BC HYDRO SSHAC LEVEL 3 PSHA STUDY METHODOLOGY

BC HYDRO SSHAC LEVEL 3 PSHA STUDY METHODOLOGY M. W. McCann, Jr. 1, K. Addo 2 and M. Lawrence 3 ABSTRACT BC Hydro recently completed a comprehensive Probabilistic Seismic Hazard Analysis (PSHA) to evaluate the ground motion hazard for a land area of about 1.5 million km 2 that contains a portfolio of assets, including forty one dams and appurtenant structures. The PSHA was carried out following the methodology developed by the Senior Seismic Hazard Analysis Committee (SSHAC) and was conducted with the intent that it would be technically defensible and stable for 10 to 15 years following its development. It is the first non-nuclear project to adopt and fully implement a SSHAC Level 3 analysis. The SSHAC guidelines provide a framework for conducting a PSHA project that defines the organizational structure and responsibilities of the project team, the goals for the analysis, and guidance for evaluating epistemic uncertainties. The SSHAC process recognizes that, as a result of epistemic uncertainties, the inputs to a PSHA are multi-valued. Uncertainties in the assessment of future ground motions stem from incomplete data and scientific understanding (epistemic uncertainties) and from process variability (aleatory uncertainties). Identifying and quantifying these uncertainties requires clear, transparent, logical and documented evaluation and integration by experts. 1 President, Jack R. Benjamin & Associates, Inc., 530 Oak Grove Avenue, Suite 202, Menlo Park, CA 94025 2 Specialist Engineer, BC Hydro Engineering, 6911 Southpoint Drive, Burnaby, BC, V3N 4X8 3 Specialist Engineering Geologist, BC Hydro Engineering, 6911 Southpoint Drive, Burnaby, BC, V3N 4X8 McCann, Jr., MW, Addo, K, Lawrence M. BC Hydro SSHAC Level 3 Study Methodology. Proceedings of the 10 th National Conference on Earthquake Engineering, Earthquake Engineering Research Institute, Anchorage, AK, 2014.

BC Hydro SSHAC Level 3 PSHA Study Methodology M. W. McCann, Jr. 2, K. Addo 2 and M. Lawrence 3 ABSTRACT BC Hydro recently completed a comprehensive Probabilistic Seismic Hazard Analysis (PSHA) to evaluate the ground motion hazard for a land area of about 1.5 million km 2 that contains a portfolio of assets, including forty one dams and appurtenant structures. The PSHA was carried out following the methodology developed by the Senior Seismic Hazard Analysis Committee (SSHAC) and was conducted with the intent that it would be technically defensible and stable for 10 to 15 years following its development. It is the first non-nuclear project to adopt and fully implement a SSHAC Level 3 analysis. The SSHAC guidelines provide a framework for conducting a PSHA project that defines the organizational structure and responsibilities of the project team, the goals for the analysis, and guidance for evaluating epistemic uncertainties. The SSHAC process recognizes that, as a result of epistemic uncertainties, the inputs to a PSHA are multi-valued. Uncertainties in the assessment of future ground motions stem from incomplete data and scientific understanding (epistemic uncertainties) and from process variability (aleatory uncertainties). Identifying and quantifying these uncertainties requires clear, transparent, logical and documented evaluation and integration by experts. Introduction BC Hydro had been involved in the evaluation of seismic hazards for its dam projects for decades. The evolution of its practices paralleled those in the hydropower industry, including performing PSHAs using the method introduced by Cornell [1]. These studies were typically performed on a site-by-site basis as project needs required. However, with the advancement of PSHA methods, the increased availability of geologic and seismologic data, and recommendations made by a review panel, it became clear there was a need for an evaluation that addressed the seismic hazard for the entire BC Hydro service region, and included a modern comprehensive assessment whose results would be stable for 10 to 15 years. 1 President, Jack R. Benjamin & Associates, Inc., 530 Oak Grove Avenue, Suite 202, Menlo Park, CA 94025 2 Specialist Engineer, BC Hydro Engineering, 6911 Southpoint Drive, Burnaby, BC, V3N 4X8 3 Specialist Engineering Geologist, BC Hydro Engineering, 6911 Southpoint Drive, Burnaby, BC, V3N 4X8 McCann, Jr. M. W., Addo, K, Lawrence M. BC Hydro SSHAC Level 3 Study Methodology. Proceedings of the 10 th National Conference on Earthquake Engineering, Earthquake Engineering Research Institute, Anchorage, AK, 2014.

