EARTHQUAKE HAZARD ANALYSIS: MUSKRAT DAMSITE, LOWER CHURCHILL, LABRADOR. Draft Report: May 22, 2014

Transcription

1 EARTHQUAKE HAZARD ANALYSIS: MUSKRAT DAMSITE, LOWER CHURCHILL, LABRADOR Draft Report: May 22, 2014 By: Gail M. Atkinson, Ph.D. Engineering Seismologist For: SNC Lavalin Ltd. 1

2 TABLE OF CONTENTS Executive Summary 1 Introduction Seismic Hazard Analysis Method Overview Treatment of Uncertainty 6 2. Input Parameters for Seismic Hazard Analysis Seismic source models Magnitude recurrence relations Ground motion prediction equations 12 - Results of Seismic Hazard Analysis 14.1 Mean-hazard UHS and Sensitivity 14.2 Results for NEHRP A site conditions (hard-rock) and for soil sites 18. Deaggregation 21 4 References 2 5 List of Abbreviations 25 Appendix A, EqHaz1 Inputs 26 A.1 H2 model 26 A.2 HY (hybrid) model 45 A. R2 model 69 Appendix B Deaggregation information for 1/10,000 p.a. by magnitudedistance bin 80 2

3 SEISMIC HAZARD ANALYSIS: MUSKRAT DAMSITE, LOWER CHURCHILL, LABRADOR Executive Summary A seismic hazard assessment was performed for the Muskrat Falls site in the Lower Churchill region of Labrador (approximate location 5.25N 60.77W). The analysis determines the expected earthquake ground motions over a range of probability levels, including 1/1000, 1/2475, 1/5000 and 1/10,000. The ground motions are calculated for stiff soil to rock site conditions, which typically have shear-wave velocities ~ 760 m/s near the surface. This site condition corresponds to NEHRP B/C boundary in the standard NEHRP (National Earthquake Hazard Reduction Program) site classification scheme. In the NEHRP classification, NEHRP A sites are those with average shear-wave velocity in the top 0 m >1500 m/s; NEHRP B and C correspond to average shear-wave velocities in the top 0 m of m/s and m/s, respectively, while NEHRP D (stiff soil) corresponds to average shear-wave velocities of m/s. Soils amplify ground motions, with softer conditions producing greater amplifications, especially at longer periods. Factors are provided that give average amplifications that would be expected for different soil types, relative to the results for the reference B/C condition. The results of the probabilistic analyses can be summarized in simplified terms as follows. At the probability level of 1/10,000, the expected peak ground acceleration (PGA, from natural earthquakes) for the reference site condition (NEHRP B/C) at Muskrat is approximately 0.06g. The ground motions at this probability level (1/10,000) correspond approximately to local earthquakes of M5.0 to 6.5 at distances from 0 to 100 km, for frequencies > 2 Hz. At long periods (f<2 Hz), motions correspond to those that would be expected for a major regional earthquake (M6.5>) at distances up to 00 km away, in either the offshore or St. Lawrence seismic zones.

