontal G

MRS (g)

Frequency (Hz)

Turkey PointGMRS Sensitivity

TP‐MidC

TP‐MidC‐mod1

TP‐Base Case

TP‐Gulf

TP‐Gulf‐mod1

Figure 7. Horizontal GMRS for Caribbean sources for five GMPE models for the Turkey Point Units 6 & 7 site.

These two supplemental sensitivity analyses indicate that the suite of GMPEs used for the Turkey Point Units 6 & 7 FSAR to characterize contributions to hazard at the site from Caribbean earthquakes, and the resulting GMRS at the site from these earthquakes, are conservative.

SSHAC Study Summary

The GMPE development reflects the uniquely challenging situation that exists with regard to the seismic characterization of Cuba and the adjacent Northern Caribbean region in that, for this region, empirical ground motion data is limited or unavailable, there are currently no calibrated GMPEs predicting ground motions in southern Florida arising from earthquakes in Cuba or the Northern Caribbean, and there are no experts identified that can be characterized as proponents as defined in SSHAC (1997).

SSHAC (1997) characterizes a Level 2 study as one in which the Technical Integrator (TI) reviews the literature and then interacts with proponents and resource experts to identify issues and interpretations and, on the basis of these interactions, “estimates community distribution,” i.e., develops “a representation…of the diversity of interpretations and their uncertainties” (see Table 3-1 and Sections 3.1.3.5 and 3.2.1 of SSHAC [1997]). SSHAC

DRAFT


(1997) also acknowledges (in Section 3.1.3.3) that the “choice of the level of [study] is often driven by…the amount of resources available for the study.”

The process described in the Initial GMPE Development Summary above is generally consistent with the conduct of a Level 2 study as described in SSHAC (1997) where it states (in section 3.2.1): “…the TI would communicate with the authors of published studies and other local experts who have expertise in the region or in regional ground motions…to hear and understand the technical positions taken by various proponents of particular hypotheses….In effect, the TI is…attempting to provide an overall assessment that would represent the informed scientific community’s view of the subject, if the community were to make such an assessment.”

Section 3.2.1 of SSHAC (1997) goes on to state that a “key aspect of the TI approach is the use of peer review to assure that the process followed was adequate and to ensure that the results provide a reasonable representation of the diversity of views of the technical community.” As again discussed above in the Initial GMPE Development Summary section of this response, the members of the Turkey Point Units 6 & 7 Technical Advisory Group (TAG) acted as participatory peer reviewers for the study, with the proposed GMPE development methodology and resultant outcomes reviewed with the TAG in three meetings and an additional TAG review after the third meeting. The TAG provided both technical and process reviews, and the methodology applied and the results obtained in developing initial GMPEs for the Caribbean seismic sources were ultimately endorsed by the TAG members, i.e., the TAG members provided, “an affirmation that the particular project [met their] standards of quality, thoroughness, and validity,” as is expected per Section 3.1.3.5 of the SSHAC Report

The challenge which the Bechtel TI team still faced in fully conforming to the expectations for a SSHAC Level 2 study as set forth in SSHAC (1997) was that the team found “various proponents” do not exist and that there is no available “informed scientific community” having direct expertise in appropriate GMPEs in the region at issue. While Professor Motazedian was consulted as a resource expert, he could not be considered a proponent as characterized in SSHAC (1997), and Professor Motazedian confirmed there are no other researchers who could be considered proponents for the purposes of a SSHAC study as there is no one, including Professor Motazedian, who has published a relationship for the attenuation of motion between Caribbean earthquake sources and the southeastern US.

Because no proponents could be identified, an additional effort was made in response to this RAI to engage ground motion experts who have otherwise developed and evaluated ground motion attenuation relationship more generally. A total of six such ground motion experts agreed to participate in the SSHAC study. These six experts were:

Dr. Norman Abrahamson (Consultant) Dr. Yousef Bozorgnia (University of California, Berkeley, Pacific Earthquake

Engineering Research Center) Dr. Kenneth Campbell (EQECAT) Dr. Shahram Pezeshk (University of Memphis) Dr. Paul Somerville (URS Corporation) Dr. Robert Youngs (AMEC-Geomatrix)

DRAFT


It should be noted that Dr. Robert Youngs also served as part of the TAG committee for Turkey Point Units 6 & 7.

The TI team of Drs. Gregor and Tavakoli developed a background technical summary note and questionnaire for distribution to the six ground motion experts. This technical note and questionnaire are contained in Enclosure 1.

The summary note and questionnaire asked seven specific questions of the experts dealing with the Caribbean GMPE models used in the PSHA. The actual responses from the experts are contained in Enclosure 1. Presented here are the excerpted and/or summarized experts’ responses for each of the seven questions as interpreted by the TI team members. Note that sentences and responses given in quotes below are taken directly from the expert’s response and are not modified or summarized.

Question 1: Are the newly developed Caribbean GMPEs acceptable for modeling ground motions in the Caribbean region for the project site located in Southern Florida for the PSHA study? Dr. N. Abrahamson Based on the comparison with the IRIS empirical data, any concerns about the

model or uncertainty parameters is secondary. Based on the comparison with the empirical data, the model appears to be predicting higher ground motions.

Dr. Y. Bozorgnia “Based on the review of the materials submitted by Bechtel, the newly developed GMPEs are acceptable, and the process of developing the GMPEs is rational and logical.”

Dr. K. Campbell “Yes, it is my opinion that the newly developed Caribbean GMPEs are generally acceptable, albeit possibly conservative at very large distances, for modeling ground motions in the Caribbean region for a site located in Southern Florida for the PSHA study…”

Dr. S. Pezeshk “My general comment is that the newly developed Caribbean GMPEs are acceptable and possibly conservative for modeling ground motions in the Caribbean region….”

Dr. P. Somerville “From a regulatory viewpoint I expect that they are acceptable because they are conservative when compared to the data….”

Dr. R. Youngs “The models developed from simulation appear to be reasonable. The comparisons with recorded motions indicate that they may slightly over estimate ground motions, which is OK.”

Question 2: Are the selected input parameters acceptable for their use? Dr. N. Abrahamson This question is not as significant based on the comparison of the empirical data to

the GMPE models. Dr. Y. Bozorgnia “Yes, especially considering the work of Motazedian and Atkinson (2005). Also,

using Kappa=0.006 sec for the reference rock is consistent with the latest preliminary findings of NGA-East project.”

Dr. K. Campbell “Yes, I am not aware of any other parameters that might be more acceptable for use in deriving a GMPE for the Caribbean region…”

Dr. S. Pezeshk The comparison of the empirical data with the Gulf of Mexico earthquake indicates that the parameters are acceptable for the capturing of the recorded ground motion data.

Dr. P. Somerville “Not relevant, see Comment on Q5 below.” Dr. R. Youngs A comparison should be made between the Chen and Atkinson (2002) amplification

functions and those used to develop the CEUS models to confirm their consistency.

DRAFT


Question 3: Is the recommended sigma value (i.e., 0.645 natural log units) acceptable? Dr. N. Abrahamson This question is not as significant based on the comparison of the empirical data to

the GMPE models. Dr. Y. Bozorgnia “… On average sense, this number is a reasonable number; however, the standard

deviation of GMPEs can be a function of frequency….” Dr. K. Campbell “Yes, I believe that the recommended sigma value of 0.645 natural log units (0.28

common log units) is acceptable as long as sufficient epistemic uncertainty is captured in the model.”

Dr. S. Pezeshk “…In response to question 3, the aleatory uncertainty with an average value of 0.645 in natural log units is reasonable to capture the uncertainties in input seismological parameters discussed above for all frequencies in the Caribbean region.”

Dr. P. Somerville “Yes” Dr. R. Youngs “There does not appear to be a strong basis for the 0.645 and it is intended for use in

Puerto Rico. A more preferable model with a stronger basis may be the EPRI (2006) model for CENA or some simplification of that model, perhaps using the final NGA (2008) results.”

Question 4: Is the recommended weighting based on the EPRI (2004) class weights acceptable?

Dr. N. Abrahamson This question is not as significant based on the comparison of the empirical data to the GMPE models.

Dr. Y. Bozorgnia “The explanation on the assigned weights in Table 3 is reasonable.”

Dr. K. Campbell “Yes, the weighting scheme based on class weights seems reasonable …”

Dr. S. Pezeshk “The recommended weighting based on the EPRI (2004) class weights is reasonable and acceptable.”

Dr. P. Somerville “Not relevant, see Comment on Q5 below.”

Dr. R. Youngs “Yes, roughly equal weights for 1 versus 2 corners is reasonable.”

Question 5: Are there any other GMPE models which should be considered for the PSHA contribution from Caribbean sources? Dr. N. Abrahamson No response. Dr. Y. Bozorgnia “Not that I know of.” Dr. K. Campbell Subduction GMPE models have a noted different rate of attenuation for forearc as

opposed to backarc site locations with the backarc stations having a greater rate of attenuation. It is recommended to make a comparison of subduction GMPE models with the empirical data to show that the subduction GMPE models would predict lower ground motions. If, however, a faster rate of attenuation is not shown, then subduction GMPE should be considered for the PSHA.

Dr. S. Pezeshk Recommendation of comparison plots with the suite of NGA (2008) empirical GMPEs and the MA (2005) model. In addition, comparisons between intermediate magnitude earthquakes (Mw 5-6) occurring in northern region of the island of Cuba could be added to the data comparison to better understand the variability of ground motions between Cuba and southern Florida.

Dr. P. Somerville “None exist now; it is preferable to develop an empirically calibrated model. See comment on Q5 below.”

Dr. R. Youngs “The USGS uses WNA GMPEs for modeling ground motions in Puerto Rico. It may be useful to compare the predictions from the NGA models adjusted to CEUS hard rock conditions using published factors (e.g., Atkinson and Boore, BSSA 2011) with the developed models at least in distances of about 200 km to evaluate whether they predict higher or lower ground motions.”

DRAFT


Question 6: Is either the Mid-Continent or Gulf Coast region GMPE from EPRI (2004) more appropriate for use in the evaluation of the contribution to PSHA at a site in southern Florida from Caribbean seismic sources? Dr. N. Abrahamson No response. Dr. Y. Bozorgnia “Based on the sensitivity analysis of five regional events reported in the Bechtel

report, it seems that EPRI Gulf Coast GMPEs are more appropriate…” Dr. K. Campbell Recommendation that the Gulf Coast EPRI (2004) be used to capture the epistemic

uncertainty in ground motion models based on the empirical data comparison with the Gulf of Mexico earthquake. In addition, it is not recommended to use the Mid-Continent EPRI (2004) model for Caribbean sources.

Dr. S. Pezeshk Based on the importance of the Q model in controlling the ground motions especially at large distances, the Caribbean GMPEs are between the Mid-Continent and Gulf Coast models. The comparison with the empirical data indicates that the Caribbean GMPE models are closer to the Gulf Coast GMPEs than the Mid-Continent GMPEs from the EPRI (2004) study.

Dr. P. Somerville “The Gulf Coast region GMPE is more appropriate; see Comment on Q5 below.” Dr. R. Youngs “I do not see any reason to consider the EPRI (2004) models to be more appropriate

than the Caribbean model developed for the Turkey Point study. The comparisons with data shown in the figures confirm that.”

DRAFT


Question 7: Do you have any other comments? Dr. N. Abrahamson None. Dr. Y. Bozorgnia For the sensitivity analysis with the IRIS data, the correction/adjustment of the

empirical data for a consistent Vs=2.83km/sec should be performed. In addition, residual plots should be computed along with the event terms and their uncertainty.

Dr. K. Campbell The applicability of the regional parameters from the MA (2005) study for distances beyond 500 km must be adequately captured in the epistemic uncertainty. Presumably this is captured in the epistemic uncertainty in the Q model. In addition, any potential updates to the MA (2005) and or more recent datasets and studies should be investigated. The terms GMPE and attenuation models should be differentiated with the term attenuation being used to describe actual attenuation properties such as Q and geometrical spreading. The assumption that based on a simplified site response analysis that the adjustments for the station-specific site conditions to the more stable CEUS hard rock site conditions would result in the reduction of the observed empirical ground motions might be true for relatively low frequencies, however, it may not be true for relatively high frequencies where a very small kappa at the project site might result in higher ground motions. The deaggregation histograms shows the Caribbean sources are important contributors to the seismic hazard at low structural and exceedance frequencies. What do the results show for the higher structural frequencies?

Dr. S. Pezeshk The shear wave velocity at the source is noted as 3.6 km/sec for the Caribbean study; however, the amplification factors taken from Chen and Atkinson (2002) were developed for a shear wave velocity of 3.8 km/sec. Thus the amplification factors used in the analysis should be adjusted for this difference in the shear wave velocity. An explanation on how the values listed in Table 2 is needed. If there are records available from intermediate magnitude events recorded between Cuba and southern Florida, they should be added to the empirical data comparison analysis. The focal mechanism and depths for the events considered should be included.

Dr. P. Somerville None. Dr. R. Youngs (a) “The presentation of the comparisons on Figures 1-7 makes it very difficult to

distinguish among the various relationships. It would perhaps be more useful to compare the median models of the EPRI (2004) clusters with the three base case Caribbean relationships.

(b) It is not stated why the EPRI (2004) Cluster 4 model is not included in the comparisons. It is appropriate for large magnitude earthquakes, which are the type of earthquakes from the Caribbean sources that contribute to the hazard at Turkey Point.

(c) Some description of the site conditions at the Florida DWPF stations should be included. This station has the appropriate path for Caribbean earthquakes recorded in Florida, but what about site conditions or expected amplification from hard rock?”

In reviewing and summarizing the responses from the experts it is concluded that the use of the Caribbean GMPE models in the PSHA is acceptable. As noted by the six experts, based on the comparison between the Caribbean GMPE model and the empirical IRIS data, these GMPE models may be conservative in their estimation of ground motions in the

DRAFT


region. In addition the assignment of an aleatory uncertainty value of 0.645 in natural log units was acceptable based on the consensus of the six experts.

As part of the expert’s responses to the questionnaire, additional comments and requests were presented by each expert (e.g., see the responses to Question 7). Specific elements of ground motion sensitivity analyses, such as the lack of any site condition information for the IRIS stations, any necessary adjustments for potentially different site amplification factors, a frequency dependent and/or a different sigma model, comparisons with other GMPE models, or the development of a new empirically-calibrated GMPE model, were noted by the experts. In consideration of the suite of responses from the experts, the TI team concluded that, although the various comments and requests were of technical interest -- in some cases, requiring an academic approach involving detailed investigations and extensive data acquisition and study to satisfy the request – the consensus of the experts was, nevertheless, that the Caribbean GMPE models were acceptable for use in the PSHA for Turkey Point Units 6 & 7 site, even as the IRIS empirical data suggest some conservatism.

When asked about the applicability of the EPRI Mid-continent or Gulf Coast models, the consensus of the six experts was that the suite of Gulf Coast GMPE models are closer to the empirical data than the Mid-continent suite. However, this did not indicate that the EPRI Gulf Coast models are preferable over the Caribbean GMPE models. Based on this conclusion with regards to the two EPRI ground motion models and the current sensitivity studies which showed that the use of the EPRI Gulf Coast GMPE in the PSHA resulted in lower GMRS ground motions, it is concluded that the use of the Caribbean GMPE models is conservative in the estimation of the GMRS ground motions for the Turkey Point Units 6 & 7 site.

Based on the polling and summary of the responses from the six experts the ultimate use of the specific Caribbean GMPEs in the PSHA for the Turkey Point Units 6 & 7 site location for seismic sources in the Cuba and Caribbean region falls within the range of the informed technical community and actually may produce slightly larger ground motions than would be estimated from the center, body, and range of the informed technical community. The experts’ opinions support the development of the final Caribbean GMPEs. There were no conflicting opinions among the experts regarding the suitability of the final Caribbean GMPEs for use in the PSHA analysis for the Turkey Point Units 6 & 7 FSAR.

DRAFT


Part c: Please provide copies of the following supporting calculations: Report #: 25409-000-K0C-0000-00009, Report#: 25409-000-K0C-0000-00024, Report #: 25409-000-K0C-0000-00034 to enable the staff to evaluate the technical details of the final GMPEs.

The requested calculations are available for inspection in the Reading Room.

As a result of additional studies conducted for this RAI, three new calculations were performed to document ground motion attenuation curve plots from the EPRI (2004) and Caribbean GMPEs, provide a comparison of IRIS empirical data with GMPE models for the Caribbean region, and to evaluate the sensitivity of GMRS at the Turkey Point site to five sets of GMPEs used for Caribbean sources. These calculations are also available for inspection in the Reading Room.

Part d: In order for the staff to be able to compare the new Caribbean GMPEs with the 2004/2006 EPRI mid-continent GMPEs, please provide plots showing both ground motion models for earthquake magnitudes of 6.0, 7.0, and 8.0 in the distance range of 200 km to 1000 km at all seven frequencies defined in the EPRI 2004 and 2006 GMPEs.

Mean ground motion attenuation curve plots are provided for the EPRI Mid-continent (FSAR References 242 and 203) and Caribbean ground motion prediction equations (GMPEs) in Figures 8 – 14 for the seven defined frequencies. Note that EPRI (FSAR Reference 203) is only a recommendation for the associated aleatory uncertainty and therefore does not impact the comparison plots of the weighted mean ground motion attenuation curve. The plotted mean attenuation curve is the weighted mean of the individual median attenuation curves as defined in EPRI (FSAR Reference 242) and for the Caribbean GMPE models. These attenuation plots are provided for three specific moment magnitude (Mw) values: 6, 7, and 8 for distances between 200km and 1,000km. Note that the EPRI (FSAR Reference 242) ground motion attenuation model is defined as a function of epicentral distance, whereas the Caribbean ground motion attenuation model is defined as a function of hypocentral distance. For these comparison plots, the attenuation curves are plotted as a function of epicentral distance and for the Caribbean GMPE curves an assumed hypocentral depth of 8 km was used.

DRAFT


0.00001

0.0001

0.001

0.01

0.1

1

100 1000

Sp

ec

tra

l Ac

ce

lera

tio

n (g

)

Epicentral Distance (km)

Attenuation Models: PGA

EPRI MidC (2004) M8

Caribbean M8

EPRI MidC (2004) M7

Caribbean M7

EPRI MidC (2004) M6

Caribbean M6

Figure 8. Comparison of EPRI Mid-continent (FSAR Reference 242) mean (dashed lines) and Caribbean mean (solid lines) attenuation curves for magnitudes 6, 7, and 8 for PGA. DRAFT


0.00001

0.0001

0.001

0.01

0.1

1

100 1000

Sp

ec

tra

l Ac

ce

lera

tio

n (g

)


Attenuation Models: 25Hz

EPRI MidC (2004) M8

Caribbean M8

EPRI MidC (2004) M7

Caribbean M7

EPRI MidC (2004) M6

Caribbean M6

Figure 9. Comparison of EPRI Mid-continent (FSAR Reference 242) mean (dashed

lines) and Caribbean mean (solid lines) attenuation curves for magnitudes 6, 7, and 8 for 25 Hz. DRAFT


0.00001

0.0001

0.001

0.01

0.1

1

100 1000

Sp

ec

tra

l Ac

ce

lera

tio

n (g

)



EPRI MidC (2004) M8

Caribbean M8

EPRI MidC (2004) M7

Caribbean M7

EPRI MidC (2004) M6

Caribbean M6




0.00001

0.0001

0.001

0.01

0.1

1

100 1000

Sp

ec

tra

l Ac

ce

lera

tio

n (g

)


Attenuation Models: 2.5Hz

EPRI MidC (2004) M8

Caribbean M8

EPRI MidC (2004) M7

Caribbean M7

EPRI MidC (2004) M6

Caribbean M6


lines) and Caribbean mean (solid lines) attenuation curves for magnitudes 6, 7, and 8 for 2.5 Hz. DRAFT


0.00001

0.0001

0.001

0.01

0.1

1

100 1000

Sp

ec

tra

l Ac

ce

lera

tio

n (g

)



EPRI MidC (2004) M8

Caribbean M8

EPRI MidC (2004) M7

Caribbean M7

EPRI MidC (2004) M6

Caribbean M6




0.00001

0.0001

0.001

0.01

0.1

1

100 1000

Sp

ec

tra

l Ac

ce

lera

tio

n (g

)



EPRI MidC (2004) M8

Caribbean M8

EPRI MidC (2004) M7

Caribbean M7

EPRI MidC (2004) M6

Caribbean M6

Figure 14. Comparison of EPRI Mid-continent (FSAR Reference 242) mean (dashed lines) and Caribbean mean (solid lines) attenuation curves for magnitudes 6, 7, and 8 for 0.5 Hz

Part e: Discuss evidence, if any, that seismic source scaling varies regionally and/or between source types in the Caribbean. For example, is there any evidence that stress parameter varies systematically between the northern Hispaniola sources, the Caribbean plate-boundary transform fault sources, and the Cuba sources? If so, what are the implications for the attenuation models and hazard calculated at TPNPP?