In view of these considerations, BC Hydro elected to conduct a SSHAC Level 3 analysis for its service region [2, 3]. While the primary motivation for the study was the safety and reliability of its dams, a related motivation was the need to meet the broader needs of managing its overall infrastructure portfolio that included hydropower production sites, transmission facilities, office buildings, etc. Because of the critical nature of its dams, the PSHA placed particular focus and a high level of detail in the seismic source characterization model development in parts of the province where BC Hydro dams are located; a total of 42 sites (41 existing dam sites and one proposed site). As such nearly the entire province and parts of Alberta, and the Northwestern U.S. were part of the study region [4]. The PSHA team consisted of a contingent of earth scientists and engineers grouped into six entities project leads (PL), conceptual tectonic framework (CTF), seismic source characterization (SSC), ground motion characterization (GMC), implementation and computations (IMC) and a participatory peer review panel (PPRP) who oversaw the expert elicitation process and the technical evaluations that were conducted. This paper presents the perspective and needs of an owner of a major lifeline infrastructure system in conducting a study of this scale. Topical areas that are discussed include the SSHAC process, PSHA methodology, major features of the PSHA including new methodological and implementation developments and study results. Owner s Perspective and Needs BC Hydro is the owner and operator of a major electric power infrastructure system in British Columbia whose assets provide lifeline services to the province and at the same time pose significant economic and public safety risks in the event of dam failure resulting in uncontrolled release of reservoir. In addition, due to the proximity of multiple BC Hydro dams to the Cascadia subduction zone which is capable of generating magnitude 9+/- earthquakes, many project sites may experience high ground motions simultaneously. In addition to the need to manage its facilities in a seismically active environment, BC Hydro recognized that its approach to seismic hazard assessment on a site-by-site basis proved to be problematic from the perspective of technical consistency and stability. In addition, there had been considerable advancement of the evaluation of seismic hazards with the introduction of the SSHAC process that had not yet been integrated into BC Hydro s standard of practice. With this background and the findings of an internal review board s recommendations, BC Hydro elected to conduct a SSHAC Level 3 analysis for the entire BC Hydro service region that met the following general objectives: Technical stability of the PSHA model and the stability of the PSHA results over time (for a period of 10 to15 years). Transparency of the modelling and evaluations that were conducted that supports BC Hydro s staff capability to understand and interpret the basis for the PSHA inputs, and as might be required in the future to refine or update elements of the PSHA inputs.

