Final Report: Seismic Hazard Assessment at the PGDP

UK/KRCEE Doc #: P11.6 2007 Final Report: Seismic Hazard Assessment at the PGDP Prepared by Kentucky Research Consortium for Energy and Environment 233 Mining and Minerals Building University of Kentucky, Lexington, KY 40506-0107 Prepared for United States Department of Energy Portsmouth/Paducah Project Office Acknowledgment: This material is based upon work supported by the Department of Energy under Award Number DE-FG05-03OR23032. June 2007 UK/KRCEE Doc #: P11.6 2007 Final Report: Seismic Hazard Assessment at the PGDP By Zhenming Wang Kentucky Geological Survey 228 Mining and Mineral Resources Building University of Kentucky Lexington, Kentucky 40506 Edward W. Woolery Department of Earth and Environmental Sciences 309 Slone Research Building University of Kentucky Lexington, Kentucky 40506 June 2007 Executive Summary Selecting a level of seismic hazard at the Paducah Gaseous Diffusion Plant for policy considerations and engineering design is not an easy task because it not only depends on seismic hazard, but also on seismic risk and other related environmental, social, and economic issues. Seismic hazard is the main focus. There is no question that there are seismic hazards at the Paducah Gaseous Diffusion Plant because of its proximity to several known seismic zones, particularly the New Madrid Seismic Zone. The issues in estimating seismic hazard are (1) the methods being used and (2) difficulty in characterizing the uncertainties of seismic sources, earthquake occurrence frequencies, and ground-motion attenuation relationships. This report summarizes how input data were derived, which methodologies were used, and what the hazard estimates at the Paducah Gaseous Diffusion Plant are. Three seismic sources (the New Madrid Seismic Zone, the Wabash Valley Seismic Zone, and the background seismicity) were identified and characterized. Four ground-motion attenuation relationships were used in this project. Probabilistic seismic hazard analysis (PSHA) and deterministic seismic hazard analysis (DSHA) were performed. A panel of six members, who are experts in geology, seismology, earthquake engineering, and statistics, provided a review of the report. The review comments and responses are included as appendices. In PSHA, seismic hazard is defined as the annual probability of a ground motion being exceeded. The inverse of the annual probability of exceedance is defined as the return period. Therefore, seismic hazard is also defined as a ground motion being exceeded in a return period. PSHA calculates seismic hazard from all earthquake sources in consideration, and implicitly incorporates uncertainty in earthquake size and location and ground motion. Figure E-1 shows the peak ground acceleration (PGA) hazard curve on hard rock at the Paducah Gaseous Diffusion Plant calculated in this study. Table E-1 lists ground-motion hazards on hard rock for the Paducah Gaseous Diffusion Plant at several return periods. PGA Hazard Curve 1.0E-02 Ann. Prob. of Exc. 1.0E-03 1.0E-04 1.0E-05 1.0E-06 1.0E-07 1.0E-08 0.01 0.1 PGA (g) 1 10 Figure E-1. PGA hazard curve on hard rock at the Paducah Gaseous Diffusion Plant. i Table E-1. PSHA ground-motion hazards on hard rock at the Paducah Gaseous Diffusion Plant. Ann. Prob. Exc. 0.004 0.002 0.001 0.0004 0.0002 Return Period (years) 250 500 1,000 2,500 5,000 Exc. Prob. in 50 years (%) 18 10 5 2 1 PGA (g) 0.09 0.18 0.29 0.49 0.62 0.2s PSA (g) 0.10 0.21 0.40 0.68 0.90 1.0s PSA (g) 0.01 0.03 0.09 0.16 0.23 In DSHA, seismic hazard is defined as the ground motion(s) from a single or several earthquakes that are expected to produce maximum values (impacts) at a site. DSHA emphasizes ground-motion hazard from an individual earthquake (scenario), such as the maximum credible earthquake (MCE) or maximum considered earthquake, and explicitly determines ground-motion hazard with a level of uncertainty. DSHA results show that the large earthquakes in the New Madrid Seismic Zone dominate seismic hazard at the Paducah Gaseous Diffusion Plant. Table E-2 lists ground-motion hazards estimated for the Paducah Gaseous Diffusion Plant from large earthquakes in the New Madrid Seismic Zone. The return period for these ground motions is about 500 to 1,000 years. Table E-2. DSHA ground-motion hazards on hard rock at the Paducah Gaseous Diffusion Plant. PGA 0.2s PSA 1.0s PSA Median (g) 0.25 0.39 0.12 Median +1σln,y (g) 0.51 0.80 0.24 Median +2σln,y (g) 1.03 1.65 0.51 1.5 Median (g) 0.38 0.59 0.18 Return period: 500 to 1,000 years The results from this project show that PSHA and DSHA could provide significantly different hazard estimates for the Paducah Gaseous Diffusion Plant. DSHA provides a ground-motion hazard with a level of uncertainty (Table E-2) from the large earthquake in the New Madrid Seismic Zone, whereas PSHA provides a range of ground-motion hazards (Fig. E-1) from all earthquakes being considered. Table E-3 lists recommended ground motions for engineering design consideration for facilities at the Paducah Gaseous Diffusion Plant. All the hazard estimates are on hard rock, and no amplification by the near-surface soils is considered in this report. Table E-3. Recommended ground motions on hard rock at the Paducah Gaseous Diffusion Plant PSHA PGA 0.2s 1.0s Facility DSHA (g) PSA PSA Return Exc. Prob. (g) (g) Period in 50 years (years) (%) Ordinary Median 1,000 5 0.27 0.40 0.10 Important Median + one standard deviation 2,500 2 0.50 0.80 0.20 ii Table of Contents Executive Summary ……………………………………………………………………...i Table of Contents …………………………………………………………….................iii List of Acronyms ……………………………………………………………................. iv List of Figures ………..……………………………………….………………………....v List of Tables ………………………………………………….………………………. vii 1.0 Introduction………………………………………………………………..…….1 2.0 Methodology …………………………………………………………………….4 2.1 Probabilistic Seismic Hazard Analysis ………….……………………….4 2.2 Deterministic Seismic Hazard Analysis ……………..………………....10 3.0 Seismic Sources ………………………………………………………………..12 3.1 New Madrid Seismic Zone ……………………………………………..12 3.2 Wabash Valley Seismic Zone …………………………………………...17 3.3 Background Seismicity ……....……………………………………….…21 4.0 Magnitude – Recurrence Relationship ………………………………………24 4.1 New Madrid Seismic Zone ……………………………………………..24 4.2 Wabash Valley Seismic Zone …………………………………………..26 4.3 Local Source Zone …………..……………………………………….…26 5.0 Ground-Motion Attenuation Relationship …………………………….…….29 6.0 Results ………………………………………………………………………….37 6.1 PSHA Results…………………………………………………………..38 6.2 DSHA Results…………………………………………………………..41 7.0 Conclusion and Recommendations……………………….............................43 8.0 References ……………………………………………………………………..46 9.0 Appendix A …………………………………………………………………..A-1 10.0 Appendix B ………………………………………………………………......B-1 11.0 Appendix C ………………………………………………………………......C-1 12.0 Appendix D ………………………………………………………………......D-1 iii List of Acronyms DSHA EPRI GPS M MCE MMI NMSZ PDF PE PGA PGDP PGV PSA PSHA SHA USGS WVSZ Deterministic Seismic Hazard Analysis Electric Power Research Institute Global Positioning System Magnitude Maximum Credible Earthquake Modified Mercalli Intensity New Madrid Seismic Zone Probability Density Function Probability of Exceedance Peak Ground Acceleration Paducah Gaseous Diffusion Plant Peak Ground Velocity Probabilistic Safety Assessment Probabilistic Seismic Hazard Analysis Seismic Hazard Analysis United States Geological Survey Wabash Valley Seismic Zone iv List of Figures Figure 1. Earthquakes with magnitude 5 or greater since 1900 in California and the New Madrid Seismic Zone ..........................................................................2 Figure 2. Ground-motion attenuation relationship...........................................................6 Figure 3. Hazard curve at a site 40 km from the source for a characteristic earthquake of M7.7 with a recurrence time of 500 years in the New Madrid Seismic Zone. ................................................................................................................7 Figure 4. PGA hazard curves showing the effect of ground-motion uncertainty ............9 Figure 5. PGA hazard curves from the Yucca Mountain project ..................................10 Figure 6. Seismicity between 1974 and 2005 in the central United States ....................13 Figure 7. New Madrid faults. Pseudo-faults (red) were used in the 1996 and 2002 USGS seismic hazard maps ...................................................................15 Figure 8. Bouguer gravity anomaly and 1974–94 earthquake epicenters and the New Madrid Rift Complex .............................................................................16 Figure 9. Dense seismic network and earthquakes recorded between January 2003 and June 2005 in the Jackson Purchase Region.....................................17 Figure 10. Epicentral locations of the damaging earthquakes in the Wabash Valley Seismic Zone ..................................................................................................19 Figure 11. Earthquakes and faults in the lower Wabash Valley ......................................20 Figure 12. New Madrid and Wabash Valley Seismic Zone.............................................21 Figure 13. Seismic sources that were considered in the national seismic hazard maps...22 Figure 14. Background earthquakes (Mmax) used in the national seismic hazard maps...22 Figure 15. Maximum background earthquakes in Kentucky ...........................................23 Figure 16. Magnitude-frequency relationship in the New Madrid Seismic Zone............25 Figure 17. Magnitude-frequency (Gutenburg-Richter) curve for the New Madrid Seismic Zone. .................................................................................................25 Figure 18. Magnitude-frequency relationship in the Wabash Valley Seismic Zone .......27 Figure 19. Magnitude-frequency (Gutenburg-Richter) curve for the Wabash Valley Seismic Zone. .................................................................................................27 Figure 20. Recorded earthquakes with magnitude greater than 2.5 surrounding PGDP between 1978 and 2006 .......................................................................28 Figure 21. Magnitude-frequency (Gutenburg-Richter) curve for the local source ..........28 Figure 22. Horizontal uncorrected PGA vs. distance to the fault for the Parkfield earthquake of September 28, 2004 .................................................................30 Figure 23. Strong-motion stations and accelerograms recorded in the 2004 M6.0 Parkfield earthquake, for the east-west component........................................31 Figure 24. Magnitude-distance-region distribution of selected recordings .....................33 Figure 25. Log values of horizontal component 5% pseudo-acceleration at frequencies 0.5, 1, and 5 Hz, and PGA, for ENA rock sites. .............................................34 Figure 26. PGA attenuation relationships at hard rock for an M7.5 earthquake in the central United States .......................................................................................35 Figure 27. PGA attenuation relationships at hard rock for an M5.0 earthquake in the central United States .......................................................................................36 Figure 28. The 0.2s PSA attenuation relationships used in this study at hard rock for an M7.5 earthquake in the central United States ......................................36 v Figure 29. The 1.0s PSA attenuation relationships used in this study at hard rock for an M7.5 earthquake in the central United States ......................................37 Figure 30. Seismic sources for the Paducah Gaseous Diffusion Plant ............................38 Figure 31. PGA hazard curve on hard rock at the Paducah Gaseous Diffusion Plant .....40 Figure 32. The 0.2s PSA hazard curve on hard rock at the Paducah Gaseous Diffusion Plant................................................................................................40 Figure 33. The 1.0s PSA hazard curve on hard rock at the Paducah Gaseous Diffusion Plant................................................................................................41 Figure 34. Isoseismal map of the February 7, 1812, New Madrid earthquake ................45 vi List of Tables Table 1. Table 2. Table 3. Table 4. Table 5. Table 6. Table 7. Table 8. Table 9. Table 10. Table 11. Table 12. Table 13. Table 14. Table 15. Comparisons between California and western Kentucky.................................2 Parameters of earthquakes ..............................................................................16 Damaging earthquakes in the Wabash Valley Seismic Zone .........................18 Earthquakes with magnitude equal to or greater than M4.0 in the New Madrid Seismic Zone......................................................................................26 Median ground motions for an M7.7 New Madrid earthquake at 15 km for a hard-rock site from several attenuation relationships ............................32 Input parameters and weights being used in PSHA for Paducah Gaseous Diffusion Plant................................................................................................39 Ground-motion values on hard rock at Paducah Gaseous Diffusion Plant ....39 Median PGA on hard rock at the Paducah Gaseous Diffusion Plant from the three seismic sources ................................................................................42 PGA at the Paducah Gaseous Diffusion Plant from the characteristic earthquake in the New Madrid Seismic Zone.................................................42 The 0.2s PSA at the Paducah Gaseous Diffusion Plant from the characteristic earthquake in the New Madrid Seismic Zone ..........................42 The 1.0s PSA at the Paducah Gaseous Diffusion Plant from the characteristic earthquake in the New Madrid Seismic Zone ..........................42 Ground-motion hazards on hard rock at the Paducah Gaseous Diffusion Plant determined by PSHA.............................................................................43 Ground-motion hazards on hard rock at the Paducah Gaseous Diffusion Plant determined by DSHA ............................................................................43 Recommended ground motions on hard rock at the Paducah Gaseous Diffusion Plant ....................................................................................................... 44 Comparison of mean PGA estimates on hard rock at the Paducah Gaseous Diffusion Plant determined from PSHA.........................................................45 vii 1.0. Introduction Federal agencies including the Federal Emergency Management Agency and the Environmental Protection Agency, State agencies such as the Kentucky Environmental and Public Protection Cabinet, and other government and private organizations such as the American Association of State Highway and Transportation Officials and the Building Seismic Safety Council use seismic-hazard maps produced by the U.S. Geological Survey (Frankel and others 1996, 2002) for seismic safety regulations and engineering design. The maps currently being used show the ground motions with 2 percent probability of exceedance (PE) in 50 years. These maps predict very high ground motion in many counties in western Kentucky: peak ground acceleration (PGA) of 1.0g or higher. These high ground-motion estimates affect everything in western Kentucky from building a single-family home to environmental clean-up at the Paducah Gaseous Diffusion Plant (PGDP) at the Superfund site. For example, it would be difficult for the U.S. Department of Energy to obtain a permit from Federal and State regulators to construct a landfill at the Paducah Gaseous Diffusion Plant if the USGS maps with 2 percent PE in 50 years are considered. The Structural Engineers Association of Kentucky (SEAOK 2002) found that if the International Residential Code of 2000, which was based on the 1996 USGS maps with 2 percent PE in 50 years, is adopted in Kentucky without revision, constructing residential structures in westernmost Kentucky, including Paducah, would be impossible without enlisting a design professional. Figure 1 shows a schematic comparison of seismic hazard for the New Madrid Seismic Zone and southern California on two time scales (Stein and others 2003). The 2000 International Building Code (IBC-2000);(ICC 2000), based on the 1996 USGS maps with 2 percent PE in 50 years, requires a design Peak Ground Acceleration (PGA) of about 0.6g in Paducah and about 0.8g at the Paducah Gaseous Diffusion Plant. Currently, the highest building-design PGA used in California (UBC-97) is capped at about 0.4g. These high design ground motions for western Kentucky are not consistent with the level of seismic activity. Although earthquakes are occurring in Kentucky and surrounding states, especially in the well-known New Madrid Seismic Zone where at least three large earthquakes (M7.0–8.0) occurred in 1811–1812, earthquake recurrence rates are much lower in the region than in California, the Pacific Northwest, and Alaska. Table 1 compares the basic geological and seismological observations and design PGA in California and western Kentucky. These comparisons clearly show that the higher design ground motion in western Kentucky may not be warranted. Selecting a level of seismic hazard for policy considerations and engineering design is very complicated. The selection of the level of seismic hazard not only depends on seismic hazard itself, but also on seismic risk and other related environmental, social, and economic issues. Seismic hazard assessment is the main focus. The objectives of this project were to gain a better understanding of the seismic hazard assessment at the Paducah Gaseous Diffusion Plant and its surrounding area, and to communicate the hazard information more effectively to users and policy makers. In order to achieve these objectives, the following tasks were established: 1) micro-seismicity observation in Paducah area, 2) thorough literature review, 3) seismic source characterization, 4) 1 probabilistic seismic hazard analysis (PSHA), 5) deterministic seismic hazard analysis (DSHA), 6) preliminary report, 7) panel review, and 8) final report (see Appendix A). The focus of this project was to review the methodology and data used by the U.S. Geological Survey due to the broad implication of the U.S. Geological Survey’s seismic hazard assessments. As a result, a series of activities were carried out, including workshops, professional conferences, publications, and personal meetings and communications (Wang, 2003a and b 2005a, b, c, and d, 2006a and b, in press; Wang and others 2003a and b, 2004a and b, 2005; Cobb, 2004, 2006; Wang and Ormsbee 2005). Figure 1. Schematic comparison of seismic hazard for the New Madrid Seismic Zone and southern California on two time scales. Circles mark area of shaking with acceleration > 0.2g (Stein and others 2003). Table 1. Design ground motion, geology, and seismicity comparisons between California and western Kentucky. California Design PGA Geology Seismicity ≤0.4g ≤0.7g (UBC97) (CALTRAN) San Andreas Fault Displacement≥20 mm/y High M7–8: ~100y M6–7: ~20-50y 2 Western Kentucky ≥0.4g ≥0.6g (IBC-2000) (bridge) New Madrid Fault Displacement≤2 mm/y Low M7–8: ~500y or longer M6–7: ? A review panel consisting of national and international experts on geology, seismology, engineering seismology, and engineering was formed to review a preliminary report that summarized the results from tasks 2) through 5). A statistician was added to the panel under the suggestion of other members of the panel (Appendix B). The review was divided into two parts, individual review (three days) and a panel review (one day) held in Lexington, Kentucky. The preliminary report was submitted to the members in late February 2007 for the individual review. The written comments on the preliminary report and responses provided by the members are included in this report as Appendix C. Subsequently, a panel review meeting was held on April 30, 2007 to discuss the preliminary report with focus on 1) ground-motion attenuation relationship – uncertainty, dependency, and hazard calculation in PSHA, 2) seismic hazard analysis (SHA) – temporal and spatial measurements, uncertainties, and quantification, and 3) seismic hazard assessment for PGDP – input parameters: sources, occurrence frequency, and ground motion attenuation. Even though there was not enough time to fully discuss all issues, the panel reached some consensus. These include: 1. The ground-motion hazards with a 2,500 return period estimated by the U.S. Geological Survey (Frankel and others 1996, 2002) are conservative. 2. PSHA, as a methodology, is the common approach for seismic hazard assessment, but some improvements are needed. 3. It is difficult to provide an estimate of seismic hazard for the Paducah Gaseous Diffusion Plant because a reasonable estimate is subjective. The recommendations from the review panel at the meeting on April 30, 2007 were: 1. to perform a PSHA with some discussions for improvements, 2. to perform a DSHA, 3. to revise the local source zone. A draft final report was completed according to the recommendations and sent to the members of the review panel for final review on May 11, 2007. The comments on the final draft report from members of the review panel is included in Appendix D. Appendix D also includes responses to the members’ comments. 3 2.0. Methodology Two methods, probabilistic seismic hazard analysis (PSHA) and deterministic seismic hazard analysis (DSHA), are commonly used for seismic-hazard assessment. PSHA and DSHA follow similar steps in estimation of seismic hazard (Reiter 1990; Kramer 1996): (1) Determination of earthquake sources (2) Determination of earthquake occurrence frequencies by selecting controlling earthquake(s): the maximum magnitude, maximum credible, or maximum considered earthquake (3) Determination of ground-motion attenuation relationships (4) Determination of seismic hazard. The differences between PSHA and DSHA are in step (4), on how to define and calculate seismic hazard. In PSHA, seismic hazard is defined as the annual probability of a ground motion being exceeded at a site (National Research Council, 1988; SSHAC, 1997; Frankel, 2004; McGuire 2004). The reciprocal of the annual probability of exceedance is called the return period and has been interpreted and used as “the mean (average) time between occurrences of a certain ground motion at a site” (McGuire 2004). Therefore, seismic hazard can also be expressed as a ground motion being exceeded in a specific return period such as 500, 1,000, or 2,500 years. PSHA calculates seismic hazard from all earthquake sources and considers the uncertainty in the number, size, and location of future earthquakes and ground motion (i.e., considers the possibility that ground motion at a site could be different for different earthquakes of the same magnitude at the same distance, because of differences in source parameter, path, and site condition) (Cornell 1968, 1971). The end results from PSHA are seismic hazard curves: a relationship between a ground-motion parameter (i.e., peak ground acceleration [PGA], peak ground velocity [PGV], or response acceleration at certain periods) and its annual probability of exceedance or return period. In DSHA, seismic hazard is defined as the ground motion(s) from a single or several earthquakes that have maximum values (impacts) at a site (Reiter 1990; Krinitzsky 2002). DSHA emphasizes the ground motion from an individual earthquake, such as the maximum credible or maximum considered earthquake, maximum probable earthquake, or design basis earthquake. Although the determination of recurrence interval of ground motion is not required and often not emphasized in DSHA, it is equal to the recurrence interval of an individual earthquake (Wang and others 2004). 2.1. PSHA PSHA was originally developed to derive theoretical ground-motion hazard curves at a site where there are little or no observations (Cornell 1968). Later, Cornell (1971) extended his method to incorporate ground-motion uncertainty (i.e., the possibility that ground motion at a site could be different for different earthquakes of the same magnitude at the same distance, because of differences in source parameters and path 4 effects).-. According to Cornell (1968, 1971) and McGuire (1995, 2004), the heart of PSHA is γ ( y ) = ∑ vP[Y ≥ y ] y = ∑ v ∫∫ {1 − ∫ 0 1 2π σ ln, y , (ln y − ln y mr ) 2 ] (ln )} ( ) ( ) exp[− d y f m f r dmdr M R 2 2σ ln, y (1) where ν is the activity rate, fM(m) and fR(r) are the probability density function (PDF) of earthquake magnitude M and epicentral (or focal) distance R, respectively, and ymr and σln,y are the median and standard deviation at m and r. The fM(m) and fR(r) were introduced to account for the uncertainty of earthquake magnitude and distance, respectively (Cornell 1968, 1971; McGuire 2004). The ymr and σln,y are determined by the ground-motion attenuation relationship (Campbell 1981, 2003; Joyner and Boore 1981; Abrahamson and Silva 1997; Toro and others 1997; EPRI 2003; Atkinson and Boore 2006). Ground motion Y is generally modeled as a function of M and R with uncertainty E (capital epsilon): ln(Y ) = f ( M , R) + E . (2) The uncertainty E is modeled as a normal distribution with a zero mean and standard deviation σln,Y (Campbell 1981, 2003; Joyner and Boore 1981; Abrahamson and Silva 1997; Toro and others, 1997; EPRI, 2003; Atkinson and Boore, 2006). In other words, the uncertainty of ground motion Y is modeled as a log-normal distribution, therefore, equation (2) can be rewritten as ln(Y ) = f ( M , R) + nσ ln,Y , (3) where n (a constant) is a number of standard deviations measured as the difference relative to the median ground motion f(M,R) (Fig. 2). According to Benjamin and Cornell (1970) and Mendenhall and others (1986), if and only if M, R, and E are independent random variables, the joint probability density function of M, R, and E is f M , R , E (m, r , ε ) = f M (m) f R (r ) f E (ε ) , (4) where fE(ε) is the PDF of E. The exceedance probability P[Y≥y] is P[Y ≥ y ] = ∫∫∫ f M , R , E (m, r , ε ) H [ln Y (m, r , ε ) − ln y ]dmdrdε = ∫∫∫ f M (m) f R (r ) f E (ε ) H [ln Y (m, r , ε ) − ln y ]dmdrdε , (5) where H[lnY(m,r,ε)-lny] is the Heaviside step function, which is zero if lnY(m,r,ε) is less than lny, and 1 otherwise (McGuire 1995). E follows a normal distribution, equation (5) can be rewritten as 5 P[Y ≥ y ] = ∫∫ {∫ f E (ε ) H [ln Y (m, r , ε ) − ln y ]dε } f M (m) f R (r )dmdr y = ∫∫ {1 − ∫ 0 1 2π σ ln, y exp[− , (ln y − ln y mr ) 2 ]d (ln y )} f M (m) f R (r )dmdr 2 2σ ln, y (6) where ln ymr=f(m,r). This results in equation (1), the heart of PSHA (Cornell 1968, 1971; McGuire 1995, 2004). Figure 2. Ground-motion attenuation relationship. The return period (Trp) is the inverse of the annual probability of exceedance (1/γ): Trp ( y ) = 1 = γ ( y) 1 y ∑ v ∫∫{1 − ∫ 0 1 2π σ ln, y (ln y − ln y mr ) 2 ]d (ln y )} f M (m) f R (r )dmdr exp[− 2 2σ ln, y . (7) 6 If all seismic sources are characteristic, the return period is Trp ( y ) = 1 (ln y − ln y mr ) 2 1 1 [ 1 )d (ln y )] exp( − − ∑ T ∫ 2π σ 2 σ 2 y ln, 0 ln, y y , (8) where T is the average recurrence interval of the characteristic earthquake (Mc) at distance Rc. For a single characteristic source, equation (8) becomes T Trp ( y ) = y 1− ∫ 0 1 2π σ ln,c (ln y − ln y c ) 2 )d (ln y ) exp(− 2 2σ ln, c . (9) Figure 3 shows how a PGA hazard curve is constructed at a site 40 km from the source with a single characteristic earthquake of M7.7 and a recurrence time of 500 years in the New Madrid Seismic Zone. Annual Probability of Exceedance -2 10 5% 2% 0.002 (1/500) 0.002x0.5 -3 10 0.002x0.2 -4 10 0.59 0.7 ε=0. 84 Probability Density 0.6 0.5 0.4 median 0.3 0.2 0.1 0 -1 10 0 0.36 10 PGA (g) Figure 3. Hazard curve at a site 40 km from the source for a characteristic earthquake of M7.7 with a recurrence time of 500 years in the New Madrid Seismic Zone. The median ground motion (µ) is 0.36g, and the standard deviation (σln) is 0.60. ε=(ln y –ln µ)/σln (Wang and others 2005). 7 The main purpose of PSHA is to directly incorporate uncertainty in earthquake size, frequency, location, and ground-motion. As demonstrated above, ground-motion uncertainty is implicitly incorporated and becomes an integral part of PSHA. Other uncertainties are incorporated explicitly through logic trees, by which different weights are assigned manually to a set of expert estimates for each input parameter (SSHAC 1997). These implicit and explicit incorporations of the uncertainty also have disadvantages. One such disadvantage, recognized by the first thorough review of PSHA by the committee chaired by Aki (National Research Council 1988), is the individual earthquake (single physical event) is lost “because the aggregated results of PSHA are not always easily related to the inputs.” In other words, “the concept of a ‘design earthquake’ is lost; i.e., there is no single event (specified, in simplest term, by a magnitude and distance) that represents the earthquake threat at, for example, the 10,000-yr ground-motion level (which we call the ‘target ground motion’)” (McGuire 1995). McGuire (1995) also proposed a methodology (de-aggregation) to seek the “design earthquake.” Another disadvantage is that uncertainty, ground-motion uncertainty in particular, becomes a controlling factor in PSHA. This can be seen clearly in recent studies (Frankel, 2004; Wang and others 2003b, 2005; Bommer and Abrahamson 2006), at low annual frequencies of exceedance (less than 10-4) in particular. Figure 4 shows how the computed hazard varies with truncation of standard deviation (Bommer and Abrahamson 2006). This is the reason that PSHA could result in extremely high ground motion (10g PGA or higher) if a long return period (100,000,000 years) is considered for facilities at the nuclear waste repository site in Yucca Mountain, Nev. (Stepp and others 2001; Abrahamson and Bommer, 2005; McGuire and others 2005; Musson 2005). As shown in Figure 5, a PGA of 11g would be the result at the nuclear waste repository site in Yucca Mountain, Nev., if a return period of 100,000,000 years is considered (Abrahamson and Bommer 2005). A significantly higher ground motion would have to be considered in reevaluation of the nuclear power plants in Switzerland if a return period of 10,000,000 to 100,000,000 years is considered (Klügel 2005; Scherbaum and others 2005). Bommer and Abrahamson (2006) attributed these high ground-motion estimates directly to the way the ground-motion uncertainty is treated in PSHA (Fig. 4). 