Based on the relatively moderate to high levels of observed seismicity in the Caribbean region, the investigation into any potential seismic source variation on either a local or regional scale could, in theory, be studied. However, based on the limited seismic networks in the region, only one such study was discovered in the literature. This study by Motazedian and Atkinson (FSAR Reference 287) studied a dataset of ground motion recordings in and around the island of Puerto Rico to develop a seismic source model including the stress parameter for the region. As noted by Motazedian and Atkinson (FSAR Reference 287), this ground motion dataset contains both crustal and subduction earthquakes but could not be subdivided based on the earthquake type. No studies were found during a literature review that specifically addresses any potential stress parameter

DRAFT


variation between earthquakes associated with the northern Hispaniola, Caribbean plate-boundary transform faults, and Cuba sources.

Motazedian and Atkinson (FSAR Reference 287) compared their attenuation model based on the Puerto Rico data and associated stress parameter with ground motion prediction equations (GMPEs) for the central and eastern United States (CEUS), California, and subduction ground motion models. Note that the CEUS and California models are based on a similar stochastic ground motion methodology that was used by Motazedian and Atkinson (FSAR Reference 287) whereas the subduction model is based on empirical data. Their conclusion was that the Puerto Rico based model is similar to the CEUS and California models and based on their attenuation curve plots, their model predicts significantly larger (i.e., larger by factors of 5 – 10 or greater) ground motions than the empirically-based subduction GMPE.

Tectonically, the larger Caribbean region consists of both crustal faults and subduction zones. As noted in Motazedian and Atkinson (FSAR Reference 287), their dataset most likely contained ground motions from both types of earthquakes. However, for the PSHA, the seismic sources around the island of Puerto Rico were not modeled based on their significantly large distance from the site. Garcia et al. (2012) present a regional tectonic characterization of the Caribbean region based on Flinn-Engdahl geographical regions. For the immediate region around Cuba and the southern Florida, the tectonics are classified as being active regions without subduction tectonics. Subduction zone tectonics are noted for the eastern side of the island of Hispaniola and further east to the island of Puerto Rico.

For the probabilistic seismic hazard analysis (PSHA) performed for Turkey Point Units 6 & 7 site, a total of 10 seismic sources were considered for the Caribbean region. Of these 10 sources the two Northern Hispaniola sources (i.e., western and eastern) are identified as being associated with the tectonic subduction zone in the region. Based on the comparison plots given in the Motazedian and Atkinson study (FSAR Reference 287) and the large distance from the Turkey Point Units 6 & 7 site to these individual subduction seismic sources, the use of an empirically derived GMPE for the subduction seismic source zones would lead to lower ground motions than the use of GMPEs based on the Motazedian and Atkinson (FSAR Reference 287) source and attenuation scaling. Thus the PSHA results are understood to have higher ground motion results than if a subduction GMPE was used for those specific Caribbean subduction seismic source zones.

This response is PLANT SPECIFIC.

References:

Ambraseys, N. N., Simpson, K. A., and Bommer, J. J. [1996] “Prediction of horizontal response spectra in Europe,” Earthquake Engineering and Structural Dynamics, 25, 371–400.

Atkinson, G., Boore, D., “Empirical Ground Motion Relation for Subduction Zone Earthquakes and their Application to Cascadia and other regions,” Bulletin of the Seismological Society of America, v. 93, no. 4, pp. 1703-1729, 2003.

DRAFT


Atkinson, G., Silva, W., “Stochastic Modeling of California Ground Motions,” Bulletin of the Seismological Society of America, v. 90, no. 2, pp. 255-274, 2000.

Dahle, A., Climent, A., Taylor, W., Bungum, H., Santos, P., Ciudad Real, M., Linholm, C., Strauch, W., and Segura, F. [1995] “New spectral strong motion attenuation models for Central America,” Proceedings of the Fifth International Conference on Seismic Zonation, (Nice, France), vol. II, pp. 1005–1012.

Garcia, D., D.J. Wald, and M.G. Hearne (2012). “A Global Earthquake Discrimination Scheme to Optimize Ground-Motion Prediction Equation Selection,” Bulletin of the Seismological Society of America, v. 102, no. 1, pp. 185-203.

Nuclear Regulatory Commission (2007). “Regulatory Guide 1.208 – A Performance-Based Approach to Define the Site-Specific Earthquake Ground Motion”, U.S. Nuclear Regulatory Commission, Washington, DC, March 2007.

Senior Seismic Hazard Analysis Committee [SSHAC] (1997). "Recommendations for Probabilistic Seismic Hazard Analysis: Guidance on Uncertainty and Use of Experts," NUREG/CR- 6372, Lawrence Livermore National Laboratory, UCRL-ID-122160.

ASSOCIATED COLA REVISIONS:

The following COLA changes are identified as a result of this response:

The text in FSAR Subsection 2.5.2.4.5.2 will be revised as follows in a future update of the FSAR:

2.5.2.4.5.2 New Attenuation Models for the Cuba and Caribbean Region

The incorporation of additional seismic sources developed for the Caribbean region in the PSHA requires applicable ground motion attenuation models. Although much of the Caribbean has experienced large, damaging earthquakes, there are very few recorded strong ground motion data from the region, and this has prevented the development of a regional empirical ground motion attenuation relationship, specifically one for attenuation of ground motion between sources in the northern Caribbean and the Turkey Point Units 6 & 7 site location in southern Florida. Moreover, the use of ground motion prediction equations (GMPE's) from other regions, such as the 2004 EPRI (Reference 242) GMPE's, cannot be uncritically adopted for PSHA analysis because of the observed differences in crustal properties between the CEUS and Caribbean.

No studies have been found modeling attenuation of strong ground motion between earthquakes in the Caribbean and sites in the CEUS so that no experts could be identified who could be characterized as a proponent, which NUREG/CR-6372 (SSHAC, 1997; Reference 318) defines as “an expert who advocates a particular hypothesis or technical position.” The absence of a proponent or proponents made it difficult to categorize the level of the study per the SSHAC guidelines. However, what has been done in this regard is otherwise fully consistent with industry practice in general and the SSHAC guidelines in particular.

Initial development of a suite of specific GMPEs was performed by, what in SSHAC nomenclature would be called, the Technical Integration team of two Bechtel

DRAFT


seismologists: Drs. Nick Gregor and Behrooz Tavakoli, both with significant experience in generating and evaluating GMPEs. Peer review was provided by a Technical Advisory Group (TAG) at several points throughout the GMPE development process. A resource expert, Dr. Dariush Motazedian of Carleton University, Ottawa, Canada, was also consulted during this time.

The GMPE development reflects the uniquely challenging situation that exists with regard to the seismic characterization of Cuba and the adjacent Northern Caribbean region in that, for this region, empirical ground motion data is limited or unavailable, there are currently no calibrated GMPEs predicting ground motions in southern Florida arising from earthquakes in Cuba or the Northern Caribbean, and there are no experts identified that can be characterized as proponents as defined in SSHAC (1997).

SSHAC (1997) characterizes a Level 2 study as one in which the Technical Integrator (TI) reviews the literature and then interacts with proponents and resource experts to identify issues and interpretations and, on the basis of these interactions, “estimates community distribution,” i.e., develops “a representation…of the diversity of interpretations and their uncertainties” (see Table 3-1 and Sections 3.1.3.5 and 3.2.1 of SSHAC [1997]). SSHAC (1997) also acknowledges (in Section 3.1.3.3) that the “choice of the level of [study] is often driven by…the amount of resources available for the study.”

Initially a literature review was performed with the goal of retrieving acceptable GMPEs for the Cuba and Caribbean region. However, this literature review only retrieved one GMPE developed recently by Motazedian and Atkinson (Reference 287). In addition to the interaction with the TAG members, correspondence was conducted with Professor Motazedian during the initial development of the Caribbean GMPEs. These initial technical discussions were for the possible application of the published Motazedian and Atkinson (Reference 287) GMPE for the PSHA study. However, based on the limitations of this GMPE (e.g., incomplete suite of necessary spectral frequencies, limited application for distances greater than 500 km, and site-specific ground conditions of soft rock), it was determined that this published GMPE was not directly acceptable for use in the PSHA.

The process described is generally consistent with the conduct of a Level 2 study as described in SSHAC (1997) where it states (in section 3.2.1): “…the TI would communicate with the authors of published studies and other local experts who have expertise in the region or in regional ground motions…to hear and understand the technical positions taken by various proponents of particular hypotheses….In effect, the TI is…attempting to provide an overall assessment that would represent the informed scientific community’s view of the subject, if the community were to make such an assessment.”

Motazedian and Atkinson (Reference 287) analyzed a dataset of approximately 300 earthquakes recorded by stations in Puerto Rico. This dataset spanned the Mw range of 3 to 5.5 and distances from about 20 km to 500 km. They acknowledge that their ground motion dataset consisted of recorded ground motions from both crustal and subduction zone earthquakes and that the separation of the earthquakes used in

DRAFT


their dataset into crustal and subduction events was not possible because of the limited station coverage for the region. Based on these data, Motazedian and Atkinson developed a set of regional anelastic attenuation and source parameters. Finally, Motazedian and Atkinson used these regional parameters within a stochastic simulation process to create an artificial dataset for larger earthquakes at near distances to fit a GMPE to these generated data.

To address this possible concern about the influence of using both subduction and crustal events to estimate the regional attenuation and source parameters, Motazedian and Atkinson (Reference 287) provide a comparison of their GMPE with representative GMPEs for the CEUS (Reference 210), California (Reference 345), and an empirically-based GMPE based on the global ground motion dataset for subduction zones (Reference 344). They conclude that, overall the Puerto Rico relations agree well with the stochastic relations for California and eastern North America and are quite different from those calculated by the global subduction relations. This comparison and noted results provide a technical justification for using the source and attenuation parameters from the Puerto Rico ground motion dataset (Reference 287) as the starting point for the development of applicable GMPEs for the Caribbean seismic sources.

The Motazedian and Atkinson (Reference 287) study focused on GMPEs most useful for the evaluation of PSHA results for Puerto Rico. For the purposes of an evaluation of potential contributions to PSHA at the Turkey Point Units 6 & 7 site from Caribbean earthquakes, GMPE results from somewhat larger and significantly more distant earthquakes are needed.

Beginning with the suite of regional anelastic attenuation and source parameters of Motazedian and Atkinson, and following the stochastic simulation methodology, region-specific attenuation models for the Cuba and Caribbean region were developed. Ground motions were estimated for seven spectral frequencies on hard rock for earthquakes with magnitudes between Mw 4.75 and 8.75 and for a distance range of 150 to 2000 kilometers.

To account for the expected uncertainty associated with the application of regional attenuation and source parameters estimated from earthquakes in and around Puerto Rico and their application to other regions within the Caribbean (e.g., Cuba), the stress parameter of the source and the anelastic attenuation models from Motazedian and Atkinson (Reference 287) were varied in the stochastic ground motion simulation analysis. The regional attenuation and source parameters from the Motazedian and Atkinson (Reference 287) study and the specific values used for the development of the suite of applicable attenuation models are listed in Table 2.5.2-231. The variation of the stress parameter was defined to be normally distributed with a standard deviation (sigma value) of 0.7 (in natural log units) given in the EPRI 1993 study (Reference 244). The variation of the regional anelastic attenuation constant term was based on an assumed sigma value of 0.4 (in natural log units) given in the 2003 Silva et al., study (Reference 342). In addition, three separate seismic source models were used in the analysis: single corner with constant stress parameter (1CC model), single corner with magnitude dependent stress parameter (1CV model), and double corner source (2C model) based on the analysis of CEUS ground motion data. These three seismic source models are part of the larger set of source models that are included in the 2004 EPRI (Reference 242) ground motion models.

DRAFT


A linear regression was performed on the simulated datasets to estimate the regression coefficients for the Caribbean regional GMPEs for use in the PSHA. A nonlinear GMPE functional form and regression, generally needed to successfully predict saturation of strong ground motion at small distances, were not required because of the large minimum distance of 150 kilometers separating Caribbean earthquakes from the Turkey Point Units 6 & 7 site. An aleatory sigma value of 0.645 (in natural log units) was selected following the Motazedian and Atkinson study (Reference 287) and was assigned to each Caribbean attenuation model for use in the PSHA for all frequencies.

To capture the epistemic uncertainty in ground motion models in the hazard analysis, a number of GMPEs from the different seismic source models were included along with model-dependent weights. The weights for these new attenuation models were assessed based on the family class weights used in the 2004 EPRI ground motion model study and the family class (that is, 1CC, 1CV, or 2C) of the seismic source model (Reference 242). Figure 2.5.2-255 shows the suite of Caribbean PGA attenuation curves for a magnitude Mw 7 earthquake over the applicable distance range of 150 - 2000 kilometers. Within a given seismic source model type (e.g., single corner constant stress parameter), the difference between the attenuation models based on the low, base, and high stress parameter values resulted in a constant scaling of the attenuation curves. This scalar variation was captured by combining the datasets from these stress parameter values for the regression analysis leading to a combined suite of nine ground motion attenuation models (i.e., heavy lines) that were adopted for use in the PSHA with the Caribbean seismic sources. As a check, the complete suite of attenuation curves is shown in Figure 2.5.2-255 and the resulting nine adopted attenuation curves span the general range of values from the complete suite of curves. Figures 2.5.2-256 and 2.5.2-257 plot the ground motion attenuation curves for the periods of 0.1 and 1.0 seconds, respectively.

At the suggestion of TAG members over the course of the three TAG meetings, a sensitivity analysis was performed to examine the effect on epistemic uncertainty of alternative attenuation relationships for use in the PSHA. The alternative relationships considered adopted a double corner (2C) seismic source model, such as might be expected to occur in a more active tectonic environment such as the western United States (WUS) rather than the double corner seismic source model of the less active tectonic environment of the CEUS, and a Gulf Coast region lower amplitude but higher (less rapidly attenuating) anelastic attenuation factor (Q) model rather than the Puerto Rico region-specific (higher amplitude but more rapidly attenuating) Q model from Motazedian and Atkinson (Reference 287) recognizing that much of the propagation path from the Caribbean sources to the Turkey Point Units 6 & 7 site is through the Gulf Coast crust. It was found that adoption of these alternatives (i.e., different suite of regional attenuation and seismic source parameter values) led to ground motion values that were equal to (at large distances based on the anelastic attenuation rates) or lower than (based on the different magnitude scaling from the WUS- based double corner model) the suite of original nine new ground motion attenuation models adopted for the Cuba and Caribbean region. A comparison of the weighted combination of the original nine GMPE models and the inclusion of these additional sensitivity models resulted in a slightly lower weighted mean ground motion attenuation curve over the magnitude and distance range needed for the PSHA. Therefore, their incorporation into the final PSHA results would have slightly lowered the already low hazards, and

DRAFT


thus, the use of the original nine GMPE models was accepted by the TAG members because the inclusion of these additional GMPE models would be expected to lead to lower, and less conservative, ground motion results.

After the suite of GMPEs was developed, two additional sensitivity analyses were performed to further validate the technical assessment of the stochastic Caribbean GMPE models developed for Turkey Point Units 6 & 7. These analyses sought additional empirical data with which to compare the GMPEs for the Turkey Point Units 6 & 7 FSAR and examined the effect of alternate suites of GMPEs on the PSHA results at the site. The results of these two additional sensitivity studies are presented here.

The first supplemental sensitivity analysis compared the suite of GMPEs used in the Turkey Point Units 6 & 7 PSHA with empirical regional ground motion data. To assist in the technical evaluation of the current Caribbean GMPE models developed for the Turkey Point Units 6 & 7 FSAR a comparison of empirical ground motions from five regional earthquakes occurring since 2004 in the Gulf of Mexico and northern Caribbean region were analyzed. These earthquakes were selected from among those available as being most representative of earthquakes whose tectonic settings and locations might be expected to be like those of future earthquakes contributing to the overall PSHA at the site under the source model adopted. Specifically, these were shallow crustal earthquakes in the northern Caribbean or Gulf of Mexico.

Regional broadband empirical data from selected Incorporated Research Institutions for Seismology (IRIS) stations were obtained, processed and compared to the suite of Caribbean GMPE models and the suite of EPRI (References 242 and 203) GMPE models for both the Mid-continent and Gulf Coast regional models. Based on these comparisons, a technical assessment can be made on the applicability of using the Caribbean GMPE for the Turkey Point Units 6 & 7 PSHA.

Note that the current PSHA results were developed using the suite of EPRI (References 242 and 203) Mid-continent GMPE models for other non-Caribbean sources. In this comparison all of the GMPE attenuation curves are for the assumed CEUS hard rock site conditions with Vs= 2.83 km/sec, whereas the empirical IRIS data are for the individual unknown site conditions of each station which is expected to be less than Vs= 2.83 km/sec. Based on simplified site response analysis it would be expected that an adjustment for the station specific site conditions to the more stable CEUS hard rock site conditions would require a reduction factor in the observed empirical ground motion values.

The current deaggregation of seismic hazard at the Turkey Point Units 6 & 7 site from all sources, Cuban, Caribbean, and southeast United States, is shown in Figures 2.5.2-226 and 2.5.2-228 for longer period motions (for which the relative contribution of the larger, more distant Cuban and Caribbean sources would be expected to have their greatest relative contribution) and for the 10-4 and 10-5 mean annual frequency of occurrence probabilities used to develop design ground motions under current NRC regulatory guidance. The relative contribution for the higher frequency cases of 5 and 10 Hz (Figures 2.5.2-227 and 2.5.2-229) from the more distant Caribbean sources is significantly reduced relative to the closer local

DRAFT


seismic sources. For this sensitivity comparison, the empirical ground motions for 1 and 2.5 Hz are presented for each of the five regional earthquakes and ground motion prediction equations.

For each event, acceleration response spectra for a spectral damping of 5 percent at each station were computed. The geometric mean of the two horizontal components was computed and amplitudes for the 1 and 2.5 Hz spectral frequencies were generated and compared to the suite of GMPE models (i.e., both of the EPRI, 2004 GMPE models and the Caribbean GMPE models).

Note that, based on the hypocentral location and geographical location relative to tectonic plate boundary, all five of these earthquakes are considered to be shallow crustal events and are not associated with any regional subduction zones. For each event, a standard time history processing methodology was applied with the final results being a dataset of the acceleration response spectra for a spectral damping of 5 percent at each station. The geometric mean of the two horizontal components was computed and comparison plots for the 1 and 2.5 Hz spectral frequencies were generated showing the empirical data and the suite of GMPE models (i.e., both of the EPRI, 2004 GMPE models and the Caribbean GMPE models). Both the individual attenuation curves and the weighted mean attenuation curve for a given set are shown in the comparison plots for each spectral frequency and earthquake.

These five events were as follows:

12/14/2004 – Caribbean Sea Region (Mw6.8, 19.05 N, -81.52 W, hypocenter depth 12.0 km)

This event occurred in the Caribbean Sea region and its epicenter is shown in Figure 2.5.2-258a along with the Turkey Point Units 6 & 7 site location, and the location of the IRIS stations that recorded this earthquake. Based on the observed station distribution, the IRIS station DWPF located in central Florida has the most applicable source to site path for this comparison. The attenuation curves and empirical data are shown in Figures 2.5.2-258b and 2.5.2-258c for the 1 and 2.5 Hz spectral frequencies. The data point from the DWPF station is highlighted as a solid blue circle. Overall the empirical data falls below the median Caribbean attenuation curve (heavy red line) and has values which are in the lower distribution range of Caribbean attenuation curves.