4 1 - INTRODUCTION This report presents a seismic hazard assessment for the Muskrat Falls site in the Lower Churchill region of Labrador, for annual exceedence probabilities in the range from 1/1000 to 1/10,000. The analysis determines the likelihood of ground motion at the site by considering the magnitudes, rates of occurrence, and locations of earthquakes throughout the region, using the probabilistic Cornell-McGuire method. The method is widely used throughout North America and forms the basis for seismic zoning maps in building codes in Canada (Adams and Halchuk, 200; Adams and Halchuk, 201). This assessment represents an update and refinement of the type of estimate provided in the 2010 National Seismic Hazard maps by the Geological Survey of Canada (GSC, Adams and Halchuk, 200), and also updates a previous study for the Muskrat region by Atkinson (2009); the results of this study consider the effects of major uncertainties on the hazard at Muskrat, and incorporate up-to-date information on seismicity and ground motion prediction equations (GMPEs), which have evolved considerably over the last 10 years (e.g. see Atkinson and Goda, 2011 for discussion). The analysis does not include any local information on specific faults or geological structures. Rather it is assumed that there are no such local features that would affect the overall regional hazard estimates; thus an implicit assumption is that there is no evidence of faults that have moved in geologically recent times (last 10,000 years) in the site area. This assumption can be refined at a more detailed analysis stage if warranted in light of site-specific geologic information. For example, if a local fault with recent offset was identified, then this fault would delineate a local fault-based source zone, with geological information on the dates and extent of movement being used to define a recurrence relation for the fault. Microseismic studies of any such identified sources could also be conducted. However, such features are very rare in eastern Canada, and it is thus very unlikely that they will be identified in the site area. The analysis addresses natural seismicity, and does not address the probability of reservoir-induced seismicity or other potential induced seismicity sources, if any. In analyzing the engineering effects of ground motion, both the amplitude and frequency content of the vibrations are important. Therefore the seismic ground motions are expressed using the response spectrum (PSA(f)), which shows the maximum acceleration that a simple structure would experience as a function of its natural frequency. The response spectrum result is a Uniform Hazard Spectrum (UHS), in which the amplitude for each frequency corresponding to a specified exceedence probability is provided. The peak ground acceleration (PGA) for this probability is also estimated, as is the peak ground velocity (PGV). The frequency associated with the PGA varies, but in general the PGA is associated with high-frequency motions (near 10 Hz); the PGV is associated with motions near 2 Hz. The UHS results of this study are presented in the figures and tables provided in Section. Time histories of ground motion that match the UHS for specified probability levels may be developed in a later phase of the project if needed. The time histories may be derived by scaling or modifying real earthquake records that are appropriate for rock 4

5 sites in similar low-seismicity environments, for magnitude-distance ranges that dominate the hazard at Muskrat. The Muskrat site is located within a broad zone of scattered seismicity. We perform hazard calculations at the site for NEHRP B/C conditions (near-surface shear-wave velocity of 760 m/s); this is the reference site condition for the results. These results would apply to facilities at or near the site, provided it is founded on NEHRP B/C. The results can be adjusted for site class A to be applicable to hard rock site conditions (shearwave velocity >1500 m/s), or for soil sites classes C or D, using conversion factors that represent average site amplifications for such conditions. The conversion factors for correcting results to site class A, C, D are given herein. 2 - SEISMIC HAZARD ANALYSIS METHOD 2.1 Overview Seismic hazard analyses in eastern Canada are based on probabilistic concepts which allow incorporation of both geologic interpretations of seismic potential and statistical data regarding the locations and sizes of past earthquakes. The Cornell-McGuire method (Cornell, 1968; McGuire, 1976, 1977, 2004) has proven particularly well-suited to calculate expected ground motions for a wide range of seismic hazard environments, offering flexibility in the consideration of spatial and temporal characteristics of regional earthquake occurrence, and the basic physics of the earthquake process. In general, it is difficult to correlate seismicity with specific faults. Earthquakes typically occur at depths of 5 to 20 km, on faults that have no surface expression. Furthermore, faults mapped on the surface in eastern Canada were formed hundreds of millions of years ago, and may bear little relation to current seismic activity. Thus there is no clear-cut relationship between observed faults and seismicity. The spatial distribution of earthquakes is described by defining seismic source zones (areas which may contain groups of faults) on the basis of seismicity and seismotectonic interpretations; the earthquake potential of these zones is generally assumed to be uniform (assumption of non-uniformity, where the past events clusters drive seismicity pattern, result in slightly different solutions). The frequency of earthquake occurrence within each source zone is described by a magnitude recurrence relationship, truncated at an upper magnitude bound, Mx. Earthquake ground motion prediction equations (GMPEs) provide the link between the occurrence of earthquakes of various magnitudes and the resulting ground motion levels at any site of interest. The probability of exceeding a specified level of ground motion at a site can then be calculated by summing up the hazard contributions over all magnitudes and distances, including all source zones. In most cases, including this study, the hazard is dominated by contributions from the source zone within which the site is located. The hazard integral sums up the likelihood of earthquakes at all distances within all source zones, assuming that earthquakes are distributed randomly in space across the source zone (or alternatively by 5