8 Figure 4. PGA hazard curves showing the effect of ground-motion uncertainty (Bommer and Abrahamson 2006). 9 Figure 5. PGA hazard curves from the Yucca Mountain project (Abrahamson and Bommer 2005). 2.2. DSHA As discussed earlier, there is a fundamental difference between PSHA and DSHA in how to define and calculate seismic hazard. DSHA emphasizes the ground motion from an individual earthquake, such as the maximum credible earthquake and maximum probable earthquake. The steps outlined in Reiter (1990) and Krinitzsky (1995, 2002) are herein used to derive ground motions at the Paducah Gaseous Diffusion Plant. The advantage of DSHA is that ground motion is directly related to an earthquake, specified by a magnitude and distance. The uncertainty, including ground-motion uncertainty, is explicitly expressed in the results from DSHA. The advantages of DSHA are 1) “an easily understood and transmitted method of estimating seismic hazard” and 2) “clear to the analyst (earth scientist), the user (engineer) and those elements of the general public who are interested in nuclear power plant safety or earthquake related problems” (Reiter, 1990). 10 DSHA also has disadvantages. One such disadvantage is that “it (DSHA) does not take into account the inherent uncertainty in seismic hazard estimation” (Reiter 1990). The other disadvantage is that “frequency of occurrence is not explicitly taken into account” (Reiter 1990). In other words, DSHA does not carry units of time. As pointed out by Hanks and Cornell (1994), however, “it is generally possible to associate recurrence interval information with plausible deterministic earthquakes.” The plausible deterministic earthquakes are always associated with a recurrence interval, so in this sense DSHA actually does carry a unit of time (Wang and others 2004). 11 3.0. Seismic Sources The causes of intraplate earthquakes in the central United States are not well understood (Braile and others 1986; Zoback, 1992; Newman and others 1999; Kenner and Segall 2000). Two hypotheses have been proposed to explain this seismicity: (1) selective reactivation of preexisting faults by local variations in pore pressure, fault friction, and/or strain localization along favorably orientated lower-crustal ductile shear zones formed during earlier deformation (Zoback and others 1985) and (2) local stress perturbations that may produce events incompatible with the regional stress field (Zoback and others 1987). In the central and eastern United States, the regional stress field is reasonably well known from well-constrained focal mechanisms (see, for example, Herrmann and Ammon 1997), yet the link between the stress field and the contemporary seismicity remains enigmatic. In fact, many dramatically different seismic source zones have been proposed and used in the seismic-hazard estimates for the central United States (EPRI 1988; Bernreuter and others, 1989; REI, 1999; Geomatrix Consultants Inc. 2004). Seismic source zones considered in this study are discussed below. 3.1. New Madrid Seismic Zone 3.1.1. New Madrid Faults The New Madrid Seismic Zone is a tightly clustered pattern of earthquake epicenters that extends from northeastern Arkansas into northwestern Tennessee and southeastern Missouri (Fig. 6). Earthquakes along the northeast-trending alignment of earthquakes in northeastern Arkansas and those events in southeastern Missouri between New Madrid and Charleston, Mo., are predominantly right-lateral strike-slip events. The earthquakes along the northwestern trend of seismicity extending from near Dyersburg, Tenn., to New Madrid, Mo., are predominantly thrust events. Focal depths of the earthquakes in the New Madrid Seismic Zone typically range between 5 and 15 km (Chiu and others, 1992). Even though they have been well studied, the locations and maximum magnitude of the New Madrid faults are still uncertain. This is illustrated in the USGS national hazard maps (Frankel and others 1996, 2002). According to Frankel and others (1996), “to calculate the hazard from large events in the New Madrid area we considered three parallel faults in an S-shaped pattern encompassing the area of highest historic seismicity. These are not meant to be actual faults; they are simply a way of expressing the uncertainty in the source locations of large earthquakes such as the 1811–12 sequence. The extent of these fictitious faults is similar to those used in Toro and others (1992). We assumed a characteristic rupture model with a characteristic moment magnitude M of 8.0, similar to the estimated magnitudes of the largest events in 1811–12 (Johnston, 1996a, b). A recurrence time of 1000 years for such an event was used as an average value, considering the uncertainty in the magnitudes of prehistoric events.” These parameters for the New Madrid Seismic Zone were used in the 1996 USGS national hazard maps (Frankel and others 1996). In 12 the 2002 USGS national hazard maps used quite different parameters for the New Madrid Seismic Zone (Frankel and others 2002): ‘The 2002 update incorporates a shorter mean recurrence time for characteristic earthquakes in New Madrid than was used in the 1996 maps, as well as a smaller median magnitude than that applied in 1996. A logic tree was developed for the characteristic magnitude (Mchar) and the configuration of the sources of the characteristic earthquakes, where the uncertainty in location is described by using three fictitious fault sources as in the 1996 maps. A mean recurrence time of 500 years for characteristic earthquakes is used in the calculations (Cramer 2001). This was based on the paleoliquefaction evidence of two to three previous sequences prior to the 1811-12 events (Tuttle and Schweig 2000).” New Madrid Seismic Zone Figure 6. Seismicity between 1974 and 2005 in the central United States. As shown in Figure 7, the northern extension of the New Madrid faults has a significant effect on seismic-hazard estimates at the Paducah Gaseous Diffusion Plant. Although many researchers have postulated that the New Madrid faults probably extend northeast into the Jackson Purchase Region in western Kentucky, even into southern Illinois (Wheeler 1997; REI, 1999), consistent geologic and seismologic evidence indicate that a northwest-trending structure separates the Southern Illinois Seismic Zone from the New Madrid zone (Braile and others 1997; Wheeler 1997). This is evident in Figure 8, which shows the Bouguer gravity anomaly and 1974–94 earthquake epicenters in the New 13 Madrid region (Braile and others 1997). As suggested by Wheeler (1997), the northeast extensions of the New Madrid faults can be substantiated by further seismic network monitoring. Recent studies (Wang and others, 2003a; Anderson and others, 2005; Horton and others, 2005) indicate that the New Madrid faults may not extend northeast into the Jackson Purchase Region. A dense seismic network of nine stations was installed in the Jackson Purchase Region (Fig. 9) in late 2002 (Wang and others 2003b). Table 2 lists the earthquakes recorded by the dense seismic network between January 2003 and June 2005 (Anderson and others 2005). The focal depths of these earthquakes are all less than 10 km. The June 6, 2003, Bardwell, Ky., event (MW4.0) is extremely shallow, only about 2 km, with southeast-northwest maximum compression (Horton and others 2005). These short-period and dense network observations suggest that the characteristics of earthquakes in the Jackson Purchase Region are different from those of earthquakes in the central New Madrid Seismic Zone. Thus, there is no evidence (microseismicity) to support the northeast extension of the New Madrid faults into the Jackson Purchase Region. The study by Baldwin and others (2005) showed that the New Madrid North faults are coincident with current seismicity in southeastern Missouri, which is consistent with the findings of Johnston and Schweig (1996). In addition, detailed coring data collected near the Paducah Gaseous Diffusion Plant show no evidence for Holocene (<11,000 years) displacement along previously interpreted faults underlying the site (William Lettis & Associates Inc. 2006). Thus, no geologic evidence suggests the New Madrid faults extend northeast into the Jackson Purchase Region, particularly near the Paducah Gaseous Diffusion Plant site. For this project, we used the locations of the New Madrid faults determined by Johnston and Schweig (1996), which are consistent with more recent studies (Wang and others, 2003a; Anderson and others 2005; Baldwin and others 2005; Horton and others 2005). 3.1.2. Maximum Magnitude The other large uncertainty for the New Madrid Seismic Zone is the estimate of the maximum magnitude. A single moment magnitude of M8.0 was used in the 1996 national maps (Frankel and others 1996), whereas an Mchar logic tree was used in the 2002 national maps for the New Madrid Seismic Zone: M7.3 (0.15 wt), M7.5 (0.2 wt), M7.7 (0.5 wt), M8.0 (0.15 wt) (Frankel and others, 2002). More recent studies (Hough and others 2000; Mueller and Pujol 2001; Bakun and others 2003) suggest that the magnitude is about M7.2 to 7.5. GPS observations also suggest a similar magnitude (~M7) (Newman and others 1999, Calais and others 2006). Although the uncertainties in the locations of the New Madrid faults and the associated maximum magnitude are large, there is a general agreement among scientists that the location of the New Madrid faults outlined by Johnston and Schweig (1996) is more appropriate for seismic hazard assessment (Cramer 2004; Geomatrix Consultants Inc. 2004; Windeler 2006). Recent studies also suggest that the maximum magnitude for the 14 New Madrid Seismic Zone is lower M7 (Newman and others 1999; Hough and others, 2000; Mueller and Pujol 2001; Bakun and others 2003). In this report, we used the location of the New Madrid faults given by Johnston and Schweig (1996) (Fig. 7) with a mean maximum magnitude of M7.5. As shown in Figure 7, the distance between the site (PGDP) and the New Madrid faults (blue lines) are much shorter than the one between the site and the faults (red lines) used in the national hazard maps (Frankel and others 1996, 2002). PGDP Figure 7. New Madrid faults (Cramer 2004). Pseudo-faults (lines in red) were used in the 1996 and 2002 USGS seismic hazard maps (Frankel and others 1996, 2002). 15 Figure 8. Bouguer gravity anomaly and 1974–94 earthquake epicenters and the New Madrid Rift Complex (Braile and others 1997). Table 2. Parameters of earthquakes (Anderson and others 2005). Date Time 06/06/03 08/26/03 02/12/04 06/20/05 06/20/05 12:29:34 2:26:58 6:49:49 2:00:32 12:21:42 Lat. 36.870 37.100 37.110 36.930 36.920 Long. -88.980 -88.680 -88.960 -88.990 -89.000 16 Depth 2.6 1.9 27.2 9.8 21.0 Magnitude 4 3.1 2.4 2.7 3.6 Depth (UK) 1.5 2.0 9.8 8.7 8.9 Figure 9. Seismic network and earthquakes (stars) recorded between January 2003 and June 2005 in the Jackson Purchase Region (Anderson and others 2005). Triangle – shortperiod seismic station; Circle – strong motion station. 3.2. Wabash Valley Seismic Zone Nuttli and Herrmann (1978) first proposed the Wabash Valley Seismic Zone on the basis of (1) the number of earthquakes, (2) the occurrence of five ≥5 mb,Lg earthquakes in the area between 1875 and 1975, and (3) the presence of the Wabash Valley Fault Zone. The boundaries of the Wabash Valley Seismic Zone as drawn by Wheeler and Frankel (2000) are shown in Figure 10. Also included in Figure 10 are the epicentral locations of the damaging Modified Mercalli Intensity Scale (MMI≥VI) earthquakes in the seismic zone (Stover and Coffman 1993) and the location of the 5.1 mb,Lg September 27, 1909, earthquake (10) that occurred just north of the seismic zone. Dates, times, and epicentral locations of the damaging earthquakes shown in Figure 10 are listed in Table 3. Unlike the seismicity in the New Madrid Seismic Zone, where there is a well-defined pattern, seismicity in the Wabash Valley Seismic Zone is diffuse over a broad area. 17 Despite the number of damaging earthquakes in the Wabash Valley Seismic Zone, there has never been an adequate number of permanent seismic stations in the seismic zone to derive well-constrained focal depths or focal mechanisms. As previously indicated, of the 18 events listed in Table 3, the only events for which well-determined focal depths and focal mechanisms have been estimated are events 15 through 18. These four earthquakes were large enough to generate sufficient surface-wave data that their focal depths and focal mechanisms could be estimated using the radiation pattern of their Rayleigh and Love waves (Herrmann and Ammon 1997). Table 3. Damaging earthquakes in the Wabash Valley Seismic Zone. 1. 2. 3. Event No. Date (Mo-Day-Yr) Time (GMT) Lat./Long. (°N/°W) Magnitude mb,Lg1 Mw2 Depth3 (km) 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. July 5, 1827 Aug. 7, 1827 Aug. 7, 1827 Sep. 25, 1876 Sep. 25, 1876 Feb. 6, 1887 July 27, 1891 Sep. 27, 1891 Apr. 30, 1899 Sep. 27, 1909 Nov. 27, 1922 Apr. 27, 1925 Sep. 2, 1925 Nov. 8, 1958 Nov. 9, 1968 Apr. 3, 1974 June 10, 1987 June 18, 2002 4:30 7:00 6:00 6:15 22:15 2:28 4:55 2:05 9:45 3:31 4:05 11:56 2:41 17:01 23:05 23:48 18:37 38.0/87.5 38.0/88.0 38.0/88.0 38.5/87.8 38.5/87.8 38.7/87.5 37.9/87.5 38.25/88.5 38.5/87.4 39.8/87.2 37.8/88.5 38.2/87.8 37.8/87.5 38.44/88.01 37.91/88.37 38.55/88.07 38.71/87.95 37.98/87.78 4.8 4.8 4.7 4.5 4.8 4.6 4.1 5.5 4.9 5.1 4.8 4.8 4.6 4.4 5.5 4.5 5.1 4.9 22 14 10 17-19 4.4 4.4 4.3 4.1 4.4 4.2 3.7 5.3 4.6 4.8 4.4 4.4 4.2 4.0 5.3 4.3 5.0 4.5 Magnitudes (mb,Lg) are from Stover and Coffman (1993) except for events 8 and 15. The 5.5 mb,Lg for event 17, the November 9, 1968, southern Illinois event, is more generally accepted than the 5.3 mb,Lg given by Stover and Coffman (1993). The mb,Lg magnitude, seismic moment, and epicentral location for event 18 are preliminary estimates based on data from the University of Kentucky Seismic and Strong-Motion Network and R. Herrmann at St. Louis University (personal communication). Except for events 15, 16, and 17, moment magnitudes (Mw) were derived using the mb to seismic moment (Mo) to moment magnitude conversion. Moment magnitudes of events 17, 18, and 19 were calculated using the seismic moments given in Herrmann and Ammon (1997). Focal depths are from Herrmann and Ammon (1997), except for event 18, which is based on a personal communication from R.B. Herrmann. 18 18 Figure 10. Epicentral locations of the felt earthquakes in the Wabash Valley Seismic Zone. The largest instrumentally recorded historical earthquake in the Wabash Valley Seismic Zone is the November 9, 1968, earthquake (event 15 in Table 3). McBride and others (2002) believed that the November 9, 1968, earthquake occurred as a result of the reactivation of a fault plane within a series of moderately dipping lower-crustal reflectors that are decoupled from the overlying Paleozoic structure. The June 18, 2002, Darmstadt, Ind., earthquake (M4.6) was well located (Table 3). Kim (2003) also believed that the June 18, 2002, earthquake occurred as a result of the reactivation of a fault within the Wabash Valley Fault System (Fig. 11). The Wabash Valley Fault System (Fig. 11) is a series of north–northeast-trending normal faults with right-lateral offsets across the Herald-Pillipstown and the New Harmony Faults. The locations and extent of faulting are well known from the extensive set of drill logs and seismic-reflection lines acquired for oil and gas exploration. The Albion- 19 Ridgeway and New Harmony Faults is the Grayville Graben, so named by Sexton and others (1996) and shown by Bear and others’ (1997) as exhibiting Cambrian extensional slip. Based on Bear and others’ (1997) interpretation of the fault movement, Wheeler and Cramer (2002) identified the Grayville Graben as Iapetan and considered the graben and the Wabash Valley Fault System non seismogenic. Woolery (2005) found that the Hovey Lake Fault (one of the Wabash Valley faults) moved as recently as approximately 37,000 years before the present. Figure 11. Earthquakes and faults in the lower Wabash Valley. As discussed above, there is no clear evidence to directly link any of the earthquakes in the Wabash Valley Seismic Zone to a specific fault. Thus, the Wabash Valley Seismic Zone was treated as an areal source in the seismic hazard analyses (Frankel and others, 1996, 2002; Wheeler and Frankel 2000). The maximum magnitude of M7.5 was assigned to the zone in the national seismic hazard maps (Frankel and others 1996, 2002; Wheeler and Frankel, 2000), and was based on the magnitude estimates from paleoliquefaction studies by Obermeier and others (1991, 1993), Munson and others 20 (1995, 1997), and Pond and Martin (1997). Recent studies by Street and others (2004) and Olson and others (2005), however, suggest that the best estimates of those paleoearthquakes are in the range of 6.2 to 7.3. The Tri-State Seismic Source Zone, one of the alternative source zones suggested by Wheeler and Cramer (2002) for the Wabash Valley Seismic Zone, was used in this study. We assigned a mean maximum magnitude of M6.8 to the Wabash Valley Seismic Zone (Fig. 12) based on these studies (Street and others, 2004; Olson and others 2005). Figure 12. New Madrid faults and Wabash Valley Seismic Zone. 3.3. Background Seismicity Earthquakes have occurred throughout Kentucky and surrounding states, many of them not associated with any known seismic zone or geologic/tectonic feature. For example, the February 28, 1854, earthquake (mb,Lg4.0) in central Kentucky is not associated with any known seismic zone. Many earthquakes were recorded by the University of Kentucky Seismic and Strong-Motion Network since 1984 (Street and Wang 2003). These earthquakes were called background seismicity (Street and others, 1996). Contribution to seismic hazard from the background seismicity was considered with smoothed spatial seismicity at grid points in the central and eastern United States (Frankel and others, 1996, 2002) (Fig. 13). A uniform background zone (Fig. 14) was also considered to account for the large earthquakes in the central and eastern United States (Frankel and others 1996, 2002). Although magnitude is large (M7.0 and 7.5), the 21 large background earthquakes have no contribution to the seismic hazard because of (1) a large-area source zone and (2) a longer recurrence interval (more than 10,000 years) in the national seismic hazard maps (Wang 2003a). Therefore, the use of these large background earthquakes is not necessary (Wang 2003a). Figure 13. Seismic sources that were considered in the national seismic hazard maps (Frankel and others 2002). Figure 14. Background earthquakes (Mmax) used in the national seismic hazard maps (Frankel and others 2002). 22 In this study, we adopted a method used by Street and others (1996). Based on historical and instrumental records, Street and others (1996) proposed a mean maximum magnitude for the background seismicity in the eight counties in western Kentucky (Ballard, Carlisle, Fulton, Graves, Hickman, Livingston, Marshall, and McCracken) of 5.3 mb,Lg (M5.0) (Figure 15). This magnitude is based on moderate-size historical events, and occasional events recorded by the University of Kentucky Seismic and Strong-Motion Network (Wang and others, 2003a). Within the eight counties, many earthquakes measuring 3.0 mb,Lg or larger have been recorded, such as the Bardwell, Kentucky earthquake (M4.0), which caused some damage. The focal depths for the small earthquakes in the area are generally in the range of 5 to 20 km. Assuming an epicentral distance of 10 km and focal depth of 10 km, the shortest distance from this local source is 14 km. The shortest distance of 15 km was used for this project. A point source of M5.0 earthquake with a distance of 15 km was considered to account for hazard contribution from the background seismicity. mb,Lg5.3 (EE) mb,Lg5.5 (PE) mb,Lg5.7 (MCE) mb,Lg5.3 mb,Lg4.5 mb,Lg4.7 5 Figure 15. Maximum background earthquakes in Kentucky (Street and others 1996). 23 4.0. Magnitude-Recurrence Relationship In the central United States, seismicity rate is relatively low in comparison with California and there are no instrumental recordings on strong and large earthquakes (M>6.0). There are only two post 1811-1812 strong events (6.0<M<6.5): the 1843 Marked Tree, Ark., and the 1895 Charleston, Mo., earthquake (M6.0). Bakun and others (2003) recently suggested that the 1895 Charleston, Mo., earthquake was located in southern Illinois, about 100 km north of Charleston (not in the New Madrid Seismic Zone). The 1811-12 New Madrid earthquakes were great events (7.0<M<8.0) and are of safety concern in the area. The instrumental and historical records are insufficient to construct the magnitude-occurrence relationships in the central United States, so prehistoric records (paleoliquefaction) had to be used (Tuttle and others 2002) (Frankel and others 1996, 2002). Figures 16 and 18 show the magnitude-occurrence relationships for the New Madrid Seismic Zone (Frankel and others, 1996) and the Wabash Valley Seismic Zone (Wheeler and Cramer 2002) based on instrumental, historical, and paleoliquefaction records. 4.1. New Madrid Seismic Zone As shown in Figure 16, the annual rate derived from instrumental and historical earthquakes is not consistent with that derived from paleoliquefaction records. Figure 16 also shows that there is a lack of strong earthquakes of M6.0 to 7.0, or an earthquake deficit, in the New Madrid Seismic Zone. A b-value of 0.95 was used in the USGS national seismic hazard mapping for the central United States (Frankel and others, 1996, 2002). Based on the a and b values determined from instrumental and historical records, the annual occurrence rate of a M7.5 earthquake is less than 0.0001 (recurrence interval is longer than 10,000 years) in the New Madrid Seismic Zone (Fig. 16). Paleoliquefaction records, however, reveal an annual occurrence rate of about 0.002 (recurrence interval of about 500 years) for a M7.5 earthquake in the New Madrid Seismic Zone. A recent study by Holbrook and others (2006) suggests that earthquakes may be temporally clustered on millennial scales and that these large earthquakes have been treated as characteristic events (Frankel and others, 1996, 2002; Geomatrix Consultants Inc. 2004). Table 4 lists instrumental and historical earthquakes with magnitude equal to or greater than M4.0 known to have occurred in the New Madrid Seismic Zone (Bakun and Hopper, 2004). Figure 17 shows the Gutenberg-Richter curve for earthquakes with magnitudes between M4.0 and M5.0 in the New Madrid Seismic Zone (Table 4). The a and b values were estimated to be about 3.15 and 1.0, respectively, from earthquakes with magnitudes between M4.0 and M5.0 (Fig. 17). The b value of 1.0 is consistent with that used in the national seismic hazard maps (Frankel and others, 1996, 2002) (Fig. 16). As shown in Figures 16 and 19, if the a and b values are used to extrapolate large earthquakes (M≥6.0) in the New Madrid Seismic Zone, the recurrence interval for large earthquake would be quite long, about 700 years for M6.0, 7,000 years for M7.0, and 70,000 years for M8.0 earthquakes. This is why the large earthquakes (M≥7.0) in the New Madrid Seismic Zone are treated as characteristic. In this study, we assigned a magnitude of M7.5 with a mean recurrence interval of 500 to 1,000 years for the characteristic event along the New 24 Madrid faults. The mean recurrence interval of 500 to 1,000 years is based on the geological studies (Tuttle and others 2002; Holbrook and others 2006). Figure 16. Magnitude-frequency relationship in the New Madrid Seismic Zone (Frankel and others 1996). 1.0000 Log (N) 0.1000 0.0100 0.0010 0.0001 4.0 5.0 6.0 magnitude 7.0 8.0 Figure 17. Magnitude-frequency (Gutenburg-Richter) curve for the New Madrid Seismic Zone. Diamond–historical rate, triangle–geological rate. 25 Table 4. Earthquakes with magnitude equal to or greater than M4.0 in the New Madrid Seismic Zone (Bakun and Hopper 2004). Date 1811-12-16 1811-12-16 "dawn" 1812-01-23 1812-02-07 1843-01-05 1843-02-17 1865-08-17 1878-11-19 1883-01-11 1903-11-04 1923-10-28 1927-05-07 1938-09-17 1962-02-02 1963-03-03 1970-11-17 1976-03-25a 1976-03-25b 1991-05-04 2003-04-30 2003-06-06 Latitude 36.00 36.25 36.80 36.30 35.90 35.90 35.54 35.65 36.80 36.59 35.54 35.65 35.55 36.37 36.64 35.86 35.59 35.60 36.56 35.920 36.87 Longitude -89.96 -89.50 -89.50 -89.40 -89.90 -89.90 -90.40 -90.25 -89.50 -89.58 -90.40 -90.25 -90.37 -89.51 -90.05 -89.95 -90.48 -90.50 -89.80 -89.920 -88.98 M 7.6 7.0 7.5 7.8 6.2 4.2 4.7 5.0 4.2 4.7 4.1 4.5 4.4 4.2 4.7 4.1 4.6 4.2 4.1 4.0 4.0 4.2. Wabash Valley Seismic Zone The paleoliquefaction studies by Obermeier and others (1991, 1993), Munson and others (1995, 1997), and Pond and Martin (1997) suggest a mean recurrence interval of about 5,000 years for the large prehistoric earthquakes in the Wabash Valley Seismic Zone. As shown in Figure 18, this recurrence interval is consistent with the intervals projected from the seismicity of small and moderate earthquakes (≤M5.0) (Wheeler and Cramer 2002). Figure 19 shows the Gutenburg-Richter curve for the Wabash Valley Seismic Zone based on Bakun and Hopper (2004) data (a=3.0, b=1.0). We derived a mean recurrence interval of about 4,000 years for an earthquake with magnitude of M6.8 or greater from Figure 19. This recurrence interval is consistent with the geologic data (Obermeier and others 1991, 1993; Munson and others, 1995, 1997; Pond and Martin, 1997) and was used for the Wabash Valley Seismic Zone in this report. 4.3. Background Seismicity The occurrence frequency of the maximum earthquake for the background earthquake was determined from the earthquakes with magnitude greater than M2.5 surrounding the Paducah Gaseous Diffusion Plant (Fig. 20). This is similar to the smoothed seismicity that was used in the national seismic hazard maps (Frankel and others 2002). The a and b 26 values were estimated to be 2.56 and 0.97, respectively (Fig. 21). The mean recurrence interval is projected to be about 200 years for an M5.0 earthquake. Figure 18. Magnitude-frequency relationship in the Wabash Valley Seismic Zone (Wheeler and Cramer 2002). 1.00000 0.10000 N 0.01000 0.00100 0.00010 0.00001 3.00 4.00 5.00 6.00 7.00 8.00 Magnitude Figure 19. Magnitude-frequency (Gutenburg-Richter) curve for the Wabash Valley Seismic Zone. Diamond–historical rate, triangle–geological (paleoliquefaction) rate. 27 PGDP Figure 20. Recorded earthquakes with magnitude greater than 2.5 surrounding the Paducah Gaseous Diffusion Plant between 1978 and 2006. 10.000 N 1.000 0.100 0.010 0.001 2 3 4 Magnitude 5 6 Figure 21. Magnitude-frequency (Gutenburg-Richter) curve for the background seismicity. 28 5.0. Ground-Motion Attenuation Relationship As shown in Figure 2, the ground-motion attenuation relationship describes a spatial relationship between a ground-motion parameter (i.e., PGA, PGV, MMI, or PSA at different periods) and earthquake magnitude and source-to-site distance with uncertainty (equation [2] or [3]). This can be demonstrated through the following example of how the ground-motion attenuation relationship is modeled. Figure 22 shows horizontal uncorrected PGA vs. distance to the fault (RRUP) and five ground-motion attenuation relationships for the Parkfield earthquake of September 28, 2004 (Shakal and others 2006). Figure 23 shows strong-motion stations and accelerograms recorded in the 2004 M6.0 Parkfield earthquake for the east-west component (Shakal and others 2006). As shown in Figure 23, source-to-site distance is measured as the shortest distance to the fault rupture (RRUP), not the epicentral distance (REPI). Figure 23 also shows that the epicentral distances are quite different from the rupture distances. RRUP is about 4 and 2 km for stations FZ11 and FZ16, and REPI is about 10 and 15 km, respectively. As shown in Figure 22, a different set of parameters (i.e., f(M,R) and σln,Y) would result if the epicentral distance was used (blue diamond). This shows that the ground-motion attenuation relationship, equation [2] or [3], depends on how earthquake source (i.e., point vs. finite), source-to-site distance (i.e., RRUP, RJB, REPI, or RHYP), and site conditions (i.e., rock vs. soil) are considered. In addition, many different functional forms are being used by different modelers. For example, Atkinson and Boore (2006) used the following functional form on hard rock for the central and eastern United States: log( PSA) = c1 + c 2 M + c3 M 2 + (c 4 + c5 M ) f1 + (c6 + c7 M ) f 2 + (c8 + c9 M ) f 0 + c10 Rcd + nσ log,PSA (10) where f0 = max (log(R0/Rcd), 0); f1 = min (log Rcd, log R1); f2 = max (log (Rcd/R2), 0); Rcd = the closet distance to the fault (RRUP); R0 = 10 km; R1 = 70 km; R2 = 140 km. And Silva and others (2002) used the functional form: ln(Y ) = c1 + c 2 M + (c6 + c7 M ) ln(R + e c4 ) + c10 ( M − 6) 2 + nσ ln,Y (11) where R is the closest distance to the surface projection of the rupture surface (RJB). Therefore, the ground-motion attenuation relationship depends not only on the functional form and associated constants being used, but also on how earthquake source (i.e., point vs. finite), source-to-site distance (i.e., RRUP, RJB, REPI, or RHYP), and site conditions (i.e., rock vs. soil) are considered. In other words, there may be a dependency between the statistical parameters (i.e., constants and standard deviation) and the variables (i.e., M and R). In fact, many researchers (Youngs and others 1995, 1997; Abrahamson and Silva 29 1997; Sadigh and others, 1997; Toro and others 1997; Campbell 2003; Akkar and Bommer 2007) have found that ground-motion uncertainty depends on M or R, or both. As discussed earlier, however, ground-motion uncertainty is treated as an independent random variable in PSHA (Cornell 1968, 1971; McGuire 1976, 1995, 2004). The dependency between the statistical parameters in the ground-motion attenuation relationship needs to be explored further, because it may have a significant impact on hazard estimates (Carroll 2003; Wang and Zhou in press). FZ16 FZ11 Figure 22. Horizontal uncorrected PGA vs. distance to the fault for the Parkfield earthquake of September 28, 2004 (Shakal and others 2006). Blue diamonds are plots for stations FZ11 and FZ16 if the epicentral distance is measured. 30 RRUP REPI REPI Figure 23. Strong-motion stations and accelerograms recorded in the 2004 M6.0 Parkfield earthquake for the east-west component (Shakal and others 2006). RRUP - the closet distance to fault rupture; REPI – epicentral distance. 