09/10/2006 – Gulf of Mexico (Mw5.9, 26.32 N, -86.84 W, hypocenter depth 29.6 km)

This event occurred in the Gulf of Mexico and its epicenter is shown in Figure 2.5.2-259a along with the Turkey Point Units 6 & 7 site location, and the location of the IRIS stations that recorded this earthquake. Based on the observed station distribution, the IRIS station DWPF located in central Florida has the most applicable source to site path for this comparison. The distribution of stations in the central United States have a less applicable travel path azimuth and may show different attenuation properties based on these different tectonic travel paths. The attenuation curves and empirical data are shown in Figures 2.5.2-259b and 2.5.2-259c for the 1 and 2.5 Hz spectral frequencies. The data point from the DWPF station is highlighted as a solid blue circle. In general the

DRAFT


empirical observations fall within the range of the Caribbean attenuation curves with the single exception of station LRAL (i.e., at a distance of about 750 km) for 2.5 Hz in which the empirical data exceeds the highest Caribbean attenuation curve. The empirical observation has not been adjusted to a common site condition of Vs= 2.83 km/sec and it could be expected that this empirical data point could possibly be reduced if a site amplification adjustment were to be applied. It can also be concluded from the comparison plots in Figures 2.5.2-259b and 2.5.2-259c, that the distribution of the current Caribbean attenuation curves adequately captures the range of the empirical data. In addition, the observation from the DWPF station is in the lower range of the Caribbean attenuation curves.

02/04/2007 – Cuba Region (Mw6.2, 19.49 N, -78.34 W, hypocenter depth 12.0 km)

This event occurred south of the island of Cuba and its epicenter is shown in Figure 2.5.2-260a along with the Turkey Point Units 6 & 7 site location, and the location of the IRIS stations that recorded this earthquake. Only three IRIS stations were analyzed from this earthquake and their station locations are shown in Figure 2.5.2-260a. Based on the observed station distribution, the IRIS station DWPF located in central Florida has the most applicable source to site path for this comparison. The attenuation curves and empirical data are shown in Figures 2.5.2-260b and 2.5.2-260c for the 1 and 2.5 Hz spectral frequencies. The data point from the DWPF station is highlighted as a solid blue circle. Overall the empirical data falls below the median Caribbean attenuation curve (heavy red line) and has values which are in the lower distribution range of Caribbean attenuation curves. In addition, the observation from the DWPF station is similar to the lowest Caribbean attenuation curve or lower.

05/28/2009 – North of Honduras (Mw7.3, 16.50 N, -87.17 W, hypocenter depth 12.0 km)

This event occurred north of Honduras in Central America and is the largest earthquake in the suite of five events analyzed in this sensitivity study. The location of its epicenter is shown in Figure 2.5.2-261a along with the Turkey Point Units 6 & 7 site location, and the location of the three IRIS stations that were analyzed. Based on the azimuths from the earthquake to the three IRIS stations, none of the associated seismic ray travel paths are ideal for this comparison study. The attenuation curves and empirical data are shown in Figures 2.5.2-261b and 2.5.2-261c for the 1 and 2.5 Hz spectral frequencies. The comparisons provided in the figures indicate that the empirical data from this earthquake are lower than any of the Caribbean attenuation curves.

01/12/2010 – Haiti (Mw7.0, 18.61 N, -72.62 W, hypocenter depth 12.0 km)

The Haiti earthquake is the most recent event in this suite of five earthquakes analyzed. The epicentral location and the suite of IRIS stations analyzed are shown in Figure 2.5.2-262a. Based on the observed station distribution, the IRIS station OTAV located in between the south part of Florida and the island of

DRAFT


Cuba has the most applicable source to site path for this comparison. Note that the data from the DWPF was not available for this earthquake. The attenuation curves and empirical data are shown in Figures 2.5.2-262b and 2.5.2-262c for the 1 and 2.5 Hz spectral frequencies. The data point from the OTAV station is highlighted as a solid blue circle. Overall the empirical data falls in the lower range or lower than the Caribbean attenuation curves. In addition, the observation from the OTAV station is significantly lower than the entire range of the Caribbean attenuation curves.

The results of these comparisons demonstrated that the suite of Caribbean GMPE models used in the current PSHA predicts larger ground motions on average than the observed empirical data from these five earthquakes for the spectral frequencies of 1 and 2.5 Hz. In addition, for the subset of data from just those stations that have a more appropriate source to site travel paths in particular this conclusion can be extended to state that the use of the Caribbean GMPE models should provide conservative (i.e., higher) ground motion predictions compared to the available empirical data from the region.

The second supplemental sensitivity determined the sensitivity of the ground motion response spectrum (GMRS) at the Turkey Point Units 6 & 7 site to the GMPEs used for the Caribbean seismic sources, which include the Cuba areal source plus nine fault sources. Five sets of seismic hazard calculations for five different GMPE suites were used to develop the corresponding GMRS. The GMPEs used for Caribbean seismic sources are: those of the base case which is (1) the Caribbean GMPE models developed for the Turkey Point Units 6 & 7 FSAR (Subsection 2.5.2.4.5.2), (2) EPRI Mid-continent region (References 242 and 203) equations, (3) EPRI Mid-continent region “mod1” (References 242 and 203) equations, (4) EPRI Gulf region (References 242 and 203) equations, and (5) EPRI Gulf region “mod 1” (References 242 and 203) equations. The modification of the EPRI GMPEs for both the Mid-continent and Gulf Coast cases was to exclude the GMPE which predicts significantly higher ground motions for large distance, such as those from the contributing Cuba and Caribbean sources. All seismic hazard calculations were made for the hard rock site conditions and include the other non-Caribbean sources in the PSHA. Thus the observed differences are based solely on the use of different GMPE models for the Caribbean seismic sources. For each of the five cases, ground motion values were estimated from the mean hazard curves for the seven standard spectral frequencies. Site specific horizontal GMRS spectra are plotted in Figure 2.5.2-263 and are developed using the site amplification factors. In general, the GMRS results using the EPRI Gulf Coast GMPEs are equal to or lower than the results using the Caribbean GMPE (i.e., indicated as Base Case in Figure 2.5.2-263). For the EPRI Mid-continent GMPE, the opposite result is concluded in which the GMRS values exceed the GMRS values using the Caribbean GMPE models, however, based on the previous additional sensitivity, the EPRI Mid-continent models predicts higher ground motions than the empirical ground motions and as such may not be applicable for the modeling of ground motion attenuation for seismic sources in this Caribbean region and the Turkey Point Units 6 & 7 site location in southern Florida. In addition the tectonic structure and potential attenuation of ground motions for Caribbean sources might be more consistent with the EPRI Gulf Coast GMPE models

DRAFT


based on the southern region of Florida being located in the Gulf Coast tectonic region of the eastern United States and based on the results shown in Figure 2.5.2-263, the use of the EPRI Gulf Coast GMPE for the PSHA gives similar or lower GMRS values.

These two supplemental sensitivity analyses indicate that the suite of GMPEs used to characterize contributions to hazard at the site from Caribbean earthquakes, and the resulting GMRS at the site from these earthquakes, are conservative.

Next, six ground motion attenuation experts were asked to review and comment on both the methodology and results of the Caribbean GMPE models. These six experts were:

• Dr. Norman Abrahamson (Consultant)

• Dr. Yousef Bozorgnia (University of California, Berkeley, PEER)

• Dr. Kenneth Campbell (EQECAT)

• Dr. Shahram Pezeshk (University of Memphis)

• Dr. Paul Somerville (URS Corporation)

• Dr. Robert Youngs (AMEC-Geomatrix)

In reviewing and summarizing the responses from the experts it is concluded that the use of the Caribbean GMPE models in the PSHA is acceptable. As noted by the six experts, based on the comparison between the Caribbean GMPE model and the empirical IRIS data, these GMPE models may be conservative in their estimation of ground motions in the region. In addition the assignment of an aleatory uncertainty value of 0.645 in natural log units was acceptable based on the consensus of the six experts.

When asked about the applicability of the EPRI Mid-continent or Gulf Coast models, the consensus of the six experts was that the suite of Gulf Coast GMPE models are closer to the empirical data than the Mid-continent suite. However, this did not indicate that the EPRI Gulf Coast models are preferable over the Caribbean GMPE models. Based on this conclusion with regards to the two EPRI ground motion models and the current sensitivity studies which showed that the use of the EPRI Gulf Coast GMPE in the PSHA resulted in lower GMRS ground motions, it is concluded that the use of the Caribbean GMPE models is conservative in the estimation of the GMRS ground motions for the Turkey Point Units 6 & 7 site.

Mean ground motion attenuation curve plots are provided for the EPRI Mid-continent (References 242 and 203) and Caribbean ground motion prediction equations (GMPEs) in Figures 2.5.2-264 – 2.5.2-270 for the seven defined frequencies. Note that EPRI (Reference 203) is only a recommendation for the associated aleatory uncertainty and therefore does not impact the comparison plots of the weighted mean ground motion attenuation curve. The plotted mean attenuation curve is the weighted mean of the individual median attenuation curves as defined in EPRI (Reference 242) and for the Caribbean GMPE models. These attenuation plots are provided for three specific Mw values: 6, 7, and 8 for distances between 200km and 1,000km. Note that the EPRI (Reference 242) ground motion attenuation model is defined as a function of epicentral distance, whereas the Caribbean ground motion

DRAFT


attenuation model is defined as a function of hypocentral distance. For these comparison plots, the attenuation curves are plotted as a function of epicentral distance and for the Caribbean GMPE curves an assumed hypocentral depth of 8 km was used.

The following references will be added to FSAR Subsection 2.5.2.7

344. Atkinson, G., Boore, D., "Empirical Ground Motion Relation for Subduction Zone Earthquakes and their Application to Cascadia and other regions," Bulletin of the Seismological Society of America, v. 93, no. 4, pp. 1703 - 1729, 2003.

345. Atkinson, G., Silva, W., "Stochastic Modeling of California Ground Motions," Bulletin of the Seismological Society of America, v. 90, no. 2, pp. 255 - 274, 2000.

DRAFT


The following new figures will be added to FSAR Section 2.5.2:

Figure 2.5.2-258a

Map showing the earthquake location for the 12/14/2004 Caribbean Sea Region earthquake (Mw6.8), the IRIS station locations used in the analysis (shown as

black triangles in this and similar subsequent figures), and Turkey Point Units 6 & 7 site location.

DRAFT


Figure 2.5.2-258b

Comparison of Caribbean GMPE (red lines), EPRI Mid-continent (blue lines) and Gulf Coast region (green lines) GMPE and empirical IRIS processed data for a spectral

frequency of 1 Hz

0.000001

0.00001

0.0001

0.001

0.01

0.1

1

100 1000 10000

Sp

ec

tra

l Ac

ce

lera

tio

n (g

)


Attenuation Models: 1Hz, 12/14/2004, M6.8







IRIS Data

DRAFT


Figure 2.5.2-258c


frequency of 2.5 Hz.

0.000001

0.00001

0.0001

0.001

0.01

0.1

1

100 1000 10000

Sp

ec

tra

l Ac

ce

lera

tio

n (g

)



DRAFT


Figure 2.5.2-259a

Map showing the earthquake location for the 09/10/2006 Gulf of Mexico earthquake (Mw5.9), the IRIS station locations used in the analysis and Turkey Point

Units 6 & 7 site location.

DRAFT


Figure 2.5.2-259b

Comparison of Caribbean GMPE (red lines), EPRI Mid-continent (blue lines) and Gulf

Coast region (green lines) GMPE and empirical IRIS processed data for a spectral frequency of 1 Hz.

0.000001

0.00001

0.0001

0.001

0.01

0.1

1

100 1000 10000

Sp

ec

tra

l Ac

ce

lera

tio

n (g

)



DRAFT


Figure 2.5.2-259c


Coast region (green lines) GMPE and empirical IRIS processed data for a spectral frequency of 2.5 Hz.

0.000001

0.00001

0.0001

0.001

0.01

0.1

1

100 1000 10000

Sp

ec

tra

l Ac

ce

lera

tio

n (g

)


Attenuation Models: 2.5Hz, 09/10/2006, M5.9







IRIS Data

DRAFT


Figure 2.5.2-260a

Map showing the earthquake location for the 02/04/2007 Cuba Region earthquake

(Mw6.2), the IRIS station locations used in the analysis and Turkey Point Units 6 & 7 site location

DRAFT


Figure 2.5.2-260b


frequency of 1 Hz.

0.000001

0.00001

0.0001

0.001

0.01

0.1

1

100 1000 10000

Sp

ec

tra

l Ac

ce

lera

tio

n (g

)


Caribbean Attenuation Models: 1Hz, 02/04/2007, M6.2


DRAFT


Figure 2.5.2-260c



0.000001

0.00001

0.0001

0.001

0.01

0.1

1

100 1000 10000

Sp

ec

tra

l Ac

ce

lera

tio

n (g

)




DRAFT


Figure 2.5.2-261a

Map showing the earthquake location for the 05/28/2009 North of Honduras

earthquake (Mw7.3), the IRIS station locations used in the analysis and Turkey Point Units 6 & 7 site location.

DRAFT


Figure 2.5.2-261b



0.000001

0.00001

0.0001

0.001

0.01

0.1

1

100 1000 10000

Sp

ec

tra

l Ac

ce

lera

tio

n (g

)



DRAFT


Figure 2.5.2-261c


Coast region (green lines) GMPE and empirical IRIS processed data for a spectral frequency of 2.5 Hz

0.000001

0.00001

0.0001

0.001

0.01

0.1

1

100 1000 10000

Sp

ec

tra

l Ac

ce

lera

tio

n (g

)



Mean 1CC-QBase-SDAll 1CC-QHigh-SDAll 1CC-QLow-SDAll 1CV-QBase-SDAll 1CV-QHigh-SDAll1CV-QLow-SDAll 2C-QBase 2C-QHigh 2C-QLow EPRI(Mean) MidCont EPRI(1) MidContEPRI(2) MidCont EPRI(3) MidCont EPRI(4) MidCont EPRI(5) MidCont EPRI(6) MidCont EPRI(7) MidContEPRI(8) MidCont EPRI(9) MidCont EPRI(Mean) Gulf EPRI(1) Gulf EPRI(2) Gulf EPRI(3) GulfEPRI(4) Gulf EPRI(5) Gulf EPRI(6) Gulf EPRI(7) Gulf EPRI(8) Gulf EPRI(9) GulfIRIS Data

DRAFT


Figure 2.5.2-262a

Map showing the earthquake location for the 01/12/2010 Haiti earthquake (Mw7.0), the IRIS station locations used in the analysis and Turkey Point Units 6 & 7 site location.

DRAFT


Figure 2.5.2-262b



0.00001

0.0001

0.001

0.01

0.1

1

100 1000 10000

Sp

ec

tra

l Ac

ce

lera

tio

n (g

)


01/12/2010 Earthquake: 1Hz, 01/12/2010, M7

Mean

1CC-QBase-SDAll

1CC-QHigh-SDAll

1CC-QLow-SDAll

1CV-QBase-SDAll

1CV-QHigh-SDAll

1CV-QLow-SDAll

2C-QBase

2C-QHigh

2C-QLow

EPRI(Mean) MidCont

EPRI(1) MidCont

EPRI(2) MidCont

EPRI(3) MidCont

EPRI(4) MidCont

EPRI(5) MidCont

EPRI(6) MidCont

EPRI(7) MidCont

EPRI(8) MidCont

EPRI(9) MidCont

EPRI(Mean) Gulf

EPRI(1) Gulf

EPRI(2) Gulf

EPRI(3) Gulf

EPRI(4) Gulf

EPRI(5) Gulf

EPRI(6) Gulf

EPRI(7) Gulf

EPRI(8) Gulf

EPRI(9) Gulf

IRIS Data

DRAFT


Figure 2.5.2-262c



0.00001

0.0001

0.001

0.01

0.1

1

100 1000 10000

Sp

ec

tra

l Ac

ce

lera

tio

n (g

)


01/12/2010 Earthquake: 2.5Hz, 01/12/2010, M7

Mean

1CC-QBase-SDAll

1CC-QHigh-SDAll

1CC-QLow-SDAll

1CV-QBase-SDAll

1CV-QHigh-SDAll

1CV-QLow-SDAll

2C-QBase

2C-QHigh

2C-QLow

EPRI(Mean) MidCont

EPRI(1) MidCont

EPRI(2) MidCont

EPRI(3) MidCont

EPRI(4) MidCont

EPRI(5) MidCont

EPRI(6) MidCont

EPRI(7) MidCont

EPRI(8) MidCont

EPRI(9) MidCont

EPRI(Mean) Gulf

EPRI(1) Gulf

EPRI(2) Gulf

EPRI(3) Gulf

EPRI(4) Gulf

EPRI(5) Gulf

EPRI(6) Gulf

EPRI(7) Gulf

EPRI(8) Gulf

EPRI(9) Gulf

IRIS Data

DRAFT


Figure 2.5.2-263

Horizontal GMRS for Caribbean sources for five GMPE models

for the Turkey Point Units 6 & 7 site.

0.01

0.1

1

0.1 1 10 100

Horizontal G

MRS (g)

Frequency (Hz)

Turkey PointGMRS Sensitivity

TP‐MidC

TP‐MidC‐mod1

TP‐Base Case

TP‐Gulf

TP‐Gulf‐mod1

DRAFT


Figure 2.5.2-264

Comparison of EPRI Mid-continent (Reference 242) mean (dashed lines) and

Caribbean mean (solid lines) attenuation curves for magnitudes 6, 7, and 8 for PGA.

0.00001

0.0001

0.001

0.01

0.1

1

100 1000

Sp

ec

tra

l Ac

ce

lera

tio

n (g

)


Attenuation Models: PGA

EPRI MidC (2004) M8

Caribbean M8

EPRI MidC (2004) M7

Caribbean M7

EPRI MidC (2004) M6

Caribbean M6

DRAFT


Figure 2.5.2-265


Caribbean mean (solid lines) attenuation curves for magnitudes 6, 7, and 8 for 25 Hz.

0.00001

0.0001

0.001

0.01

0.1

1

100 1000

Sp

ec

tra

l Ac

ce

lera

tio

n (g

)



EPRI MidC (2004) M8

Caribbean M8

EPRI MidC (2004) M7

Caribbean M7

EPRI MidC (2004) M6

Caribbean M6

DRAFT


Figure 2.5.2-266



0.00001

0.0001

0.001

0.01

0.1

1

100 1000

Sp

ec

tra

l Ac

ce

lera

tio

n (g

)



EPRI MidC (2004) M8

Caribbean M8

EPRI MidC (2004) M7

Caribbean M7

EPRI MidC (2004) M6

Caribbean M6

DRAFT


Figure 2.5.2-267

Comparison of EPRI Mid-continent (Reference 242) mean (dashed lines) and Caribbean mean (solid lines) attenuation curves

for magnitudes 6, 7, and 8 for 5 Hz.

0.00001

0.0001

0.001

0.01

0.1

1

100 1000

Sp

ec

tra

l Ac

ce

lera

tio

n (g

)



EPRI MidC (2004) M8

Caribbean M8

EPRI MidC (2004) M7

Caribbean M7

EPRI MidC (2004) M6

Caribbean M6

DRAFT


Figure 2.5.2-268


Caribbean mean (solid lines) attenuation curves for magnitudes 6, 7, and 8 for 2.5 Hz.

0.00001

0.0001

0.001

0.01

0.1

1

100 1000

Sp

ec

tra

l Ac

ce

lera

tio

n (g

)



EPRI MidC (2004) M8

Caribbean M8

EPRI MidC (2004) M7

Caribbean M7

EPRI MidC (2004) M6

Caribbean M6

DRAFT


Figure 2.5.2-269



0.00001

0.0001

0.001

0.01

0.1

1

100 1000

Sp

ec

tra

l Ac

ce

lera

tio

n (g

)



EPRI MidC (2004) M8

Caribbean M8

EPRI MidC (2004) M7

Caribbean M7

EPRI MidC (2004) M6

Caribbean M6

DRAFT


Figure 2.5.2-270


Caribbean mean (solid lines) attenuation curves for magnitudes 6, 7, and 8 for 0.5 Hz.