6 using a spatial smoothing algorithm to consider the historical distribution of seismicity patterns more closely). To obtain ground motion levels or earthquake response spectra for a specified probability, calculations are repeated for a number of ground motion values, for all desired ground motion parameters, and interpolation is used to determine the relationship between ground-motion amplitude and annual probability. The Cornell-McGuire framework has been well-accepted in all parts of North America. In Canada, it forms the basis for the seismic hazard maps in the National Building Code of Canada (NBCC 1985 and beyond), and is the usual basis for seismic hazard evaluations of all important engineered structures. The results are generally expressed as a Uniform Hazard Spectrum (UHS), in which the amplitude for each frequency corresponding to a specified target probability is provided. The peak ground acceleration (PGA) and velocity (PGV) for the target probability may also be estimated. When time histories of ground-motion are required for use in engineering analyses, these may be derived to be consistent with the expected ground motion characteristics of the UHS for the target probability. The analysis methods used to generate UHS results and time histories are described in more detail by McGuire (2004). The calculations are carried out using the EqHaz program set developed by Assatourians and Atkinson (201) ( This program performs probabilistic seismic-hazard analysis (PSHA) by the Monte Carlo simulation method (Musson, 1998, 1999, 2000, 2012a; Hong and Goda, 2006; Assatourians and Atkinson, 201). The program can handle areal and fault sources with magnituderecurrence statistics as described by Gutenberg Richter (untruncated, truncated, or tapered), or a user-specified discretized cumulative or incremental magnitude-recurrence distribution. Ground-motion amplitudes are modeled as user-specified functions of magnitude and distance (in a table format). Both epistemic and aleatory uncertainty in the key input parameters can be modeled. The user may treat these uncertainty sources as equivalent (in terms of our ability to predict future ground motions), enabling the treatment of ground motions realized over a long simulated catalog as an extreme-value statistical problem. Alternatively, the user may treat epistemic uncertainty separately using confidence fractiles on the input parameters, which has been traditional practice (e.g., McGuire, 2004). The mean-hazard results, which are the focus of this study, are not sensitive to this choice. 2.2 Treatment of Uncertainty It has long been recognized that seismic hazard analyses are subject to greater uncertainties than those associated with most environmental phenomena. Two types of uncertainty exist: random uncertainty due to the physical variability of earthquake processes model uncertainty due to incomplete knowledge concerning the processes governing earthquake occurrence and ground motion generation (eg. uncertainties in input parameters to hazard analysis). 6

7 The first type of uncertainty is incorporated directly into the Cornell-McGuire analysis framework, and is included in a standard best-estimate seismic hazard result. The second type of uncertainty implies a spread of possible results about those that might be considered a best estimate. This type of uncertainty can cause differences in results, among alternative hypotheses, of factors of more than two. It also implies that, as new information on seismic hazard becomes available (through seismic monitoring and research) hazard estimates may change significantly from those developed at an earlier time. Seismic hazard analysis procedures have been developed to formally evaluate the level of model uncertainty (sometimes referred to as epistemic uncertainty) in hazard analyses. A logic tree approach is often used to represent each input parameter by a simple probability distribution, thereby producing a family of possible output hazard curves, with associated weights (McGuire, 2004). Such an approach has been used in hazard analyses for critical engineered structures such as dams and nuclear power plants, and has also been used in the national seismic hazard maps (Adams and Halchuck, 200; 201). The logic tree approach is simply a way of formalizing consideration of the implications of alternative assumptions. The analysis in this report incorporates both random variability in earthquake locations and ground motions, and model uncertainty. Model uncertainty is incorporated by considering a range of plausible alternatives for the seismotectonic source model, the magnitude recurrence parameters (including maximum magnitude) and the GMPEs. A mean-hazard UHS is provided, for probability levels of 1/1000, 1/2500, 1/5000 and 1/10,000 p.a. 2. Input Parameters for Seismic Hazard Analysis The input parameters for the seismic hazard analysis include the seismic source zonation, the magnitude recurrence parameters and maximum earthquake magnitude for each source zone, and the GMPEs for response spectra at several vibration frequencies and PGA and PGV Seismic source models The first step in the seismic hazard analysis is the definition of seismogenic source zones. Figure 1 shows seismicity of the region through the end of 201, as obtained from the Canadian Composite Seismicity Catalog of Fereidoni et at. (2011), with added information from 2010 through 201 as obtained from the website of the Geological Survey of Canada. All events in the catalog are given in moment magnitude (M), where M was converted from other magnitudes as necessary, as described by Fereidoni et al. (2011). The Muskrat site location is in an area of very sparse seismicity, with active regions such as those offshore or in the Lower St. Lawrence being too far distant (00 km) to cause significant hazard, except possibly at long periods (as long-period motions decay slowly). At intermediate-to-high frequencies, the hazard will be dominated by the sparse local seismicity. 7