31 One of the fundamental differences between assessing seismic hazard in the western and central United States is in the ground-motion attenuation relationship (Wang and others 2005). The attenuation relationships developed for California are based on observations, such as those by Abrahamson and Silva (1997), Boore and others (1997), Sadigh and others (1997), Boore and Atkinson (2006), Campbell and Bozorgnia (2006), and Chiou and Youngs (2006). Figure 24 shows the world-wide data used for development of ground-motion attenuation relationship for the PEER-Lifelines Next Generation Attenuation of Ground Motion (NGA) Project by Chiou and Youngs (2006). In contrast, all the attenuation relationships currently available for the central United States are based on theoretical models with very limited observations (Frankel and others, 1996; Toro and others, 1997; Somerville and others, 2001; Silva and others, 2002; Campbell, 2003; EPRI, 2003; Atkinson and Boore, 2006). Figure 25 shows the simulated data being used for the ground-motion attenuation analysis by Atkinson and Boore (2006) for the central and eastern United States. This significant difference results in differences in ground-motion uncertainties (in both median and standard deviation) for the central United States. As shown by Frankel (2004), the median ground motions for California vary only slightly between proposed attenuation relationships. For example, PGA ranges from 0.30 to 0.38g between four attenuation relationships for a M7.8 earthquake at 15 km in San Francisco (Frankel, 2004). For comparison, Table 5 lists the median ground motions (PGA) for a M7.7 earthquake at 15 km from the New Madrid Seismic Zone with five attenuation relationships. The range of the median PGA in the central United States is between 0.69 and 1.20g. Similarly, Frankel (2004) showed a large range of median ground motions, especially near-source (<30 km). The theoretical models predict higher median ground motions (PGA and 5 Hz S.A.) in the central United States than the ones in the west for a similar earthquake. Thus, the theoretical models predict not only higher median ground motion in comparison with a similar magnitude in the West, but also a larger range (uncertainty). Some theoretical models also predict higher standard deviations in the central and eastern United States than in the west. Recent studies suggest that the standard deviation should be similar in the two regions (Atkinson and Boore 2006). Table 5. Median ground motions for a M7.7 New Madrid earthquake at 15 km for a hardrock site from several attenuation relationships. Frankel and Toro and Atkinson and Campbell Somerville and others (1996) others Boore (1995) (2003) others (2001) (1997) PGA (g) 1.20 0.90 0.90 0.91 0.69 32 Figure 24. Magnitude-distance-region distribution of selected recordings (Chiou and Youngs 2006). 33 Figure 25. Log values of horizontal component 5% pseudo-acceleration at frequencies 0.5, 1, and 5 Hz, and PGA, for rock sites in eastern North America. Dots show PSA from simulations, including aleatory uncertainty, for M 5 (light) and M 8 (dark). Solid lines show predicted amplitudes from regression equations developed from a simulated database for M 5, 6, 7, and 8 (Atkinson and Boore 2006). Use of different attenuation relationships will result in different ground-motion estimates, for near-source (10 to 30 km) in particular. As stated by Frankel and others (2002), “significant differences between the 1996 and 2002 maps are caused by the inclusion of additional attenuation relations in the 2002 maps. In 1996, we used the attenuation relations of Toro et al. (1997) and Frankel et al. (1996), which were assigned equal weight. For the 2002 maps we have added the attenuation relations of Atkinson and Boore (1995), Somerville et al. (2001) and Campbell (2003).” As concluded by SSHAC (1997), “one key source of difficulty is failure to recognize that 1) there is not likely to be ‘consensus’ (as the work is commonly understood) among the various experts and 2) no 34 single interpretation concerning a complex earth-sciences issue is the ‘correct’ one.” There is no consistent or unique way to chose ground-motion attenuation relationships for seismic hazard analysis. Recent studies have shown that ground motion at near-source has been over-predicted, however (USGS/NRC Workshop 2005; Atkinson and Boore 2006), even on the West Coast, where ground motion was overly predicted at near-source (Abrahamson, 2006; Boore and Atkinson, 2006; Campbell and Bozorgnia, 2006; Chiou and Youngs 2006). There is a consensus that many current attenuation relationships predict too high ground motion at near-source, particularly Frankel and others’(1996) attenuation relationship, in the central and eastern United States (USGS/NRC Workshop 2005). Figure 26 shows some of the ground-motion attenuation relationships for a M7.5 earthquake in the central United States. As shown in the figure, the Frankel and others (1996) attenuation relationship predicts higher PGA at near-source between 10 and 50 km. Figure 27 shows some of the ground-motion attenuation relationships for a M5.0 earthquake in the central United States. In this report, we used the ground-motion attenuation relationships of Somerville and others (2001), Silva and others (2002), Campbell (2003), and Atkinson and Boore (2006). These attenuation relationships represent different approaches (i.e., finite source/green function, double-corner, and hybrid methods). Figures 28 and 29 show 0.2s and 1.0s response accelerations of the four attenuation relationships for a M7.5 earthquake in the central United States. The rupture distance is used throughout this report. 35 10.00 1.00 PGA (g) Silva-DC-S Somerville 0.10 AB06 Campbell03 Frankel 0.01 0.00 1 10 100 Distance (km) 1000 Figure 26. PGA attenuation relationships at hard rock for an M7.5 earthquake in the central United States. 10.00 1.00 PGA (g) Silva-DC-S 0.10 AB06 Campbell03 0.01 0.00 1 10 100 1000 Distance (km) Figure 27. PGA attenuation relationships at hard rock for a M5.0 earthquake in the central United States. 36 10.00 0.2s PSA (g) 1.00 Silva-DC-S Somerville 0.10 AB06 Campbell03 0.01 0.00 1 10 100 Distance (km) 1000 Figure 28. The 0.2s PSA attenuation relationships used in this study at hard rock for a M7.5 earthquake in the central United States. 1.00 0.10 1.0s PSA (g) Silva-DC-S Somerville AB06 Campbell03 0.01 0.00 1 10 100 Distance (km) 1000 Figure 29. The 1.0s PSA attenuation relationships used in this study at hard rock for a M7.5 earthquake in the central United States. 37 6.0. Results Three seismic sources affect the Paducah Gaseous Diffusion Plant: the New Madrid faults, Wabash Valley Seismic Zone, and small earthquakes nearby (Fig. 30). The mean distances from the plant to the New Madrid faults and Wabash Valley Seismic Zone are 40 and 60 km, respectively. The source-to-site distance from the New Madrid faults is treated as characteristic, and is similar to the characteristic source used in the national hazard mapping and other studies (Frankel and others 1996, 2002; Geomatrix Consultants Inc. 2004). The Wabash Valley Seismic Zone is a large areal source. As shown in Figure 20, local earthquakes around Paducah may also contribute to the hazard. In this project, we used a point source at 15 km with a maximum magnitude of M5.0 (Fig. 30) to account for local earthquakes for the Paducah Gaseous Diffusion Plant. Mmax=6.8 TC=4,000 yrs. Background Seismicity Mmax=5.0 R=15 km Mmax=7.5 TC=500-1000 yrs. Figure 30. Seismic sources for the Paducah Gaseous Diffusion Plant. 38 6.1. PSHA Results The ground-motion uncertainty is inherently a part of PSHA, and other uncertainties, such as fault location, which are treated with logic trees, by which different weights are assigned manually to a set of expert estimates for each input parameter (SSHAC 1997). In this project, the weights in Table 6 were used to account for the uncertainties in location, magnitude, recurrence-interval, and attenuation relationship. It has been shown that the ground-motion hazard at site in the New Madrid Seismic Zone can be estimated with a single equivalent magnitude and distance (Frankel 2004). The de-aggregation analysis also shows that ground motion hazard in Paducah can be approximated by a single equivalent magnitude and distance (Petersen, 2005). Although this analysis (Table 6) is not a standard PSHA, it can provide a good estimate (Frankel 2004; Petersen 2005) and is easy to understand. The hazard curves for PGA, 0.2s PSA, and 1.0s PSA are shown in Figures 31 through 33. Table 7 lists ground-motion values on hard rock at several annual probabilities of exceedance at the Paducah Gaseous Diffusion Plant. Table 6. Input parameters and weights used here in our PSHA for the Paducah Gaseous Diffusion Plant. Source Mmax Recurrence Attenuation Distance (km) (mean) interval (yrs.) (mean) (mean) NMSZ 7.5 500 (0.75) 40 AB-06 (0.25) (characteristic) 1,000 (0.25) Campbell-03 (0.25) Silva-DC-S (0.25) Somerville (0.25) WVSZ 6.8 4000 (1.0) 60 AB-06 (0.25) (areal) Campbell-03 (0.25) Silva-DC-S (0.25) Somerville (0.25) 5.0 200 (1.0) 15 AB-06 (0.33) Background Seismicity Campbell-03 (0.33) (point) Silva-DC-S (0.33) Table 7. Mean ground-motion hazards on hard rock at the Paducah Gaseous Diffusion Plant. Ann. Prob. Exc. Return Period PGA 0.2s PSA 1.0s PSA (years) (g) (g) (g) 0.004 250 0.09 0.10 0.01 0.002 500 0.18 0.21 0.03 0.001 1,000 0.29 0.40 0.09 0.0004 2,500 0.49 0.68 0.16 0.0002 5,000 0.62 0.90 0.23 39 PGA Hazard Curve 1.0E-02 Ann. Prob. of Exc. 1.0E-03 1.0E-04 1.0E-05 1.0E-06 1.0E-07 1.0E-08 0.01 0.1 PGA (g) 1 10 Figure 31. Mean PGA hazard curve on hard rock at the Paducah Gaseous Diffusion Plant. 0.2s PSA Hazard Curve 1.0E-02 Ann. Prob. of Exc. 1.0E-03 1.0E-04 1.0E-05 1.0E-06 1.0E-07 1.0E-08 0.01 0.1 1 10 0.2s PSA (g) Figure 32. Mean 0.2s PSA hazard curve on hard rock at the Paducah Gaseous Diffusion Plant. 40 1.0s PSA Hazard Curve 1.0E-02 Ann. Prob. of Exc. 1.0E-03 1.0E-04 1.0E-05 1.0E-06 1.0E-07 1.0E-08 0.01 0.1 1.0s PSA (g) 1 10 Figure 33. Mean 1.0s PSA hazard curve on hard rock at the Paducah Gaseous Diffusion Plant. 6.2. DSHA Results Table 8 lists the median PGA values for the three sources affecting the Paducah Gaseous Diffusion Plant (Fig. 30), using the four attenuation relationships. As shown in Table 8, the characteristic earthquake in the New Madrid Seismic Zone dominates the hazard at the plant. Tables 9, 10, and 11 list PGA and 0.2s and 1.0s PSA hazards at the Paducah Gaseous Diffusion Plant from the characteristic earthquake in the New Madrid Seismic Zone for the four ground-motion attenuation relations (Somerville and others 2001; Silva and others 2002; Campbell 2003; Atkinson and Boore 2006). The return period for these ground motions is about 500 to 1,000 years, the same as the recurrence interval of the characteristic earthquake in the New Madrid Seismic Zone. 41 Table 8. Median PGA on hard rock at the Paducah Gaseous Diffusion Plant from the three seismic sources. Source AB-06 Campbell-03 Silva-DC-S Somerville (g) (g) (g) (g) NMSZ 0.14 0.28 0.29 0.29 WVSZ 0.04 0.09 0.11 0.10 Background 0.11 0.17 0.10 n/a Seismicity Table 9. PGA at the Paducah Gaseous Diffusion Plant from the characteristic earthquake in the New Madrid Seismic Zone. Median (g) Median +1σln,y Median+2σln,y 1.5 Median (g) (g) (g) AB-06 0.14 0.28 0.56 0.21 Campbell-03 0.28 0.55 1.08 0.42 Silva-DC-S 0.29 0.67 1.55 0.44 Somerville 0.29 0.52 0.94 0.44 Average 0.25 0.51 1.03 0.38 Table 10. The 0.2s PSA at the Paducah Gaseous Diffusion Plant from the characteristic earthquake in the New Madrid Seismic Zone. Median (g) Median+1σln,y Median +2σln,y 1.5 Median (g) (g) (g) AB-06 0.23 0.46 0.92 0.35 Campbell-03 0.40 0.82 1.68 0.60 Silva-DC-S 0.43 0.99 2.29 0.65 Somerville 0.51 0.93 1.71 0.77 Average 0.39 0.80 1.65 0.59 Table 11. The 1.0s PSA at the Paducah Gaseous Diffusion Plant from the characteristic earthquake in the New Madrid Seismic Zone. Median (g) Median+1σln,y Median +2σln,y 1.5 Median (g) (g) (g) AB-06 0.07 0.14 0.28 0.11 Campbell-03 0.15 0.31 0.65 0.23 Silva-DC-S 0.09 0.21 0.51 0.14 Somerville 0.15 0.30 0.60 0.23 Average 0.12 0.24 0.51 0.18 42 7.0. Conclusion and Recommendations Estimating seismic hazard at the Paducah Gaseous Diffusion Plant is difficult because of the lack of instrumental ground-motion observations from large earthquakes in the region. Three seismic sources (i.e., the New Madrid Seismic Zone, the Wabash Valley Seismic Zone, and background seismicity) were characterized based on currently available information on geology and seismology in the central United States. Four ground-motion attenuation relationships were chosen and used for evaluating groundmotion hazard on hard rock at the plant. Probabilistic seismic hazard analysis and deterministic seismic hazard analysis were performed for the plant. Table 12 lists ground-motion hazards derived from PSHA at several commonly considered return periods. Table 13 lists ground-motion hazards with associated uncertainty derived from DSHA. Table 12. Ground-motion hazards on hard rock at the Paducah Gaseous Diffusion Plant determined by PSHA. Ann. Prob. Return Period Exc. Prob. in 50 PGA 0.2s PSA 1.0s PSA Exc. (years) years (%) (g) (g) (g) 0.004 250 18 0.09 0.10 0.01 0.002 500 10 0.18 0.21 0.03 0.001 1,000 5 0.29 0.40 0.09 0.0004 2,500 2 0.49 0.68 0.16 0.0002 5,000 1 0.62 0.90 0.23 Table 13. Ground motion hazards on hard rock at the Paducah Gaseous Diffusion Plant determined by DSHA. Average 1.5 Average Median Average Median Average Median (g) (g) +1σln,y (g) Median +2σln,y (g) PGA 0.25 0.51 1.03 0.38 0.2s PSA 0.39 0.80 1.65 0.59 1.0s PSA 0.12 0.24 0.51 0.18 These results show that PSHA and DSHA utilize the same geological and seismological parameters, but produce quite different estimates of ground motion at the Paducah Gaseous Diffusion Plant because of the differences in defining the seismic hazard. In PSHA, seismic hazard is defined as the return period (or annual probability of exceedance) having a ground motion larger than a specific value. PSHA calculates seismic hazard from all earthquake sources in consideration, and incorporates uncertainty in earthquake size and location and ground motion implicitly. In DSHA, seismic hazard is defined as the ground motion(s) from a single or several earthquakes that have maximum values (impacts) at a site. DSHA emphasizes the ground motion from an individual earthquake, such as the maximum credible earthquake or maximum probable earthquake, and explicitly determines ground-motion hazard with a level of uncertainty. 43 What level of ground motion should be considered for engineering design of a facility at the Paducah Gaseous Diffusion Plant? The answer to this question is complicated and depends on many factors, such as which methodology is used, what type of facility is being considered, and what environment is being considered. There should be a scientific basis in selecting a design ground motion. It is well understood that large earthquakes, similar to the 1811-1812 New Madrid events, in the New Madrid Seismic Zone pose the biggest hazard in the central United States, at the Paducah Gaseous Diffusion Plant in particular. This study shows that the best estimate (mean) of PGA is about 0.25g at the Paducah Gaseous Diffusion Plant from the New Madrid earthquakes (Table 13). This estimate is consistent with the limited MMI data (Fig. 34). Figure 34 shows that the Paducah Gaseous Diffusion Plant site experienced a MMI VIII intensity, which is equivalent to a PGA of 0.20 to 0.30 g (Bolt 1993; Atkinson and Kaka 2007). This suggests that the PGA level of 0.25 to 0.3g would be appropriate for engineering design of ordinary buildings and facilities at the site and surrounding areas. Therefore, the ground motion with 1,000-year return period, derived from PSHA (Table 12), would be appropriate for engineering design of ordinary buildings and facilities. This is why the ground motion with 1,000-year return period, produced by the US Geological Survey (Frankel and others 1996), was proposed and selected as the basis for seismic design of residential buildings in western Kentucky (SEAOK 2002). The ground motion with 1,000-year return period has also been considered as the upper level ground motion for seismic retrofit of highway structures in the central and eastern United States (FHWA 2006). For other important facilities, the DSHA ground motion with one standard deviation (0.51g PGA) might be considered (Table 13). This ground motion (0.51g PGA) is similar to the ground motion (0.49g PGA) with a 2,500-year return period derived from PSHA (Table 12). Table 14 lists the recommended ground motions for design consideration for facilities at the Paducah Gaseous Diffusion Plant. The results from our PSHA are consistently lower than those from the national seismic hazard maps (Frankel and others 2002) and the site-specific study by REI (1999) at the same return periods (Table 15). These differences result from the difference of the input parameters, particularly the location of the New Madrid faults (Fig. 7), a smaller mean magnitude (M7.5) for the characteristic earthquake in the New Madrid Seismic Zone, and use of a lower ground motion attenuation relationship. Table 14. Recommended ground motions on hard rock at the Paducah Gaseous Diffusion Plant. PSHA Facility DSHA Ordinary Important Median Median + one standard deviation Return Period (years) 1,000 Exc. Prob. in 50 years (%) 5 2,500 2 44 PGA (g) 0.2s PSA (g) 1.0s PSA (g) 0.27 0.40 0.10 0.50 0.80 0.20 Figure 34. Isoseismal map of the February 7, 1812, New Madrid earthquake (Hough and others 2000). Table 15. Comparison of mean PGA estimates on hard rock at the Paducah Gaseous Diffusion Plant determined from PSHA. Return Period This study USGS -20021) REI –1999 (years) (g) (g) (g) 250 0.09 0.08 0.10 500 0.18 0.24 0.20 1,000 0.29 0.55 0.38 2,500 0.49 0.95 0.78 5,000 0.62 1.24 1.15 1) USGS values were converted from PGA for soft rock by a factor of 1.52. 45 8.0. References Abrahamson, N., 2006, Lesson learned from ground rupture and strong motion [abstract]: Seismological Research Letters, v. 77, p. 225. Abrahamson, N.A., and Bommer, J.J., 2005, Probability and uncertainty in seismic hazard analysis: Earthquake Spectra, v. 21, p. 603–607. Abrahamson, N.A., and Silva, W.J., 1997, Empirical response spectral attenuation relations for shallow crustal earthquakes: Seismological Research Letters, v. 68, p. 94-127. Akkar, S., and Bommer, J.J., 2007, Empirical prediction equations for peak ground velocity derived from strong-motion records from Europe and the Middle East: Bulletin of the Seismological Society of America, v. 97, 511–532. Anderson, C., Wang, Z., and Woolery, E.W., 2005, Observed Seismicity in the Jackson Purchase Region of Western Kentucky between January 2003 and June 2005: the 77th Annual Meeting of the Eastern Section of the Seismological Society of America, Memphis, Tennessee, October 3-4, 2005. Atkinson, G.M., and Kaka, S.I., 2007, Relationship between felt intensity and instrumental ground motion in the central United States and California: Bulletin of the Seismological Society of America, v. 97, p. 497–510. Atkinson, G.M., and Boore, D.M., 2006, Earthquake ground-motion predictions for eastern North America: Bulletin of the Seismological Society of America, v. 96, p. 2181–2205. Atkinson, G., and Boore, D., 1995, New ground-motion relations for eastern North America: Bulletin of the Seismological Society of America, v. 85, p. 17–30. Bakun, W.H., and Hopper, M.G., 2004, Historical seismic activity in the central United States: Seismological Research Letters, v. 75, p. 564–574. Bakun, W.H., Johnston, A.C., and Hopper, M.G., 2003, Estimating locations and magnitudes of earthquakes in eastern North America from modified Mercalli intensities: Bulletin of the Seismological Society of America, v. 93, p. 190–202. Baldwin, J.N., Harris, J.B., Van Arsdale, R.B., Givler, R., Kelson, K.I., Sexton, J.L., Lake, M., 2005, Constraints on the location of the late Quaternary Reelfoot and New Madrid North Faults in the northern New Madrid Seismic Zone, central United States: Seismological Research Letters, v. 76, p. 772–789. Bear, G.W., Rupp, J.A., and Rudman, A.J., 1997, Seismic interpretations of the deep structure of the Wabash Valley Fault System: Seismological Research Letters, v. 68, p. 624–640. Benjamin, J.R., and Cornell, C.A., 1970, Probability, statistics, and decision for civil engineers: New York, McGraw-Hill, 684 p. Bernreuter, D.L., Savy, J.B., Mensing, R.W., and Chen, J.C., 1989, Seismic hazard characterization of 69 nuclear plant sites east of the Rocky Mountains, Report NUREG/CR-5250, v. 1–8. Bolt, B., 1993, Earthquakes, W.H. Freeman and Company, San Francisco, CA, 241 p. Bommer, J.J., and Abrahamson, N.A., 2006, Why do modern probabilistic seismichazard analyses often lead to increased hazard estimates?: Bulletin of the Seismological Society of America, v. 96, p. 1976–1977. 46 Boore, D.M., and Atkinson, G.M., 2006, Boore-Atkinson provisional NGA empirical ground-motion model for the average horizontal component of PGA, PGV and SA at spectral periods of 0.05, 0.1, 0.2, 0.3, 0.5, 1, 2, 3, 4, and 5 seconds: v. 1.7, 63 p. Boore, D.M., Joyner, W.B., and Fumal, T.E., 1997, Equations for estimating horizontal response spectra and peak acceleration from western North American earthquakes: A summary of recent work: Seismological Research Letters, v. 68, p.128–153. Braile, L.W., Hinze, W.J., and Keller, G.R., 1997, New Madrid seismicity, gravity anomalies, and interpreted ancient rift structure: Seismological Research Letters, v. 63, p. 599–610. Braile, L.W., Hinze, W.J., Keller, G.R., Lidiak, E.G., and Sexton, J.L., 1986, Tectonic development of the New Madrid Rift Complex, Mississippi Embayment, North America: Tectonophysics, v. 131, p. 1–21. Building Seismic Safety Council (BSSC), 2004, NEHRP recommended provisions for seismic regulations for new buildings [2003 ed.]: Federal Emergency Management Agency, FEMA 450, 340 p. Building Seismic Safety Council (BSSC), 1998, NEHRP recommended provisions for seismic regulations for new buildings [1997 ed.]: Federal Emergency Management Agency, FEMA 302, 337 p. Calais, E., Han, J.Y., DeMets, C., and Nocquet, J.M., 2006, Deformation of the North American plate interior from a decade of continuous GPS measurements: Journal of Geophysical Research, v. 111, p. Campbell, K.W. and Bozorgnia, Y., 2006, Campbell-Bozorgnia NGA empirical ground motion model for the average horizontal component of PGA, PGV, PGD and SA at selected spectral periods ranging from 0.01–10.0 seconds: v. 1.0, 193 p. Campbell, K.W., 2003, Prediction of strong ground motion using the hybrid empirical method and its use in the development of ground-motion (attenuation) relations in eastern North America: Bulletin of the Seismological Society of America, v. 93, p. 1012–1033. Campbell, K.W., 1981, Near-source attenuation of peak horizontal acceleration: Bulletin of the Seismological Society of America, v. 71, p. 2039- 2070. Carroll, R., 2003, Variances are not always nuisance parameters. Biometrics, v. 59, p. 211-220. Chiou, B.S.-J., and Youngs, R.R., 2006, Chiou and Youngs PEER-NGA empirical ground motion model for the average horizontal component of peak acceleration and pseudo-spectral acceleration for spectral periods of 0.01 to 10 seconds, 71 p. Chiu, J.-M., Johnston, A.C., and. Yang, Y.T, 1992, Imaging the active faults of the Central New Madrid Seismic Zone using PANDA array data: Seismological Research Letters, v. 63, p. 375–393. Cobb, J.C., 2006, USGS National Seismic Hazard Maps: Kentucky Issues: earthquake.usgs.gov/research/hazmaps/whats_new/workshops/CEUS_workshop.p hp [accessed 5/31/2007]. Cobb, J.C., 2004, USGS National Seismic Hazards Mapping Program and States: De facto Setting Public Policy for States: A presentation to the USGS Science Earthquake Studies Advisory Committee, June 3, 2004, Memphis, Tenn. 47 Cornell, C.A., 1971, Probabilistic analysis of damage to structures under seismic loads, in Howells, D.A., Haigh, I.P., and Taylor, C., eds., Dynamic waves in civil engineering: Proceedings of a conference organized by the Society for Earthquake and Civil Engineering Dynamics: New York, John Wiley, p. 473–493. Cornell, C.A., 1968, Engineeing seismic risk analysis: Bulletin of the Seismological Society of America, v. 58, p. 1583–1606. Cramer, C.H., 2004, Earthscope and PSHA 2004: www.ceri.memphis.edu/usgs/products/ ppt_db2.html [accessed 9/20/2005]. Cramer, C.H., 2001, The New Madrid Seismic Zone: Capturing uncertainty in seismic hazard analyses: Seismological Research Letters, v. 72, p. 664–672. Electric Power Research Institute (EPRI), 2003, CEUS ground motion project, model development and results: Report 1008910, 67 p. Electric Power Research Institute (EPRI), 1988, Seismic hazard methodology for central and eastern United States: EPRI NR-4726, 10 v. Federal Highway Administration (FHWA), 2006, Seismic Retrofitting Manual for Highway Structures: Part 1 - Bridges, U.S. Department of Transportation, Federal Highway Administration, Publication No. FHWA-HRT-06-032, 656 p. Frankel, A., 2004, How can seismic hazard around the New Madrid Seismic Zone be similar to that in California?: Seismological Research Letters, v. 75, p. 575–586. Frankel, A., Mueller, C., Barnhard, T., Perkins, D., Leyendecker, E., Dickman, N., Hanson, S., and Hopper, M., 1996, National seismic hazard maps: Documentation June 1996: U.S. Geological Survey Open-File Report 96-532, 110 p. Frankel, A.D., Petersen, M.D., Mueller, C.S., Haller, K.M., Wheeler, R.L., Leyendecker, E.V., Wesson, R.L., Harmsen, S.C., Cramer, C.H., Perkins, D.M., and Rukstales, K.S., 2002, Documentation for the 2002 update of the national seismic hazard maps: U.S. Geological Survey Open-File Report 02-420, 33 p. Geomatrix Consultants Inc., 2004, Dam safety seismic hazard assessment: Tennessee Valley Authority, Project 9223, 137 p. Hanks,T.C., and Cornell, C.A., 1994, Probabilistic seismic hazard analysis: A beginner’s guide, in Proceedings of the Fifth Symposium on Current Issues Related to Nuclear Power Plant Structures, Equipment and Piping, pp. I/1-1–I/1-17, N. C. State Univ., Raleigh, North Carolina. Herrmann, R.B., and Ammon, C.J., 1997, Faulting parameters of earthquakes in the New Madrid, Missouri, region: Engineering Geology, v. 46, p. 299–311. Holbrook, J., Autin, W.J., Rittenour, T.M., Marshak, S., and Goble, R.J., 2006, Stratigraphic evidence for millennial-scale temporal clustering of earthquakes on a continental-interior fault: Holocene Mississippi River floodplain deposits, New Madrid Seismic Zone, USA: Tectonophysics, v. 420, p. 431–454. Horton, S.P., Kim, W., and Withers, M., 2005, The 6 June 2003 Bardwell, Kentucky, earthquake sequence: Evidence for a locally perturbed stress field in the Mississippi Embayment: Bulletin of the Seismological Society of America, v. 95, p. 431–445. Hough, S.E., Armbruster, J.G., Seeber, L., and Hough, J.F., 2000, On the modified Mercalli intensities and magnitudes of the 1811–12 New Madrid earthquakes: Journal of Geophysical Research, v. 105, p. 23,839–23,864. International Code Council (ICC), 2000, International Building Code: 678 p. 48 Johnston, A.C., 1996a, Seismic moment assessment of stable continental earthquake, III., 1811–182 New Madrid, 1886 Charleston, and 1755 Lisbon: Geophys. J. Int., v. 126, p. 314–344. Johnston, A.C., 1996b, Seismic moment assessment of earthquakes in stable continental regions—I. Instrumental seismicity: Geophys. J. Int., v. 124, p. 381–414. Johnston, A.C., and Schweig, E.S., 1996, The enigma of the New Madrid earthquakes of 1811–1812: Annual Review of Earth and Planetary Sciences, v. 24, p. 339–384. Joyner, W.B., and Boore, D.M., 1981, Peak horizontal acceleration and velocity from strong-motion records including records from the 1979 Imperial Valley, California, earthquake: Bulletin of the Seismological Society of America, v. 71, p. 2,011–2,038. Kenner, S.J., and Segall, P., 2000, A mechanical model for intraplate earthquakes: Application to the New Madrid Seismic Zone: Science, v. 289, p. 2,329–2,332. Kim, W., 2003, June 18, 2002, Evansville, Indiana Earthquake: Reactivation of ancient rift in the Wabash Valley Fault Zone?: Bulletin of the Seismological Society of America, v. 93, p. 2201–2211. Klügel, J.-U., 2005, Problems in the application of the SSHAC probability method or assessing earthquake hazards at Swiss nuclear power plants: Engineering Geology, v. 78, p. 285–307. Kramer, S.L., 1996, Geotechnical earthquake engineering: Upper Saddle River, N.J., Prentice Hall, 653 p. Krinitzsky, E.L., 2002, How to obtain earthquake ground motions for engineering design: Engineering Geology, v. 65, p. 1–16. Krinitzsky, E.L., 1995, Deterministic versus probabilistic seismic hazard analysis for critical structures: Engineering Geology, v. 40, p. 1–7. McBride, J.H., Hildenbrand, T.G., Stephenson, W.J., and Potter, C.J., 2002, Interpreting the earthquake source of the Wabash Valley Seismic Zone (Illinois, Indiana, and Kentucky) from seismic reflection, gravity, and magnetic intensity data: Seismological Research Letters, v. 73, p. 660–686. McGuire, R.K., Cornell, C.A., and Toro, G.R., 2005, The case for using mean seismic hazard: Earthquake Spectra, v. 21, p. 879–886. McGuire, R.K., 2004, Seismic hazard and risk analysis: Earthquake Engineering Research Institute, MNO-10, 221 p. McGuire, R.K., 1995, Probabilistic seismic hazard analysis and design earthquakes: Closing the loop: Bulletin of the Seismological Society of America, v. 85, p. 1275–1284. McGuire, R.K., 1976, FORTRAN computer program for seismic risk analysis: U.S. Geological Survey Open-File Report 76-67. Mendenhall, W., Scheaffer, R.L., and Wackerly, D.D., 1986, Mathematical statistics with applications: Boston, Duxbury Press, 750 p. Mueller, K., and Pujol, J., 2001, Three-dimensional geometry of the Reelfoot blind thrust: Implications for moment release and earthquake magnitude in the New Madrid Seismic Zone: Bulletin of the Seismological Society of America, v. 91, p. 1563–1573. Munson, P.J., Munson, C.A., and Pond, E.C., 1995, Paleoliquefaction evidence for a strong Holocene earthquake in south-central Indiana: Geology, v. 23, p. 325–328. 