0.00001

0.0001

0.001

0.01

0.1

1

100 1000

Sp

ec

tra

l Ac

ce

lera

tio

n (g

)



EPRI MidC (2004) M8

Caribbean M8

EPRI MidC (2004) M7

Caribbean M7

EPRI MidC (2004) M6

Caribbean M6

DRAFT


ASSOCIATED ENCLOSURES:

Enclosure A – SSHAC Caribbean GMPE Questionnaire

DRAFT

Proposed Turkey Point Units 6 and 7 Docket Nos. 52-040 and 52-041 FPL Draft Revised Response to NRC RAI No. 02.05.02-2 (eRAI 5896) Enclosure A Page 1 of 57

Enclosure A

Turkey Point Units 6 and 7

COL Application

SSHAC Caribbean GMPE Questionnaire

DRAFT


We have developed a suite of ground motion prediction equations (GMPEs) for use in a probabilistic seismic hazard analysis (PSHA) for an electrical generation plant located in southern Florida. The PSHA incorporates contributions to seismic hazard from seismic sources both in the immediate region of the southeastern United States and, as well, seismic sources located at distances greater than 1,000 km in the Caribbean region.

The newly developed GMPE models that are the subject of this questionnaire are applied only to Caribbean seismic sources.

Our motivation in developing these GMPEs is the lack of a suite of region-specific empirical GMPE(s) that could be used within the PSHA study for large magnitude earthquakes (i.e., magnitudes greater than 7) and over the large distance range (i.e., 150 – 2,000km) needed. We employed a stochastic simulation procedure to generate a dataset on which we performed a regression analysis. Critical to any stochastic simulation procedure is the selection of the necessary seismological input parameter values. Our selected values are listed in Table 1 and are based on the empirical ground motion study for the island of Puerto Rico (Motazedian and Atkinson, 2005) along with an uncertainty on the stress parameter and quality factor models.

Table 1. Regional attenuation and source parameters used in the Motazedian and Atkinson (2005) study for the simulation of ground motions.

Parameter Simulation Values Stress Parameter 65 bars (Low Case)1

130 bars (Base Case) 260 bars (High Case)1

Geometrical Spreading 1/R for R<75km 1.0 for 75<R<100km

1/SQRT(R) for R>100km

Quality Factor (Q) Model 241 f^0.59 (Low Case)2

359 f^0.59 (Base Case) 536 f^0.59 (High Case)2

Path Duration Atkinson and Boore (1995) Model with hinge points at 75 and 100 km

Site Amplification Chen and Atkinson (2002) CEUS Hard Rock

Kappa 0.006 sec (EPRI, 1993) Shear Wave Velocity (Vs)

at the Source 3.6 km/sec

Density 2.8 g/cm3

1 Based on assumed sigma of 0.7 (natural log units) taken from EPRI, 1993. 2 Based on assumed sigma of 0.4 (natural log units) taken from Silva et al., 2003


DRAFT

We also considered three source spectra models: single corner with constant stress parameter (Boore, 2005), single corner with variable stress parameter (Silva et al., 2003) and double corner stress parameter (Atkinson and Boore, 1995) source models. For the variable stress parameter model, the magnitude dependence is given in Table 2.

Table 2. Variable stress parameter values used in the simulation for the single corner variable stress parameter model (from Silva et al., 2003) .

Magnitude Low Stress Parameter

Base Stress Parameter

High Stress Parameter

4.75 65 130 260 5.00 65 130 260 5.25 65 130 260 5.50 65 130 260 5.75 61 122 244 6.00 57 114 228 6.25 53 106 212 6.50 49 98 196 6.75 46 91 182 7.00 43 85 170 7.25 40 79 158 7.50 37 73 146 7.75 35 69 138 8.00 33 65 130 8.25 31 61 122 8.50 29 57 114 8.75 29 57 114

The results for the several stress parameter cases for the single corner scaling models were found to vary about the base case and were combined for a given Q model variation. These combined stress drop attenuation curves are designated “SDAll” below. For the double corner [2C] source model, the variation in stress parameter is not included because this source model is not stress parameter dependent.

GMPE models were developed for the seven spectral frequencies used in the PSHA: PGA, 25Hz, 10Hz, 5Hz, 2.5Hz, 1Hz, and 0.5Hz. These newly developed GMPEs are applicable for CEUS hard rock site conditions (i.e., Vs=2.8km/sec). A recommended total sigma of 0.645 in natural log units was proposed based on the estimated sigma value given in the Motazedian and Atkinson (2005) study.

For the suite of nine ground motion models (three source models and three Q factor models), a non-uniform weighting scheme was recommended. The single corner [1C] source model (both constant stress parameter [1CC] and variable [1CV] stress parameter) fall into Cluster 1 of the EPRI (2004) ground motion study for General Area seismic sources, while the double corner [2C] model falls into Cluster 2. The third Cluster in the EPRI (2004) study is based on hybrid empirical attenuation models, which we have not considered. The relative weights (from Table 5, EPRI 2004) are listed as:


DRAFT

Cluster 1 = 0.3512 Cluster 2 = 0.3985 Cluster 3 = 0.2503 Total = 1.0

Given these values, and specifically the nearly equal weights of Clusters 1 and 2, we have recommended that the relative weights between the single corner source models and double corner source model be weighted equally. Within the single corner source model case, the constant [1CC] and variable [1CV] stress parameter cases are also considered equally weighted. Finally, because of the large uncertainty in the regional anelastic attenuation, the relative weighting between the three Q low, base, and high models was assigned to be equal (i.e., 0.333, 0.334, 0.333). Combining these weights the weights for each of the nine recommended models are as listed in Table 3.

Table 3. Summary of attenuation tables listing regression coefficients and associated weights for each of nine recommended models.

Attenuation Model Model Weight 1CC, Q Base 0.08341CC, Q High 0.0833 1CC, Q Low 0.08331CV, Q Base 0.0834 1CV, Q High 0.0833 1CV, Q Low 0.08332C, Q Base 0.16672C, Q High 0.16672C, Q Low 0.1666

A comparison is presented in Figures 1 – 7 of the newly developed Caribbean GMPE (red curves) with the EPRI (2004) Mid-Continent (blue curves) and Gulf Coast (red curves) regional GMPE models. Attenuation curves are presented for magnitudes 7 and 8 between distances of 200 – 1,000 km. For each of the three GMPE models, the individual models are graphically plotted as thin color coded lines with the weighted mean attenuation curve being shown as the heavy color coded line.


DRAFT

0.00

001

0.00

01

0.00

1

0.010.

11 100

1000

Spectral Acceleration (g)

Hyp

ocen

tral D

ista

nce

(km

)

Car

ibbe

an A

ttenu

atio

n M

odel

s: P

GA

, M7

Mea

n1C

C-Q

Base

-SD

All

1CC

-QH

igh-

SDAl

l1C

C-Q

Low

-SD

All

1CV-

QBa

se-S

DAl

l1C

V-Q

Hig

h-SD

All

1CV-

QLo

w-S

DAl

l2C

-QBa

se2C

-QH

igh

2C-Q

Low

EPR

I(Mea

n) M

idCo

ntEP

RI(1

) Mid

Con

tEP

RI(2

) Mid

Con

tEP

RI(3

) Mid

Con

tEP

RI(4

) Mid

Con

tEP

RI(5

) Mid

Con

tEP

RI(6

) Mid

Con

tEP

RI(7

) Mid

Con

tEP

RI(8

) Mid

Con

tEP

RI(9

) Mid

Con

tEP

RI(M

ean)

Gul

fEP

RI(1

) Gul

fEP

RI(2

) Gul

fEP

RI(3

) Gul

fEP

RI(4

) Gul

fEP

RI(5

) Gul

fEP

RI(6

) Gul

fEP

RI(7

) Gul

fEP

RI(8

) Gul

fEP

RI(9

) Gul

f

Figu

re 1

a. M

agni

tude

7 a

ttenu

atio

n cu

rve

for P

GA

for t

he C

arib

bean

(red

cur

ves)

, Mid

-Con

tinen

t (bl

ue c

urve

s), a

nd G

ulf C

oast

(g

reen

cur

ves)

regi

on G

MPE

mod

els.

Fi

le: \

\am

ers\

corp

\eng

tech

\GH

ES\G

HES

\FPL

_Tur

keyP

tCO

L\G

S\A

ttenu

atio

n\SS

HA

C-L

evel

2-G

MPE

\Com

pare

_PG

A.x

ls, C

hart-

M7


DRAFT

0.00

001

0.00

01

0.00

1

0.010.

11 100

1000


Hyp

ocen

tral D

ista

nce

(km

)

Car

ibbe

an A

ttenu

atio

n M

odel

s: P

GA

, M8

Mea

n1C

C-Q

Base

-SD

All

1CC

-QH

igh-

SDAl

l1C

C-Q

Low

-SD

All

1CV-

QBa

se-S

DAl

l1C

V-Q

Hig

h-SD

All

1CV-

QLo

w-S

DAl

l2C

-QBa

se2C

-QH

igh

2C-Q

Low

EPR

I(Mea

n) M

idCo

ntEP

RI(1

) Mid

Con

tEP

RI(2

) Mid

Con

tEP

RI(3

) Mid

Con

tEP

RI(4

) Mid

Con

tEP

RI(5

) Mid

Con

tEP

RI(6

) Mid

Con

tEP

RI(7

) Mid

Con

tEP

RI(8

) Mid

Con

tEP

RI(9

) Mid

Con

tEP

RI(M

ean)

Gul

fEP

RI(1

) Gul

fEP

RI(2

) Gul

fEP

RI(3

) Gul

fEP

RI(4

) Gul

fEP

RI(5

) Gul

fEP

RI(6

) Gul

fEP

RI(7

) Gul

fEP

RI(8

) Gul

fEP

RI(9

) Gul

f

Figu

re 1

b. M

agni

tude

8 a

ttenu

atio

n cu

rve

for P

GA

for t

he C

arib

bean

(red

cur

ves)

, Mid

-Con

tinen

t (bl

ue c

urve

s), a

nd G

ulf C

oast

(g

reen

cur

ves)

regi

on G

MPE

mod

els.

Fi

le: \

\am

ers\

corp

\eng

tech

\GH

ES\G

HES

\FPL

_Tur

keyP

tCO

L\G

S\A

ttenu

atio

n\SS

HA

C-L

evel

2-G

MPE

\Com

pare

_PG

A.x

ls, C

hart-

M8


DRAFT

0.00

001

0.00

01

0.00

1

0.010.

11 100

1000


Hyp

ocen

tral D

ista

nce

(km

)

Car

ibbe

an A

ttenu

atio

n M

odel

s: 2

5Hz,

M7

Mea

n1C

C-Q

Base

-SD

All

1CC

-QH

igh-

SDAl

l1C

C-Q

Low

-SD

All

1CV-

QBa

se-S

DAl

l1C

V-Q

Hig

h-SD

All

1CV-

QLo

w-S

DAl

l2C

-QBa

se2C

-QH

igh

2C-Q

Low

EPR

I(Mea

n) M

idCo

ntEP

RI(1

) Mid

Con

tEP

RI(2

) Mid

Con

tEP

RI(3

) Mid

Con

tEP

RI(4

) Mid

Con

tEP

RI(5

) Mid

Con

tEP

RI(6

) Mid

Con

tEP

RI(7

) Mid

Con

tEP

RI(8

) Mid

Con

tEP

RI(9

) Mid

Con

tEP

RI(M

ean)

Gul

fEP

RI(1

) Gul

fEP

RI(2

) Gul

fEP

RI(3

) Gul

fEP

RI(4

) Gul

fEP

RI(5

) Gul

fEP

RI(6

) Gul

fEP

RI(7

) Gul

fEP

RI(8

) Gul

fEP

RI(9

) Gul

f

Figu

re 2

a. M

agni

tude

7 a

ttenu

atio

n cu

rve

for 2

5 H

z fo

r the

Car

ibbe

an (r

ed c

urve

s), M

id-C

ontin

ent (

blue

cur

ves)

, and

Gul

f Coa

st

(gre

en c

urve

s) re

gion

GM

PE m

odel

s.

File

: \\a

mer

s\co

rp\e

ngte

ch\G

HES

\GH

ES\F

PL_T

urke

yPtC

OL\

GS\

Atte

nuat

ion\

SSH

AC

-Lev

el2-

GM

PE\C

ompa

re_2

5Hz.

xls,

Cha

rt-M

7


DRAFT

0.00

001

0.00

01

0.00

1

0.010.

11 100

1000


Hyp

ocen

tral D

ista

nce

(km

)

Car

ibbe

an A

ttenu

atio

n M

odel

s: 2

5Hz,

M8

Mea

n1C

C-Q

Base

-SD

All

1CC

-QH

igh-

SDAl

l1C

C-Q

Low

-SD

All

1CV-

QBa

se-S

DAl

l1C

V-Q

Hig

h-SD

All

1CV-

QLo

w-S

DAl

l2C

-QBa

se2C

-QH

igh

2C-Q

Low

EPR

I(Mea

n) M

idCo

ntEP

RI(1

) Mid

Con

tEP

RI(2

) Mid

Con

tEP

RI(3

) Mid

Con

tEP

RI(4

) Mid

Con

tEP

RI(5

) Mid

Con

tEP

RI(6

) Mid

Con

tEP

RI(7

) Mid

Con

tEP

RI(8

) Mid

Con

tEP

RI(9

) Mid

Con

tEP

RI(M

ean)

Gul

fEP

RI(1

) Gul

fEP

RI(2

) Gul

fEP

RI(3

) Gul

fEP

RI(4

) Gul

fEP

RI(5

) Gul

fEP

RI(6

) Gul

fEP

RI(7

) Gul

fEP

RI(8

) Gul

fEP

RI(9

) Gul

f

Figu

re 2

b. M

agni

tude

8 a

ttenu

atio

n cu

rve

for 2

5 H

z fo

r the

Car

ibbe

an (r

ed c

urve

s), M

id-C

ontin

ent (

blue

cur

ves)

, and

Gul

f Coa

st

(gre

en c

urve

s) re

gion

GM

PE m

odel

s.

File

: \\a

mer

s\co

rp\e

ngte

ch\G

HES

\GH

ES\F

PL_T

urke

yPtC

OL\

GS\

Atte

nuat

ion\

SSH

AC

-Lev

el2-

GM

PE\C

ompa

re_2

5Hz.

xls,

Cha

rt-M

8


DRAFT

0.00

001

0.00

01

0.00

1

0.010.

11 100

1000


Hyp

ocen

tral D

ista

nce

(km

)

Car

ibbe

an A

ttenu

atio

n M

odel

s: 1

0Hz,

M7

Mea

n1C

C-Q

Base

-SD

All

1CC

-QH

igh-

SDAl

l1C

C-Q

Low

-SD

All

1CV-

QBa

se-S

DAl

l1C

V-Q

Hig

h-SD

All

1CV-

QLo

w-S

DAl

l2C

-QBa

se2C

-QH

igh

2C-Q

Low

EPR

I(Mea

n) M

idCo

ntEP

RI(1

) Mid

Con

tEP

RI(2

) Mid

Con

tEP

RI(3

) Mid

Con

tEP

RI(4

) Mid

Con

tEP

RI(5

) Mid

Con

tEP

RI(6

) Mid

Con

tEP

RI(7

) Mid

Con

tEP

RI(8

) Mid

Con

tEP

RI(9

) Mid

Con

tEP

RI(M

ean)

Gul

fEP

RI(1

) Gul

fEP

RI(2

) Gul

fEP

RI(3

) Gul

fEP

RI(4

) Gul

fEP

RI(5

) Gul

fEP

RI(6

) Gul

fEP

RI(7

) Gul

fEP

RI(8

) Gul

fEP

RI(9

) Gul

f

Figu

re 3

a. M

agni

tude

7 a

ttenu

atio

n cu

rve

for 1

0 H

z fo

r the

Car

ibbe

an (r

ed c

urve

s), M

id-C

ontin

ent (

blue

cur

ves)

, and

Gul

f Coa

st

(gre

en c

urve

s) re

gion

GM

PE m

odel

s.

File

: \\a

mer

s\co

rp\e

ngte

ch\G

HES

\GH

ES\F

PL_T

urke

yPtC

OL\

GS\

Atte

nuat

ion\

SSH

AC

-Lev

el2-

GM

PE\C

ompa

re_1

0Hz.

xls,

Cha

rt-M

7


DRAFT

0.00

001

0.00

01

0.00

1

0.010.

11 100

1000


Hyp

ocen

tral D

ista

nce

(km

)

Car

ibbe

an A

ttenu

atio

n M

odel

s: 1

0Hz,

M8

Mea

n1C

C-Q

Base

-SD

All

1CC

-QH

igh-

SDAl

l1C

C-Q

Low

-SD

All

1CV-

QBa

se-S

DAl

l1C

V-Q

Hig

h-SD

All

1CV-

QLo

w-S

DAl

l2C

-QBa

se2C

-QH

igh

2C-Q

Low

EPR

I(Mea

n) M

idCo

ntEP

RI(1

) Mid

Con

tEP

RI(2

) Mid

Con

tEP

RI(3

) Mid

Con

tEP

RI(4

) Mid

Con

tEP

RI(5

) Mid

Con

tEP

RI(6

) Mid

Con

tEP

RI(7

) Mid

Con

tEP

RI(8

) Mid

Con

tEP

RI(9

) Mid

Con

tEP

RI(M

ean)

Gul

fEP

RI(1

) Gul

fEP

RI(2

) Gul

fEP

RI(3

) Gul

fEP

RI(4

) Gul

fEP

RI(5

) Gul

fEP

RI(6

) Gul

fEP

RI(7

) Gul

fEP

RI(8

) Gul

fEP

RI(9

) Gul

f

Figu

re 3

b. M

agni

tude

8 a

ttenu

atio

n cu

rve

for 1

0 H

z fo

r the

Car

ibbe

an (r

ed c

urve

s), M

id-C

ontin

ent (

blue

cur

ves)

, and

Gul

f Coa

st

(gre

en c

urve

s) re

gion

GM

PE m

odel

s.

File

: \\a

mer

s\co

rp\e

ngte

ch\G

HES

\GH

ES\F

PL_T

urke

yPtC

OL\

GS\

Atte

nuat

ion\

SSH

AC

-Lev

el2-

GM

PE\C

ompa

re_1

0Hz.

xls,

Cha

rt-M

8


DRAFT

0.00

001

0.00

01

0.00

1

0.010.

11 100

1000


Hyp

ocen

tral D

ista

nce

(km

)

Car

ibbe

an A

ttenu

atio

n M

odel

s: 5

Hz,

M7

Mea

n1C

C-Q

Base

-SD

All

1CC

-QH

igh-

SDAl

l1C

C-Q

Low

-SD

All

1CV-

QBa

se-S

DAl

l1C

V-Q

Hig

h-SD

All

1CV-

QLo

w-S

DAl

l2C

-QBa

se2C

-QH

igh

2C-Q

Low

EPR

I(Mea

n) M

idCo

ntEP

RI(1

) Mid

Con

tEP

RI(2

) Mid

Con

tEP

RI(3

) Mid

Con

tEP

RI(4

) Mid

Con

tEP

RI(5

) Mid

Con

tEP

RI(6

) Mid

Con

tEP

RI(7

) Mid

Con

tEP

RI(8

) Mid

Con

tEP

RI(9

) Mid

Con

tEP

RI(M

ean)

Gul

fEP

RI(1

) Gul

fEP

RI(2

) Gul

fEP

RI(3

) Gul

fEP

RI(4

) Gul

fEP

RI(5

) Gul

fEP

RI(6

) Gul

fEP

RI(7

) Gul

fEP

RI(8

) Gul

fEP

RI(9

) Gul

f

Figu

re 4

a. M

agni

tude

7 a

ttenu

atio

n cu

rve

for 5

Hz

for t

he C

arib

bean

(red

cur

ves)

, Mid

-Con

tinen

t (bl

ue c

urve

s), a

nd G

ulf C

oast

(gre

encu

rves

) reg

ion

GM

PE m

odel

s.