8 In this study the source models developed by the Geological Survey of Canada (Adams et al., 2012) and proposed for implementation in the 2015 version of National Building Code of Canada are used. The overall concept of this model is very similar to that described by Atkinson and Goda (2011), though differing in some of its details. It comprises three alternative source models, as described in the following and shown in Figure 2. These models are respectively weighted 0.4, 0.4 and 0.2: a historical model (H2), which is composed of conventional areal sources defined on the basis of historical seismicity clusters. This model comprises a number of relatively small source zones together with a few large background zones. Though the zone boundaries are largely chosen to enclose seismicity clusters, there is also some account taken of broad regional geological features such as the Iapetan rifted margin. Earthquakes in the H2 model may have maximum magnitudes (Mx) up to the largest observed for continental interior regions, M~7.5. a hybrid model (HY), which assumes that the zones of historical seismicity clusters host earthquakes up to M 6.8 only. Larger events are modeled using broader, overlying regional seismotectonic sources, for events larger than magnitude 6.8 up to the same maximum magnitudes as for the H2 model. This follows the concept used in Atkinson and Goda (2011), with some differences in details, and represents an updating of the R model concept proposed by Adams and Halchuk (200). The HY model uses all of the sources in the H2 model (for earthquakes up to magnitude 6.8) together with 5 large seismotectonic sources, which are intended to capture the occurrence of rare, large earthquakes. To avoid double counting, the Mx for most of the H2 zones was capped at 6.8. However, because the 5 large seismotectonic sources do not cover the entire areal extent of the H2 zones the Mx in nonoverlapped zones were retained at their original H2 values. a regional model (R2) that maintains the philosophy and the basic geometry of the R model of Adams and Halchuk (200) as implemented in the 2010 NBCC. 8

9 Figure 1 Recorded seismicity (M>2) through 201 in the Muskrat area. 9

10 ECM LIR a b c Figure 2 Areal source model as proposed by Adams and Halchuk (201), with the location of the site shown by the symbol. The total model is the sum of H2(a), hybrid(b), and R2(c) models with 40%, 40%, and 20% weights respectively Magnitude Recurrence Relations Recurrence data, expressing the relative frequency of occurrence of earthquakes within a zone as a function of magnitude, can generally be fit to the Gutenberg-Richter relation: Log N(M) = a b M 10

11 where N(M) is the number of events per annum of magnitude M, M is moment magnitude, and a and b are the rate and slope of the relation. In most parts of the world, b values are in the range from 0.8 to 1, while a values vary widely depending on the activity level of the region. For this study, as we have adopted the recently-developed source zone models of the GSC, we likewise adopt their magnitude-recurrence relations for all zones (from Adams and Halchuk, 201). These recurrence relations were calculated with a maximum likelihood algorithm, using the seismicity data of the SHEEF2011 moment-magnitude catalog of the GSC (Adams and Halchuk, 201). The SHEEF2011 catalog (Geological Survey of Canada) is very similar to the Canadian Composite Catalog of Fereidoni et al. (2011). The magnitude recurrence relation for the Labrador Iapetan Rift source zone (LIR zone in H2 and Hybrid models, in which the Muskrat is located) is shown in Figure as an example. The complete set of seismicity parameters for all zones of the three models, as implemented in this study, are given in the appendix. These follow the recurrence relations and maximum magnitudes used by the GSC. For the minimum magnitude, we used a value of 4.8, consistent with the GSC value. Figure Recurrence Relations for the LIR zone of H2 and Hybrid models, showing three alternative sets of recurrence parameters to represent epistemic uncertainty (blue is the middle alternative, green and red are low and high alternatives, respectively. Each one is associated with three alternative Mx values. 11