49 Munson, P.J., Obermeier, S.F., Munson, C.A., and Hajic, E.R., 1997, Liquefaction evidence for Holocene and latest Pleistocene earthquakes in southern halves of Indiana and Illinois: A preliminary overview: Seismological Research Letters, v. 68, p. 521–536. Musson, R.M.W., 2005, Against fractiles: Earthquake Spectra, v. 21, p. 887–891. National Research Council, 1988, Probabilistic seismic hazard analysis, report of the Panel on Seismic Hazard Analysis: Washington, D.C., National Academy Press, 97 p. Newman, A., Stein, S., Weber, J., Engeln, J., Mao, A., and Dixon, T., 1999, Slow deformation and low seismic hazard at the New Madrid Seismic Zone: Science, v. 284, p. 619–621. Nuclear Regulatory Commission, 1997, Regulatory Guide 1.165, Identification and characterization of seismic sources and determination of safe shutdown earthquake ground motion. Nuttli, O.W., and Herrmann, R.B., 1978, Credible earthquakes in the central United States, in State-of-the art for assessing earthquake hazards in the United States: U.S. Army Corps of Engineers Geotechnical Lab Report 12, 99 p. Obermeier, S.F., Martin, J.R., Frankel, A.D., Youd, T.L., Munson, P.J., Munson, C.A., and Pond, E.C., 1993, Liquefaction evidence for one or more strong Holocene earthquakes in the Wabash Valley of southern Indiana and Illinois: US Geological Survey Professional Paper 1536, 27 p. Obermeier, S.F., Bleuer, N.R., Munson, C.A., Munson, P.J., Martin, W.S., McWilliams, K.M., Tabacznski, D.A., Odum, J.K., Rubin, M., and Eggert, D.L., 1991, Evidence of strong earthquake shaking in the lower Wabash Valley from prehistoric liquefaction features: Science, v. 251, p. 1061–1063. Olson, S.M., Green, R.A., and Obermeier, S.F., 2005, Revised magnitude-bound relation for the Wabash Valley Seismic Zone of the central United States: Seismological Research Letters, v. 76, p. 756–771. Petersen, M., 2005, National seismic hazard maps:Issues for Kentucky, in Wang, Z., comp., Better Understanding and Communication of the National Seismic Hazard Maps: Summary of USGS-KGS Meeting on Seismic Hazard Assessment in Western Kentucky, Kentucky Geological Survey, ser. 12, Special Publication 7, 47 p. Pond, E.C. and Martin, J.R., 1997, Estimated magnitudes and accelerations associated with prehistoric earthquakes in the Wabash Valley of the central United States: Seismological Research Letters, v. 68, p. 611–623. Reiter, L., 1990, Earthquake hazard analysis: New York, Columbia University Press, 254p. Risk Engineering Inc. (REI), 1999, Updated probabilistic seismic hazard analysis for the Paducah Gaseous Diffusion Plant, Paducah, Kentucky: Final report, 124 p. Sadigh, K., Chang, C.Y., Egan, J.A., Makdisi, F., and Youngs, R.R., 1997, Attenuation relationships for shallow crustal earthquakes based on California strong motion data: Seismological Research Letters, v. 68, p.180–189. 50 Scherbaum, F., Bommer, J., Bungum, H., Cotton, F., and Abrahamson, N., 2005, Composite ground-motion models and logic trees: Methodology, sensitivities, and uncertainties: Bulletin of the Seismological Society of America, v. 95, p. 1575– 1593. Senior Seismic Hazard Analysis Committee (SSHAC), 1997, Recommendations for probabilistic seismic hazard analysis: Guidance on uncertainty and use of experts: Lawrence Livermore National Laboratory, NUREG/CR-6372, 81 p. Sexton, J.L., Braile, L.W., Hinze, W.J., and Campbell, M.J., 1996, Seismic reflection profiling studies of a buried Precambrian rift beneath the Wabash Valley Fault Zone: Geophysics, v. 51, p. 640–660. Shakal, A., Haddadi, H., Graizer, V., Lin, K., and Huang, M., 2006, Some key features of the strong-motion data from the M 6.0 Parkfield, California, earthquake of 28 September 2004: Bulletin of the Seismological Society of America, v. 96, p. S90– 118. Silva, W., Gregor, N., and Darragh, R., 2002, Development of regional hard rock attenuation relationships for central and eastern North America: Pacific Engineering and Analysis, 311 Pomona Ave., El Cerrito, CA 94530, 57 p. Somerville, P., Collins, N., Abrahamson, N., Graves, R., and Saikia, C., 2001, Ground motion attenuation relations for the central and eastern United States: Final report to U.S. Geological Survey, 16 p. Stein, S., Tomasello, J., and Newman, A., 2003, Should Memphis build for California’s earthquakes?: EOS, Transactions of the American Geophysical Union, v. 84, p. 177, 184–185. Stepp, J.C., Wong, I., Whitney, J., Quittmeyer, R., Abrahamson, N., Toro, G., Youngs, R., Coppersmith, K., Savy, J., Sullivan, T., and Yucca Mountain PSHA project members, 2001. Probabilistic seismic hazard analysis for ground motions and fault displacements at Yucca Mountain, Nevada: Earthquake Spectra, v. 17, p. 113–151. Stover, C.W., and Coffman, J.L., 1993, Seismicity of the United States, 1568–1989 (rev.): U.S. Geological Survey Professional Paper 1527, 418 p. Street, R.L., Bauer, R.A., and Woolery, E.W., 2004, Short note: Magnitude scaling of prehistorical earthquakes in the Wabash Valley Seismic Zone of the central United States: Seismological Research Letters, v. 75, p. 637–641. Street, R., and Wang, Z., 2003, Analysis of strong motion records from the University of Kentucky accelerometers in the New Madrid Seismic Zones: 1990 through 2001: Report to the U.S. Geological Survey, Award Number 02HQGR0016, 10 p. Street, R., Wang, Z., Harik, I., and Allen, D.,1996, Source zones, recurrence rates, and time histories for earthquakes affecting Kentucky: Kentucky Transportation Center, University of Kentucky, Research Report KCT-96-4, 187 p. Structural Engineers Association of Kentucky (SEAOK), 2002, White Paper on Review of the 2002 Kentucky Residential Code (2nd ed.), SEAOK Document WP-01-2.1, 66 p. Toro, G.R., Abrahamson, N.A., and Schneider, J.F., 1997, Model of strong ground motions from earthquakes in central and eastern North America: Best estimates and uncertainties: Seismological Research Letters, v. 68, p. 41–57. 51 Toro, G.R., Silva, W.J., McGuire, R.K., and Herrmann, R.B., 1992, Probabilsitic seismic hazard mapping of the Mississippi Embayment: Seismological Research Letters, v. 63, p. 449–475. Tuttle, M.P., Schweig, E.S., Sims, J.D., Lafferty, R.H., Wolf, L.W., and Haynes, M.L., 2002, The earthquake potential of the New Madrid Seismic Zone: Bulletin of the Seismological Society of America, v. 92, p. 2080–2089. Tuttle, M.P. and Schweig, E.S., 2000, Earthquake potential of the New Madrid Seismic Zone [abs.]: EOS, Transactions of the American Geophysical Union, v. 81, p. S308–309. U.S. Geological Survey (USGS)/Nuclear Regulatory Commission (NRC), 2005, Workshop on attenuation of ground motion in the central and eastern United States: CD–ROM. Wang, Z., in press, Seismic Hazard and Risk Assessment in the Intraplate Environment: The New Madrid Seismic Zone of the Central United States: GSA Book SPE425: Continental Intraplate Earthquakes: Science, Hazard, and Policy Issues, ed. Seth Stein and Stéphane Mazzotti. Wang, Z., 2006a, Understanding Seismic Hazard and Risk Assessments: An Example in the New Madrid Seismic Zone of the Central United States, Proceedings of the 8th National Conference on Earthquake Engineering, April 18-22, 2006, San Francisco, CA, Paper no. 416. Wang, Z., 2006b, Addendum to the PEGASOS Follow-Up Workshop, Swissnuclear workshop on the refinement of the PEGASOS Project, November 28-29, 2006, Zurich, Swizzerland. Wang, Z., 2005a, Comment on J.U. Klügel’s: Problems in the Application of the SSHAC Probability Method for Assessing Earthquake Hazards at Swiss Nuclear Power Plants, in Engineering Geology, vol. 78, pp. 285-307, Engineering Geology, v. 82, p. 86-88. Wang, Z., 2005b, Reply to “Comment on ‘Comparison between probabilistic seismic hazard analysis and flood frequency analysis’ by Zhenming Wang and Lindell Ormsbee” by R. M. W. Musson: EOS, Transactions of the American Geophysical Union, v. 86, p. 354. Wang, Z., 2005c, Reply to “Comment on ‘Comparison between probabilistic seismic hazard analysis and flood frequency analysis’ by Zhenming Wang and Lindell Ormsbee” by Thomas L Holzer: EOS, Transactions of the American Geophysical Union, v. 86, p. 303. Wang, Z. (compiler), 2005d, Better Understanding and Communication of the National Seismic Hazard Maps: Summary of USGS-KGS Meeting on Seismic Hazard Assessment in Western Kentucky, Kentucky Geological Survey, ser. 12, Special Publication 7, 47 p. Wang, Z., 2003a, Summary for the USGS NEHRP Hazard Maps, in Wang, Z., comp., the Kentucky NEHRP Seismic Hazard and Design Maps Workshop proceedings: Kentucky Geological Survey, ser. 12, Special Publication 5, p. 16–18. Wang, Z. (compiler), 2003b, Proceedings of The Kentucky NEHRP Seismic Hazards and Design Maps Workshop, November 18, 2002, Lexington, Kentucky, Kentucky Geological Survey, ser. 12, Special Publication 5, 38p. 52 Wang, Z., and Ormsbee, L., 2005, Comparison between probabilistic seismic hazard analysis and flood frequency analysis: EOS, Transactions of the American Geophysical Union, v. 86, p. 45, 51-52. Wang, Z., and Zhou, M., in press, Comment on “Why Do Modern Probabilistic Seismic Hazard Analyses Often Lead to Increased Hazard Estimates?” by Julian J. Bommer and Norman A. Abrahamson: Bulletin of the Seismological Society of America. Wang, Z., Woolery, E.W., Shi, B., and Kiefer, J.D., 2005, Comment on “How Can Seismic Hazard around the New Madrid Seismic Zone Be Similar to that in California?” by Arthur Frankel: Seismological Research Letters, v. 76, p. 466– 471. Wang, Z., Woolery,E.W., B. Shi, and J.D. Kiefer, 2004a, Reply to Comment on “Communicating with Uncertainty: A Critical Issue with Probabilistic Seismic Hazard Analysis” by R.M.W. Musson: EOS, Transactions of the American Geophysical Union, v. 85, p. 236. Wang, Z., Woolery, E.W., Shi, B., and Kiefer, J.D., 2004b, Reply to “Comment on ‘Communicating with Uncertainty: A Critical Issue with Probabilistic Seismic Hazard Analysis’”by C.H. Cramer: EOS, Transactions of the American Geophysical Union, v. 85, p. 283, 286. Wang, Z., Woolery, E.W., and Shi, B., 2003a, Observed seismicity (earthquake activity) in the Jackson Purchase Region of western Kentucky: January through June 2003: Kentucky Geological Survey, ser. 12, Special Publication 6, 16 p. Wang, Z., Woolery, E.W., Shi, B., and Kiefer, J.D., 2003b, Communicating with uncertainty: A critical issue with probabilistic seismic hazard analysis: EOS, Transactions of the American Geophysical Union, v. 84, p. 501, 506, 508. Wheeler, R.L., 1997, Boundary separating the seismically active Reelfoot Rift from the sparsely seismic Rough Creek Graben, Kentucky and Illinois: Seismological Research Letters, v. 63, p. 586–598. Wheeler, R.L., and Cramer, C.H., 2002, Updated seismic hazard in the southern Illinois Basin – Geological and geophysical foundations for use in the 2002 USGS national seismic-hazard maps: Seismological Research Letters, v. 73, p. 776-791. Wheeler, R.L., and Frankel, A., 2000, Geology in the 1996 USGS seismic-hazard maps, central and eastern United States: Seismological Research Letters, v.71, p. 273– 282. William Lettis & Associates Inc., 2006, Investigation of Holocene faulting at proposed C-746-U landfill expansion, Paducah Gaseous Diffusion Plant, Paducah, Kentucky: 66 p. Windeler, D., 2006, USGS workshop on the 2007 CEUS hazard map update: Some “user perspectives” from insurance loss modeling: earthquake.usgs.gov/research/ hazmaps/whats_new/workshops/CEUS_workshop.php [accessed 5/31/2007]. Woolery, E.W., 2005, Geophysical and geological evidence of Neotectonic deformation along the Hovey Lake fault, lower Wabash Valley Fault System: Bulletin of the Seismological Society of America, v. 95, p. 1193–1201. Youngs, R.R., Chiou, S.J., Silva, W.J., and Humphrey, J.R., 1997, Strong ground motion attenuation relationship for subduction zone earthquake: Seismological Research Letters, v. 68, p. 58–73. 53 Youngs, R.R., Abrahamson, N., Makdisi, F.I., and Sadigh, K., 1995, Magnitudedependent variance of peak ground acceleration: Bulletin of the Seismological Society of America, v. 85, p. 1161–1176. Zoback, M.L., 1992, Stress field constraints on intraplate seismicity in eastern North America: Journal of Geophysical Research, v. 97, p. 11,761–11,782. Zoback, M.D., Prescott, W.H., and Kroeger, S.W., 1985, Evidence for lower crustal ductile strain localization in southern New York: Nature, v. 317, p. 705-707. Zoback, M.D., Zoback, M.L., Mount, V.S., Suppe, J., Eaton, J.P., Eaton, J.P., Healy, J.H., Oppenheimer, D., Reasenberg, P., Jones, L., Raleigh, C.B., Wong, I.G., Scotti, A., and Wentworth, O.C., 1987, New evidence on the state of stress of the San Andreas fault system: Science, v. 238, p. 1105-1111. 54 Appendix A – Project Proposal TITLE: Enhancing Earthquake Monitoring and Assessing Seismic Hazard for the Paducah Gaseous Diffusion Plant, Paducah, Kentucky GOALS The goals for this proposal are: 1) to better monitor and locate earthquakes in the area and; 2) to provide an independent and peer reviewed ground motion hazard assessment for the Paducah Gaseous Diffusion Plant (PGDP). OBJECTIVES In the central United States, the best information for determining seismogenic faults (faults that are capable of generating earthquakes) is seismicity (earthquake activity). Until recently, the lack of seismic stations in the area has precluded any definitive determination of the active faults in the area. Earthquakes occur periodically in area surrounding PGDP, for example, the August 26, 2003 west Paducah earthquake (mb 3.2). In order to better monitor and locate earthquakes, a temporary seismic network has been deployed in the area, with support from the Commonwealth of Kentucky. The seismic stations are still not dense enough to accurately locate earthquakes in area surround PGDP, however. Additional seismic stations are needed. Federal government agencies, including the U.S. Nuclear Regulatory Commission, use the seismic hazard maps developed by the U.S. Geological Survey for seismic safety regulations. These maps are based on a 2 percent probability that a ground motion will be exceeded in 50 years (2,500-year return period). The maps predict very high ground motion for the area surrounding the PGDP. These high seismic hazard estimates for the area have a significant impact on seismic regulations and engineering designs for facilities at the PGDP. The seismic hazard at the PGDP has also been estimated by many other public and private organizations. The results are significantly different among these estimates. The tasks for this study include: Task 1. Micro-seismicity observation in Paducah area. We propose to complete two seismic stations, one at the PGDP and the other in Paducah, and to install three new seismic stations in the area. These stations, combined with the eight existing seismic stations, will enhance our capability to monitor micro-seismicity in the area. Task 2. Thorough literature review. There are many new developments and data on A-1 seismic hazard assessment methodology, geology, and seismology locally, regionally, and nationally. The focus will be on the new geological and geophysical investigations in the area. The literature review will ensure the use of the best data and methodology. Task 3. Seismic source Characterization. Based on the information derived from Task 1 and 2, the seismic sources in and around the PGDP and their characteristics will be defined. Task 4. Probabilistic seismic hazard analysis (PSHA). PSHA will be performed based on the seismic source data from Task 3. Task 5. Deterministic seismic hazard analysis (DSHA). DSHA will be performed based on the seismic source data from Task 3. Task 6. Preliminary report. Task 7. Panel review. A 5-member review panel consisting of national and international experts will be formed to review the preliminary report. Task 8. Final report. A-2 Appendix B – Agenda of the Independent Technical Review Meeting Independent Technical Review For Final Research Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant By Zhenming Wang and Edward W. Woolery Date: April 30, 2007 Place: Room 102, Mining and Mineral Resources Building, UK campus B-1 Agenda 8:00-8:10 Introduction Ed Woolery 8:10-8:20 Nature of the Project Lindell Ormsbee (Director, Kentucky Water Resource Institute) 8:20-8:30 Issues Related to Seismic Hazard Assessment in Western Kentucky Jim Cobb (Director, Kentucky Geological Survey) 8:30-10:00 Ground Motion Attenuation Relationship – uncertainty, dependency, and hazard calculation in PSHA (focus) Woolery/Wang 8:30-9:00 Presentation Zhenming Wang 9:00-9:50 Q/A and Discussion Panel members 9:50-10:00 Q/A from attendees 10:00-10:30 Coffee Break 10:30-12:00 Seismic Hazard Analysis (SHA) – temporal and spatial measurements, uncertainties, and quantification (focus) Woolery/Wang 10:30-11:00 Presentation Zhenming Wang 11:00-11:50 Q/A and Discussion Panel members 11:50-12:00 Q/A from attendees 12:00-13:00 Lunch Break 13:00-14:30 Seismic Hazard Assessment for PGDP – input parameters: sources, occurrence frequency, and ground motion attenuation (focus) Woolery/Wang 13:00-13:30 Presentation Zhenming Wang 13:30-14:20 Q/A and Discussion Panel members 14:20-14:30 Q/A from attendees 14:30-14:45 Break 14:45-16:00 Discussion and Summary Wang/Woolery B-2 Members of Independent Technical Review Panel 1. Roy B. Van Arsdale 8872 Farmoor Road Germantown, TN 38139 E-mail: [email protected] Phone: 901-678-4356 2. Gail Atkinson 280 McLean Ave. Arnprior, ON K7S 3T2 Canada E-mail: [email protected] Phone: 613-623-3240 3. James E. Beavers James E Beavers Consultants 6318 Beaver Ridge Road Knoxville, TN 37931 E-mail: [email protected] Phone: 865-690-8936 4. Kenneth W. Campbell 1030 NW 161st Place Beaverton, OR 97006 E-mail: [email protected] Phone: 503-533-4359 5. Leon Reiter 1960 Dundee Road Rockville, MD 20850 E-mail: [email protected] Phone: 301- 279-0509 6. Mai Zhou Department of Statistics 849 POT University of Kentucky Lexington, KY 40506-0027 E-mail: [email protected] Phone: 859-257-6912 B-3 Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 Reviewer Document / Section or Page # 1 Roy VanArsdale General 2 Roy VanArsdale General 3 Roy VanArsdale General Response to Comment Comment Although the work of Tuttle et al. is the most recent to address earthquake recurrence in the New Madrid seismic zone, an earlier article came to the same conclusion. Kelson et al. (1996) concluded that the recurrence interval on Reelfoot fault earthquakes in between 400 and 500 years. This is significant because the earthquake recurrence interval is tied to a specific fault. I did not see any treatment of multiple large earthquakes occurring on the New Madrid seismic zone like that which occurred in 1811-1812. Tuttle et al. (2002) address this and there is also evidence for this clustering in Van Arsdale et al. (1998). Does this clustering of large earthquakes not affect your results? There is a large hole in our basement data at the north end of Reelfoot Rift. We really do not know how the Reelfoot Rift links with the Rough Creek Graben. I have a Ph.D. student (Ryan Csontos) who just completed his dissertation in which he took a stab at this. It appears that the Precambrian crystalline basement rises between the northern end of the Reelfoot Rift and southern end of Rough Creek Graben. Ryan interpreted the Reelfoot fault to be a normal fault at depth which forms a step up and out of the Reelfoot Rift. In his model, the Reelfoot fault is an inverted normal fault. Another issue about the structure id the strike of fault in this transition zone. Do the faults continue N45E or do they curve and merge with more easterly Rough Creek Graben faults? This should have a bearing on the stress on these faults from the N60W regional maximum stress C-1 A recent study by Holbrook et al. (2006) indicates the earthquake recurrence interval of about 1,000 years for the same fault. This is the reason that a range of recurrence interval, 500 to 1,000 years, is considered. Seismic hazard is defined as an earthquake of magnitude M or greater (cumulative) or ground motion generated by the earthquake at a site vs. mean recurrence interval (or return period for ground motion). Seismic risk is defined as the probability of at least one occurrence of M or greater earthquake (cumulative) or the ground motion at a site over a period. The clustering is considered, and will not have effect on your results. These are very good comment. The questions need to be addressed through future studies. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 Reviewer Document / Section or Page # Response to Comment Comment 4 Roy VanArsdale General You do not address the large number of earthquakes that trend into western Kentucky inllustrated in Figure 11. The data quality, in terms of magnitude, location, and focal depth, for earthquakes before 2003 in western Kentucky is very poor due to the lack of seismic stations. Based on those earthquakes, wheeler (1997) suggested the northeast extensions of the New Madrid faults, but also suggested that that can be substantiated by further seismic network monitoring. Recent studies by more dense network (Wang and others, 2003a; Horton and others, 2005; Anderson and others, 2005) shows consistent difference between the earthquakes in the New Madrid and those in the Jackson Purchase Region, indicating that the New Madrid faults may not extend northeast into western Kentucky. There is no geologic evidence indicating the extension in the Jackson Purchase Region. On the other hand, there are geologic evidences showing the northeast extensions of the New Madrid faults on Missouri side, such as Baldwin and others (2005). 5 Roy VanArsdale Executive Summery How can you have high seismic risk without seismic hazard? 6 Roy VanArsdale Chapter 3, Section 1 (page 21) 7 Roy VanArsdale Roy VanArsdale Roy VanArsdale Chapter 3, Section 1 (page 21) Chapter 3, Section 1 (page 21) Chapter 3, Section 1 (page 24) The Reel rift –Rough Creek graben-Rome trough is commonly considered to be one large perhaps discontinuous Cambrian rift What about dense seismicity in W Ky in Fig. 11? In the report, we state “High seismic hazard does not necessarily mean high seismic risk, and vice versa.” This means that low seismic hazard does not necessarily mean low seismic risk or there could be high seismic risk even though seismic hazard is low. If there is no seismic hazard, there is no seismic risk. This can be illustrated through following examples: 1) Mojave desert has high seismic hazard (frequent large earthquakes, such as Hechtor Mine earthquake), but has low seismic risk because few exposures (people and property). 2) San Simeon area has relative low seismic hazard (compare to Mojave desert), but has higher seismic risk because high expotures. This is good comment. There relationship between them in Quaternary, particularly in Holocene, is not clear which has impact on seismic hazard assessment See response to comment #4. True, also a black hole of no data See response to comment #3. Why not through 2006? All earthquakes up to March 2007 will be included. 8 9 C-2 Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 1 Reviewer Gail Atkinson Document / Section or Page # General Response to Comment Comment This report deals with seismic hazards to the Paducah gaseous diffusion plant and the methodology by which they should be assessed. The report is clearly written and easy to follow for the most part, but the reasoning used to propose an alternative methodology is flawed. The report does not actually provide a probabilistic seismic hazard assessment (PSHA) for the site. Rather it is focused on providing arguments as to why PSHA may not be applicable. I did not find these arguments convincing. 1) PSHA is a well accepted technique throughout the world, and the subject of many knowledgeable and definitive articles and textbooks by leading scientists and engineers over the last 40 years. In my view it has a much sounder basis than the new methodology proposed here, which is a hybrid approach (elements of deterministic and probabilistic methodologies) that has been termed Seismic Hazard Assessment (SHA). 2) The proposed methodology (SHA) is seriously flawed, as discussed in the points below. C-3 These general comments can be summarized into two questions 1) Is PSHA appropriate even though it has been used for seismic hazard assessment for three decades? 2) Is the proposed methodology (SHA) seriously flawed? The answer to the question 1 is clear: PSHA may not be appropriate for seismic hazard assessment because it contains a mathematical error in its formulation: incorrectly treating the ground-motion uncertainty as an independent random variable. The ground-motion uncertainty is an explicit or implicit dependent variable as it is modeled in the groundmotion attenuation relationship. The mathematical error results in double/triple counts of uncertainties in earthquake magnitude and source-to-site distance. The mathematical error also results in mixing temporal measurement (occurrence of an earthquake and its consequence [ground motion] at a site) with spatial measurement (ground-motion variability due to the source, path, and site effects). The results from a PSHA study are artifact. The answer to the question 2 is also clear: SHA is appropriate because 1) it was peer reviewed (paper no. 416, Proceedings of the 8th U.S. National Conference on Earthquake Engineering, April 18-22, 2006; Chapter 24, GSA Special Publication 425, in press), 2) it is analogous to flood and wind hazard analyses for engineering design, and 3) it is similar to the Milne-Davenport approach (1969) and Stein and others (2005, 2006). Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 2 Reviewer Gail Atkinson Document / Section or Page # General Response to Comment Comment I also question why an entirely new methodology would be proposed in the context of a specific engineering project. For engineering projects, it is generally considered important to follow accepted practice. I appreciate that the motivation for such an approach arises from the consideration that PSHA suggests large ground motions at low probabilities for many regions of the CUS influenced by the New Madrid seismic zone and other nearby sources. However I do not believe that the methodology proposed is a correct way to deal with these issues. Depending on the regulatory requirements that may apply, there could be other approaches to dealing with the site issues that would be more defendable. Just as an illustrative example (not a recommendation), it may be considered acceptable to find the probabilistic ground motions associated with each potential source separately (New Madrid, Wabash, Background), for some target probability – one might then say, for example, that the facilities can accommodate the 2%/50 year motions from each of the potential sources, while recognizing that this is not the total probability of receiving the ground motions. (The implicit rationale would be that the facility is not expected to be able to withstand a significant event from more than one potential source during its lifetime.) I emphasize that this is not a proposed solution, just a discussion point, and that this argument may not be applicable depending on whether there are specific reliability targets for the project. C-4 As shown in the report and response to comment #1, the results from a PSHA study are all artifacts, and may not be appropriate for seismic hazard assessment. As demonstrated by Harris (ATC-USGS hazard workshop, 2006), return period derived from PSHA is interpreted and used as mean recurrence interval (MRI) and compared with those of wind, snow, and other hazards. However, the return period is not equal or equivalent to MRI. The proposed approach is not new, but a re-introduction of an old one (Milne and Davenport, 1969) with addition of uncertainty. Return period derived from the proposed approach is identical to MRI derived from wind, flood, and other hazard analysis. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 3 Reviewer Gail Atkinson Document / Section or Page # General Comment Response to Comment The proposed methodology is really a recasting of the concept of the “Maximum Credible Earthquake”, in which specific source scenarios (New Madrid, Wabash, Background) are assigned in terms of a fixed distance and a subjective maximum magnitude. The casting of a recurrence relation for each source into a probabilistic ground-motion distribution only applies for the specific distance and maximum magnitude. In the case of background earthquakes and poorly-understood sources (such as Wabash), the maximum magnitude and distance are arbitrary. The maximum magnitudes for the Background (Mx=5) and Wabash (Mx=6.8) sources are not justified. The results of this proposed methodology will be very sensitive to the assigned maximum magnitude and distance. The derived ground-motion probabilities are not correct as they do not consider that for each of the considered scenarios, there is a significant probability of a larger event at a closer location. They also do not properly account for the effect of sigma on ground-motion probability. The variability of actual ground motions about the predicted median increases the frequency of exceedence of any given ground motion level, as is shown in the appended illustration. Thus no probabilistic ground-motion distribution is actually obtained by this method. These comments are really about how to treat temporal and spatial uncertainties (variability) of earthquake. First, the temporal and spatial uncertainties are two intrinsic, but fundamentally different measures, and must be treated separately. PSHA mixes the temporal uncertainty with the spatial one (this is the result of incorrect formulation of PSHA), i.e. using the ground motion uncertainty to extrapolate the frequency (temporal measure). The proposed approach treats the temporal and spatial uncertainties separately. The “Maximum Credible Earthquake” is the best estimate (mean) of the maximum earthquake in a source zone, not subjective one. The maximum magnitude for Wabash (Mx=6.8) source is based on the most recent studies (Street and others, 2004; Olson and others, 2005). The maximum magnitudes for the background (Mx=5) source is somewhat subjective. The distances or source boundaries (Wabash) are more subjective, These subjective determinations of magnitude and boundaries are consistent with current practice in the region. 4 Gail Atkinson Specific (P 1.2) It would be useful to discuss what regulatory requirements, if any, apply to the Plant – is there a specified target probability, for example? This is more relevant than the general issue of the 2% in 50 year maps and their possible implications for buildings and other projects in the region. There is no specific target probability or regulatory requirement. This report has a general implication for engineering design and policy consideration in Kentucky. 