File

: \\a

mer

s\co

rp\e

ngte

ch\G

HES

\GH

ES\F

PL_T

urke

yPtC

OL\

GS\

Atte

nuat

ion\

SSH

AC

-Lev

el2-

GM

PE\C

ompa

re_0

5Hz.

xls,

Cha

rt-M

7


DRAFT

0.00

001

0.00

01

0.00

1

0.010.

11 100

1000


Hyp

ocen

tral D

ista

nce

(km

)

Car

ibbe

an A

ttenu

atio

n M

odel

s: 5

Hz,

M8

Mea

n1C

C-Q

Base

-SD

All

1CC

-QH

igh-

SDAl

l1C

C-Q

Low

-SD

All

1CV-

QBa

se-S

DAl

l1C

V-Q

Hig

h-SD

All

1CV-

QLo

w-S

DAl

l2C

-QBa

se2C

-QH

igh

2C-Q

Low

EPR

I(Mea

n) M

idCo

ntEP

RI(1

) Mid

Con

tEP

RI(2

) Mid

Con

tEP

RI(3

) Mid

Con

tEP

RI(4

) Mid

Con

tEP

RI(5

) Mid

Con

tEP

RI(6

) Mid

Con

tEP

RI(7

) Mid

Con

tEP

RI(8

) Mid

Con

tEP

RI(9

) Mid

Con

tEP

RI(M

ean)

Gul

fEP

RI(1

) Gul

fEP

RI(2

) Gul

fEP

RI(3

) Gul

fEP

RI(4

) Gul

fEP

RI(5

) Gul

fEP

RI(6

) Gul

fEP

RI(7

) Gul

fEP

RI(8

) Gul

fEP

RI(9

) Gul

f

Figu

re 4

b. M

agni

tude

8 a

ttenu

atio

n cu

rve

for 5

Hz

for t

he C

arib

bean

(red

cur

ves)

, Mid

-Con

tinen

t (bl

ue c

urve

s), a

nd G

ulf C

oast

(gre

encu

rves

) reg

ion

GM

PE m

odel

s.

File

: \\a

mer

s\co

rp\e

ngte

ch\G

HES

\GH

ES\F

PL_T

urke

yPtC

OL\

GS\

Atte

nuat

ion\

SSH

AC

-Lev

el2-

GM

PE\C

ompa

re_0

5Hz.

xls,

Cha

rt-M

8


DRAFT

0.00

001

0.00

01

0.00

1

0.010.

11 100

1000


Hyp

ocen

tral D

ista

nce

(km

)

Car

ibbe

an A

ttenu

atio

n M

odel

s: 2

.5H

z, M

7

Mea

n1C

C-Q

Base

-SD

All

1CC

-QH

igh-

SDAl

l1C

C-Q

Low

-SD

All

1CV-

QBa

se-S

DAl

l1C

V-Q

Hig

h-SD

All

1CV-

QLo

w-S

DAl

l2C

-QBa

se2C

-QH

igh

2C-Q

Low

EPR

I(Mea

n) M

idCo

ntEP

RI(1

) Mid

Con

tEP

RI(2

) Mid

Con

tEP

RI(3

) Mid

Con

tEP

RI(4

) Mid

Con

tEP

RI(5

) Mid

Con

tEP

RI(6

) Mid

Con

tEP

RI(7

) Mid

Con

tEP

RI(8

) Mid

Con

tEP

RI(9

) Mid

Con

tEP

RI(M

ean)

Gul

fEP

RI(1

) Gul

fEP

RI(2

) Gul

fEP

RI(3

) Gul

fEP

RI(4

) Gul

fEP

RI(5

) Gul

fEP

RI(6

) Gul

fEP

RI(7

) Gul

fEP

RI(8

) Gul

fEP

RI(9

) Gul

f

Figu

re 5

a. M

agni

tude

7 a

ttenu

atio

n cu

rve

for 2

.5 H

z fo

r the

Car

ibbe

an (r

ed c

urve

s), M

id-C

ontin

ent (

blue

cur

ves)

, and

Gul

f Coa

st(g

reen

cur

ves)

regi

on G

MPE

mod

els.

Fi

le: \

\am

ers\

corp

\eng

tech

\GH

ES\G

HES

\FPL

_Tur

keyP

tCO

L\G

S\A

ttenu

atio

n\SS

HA

C-L

evel

2-G

MPE

\Com

pare

_02p

5Hz.

xls,

Cha

rt-M

7


DRAFT

0.00

001

0.00

01

0.00

1

0.010.

11 100

1000


Hyp

ocen

tral D

ista

nce

(km

)

Car

ibbe

an A

ttenu

atio

n M

odel

s: 2

.5H

z, M

8

Mea

n1C

C-Q

Base

-SD

All

1CC

-QH

igh-

SDAl

l1C

C-Q

Low

-SD

All

1CV-

QBa

se-S

DAl

l1C

V-Q

Hig

h-SD

All

1CV-

QLo

w-S

DAl

l2C

-QBa

se2C

-QH

igh

2C-Q

Low

EPR

I(Mea

n) M

idCo

ntEP

RI(1

) Mid

Con

tEP

RI(2

) Mid

Con

tEP

RI(3

) Mid

Con

tEP

RI(4

) Mid

Con

tEP

RI(5

) Mid

Con

tEP

RI(6

) Mid

Con

tEP

RI(7

) Mid

Con

tEP

RI(8

) Mid

Con

tEP

RI(9

) Mid

Con

tEP

RI(M

ean)

Gul

fEP

RI(1

) Gul

fEP

RI(2

) Gul

fEP

RI(3

) Gul

fEP

RI(4

) Gul

fEP

RI(5

) Gul

fEP

RI(6

) Gul

fEP

RI(7

) Gul

fEP

RI(8

) Gul

fEP

RI(9

) Gul

f

Figu

re 5

b. M

agni

tude

8 a

ttenu

atio

n cu

rve

for 2

.5 H

z fo

r the

Car

ibbe

an (r

ed c

urve

s), M

id-C

ontin

ent (

blue

cur

ves)

, and

Gul

f Coa

st(g

reen

cur

ves)

regi

on G

MPE

mod

els.

Fi

le: \

\am

ers\

corp

\eng

tech

\GH

ES\G

HES

\FPL

_Tur

keyP

tCO

L\G

S\A

ttenu

atio

n\SS

HA

C-L

evel

2-G

MPE

\Com

pare

_02p

5Hz.

xls,

Cha

rt-M

8


DRAFT

0.00

001

0.00

01

0.00

1

0.010.

11 100

1000


Hyp

ocen

tral D

ista

nce

(km

)

Car

ibbe

an A

ttenu

atio

n M

odel

s: 1

Hz,

M7

Mea

n1C

C-Q

Base

-SD

All

1CC

-QH

igh-

SDAl

l1C

C-Q

Low

-SD

All

1CV-

QBa

se-S

DAl

l1C

V-Q

Hig

h-SD

All

1CV-

QLo

w-S

DAl

l2C

-QBa

se2C

-QH

igh

2C-Q

Low

EPR

I(Mea

n) M

idCo

ntEP

RI(1

) Mid

Con

tEP

RI(2

) Mid

Con

tEP

RI(3

) Mid

Con

tEP

RI(4

) Mid

Con

tEP

RI(5

) Mid

Con

tEP

RI(6

) Mid

Con

tEP

RI(7

) Mid

Con

tEP

RI(8

) Mid

Con

tEP

RI(9

) Mid

Con

tEP

RI(M

ean)

Gul

fEP

RI(1

) Gul

fEP

RI(2

) Gul

fEP

RI(3

) Gul

fEP

RI(4

) Gul

fEP

RI(5

) Gul

fEP

RI(6

) Gul

fEP

RI(7

) Gul

fEP

RI(8

) Gul

fEP

RI(9

) Gul

f

Figu

re 6

a. M

agni

tude

7 a

ttenu

atio

n cu

rve

for 1

Hz

for t

he C

arib

bean

(red

cur

ves)

, Mid

-Con

tinen

t (bl

ue c

urve

s), a

nd G

ulf C

oast

(gre

encu

rves

) reg

ion

GM

PE m

odel

s.

File

: \\a

mer

s\co

rp\e

ngte

ch\G

HES

\GH

ES\F

PL_T

urke

yPtC

OL\

GS\

Atte

nuat

ion\

SSH

AC

-Lev

el2-

GM

PE\C

ompa

re_0

1Hz.

xls,

Cha

rt-M

7


DRAFT

0.00

001

0.00

01

0.00

1

0.010.

11 100

1000


Hyp

ocen

tral D

ista

nce

(km

)

Car

ibbe

an A

ttenu

atio

n M

odel

s: 1

Hz,

M8

Mea

n1C

C-Q

Base

-SD

All

1CC

-QH

igh-

SDAl

l1C

C-Q

Low

-SD

All

1CV-

QBa

se-S

DAl

l1C

V-Q

Hig

h-SD

All

1CV-

QLo

w-S

DAl

l2C

-QBa

se2C

-QH

igh

2C-Q

Low

EPR

I(Mea

n) M

idCo

ntEP

RI(1

) Mid

Con

tEP

RI(2

) Mid

Con

tEP

RI(3

) Mid

Con

tEP

RI(4

) Mid

Con

tEP

RI(5

) Mid

Con

tEP

RI(6

) Mid

Con

tEP

RI(7

) Mid

Con

tEP

RI(8

) Mid

Con

tEP

RI(9

) Mid

Con

tEP

RI(M

ean)

Gul

fEP

RI(1

) Gul

fEP

RI(2

) Gul

fEP

RI(3

) Gul

fEP

RI(4

) Gul

fEP

RI(5

) Gul

fEP

RI(6

) Gul

fEP

RI(7

) Gul

fEP

RI(8

) Gul

fEP

RI(9

) Gul

f

Figu

re 6

b. M

agni

tude

8 a

ttenu

atio

n cu

rve

for 1

Hz

for t

he C

arib

bean

(red

cur

ves)

, Mid

-Con

tinen

t (bl

ue c

urve

s), a

nd G

ulf C

oast

(gre

encu

rves

) reg

ion

GM

PE m

odel

s.

File

: \\a

mer

s\co

rp\e

ngte

ch\G

HES

\GH

ES\F

PL_T

urke

yPtC

OL\

GS\

Atte

nuat

ion\

SSH

AC

-Lev

el2-

GM

PE\C

ompa

re_0

1Hz.

xls,

Cha

rt-M

8


DRAFT

0.00

001

0.00

01

0.00

1

0.010.

11 100

1000


Hyp

ocen

tral D

ista

nce

(km

)

Car

ibbe

an A

ttenu

atio

n M

odel

s: 0

.5H

z, M

7

Mea

n1C

C-Q

Base

-SD

All

1CC

-QH

igh-

SDAl

l1C

C-Q

Low

-SD

All

1CV-

QBa

se-S

DAl

l1C

V-Q

Hig

h-SD

All

1CV-

QLo

w-S

DAl

l2C

-QBa

se2C

-QH

igh

2C-Q

Low

EPR

I(Mea

n) M

idCo

ntEP

RI(1

) Mid

Con

tEP

RI(2

) Mid

Con

tEP

RI(3

) Mid

Con

tEP

RI(4

) Mid

Con

tEP

RI(5

) Mid

Con

tEP

RI(6

) Mid

Con

tEP

RI(7

) Mid

Con

tEP

RI(8

) Mid

Con

tEP

RI(9

) Mid

Con

tEP

RI(M

ean)

Gul

fEP

RI(1

) Gul

fEP

RI(2

) Gul

fEP

RI(3

) Gul

fEP

RI(4

) Gul

fEP

RI(5

) Gul

fEP

RI(6

) Gul

fEP

RI(7

) Gul

fEP

RI(8

) Gul

fEP

RI(9

) Gul

f

Figu

re 7

a. M

agni

tude

7 a

ttenu

atio

n cu

rve

for 0

.5 H

z fo

r the

Car

ibbe

an (r

ed c

urve

s), M

id-C

ontin

ent (

blue

cur

ves)

, and

Gul

f Coa

st(g

reen

cur

ves)

regi

on G

MPE

mod

els.

Fi

le: \

\am

ers\

corp

\eng

tech

\GH

ES\G

HES

\FPL

_Tur

keyP

tCO

L\G

S\A

ttenu

atio

n\SS

HA

C-L

evel

2-G

MPE

\Com

pare

_00p

5Hz.

xls,

Cha

rt-M

7


DRAFT

0.00

001

0.00

01

0.00

1

0.010.

11 100

1000


Hyp

ocen

tral D

ista

nce

(km

)

Car

ibbe

an A

ttenu

atio

n M

odel

s: 0

.5H

z, M

8

Mea

n1C

C-Q

Base

-SD

All

1CC

-QH

igh-

SDAl

l1C

C-Q

Low

-SD

All

1CV-

QBa

se-S

DAl

l1C

V-Q

Hig

h-SD

All

1CV-

QLo

w-S

DAl

l2C

-QBa

se2C

-QH

igh

2C-Q

Low

EPR

I(Mea

n) M

idCo

ntEP

RI(1

) Mid

Con

tEP

RI(2

) Mid

Con

tEP

RI(3

) Mid

Con

tEP

RI(4

) Mid

Con

tEP

RI(5

) Mid

Con

tEP

RI(6

) Mid

Con

tEP

RI(7

) Mid

Con

tEP

RI(8

) Mid

Con

tEP

RI(9

) Mid

Con

tEP

RI(M

ean)

Gul

fEP

RI(1

) Gul

fEP

RI(2

) Gul

fEP

RI(3

) Gul

fEP

RI(4

) Gul

fEP

RI(5

) Gul

fEP

RI(6

) Gul

fEP

RI(7

) Gul

fEP

RI(8

) Gul

fEP

RI(9

) Gul

f

Figu

re 7

b. M

agni

tude

8 a

ttenu

atio

n cu

rve

for 0

.5 H

z fo

r the

Car

ibbe

an (r

ed c

urve

s), M

id-C

ontin

ent (

blue

cur

ves)

, and

Gul

f Coa

st(g

reen

cur

ves)

regi

on G

MPE

mod

els.

ile: \

\am

ers\

corp

\eng

tech

\GH

ES\G

HES

\FPL

_Tur

keyP

tCO

L\G

S\A

ttenu

atio

n\SS

HA

C-L

evel

2-G

MPE

\Com

pare

_00p

5Hz.

xls,

Cha

rt-M

8


DRAFT

Following the development of the 9 Caribbean GMPE models, an additional sensitivity analysis was performed to assess the applicability of the GMPE models for the modeling of ground motion attenuation for the Caribbean region. Summary results of this analysis are presented below.

The sensitivity analysis compares the GMPEs developed for the project with empirical ground motion data from five regional earthquakes that have occurred since 2004 in the Gulf of Mexico and Caribbean region. Regional broadband empirical data from selected IRIS stations were obtained, processed, and compared with the suite of Caribbean GMPE models and with both the Mid-Continent and Gulf Coast suite of EPRI (2004) GMPE regional models.

In this comparison all of the GMPE attenuation curves are for the assumed CEUS hard rock site conditions with Vs=2.83 km/sec, whereas the empirical IRIS data are for the whatever individual site conditions exist at each station, which were not obtainable at this time. Based on simplified site response analysis it would be expected that adjustments for the station-specific site conditions to the more stable CEUS hard rock site condition would result in some reduction in the observed empirical ground motion values.

The current deaggregation of seismic hazard at the project site from all sources, Cuban, Caribbean, and southeast United States, is shown in Figure 8 for longer period motions (for which the relative contribution of the larger, more distant Cuban and Caribbean sources would be expected to have their greatest contribution) and for the 10-4 and 10-5 mean annual frequency of exceedance probabilities used to develop design ground motions under current NRC regulatory guidance. The relative contribution for the higher frequency case (not shown) of 5 and 10 Hz from the more distant Caribbean sources is significantly reduced compared to the contribution to hazard from the closer local seismic sources. For this sensitivity comparison, the empirical ground motions for 1 and 2.5 Hz are presented for each of the five regional earthquakes and ground motion prediction equations.

Figure 8. Magnitude and Distance Deaggregation for 1 and 2.5 Hz at 10-4 and 10-5 Annual Frequencies of Exceedance.


DRAFT

The five regional earthquakes that have been analyzed are:

� 12/14/2004 – Caribbean Sea Region (Mw6.8) � 09/10/2006 – Gulf of Mexico (Mw5.9) � 02/04/2007 – Cuba Region (Mw6.2) � 05/28/2009 – North of Honduras (Mw7.3) � 01/12/2010 – Haiti (Mw7.0)

For each event, a standard time history processing methodology was applied with the final results being a dataset of the acceleration response spectra for a spectral damping of 5% at each station. The geometric mean of the two horizontal components was computed and comparison plots for the 1 and 2.5 Hz spectral frequencies were generated showing the empirical data and the suite of GMPE models (i.e both of the EPRI, 2004 GMPE models and the Caribbean GMPE models). Both the individual attenuation curves and the weighted mean attenuation curve for a given set are shown in the comparison plots for each spectral frequency and earthquake.

Caribbean Sea Region (Mw6.8), 12/14/2004:

This event occurred in the Caribbean Sea region and its epicenter is shown in Figure 9a along with the FPL project site location, and the location of the IRIS stations that recorded this earthquake. Based on the observed station distribution, the IRIS station DWPF located in central Florida has the most applicable source to site path for this comparison. The large majority of stations located in central America have the opposite travel path azimuth and may show different attenuation properties based on these different tectonic travel paths. The attenuation curves and empirical data are shown in Figures 9b and 9c for the 1 and 2.5 Hz spectral frequencies. The data point from the DWPF station is highlighted as a solid blue circle. Overall the empirical data falls below the median Caribbean attenuation curve (heavy red line) and has values which are in the lower distribution of range of Caribbean attenuation curves. In addition, the observation from the DWPF station is lower than the entire range of the Caribbean attenuation curves.


DRAFT

Figure 9a. Map showing the earthquake location for the 12/14/2004 Caribbean Sea Region earthquake (Mw6.8), the IRIS station locations used in the analysis and FPL project site location.


DRAFT

0.000001

0.00001

0.0001

0.001

0.01

0.1

1

100 1000 10000

Spec

tral A

ccel

erat

ion

(g)


Attenuation Models: 1Hz, 12/14/2004, M6.8Mean 1CC-QBase-SDAll 1CC-QHigh-SDAll 1CC-QLow-SDAll 1CV-QBase-SDAll






IRIS Data

Figure 9b. Comparison of Caribbean GMPE (red lines), EPRI Mid-Continent (blue lines) and Gulf Coast region (green lines) GMPE and empirical IRIS processed data for a spectral frequency of 1 Hz. The DWPF data point is shown as a blue-filled circle.

0.000001

0.00001

0.0001

0.001

0.01

0.1

1

100 1000 10000

Spec

tral A

ccel

erat

ion

(g)



Figure 9c. Comparison of Caribbean GMPE (red lines), EPRI MidContinent (blue lines) and Gulf Coast region (green lines) GMPE and empirical IRIS processed data for a spectral frequency of 2.5 Hz. The DWPF data point is shown as a blue-filled circle.


DRAFT

Gulf of Mexico (Mw5.9), 09/10/2006:

This event occurred in the Gulf of Mexico and its epicenter is shown in Figure 10a along with the FPL project site location, and the location of the IRIS stations that recorded this earthquake. Based on the observed station distribution, the IRIS station DWPF located in central Florida has the most applicable source to site path for this comparison. The distribution of stations in the central United States have a less applicable travel path azimuth and may show different attenuation properties based on these different tectonic travel paths. The attenuation curves and empirical data are shown in Figures 10b and 10c for the 1 and 2.5 Hz spectral frequencies. The data point from the DWPF station is highlighted as a solid blue circle. In general the empirical observations fall within the range of the Caribbean attenuation curves with the single except of station LRAL (i.e., at a distance of about 750km) for 2.5 Hz in which the empirical data exceeds the highest Caribbean attenuation curve. As noted earlier, the empirical observation has not been adjusted to a common site condition of Vs=2.83km/sec and it could be expected that this empirical data point could possibly be reduced if a site amplification adjustment were to be applied. It can also be concluded from the comparison plots in Figures 10b and 10c, that the distribution of the current Caribbean attenuation curves adequately captures the range of the empirical data. In addition, the observation from the DWPF station is in the lower range of the Caribbean attenuation curves.