12 In addition to using the seismicity parameters for uniformly-distributed seismicity within all zones, from the GSC model, we also consider the implication of a smoothed seismicity model, in which the distribution of events within each of the zones is assumed to follow that of the historical seismicity catalog (as based on the Canadian Composite Catalog). This provides a test of the sensitivity of results to the spatial distribution of events, and its unevenness in some zones. The algorithm is as described by Assatourians and Atkinson (201). The completeness intervals used in interpreting the historical seismicity distribution for this exercise are as follows (where it is assumed that the historical catalog is complete for each magnitude level beginning with the years quoted): Year to begin statistics: M2.5 M2.8 M. M4. M4.8 M Ground motion prediction equations Ground motions on the reference ground condition of B/C boundary are given in this analysis by the ground-motion prediction equations (GMPEs) of Atkinson and Adams (201) for eastern North America (ENA). The equations provide peak ground acceleration (PGA) and velocity (PGV), as well as response spectra (PSA, 5% damped horizontal component) as a function of moment magnitude and hypocentral distance. The hypocentral distance version of the Atkinson and Adams equations is used, as the events are treated as point sources in the hazard analysis, with appropriate conversions from finite-source to point-source distance metrics implemented as described by Atkinson and Adams (201). Hypocentral distance was converted to epicentral distance assuming an average focal depth of 10 km. To consider epistemic uncertainty in the median GMPEs, Atkinson and Adams define a middle, lower, and upper suite of GMPEs for input to the hazard analysis. These alternatives were defined by considering alternative proposed GMPEs for ENA, and the data that guided them. We considered these three alternatives with weights of 0.5 (middle) and 0.25 (lower, upper). The considered suite of groundmotion relations is shown on Figure 4 for NEHRP B/C site conditions (random horizontal component). Conversion of results to other site conditions is discussed in Section. On Figure 5, we provide an example of the GMPE plotted versus frequency, for a typical scenario event of M6.5 at 50 km. 12

13 Figure 4 Ground-motion prediction equations for ENA (B/C) considered in this study, at two samples frequencies (5 Hz on left, 0.5 Hz on right), for events of M=5, 6, 7, as a function of epicentral distance. The solid lines show the middle equation, while the dashed lines show the upper and lower curves to express epistemic uncertainty. The equations were developed by Atkinson and Adams (201), considering the mean and standard deviation of estimates from alternative GMPE models, as indicated by the symbols. After atkinson and Adams,

14 Figure 5 Middle and Upper branches of the GMPE for ENA (B/C) considered in this study, evaluated for an M6.5 earthquake at 50km epicentral distance. Random uncertainty in the relations was modeled by a lognormal distribution of ground motion amplitudes about these median relations, with a standard deviation of 0.2 log (base 10) units for high frequencies, increasing to 0.27 units at low frequencies. This random uncertainty is consistent with recent studies (eg. Atkinson, 2012; Atkinson and Adams, 201). - RESULTS OF SEISMIC HAZARD ANALYSIS.1 Mean-hazard UHS and Sensitivity Using the input parameters given in the previous section, the PGA, PGV and response spectra were computed for a range of probabilities using the Monte Carlo simulation method, as described by Assatourians and Atkinson (201), with the EqHaz software suite. The appendix contains the input files to EqHaz for the analysis. The values of PGA and PSA (5% damped), for the horizontal component of motion on the reference B/C condition are displayed as a function of probability on Figure 6. These curves were obtained using the uniform seismicity approach (in which the seismicity is distributed uniformly within each source zone); this approach is the one adopted by the GSC for the national seismic hazard maps. 14

15 Figure 6. Hazard curves of motions at Muskrat calculated for NEHRP site B/C. The black horizontal lines correspond to probability levels of 1/1000, 1/2475, 1/5000, and 1/10,000. On Figure 7, the contributions to hazard are illustrated for the Hybrid model, which gives a total hazard curve that is very similar to the weighted results in Figure 6 from all models. Thus Figure 7 provides a good representation of where the hazard contributions originate. It is observed that at low frequencies (1 Hz), the dominant hazard contribution is the offshore ECM zone, which has a relatively high rate of large magnitude events. At high 15

16 frequencies (10 Hz), the dominant source of hazard is the LIR zone in which the site is located. Figure 7. Hazard curves at Muskrat calculated for NEHRP site B/C, shown by source zone for the Hybrid model. The black dotted line shows the total hazard curve considering all sources. The largest contribution to 10,000yr return period motion at 1Hz comes from the source ECM, while at 10Hz it comes from the source LIR (both sources are labeled in Figure 2). Figure 8 summarizes the mean-hazard UHS results for selected return periods from 1000 to 10,000 years (where the return period is the inverse of the annual probability), for the reference site condition (NEHRP B/C). These results are listed in Table 1, for the model assuming uniform distribution of seismicity within all source zones. For comparison, Figure 8 also shows the sensitivity of results to whether we assume a uniform or smoothed distribution of seismicity within the source zones. This check shows that use of the smoothed seismicity approach would reduce the hazard estimates significantly, in comparison to the uniform seismicity approach results as listed in Table 1. This is because very few events have been observed in the area of the site, raising the possibility that the hazard estimates smoothed over the entire broad regional zones is conservative, especially at high frequencies. It is noted, however, that the hazard in the area is very low, and thus difficult to determine with precision. Moreover, the uniform seismicity approach is being used by the GSC for the national seismic hazard maps. We therefore recommend the use of the uniform seismicity values, as given in Table 1, for this site. 16