5 Gail Atkinson Specific (P 1.6) There is no need to discuss the USGS maps if they are not required by the applicable code – this is not really relevant and should be deleted. With universally referred in government regulations, codes, and other relevant documents, the USGS maps have to be discussed. Revision will be done to reflect these and add more explanations C-5 Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 Reviewer Document / Section or Page # Response to Comment Comment 6 Gail Atkinson Specific (P 2) Fig. 1 is not relevant due to the very short time span (1 week) – if you want to illustrate the known seismicity of the country from a hazards viewpoint, plot something like all damaging earthquakes in the historic record. Revised to use Stein and others (2003). 7 Gail Atkinson Specific (P 3.3) The focus seems unbalanced – when performing an assessment of seismic hazard for a specific site, it would not be the appropriate venue to review the national seismic hazard maps of the USGS, nor to propose a new methodology. As described in the responses to comments 4 and 5 (universally referred in government regulations, codes, and other relevant documents, and general implication for engineering design and policy consideration in Kentucky), the USGS hazard maps have to be discussed and reviewed. Specific (P 5) (and throughout) The definition of risk versus hazard used in this report does not follow the accepted convention. There was initial confusion between the terms hazard and risk in the early days of seismic hazard methodology. However, it is now nearly universal usage that seismic hazard refers to the likelihood of receiving seismic ground motions (or other seismic effects), while seismic risk is the product of the hazard and the consequence (exposure or vulnerability). Thus a site with moderate seismicity but a hazardous or critical facility may pose a high seismic risk, while a site with high seismicity but few facilities may have low seismic risk. 8 9 Gail Atkinson Gail Atkinson Specific (P 6) This page deals with hazard, not risk. A magnitude recurrence curve is not on its own relevant to hazard, as events need to be associated with distances to determine ground motions. C-6 As shown, PSHA is mathematically incorrect; an alternative needs to be developed. The definition of hazard and risk used in this report follows the accepted convention, particularly in engineering (hydraulic, flood, wind, and snow). Seismic hazard describes phenomena, such as surface rupture, ground motion, ground-motion amplification, liquefaction, and induced landslides, generated by earthquakes that have potential to cause harm. Seismic risk, on the other hand, describes the likelihood (chance) of experiencing a specified level of seismic hazard in a given time exposure. These definitions are also consistent with those of McGuire (2004) and Reiter (1990). As defined by McGuire (2004), seismic hazard is “a property of an earthquake that can cause damage and loss. A PSHA determines the frequency (the number of events per unit of time) with which a seismic hazard will occur,” seismic risk is “the probability that some humans will incur loss or that their built environment will be damaged. These probabilities usually represent a level of loss or damage that is equaled or exceeded over some time period.” A similar definition was described by Reiter (1990),”seismic hazard describes the potential for dangerous, earthquake-related natural phenomena such as ground shaking, fault rupture, or soil liquefaction; seismic risk is the probability of occurrence of these consequences.” As defined, “seismic hazard is a property of an earthquake that can cause damage and loss,” a magnitude recurrence curve is a hazard curve because an M6.0 earthquake can cause damage and loss. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 Reviewer Document / Section or Page # Response to Comment Comment 10 Gail Atkinson Specific (P 7) The discussion of flood hazards is not relevant. 11 Gail Atkinson Specific (P 8) The description of seismic hazard versus risk is not a correct description of these concepts as they are used today. Furthermore, the discussion of seismic risk is not required here, as the report is dealing with seismic hazard. 12 Gail Atkinson Specific (P 11) The arguments presented regarding Eqn 4 are not convincing. The issue of E being independent of M and R is not central, in my view. Furthermore, E is in fact largely independent of M and R, as shown by recent ground motion databases (PEER/NGA). The opposing references cited are largely taken out of context - there are many analyses, authored by the same sources cited on this page, to show that E does not depend strongly, if at all, on M and R. The conclusions reached on the validity of Equations 4 and 9 are not justified. 13 Gail Atkinson Specific (P 15.5) There is no suggestion in the cited papers that ground motions will occur in 108 years. The arguments advanced here are not correct, nor do they appear relevant. 14 Gail Atkinson Specific (P 15.8) This reasoning is not correct. PSHA is simply a compound probability, like any other compound probability. Space and time are both relevant in determining the likelihood of receiving strong ground motion at a site. 15 Gail Atkinson Specific (P 16) A hazard assessment for a critical facility is not the place to introduce a new trial methodology, in my view. C-7 It is relevant because seismic hazard and risk analyses were developed based on analogy to flood, wind, and other analyses. The description of seismic hazard versus risk is consistent through this report. The discussion of seismic risk will help to understand why and how we do seismic hazard analysis. R in Eq. 4 is focal distance (Cornell, 1968). In the ground motion attenuation relationships, R is measured as rupture, JB, or seismogenic distance. The ground motion standard deviation will be different if different R is used (R dependent). fR(r) in Eq. 4 is to account for the uncertainty of focal point (distribution). The uncertainty of focal point is accounted in part by the uncertainty of ground motion because R is measured as a single distance (rupture, JB, or seismogenic) regardless focal distance. Eq.4 counts the distance uncertainty, at least some portion, twice. Similarly, fM(m) in Eq. 4 is to account for the uncertainty of magnitude (distribution). Also similarly, the ground motion standard deviation is dependent of M. Again, Eq.4 counts the magnitude uncertainty, at least some portion, twice. As defined by McGuire (2004), return period is the mean (average) time between occurrences of a seismic hazard. The reciprocal of return period is frequency. “PSHA determines the frequency (the number of events per unit of time) with which a seismic hazard will occur” McGuire (2004). The same interpretation was also given by (Frankel, 2004, 2005; Holzer, 2005) The temporal and spatial uncertainties are two intrinsic, but fundamentally different measures, and must be treated separately. This report is not necessary for a critical facility. The main goal of this report is to conduct scientific research on the methodologies, geological and seismological parameters, and the results related to the Paducah Gaseous Diffusion Plant and the region. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 Reviewer Document / Section or Page # Comment Response to Comment As shown earlier, ground motion uncertainty is a dependent of magnitude and distance. The uncertainty in the location of a future event is considered by confident level. 16 Gail Atkinson Specific (P 18) This SHA hazard curve is inherently limited in scope and applicability. It assumes a fixed distance to a single source, with no uncertainty in the location of a future events being considered. It is simply a transformation of the GutenbergRichter relation (Fig. 2), with a discontinuity imposed at M=5.5. 17 Gail Atkinson Specific (P 20) Include the location of Paducah on Fig. 11, 12. Note that this discussion highlights the fact that the location of a New Madrid event is uncertain, not fixed. 18 Gail Atkinson Specific (P 24.8) The reference to Fig. 10 is incorrect (Fig. 15?) 19 Gail Atkinson Specific (P 26) Include Paducah location on Fig. 15 20 Gail Atkinson Specific (P 28) A maximum magnitude of M6.8 cannot be arbitrarily assigned to the Wabash source in this way. This is a subjective “MCE” with an unknown exceedence probability. It has no physical basis as a limit on magnitude. 21 Gail Atkinson Specific (P 28.5) Add lat, lon to figure 17. 22 Gail Atkinson Specific (P 29) A maximum magnitude of 5 cannot be credibly assigned anywhere in the world. This would imply we have identified all capable faults in the crust with spatial scales of about 1 km or more, and ruled out earthquake motion on any of them. There is no physical basis for such a claim. Worldwide experience has demonstrated time and again that large earthquakes happen, albeit with low recurrence rates, even in stable regions that appear to be nearly aseismic. Assigning Mx=5 to background seismicity is not justified. C-8 See response to the general comment #3. See response to the general comment #3. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 Reviewer Document / Section or Page # Response to Comment Comment 23 Gail Atkinson Specific (P 30) Figure 20 demonstrates that the possibility of a large local earthquake (M6 to 6.5) is not a negligible contributor to hazard. Why is there no contribution from M5 to 5.5 shown on this figure? 24 Gail Atkinson Specific (P 33) It is possible on Fig. 22 that we are seeing a temporary deficit of moderate events due to the after-effects of the 1811-1812 sequence. 25 Gail Atkinson Specific (P 36) The data points for the GR relation for the background seismicity need to be shown. For all zones, the report should clearly show the zone boundaries that are associated with the magnitude recurrence relations. The completeness of the catalogue used needs to be discussed. The conversions from local magnitude scales to moment magnitude need to be presented. Will revise. 26 Gail Atkinson Specific (P 37.4) Discuss why ENA ground motions are higher than CA motions, and point out that this only applies at high frequencies. Will revise. 27 Gail Atkinson Specific (P 37.5) The differences in standard deviation are exaggerated. Most recent studies suggest that sigma should be similar in ENA and CA (eg. EPRI, 2004; Atkinson and Boore, 2006) – about 0.25 to 0.30 in log(10) units (cite units when discussing sigma) in the general case. The differences in standard deviation are in the range of 0.6 to 0.8 (in ln). C-9 This figure is from Peterson (2005) showing that there are earthquakes closer to the site Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 Reviewer Document / Section or Page # Response to Comment Comment 28 Gail Atkinson Specific (P 37.6) The AB95 and F96 relations (and arguably also the T97 relations) do not apply well to large finite sources like New Madrid for which a point source is a poor model. You may wish to quote only finite-fault models. The recent Atkinson and Boore (2006) ENA model uses a finite-fault source. It predicts a PGA of approximately 0.7g for the cited distance of 15 km from an M7.7 event on hard rock. Thus the relevant estimates of median PGA for hard rock, in my view, range from 0.7g (AB06 and S01) to 0.9g (C03). Good comment. Will revise. 29 Gail Atkinson Specific (P 38) Discuss distance measures used in the plots. Have they all been converted to one measure? Note that AB06 is for distance to fault, so in the case of moderate events this is likely to always be greater than a few km (eg. an M5.0 earthquake would likely correspond to about Dfault=10 km at the epicenter). Will revise. 30 Gail Atkinson Specific (P 40) The results are an incorrect assessment of the hazards from these sources, as they do not consider uncertainty in location, nor are the assumed maximum magnitudes for local sources reasonable. The local M5.0 at 15 km is particularly arbitrary. The nearest location for both NMSZ and WVSZ are subject to uncertainty, as are their maximum magnitudes (and recurrence intervals). Note that the combination does not consider the additive nature of the ground motion probabilities. The uncertainties in location and maximum magnitude are considered in the confident level because uncertainty in ground motion is dependent of both of them. Otherwise, uncertainties will be counted twice or three times (in PSHA). Fig. 32 is not actually the probability of exceedence of the ground motions, as the given probabilities relate only to a specific subset (given distance). The effects of sigma on increasing expected ground motion are not included for a given probability are not included. Effects of maximum magnitude on truncating the ground-motion estimates are apparent. Note that the likely importance of the Background seismicity, if extended down to accommodate larger events than assigned Mx=5, is apparent. The annual probability of exceedance (i.e., frequency by McGuire [2004]) is temporal measure. The sigma (ground motion) is spatial measure. The temporal and spatial measures should not be mixed together. 31 Gail Atkinson Specific (P 42) C-10 Also see the response to the general comment #3 on M5.0 and the distance. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 Reviewer Document / Section or Page # Response to Comment Comment 32 Gail Atkinson Specific (P 44.2) The definition of seismic risk given is not correct. See response to comment #8. 33 Gail Atkinson Specific (P 44.5) The conclusion regarding PSHA is not correct. See response to comment #1. 34 Gail Atkinson Specific (P 45) The suggested methodology is seriously flawed and will not result in a defendable estimate of seismic hazard. This could be demonstrated by a Monte Carlo simulation without resort to the PSHA equations. See response to comment #1, 2, and 3. 35 Gail Atkinson Specific (P 46) The ground motions presented can only be considered as judgmental scenario motions, without any associated probabilities. They are not a quantitative hazard calculation. The likelihood of exceeding the motions could be assessed by performing a PSHA using accepted methodologies. Note that the motions are for bedrock, and are likely to be significantly modified by site response. The proposed approach considers separately the associated uncertainties (probabilities) in time and space. 36 Gail Atkinson Appendix The example shows the problem associated with mixing the temporal measure with spatial one. The examples you shown are all “deterministic” interpretation. The probability that PGA exceeds 0.1g is 84 percent if M7 event occurs. An event with 84 percent probability of occurrence is not necessary to occur (statistics), but is interpreted as sure to occur (one event). Similarly, the probability that PGA exceeds 0.1g is 50 percent if M6 event occurs. If earthquake occurrences follow Poisson distribution, the probability that at least one PGA exceeds 0.1g is about 99.3 percent if 10 M6 events occur. This can not be interpreted as 10 events (PGA exceeds 0.1g). The probability that PGA exceeds 0.1g is 16 percent if M5 event occurs. The probability that at least one PGA exceeds 0.1g is about 99.99999 percent if 100 M5 events occur. This can not be interpreted as no event (PGA exceeds 0.1g) to occur. C-11 Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 1 Reviewer Document / Section or Page # Comment Response to Comment Jim Beavers comment #1 (Pi, 1st paragraph) LL Line 7, it is stated: Seismic risk, on the other handed, describes the likelihood (chance) of experiencing a specified level of seismic hazard . . .” In hydraulic engineering, risk can be defined as the probability of a peak discharge being exceeded in a time period, such as 1% of 10,000 cfs being exceeded in one year (Gupta, 1989). Similarly, Comment: I do not think I would call this Seismic Risk. Risk is a concept that denotes a potential negative impact to an asset or some characteristic of value that may arise from some present process or future event. In everyday usage, "risk" is often used synonymously with the probability of a loss. What you are talking about here is frequency of occurrence. I have a risk of an earthquake causing my historic building in Urbana, Illinois to collapse. Thus, I passed this risk on to my insurance company. According to McGuire (2004), seismic hazard is “a property of an earthquake that can cause damage and loss. A PSHA determines the frequency (the number of events per unit of time) with which a seismic hazard will occur.” Because magnitude is a property of an earthquake, the larger magnitude, the higher potential to cause harm, a magnitude M or greater with a MRI is seismic hazard. Similarly, MMI or ground motion at a site is a property of an earthquake, MMI VIII (or PGA 0.25-0.30g) or greater with a return period is seismic hazard. MMI VIII is described to have a considerable damage to ordinary buildings. Consequently, a considerable damage or greater to ordinary buildings at a site with a return period is seismic hazard, too. Therefore, measurements of seismic hazard can be different, from magnitude to damage (loss) level to buildings, and one measure can be converted to another through a statistical relationship (i.e., ground motion attenuation and fragility curve). As defined by McGuire (2004), seismic risk is “the probability that some humans will incur loss or that their built environment will be damaged. These probabilities usually represent a level of loss or damage that is equaled or exceeded over some time period.” A similar definition was described by Reiter (1990),” seismic risk is the probability of occurrence [in time] of these consequences.” From these definitions, seismic risk is quantified by three elements: probability, a level of consequence (damage or loss), and time. Because damage or loss is also a property (measure) of an earthquake, the likelihood (probability) of its (M or greater) occurrence during a specific time period is risk. C-12 Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 Reviewer Document / Section or Page # 2 Jim Beavers Page i, Paragraph 2. Last sentence it is stated: “Temporal and spatial uncertainties are of different characteristics and must be considered separately in hazard assessment.” I think I disagree with this statement. 3 Jim Beavers Since this subject is quite controversial I, as a reviewer, will be expecting to see considerable detail in the report about how this process is better than the PSHA process, sort of a one on one comparison. This report has a detailed description and discussion on PSHA and SHA. There are also several references on these. 4 Jim Beavers Page i, Starting in Paragraph 3. Line 16 it states: “There is a mathematical error in the . . .” This discussion is continued through paragraph 4. Page 1, Paragraph 1. Line 13, it states: “For example, it would not be feasible for the U.S. Department of Energy to obtain a permit from Federal and State regulators to construct a landfill at the Paducah Gaseous Diffusion Plant . . .” I do not believe you can say this, because we do not officially know that it is not feasible. Where is the feasibility study that says it is not feasible? In fact, the CERCLA Cell report for Site A had a peak ground motion design value of 0.48g. The CERCLA Cell project was stopped for political reasons not technical. This statement reflects the fact that Kentucky Solid Waste Division refused to issue the permit by citing the USGS hazard estimate. Response to Comment Comment C-13 Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 Reviewer Document / Section or Page # Response to Comment Comment 5 Jim Beavers Page 1, Paragraph 2. Line 4, it is stated: Currently, the highest building design PGA used in California (UBC97) is capped at about 0.4g. This is true; however, I believe that this capped value will be removed shortly because it truly underestimates the hazard in California. This cap was imposed by a bunch of engineers in the mid 1980’s. With the deterministic cap and the NGA attenuation relationships (near-source saturation), this cap (0.4g) may still be valid. 6 Jim Beavers Page 1, Paragraph 2. Line 12 it is stated: “It clearly shows that the higher design ground motion in western Kentucky does not make sense scientifically. This is where you are going to have to show why the PSHA is indeed the incorrect way to consider the uncertainties, to convince me and others. Comparing earthquake activity in California to the New Madrid Seismic Zone (NMSZ) may make sense to the layman, but I can see where the PSHA approach might make sense, especially with the body of literature out there that continues to support PSHA, especially the EPRI and LLNL methodologies. I really believe the DSHA does not consider all of the uncertainty. I have had a lot of discussion on this DSHA with Ellis Krinitzky and was not convinced that DSHA considered all of the uncertainty. However, in the 70’s and 80’s I would look at the PSHA approach to seismic hazard and then the DSHA approach and then make a judgmental decision on what the seismic design basis should be for a DOE facility. As shown in this report, there is a mathematical problem: treating the groundmotion uncertainty as an independent random variable. As it modeled in modern attenuation relationships, the ground-motion uncertainty is not an independent random variable. 7 Jim Beavers Page 3, Paragraph 1. This paragraph is right on target. 8 Jim Beavers Page 3, Paragraph 2. First line it is stated: “Objectives of this project are . . .” I would think one objective would be to clearly show why the PSHA approach overstates the seismic hazard. C-14 With this mathematical problem, PSHA is difficult to understand and use. PSHA approach may understates the seismic hazard. For example, the ground motions with 500-year return period is considered to be low in the New Madrid area. The end result from PSHA is a hazard curve from which one could not tell it is a high or low estimate. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 9 Reviewer Document / Section or Page # Comment Jim Beavers Page 4, Paragraph 1. First sentence it is stated: “Two methods . . .” I currently do not believe that both PSHA and DSHA are commonly used today. I think the use of PSHA for outweighs the use of DSHA. In a recent correspondence with from John Schneider (Geoscience Australia), he states: Response to Comment This is an interesting comment. “In fact, in many instances, the deaggregation of a probabilistic analysis has been used to identify and justify specific scenarios, which are in effect deterministic solutions.” I find it puzzling that there is still a debate over this issue. In my view PSHA is merely a means of formally accounting for uncertainty. I can’t imagine why anyone would have any philosophical objection to that! In fact, in many instances, the deaggregation of a probabilistic analysis has been used to identify and justify specific scenarios, which are in effect deterministic solutions. In short, I don’t know anyone apart from Ellis in the deterministic camp. 10 Jim Beavers Page 4, Paragraph 4. Last sentence states; “Wang (2004) . . .” I assume Wang (2004) is Wang, Z. (in press) reference document at bottom of page 53 or is it Wang et al. 2004. Also on page 53 you have a Wang 2003 with no title or reference. In addition, I would suggest you list your reference based on name and then earliest date, i.e., Wang 2003 would come before Wang 2004 in your reference list. The reference list needs to be verified, e.g. later in the report you reference Wheeler 1997 and site SRL Vol. 63 which should be Vol. 68. These will corrected. 11 Jim Beavers Page 5, Paragraph 2. Second sentence it states: “The probability that no earthquake will . . .” Suggest this say: “The probability that no such earthquakes will . . .” Revised. C-15 Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 Reviewer Document / Section or Page # 12 Jim Beavers Page 5, Paragraph 2. Fourth sentence it states: “Equation (3) shows the relationship between seismic risk, . . .with X percent in PE in Y years, and seismic hazard, expressed . . .” 13 Jim Beavers Page 5 Paragraph 2 continued on page 6. Last sentence it states: “Equation (3) also shows that the probability p shows . . . and has no relation to spatial characteristics of the hazard . . .” 14 Jim Beavers Page 6, Paragraph 2. General Response to Comment Comment In the introduction, I think you need to clearly state what is meant by seismic risk and seismic hazard in the introduction and stick with that notation throughout the document. See Comment on Item 1. What you are calling seismic risk I still see as frequency because a 10% chance in 50 years has a frequency on the average over hundreds of thousands of years every 475 years. In addition, changing time interval notation in Equation (3) from t to Y could leave the reader confused. Another example of using the words “seismic risk” is the Ohio Department of Natural Resources where they state: “The brief historic record of Ohio earthquakes suggests a risk of moderately damaging earthquakes in the western, northeastern, and southeastern parts of the state.” Here the risk is in terms of potential damage. This is true, however we are only talking about PE of a magnitude M or greater in a certain source zone. However, to mitigate effects of the hazard’s occurrence I must design my building for a peak ground acceleration or spectral value. Thus, I have to know where the earthquake is going to occur because of the attenuation factors which are directly spatially related. Even in a DSHA, I still have to put the earthquake some place to get my design values. In the old days we put it right under our site. Seismic hazard and risk are two different concepts. They have been used interchangeably quite often. The attempt in this report is to distinguish and use them consistently. I basically agree with this paragraph assuming your Gutenburg –Richter curve represents the earthquake activity of the NMSZ. However, when I got to pages 16 through 18 I realized that you had labeled Figure 2 wrong. The abscissa should be labeled N, not Log (N). See also Figure 23. Corrected. C-16 Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 Reviewer Document / Section or Page # Response to Comment Comment 15 Jim Beavers Page 6, Paragraph 2. Last sentence it states: “The risk posed by . . .” This is still frequency to me. By common definition, a frequency is used to describe how often an event occurs, is not a probability to occur over a time period. 16 Jim Beavers Page 6, Paragraph 3. First sentence states: “In practice . . .” To me this is where the spatial aspects come into the equation. Agree. 17 Jim Beavers Page 6, Paragraph 3 continued on Page 7. Fifth sentence states: “From Figure 3 a mean annual . . .” There needs to be some definition of Pf before it is introduced here. Revised. 18 Jim Beavers Page 6, Paragraph 3 continued on Page 7. Seventh sentence states: “Similarly, the ground motion and their MRI’s at a site . . .” Here you are going from Equation (3), which you justified on page 5 as the probability of earthquakes equal to or greater than a specific size (M) with X percent PE in Y years, and . . ., which I agree with, and now all of a sudden you are implying that it is equally compatible to replace (M) with ground motions. I do not think you can do this???? It is a simple mathematics. From equation (6), ln(Y ) = f ( M , R) + nσ ln,Y , (6) We have M = g ( R, ln Y , nσ ln Y ) . (16) Combining equation (16) with equation (15) τ= 1 = e − 2.303a + 2.303bM N (15) Results in τ= C-17 1 = e − 2.303a + 2.303bg ( R ,ln Y ,nσ ln Y ) N (17) Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 Reviewer Document / Section or Page # 19 Jim Beavers The Milne and Davenport attenuation curves do consider only an estimated value and show no concept of the uncertainty in the ground motions. See Bommer and Abrahamson 2006. This is true and is addressed in this report, the equation (17). 20 Jim Beavers Page 6, Paragraph 3 continued on Page 7. Seventh sentence it states: “Similarly, the ground motions and their MRI’s . . .” Page 6, Paragraph 3 continued on Page 7. Eighth sentence it states: “An empirical method, which is identical to the empirical floodhazard analysis . . .” . . .” The Milne and Davenport paper is just that, an empirical paper that uses an attenuation equation that has no uncertainty and basically is a measured methodology using assumptions that most would not be considered appropriate today. As a result, I think this approach may underestimate the seismic hazard. The proposed approach is to consider the uncertainty. A similar approach has also been proposed by Stein and others (2005, 2006). 21 Jim Beavers This is correct if you use equation 3; however, I question using equation 3 for PGA especially when you based your justification for equation 3 on probability of earthquakes equal to or greater than a specific size (M) with X percent PE in Y years. See Comment 17. See response to comment #18. 22 Jim Beavers Page 6, Paragraph 3 continued on Page 7. The eleventh sentence states: For a building with an exposure . . .” Page 8, Paragraph 1. Third sentence it states: “Seismic risk, on the other hand, describes a probability of . . .” Again, I am having some trouble calling this seismic risk. I think of it as a frequency. See response to comment #15. Response to Comment Comment C-18 Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 Reviewer Document / Section or Page # Comment 23 Jim Beavers Page 8, Paragraph 1. Fourth Sentence it states: Seismic risk not only depends on seismic hazard . . .used to describe the occurrences of earthquakes” I agree that seismic risk depends on seismic hazard, exposure and model. My problem is the model where with the leap of faith from justifying the Poisson model (equation 3) based on the on probability of earthquakes equal to or greater than a specific size (M) with X percent PE in Y years and then saying that is the same for ground motion. See Comments 11, 12, and 17. To introduce the ground motion parameter requires a spatial element as noted in Comment 12 and 15. See response to comment #18. 