Figure 10a. Map showing the earthquake location for the 09/10/2006 Gulf of Mexico earthquake (Mw5.9), the IRIS station locations used in the analysis and FPL project site location.


DRAFT

0.000001

0.00001

0.0001

0.001

0.01

0.1

1

100 1000 10000

Spec

tral A

ccel

erat

ion

(g)




0.000001

0.00001

0.0001

0.001

0.01

0.1

1

100 1000 10000

Spec

tral A

ccel

erat

ion

(g)



Figure 10c. Comparison of Caribbean GMPE (red lines), EPRI Mid-Continent (blue lines) and Gulf Coast region (green lines) GMPE and empirical IRIS processed data for a spectral frequency of 2.5 Hz. The DWPF data point is shown as a blue-filled circle.


DRAFT

Cuba Region (Mw6.2), 02/04/2007:

This event occurred south of the island of Cuba and its epicenter is shown in Figure 11a along with the FPL project site location, and the location of the IRIS stations that recorded this earthquake. Only three IRIS stations were analyzed from this. Based on the observed station distribution, the IRIS station DWPF located in central Florida has the most applicable source to site path for this comparison. The attenuation curves and empirical data are shown in Figures 11b and 11c for the 1 and 2.5 Hz spectral frequencies. The data point from the DWPF station is highlighted as a solid blue circle. Overall the empirical data falls below the median Caribbean attenuation curve (heavy red line) and has values which are in the lower distribution of range of Caribbean attenuation curves. In addition, the observation from the DWPF station is similar to the lowest Caribbean attenuation curve or lower.

Figure 11a. Map showing the earthquake location for the 02/04/2007 Cuba Region earthquake (Mw6.2), the IRIS station locations used in the analysis and FPL project site location.


DRAFT

0.000001

0.00001

0.0001

0.001

0.01

0.1

1

100 1000 10000

Spec

tral A

ccel

erat

ion

(g)


Caribbean Attenuation Models: 1Hz, 02/04/2007, M6.2Mean 1CC-QBase-SDAll 1CC-QHigh-SDAll 1CC-QLow-SDAll 1CV-QBase-SDAll1CV-QHigh-SDAll 1CV-QLow-SDAll 2C-QBase 2C-QHigh 2C-QLowEPRI(Mean) MidCont EPRI(1) MidCont EPRI(2) MidCont EPRI(3) MidCont EPRI(4) MidContEPRI(5) MidCont EPRI(6) MidCont EPRI(7) MidCont EPRI(8) MidCont EPRI(9) MidContEPRI(Mean) Gulf EPRI(1) Gulf EPRI(2) Gulf EPRI(3) Gulf EPRI(4) GulfEPRI(5) Gulf EPRI(6) Gulf EPRI(7) Gulf EPRI(8) Gulf EPRI(9) GulfIRIS Data


0.000001

0.00001

0.0001

0.001

0.01

0.1

1

100 1000 10000

Spec

tral A

ccel

erat

ion

(g)


Caribbean Attenuation Models: 2.5Hz, 02/04/2007, M6.2Mean 1CC-QBase-SDAll 1CC-QHigh-SDAll 1CC-QLow-SDAll 1CV-QBase-SDAll1CV-QHigh-SDAll 1CV-QLow-SDAll 2C-QBase 2C-QHigh 2C-QLowEPRI(Mean) MidCont EPRI(1) MidCont EPRI(2) MidCont EPRI(3) MidCont EPRI(4) MidContEPRI(5) MidCont EPRI(6) MidCont EPRI(7) MidCont EPRI(8) MidCont EPRI(9) MidContEPRI(Mean) Gulf EPRI(1) Gulf EPRI(2) Gulf EPRI(3) Gulf EPRI(4) GulfEPRI(5) Gulf EPRI(6) Gulf EPRI(7) Gulf EPRI(8) Gulf EPRI(9) GulfIRIS Data

Figure 11c. Comparison of Caribbean GMPE (red lines), EPRI Mid-Continent (blue lines) and Gulf Coast region (green lines) GMPE and empirical IRIS processed data for a spectral frequency of 2.5 Hz. The DWPF data point is shown as a blue-filled circle.


DRAFT

North of Honduras (Mw7.3), 05/28/2009:

This event occurred north of Honduras in Central America and is the largest earthquake in the suite of five events analyzed in this sensitivity study. The location of its epicenter is shown in Figure 12a along with the FPL project site location, and the location of the three IRIS stations that were analyzed. Based on the azimuths from the earthquake to the three IRIS stations, none of the associated seismic ray travel paths are ideal for this comparison study. The attenuation curves and empirical data are shown in Figures 12b and 12c for the 1 and 2.5 Hz spectral frequencies. The comparisons provided in the figures indicate that the empirical data from this earthquake are lower than any of the Caribbean attenuation curves.

Figure 12a. Map showing the earthquake location for the 5/28/2009 Caribbean Sea Region earthquake (Mw7.3), the IRIS station locations used in the analysis and FPL project site location.


DRAFT

0.000001

0.00001

0.0001

0.001

0.01

0.1

1

100 1000 10000

Spec

tral A

ccel

erat

ion

(g)



Figure 12b. Comparison of Caribbean GMPE (red lines), EPRI Mid-Continent (blue lines) and Gulf Coast region (green lines) GMPE and empirical IRIS processed data for a spectral frequency of 1 Hz.

0.000001

0.00001

0.0001

0.001

0.01

0.1

1

100 1000 10000

Spec

tral A

ccel

erat

ion

(g)


Caribbean Attenuation Models: 2.5Hz, 05/28/2009, M7.3Mean 1CC-QBase-SDAll 1CC-QHigh-SDAll 1CC-QLow-SDAll 1CV-QBase-SDAll 1CV-QHigh-SDAll1CV-QLow-SDAll 2C-QBase 2C-QHigh 2C-QLow EPRI(Mean) MidCont EPRI(1) MidContEPRI(2) MidCont EPRI(3) MidCont EPRI(4) MidCont EPRI(5) MidCont EPRI(6) MidCont EPRI(7) MidContEPRI(8) MidCont EPRI(9) MidCont EPRI(Mean) Gulf EPRI(1) Gulf EPRI(2) Gulf EPRI(3) GulfEPRI(4) Gulf EPRI(5) Gulf EPRI(6) Gulf EPRI(7) Gulf EPRI(8) Gulf EPRI(9) GulfIRIS Data

Figure 12c. Comparison of Caribbean GMPE (red lines), EPRI Mid-Continent (blue lines) and Gulf Coast region (green lines) GMPE and empirical IRIS processed data for a spectral frequency of 2.5 Hz.


DRAFT

Haiti (Mw7.0), 01/12/2010:

The Haiti earthquake is the most recent event in this suite of five earthquakes analyzed. The epicentral location and the suite of IRIS stations analyzed are shown in Figure 13a. Based on the observed station distribution, the IRIS station OTAV located in between the south part of Florida and the island of Cuba has the most applicable source to site path for this comparison. Note that the data from the DWPF was not available for this earthquake. The attenuation curves and empirical data are shown in Figures 13b and 13c for the 1 and 2.5 Hz spectral frequencies. The data point from the OTAV station is highlighted as a solid blue circle. Overall the empirical data falls in the lower range or lower than the Caribbean attenuation curves. In addition, the observation from the OTAV station is significantly lower than the entire range of the Caribbean attenuation curves.

Figure 13a. Map showing the earthquake location for the 01/12/2010 Haiti earthquake (Mw7.0), the IRIS station locations used in the analysis and FPL project site location.


DRAFT

0.00001

0.0001

0.001

0.01

0.1

1

100 1000 10000

Spec

tral A

ccel

erat

ion

(g)


01/12/2010 Earthquake: 1Hz, 01/12/2010, M7Mean1CC-QBase-SDAll1CC-QHigh-SDAll1CC-QLow-SDAll1CV-QBase-SDAll1CV-QHigh-SDAll1CV-QLow-SDAll2C-QBase2C-QHigh2C-QLowEPRI(Mean) MidContEPRI(1) MidContEPRI(2) MidContEPRI(3) MidContEPRI(4) MidContEPRI(5) MidContEPRI(6) MidContEPRI(7) MidContEPRI(8) MidContEPRI(9) MidContEPRI(Mean) GulfEPRI(1) GulfEPRI(2) GulfEPRI(3) GulfEPRI(4) GulfEPRI(5) GulfEPRI(6) GulfEPRI(7) GulfEPRI(8) GulfEPRI(9) GulfIRIS Data

Figure 13b. Comparison of Caribbean GMPE (red lines), EPRI Mid-Continent (blue lines) and Gulf Coast region (green lines) GMPE and empirical IRIS processed data for a spectral frequency of 1 Hz. The OTAV data point is shown as a blue-filled circle.

0.00001

0.0001

0.001

0.01

0.1

1

100 1000 10000

Spec

tral A

ccel

erat

ion

(g)


01/12/2010 Earthquake: 2.5Hz, 01/12/2010, M7Mean1CC-QBase-SDAll1CC-QHigh-SDAll1CC-QLow-SDAll1CV-QBase-SDAll1CV-QHigh-SDAll1CV-QLow-SDAll2C-QBase2C-QHigh2C-QLowEPRI(Mean) MidContEPRI(1) MidContEPRI(2) MidContEPRI(3) MidContEPRI(4) MidContEPRI(5) MidContEPRI(6) MidContEPRI(7) MidContEPRI(8) MidContEPRI(9) MidContEPRI(Mean) GulfEPRI(1) GulfEPRI(2) GulfEPRI(3) GulfEPRI(4) GulfEPRI(5) GulfEPRI(6) GulfEPRI(7) GulfEPRI(8) GulfEPRI(9) GulfIRIS Data

Figure 13c. Comparison of Caribbean GMPE (red lines), EPRI Mid-Continent (blue lines) and Gulf Coast region (green lines) GMPE and empirical IRIS processed data for a spectral frequency of 2.5 Hz. The OTAV data point is shown as a blue-filled circle.


DRAFT

Based on this sensitivity analysis, we conclude that the suite of Caribbean GMPE models used in the current PSHA predicts larger ground motions on average than the observed empirical data from these five earthquakes for the spectral frequencies of 1 and 2.5 Hz. In addition, given the subset of data from just those stations which have a more appropriate source to path seismic ray path this conclusion can be extended to state that the use of the Caribbean GMPE models should provide conservative (i.e., higher) ground motion predictions compared to the observed empirical data from the region.

The sensitivity analysis indicates to us that the suite of GMPEs developed for the FPL FSAR to characterize contributions to hazard at the southern Florida site from Caribbean earthquakes is more like the EPRI Gulf Coast GMPEs than the EPRI Mid-Continent GMPEs and that the contribution to design ground motions from Caribbean earthquakes of a given magnitude and distance is conservatively represented by incorporation of the developed project-specific GMPEs.


DRAFT

References

Atkinson, G. M. and D. M. Boore (1995). “New ground motion relations for eastern North America,” Bull. Seism. Soc. Am., Vol. 85, No. 1, p. 17-30.

Boore, D. M. (2005). “SMSIM – Fortran Programs for Simulating Ground Motions from Earthquakes Version 2.3 – A Revision of OFR 96-80-A”, USGS Open File Report 00-509.

Chen, S. Z. and G. M. Atkinson (2002). “Global Comparisons of Earthquake Source Spectra,” Bull. Seis. Soc. Am., Vol. 92, No. 3, p. 885-895.

Electric Power Research Institute [EPRI] (1993) “Guidelines for determining design basis ground motions,” EPRI TR-102293, Vol. 1-5, Palo Alto, California.

Electric Power Research Institute [EPRI] (2004) CEUS Ground Motion Project Final Report, Elec. Power Res. Inst, Technical Report 1009684, dated December 2004.

Motazedian, D. and G. Atkinson (2005). Ground-Motion relations for Puerto Rico. Geological Soc. America, Special Paper 385, p. 61-80.

Silva, W., N. Gregor, and R. Darragh (2003). Development Of Regional Hard Rock Attenuation Relations For Central And Eastern North America, Mid-Continent And Gulf Coast Areas. Pacific Engineering, http://www.pacificengineering.org/, dated August 13, 2003.


DRAFT

Questions

1. Are the newly developed Caribbean GMPEs acceptable for modeling ground motions in the Caribbean region for the project site located in Southern Florida for the PSHA study?

2. Are the selected input parameters acceptable for their use? 3. Is the recommended sigma value (i.e., 0.645 natural log units) acceptable? 4. Is the recommended weighting based on the EPRI (2004) class weights acceptable? 5. Are there any other GMPE models which should be considered for the PSHA contribution

from Caribbean sources? 6. Is either the Mid-Continent or Gulf Coast region GMPE from EPRI (2004) more

appropriate for use in the evaluation of the contribution to PSHA at a site in southern Florida from Caribbean seismic sources?

7. Do you have any other comments?


DRAFT

1

Gregor, Nicholas

From: Litehiser, JoeSent: Tuesday, October 30, 2012 8:18 AMTo: norman abrahamsonCc: Gregor, Nicholas; Tavakoli, BehroozSubject: FPL Questionnaire.Attachments: CaribbeanGMPE_Questionaire.pdf

Norm�–�My�brief�take�away�from�our�conversation�last�Friday:�I�heard�that�you�believed��that�Nick’s��IRIS�empirical�data�was�the�most�convincing�and�relevant�part�of�our�argument�overriding�any�concern�or�uncertainty�about�model�parameters.��Further,�you�said�that�uncertainty�about�subsurface�conditions�on�IRIS�results�would�not�amount�to�more�than�a�factor�of�about�two�–�small�relative�to�other�uncertainties.��Finally,�you��said�that�the�earthquakes�we�used�for�IRIS�data�are�all�shallow�crustal�earthquakes�and�that�this�point�needs�to�be�emphasized�in�our�response.��Given�this�and�the�seven�questions�below,�I�hope�to�call�you�at�7�PM�Eastern�time�this�evening.��If�there�is�anything,�however�brief,�you�can�sketch�in�for�the�specific�questions�that�would�be�very�helpful.��You�have�the�original�questionnaire�but�here�it�is�again�for�your�convenience.��Thanks��Joe�

�Question�1:�Are�the�newly�developed�Caribbean�GMPEs�acceptable�for�modeling�ground�motions�in�the�Caribbean�

region�for�the�project�site�located�in�Southern�Florida�for�the�PSHA�study?�N.�Abrahamson�

Response� �Notes� �Action�Item� �

Question�2:�Are�the�selected�input�parameters�acceptable�for�their�use?�N.�Abrahamson�


Question�3:�Is�the�recommended�sigma�value�(i.e.,�0.645�natural�log�units)�acceptable?�N.�Abrahamson�

Response� �Notes� �Action�Item� �Action�Item� �

Question�4:�Is�the�recommended�weighting�based�on�the�EPRI�(2004)�class�weights�acceptable?�Question�5:�Are�there�any�other�GMPE�models�which�should�be�considered�for�the�PSHA�contribution�from�Caribbean�

sources?�N.�Abrahamson�


Question�6:�Is�either�the�Mid�Continent�or�Gulf�Coast�region�GMPE�from�EPRI�(2004)�more�appropriate�for�use�in�the�evaluation�of�the�contribution�to�the�PSHA�at�a�site�in�southern�Florida�from�Caribbean�seismic�sources?�

N.�Abrahamson�


Question�7:�Do�you�have�any�other�comments?�N.�Abrahamson�


�


DRAFT

1��

Review�of�Bechtel�Caribbean�GMPE�Report�

By:�Yousef�Bozorgnia�

August�2,�2012�

I�reviewed�the�Bechtel�report�on�development�of�the�Caribbean�GMPEs�for�use�for�Southern�Florida.�Below�are�the�specific�questions�stated�in�the�report�and�my�reply.�

1.�Are�the�newly�developed�Caribbean�GMPEs�acceptable�for�modeling�ground�motions�in�the�Caribbean�region�for�the�project�site�located�in�Southern�Florida�for�the�PSHA�study?��Based�on�the�review�of�the�materials�submitted�by�Bechtel,�the�newly�developed�GMPEs�are�acceptable,�and�the�process�of�developing�the�GMPEs�is�rational�and�logical.��2.�Are�the�selected�input�parameters�acceptable�for�their�use?��Yes,�especially�considering�the�work�of�Motazedian�and�Atkinson�(2005).�Also,�using�Kappa=0.006�sec�for�the�reference�rock�is�consistent�with�the�latest�preliminary�findings�of�NGA�East�project.��A�word�of�caution:�Chen�and�Atkinson�(2002)�have�a�table�of�site�amplifications�for�CEUS�Hard�Rock�that�lists�both�Amp(f)��and��Amp(f)*exp(�pi*Kappa*f).�For�the�CEUS�Hard�Rock,�Chen�and�Atkinson�used�Kappa=0.002�sec.�If�in�the�Caribbean�study,�values�of��Amp(f)*exp(�pi*Kappa*f)�were�directly�taken�from�Chen�and�Atkinson�(2002),�theses�values�should�be�adjusted�for�Kappa=0.006�sec.��3.�Is�the�recommended�sigma�value�(i.e.,�0.645�natural�log�units)�acceptable?��This�value�is�based�on�Motazedian�and�Atkinson�(2005)�that�stated:�“The�standard�deviation�of�the�distribution�of�the�residuals�for�intermediate�frequencies�on�average�is�~0.28�in�log10�units.”�(which�is�equivalent�of�0.645�on�natural�log�units�).�As�stated�in�Motazedian�and�Atkinson�(2005),�this�sigma�is�an�average�standard�deviation�for�intermediate�frequencies.�On�average�sense,�this�number�is�a�reasonable�number;�however,�the�standard�deviation�of�GMPEs�can�be�a�function�of�frequency.�A�simple�approximate�sensitivity�analysis�can�be�performed�to�estimate�a�frequency�dependent�sigma�and�its�impact�on�the�results.��4.�Is�the�recommended�weighting�based�on�the�EPRI�(2004)�class�weights�acceptable?��The�explanation�on�the�assigned�weights�in�Table�3�is�reasonable.��5.�Are�there�any�other�GMPE�models�which�should�be�considered�for�the�PSHA�contribution�from�Caribbean�sources?��Not�that�I�know�of.��


DRAFT

2��

6.�Is�either�the�Mid�Continent�or�Gulf�Coast�region�GMPE�from�EPRI�(2004)�more�appropriate�for�use�in�the�evaluation�of�the�contribution�to�PSHA�at�a�site�in�southern�Florida�from�Caribbean�seismic�sources?��Based�on�the�sensitivity�analysis�of�five�regional�events�reported�in�the�Bechtel�report,�it�seems�that�EPRI�Gulf�Coast�GMPEs�are�more�appropriate.�However,�it�should�be�noted�that�the�recorded�empirical�data�were�not�corrected�for�local�site�effects�to�correspond�the�spectral�ordinates�of�the�recorded�motions�to�Vs=2.83�km/sec.��

7.�Do�you�have�any�other�comments?��

(a)�In�the�sensitivity�analysis,�the�site�corrections�at�the�recording�sites�to�correspond�the�spectral�ordinates�to�Vs=2.83�km/sec�should�be�carried�out.�Due�to�lack�of�specific�soil�date�at�the�recording�site,�the�site�correction�needs�to�be�done�in�an�approximation�manner.�There�are�different�methods�of�approximating�site�correction�that�may�be�employed.�

(b)�In�the�sensitivity�analysis,�it�would�be�useful�to�show�the�residual�plots�rather�than�the�predicted�curves�superimposed�over�the�recorded�data.�In�this�process,�it�would�be�also�useful�to�estimate�the�“event�terms”�(and�their�uncertainties)�for�the�earthquakes�examined�in�the�report.�

�

�

�


DRAFT

Response to SSHAC Caribbean GMPE QuestionnaireKenneth W. Campbell, Ph.D.