17 Figure 8 Mean-hazard horizontal-component UHS (5% damped) for Muskrat at a range of target probabilities. All for NEHRP B/C. Solid lines show results assuming uniform seismicity within source zones; dotted lines show results if a smoothed seismicity model is used. The peak ground acceleration (PGA) is plotted for reference on Figure 8 at a frequency of 90 Hz, but a curve between 20 Hz and 90 Hz is not plotted (no spectral values were calculated for frequencies above 20 Hz). The PGA refers to the maximum acceleration of the ground shaking during the seismic event (ie. the peak amplitude on a free-field record of ground acceleration versus time) it does not have an actual associated frequency, as the frequency at which the PGA occurs will depend on the earthquake magnitude and distance. The response spectrum shows the maximum acceleration of a damped singledegree-of-freedom oscillator, when subjected to the input record of ground acceleration versus time. Oscillators with a high natural frequency will respond to input ground motions that are rich in high frequency content, while oscillators with low natural frequency will respond more strongly to input ground motions that are rich in low frequency content. 17

18 Table 1 Mean-Hazard Ground Motions for Muskrat, for 5% damped horizontalcomponent PSA, PGA (cm/s 2 ) and PGV (cm/s), for NEHRP B/C site conditions, for a range of annual probabilities. Frequency (Hz) 1/1000 1/2475 1/5000 1/ PGA PGV Results for NEHRP A site conditions (hard-rock) and for soil sites The seismic hazard results have been obtained for NEHRP B/C site conditions, as given in Table 1. To obtain the corresponding ground motions on NEHRP A (hard rock, shear-wave velocity>1500 m/s) from the NEHRP B/C results, we must consider the amplification that is expected for B/C sites, relative to A. We adopt the factors given in Atkinson and Adams (201), which were taken from the study of Atkinson and Boore (2006). The adopted factors expressing the amplification of B/C sites relative to hard rock are listed in Table 2. 18

19 Table 2 Amplification factors to go from NEHRP A results (hard rock) to NEHRP B/C results (shear-wave velocity 760 m/s in upper 0 m). From Atkinson and Adams (201). Frequency Amplification (A to B/C) PGA 10 ( log(Repi)) PGV 1.2 For most ground motion parameters, the factors of Table 2 may simply be divided out of the results of Table 1 to obtain motions for NEHRP A site conditions. For PGA, we modified the GMPE to correspond to values on Class A (applying the distance-dependent conversion of Table 2 from B/C to A), then recomputed the hazard. The resulting motions for hard-rock sites (NEHRP A) are shown in Table. Table Weighted-Mean-Hazard Ground Motions for Muskrat, for 5% damped horizontal-component PSA, PGA (cm/s 2 ) and PGV (cm/s), for NEHRP A site conditions, for a range of annual probabilities. Frequency (Hz) 1/1000 1/2475 1/5000 1/10, PGA PGV

20 We note that the results listed in Table can be directly compared with the previous seismic hazard results for Muskrat, which were for hard rock, by Atkinson (2009). Figure 9 presents this comparison with the previous study. Overall, the results are very similar to those obtained previously, in spite of the many changes in the input parameters. At high frequencies, the present study estimates lower ground motions that previously by a modest amount, while at low frequencies the results are very nearly the same. Figure 9. Comparison of UHS calculated for site class A in this study (black lines) in comparison to results of Atkinson (2009). Results have been provided above for both B/C and hard rock site conditions. If estimates are needed for softer site conditions, they can be derived by amplifying the motions given in Table 1 for B/C, to the appropriate site condition. This can either be done by site specific studies, or by the application of generic soil amplification factors. In Table 4, we present site amplification factors for typical NEHRP C and D soils, based on the equations of Boore and Atkinson (2008). The factors are given to amplify from B/C to NEHRP C (assumed 450 m/s) and D (assumed shear-wave velocity = 250 m/s), for an assumed level of shaking corresponding to PGA<0.1g. Note that due to the low ground-motion levels involved, soil response is essentially linear, which means the same factors can be applied for all probabilities. 20