Provide training for BC Hydro staff in areas such as the SSHAC process, PSHA, and implementation of the PSHA model. Establish an in-house capability to implement and use the PSHA (software and inputs) for other BC Hydro assets. To implement the SSHAC process, BC Hydro formed a project team of internal staff and external experts to conduct the study. Stability in the PSHA results is important to BC Hydro due to the value of the investment in the PSHA product and the impact that changes in the hazard results have on facility safety assessments and future costs of seismic modifications that may be needed. Study Region and Tectonic Complexity The study region covers most of BC, part of the northwestern US and part of eastern Alberta. The PSHA was a major programmatic and technical undertaking given the size and tectonic diversity of the region. Eastern BC and Alberta are in the stable continental interior (Fig. 1), whereas the western edge of the study region is located along the active plate margin of the Cascadia subduction zone interface which may generate up to M 9 earthquakes below Vancouver Island and southwestern BC. Other factors that contributed to the study complexity included: 1. The availability of data to perform the PSHA is very heterogeneous across the province. For much of the study region, there is limited data or large gaps in basic data. For instance, in the eastern part of the study region the historic earthquake catalogue is limited and incomplete due to low seismicity rates and limited seismic monitoring. 2. Due to low seismicity rates in much of BC there are few strong motion recording stations and thus a limited number of strong motion recordings. 3. Increasing sophistication in the modeling and parameterization of a PSHA requires detailed specification of a number of seismic source properties that were not previously considered in past BC Hydro PSHA studies. 4. There are elements of the PSHA that required new development to support the project objectives. These include improvements in seismic source characterization and modeling capabilities, the use in the ground motion evaluation of the single-station sigma modeling approach, and the need to develop a new subduction ground motion prediction model. SSHAC Process The original SSHAC process was developed in the mid-1990s and published in 1997 [2]. It was developed as a result of critical need to establish a process for conducting scientific evaluations in which inputs must be derived from expert elicitations. Based on a decade of experience implementing the SSHAC process, additional guidance was developed by the US Nuclear Regulatory Agency (USNRA) [4]. This new guidance states the overall goal for a PSHA in the following terms:

The fundamental goal of a SSHAC process is to carry out properly and document completely the activities of evaluation and integration, defined as: Evaluation: The consideration of the complete set of data, models, and methods proposed by the larger technical community that are relevant to the hazard analysis. Integration: Representing the center, body, and range of technically defensible interpretations in light of the evaluation process (i.e., informed by the assessments of existing data, models, and methods). The assessment of earthquake ground motion hazard (seismic hazard curves) at a site and the quantification of epistemic uncertainty in the estimate of ground motion hazards are directly tied to the development of the PSHA inputs (i.e., SSC and GMC). The SSHAC goal recognizes there are epistemic uncertainties associated with the development of the inputs to the PSHA that is attributable to limited or incomplete datasets and diversity in scientific understanding, and process variability (aleatory uncertainties). Achieving the SSHAC goal requires that evaluations be conducted such that a complete understanding of the stateof-knowledge of the technical Figure 1. Extent and tectonic diversity of the BC community is achieved and sources of Hydro PSHA study region. uncertainty are identified and modeled in a technically sound, complete and transparent fashion. In the context of a dynamic scientific environment, stability of the PSHA inputs (i.e., SSC and GMC) is defined in the context of the SSHAC goal. The inputs are judged to be stable (and thus the PSHA results as well) if future modifications (due to new developments in the next 10 to15 years) based on new data or scientific interpretations falls within the body and range of the epistemic uncertainty distributions captured in the PSHA inputs. While it cannot be guaranteed that all future interpretations will be explicitly represented in the composite uncertainty distribution that is developed, it is reasonable to expect that if the composite distribution is truly determined, the likelihood that new interpretations will fall outside the distribution should be small.

PSHA Model The PSHA is based on the model developed principally by Cornell [1]. The Cornell model is the basis for estimating the future, random occurrence of earthquake ground motions at a site; it is the aleatory model for seismic hazard. Aleatory Model The occurrence of earthquakes in a seismic source zone or on a fault source is modeled as a Poisson process. The probability that a ground motion parameter "Z" exceeds a specified value "z" in a time period "t" is given by: P( Z ν ( z) t 0 > z t = t ) = 1. e (1) 0 0 where ν(z) is the average annual frequency of events in which Z exceeds z per unit time (typically one year). The total average annual number of events in which Z exceeds z is obtained by summing the contributions from all sources, that is: ν ( z ) = ν i ( z) (2) n where ν i (z) is the mean annual number (or rate) of events associated with seismic source i for which Z exceeds z at the site. The parameter ν i (z) is given by the expression: o u m f > o i ( m) fi ( r m) P( Z z m, r) dr ν i ( z) = αi( m ) dm (3) m 0 where; α i (m o ) = annual frequency of earthquakes on source i above a minimum earthquake magnitude, m o ; f i (m) = probability density of earthquake size for source i for earthquakes between m o and a maximum magnitude for the source, m u ; f i (r m) = probability density for distance (r) to an earthquake of magnitude m occurring on source i, and P(Z > z m,r) = probability that given an earthquake of magnitude m that occurs a distance r from a site that the ground motion exceeds the specified level z. The SSHAC process evolved from recognition that the inputs (models and parameters) to the PSHA aleatory model (see equation 3), are uncertain, multi-valued. As such, a complete Bayesian evaluation of these uncertainties is required to provide a complete and technically defensible assessment of the seismic hazard.