Kenneth W Campbell Consulting

QUESTIONS AND ANSWERS:

Q1. Are the newly developed Caribbean GMPEs acceptable for modeling ground motions in the Caribbean region for the project site located in Southern Florida for the PSHA study?A1. Yes, it is my opinion that the newly developed Caribbean GMPEs are generally acceptable, albeit possibly conservative at very large distances, for modeling ground motions in the Caribbean region for a site located in Southern Florida for the PSHA study. However, without seeing the functional form, it is not possible to comment on the model itself, which appears to be a very simple functional form, possibly lacking the “hinged-trilinear” geometric attenuation termthat was included in the stochastic simulations. This term could be important at large distances. Only a plot of residuals would indicate whether this apparent simple functional form is appropriate. If the term was included in the GMPEs, then this statement can be ignored. Because of the large degree of epistemic uncertainty in the model parameters, travel path, and types of sources (tectonic regime), it would be better if uncertainty was applied to all of the model parameters and not just Q and source spectra type (e.g., “kappa,” in which EPRI (1993) included three alternative values for this parameter). I am also somewhat confused regarding how uncertainty in stress drop was treated. It appears that it was included as aleatory modelinguncertainty based on the description in the text (which is not used in capturing aleatory uncertainty in the model), instead of being modeled as epistemic uncertainty. If so, perhaps it would be better to include it as epistemic uncertainty to be consistent with my previous comment.

Q2. Are the selected input parameters acceptable for their use?A2. Yes, I am not aware of any other parameters that might be more acceptable for use in deriving a GMPE for the Caribbean region. However, as mentioned above, it would be more appropriate to include uncertainty in more of these parameters considering the large degree of uncertainty involved in their estimation and current application to a site in southern Florida. Iam a bit concerned that Motazedian and Atkinson (2005) used a stress drop from a calibration with California earthquakes and not derived from the data. However, the generally good comparison with the Puerto Rico data would seem to indicate that their 130-bar median value is generally acceptable. The assumption by Motazedian and Atkinson (2005) that site response at site SJG, which appears to represent a shallow soft-rock site, is the same for all of the recording sites in Puerto Rico, is of some concern, since systematic differences in site response could “bleed” into other parameters. However, there is no way of knowing if this is the case and this possibility must instead be captured as epistemic uncertainty


DRAFT

Q3. Is the recommended sigma value (i.e., 0.645 natural log units) acceptable?A3. Yes, I believe that the recommended sigma value of 0.645 natural log units (0.28 common log units) is acceptable as long as sufficient epistemic uncertainty is captured in the model.

Q4. Is the recommended weighting based on the EPRI (2004) class weights acceptable?A4. Yes, the weighting scheme based on class weights seems reasonable. There should be an explanation why the third class of models defined by EPRI (hybrid empirical models) was not considered. It is simply stated as a fact without explanation.

Q5. Are there any other GMPE models which should be considered for the PSHA contribution from Caribbean sources?A5. None of the earthquakes used to compare recorded ground motions with the Caribbean GMPEs developed in this study appeared to be subduction earthquakes. However, most recordings from subduction earthquakes represent continental travel paths and, at large distances, paths in the backarc region, which have been shown to show greater attenuation than paths in the forearc region. Therefore, subduction GMPEs are not appropriate for the primarily oceanic travel paths to the southern Florida site. To show that this is the case, it would be beneficial to compare predictions from the current set of empirical subduction GMPEs with the Caribbean GMPEs developed in this study for large earthquakes at distances of relevance to the southern Florida site to show that these subduction GMPEs would predict smaller ground motions. However, if they don’t, then they should be considered for predicting ground motions from subduction sources.

Q6. Is either the Mid-Continent or Gulf Coast region GMPE from EPRI (2004) more appropriate for use in the evaluation of the contribution to PSHA at a site in southern Florida from Caribbean seismic sources?A6. Although the limited amount of data appears to suggest that the Gulf Coast region GMPE developed by EPRI (2004) might be more appropriate at large distances, the one earthquake that occurred in the Gulf of Mexico (the closest earthquake to the site) showed relatively good agreement with the Caribbean GMPE developed in this study (ignoring differences in site response). Therefore, I would not suggest that the Gulf Coast region GMPE is more appropriate, but instead suggest that it be captured in the epistemic uncertainty model with by using it as an alternative model or insuring that the epistemic uncertainty in the model parameters include it. I would not recommend that the Mid-Continent region GMPE of EPRI (2004) be used for Caribbean earthquake sources.

Q7. Do you have any other comments?A7a. The Puerto Rico data used by Motazedian and Atkinson (2005) to estimate the model parameters were recorded at distances of 500 km and less. Therefore, use of these parameters


DRAFT

beyond 500 km represents an extrapolation and must be adequately captured as epistemic uncertainty. Presumably, this is accomplished by included epistemic uncertainty in the Q term.Since 8 years has passed since the Motazedian and Atkinson study, there is no doubt many more recordings that could be used to update the model. It might be worthwhile to contact Gail Atkinson and Dariush Motazedian to see if they have any more information or would suggest the use of different parameters from their 2005 study.

A7b. The terms “GMPE” and “attenuation curve” are used interchangeably. This latter term is no longer considered to be acceptable in the engineering seismology community, except when describing actual attenuation properties, such as Q and geometrical spreading. I would suggest that either the terms “GMPE,” “distance scaling” or “median prediction” be used as appropriate when describing the distance-scaling characteristics of the GMPE model.

A7c. The statement on Page 18 that “Based on simplified site response analysis it would be expected that adjustments for the station-specific site conditions to the more stable CEUS hard rock site condition would result in some reduction in the observed empirical ground motion values” might be true for relatively low frequencies but might not be true for relatively highfrequencies, where a very small kappa at the southern Florida site might actually result in higher ground motion.

A7d. The deaggregation histograms show that the Caribbean sources are important to (and possibly even dominating) the seismic hazard at relatively low structural and exceedance frequencies. How about at higher structural frequencies? For completeness, I would like to seedeaggregation plots at the higher structural frequencies.


DRAFT

1

SSHAC Caribbean Ground Motion Prediction Equation Questionnaire

RESPONSE TO QUESTIONS By

Shahram Pezeshk Question 1. Are the newly developed Caribbean GMPEs acceptable for modeling ground motions in the Caribbean region for the project site located in Southern Florida for the PSHA study? My general comment is that the newly developed Caribbean GMPEs are acceptable and possibly conservative for modeling ground motions in the Caribbean region. I will elaborate on my answer in much detail through answering the remaining questions. Question 2. Are the selected input parameters acceptable for their use? The point-source stochastic method will be a reasonable approach in concept to simulate ground motions as a function of earthquake magnitudes and distances, if the selected input seismological parameters for the simulations are validated using available instrumental seismic data, and are compared to ground-motion relations for other tectonically similar regions. Because of its success and simplicity, the point-source stochastic method is now widely used to predict ground motion in many regions of the world where the number of strong ground-motion recordings is limited and no empirical ground-motion relations are available. This approach has been successfully validated and applied for other regions such as western United States (WUS) and central and eastern United States (CEUS), and on average reproduces similar empirical ground-motion attenuation relationships that can be obtained by direct use of ground-motion recordings. In the Caribbean-Cuba region, where the number of ground-motion recordings is limited, and except for the Motazedian and Atkinson [MA05] (2005) relation, there are no empirical ground-motion relations available, it is possible to derive simple source models (e.g., 1CC, 1CV, 2C) and to evaluate the variation of seismological parameters (e.g., stress parameter, kappa, and Q factor) using the seismological network data. These parameters are used to scale ground-motion amplitudes with the earthquake magnitude and the source-to-site distance. However, the empirical ground-motion relations estimated from the simulated records should be compared to those calculated from the actual seismographic records in the region, and the applicability of the ground motion prediction equations (GMPEs) should be assessed by the data from other tectonically similar regions. When I started reviewing the March 5, 2012 report, I was going to recommend that the resulting GMPEs to be superimposed with some selected actual earthquakes in the Cuba and Caribbean region that were recorded by the broadband CEUS stations, Global Seismic Network (GSN), IRIS broadband data, Regional Seismic Network, and other cooperative stations in the Caribbean


DRAFT

2

region (see McNamara et al., 2012). Now, I am glad that this has been done in the most recent report that has been provided to me (the SSHAC Caribbean GMPE Questionnaire). Therefore, the comparison of the developed Caribbean GMPEs with the EPRI (2004) Mid-Continent and Gulf Coast GMPEs has provided a good way to show the appropriate ranges of input parameters such as stress parameter that scales high frequency of ground motions have been selected. The Gulf of Mexico (Mw 5.9) event is possibly a good representative of ground motions between the Cuba area and the FPL project site location that could be used to validate the selected parameters. This event shows that the selected seismological parameters are acceptable to capture the recording ground motion data (see Figures 10a and 10b). Source models and stress parameters McNamara et al. (2012) have provided valuable broadband dataset from the earthquakes in the Caribbean region (table of earthquakes and station parameters) to determine the Q-factor relations in the region. The McNarama et al. (2012) study, which has been published after the development of the Caribbean GMPEs, sheds lights on some of the modeling parameters. Boore et al. (2009) have also provided a procedure for estimating stress parameters using an updated dataset and a revised point source stochastic model. It should be noted that the stress parameter is strongly dependent on the source model used. For instance, Atkinson et al. (2009) and Boore (2009) have explained why a stress parameter of about 250 bars is needed in SMSIM [point-source simulation] to approximately match the predictions of Atkinson and Boore (2006), which is based on EXSIM [finite-fault simulation] with 140 bars, when all other model parameters are set to be equal (Campbell, 2008). The MA05 article has used a source model corresponding to an input stress parameter of 130 bars for the Puerto Rico region based on the calibrated source parameters for EXSIM. This is the EXSIM source model and the stress parameter of 130 bars may be modified based on the SMSIM source model to match the predictions of Atkinson and Silva (2000) in their stochastic ground-motion model. The earthquake ground motions in the Caribbean region is similar to the WUS region, and attenuates faster with epicentral distance than comparable-sized earthquakes in the CEUS region (see Q factor below). The stress parameters in the Caribbean region can be calibrated by the California strong-motion dataset or by the new generation of ground-motion attenuations in WUS. As a result, the stress parameters used in the Caribbean region may not capture the appropriate changes in the stress parameter. However, it is difficult to confirm them with the insufficient information provided in this report. According to Atkinson and Boore (2006), it should be noted that the effect of the stress parameter on the simulated ground-motion values is approximately independent of distance. The frequency range over which the increase in amplitudes will occur depends on earthquake magnitude (Atkinson and Boore, 2006). Increasing the stress parameter has almost no effect on amplitudes at low frequencies (long periods). Then, the amplitude will increase and reach a constant factor at high frequencies (see Aleatory and Epistemic Uncertainties).


DRAFT

3

Q factor The instrumental seismic recordings in the different tectonic environments typically show that areas of active tectonics, like the Caribbean region and the WUS region, have higher attenuation of seismic waves than the stable continental regions such as the CEUS region (Aki, 1980; Singh and Herrmann, 1983; Frankel et al., 1990; Benz et al., 1997; Erikson et al., 2004; McNamara et al., 2012). According to the McNamara et al. (2012) article, possible reasons to explain these observations in notably different tectonic environments are a highly fractured crust in tectonically active regions that effectively absorb high-frequency seismic waves, differences in crustal temperature, and variations in crustal structures that control elastic wave propagation. Therefore, the earthquake ground motions in the Caribbean region- similar to the WUS region- are attenuated faster with distance than comparable-sized earthquakes in the CEUS region. In fact, it has been observed throughout the world that Lg-phase attenuation (regional surface-wave phase observed as the dominant phase on broadband seismograms at regional distances) is higher for regions with active tectonic than for stable continental interiors (McNamara et al. 2012). Recent studies in the Caribbean region using digital broadband seismic instruments show that both direct S and Lg-phase attenuation results are most similar to the results obtained for the WUS region. Figure 1 (similar to Figure 9 of McNamara et al., 2012) illustrates the Q models in the Caribbean region and the adjacent region ranging from Jamaica and Cuba in the west to Puerto Rico and Lesser Antilles in the east compared to those in the WUS and CEUS regions. The results indicate that the local attenuation properties of the Caribbean region are consistent with values estimated for the WUS region, which is likely to be characterized by a lower Q factor (higher attenuation) than the CEUS region. Hough and Anderson (1988) as well as McNamara et al. (2012) have pointed out that the attenuation properties of Lg-phase differ from those of the direct S wave because the Lg-phase samples the entire crust including the “deeper crust,” which is likely to be characterized by the lower attenuation (higher Q factor), while the direct S waves are more controlled by the “upper crust” (lower Q factor). Therefore, the Q-factor model developed by MA05 for the Puerto Rico region using local earthquakes as deep as 200 km might give higher Q factor (lower attenuation) than the shallow crustal attenuation models in the Caribbean region. As a result, the MA05 Q-factor model and the range of uncertainties for the simulation of ground motions in the Caribbean region implies that Q factor adopted for the Caribbean crust has captured conservatively the epistemic uncertainty in Q factor. Note that the focal mechanism and the depth of actual earthquakes in the Caribbean region are significant, especially for their use to validate the Caribbean GMPEs. This information should be addressed in the report.


DRAFT

4

Figure 1. Comparison of Q-factors for three different tectonic regions; 1)

Caribbean, 2) CEUS, and 3) WUS regions.

Kappa Filter The effect of kappa on ground motions is negligible for hard-rock sites at large distances, since high frequency amplitudes are reduced through the kappa operation and are filtered at large distances. Figure 2 shows the comparison between Q-factor and kappa of 0.006 at a distance of about 100 km for the Puerto Rico area in the Caribbean region. Using the site attenuation of 0.006 (kappa), which models the near surface attenuation, there is no significant effect on the ground-motion scaling at large distances for hard-rock site compared to Q factor.

10

100

1000

10000

0.1 1 10 100

Q fa

ctor

Frequency (Hz)

Comparison of Q factors

CARIB1 CARIB2 CARIB3 CEUS1 CEUS2 WUS1 WUS2 WUS3 WUS4


DRAFT

5

Geometrical spreading In developing attenuation models, the Q factor is often determined by assuming a hinged-trilinear functional form for the geometrical spreading, in which amplitude decay varies as a function of epicentral distance. This is interpreted as reflecting the seismic waveform in the different layers of crustal structure (Atkinson and Mereu, 1992; Atkinson and Boore, 1995). At large distances, the crustal waveguide Lg-phase control seismic records, and the attenuation rate changes to R-0.5 form (Frankel, 1991; McNamara et al., 1996; Benz et al., 1997; McNamara et al., 2012). This fixed exponent of -0.5 has a main role to scale ground-motion amplitudes at large distances. MA05 and McNamara et al. (2012) found the same distance-dependent geometrical spreading and adopted a hinged-trilinear functional form specific to Caribbean region. McNamara et al. (2012) have observed that beyond 200–250 km, body waves consist of small amplitude mantle head waves and crustal Lg dominates the ground-motion records. Based on these observations, they have confirmed the attenuation rate is changed by R-0.5 at distances of 100 km and greater in the Caribbean region compared to the 130 km in the CEUS region. This difference in hinge point at large distances reflects regional differences in the crustal structures between two regions. Therefore, using a constant rate of -0.5, the ground-motion attenuations are strongly controlled not only by region-specific Q-factor relations at large distances, but also by the hinge points (Figure 3).

�Figure 2. Comparison between Q-factor and kappa relations at a distance of

about 100 km for the Puerto Rico area in the Caribbean region.

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0.1 1 10 100

Am

plitu

de

Frequancy (Hz)

Comparison between Q factor and Kappa

PR-Q factor Kappa Attenuation


DRAFT

6

Figure 3. Comparison between Q-factor and Geometrical spreading at frequency of 1Hz and distances between 100 km and 2000 km for the Caribbean region.

Comparison between the attenuation model (Q-factor and Geometrical spreading) and the resulting Caribbean GMPEs at frequency of 1 Hz (see Figures 1 through 7 of the SSHAC Caribbean GMPE Questionnaire) indicates that the Q-factor relations have strongly controlled the ground-motion amplitudes at large distances. Therefore, selection of the appropriate regional-specific Q-factor plays the main role to scale ground-motion amplitudes at large distances in the Caribbean region. As it was discussed earlier and as it is plotted in Figure 1, the Q factors have been selected conservatively. This may conclude that the observed earthquake ground-motion data in the Caribbean region should be located lower than the mean Caribbean GMPEs for the same site conditions - especially at large distances. Data provided in Figures 9b, 9c, 12b, 12c, 13b, and 13c of the SSHAC Caribbean GMPE Questionnaire show that in fact the observed earthquake ground-motion data in the Caribbean region is lower than the mean Caribbean GMPEs. It is also important to note that the observed ground motion data have not been modified for the site amplification.

0.000001

0.00001

0.0001

0.001

0.01

0.1

1

10 100 1000 10000

Am

plitu

de

Distance (km)

Comparison between Q factors and Geometrical Spreading

Qo=241 Qo=359 Qo=536


DRAFT

7

Question 3. Is the recommended sigma value (i.e., 0.645 natural log units) acceptable? Aleatory Uncertainties Atkinson and Boore (2006) has pointed out that the aleatory uncertainty due to the natural random variability in earthquake source, path and site effects is independent of magnitude and distance, with an average value of 0.30 log units for all frequencies in CEUS. This calculated variability, based on the simulation parameters, is slightly larger than typically observed values for the GMPEs in California (e.g., Boore et al., 1997; Abrahamson and Silva, 1997). The MA05 study has also calculated the residuals of the recorded Puerto Rico data versus the Puerto Rico ground motion attenuation relations, and has suggested a standard deviation of the relations of 0.28 log(10) units for all frequencies. This standard deviation is typical for ground-motion relations in many regions and should be relatively independent of region. Comparison of the variability of ground motions between eastern North America (ENA) and the California or Caribbean region show that the simulated variability may slightly overestimate the actual variability. However, the variability issue will require further earthquake data in ENA. In response to question 3, the aleatory uncertainty with an average value of 0.645 in natural log units is reasonable to capture the uncertainties in input seismological parameters discussed above for all frequencies in the Caribbean region. Epistemic Uncertainties Epistemic uncertainty is modeled by examining the influence of epistemic uncertainty in stress parameter and Q factor, which are the largest source of epistemic uncertainty. It is also evaluated through comparisons of the results of the Caribbean GMPEs with other ground motion prediction equations in the other tectonically similar regions. The Caribbean GMPEs have been developed for a stress of 130 bars based on the EXSIM model (see Question 2 – Source models and stress parameters) for the problem issue with SMSIM stress parameter matching), and the epistemic uncertainty in this value has been considered to be an order of a factor of 2. According to Atkinson and Boore (2006), it should be noted that the effect of the stress parameter on the simulated ground-motion values is approximately independent of distance. The frequency range over which the increase in amplitudes will occur depends on the earthquake magnitude. As shown in Figure 4, increasing the stress parameter has almost no effect on the amplitude at low frequencies. Then, the amplitude will increase and reach a constant factor at high frequencies. Furthermore, the effect of epistemic uncertainty in stress parameter is illustrated in Figure 4, which plots the amount by which the log amplitudes predicted by the GMPEs would need to be increased to accommodate a factor of 2 increases in stress parameter (Atkinson and Boore, 2006, personal communications with Atkinson).


DRAFT

8

Figure 4. Effect on predicted ground-motion amplitudes of increasing the stress

parameter by a factor of 2 (relative to predicted values for 140 bars), adopted from Atkinson and Boore (2006).