21 Table 4 Amplification factors to go from NEHRP B/C results (760m/s) to NEHRP C results (shear-wave velocity 450 m/s in upper 0 m) and to NEHRP D results (shear-wave velocity 250 m/s in upper 0 m) assuming pga<0.1g. Freq (Hz) Amplification Amplification (B/C to C) (B/C to D) Deaggregation To understand what types of events contribute most to the hazard of the study area, one carries out a deaggregation to assess the relative contributions to the exceedence probability of a selected ground motion parameter by magnitude and distance. In Figure 10, deaggregation plots of 1/10,000 per annum motions at 0.2Hz, 1.0Hz, 5.0Hz and PGA are shown. This shows that the high frequency motions are coming from a broad range of magnitudes at distances close to the site. By contrast, longer period motions are controlled by large events out to very large distances, in particular from more active areas offshore. Details of the deaggregation plots of Figure 10 are provided in Appendix B, where the first 21

22 columns (magnitude, distance, relative contribution to hazard at 1/10,000 p.a.) are the information plotted in Figure 10; Appendix B also shows the average epsilon for the contributions from each magnitude-distance bin, and its standard deviation, as well as the number of occurrences that are represented in the bin (from the Monte Carlo simulations of EQHAZ). Table 5 provides, for the 1/10,000 motions, the mean deaggregation results, expressing the weighted average of the magnitude, distance and epsilon contributions (where epsilon is the number of standard deviations with respect to the median GMPE; for example, if epsilon is 1.5, then the average contributions are coming from motions that are 1.5 standard deviations above the median). For high frequencies, the dominant hazard contributions have a simple distribution reflecting the local source origin of most of the hazard. At low frequencies, the dominant hazard contribution shifts to large events at regional distances. At intermediate frequencies (1 Hz), there is a bimodal distribution in which both local and distance sources contribute. Therefore Table 5 is subdivided for 1 Hz into the mean contribution for distances <240 km, and distances >240 km. Overall, it is apparent that there are two significant scenarios for hazard: an event of M6.1 to 6.5 at a distance of 90 to 100 km, and an event of M~7. at 50 to 450 km. The average duration (significant duration considering the 5% to 95% of the acceleration time series) for the mean deaggregation event is also given in Table 5; this duration was calculated as the sum of the source duration and path duration components, using the empirical results of Boore and Thompson (2014). The long duration for low-frequency motion scenarios arises from the large distances, which result in a large path duration component. However, these long-duration signals are quite weak in amplitude. Table 5 Mean deaggregation results for 1/10,000 motions M R[km] Epsilon Duration[s] PGA PSA(5.0Hz) PSA(1.0Hz), R < 240km PSA(1.0Hz), R 240km PSA(0.2Hz)

23 Figure 10. Deaggregation plots of 10000yr return period motion at Muskrat. 4 - REFERENCES Adams, J., and S. Halchuk (200). Fourth generation seismic hazard maps of Canada: Values for over 650 Canadian localities intended for the 2005 National Building Code of Canada. Geological Survey of Canada Open File pp. 2