24 Jim Beavers Jim Beavers I agree, but not for the same reasons. If there is a high seismic hazard geographic area and I build an important building in that area that costs $5 million I have a high seismic risk. However, if I build a small cattle barn that cost $500 I don not have a high seismic risk. This is why there are no nuclear power plants within a 120 mile radius of the NMSZ. I agree, but I do not agree with your supporting logic, because you are only considering frequency of magnitude of events and not the uncertainty of ground motion. This is good example showing that lower exposure (building) gives you lower risk even though hazard is high. 25 26 Jim Beavers Page 8, Paragraph 2. Second sentence states: “High seismic hazard does not mean high seismic risk . . .” Page 8, Paragraph 2. Third sentence states: “Moreover, the mitigation policy is mostly . . .” Page 8, Paragraph 2. Last sentence states: “That is why we have to spend more resources . . .” I agree with the statement but disagree with the implied reasoning. We are spending more resources and effort to mitigate seismic hazard in San Francisco because they have a greater seismic risk as a result of the built environment and population density, and they understand their seismic hazard better than those in the NMSZ. The comparisons in the report are based on the same exposure. Higher exposure makes the comparison more valid. Response to Comment C-19 The uncertainty of ground motion is considered by a level of confidence, like the flood risk. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 27 Reviewer Jim Beavers 28 Jim Beavers 29 Jim Beavers Document / Section or Page # Page 9, Table 2. Comparison of Hazard and Risk Page 10, Paragraph 3. This paragraph starts with: “According to Benjamin and Cornell . . .” Page 11, Paragraph 2. This paragraph starts with: “As demonstrated above . . .” Response to Comment Comment This table is accurate with respect to the probabilities about MRI’s of earthquakes having various magnitudes, but it doesn’t stand up for considering ground motion MRI’s. In this respect, in the Wang and Ormsbee 2005 EOS paper it is stated: “Figure 2 shows that PGA with 2% PE in 50 years is 0.97g.” It is then stated: “This PGA (0.97g) does not mean that it could occur in 2500 years: but rather that there are 0.0835, 0.0294, and 0.0086 probabilities that PGA will exceed 0.97g if each of the three earthquakes occur.” In my view it means that the probability of exceedance of a 0.97 PGA will occur on the average once every 2500 years over hundreds of thousands of years. While this is mathematically true, I currently believe that there is enough independence of the ground motion uncertainty that E can be treated as an independent variable, like M and R. In the Wang and Ormsbee 2005 EOS paper it is stated: “Figure 2 shows that PGA with 2% PE in 50 years is 0.97g.” It is then stated: “This PGA (0.97g) does not mean that it could occur in 2500 years: but rather that there are 0.0835, 0.0294, and 0.0086 probabilities that PGA will exceed 0.97g if each of the three earthquakes occur.” While this is true in the explicit sense. At the present state of knowledge, I do not see an alternative. Maybe this is why you have been having a hard time convincing others about your approach. Along those lines, it is very interesting to me that you reference Shakal and others, 2006 research on the M 6.0 Parkfield to position your justification that the ground motion uncertainty is dependent on M and R or both and at the same time Bommer and Abrahamson (2006) in the BSSA are using the M 6.0 Parkfield event to clearly show the uncertainty of ground motion for any earthquake. The key point here is that the distance being measured for a finite fault (modern attenuation), in comparison with the distance being measured for a point source. In reality I know that both the ground motion are dependent on both M and R, because if you do not have M you do not have ground motion and until you know R you do not know what levels the ground motion will be. But it looks like to me they (your nonbelievers) have a pretty good justification, so far, that the uncertainty is independent of M and R. C-20 A probability of 0.0835, 0.0294, or 0.0086 that PGA will exceed 0.97g if each of the three earthquakes occurs does not mean this will occur. No response. We agree with that “In reality I know that both the ground motion are dependent on both M and R, because if you do not have M you do not have ground motion and until you know R you do not know what levels the ground motion will be.” This will results in different formulation for hazard calculation. In other words, current PSHA has a mathematical problem. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 Reviewer Document / Section or Page # 30 Jim Beavers Page 11, Equation 10. 31 Jim Beavers Page 12, Last Paragraph. This paragraph starts with Equations (11) through (13) . . .” 32 Jim Beavers Page 12, Last Paragraph. Last sentence, here you use the term ergodic assumption. Response to Comment Comment Below this equation you describe σsource and σpath and do not describe σmodeling is there a reason for this? I do not have EPRI 2003 to verify. It appears that you are using equations (11) through (13) to show that the PSHA result in an invalid formulation. But it is not clear to me what you are trying to say. At first brush, it looks like to me you are saying the following I have equation 13 which says Trp (y) = T divided by the uncertainty of ground motion and since Trp (y) is a return period and T is the characteristic earthquake return period, they are the same so equation 13 is invalid. However in your EOS paper you imply that if I have a characteristic earthquake of return period T at some distance R and probability of exceeding a certain ground motion that the probability of the ground motion being exceed at the site of interest is (1/T)x(probability of exceedance). For characteristic earthquakes, I believe this is the correct approach if you know the distance to the site of interest. In my mind I think equation 13 is still good because Trp (y) is the return period of (y) being exceed while T is the return period of the characteristic earthquake. See earlier comments 13, 18, 21 and 23. σmodeling describes modeling uncertainty. This is also called Chaos Theory, if you look at Bommer and Abrahamson (2006), you might call the uncertainty of ground motion that because the spread is one order of magnitude based on the M 6.0 at Parkfield. The term ergodic assumption was defined by Anderson Brune (1999). C-21 These equations show the fundamentally difference between the recurrence interval (T) of an earthquake and the return period (Trp) of a ground motion that is generated by the earthquake at a site. Occurrence of a ground motion at a site must be associated with an earthquake. There would not be a ground motion at a site if there is no earthquake. However, PSHA could produce a range of return period from a single recurrence interval. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 Reviewer Document / Section or Page # 33 Jim Beavers Page 15, First Paragraph. Fifth sentence that starts with: “This interpretation fundamentally . . .” 34 Jim Beavers Page 15, First Paragraph. Eighth and ninth sentences that both start with: “Figure 9 shows . . .” 35 Jim Beavers Page 15, First Paragraph. Last sentence where it states: “In other words . . .” 36 Jim Beavers Page 15, Second Paragraph. First sentence. Response to Comment Comment I agree that in the discussion above that it kind of gets ludicrous when we go talking about a 100-million year earthquake. However, I really do not think you are changing the physical and statistical meanings except maybe to the lay person. We all know that this still remains a probability of occurrence. Going to a return period is just the nature of the Beast and we need to live with it whether we are talking about a 100 year return period in flooding or a 100 million return period in earthquakes. I guess a 100 million year return period in terms of magnitude would be an Mw of 12.0 which as I recall would split the earth in half. It is quite clear to me that if you have a M 7.7 that has an MRI of 500 years in the NMSZ and the uncertainty does exist per Campbell (June 2003), BSSA, that you will have the uncertainty show in Figure 9and if the median PGA is 0.36 g then the probability of exceeding 0.36 g, given the earthquake occurs, is (1/500)x0.5 or 0.001 annual frequency or an event that has a return period of 1000 years of during the 50 year life of a building there is a 5% chance the building will experience that kind of ground motion. In past designs, the rule was to design for a 10% chance in 50 years which is the “500 year earthquake” in better words (more accurate) an earthquake that might occur from the characteristic fault that could cause the building experience at PGA of 0.36g or more in its life time. I think you are overstating the case when you say: “. . . however, to mean that that ground motion will occur at least once in 2,500 years . . .” My question is who has been interpreting a 20 percent probability of being exceeded in 500 years as being the ground motion occurs at least once in 2,500 years? They should be interpreting it as: “on average, over hundreds of thousands of years, this ground motion will be exceeded once every 2,500 years. I think I disagree with this statement and do not support the logic you have used thus far that mixing temporal and spatial measurements is causing any kind of problem, especially you discussion in the first paragraph. See earlier comments 13, 18, 21, 23, 33, 34, and 35. C-22 Good comment. Good comments. These show the differences between PSHA and the proposed approach. According to McGuire (2004), return period is: “the mean (average) time between occurrences of a seismic hazard – for example, a certain ground motion at a site, or a certain level of damage or loss.” Frankel (2005) and Holzer (2005) interpreted exactly that way. Temporal and spatial measurements are two the most fundamental elements of the world. Mixing them one way or the other would cause problem. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 Reviewer Document / Section or Page # 37 Jim Beavers Page 15, Second Paragraph. Second sentence that starts with: “Temporal and spatial . . .” 38 Jim Beavers Page 15, Second Paragraph. Last sentence. 39 Jim Beavers Pages 16 through 18. Section 2.2 New Approach— Seismic Hazard Assessment. Response to Comment Comment I am confused. Here like you are saying that the temporal measurements (M) and spatial measurements (ground motions) are two intrinsic independent characteristics of and earthquake . . . and must be treated separately. If that is true, why can’t I consider M, R, and ground motion as independent events for PSHA? I think you are overstating the issue. I do not think of it as being inappropriate or confusing, only to the lay person or engineer that has no experience in seismic design. Based on the two DOE projects, one at Portsmouth and one at Paducah, in which I am the DOE site reviewer there are a number of engineers in the Midwest and east that are not familiar with seismic design. When I started reviewing this report, especially when I saw your table in the Executive Summary, and knowing certain issues you have had with the USGS methodology and viceversa, I thought I would do a DSHA to see what I get and how it compares to your results. Based on the PSHA work that had been done for me at Paducah (McGuire 1999 (REI 99)) and my use of the USGS methodology I had access to the deaggregations. The deaggregations for both (McGuire and USGS) show that a magnitude M 7.5 or 8 was driving the PSHA ground motions 20 kilometers (km) from the PDGP. The 20 km is based on (Johnston and VanAresdale, Appendix to REI 99). So I said: “Ok, let’s have a DSHA earthquake of M 8.0 occur 20 kilometers from the PGDP and let’s also be more realistic and have an M 8.0 occur 60 km from the PGDP where the February 11, 1812 event occurred. After I did these I decided to look at it from your perspective of 30 km. …. C-23 M, R, and ground motion at a site are not temporal measurements. Unfortunately, it happens all the times. Your analyses show how PSHA can derive different return periods for a single earthquake with a recurrence interval. If an earthquake occurs every 500 years, the ground motion generated by the earthquake at a site must also occur every 500 years. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 Reviewer Document / Section or Page # 40 Jim Beavers Page 22. Last paragraph. 41 Jim Beavers 42 Jim Beavers 43 Jim Beavers 44 Jim Beavers 45 Jim Beavers Page 20. Last paragraph. Last sentence which starts out as: “This can be seen clearly . . .” Page 21. First paragraph. Sixth sentence where it states: “These short period and dense network . . .” Page 21. Second paragraph. Last sentence that starts with: “Thus, there is no evidence . . .” Page 21. Last paragraph. Last sentence where it states: “In this report, we used the location . . .” Pages 24-28. Section 3.2. 46 Jim Beavers Page 29. First paragraph. Response to Comment Comment In this paragraph it appears that you reference Wheeler 1997 in support the NMSZ extending northeastward toward the PGDP and site Wheeler 1997 in support of it not extending toward the PGDP. This is confusing to the reader unless you quote statements made by Wheeler showing his own uncertainty on the issue. As I recall you have done this elsewhere in the document. As an engineer when I look at Figure13 it doesn’t mean a thing to me. You need to explain what I am supposed to be seeing. Also, if I look at Figure 4 of Braile et al. 97 it looks to me like Johnston and VanAresdale (REI 99) have a justification for the northeast extension. These will clarified. In what way do these observations suggest that the characteristics of earthquakes in the Jackson Purchase Region are different from those of earthquakes in the central NMSZ? In terms of stress field and focal dept. Will revise. I think the jury is still out on this. I agree with using a maximum magnitude of M 7.5. I have read this section and am not going to comment as I feel it has little bearing on PGDP. I don not have a copy of Peterson 2005 although I was at the workshop. C-24 It is summary for a meeting between KGS and USGS in Lexington. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 Reviewer Document / Section or Page # 47 Jim Beavers 48 Jim Beavers 49 Jim Beavers 50 Jim Beavers 51 Jim Beavers Page 29. First paragraph. Fourth sentence that starts with: “The use of these large background earthquakes . . .” Page 32. First paragraph. Last sentence that starts with Figures 22 and 23 show . . .” Page 32. Last paragraph. Fifth sentence that starts with: “A recent study by Holbrook and others (2006) . . .” Page 37. Last paragraph continuing on to Page 38. Ninth sentence that starts with: “As shown in the Figure, Frankel . . .” Page 40. Last paragraph. First sentence that starts with: “Figure32, shows median PGA . . .” Response to Comment Comment I believe they do if you are doing a PSHA and are needed for completeness. It has also been shown by Frankel (2004) and Petersen (2005). In this sentence you imply that Figure 23 is for the Wabash Valley Seismic Zone; however, this figure is labeled as magnitude-occurrence relationship of the NMSZ. Will correct it. Just before this sentence there seems to be some missing or misrepresenting text because at the end of the fourth sentence it states: “. . . New Madrid Seismic Zone However (Fig 22). Will revise. The figure actually shows Frankel attenuation curve at near source similar to Campbell (2003). It is AB06 that is higher in the near source. Maybe Frankel and others did not get put on the graph, because the one I first thought was Frankel and others now looks like it is Silva-DC-S The comparison should be at distance between 10 and 40 km. Frankel and others (1996) did not provide values less than 10 km. I am confused here. You talk about using Campbell (2003) attenuation equations in the earlier parts of the document and all of a sudden here, for your detail work, you say you are going to use Atkinson and Boore (2006) which, in Figure 27, has the highest near source attenuation values, but in Tables 7 through 8 you us all attenuation equations except Frankel. In this report, we used the ground-motion attenuation relationships of Somerville and others (2001), Silva and others (2002), Campbell (2003), and Atkinson and Boore (2006). Figures 29 and 30 show 0.2s and 1.0s response accelerations of the four attenuation relationships for an M7.5 earthquake in the central United States. C-25 Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 552 Reviewer Document / Section or Page # Jim Beavers Page 40. Last paragraph. Last sentence that starts with: “Tables 7, 8 and 9 list the the PGA . . .” Response to Comment Comment I am also confused as to how you got these numbers. The old building code process required seismic design of a building to be designed for an earthquake that had a 10% probability of being exceed during its assumed life. The assumed life was 50 years. For a 10% probability of exceedence in 50 years represents an event that occurs every 475 years to be exact or 500 years This turns out to be the return period of the New Madrid earthquakes as you have said and you have called them characteristic earthquakes and rightfully so. If the characteristic earthquake occurs you showed in Figure 10, page18 that the mean PGA would be 0.44g so how could your mean PGA ground motions in Table 7 be below 0.3g. You need to have more discussion in you report on how you got these numbers and the justification for it. C-26 0.44g is the median PGA for a site at 30 Km distance. Table 7 is for PGAs at a site of 45 km. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 1 Reviewer Leon Reiter Document / Section or Page # General Comment Response to Comment In general, I have found that the draft report is lacking in technical justification for a number of the methods used and the assumptions made. This is particularly true for the proposed approach called "Seismic Hazard Analysis" (SHA) and the definitions of seismic hazard and seismic risk. Some of my criticisms may be due to the draft report's lack of clarity in explaining and justifying what was done. A clearer explanation may alleviate some, but not all, of my concerns. This report is not a typical site-specific seismic hazard assessment, but a summary of scientific research on geological and seismological conditions, the methodologies, and the seismic hazard assessment related to the Paducah Gaseous Diffusion Plant and the surrounding area. Therefore, it may be reviewed in a different way than a normal sitespecific technical report. The proposed approach, SHA, is not really a new one, but an old one (Milne and Davenport, 1969) with inclusion of ground motion uncertainty. A similar approach has also been proposed by Stein and others (2005, 2006). SHA is analogous to flood, wind, and other hazard analysis and technically sound. The definition of hazard and risk used in this report follows the accepted convention, particularly in engineering (hydraulic, flood, wind, and snow). These definitions are also consistent with those of McGuire (2004) and Reiter (1990). A better explanation on the methods used and the assumptions made will be addressed. 2 Leon Reiter Specific comment #1 (P 1, 2nd paragraph) How can Figure 1 show that that higher seismic design in western Kentucky doesn't make sense when the total recording period is only one week? During one week you could be seeing the effects of a swarm that could give you an atypical increase in seismicity or seismic quiescence that would show anomalous low seismicity. If you want to make this point show a longer period of time. Revised to use Stein and others (2003) 3 Leon Reiter DSHA does not (as stated in (2)) require the determination of earthquake occurrence frequencies. True. 4 Leon Reiter Specific comment #2 (P 4, 1st paragraph) Specific comment #3 (P 4, last line) There is no Wang (2004) in the list of references. Is this Wang (2003), which is listed, but without a title? It should be Wang and others (2004). C-27 Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 5 Reviewer Leon Reiter Document / Section or Page # Specific comment #4 (P 5, Section 2.1.1) Comment Response to Comment In this section and at other locations in the text the authors introduce their definitions of seismic risk and seismic hazard. These definitions are unclear and cause confusion. The commonly accepted definitions of hazard and risk (e.g., Reiter, 1990, McGuire, 2002) define seismic hazard as those earthquake-related properties that have a potential to cause damage or loss. Seismic hazard may be described deterministically (DSHA) or probabilistically (PSHA). Seismic risk is the probability of occurrence of adverse consequences from seismic events to humans or their built environment. This fits in with the classic definition of risk (Kaplan and Garrick, 1981) stating that risk analysis answers three questions: what can go wrong, how likely is it to happen, and what are the consequences or outcomes. According to the authors (bottom of p. 5) "Equation (3) [the probability of at least one earthquake with magnitude equal to or greater than a specific size occurring in t years] shows the relationship between seismic risk, expressed in terms of an earthquake magnitude (M) with X percent PE in Y years, and seismic hazard, expressed in terms of an earthquake with a magnitude M or greater and its MRI [mean recurrence interval] in an area or along a fault." Thus, according to the authors, the magnitude of an earthquake (and its mean recurrence interval) represents the hazard and the likelihood of its occurrence during a specific time period represents the risk. These are simply different ways of expressing the same information. Risk, in this case, assumes a Poisson model of earthquake occurrence. (continue to next page) Seismic hazard and risk are two fundamentally different concepts. Seismic hazard is a natural phenomenon generated by earthquakes, such as ground motion, and is quantified by two parameters: a level of hazard and its mean return interval (MRI) or frequency. Seismic risk, on the other hand, describes a probability of occurrence of a specific level of seismic hazard over a certain time, and is quantified by three parameters: probability, a level of hazard, and exposure time. These definitions are consistent with those defined by McGuire (2004) and Reiter (1990). C-28 According to McGuire (2004), seismic hazard is “a property of an earthquake that can cause damage and loss. A PSHA determines the frequency (the number of events per unit of time) with which a seismic hazard will occur.” Because magnitude is a property of an earthquake, the larger magnitude, the higher potential to cause harm, a magnitude M or greater with a MRI is seismic hazard. Similarly, MMI or ground motion at a site is a property of an earthquake, MMI VIII (or PGA 0.25-0.30g) or greater with a return period is seismic hazard. MMI VIII is described to have a considerable damage to ordinary buildings. Consequently, a considerable damage or greater to ordinary buildings at a site with a return period is seismic hazard, too. Therefore, measurements of seismic hazard can be different, from magnitude to damage (loss) level to buildings, and one measure can be converted to another through a statistical relationship (i.e., ground motion attenuation and fragility curve). As defined by McGuire (2004), seismic risk is “the probability that some humans will incur loss or that their built environment will be damaged. These probabilities usually represent a level of loss or damage that is equaled or exceeded over some time period.” A similar definition was described by Reiter (1990),” seismic risk is the probability of occurrence [in time] of these consequences.” From these definitions, seismic risk is quantified by three elements: probability, a level of consequence (damage or loss), and time. Because damage or loss is also a property (measure) of an earthquake, the likelihood (probability) of its (M or greater) occurrence during a specific time period is risk. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 Reviewer Document / Section or Page # Response to Comment Comment There is no mention of the critical issue of consequences such as building damage or loss of life. Using their definitions, the same information is needed to define hazard and risk. The authors are using their definitions to make a point. Frankly, I am not sure why they chose these definitions and am not sure why they chose these definitions and what point they are trying to make. If they insist on this approach they should systematically explain how they differ from the classic definitions of hazard and, particularly, risk and why they are using these definitions. I have unsuccessfully attempted to find clearer definitions and rationales in some of the other papers the authors have written. 6 It is very important to mention the assumption of a Poisson model of earthquake occurrence (in time). The risk (probability) calculations throughout the report are based on this assumption. The probability will be different if a non-Poisson model of earthquake occurrence is assumed. This is one of the differences between seismic hazard and risk: in order to estimate seismic risk, we have to make an assumption on earthquake occurrence in time (Poisson or non-Poisson). Seismic hazard is estimated from observation (data). The other important parameter, exposure time, is also very important to mention here. The exposure time is a normal life time or considered time for something (building, dam, bridge, etc.) being exposed to the hazard. The exposure time and physical content (regular two-story house, concrete dam, etc.) are properties of something being exposed, but not properties of an earthquake. Therefore, seismic risk is an interaction (or so-called product) of seismic hazard and something being exposed. Thus, seismic hazard and risk are different. Will revise. 7 Leon Reiter Specific comment #5 Figure 2 and Figure 23. Vertical axis should be "N" not "Log (N)" 8 Leon Reiter Specific comment #6 (P 7) Figure 3. The authors give an example of flood hazard and say that they can convert this to risk, by using equation (3). I did a quick foray into the web looking at definitions of general, and flood, hazard and risk. These definitions make use of the classic definition I mentioned above with respect to seismic hazard and risk, i.e., adding the component of consequences (e.g., building vulnerability and loss of life). See response to the specific comment #4 9 Leon Reiter Specific comment #7 (P 9) Table 2. When MMI is used an argument could be made that this is true risk because it considers the level of damage. If MMI is ok, why not M? (response to the specific comment #4) C-29 Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 10 11 Reviewer Leon Reiter Leon Reiter Document / Section or Page # Specific comment #8 (P 11) Specific comment #9 (P 12) Comment Response to Comment 1st full paragraph. The authors raise an important point here that the uncertainties may not be independent. I am not sure whether they are correct, but it seems to me that even if they are correct it may be a necessary evil that we try to work around, but can't get rid of completely. This is something I would be happy to hear discussed by my colleagues at the review panel meeting. The authors also claim that Bommer and Abrahamson (2006) attribute the large uncertainty in Figure 6 to the use of site-fault distance rather than epicentral distance. However, Bommer and Anderson (2006) argue that the large variability reflects the variability due to wave propagation from a finite fault that is characterized only by the distance from the station to the closest point on the fault. 1st full paragraph. I don't understand how "Equations (11) through (13) demonstrate that the invalid formulation of PSHA results in extrapolation of the return period from the recurrence interval of the earthquake and the ground-motion uncertainty …or the so called ergodic assumption (Anderson and Brune, 1999.)" Anderson and Brune (1999) showed that when determining hazard for a specific scenario (e.g., x km from the San Andreas fault), the use of generalized attenuation equations based on many earthquakes may overestimate the hazard when compared to ground motion-like data (precarious rocks) that exist for that scenario. They argued that the aleatory uncertainty in the generalized attenuation equations included epistemic uncertainty that could be reduced when a specific scenario is being considered. Do the authors have any data like this that could be used to reduce the uncertainty in the Paducah hazard analysis? This could be another good topic for review panel discussion. In the ground motion attenuation relationships, R is measured as rupture, JB, or seismogenic distance. The ground motion standard deviation will be different if different R is used (R dependent). fR(r) in Eq. 4 is to account for the uncertainty of focal point (distribution). The uncertainty of focal point is accounted in part by the uncertainty of ground motion because R is measured as a single distance (rupture, JB, or seismogenic) regardless focal distance. Eq.4 counts the distance uncertainty, at least some portion, twice. C-30 Similarly, fM(m) in Eq. 4 is to account for the uncertainty of magnitude (distribution). Also similarly, the ground motion standard deviation is dependent of M. Again, Eq.4 counts the magnitude uncertainty, at least some portion, twice. These will be fully discussed at the review meeting. Ground-motion uncertainty has been separated into aleatory and epistemic parts. But it is difficult to do so, particularly in the CEUS. This will be discussed at the meeting. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 Reviewer Document / Section or Page # 12 Leon Reiter Specific comment #10 (P 15) 13 Leon Reiter Specific comment #11 (P 15) Response to Comment Comment 1st paragraph, first sentence. The authors state that geologic records of earthquakes are limited to the past 11,000 years (Holocene). This is not true. Many records go back much longer, e.g., the area around Yucca Mountain contains geologic records of earthquakes that go back many hundreds of thousand of years. 1st paragraph, last line (see also statements in the middle of the first paragraph.) I can't find the statement in Frankel (2005) which says that ground motion with a 2500 year return period will [authors' emphasis] occur at least once in 2500 years. On the contrary, Frankel (2005) talks about the ground motion being exceeded once on average [my emphasis] over 2500 years. Also, in a response to Wang and Ormsbee (2005), Holzer (2005) clearly states that the 2500 year PGA is not guaranteed [my emphasis] to occur in 2500 years. How important is this to the authors' criticism of PSHA? C-31 But not hundreds of million of years. Figures 1 and 2 in Frankel (2005) shows that (which is the acceleration that will be exceeded). Frankel’s explanation is a “deterministic” interpretation. An event with a 63% probability of occurrence may not occur, but was interpreted and shown to occur. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 14 Reviewer Leon Reiter Document / Section or Page # Specific comment #12 (P 15) Comment Response to Comment last paragraph. The authors' statement that PSHA is invalid because it inappropriately mixes temporal measurement (occurrence of an earthquake and its ground motion) and spatial variation (ground motion uncertainty due to source, path, and site effects), appears to be a key point in this report that needs to be clarified. I don't understand how spatial variation (as defined above) cannot be taken into account (if that indeed is what the authors are stating) when describing the likelihood of exceeding a given ground motion over a period of time. If there were no spatial variation, every time an earthquake occurred we would more likely know what the ground motion would be. Because there is spatial variation (much of which is assumed to be random based on current knowledge), the likelihood of reaching a certain ground motion when an earthquake of given size occurs has to be different, because of increased uncertainty, than if there were no spatial variation. Eventually I assume we will increase our knowledge of spatial variations such that we will have a better idea of what the source, path and site effects are and they won't be assumed to be random. It was stated that “the invalid formulation causes PSHA mixing the temporal measurement (occurrence of an earthquake and its consequence [ground motion] at a site) with spatial measurement (ground-motion uncertainty due to the source, path, and site effects).” C-32 Temporal and spatial measurements are two intrinsic and independent characteristics of an earthquake and its consequence (ground motion) at a site, and must be treated separately. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 15 Reviewer Leon Reiter Document / Section or Page # Specific comment #13 (P 17) Comment Response to Comment middle paragraph and Figure 10. It's very important to understand what the authors' proposed SHA does and does not do. For example, Figure 10 shows that at a given distance (30 km) from the New Madrid faults, the earthquake with an average recurrence rate of 0.004/yr, will produce ground motion whose median is 0.1g (and whose 16th percentile is about 0.04g and whose 84th percentile is about 0.22g). Ground motion contributions at 0.1g from other earthquakes with smaller or larger recurrence rates are not considered in this statement and have to be addressed in terms of earthquakes with other average recurrence rates. The statement on p. 17 that "Equation (17) describes a….hazard curve in terms of ground motion and its MRI [my emphasis] at a site." can be misleading. Thus, if one stated that the median ground motion associated with a recurrence rate of 0.004/yr was 0.1 g, it would be incorrect. A similar problem exists in the last paragraph on p. 17, although the last sentence is clearer. Both paragraphs should be reworded to make absolutely clear what SHA is and is not. The last paragraph on p. 17 also states that Figure 10 (SHA) is comparable to Figure 3 (flood hazard at Lock 4). How can this be so? I assume that the flood hazard curve is derived from annual peak discharge recorded at the same place. This includes all the uncertainty and is much simpler than having to derive magnitudes, recurrence information, and attenuation equations to determine what the seismic ground motion hazard at a given place (e.g., Paducah). Also the data base used for determining flood includes floods of different sizes and is not comparable to the SHA curve in which the peak ground motion is only associated with a given size earthquake. In SHA, temporal and spatial measures (including associated uncertainties) are considered separately. Ground motions from earthquakes with different recurrence rates should not be considered all together, particularly in the way of PSHA. This can be demonstrated from Figure 10. If there are only two characteristic earthquakes, M5.5 and M7.5, with 0.004/yr and 0.002/yr (Fig. 2) both at 30 km. At 0.22g, the confident level is 84% (16% PE) if M5.5 occurs, and 16% (84% PE) if M7.5 occurs. Here, ground motion with a confident level of 84% is compared with the one with a confident level of 16%. This comparison may not be statistically correct. Comparison for two statistic datasets should be based on the same level of confident. C-33 Equation (17) describes a hazard curve in terms of ground motion and its MRI [my emphasis] at a site has a clear physical meaning. The hazard curve directly converted from G-B curve (Eq. 15) and ground motion attenuation (Eq.16), i.e. converting the source measurement (magnitude) to the measurement (PGA) at a site at 30 km. Figure 10 (SHA) is comparable to Figure 3 (flood hazard at Lock 4) in terms of meanings, the way how the curves are constructed and used. In fact, PSHA was originally developed from analogy of flood, wind, and snow hazards (Cornell, 1968). The problem with PSHA is that there is a mathematical error (dependency of variable) in the formulation. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 16 Reviewer Leon Reiter Document / Section or Page # Response to Comment Comment Specific comment #14 (Fig. 10) Figure 10. What would the mean seismic hazard be? In the title to this figure the authors imply that the median is the same as the mean for the characteristic earthquake. This is not correct if the ground motion was derived from attenuation equations that assumed a log normal distribution. Can SHA calculate the mean hazard, which is used extensively for many regulatory purposes? This is a good point. Mean and median are different and need to be clarified. Figure 11 and other map-figures following. It would be very helpful if the authors showed the location of the Paducah facility on these maps. I think it only appears on Figure 31 and, possibly, as a yellow dot on Figure 20. last paragraph. How specifically does Figure 12 show that the northeast extension of the New Madrid faults has a significant effect on seismic hazard estimates at Paducah? How much closer to Paducah is the New Madrid fault if one assumes that there is a northeast extension? Will revise. 17 Leon Reiter Specific comment #15 (P. 20) 18 Leon Reiter Specific comment #16 (P. 20) C-34 A mean curve will be added to Fig. 10. The distance will be less than 10 km from the faults (in red) to the site. Our measurement from the faults of Johnston and Schweig (1996) to the site is about 45 km. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 Reviewer Document / Section or Page # Response to Comment Comment 19 Leon Reiter Specific comment #17 (P. 21) 1st paragraph. As stated above, the authors believe that the northeast extension is a significant issue. They have cited some evidence against its existence; however, this evidence should be laid out carefully and systematically. For example, the authors could show the location of the Jackson Purchase region with respect to the surrounding area (including the Paducah facility), the proposed extension of the New Madrid fault, the proposed northwest-trending structure, and discuss their significance. They could also show the plots of micro-seismicity (or modify the existing figures) that support the argument that the New Madrid faults don't extend into this region. A table comparing the aspects of earthquakes in the New Madrid zone, the northwest trending structure, and the Jackson Purchase/northeast extension, along with other seismological and geological evidence (as stated on p. 20) would be useful. One can then judge whether the evidence supports the claim. Do other hazard maps (e.g., Frankel 2002, Risk Engineering, 1999) make the same assumptions that the authors of this report do about the Jackson Purchase, the northwest-trending structure and the northeast extension of the New Madrid faults? If not, justify the choice. Good comment. Will revise. 20 Leon Reiter Specific comment #18 (P. 22) Figure 12. It is not clear what the blue lines represent and the basis for their definition. Do they represent faults as identified by the authors and Johnston and Schweig (1996)? Should be the same as the New Madrid faults shown in Figure 31? What are the blue boxes trending NNW supposed to represent? The blue lines represent New Madrid faults (SW, BL, NE, W, and thrust-box) and rift boundaries (ER and WR) by Johnston and Schweig (1996). The faults in Fig. 31 are the same as those of Johnston and Schweig (1996), except the thrust fault presenting by northern edge. 21 Leon Reiter Specific comment #19 (P. 24) 1st paragraph. The authors refer to Figure 10. Do they mean Figure 15? Yes, Fig. 15. C-35 Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 22 23 Reviewer Leon Reiter Leon Reiter Document / Section or Page # Specific comment #20 (P. 27) Specific comment #21 (P. 28) Response to Comment Comment top paragraph, Figure 16, and bottom paragraph. How old is "Iapetan?" What are the dotted circles in Figure 16? What is the significance of the J.T. Myers Locks and Dam shown on Figure 16? Can the Paducah facility be located on this Figure and Figure 15? (see comment 12 above). Do the authors mean to say "areal" rather than "aerial?" (see also "aerial" in paragraph 1 of p. 40) What is the rationale behind the authors' use of the Tri-State Seismic Source zone? How would the other alternative models affect the hazard calculations? I assume that a maximum magnitude of 6.8 was picked because it was midway between 6.2 and 7.3. Is this correct? The Figures 15 and 16 were taken from other reports. The references will be cited. It should be “areal”. The zone has been called by different names, such as the Wabash Valley. I prefer the Wabash Valley zone and will revise that. Different models (zone boundaries) surely affect the hazard calculations. A maximum magnitude of 6.8 was picked because it was midway between 6.2 and 7.3. 24 25 Leon Reiter Leon Reiter Specific comment #22 (P. 29) Specific comment #23 (Fig. 21) Discussion of background seismicity. The authors contend that large earthquakes (M=7.0 to 7.5?) in the background zone do not make any contribution to the hazard (citing Figure 20 taken from Petersen, 2005) and, they cause confusion. Figure 20 is not clear, but it looks like nearby (background?) magnitude 6 and 6.5 earthquakes (blue and green bars surrounding Paducah facility) are contributing to hazard. How is this consistent with the magnitude 4,7 to 5+ maximum background earthquakes shown in Figure 21? Also how do large background earthquakes "cause confusion?" Figure 21. The text states the Paducah facility is located in McCracken County shown in Figure 21. I cannot locate McCracken County on this map because the print is too small. C-36 As shown in Figs. 18 and 19, large earthquakes (M=7.0 to 7.5) in the background zone was used in the national mapping. The recurrence interval of the large earthquake is in 10,000 years or greater. In PSHA, these large earthquakes were distributed in large areas (Fig. 19) such that contributions from these large earthquakes to any site are negligible. This can be seen in Fig. 20. In other words, the large earthquakes were introduced, but have no effect on hazard calculation. Some people, even seismologists, have used Fig. 19 to generate ground motion hazard maps to show the general public and policy makers. This is clearly confusion. Fig. 20 was used to show that there is no contribution to the hazard from the large background earthquakes. Magnitude 6 and 6.5 earthquake shown in Fig. 20 was derived from the smoothed seismicity (Fig. 18) by Frankel and others (2002). The magnitude 4,7 to 5+ maximum background earthquakes shown in Figure 21 were derived from historical observations plus one standard deviation (~0.25 unit). A bigger map is needed to show county boundary. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 Reviewer Document / Section or Page # Response to Comment Comment 26 Leon Reiter Specific comment #24 (P. 32) 1st paragraph. The magnitude recurrence relationship for the Wabash Valley Seismic Zone is shown on Figures 24 and 25, not Figure 23 (as stated in the text). Correct. 27 Leon Reiter Specific comment #25 (P. 32) 2nd paragraph. Make it clear that Figure 23 itself does not come from Bakun and Hopper (2004), but rather it is based on data from that source. Also, do the authors assume that the1811-1812 events can be considered as a single, magnitude 7.5 earthquake? If so, how significant is this assumption? Will revise. Yes, we assumed that the 1811-1812 events can be considered as a single, magnitude 7.5 earthquake. In this report, seismic hazard is defined as an earthquake of magnitude M or greater (cumulative) or ground motion generated by the earthquake at a site vs. mean recurrence interval (or return period for ground motion). The cluster events are considered through the cumulative effect. It should be, but directly cited from Frankel and others (1996). 28 Leon Reiter Specific comment #26 Figure 22. Is the red curve a line drawn through individual seismicity data points? 29 Leon Reiter Specific comment #27 (P. 33) Table 5. What happened to event #6 in Bakun and Hopper (2004), the February 7, 1812, M=7.8 earthquake in the New Madrid Seismic Zone? Has this been left out of the authors' calculations? If so, justify this choice and estimate its impact. That was a mistake. Will be added. The hazard calculations will be the same. 30 Leon Reiter Specific comment #28 Figure 26. Overlay the data mentioned on p.34 that served as a basis for drawing the magnitude frequency relationship for the background seismicity. Will be added 31 Leon Reiter Specific comment #29 (P. 37) 1st paragraph. Contrary to what is stated, Table 6 contains five, not six attenuation relationships, the lowest value of which is 0.69g, not 0.46g. Also I am not clear what range of standard deviations the authors are assuming for the central U.S. Is it 0.6 to 0.8? Errors will be corrected. C-37 0.6 to 0.8 is the range of standard deviation for all attenuations in CUS. Exactly number used are based on each attenuation. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 Reviewer Document / Section or Page # Response to Comment Comment 32 Leon Reiter Specific comment #30 (P. 37) 2nd paragraph. I look for my colleagues Ken Campbell and Gail Atkinson to confirm the statement that "There is a consensus that many current attenuation relationships predict too high ground motion at near source, particularly Frankel and others attenuation relationship (USGS/NRC Workshop, 2005)." I contacted someone from the NRC who was at the workshop and the USGS organizer of the workshop and they do not remember this statement about a consensus. There is video CD for the workshop. 33 Leon Reiter Specific comment #31 Figure 27. I cannot see the symbol for the Frankel curve (referenced in the text on p. 38) on the figure. Is the high near-field curve from Atkinson and Boore (2006)? Frankel and others (1996) did not provide attenuation equations, but only a table with cut-off distance at 10 km. The comparisons were made at 10 km. 34 Leon Reiter Specific comment #32 (P. 39) 1st paragraph. Why did the authors choose these 4 attenuation relationships? Was Frankel and others relationship left out only because they felt that there was a consensus to support leaving it out, or were there other reasons? It was a “outlier” 35 Leon Reiter Specific comment #33 (P. 40) 1st paragraph. Re background seismicity, what is the justification of using a 15-km distance to the source? Also, the contributions from background seismicity shown in Figure 32 (e.g., PGA) look pretty high even though the maximum earthquake is only 5.0. On p. 29 (see also comment 22) the authors justify not using a higher magnitude cutoff by saying that higher magnitudes won't contribute much. Can they do a sensitivity test showing what the effects of having higher cutoffs would be? The focal depth is generally between 2 and 20 km in the region. We assumed a focal depth of 11 km and epicentral distance of 10 km. This results in a focal distance of 14.9 (round-up to 15) km. Figure 31. In comparing this to the blue lines in Figure 12, I am not sure why these particular New Madrid faults and lengths were chosen. Please explain. The New Madrid faults in Fig. 31 are the same in Fig. 12. See response to comment #18. 36 Leon Reiter Specific comment #34 C-38 Higher background earthquake will have (and should have) significant effect on hazard calculation. But the large background earthquakes have no effect because the way they were treated in a PSHA study. See response to comment #22 for further explanation. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 37 Reviewer Leon Reiter Document / Section or Page # Response to Comment Comment Specific comment #35 (P. 43) Tables 7, 8, and 9, when compared to Figures 27 to 30, and 32 show that what the authors did was equivalent to a deterministic scenario (M=7.5 at 45 km). The ground motion from other magnitudes and distances are not incorporated into the estimate; uncertainty at a given ground motion is shown assuming a fixed magnitude and distance. Is this what the authors wanted? If so, provide a rationale why this is acceptable? This was also discussed in comment 13. For a single characteristic source, SHA is equivalent to a deterministic scenario. See explanations to comments #8, 12, and 13. 38 Leon Reiter Specific comment #36 (P. 43) It would be highly useful if a table was made comparing these results with those of other studies that estimated seismic hazard at Paducah, e.g., Risk Engineering, Inc., (1999); Frankel and others, (2002); and any others that may exist. The authors of the report could then explain the differences between the results, the specific causes of these differences, and why their results are more valid. Although parts of this have been discussed in a general way in the text of the report, a specific discussion and evaluation of critical differences would be very helpful in evaluating this report and the novel way it approaches seismic hazard. A table comparison is not easy because hazard comparison is not only on ground motion value, but also on frequency (return period). For a single characteristic source, SHA derives a single frequency (return period), but PSHA derives a range of frequency. 39 Leon Reiter Specific comment #37 (P. 44-45) There are many important issues raised here. Comments 4, 8, 9,11, 12, 13, 14, 35, and 36 address these issues and the content of pp. 44-45 should be addressed in light of these comments. Similar concerns exist with respect to the executive summary. 1st paragraph. It should be made clear that although Reiter (1990) and Wang (2006) agree that seismic hazard and risk are different concepts they do not agree on what these concepts are. The same statement is made on p. 5, 1st paragraph. All these really come to a single question: is PSHA (Cornell-McGuire method) right? 40 Leon Reiter Specific comment #38 (P. 44) 41 Leon Reiter Specific comment #39 (P. 46) What is the basis for the authors' recommendation of using the average of the median and the median plus one standard deviation? Why not use, for example, the mean (not shown) or the one standard deviation estimate? C-39 It has been shown that PSHA is mathematically incorrect. This will be discussed thoroughly at the review meeting. See explanations to comment #4. There is confusion in terms of mean and median hazards. These will be addressed and discussed at the meeting. Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 Comment on “Seismic hazard assessment for Paducah gaseous diffusion plant” by Z. Wang and E. Woolery Mai Zhou Department of Statistics 849 Patterson Office Tower University of Kentucky Lexington, Kentucky 40506 E-mail: [email protected] Ground motion Y is generally modeled as a function of M and R with variability E in a regression model: ln(Y ) = g ( M , R) + E . (1) The variability E is modeled as a normal distribution with a zero mean and standard deviation σln,Y. In other words, the variability of ground motion Y is modeled as a log-normal distribution. Therefore, equation (1) can be rewritten as ln(Y ) = g ( M , R) + nσ ln,Y , (2) where n is a number of standard deviations measured as the difference relative to the median ground motion g(M,R). Modern PSHA is based on the following equation γ ( y ) = ∑ vP[Y ≥ y ] y = ∑ v ∫∫ {1 − ∫ 0 = ∑ ∫∫ [1 −Φ( 1 2π σ ln, y exp[− ln( y ) − g (m, r ) σ ln, y (ln(ξ ) − g (m, r )) 2 ]d (ξ )} f M (m) f R (r )dmdr 2 2σ ln, y (3) ) f M (m) f R (r )dmdr where ν is the activity rate, fM(m) and fR(r) are the probability density function (PDF) of earthquake magnitude M and site-to-source distance R, respectively, and g(m,r) and σln,y are the median and standard deviation at m and r, Φ(t) is the cumulative probability function for the standard normal random variable. Since the modeling consequences are so crucial, I would point out a few places in the PSHA calculation that I feel needs caution and a through review is perhaps needed. 1. Whether the error distribution is normal or not? Even it is normal, whether the variance of the error distribution remains a constant, as M and R changes? The systematic change of the variance, called variance structure, do not affect the estimation of the regression function g(m,r) too badly. But for exceedance probability, this variance structure is very important. 2. The estimation of σln,y, the standard deviation of E, is crucial, and is usually a harder task compared to the estimation of the regression function. If the regression function g(m,r) is not specified accurately, or if there is other systematic influence on the regression being ignored, then often the discrepancy in the regression functions are treated as error and regulated to E, thus inflating the σln,y. For example, site condition is not considered in the model. Also, if the distance R are measured with large error, the changes in ground motion due to these factors may be mixed with the intrinsic variability of E. C-40 Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated February 2007 April 2007 3. The probability density function (PDF) fM(m) and/or fR(r). The form and accuracy of these two densities affects the exceedance probability a great deal. How confident we are when we plug-in a PDF for fR(r)? The assumption of normal distribution for the error E is usually granted when a regression model is assumed. This is not critical when the purpose of the model is mainly for estimating the regression function g(M,R). Since the least squares method used in the estimation of regression function is also consistent when the error follows other type of distributions, or the variance is not constant. But we are using the model to calculate the exceedance probability, which involves the tail behavior of the error term. The assumption of normality, and the assumption of constant variance is critical. Even if the normal assumption is reasonable, its variance may depend on M, R. Only when M, R, and E are independent random variables, the joint probability density function of M, R, and E can be written as a product f M , R , E (m, r , ε ) = f M (m) f R (r ) f E (ε ) , (4) where fE(ε) is the PDF of E. The exceedance probability P[Y≥y] is P[Y ≥ y ] = ∫∫∫ f M , R , E (m, r , ε ) H [ g (m, r ) + ε − ln( y )]dmdrdε = ∫∫∫ f M (m) f R (r ) f E (ε ) H [ g (m, r ) + ε − ln( y )]dmdrdε , (5) where H[g(m,r) + ε –ln(y)] is the Heaviside step function, which is zero if g(m,r)+ε is less than ln(y), and 1 otherwise. Because E follows a normal distribution, equation (5) can be rewritten as P[Y ≥ y ] = ∫∫{∫ f E (ε ) H [ g (m, r ) + ε − ln( y )]dε } f M (m) f R (r )dmdr = ∫∫{1 − ∫ ln( y ) − g ( m , r ) = ∫∫{1 − ∫ ln( y ) 2π σ ln, y −∞ −∞ y = ∫∫ {1 − ∫ 0 1 1 2π σ ln, y 1 2π σ ln, y exp(− exp[ − ε2 )dε } f M (m) f R (r )dmdr 2 2σ ln, y (ε − g (m, r )) 2 ]dε } f M (m) f R (r )dmdr 2 2σ ln, y . (6) (ln(ξ ) − g (m, r )) 2 exp[− ]d (ln(ξ ))} f M (m) f R (r )dmdr 2 2σ ln, y Therefore, we have equation (3), the heart of modern PSHA. As demonstrated above, equation (3) is derived from the pre-condition that only if M, R, and E are independent random variables. However, if the ground-motion variability E depends on M and R, then hazard calculation, equation (3), in PSHA is not correct. C-41 Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated May 2007 June 2007 1 2 Reviewer Document / Section or Page # Response to Comment Comment Roy VanArsdale An appendix illustrating your calculations for both PSHA and DSHA. Roy VanArsdale DSHA calculation is shown in table 8 though 11. A brief discussion in your conclusion section point out This has been added. the differences between your values and the USGS values. D-1 PSHA calculation is straight forward, but very time consuming (long). We decided not to include it. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated May 2007 June 2007 Reviewer Document / Section or Page # Response to Comment Comment 1 Gail Atkinson General The subject report deals with seismic hazards to the Paducah gaseous Responses are only provided to the highlighted (yellow) ones. diffusion plant. This review deals with the Revised Version, entitled: “Final Others have been revised accordingly. Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, dated May 11, 2007. The report is clearly written and easy to follow. Technically, it is much improved over an initial draft (March 2007) that was reviewed by a review team and discussed at a team meeting in Lexington KY on April 30. The methods and conclusions of the report are now for the most part well reasoned, with a few significant exceptions that need to be remedied to make the report technically sound and defensible overall. I have listed my comments below by page and fraction (eg. 2.5 indicates the middle of page 5). All suggested changes are straightforward to implement. With the highlighted comments addressed as suggested, the report will then form a good assessment of the seismic hazard at Paducah. 2 Gail Atkinson 20.2 The use of Mx=6.8 in Wabash is inconsistent with the estimated range of M6.2 to 7.3 for paleoseismic events. The Mx for WVSZ should be at least 7.3, and possibly 7.5. See also Figure 18, which also shows higher magnitudes for paleo events. 3 Gail Atkinson 21.8 The treatment of the background source is not satisfactory. You cannot justify a low Mx (in the M5 range) anywhere in the world. Most global studies suggest Mx~7 for stable craton regions (eg. Johnston et al., 1996). You also cannot fix an arbitrary distance. This highlights one of the weaknesses of DSHA; it cannot handle background seismicity. I suggest that for the DSHA you just state that the DSHA focuses on the perceived dominant hazard source, the New Madrid Seismic Zone, and ignores other potential contributions such as the local seismicity and WVSZ, which are handled in the PSHA. D-2 We used mean values (best estimate) for any set of parameters throughout this report. Fig. 18 was the old estimate and used by the USGS (Frankel and others, 2002). As discussed in Wang (2003a), there is no contribution from those large background earthquakes because (1) a large-area source zone and (2) a longer recurrence interval (more than 10,000 years). Use of the large background earthquake only causes confusion. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated May 2007 June 2007 Reviewer Document / Section or Page # Response to Comment Comment 4 Gail Atkinson 25.9 Is Figure 20 the definition of the background zone? Show the spatial definition of this zone explicitly. 5 Gail Atkinson Figures 26-29 State the type of distance used in the plots; this is especially important as you made a big point of the types of distances and their impacts on these plots earlier in the report. 6 Gail Atkinson 37.3 The sentence, and corresponding approach “We used a point source at 15 km with a maximum magnitude of M5.0 to account for the local earthquake” is not justified. A proper areal source zone with the magnitude recurrence relation as defined from Fig. 21 should be defined and included in the PSHA, with a suitable Mx (6.5 to 7. based on global precedents). It is fine to exclude the local source from the DSHA, as long as it is properly included in the PSHA. The USGS also used the point source (grid point) to account for the seismicity (Frankel and others, 1996, 2002). 7 Gail Atkinson Figure 30 Show exactly how the local and WVSZ areal sources are defined for the PSHA. The local zone (background) is a point source at 15 km. WVSZ is an areal source shown in Fig. 30. 8 Gail Atkinson 38.1 The most important uncertainties for a logic tree in this case are the GMPEs and the source geometry. You have ignored uncertainty in the spatial definition of the source zones. This uncertainty should ideally be considered, or as a minimum you should state explicitly that you are ignoring uncertainty in the definition of the source zones. It is OK to use a singe Mx value, as long as it is sufficiently large to be above the range of interest/sensitivity to this parameter. Properly chosen, hazard results are not very sensitive to Mx. The local seismicity is not properly treated here, as noted above, and needs to be properly included in the analysis. It has been shown that a properly chosen Mx and distance can be used to quantify hazard at a site in NMSZ (Frankel, 2004; Petersen, 2005). D-3 The background seismicity was treated as a point source, which is similar to the smoothed grid seismicity in the USGS maps. Fig. 20 shows the earthquakes that was used to derive a and b values. Rrup was used throughout this report. The background seismicity was treated in a similar way to the USGS mapping (Frankel and others, 1996, 2002). Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated May 2007 June 2007 9 10 Reviewer Gail Atkinson Gail Atkinson Document / Section or Page # Table 6 44.8 Response to Comment Comment What weights were used for the GMPEs? Does Table 7 and Figs 31-33 refer to the mean-hazard PSHA results? Sensitivity to the alternative models should be shown. The presentation of the PSHA results is incomplete. Delete the entire paragraph under Table 15. You consider only probabilities to 1/2500 in the report, then appear to state at the very end that your target probability is much lower. There is no suggestion in the report that probabilities of 1/100,000,000 are of interest, and thus none of this discussion is relevant. It just detracts from the report, which should simply end after Table 15. D-4 Equal weight (0.25) was assigned to four GMPEs. Table 7 and Figs 31-33 refer to the mean hazard. No sensitivity to alternative models was carried out in this study. Deleted. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated May 2007 June 2007 Reviewer 1 Jim Beavers 2 Jim Beavers Document / Section or Page # Response to Comment Comment Zhenming, per our conversation today with regard to your PSHA PGA number (0.49g)on hard rock (USGS Type A foundation) at 2500 years we talked about three things that brought the number down from the 0.8g PGA I had calculated from the USGS (1996) B-C boundary of 1.2g and the corresponding 0.8g Risk engineering had calculated. These three things were These all make sense to me, as a result, the 0.49g seems realistic to me knowing these three items changed. To convince others that 0.49g is the right number for this study I would run a sensitivity analysis. For example, run your PSHA with just using Frankel's attenuation and see how much it raises the 0.49g. Then increase the magnitude to 8.0 and see how much further it raises it. Finally change the distance to what Art used. By then you should be closer 0.8g. This will give you a feel for what is contributing to the reduction. The only other variable that may cause the 0.49g to go up is the lower return period 500 verses 1000, but you used that anyway. In McGuire's and Frankel's 0.8g was an M 8.0 and R of 1000. Make a few comments about your sensitivity study in Section 6.1 about these contributions. This will help you down the road in case other external reviewers are brought in at the PGDP, which is highly likely to occur for the upcoming DOE CERCLA Waste Disposal Facility. Also for your top of soil numbers I would just go to the Bechtel-Jacobs 2002 report (BJC/PAD-356) and scale the soil amplification numbers from figures 7.3-1a for PGA, 7.3-1b for 0.1 sec. and 7.3-1c for 1 sec. We did not do a 0.2 sec curve. The CERCLA will have longer period motions, probably around 1 sec. It looks like your 0.49g will lower the long period motions. I took a quick look at the Bechtel-Jacobs report and with a hard rock PGA of 0.49g from figure 7.3-1a I get an amplification factor for PGA at top of soil of 0.8, from figure 7.3-1b I get an amplification factor for 0.1sec at top of soil of 1.2, and from figure 7.3-1c I get an amplification factor at 1 sec at top of soil of about 2.0. You will see in Table 8-1 we ended up with a preferred method that had amplification factors respectively of 0.73, 0.68 and 2.55. You have a little more amplification at PGA and at the 0.1 sec because of the PGA being 0.49g, But when you get out to the 1 sec period we had a 25% higher amplification because our hard rock PGA was 0.8g or 0.71g after refinement of my earlier calculations in the Bechtel-Jacobs report. D-5 The three things are: 1) the location of the New Madrid faults (further west), 2) a smaller mean magnitude (M7.5 vs. M7.7) for the characteristic earthquake in the New Madrid Seismic Zone, and 3) use of lower ground motion attenuation relationships. Soil amplification is not part of this project. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated May 2007 June 2007 Reviewer Document / Section or Page # 1 Ken Campbell General #1 2 Ken Campbell General #2 Response to Comment Comment It is not clear what role the Independent Expert Review Panel had in the study. It is very important that the roles of these reviewers be described together with such information as: (1) when and where the review meeting was held and how long the meeting lasted, (2) the amount of time that each reviewer was given to perform the review, (3) the materials provided to the reviewers for review, and (4) the recommendations that were made at the review meeting by each of the reviewers. It is also important that reasons be given why some of the recommendations of the Review Panel, both written and verbal, were not adopted in revising the report. The so-called PSHA conducted in this report is not a standard PSHA such as is done in practice. The PSHA presented in the report only takes into account the characteristic earthquake on the New Madrid Seismic Zone (NMSZ) and the maximum magnitude earthquakes on the Wabash Valley Seismic Zone (WBSZ) and the Local Source Zone (LSZ) located at specific distances to the PGDP site. This will not necessarily represent events that contribute the greatest to the probabilistic ground motion for a given probability of exceedance, because of trade-offs between the recurrence interval of the events and their magnitudes and distances. On the other hand, a true PSHA would also allow the noncharacteristic earthquakes to float within their area sources, thus allowing many events to occur farther from the PGDP site than was assumed. Of course, there would be some floating earthquakes within the LSZ that would also occur closer to the PGDP site. For a full standard PSHA, the complete recurrence curves (magnitude-frequency distributions) and distance distributions for every source should be used. Also, the epistemic uncertainty characterized by the use of multiple attenuation relationships should be included as part of the epistemic uncertainty model. D-6 Revisions have been done to address these. And other materials were also included as appendix. The probabilistic analysis carried out in this project is not a standard PSHA. As shown by Frankel (2004) and Petersen (2005), a simpler one, like the one carried out in this project, can provide a good estimate. This serves the purposes of this project: 1) to gain better understanding on the seismic hazard assessment at the Paducah Gaseous Diffusion Plant and its surrounding area, and 2) to communicate the hazard information more effectively to the users and policy makers. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated May 2007 June 2007 Reviewer Document / Section or Page # Response to Comment Comment 3 Ken Campbell General #3 It was unanimous amongst the Review Panel members that, not only should a full PSHA be done, but that the PSHA should account for epistemic uncertainty in such parameters as the characteristic and maximum magnitudes and the distances from the site to the seismic sources (in this case, the New Madrid Fault Zone and the boundaries of the WVSZ and LSZ). No such uncertainty was included in the revised report. In lieu of formally accounting for epistemic uncertainty, a series of sensitivity analyses could be used to show the sensitivity of the results to the modeling assumptions that were made. The recommendation is to perform a PSHA with some discussions for improvements. This report reflected that. More analyses, including sensitivity analysis, could be done, but there is a time constrain. 4 Ken Campbell General #4 In some cases, there is no such documentation to support a decision to use one parameter over the other. This is particularly true in CEUS. We had tried our best in this report. 5 Ken Campbell General #5 There is a general lack of documentation regarding why certain decisions were made, such as why the specific attenuation relations used in the analysis were selected and why others were excluded and why certain investigators characterizations of seismic sources were used and others were not. Without such documentation, the reader gets the impression that the selection was arbitrary and designed to achieve a certain result, even if that was not the case. Since the USGS National Seismic Hazard Mapping Project (NSHMP) will generally be considered the basis for comparison, any deviation from that Project’s hazard model should be clearly described and explained. Although the revised report has been improved considerably from the original version, there is still a perceived undercurrent of bias against PSHA that gives an impression of unprofessionalism. It is certainly appropriate to point out the weaknesses of PSHA, but they should be balanced by also discussing its strengths. DHSA also has weaknesses and strengths, but comments throughout the report tend to emphasize its strengths while emphasizing the weaknesses in PSHA. D-7 Text has revised to address the weaknesses of DSHA. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated May 2007 June 2007 Reviewer Document / Section or Page # 6 Ken Campbell Page 1. 7 Ken Campbell Page 1. 8 Ken Campbell 9 Ken Campbell Response to Comment Comment It appears that the USGS hazard maps, specifically with respect to their use in design, are being misrepresented. The ground motion values from the maps are not used directly to derive design ground motion in the NEHRP and IBC design codes. Aside from the issue of deterministic caps in the design maps, the ground motion from the hazard maps are multiplied by the site factor representing the NEHRP site class for the site of interest and this value is in turn multiplied by 2/3. For a hard rock site in the CEUS (NEHRP site class A), the site factor is 0.8 for all ground motion parameters. Therefore, the mapped value of ground motion would be multiplied by 0.8 x 2/3 = 0.53 to derive the design value, nearly a 50% reduction in ground motion. Continually referencing the mapped values is confusing and gives the impression that these mapped values are used for design. The statement that “These high design ground motions for western Kentucky are not consistent with scientific research and observations” is not justified and, in my opinion, should be deleted. Probabilistic ground motions approaching or exceeding, say, those in San Francisco, can possibly be justified given the relatively short recurrence interval of large New Madrid earthquakes (i.e., 500 years), the factor of two increase in short-period ground motion for the same magnitude and distance in the CEUS, and the lower rate of attenuation in the CEUS. The design values (0.6g and 0.8g) were reduced by a 1.5 factor. Page 8. Deaggregation methods were developed to overcome the disadvantage in the PSHA methodology that was identified by NRC (1988) and has now been accepted by practitioners and regulators alike as a valid means of developing one or more design earthquakes from PSHA results. De-aggregation is an effort in PSHA to try seeking the “design earthquake.” (revised) Page 8. It is important to mention that the second disadvantage of PSHA of obtaining excessively large ground motion values at very low probabilities of exceedance is not an issue when the results are constrained to reasonable probability levels (e.g., ≥ 2% probability of exceedance in 50 years). Even Figure 4 shows that the contribution of uncertainty caps out at 2-3 standard deviations for probabilities constrained to such levels. The ground-motion uncertainty is an integral part of PSHA, a cap on it may not statistically sound. D-8 This has been revised. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated May 2007 June 2007 Reviewer Document / Section or Page # Response to Comment Comment 10 Ken Campbell Page 10. There seems to be a clear bias against PSHA, since only its disadvantages are listed, whereas only advantages are listed for DSHA. See General Comment 5 for additional discussion of this topic. In fact, since both methods have strengths and weaknesses, there is clear justification for using both methods. Revised. 11 Ken Campbell Page 11. 12 Ken Campbell Page 13. References for the possible causes of seismicity in the NMSZ are quite old. Several new theories have been put forth since these references were written that should also be presented. The few small events that have been recorded in the Jackson Purchase region are not sufficient to justify the strong conclusion that “there is no evidence (microseismicity) to support the northeast extensions of the New Madrid faults into Jackson Purchase region.” Many more recordings would be required to justify such a conclusion. Even if true, the fault could be located just outside of the Jackson purchase region, or it could be locked and not generating earthquakes at even the microearthquake level. There are some new references, particularly from GPS. However, those could cause confusion. Those records are surely not sufficient, but at least they are real data. 13 Ken Campbell Page 13. It would be useful to show a map of the New Madrid faults that were used to define the New Madrid characteristic earthquakes in relation to the PGDP site. It is shown in Figs. 7 and 30. 14 Ken Campbell Page 20. 15 Ken Campbell Page 21. It is not clear why the so-called Tri-State Seismic Source Zone rather than other alternative source zone configurations of Wheeler and Cramer (2002) were used to represent the WVSZ. These alternative source zones would have made a valid epistemic uncertainty model. The characterization of the LSZ in terms of magnitude, distance, and focal depth distributions seems arbitrary and needs to be justified. For example, as discussed in the review meeting, Mmax (Mw) = 5.0 is too low to be a reasonable estimate of the largest earthquake that can be expected to occur in the background region surrounding the PGDP site. Based on a worldwide study, EPRI proposed that Mw = 6.3 ± 0.2 represented a reasonable estimate of maximum magnitude in non-rifted SCR crust. Alternatively, one could look at a much large region of the CEUS (and possibly eastern Canada) with tectonic conditions similar to the region around the PGDP site to come up with a more reasonable estimate of Mmax. Different names have been used for the zone in the literature. WVSZ was used throughout this report. A M8.0 or even larger earthquake can be put at the site. But, it is meaningless for hazard assessment, particularly for PSHA, if the associated recurrence interval is unknown. Determination of these earthquakes should be consistent with historical and geological data. D-9 Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated May 2007 June 2007 16 Reviewer Ken Campbell Document / Section or Page # Page 23. 17 Ken Campbell Page 28-29. 18 Ken Campbell Page 28-29. 19 Ken Campbell Page 31. Response to Comment Comment The only comprehensive study of recurrence intervals on the New Madrid fault is the paleoliquefaction studies reported by Tuttle and her co-workers. She shows evidence of at least three past sequences of large liquefaction events rivaling that in 1811- 1812 that suggests a mean recurrence interval of 500 years for such large events. The 1,000-year recurrence interval used previously by the USGS and others would appear to be longer justified. The UK statistician, Mai Zhou, who was a member of the Independent Expert Review Panel, indicated to me during the review meeting that he did not see any problem with framing the PSHA integral the way that it is, even if the standard deviation of ground motion is a function of magnitude and/or distance, as long as this function of magnitude and/or distance was included in the analysis. So any statement to the contrary should be deleted. There is no reference to studies (e.g., the recent NGA studies; Boore et al., 1997) that have concluded that the standard deviation of ground motion is not a significant function of magnitude. These newer studies should be reviewed and could possibly be used to justify a revision of the aleatory uncertainty model currently used to characterize ground motions in the CEUS. The range of median PGA values from Table 5 is 0.69-1.20, not 0.46-1.20. D-10 Here is one reference published recently: Holbrook, J., Autin, W.J., Rittenour, T.M., Marshak, S., and Goble, R.J., 2006, Stratigraphic evidence for millennialscale temporal clustering of earthquakes on a continental-interior fault: Holocene Mississippi River floodplain deposits, New Madrid Seismic Zone, USA: Tectonophysics, v. 420, p. 431–454. There are some GPS studies available, but were not used in this report. Mark Zoback also suggested 1,000 year recurrence interval at a recent EarthScope workshop. See his review comments on the preliminary report. As the way it is being modeled (finite source and global data), ground-motion uncertainty is a dependence of magnitude and distance. revised Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated May 2007 June 2007 Reviewer Document / Section or Page # 20 Ken Campbell Page 34. 21 Ken Campbell Page 38. 22 Ken Campbell Page 38. 23 Ken Campbell Page 38. Response to Comment Comment The plot of the attenuation relationships in Figures 26-29 could be deceiving. For example, the plotted relationships do not all use the same distance measure and do not represent the same site conditions. If these differences were not taken into account, then the figure is incorrect and so, too, might be the estimates of ground motion from these relationships. If these differences were corrected for, then how were the corrections done? The relationship by Frankel et al. (1996) is not that different from many of the other relationships in the distance range 10-100 km, so I don’t understand the statement to the contrary. Furthermore, the Frankel et al. relationship represents NEHRP B site conditions and, using the USGS conversions factors, should be divided by 1.53 to represent the hard rock site conditions for which estimates are sought. As mentioned in General Comment 2, Table 6 does not represent a true PSHA, since it does not include: (1) epistemic uncertainty in Mchar and Mmax, (2) epistemic uncertainty in the location of faults and the boundaries of source zones, (3) aleatory uncertainty in the characteristic magnitude of the New Madrid fault or in the exponentially distributed magnitudes of the source zones, (4) aleatory uncertainty in the locations of earthquakes distributed within the source zones, and (5) epistemic uncertainty in recurrence parameters. It is really a pseudo-deterministic model, where the only uncertainty is the aleatory uncertainty in the estimation of ground motion. Why were the specific attenuation relationships selected for use in the study? For example, why was the Silva DC-S model chosen over the other three that he has developed and used to characterize epistemic uncertainty? Was the hard rock or NEHRP BC version of the Atkinson and Boore (2006) attenuation relationship used? Were differences in distance measures between the various relationships taken into account? Were differences in site classes between the various relationships taken into account? All attenuations are for hard rock site. The distance is Rrup. No distance conversion was done. Frankel’s ground motion values was corrected by the factor 1.53. I don’t see the justification for giving the 1,000-year recurrence interval on the New Madrid fault 25% weight. As mentioned before, this estimate is no longer considered to be valid and is contradicted by the latest paleoliquefaction studies. See response to comment #16. D-11 See response to the general comment #2. All attenuations are for hard rock. The Silva DC-S model provides a reasonable value. Others represent different models, i.e., composite, double-corner, and hybrid. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated May 2007 June 2007 24 Reviewer Ken Campbell Document / Section or Page # Page 43. Response to Comment Comment Using an estimated value of PGA from an estimated value of MMI at the PGDP site for the 7 February 1812 earthquake using the simple relationship between PGA and MMI given by Bolt (1993) should not be used as justification for selecting a return period of 1,000 years for determining design ground motions for the PGDP site. New relationships between PGA and MMI, some developed specifically for the CEUS, have been published and should also be reviewed and cited. Selecting an exceedance probability (or return period) should be based on other factors as well, such as whether the risk is acceptable for the particular facility and site and whether it conforms to relevant public policy guidelines. D-12 A new reference (Atkinson and Kaka, 2006) was added. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated May 2007 June 2007 1 Reviewer Leon Reiter Document / Section or Page # General Response to Comment Comment At your request, I have reviewed the revised report on the Paducah facility by Zhenming Wang and Edward W. Woolery, and my comments follow. Similar to my review of the February 2007, draft report, I have employed the same general approach I found useful in reviewing many nuclear facilities and in the peer review of seismic hazard analyses submitted to professional journals for publication.1 This general approach emphasizes clarity and technical justification for the methods used and the assumptions made. In general, the revised report represents an improvement over the draft report in that the controversial definitions of seismic hazard and risk and the use of a new methodology (SHA) have been omitted. Most of the comments in my review of the draft report are no longer relevant or have been addressed. However my specific comments #2, 17, 20, 21, 23, 32, 33, and 36 have only been addressed partially, if at all, and they are relevant to my review of the revised report. This report is not a typical site-specific seismic hazard assessment, but a summary of scientific research on geological and seismological conditions, the methodologies, and the seismic hazard assessment related to the Paducah Gaseous Diffusion Plant and the surrounding area. Therefore, it may be reviewed in a different way than a normal site-specific technical report. The specific comments #2, 17, 20, 21, 23, 32, 33, and 36 for the early version have also been addressed in some degree. The primary difference between the draft and revised report is the addition of a PSHA and a DSHA for the Paducah facility and the introduction of a two level design basis. My comments on the new material in section 6 (Results) follow along with some new specific comments on the rest of the report. 1 Please note that my review represents my own views and not necessarily those of my past employers, the Nuclear Regulatory Commission and the U.S. Nuclear Waste Technical Review Board. D-13 Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated May 2007 June 2007 2 Reviewer Leon Reiter Document / Section or Page # Comments on section 6 (Results) 1. Response to Comment Comment It was not clear to me what are all the assumptions and input parameters behind the PSHA. Based on a May 25, 2007, e-mail exchange and subsequent telephone conversation with Dr. Wang, I draw the following conclusions. The New Madrid (NMSZ), Wabash Valley (WVSZ), and local source zones, were the only ones considered in the analysis. Only one magnitude (Mmax) for each source zone was used. Only one distance for each earthquake was used for each of the New Madrid and local source zones, while the earthquakes in the WVSZ were allowed to occur anywhere within that zone. The NMSZ allowed two different recurrence intervals for the controlling earthquake, while the WVSZ and the local source zone allowed only one recurrence interval for each of the controlling earthquakes in each source zone. Four, and in one case three, different equally weighted ground motion relationships were used assuming the standard deviation determined by the originators of the relationships. Therefore, no uncertainty was assumed in the magnitude of controlling earthquakes, the location of these earthquakes in the NMSZ and local source zone, the recurrence intervals for the controlling earthquakes in the WVSZ and local source zone. Also the effects of earthquakes smaller than Mmax in each source zone were not taken into account. A typical PSHA would address these uncertainties. Although some of these omissions may, as Dr. Wang maintains, have little or no effect upon the results, this remains to be shown. Assumptions about the local source zone may have a larger than assumed effect, particularly for PGA. Other assumptions that need further proof include the lack of presence of the northeast extension of the NMSZ and the choice of the four attenuation relationships. It would be very useful to those assessing the PSHA to have a better understanding of the bases for these assumptions and their importance. Sensitivity tests to different assumptions would be very helpful. Jim Beavers in his May 25, 2007, e-mail to Dr. Wang made a similar suggestion. Justification of some of the assumptions in the revised report by referral to the USGS studies, is not necessarily a valid approach because a seismic hazard analysis for an individual nuclear facility site may require a higher level of justification than local seismic hazard extracted from a generalized nationwide study. D-14 PSHA and DSHA in this report are not a site-specific. The main purposes are to gain better understanding on the seismic hazard assessment at the Paducah Gaseous Diffusion Plant and its surrounding area. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated May 2007 June 2007 Reviewer Document / Section or Page # The introduction of a two level design basis represents a positive step. The choice of a 1,000year return period for ordinary structures seems to have a good basis. This is not as true for the use of the DSHA for important structures. The rationale behind the choice of the median plus one standard deviation and its correlation with the PSHA is important, and needs to be laid out. However choice of design levels is not a seismological decision because it implies a certain level of risk acceptance, which is a social decision. Seismology is most useful when it provides the analysis that allows social decision-makers to make informed decisions. There is some confusion between the use of the terms "mean" and "median." Based upon my understanding of the revised report the PSHA results is a mean because it represents the average of the weights applied. (Theoretically it is still a mean even if the uncertainties are underrepresented). In the DSHA, the number used is the average of the medians, and, as far as I know, not what analysts intend when they use terms like "best estimate" or "mean." I suggest that the report identify this, as it does in some, but not all tables (e.g., Tables 15 and E-3) as the average of the medians or the medians plus one standard deviation. It is true that “Choice of design levels is not a seismological decision because it implies a certain level of risk acceptance, which is a social decision.” But seismologists need to provide hazard information that can be understood. This is our main effort. Specific comment #1 (P 2, Fig.1) Identify the location of the centers (0 km, 0 km) of the seismicity plots. The map is a schematic and cited from Stain and others (2003). No reference point was given. Specific comment #2 (P 4) third This paragraph implies that the SSE and the OBE for nuclear power plants are only determined through DSHA. This is not true. The OBE was always defined (10CFR Part 100, Appendix A) as "…that earthquake which could reasonably affect the plant site during the operation life of the plant;…" 10CFR Part 100.23 states that "…uncertainties in defining the SSE must be addressed through an appropriate analysis such as PSHA or suitable sensitivity analyses." USNRC Regulatory Guide 1.165 describes how PSHA can be used to determine the SSE. Although these terms originally have a clear meaning, they are confused. All those could cause confusion have been deleted. Why is this figure located here? As far as I can tell it is only referred to on p. 28. It is described on page 5. 3 Leon Reiter Comments on section 6 (Results) 2. 4 Leon Reiter Comments on section 6 (Results) 3. 5 Leon Reiter 6 Leon Reiter paragraph 7 Leon Reiter Response to Comment Comment Specific comment #3 (P. 6, Figure 2) D-15 Median is only applied to each ground motion attenuation relationship. Mean is for all others. Comment Number Review Comments and Reponses Report on Seismic Hazard Assessment for Paducah Gaseous Diffusion Plant, Dated May 2007 June 2007 Reviewer Document / Section or Page # Response to Comment Comment 8 Leon Reiter Specific comment #4 (P. 8, First paragraph) The report's concern about the lack of a design earthquake fails to mention that McGuire (1995) not only mentions this concern but also proposes a methodology (deaggregation) to address concern. Why isn't this discussed? De-aggregation is an effort in PSHA to try seeking the “design earthquake.” (revised) 9 Leon Reiter Specific comment #5 Reiter (2004) does not appear in the list of references. It is an abstract and deleted from the references The report introduces two terms for essentially the same phenomenon (randomly occurring nearby earthquakes); "Background Seismicity" and "Local Source Zone." It would be helpful if you made clearer the distinction and your use of these terms. These have been revised to use the background seismicity only. (P. 28-29. ) What is the point of the discussion of the different source to site distance measures in the revised report? Is anyone suggesting the use of epicentral distance in the attenuation relationships? This discussion may be a leftover from the key arguments in the draft report about whether or not distance and magnitude are independent random variables. This is really not an important issue in the revised report. There is a difference between epicentral and fault or other distances. This may be one of the areas that PSHA needs to improve. Specific comment #8 If the report does include this figure (see discussion above), the title should mention and explain the use of REPI and RRUP in the figure. Rrup is used throughout this report (revised) The final report states that ground motion at near-source has been over-predicted and references a USGS/NRC workshop in 2005 and Atkinson and Boore (2006). The USGS/NRC workshop does not appear in the list of references and Figure 26 shows that at distances less than 10 km the AB06 ground motion relationship predicts higher ground motion than the other models used in the PSHA. The term "near source" needs to be clarified to justify the report's conclusion. The basis for picking the four attenuation relationship and excluding others (e.g., Frankel) needs to presented. A CD on the workshop is available. Frankel and others (1996) only gave ground motion values from 10 km and greater. (P. 8, Second Paragraph) 10 Leon Reiter Specific comment #6 (P. 20, last paragraph) 11 Leon Reiter 12 Leon Reiter 13 Leon Reiter Specific comment #7 (P. 30, Figure 23) 14 Leon Reiter Specific comment #9 (P. 34, First paragraph) Specific comment #10 (P. 34, Second paragraph) D-16 Near-source means in this report 10-50 km. These attenuation relationships represent different approaches (i.e., finite source/green function, double-corner, and hybrid methods).