The effect of the Q-factor on the simulated ground motion values is strongly dependent of distance. As mentioned above, the MA05 Q-factor model and the range of epistemic uncertainties for the simulation of ground motions in the Caribbean region implies that Q-factor adopted for the Caribbean crust has captured conservatively the epistemic uncertainty in Q-factor (see Figures 2 and 3 and its related discussions). According to McNamara et al., 2012, while the Q-factor relations in the Caribbean region are relatively low and indicative of an active tectonic environment, Lg-phase amplitudes are considerably higher than might be expected in an oceanic region. Therefore, because of the complicated nature of the crust, the Q-factor is expected to be

0.00

0.05

0.10

0.15

0.20

0.25

4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0

Log

PSA

Fac

tor

(Rel

ativ

e to

140

Bar

s)

Moment Magnitude

0.2 Hz

0.5 Hz

1 Hz

2 Hz

5 HZ

PGA

PGV


DRAFT

highly variable throughout other parts of the Caribbean region, and then the appropriate ranges of Q factors used in the Caribbean region are acceptable. In addition, seismic hazard assessments for the Caribbean region should use a variety of ground motion prediction equations (GMPEs) obtained for other tectonically similar regions to capture the large epistemic uncertainty observed on earthquake ground-motion predictions. Question 4. Is the recommended weighting based on the EPRI (2004) class weights acceptable? The recommended weighting based on the EPRI (2004) class weights is reasonable and acceptable. Question 5. Are there any other GMPE models which should be considered for the PSHA contribution from Caribbean sources? The resulting Caribbean GMPEs should be similar to stochastic ground-motion relations obtained for other parts of United States. Differences are attributable to regional variations in ground-motion parameters and crustal structures. The Q-factor relations in the Caribbean region and the WUS region show that areas of active tectonics have higher attenuation of seismic waves than the stable continental regions such as the CEUS region. Therefore, I would recommend that the shallow crustal ground-motion attenuation relations in WUS (NGA 2008) as well as the ground-motion model of MA05, which was developed for Puerto Rico, are also plotted to provide a better estimation of ground motions in the Caribbean region for comparison. The intermediate earthquakes (Mw 5-6), in particular near the north of Cuba can be added to the actual data plots (see USGS and IRIS datasets) to better understand the variability of ground motions between the Cuba and the FPL project site. Question 6. Is either the Mid-Continent or Gulf Coast region GMPE from EPRI (2004) more appropriate for use in the evaluation of the contribution to PSHA at a site in southern Florida from Caribbean seismic sources? The Q-factor relations in the different tectonic environments (Figure 1) show that areas of active tectonics, like the Caribbean region and the WUS region, have higher attenuation of seismic waves than the stable continental regions such as the CEUS region. Therefore, it is possible that the ground-motion attenuation relations developed in the CEUS Mid-continental region may not be applicable for the estimation of ground motions in the Caribbean region. This is shown considering the observed ground-motion data in the region (see Figures 9 through 12 of the SSHAC Caribbean GMPE Questionnaire). In my opinion, as discussed earlier, the main parameter that controls the behavior of ground motions, especially at large distances, is the region-specific Q-values. As shown in Figure 1, the Q-values used in the development of ground motions in the Caribbean region lie between those


DRAFT

of the CEUS and WUS regions. Therefore, the comparison of the developed Caribbean GMPEs with the EPRI (2004) Mid-Continent and Gulf Coast GMPEs implies that the ground motion attenuation curves adopted for the Caribbean region are between those EPRI cases (see Figures 1 through 7 of the SSHAC Caribbean GMPE Questionnaire). However, the actual earthquakes show that the Caribbean GMPEs are closer to the Gulf Coast GMPES than EPRI (2004) Mid-Continent GMPEs. Another parameter that makes the ground motions in the Caribbean GMPEs deviate from the EPRI (2004) ground motion relations is differences in the hinge points of the attenuation curve that reflect regional differences in crustal structure between two regions (see Geometrical Spreading section). Assuming all parameters being equal, shifting the hinge point to 100 km --compared to 130 km for the CEUS region- leads to reduce the ground motions in the Caribbean region at large distances. Question 7. Do you have any other comments? Four minor comments:

• Table 1 of the SSHAC Caribbean GMPE Questionnaire suggests that a value of 3.6 km/sec was used to represent the shear-wave velocity at the source. Similarly, this table suggests that site amplification factors provided in Chen and Atkinson (2002) have been used. It should be noted that Chen and Atkinson (2002) used a shear-wave velocity of 3.8 km/sec; therefore, the site amplifications need to be adjusted accordingly.

• I am not clear how data for Table 2 were generated. An explanation would be needed. • More recording events, especially between Cuba and the FPL project site, should be

superimposed to the empirical ground-motion attenuation curves for the intermediate magnitude ranges between 5 and 6, if the records are available to process.

• The focal mechanism and depth of the actual earthquakes should be addressed in the report.


DRAFT

REFERENCE Abrahamson, N. A., and W. J. Silva (1997). Empirical response spectral attenuation relations for

shallow crustal earthquakes. Seismol. Research Letters, 68(1), 94–127. Aki, K. (1980). Scattering and attenuation of shear waves in the lithosphere, J. Geophys. Res. 85,

6496–6504, doi 0.1029/JB085iB11p06496. Atkinson, G. M., and D. M. Boore (1995). Ground-motion relations for eastern North America,

Bull. Seismol. Soc. Am. 85, 17–30. Atkinson, G., and D. Boore (2006). Ground motion prediction equations for earthquakes in

eastern North America, Bull. Seismol. Soc. Am. 96, 2181–2205. Atkinson, G., and R. Mereu (1992). The shape of ground motion attenuation curves in

southeastern Canada, Bull. Seismol. Soc. Am. 82, 2014–2031. Atkinson, G., K. Assatourians, D. Boore, K. Campbell, and D. Motazedian (2009). A guide to

differences between stochastic point-source and stochastic finite-fault simulations, Bull. Seismol. Soc. Am. 99, 3192–3201.

Atkinson, G. M., and W. Silva (2000). Stochastic modeling of California ground motions, Bull.

Seism. Soc. Am. 90, 255–274. Benz, H., A. Frankel, and D. Boore (1997). Regional Lg attenuation for the continental United

States, Bull. Seismol. Soc. Am. 87, 606–619. Boore, D., W.B. Joyner, and T.E. Fumal (1997). Equations for Estimating Horizontal Response

Spectra and Peak Acceleration from Western North American Earthquakes: As Summary of Recent Work, Seismol. Research Letters, 68(1), 128-153

Boore, D. (2009). Comparing stochastic point-source and finite-source ground-motion

simulations: SMSIM and EXSIM, Bull. Seismol. Soc. Am. 99, 3202–3216. Boore, D. M. (2005). “SMSIM – Fortran Programs for Simulating Ground Motions from

Earthquakes Version 2.3 – A Revision of OFR 96-80-A,” USGS Open File Report 00-509.

Campbell, K. (2008). Hybrid empirical ground motion model for PGA and 5% damped linear

elastic response spectra from shallow crustal earthquakes in stable continental regions: Example for eastern North America. Proc. 14th World Conf. Earthq. Eng., Oct. 12-17, Beijing, China, Paper S03-001.

Erickson, D., D. E. McNamara, and H. M. Benz (2004). Frequency dependent Lg Q within the

continental United States, Bull. Seismol. Soc. Am. 94, no. 5, 1630–1643,


DRAFT

Electric Power Research Institute [EPRI] (2004) CEUS Ground Motion Project Final Report, Elec. Power Res. Inst, Technical Report 1009684, dated December 2004.

Frankel, A. (1991). Mechanisms of seismic attenuation in the crust: Scattering and anelasticity in

New York state, South Africa and, southern California, J. Geophys. Res. 96, 6269–6289. Hough, S. E., and J. G. Anderson (1988). High-frequency spectra observed at Anza, California:

Implications for Q structure, Bull. Seismol. Soc. Am. 78, 692–707. McNamara, D., T. Owens, and W. Walter (1996). Propagation characteristics of Lg across the

Tibetan Plateau, Bull. Seismol. Soc. Am. 86, 457–469. McNamara, D., M. Meremonte, J. Z. Maharrey, S-L. Mildore, J. R. Altidore, D. Anglade, S. E.

Hough, D. Given, H. Benz, L. Gee, and A. Frankel (2012). Frequency-dependent seismic attenuation within the Hispaniola Island region of the Caribbean Sea, Bull. Seismol. Soc. Am. 102(2), 773-782.

Motazedian, D., and G. Atkinson (2005). Ground-motion relations for Puerto Rico, in Active

tectonics and seismic hazards of Puerto Rico, the Virgin Islands, and offshore areas, P. Mann (Editor), Geol. Soc. Am. Special Paper 385, 61–80.

Power, M., B. Chiou, N. Abrahamson, Y. Bozorgnia, T. Shantz, and C. Roblee (2008). An

overview of the NGA project, Earthquake Spectra, 24, 3–21. Singh, S., and R. B. Herrmann (1983). Regionalization of crustal code Q in the continental

United States, J. Gephys. Res. 88, 527-538.


DRAFT


Response to Questionnaire

Paul Somerville URS July 11, 2012

1. Are the newly developed Caribbean GMPEs acceptable for modeling ground motions in the Caribbean region for the project site located in Southern Florida for the PSHA study?

From a regulatory viewpoint I expect that they are acceptable because they are conservative when compared with the data. From a scientific viewpoint they do not have a sound physical basis and therefore lack predictive power, as suggested by their overprediction of the data. See Comments on Q1 and Q5 below.

2. Are the selected input parameters acceptable for their use?

Not relevant; see Comment on Q5 below

3. Is the recommended sigma value (i.e., 0.645 natural log units) acceptable?

Yes

4. Is the recommended weighting based on the EPRI (2004) class weights acceptable?

Not relevant; see Comment on Q5 below

5. Are there any other GMPE models which should be considered for the PSHA contribution from Caribbean sources?

None exist now; it is preferable to develop an empirically calibrated model. See Comment on Q5 below.

6. Is either the Mid-Continent or Gulf Coast region GMPE from EPRI (2004) more appropriate for use in the evaluation of the contribution to PSHA at a site in southern Florida from Caribbean seismic sources?

The Gulf Coast region GMPE is more appropriate; see Comment on Q5 below

7. Do you have any other comments?

Comment on Q1. Motazedian and Atkinson (2005) and the Caribbean Model

Motazedian and Atkinson (2005), page 77, state that:

“the Puerto Rico relations are based on the combination of events of all types. Crustal events have contributions to the Puerto Rico relations as well as interface and in-slab events. Due to the limited data precision for Puerto Rico events, it is not feasible to distinguish between event types in most cases. Thus, the Puerto Rico ground motions are region-specific, but the attenuation is averaged over all types of events.” The in-slab events have depths as large as 200 km.


DRAFT

The bold italics are my emphasis, and I interpret Puerto Rico as not including the whole Caribbean region per Figure 1 and my Comments on Q5. I think that this statement renders the Motazedian and Atkinson (2005) model unsuitable for use in representing Caribbean earthquake ground motions at the Florida site. This unsuitability is manifested in its large degree of overprediction of most of the recorded ground motions (see Table 1).

Comment on Q5. Alternative Empirically Calibrated Model from Recorded Ground Motion Data

The need for an alternative empirically calibrated model is suggested by the following analysis of the source characteristics of the earthquakes (listed in Table 1 and derived from source parameters provided by Nick Gregor in Appendix 1) whose recorded ground motions were used for comparison with the candidate models. The path continuity is inferred from Figure 1, in which dark blue regions indicate oceanic crust and paler blue offshore regions indicate continental crust. The path continuity is relevant because, for distances beyond about 70 km, physics-based ground motion prediction equations for crustal earthquakes are based on the propagation of waves that are trapped in the crustal waveguide; in ENA these waves are called Lg waves. Discontinuity in the waveguide due to crustal thinning or thickening can lead to drastic reduction in the efficiency of the waveguide. If the earthquake occurs below the crust (i.e. in the mantle), then the crustal waveguide is not primarily involved in controlling the seismic wave propagation to the site. In that case, the attenuation of ground motion would instead be controlled by the seismic velocity gradient of the upper mantle (the gradient returns the seismic waves back to the surface through critical reflection phenomena). I am not aware of any ground motion prediction models that address mantle earthquakes (other than ones for slab earthquakes in subduction zones, which would seem to be too deep and tectonically different to be suitable here).

Table 1. Tectonic Environment of Recorded Earthquakes

Event Depth (km) Mechanism / Tectonic Environment

Path Continuity

Recorded /Model ground motion: Gulf Coast Caribbean

Caribbean Sea

12 Strike-slip on transform fault in oceanic mantle

Very low Consistent Lower

Gulf of Mexico

29.6 Intraplate thrust in oceanic mantle

Low Higher Consistent

South of Cuba

12.6 Strike-slip on transform fault in oceanic mantle


N. of Honduras

12.0 Strike-slip on transform fault in continental - oceanic transition


Haiti 12.0 Strike-slip in continental crust Very Low Consistent Lower * recorded ground motions with respect to Gulf Coast model

From Table 1 we can see that none of the five earthquakes is a crustal earthquake with a continuous crustal path to Florida, which is the condition on which all of the considered GMPE’s is based. The first three earthquakes almost certainly occurred within the mantle, all with low or very low path continuity. The fourth occurred near a continental - oceanic transition with very low path continuity. Only the fifth, the Haiti earthquake, occurred in continental crust but it also has very low path continuity.


DRAFT

I infer that, for four of the events, the combination of very low path continuity and occurrence below the crust (except for the Haiti event) and causes the rapid rate of attenuation of their ground motions, consistent with the Gulf Coast ground motion model. Only the Gulf of Mexico event, which has low (but not very low) path continuity, has ground motions whose rate of attenuation is consistent with the Caribbean model, and even for that event, the recording in Florida is consistent with the higher of the Gulf Coast models. The relatively high ground motions of this event may reflect the high strength of the intraplate upper mantle, and may not be representative of the lower strength of the shallow upper mantle of the other three mantle events, which are all on or near plate margins.

We have thus shown that the physical basis of the Motazedian and Atkinson (2005) model renders it unsuitable for predicting the ground motions of earthquakes that occur in the mantle or whose waves propagate across path discontinuities or both, and that this lack or predictive power seems to be manifested in their overprediction of the ground motions recorded from at least four of the five earthquakes.

In view of the lack of predictive power of the Motazedian and Atkinson (2005) model and the Caribbean model derived from it, I think it would be preferable to select a model that is consistent with the recorded ground motion data. The Gulf Coast region model seems to form the best starting point for such a model.

Figure 1. Tectonics and Crustal Structure of the Caribbean Sea Region. Source: http://escweb.wr.usgs.gov/share/mooney/training%20courses.html#Caribbean; http://escweb.wr.usgs.gov/?attach_external_tab&99960320&4&0&0&0&0&Google%20Chrome%20Frame


DRAFT

Appendix 1. Inference of Tectonic Environment of Analyzed Earthquakes

12/14/2004 (M6.8)-Harvard Caribbean Sea Latitude: 19.050 Longitude: -81.520 Depth = 12.0 km Fault Plane 1 (Strike, Dip, Rake): 258.0, 84.0, -2.0 Fault Plane 2 (Strike, Dip, Rake): 349.0, 88.0, -174.0 Tectonics: Strike-slip on transform fault in Oceanic mantle? 09/10/2006 (M5.9)-GCMT Gulf of Mexico Latitude: 26.32 Longitude: -86.84 Depth = 29.6 km Fault Plane 1 (Strike, Dip, Rake): 324.0, 28.0, 117.0 Fault Plane 2 (Strike, Dip, Rake): 114.0, 65.0, 77.0 Tectonics: Thrust in oceanic mantle? 02/04/2007 (M6.2)-GCMT South of Cuba Latitude: 19.490 Longitude: -78.340 Depth = 12.6 km Fault Plane 1 (Strike, Dip, Rake): 257.0, 76.0, -9.0 Fault Plane 2 (Strike, Dip, Rake): 349.0, 81.0, -166.0 Tectonics: Strike-slip on transform fault in Oceanic mantle? 05/28/2009 (M7.3)-GCMT North of Honduras

Latitude: 16.500 Longitude: -87.170 Depth = 12.0 km (Fixed) Fault Plane 1 (Strike, Dip, Rake): 63.0, 60.0, -7.0 Fault Plane 2 (Strike, Dip, Rake): 156.0, 84.0, -150.0 Tectonics: Strike-slip on transform fault in continental – oceanic crust transition? 01/12/2010 (M7.0)-GCMT Haiti Latitude: 18.610 Longitude: -72.620 Depth = 12.0 km (Fixed) Fault Plane 1 (Strike, Dip, Rake): 250.0, 71.0, 22.0 Fault Plane 2 (Strike, Dip, Rake): 152.0, 69.0, 159.0 Tectonics: Strike-slip in Continental crust?


DRAFT

Response�to�Caribbean�GMPE�Questions�by�Robert�Youngs�

1.�Are�the�newly�developed�Caribbean�GMPEs�acceptable�for�modeling�ground�motions�in�the�Caribbean�region�for�the�project�site�located�in�Southern�Florida�for�the�PSHA�study?�

Response:�The�models�developed�from�simulation�appear�to�be�reasonable.�The�comparisons�with�recorded�motions�indicate�that�they�may�slightly�over�estimate�ground�motions,�which�is�OK.��

2.�Are�the�selected�input�parameters�acceptable�for�their�use?�

A�comparison�of�the�Chen�and�Atkinson�(2002)�amplification�functions�with�those�used�to�develop�CEUS�models�should�be�made�to�confirm�their�consistency.�You�want�the�model�to�be�predicting�ground�motions�for�site�conditions�similar�to�those�of�EPRI�(2004)�

3.�Is�the�recommended�sigma�value�(i.e.,�0.645�natural�log�units)�acceptable?�

Response:�There�does�not�appear�to�be�a�strong�basis�for�the�0.645�and�it�is�intended�for�use�in�Puerto�Rico.��A�more�preferable�model�with�a�stronger�basis�may�be�the�EPRI�(2006)�model�for�CENA�or�some�simplification�of�that�model,�perhaps�using�the�final�NGA�(2008)�results.�

4.�Is�the�recommended�weighting�based�on�the�EPRI�(2004)�class�weights�acceptable?�

Yes,�roughly�equal�weight�for�1�versus�2�corner�is�reasonable�

5.�Are�there�any�other�GMPE�models�which�should�be�considered�for�the�PSHA�contribution�from�Caribbean�sources?�

Response:�The�USGS�uses�WNA�GMPEs�for�modeling�ground�motions�in�Puerto�Rico.�It�may�be�useful�to�compare�the�predictions�from�the�NGA�models�adjusted�to�CEUS�hard�rock�conditions�using�published�factors�(e.g.�Atkinson�and�Boore,�BSSA�2011)�with�the�developed�models�at�least�in�a�distances�of�about�200�evaluate�whether�they�predict�higher�or�lower�ground�motions.�

6.�Is�either�the�Mid�Continent�or�Gulf�Coast�region�GMPE�from�EPRI�(2004)�more�appropriate�for�use�in�the�evaluation�of�the�contribution�to�PSHA�at�a�site�in�southern�Florida�from�Caribbean�seismic�sources?�

Response:�I�do�not�see�any�reason�to�consider�the�EPRI�(2004)�models�to�be�more�appropriate�than�the�Caribbean�model�developed�for�the�Turkey�Point�study.�The�comparisons�with�data�shown�in�the�figures�confirms�that.�

7.�Do�you�have�any�other�comments?�

Response:�

Yes,�see�below�


DRAFT

(a) The�presentation�of�the�comparisons�on�Figures�1�7�makes�it�very�difficult�to�distinguish�among�the�various�relationships.�It�would�perhaps�be�more�useful�to�compare�the�median�models�of�the�EPRI�(2004)�clusters�with�the�three�base�case�Caribbean�relationships.�

(b) It�is�not�stated�why�the�EPRI�(2004)�Cluster�4�model�is�not�included�in�the�comparisons.�It�is�appropriate�for�large�magnitude�earthquakes,�which�are�the�type�of�earthquakes�from�the�Caribbean�sources�that�contribute�to�the�hazard�at�Turkey�Point.�

(c) Some�description�of�the�site�conditions�at�the�Florida�DWPF�station�should�be�included.�This�station�has�the�appropriate�path�for�Caribbean�earthquakes�recorded�in�Florida,�but�what�about�site�conditions�or�expected�amplification�from�hard�rock?��


DRAFT

draft revised response for nrc rai letter no. 037 (erai ... · approach to define the site-specific...

Documents