Modeling Epistemic Uncertainties To model epistemic uncertainties in the PSHA, logic trees are used. Logic trees are applied in the individual parts of the PSHA; in the GMC and in the SSC, and to combine all the sources of uncertainty. In the SSC logic trees are used to capture: uncertainty for individual seismic sources (Fig. 2); epistemic uncertainties that are a source of dependence between seismic sources; and the combination of seismic sources to determine the total seismic hazard at a site. A master logic tree models the dependence between sources, where dependencies may be associated physical or parametric factors that are derived from the conceptual tectonic framework or from regional (e.g., location of the Cascadia subduction zone interface) or local factors associated with individual seismic sources (i.e., alternative seismic source boundaries). Implementation of the master logic tree (i.e., combination of the individual seismic source logic trees as defined in the master logic tree) produces the global SSC logic tree, which is implemented in the PSHA calculations. The global SSC logic tree includes the combination of seismic sources that are used to estimate the ground motion hazard at a site. Figure 2. Illustration of a BC Hydro SSC seismic source logic tree. Seismic Source Characterization The objective of the SSC is to use available earth science information (data, evaluations and interpretations of these data, including evaluations of others not directly involved in the

SSC) and integrate them into a composite temporal and spatial (aleatory) model of the future rate of earthquake occurrences. Due to limitations of available data and uncertainties associated with understanding the current stress regime and ongoing seismogenic processes, there are epistemic uncertainties that must be identified and quantitatively evaluated to produce a distribution on the SSC inputs to the PSHA model. The purpose of the SSC is to estimate the rate (events per year), location (spatial distribution) and size (earthquake magnitude) of future earthquake occurrences within a seismic source. Traditionally, sources of future earthquake occurrences have been defined in terms of fault sources (or in some cases fault zones) and area sources (sometimes referred to as volume sources). The highest level of specificity of a seismic source and thus the location of future earthquakes is achieved through the use of fault sources (and fault zones). A fault source is used to model a specific geologic feature for which there is evidence to characterize and estimate future earthquake occurrences, its location and rate of occurrence. Area sources are used to define broad (and potentially large) geographic areas or area sources of background seismicity. The temporal rate of earthquake occurrences is defined in terms of a recurrence relationship that quantifies the rate of occurrence of earthquakes of different magnitude. The truncated exponential and the characteristic earthquake recurrence models are typically used. The combination of recurrence models and seismic sources defines a spectrum of SSC modeling alternatives as shown in Fig. 3a. As part of the SSC data gathering and evaluation process, there was a considerable amount of geologic and seismologic information on mapped faults, moderate/large magnitude earthquakes that had not yet been associated with nearby tectonic features, and marine evidence of potential offsets that had not been investigated, etc. Within the context of the traditional spectrum of SSC modeling alternatives, much of this data could not be readily incorporated into the identification and characterization of seismic sources (Fig. 3a). In an effort to better utilize available geologic evidence new SSC modeling tools were developed. The first is the zone and embedded fault concept that provides a higher level of specificity of the characteristics of future earthquake occurrences (Fig. 3b). This concept provides an alternative modeling approach for characterizing the spatial distribution of future earthquake occurrences in a seismic source zone (an area source), that relaxes the assumption of uniform seismicity or spatial smoothed seismicity that is often used in seismic source zones [5]. In this case, the spatial smoothing is defined by the location of the embedded faults. Embedded faults are a network of tectonic features which, in aggregate, represent an alternative model for the spatial location of future earthquake occurrences within a seismic source zone. The network of embedded faults is integral to the seismic source zone and not an independent source of earthquake occurrences. The location and geometry-related characteristics of embedded faults are defined based on specific geologic features within the source zone, but other aspects of the seismic source characterization, such as deformation rates and maximum magnitudes, are derived from the attributes of the source zone in which they are located.