24 Adams, J. and S. Halchuk (201). Documentation for the fifth generation seismic hazard maps of Canada. In preparation by the Geological Survey of Canada. Assatourians, K., and Atkinson, G., (201). EqHaz: An Open-Source Probabilistic Seismic-Hazard Code Based on the Monte Carlo Simulation Approach. Seism. Res. Lett., 84, Atkinson, G. (2009). Earthquake hazard analysis: Gull and Muskrat damsites, Lower Churchill. SNC Lavalin Ltd., 22p. Atkinson, G., and J. Adams, (201). Ground motion prediction equations for application to the 2015 Canadian national seismic hazard maps. Can. J. Civ. Eng., 40, Atkinson, G. M., and Goda, K. (2011). Effects of seismicity models and new ground motion prediction equations on seismic hazard assessment for four Canadian cities, Bull. Seismol. Soc. Am. 101, Boore, D. M. and G. M. Atkinson (2008). Ground-motion prediction equations for the average horizontal component of PGA, PGV, and 5%-damped PSA at spectral periods between 0.01 s and 10.0 s, Earthquake Spectra 24, Boore, D. and E. Thompson (2014). Path durations for use in the stochastic-method simulation of ground motions. Bull. Seism. Soc. Am., submitted. Cornell, C. (1968). Engineering seismic risk analysis. Bull Seism. Soc. Am., 58, Fereidoni, A., Atkinson, G., Macias, M., and K. Goda, (2011). CCSC: A Composite Seismicity Catalog for Earthquake Hazard Assessment in Major Canadian Cities. Seism. Res. Lett., 8, Hong, H. P., and Goda, K.(2006). A comparison of seismic-hazard and risk deaggregation, Bull. Seismol. Soc. Am. 96, McGuire, R. (1976). FORTRAN computer program for seismic risk analysis. U.S. Geol. Surv. Open-file Rpt McGuire, R. (1977). Seismic design spectra and mapping procedures using hazard analysis based directly on oscillator response. Intl. J. Earthq. Eng. Struct. Dyn., 5, McGuire, R. (2004). Seismic hazard and risk analysis. EERI Monograph MNO-10. Earthq. Eng. Res. Inst., Oakland, Ca. Musson, R. M. W.(1998). On the use of Monte Carlo simulations for seismic hazard assessment, in Proc. of the sixth US National World Conference on Earthquake Engineering, 12 pp. Musson, R. M. W. (1999). Determination of design earthquakes in seismic hazard analysis through Monte Carlo simulation, J. Earthq. Eng., Musson, R. M. W.(2000). The use of Monte Carlo simulations for seismic hazard assessment in the U.K., Ann. Geofisc. 4,

25 Musson, R. M. W.(2012). PSHA validated by quasi observational means, Seismol. Res. Lett. 8, LIST OF ABBREVIATIONS CC Central Canada ENA Eastern North America GMPE ground-motion prediction equation GSC Geological Survey of Canada IRB Iapetan rifted background IRM Iapetan rifted margin M moment magnitude ML local magnitude MN Nuttli magnitude NA North America NBCC National Building Code of Canada NEHRP National Earthquake Hazard Reduction Program p.a. per annum PGA peak ground acceleration PGV peak ground velocity PSA Pseudo-acceleration, 5% damped RIS Reservoir-induced seismicity UHS- Uniform hazard spectra 25

26 APPENDIX A, EQHAZ1 INPUTS: A.1- H2 model! EqHaz1 input (H2(40%)+HY(40%)+R2(20%)! GSC H2 source model with weight of 0.4! This run will generate years of synthetic events to be used in EqHaz !Number of subcatalogue simulations and number of years of each subcatalog 7 0!Number of zones and number of faults 1 1 0!Seed value of random generation subroutine (Use 0 for random seed, integer for repeatable seed), mixing epistemic/aleatory uncertainties (Yes=1, No=0), do smooth seismicity (Yes=1, No=0)!ccsc11east.txt smooth_gat.par!name of earthquake catalog file used for smooth seismicity approach, name of parameter file used for smooth seismicity approach ADRN ADRS

27 AOBH AOH

28 BSL BSLE

29 BSLW CHA

30 CHV CLO CNH

31 COCN COCS

32 GAT GATW

33 GNS JMS

34 KIP LCH LFN

35 LIR LSP

36 MIR MNT

37 NAN OBGH

38 PEM PMQ

39 QCR SAG

40 SEB SLE

41 SVH TRR

42 WLO SCCE

43 SCCE

44

45 A.2- HY (Hybrid) model! EqHaz1 input (H2(40%)+HY(40%)+R2(20%)! GSC HY source model with weight of 0.4 for Muskrat exercise Feb.-Mar. 2014! This run will generate years of synthetic events to be used in EqHaz !Number of subcatalogue simulations and number of years of each subcatalog 44 0!Number of zones and number of faults 1 1 0!Seed value of random generation subroutine (Use 0 for random seed, integer for repeatable seed), mixing epistemic/aleatory uncertainties (Yes=1, No=0), do smooth seismicity (Yes=1, No=0)!ccsc11east.txt smooth_gat.par!name of earthquake catalog file used for smooth seismicity approach, name of parameter file used for smooth seismicity approach ADRN ADRS

46 AOBH

47 AOH BSL

48 BSLE BSLW CHA