(a) (b) Figure 3. Illustration of the SSC modeling alternatives; a) traditional modeling approach, and b) as used in the BC Hydro PSHA with the introduction of the zone and embedded fault source concept, and the alternative characteristic (AC) recurrence model.

The second is the alternative characteristic (AC) recurrence model [6] which is a described in a companion paper in these proceedings. Ground Motion Characterization In a PSHA, estimates of earthquake ground motion are made using ground motion prediction equations (GMPE) that estimate ground motions as a function of earthquake magnitude, source-site distance, style of faulting and site conditions. GMPEs provide a best estimate (median) ground motion and its aleatory variability (randomness). For the BC Hydro PSHA three sets of GMPEs are required to estimate ground motions for the different tectonic environments that comprise the study regions; Cascadia subduction zone (interface and intraslab); crustal earthquakes east of the plate margin; and the stable continental interior of eastern BC and Alberta. To take into account the epistemic uncertainty in the GMPEs, uncertainty in individual models (due to lack-of-fit to data) and differences in alternative modeling interpretations associated with different model types, variations in datasets, etc. are considered. Crustal Earthquakes In the BC Hydro PSHA, the Next Generation Attenuation (NGA) models were used to model ground motions associated with earthquakes that occur in active crustal sources; i.e., regions in western BC. For seismic sources in the stable continental interior (i.e., Alberta), alternative models available in the literature were evaluated and used. Probability weights were assigned to each alternative GMPE based on its technical merit and the evaluation by the GMC Team. Subduction Zone Earthquakes During the first GMC workshop, the team concluded that available ground motion prediction models could not be used to estimate ground motions associated with subduction zone earthquakes. The team noted the range in existing subduction ground motion models estimates is significant and many of the models had not been updated as new ground motion data became available. As a result it was judged that simply using existing models and assigning weights to them was not appropriate. Consequently, the GMC TI and the project sponsor concluded that an effort would be undertaken to develop a new subduction ground motion prediction equation. Aleatory Variability Single Station Sigma The aleatory variability of ground motions (the aleatory logarithmic standard deviation in GMPEs) is generally determined from residuals of ground motions as derived from the statistical analysis carried out to estimate the parameters of the GMPE. There are a number of factors that contribute to this variability, including the random differences between earthquakes of the same magnitude (i.e., dynamic stress drop, fault rupture characteristics), differences in propagation path between the earthquake source and strong-motion recording stations (even for recordings

from the same earthquake), and variations in near-surface geologic conditions between strongmotion recording stations. Recent studies of ground motion variability have estimated a component of the total variability that is attributable to the variation in site-response. These studies, which have examined strong-motion datasets where individual recording sites have multiple ground motion recordings, have shown the variability at individual sites, referred to as the single station sigma is lower than the traditional or total aleatory standard deviation that is estimated in ground motion regression analysis [7]. In this project, the single-station sigma was used in the PSHA calculations. A key advantage of using the single-station sigma is that it provided a framework for using site-specific information in the evaluation of seismic ground motions at a site. For individual sites, it offers the flexibility to incorporate into the hazard results, the effects of site-specific amplification, variability, and epistemic uncertainties associated with the site response effects. The diversity of the seismic hazard at BC Hydro facility locations is reflected in the mean seismic hazard curves for peak ground acceleration (PGA) for sites that span the study region (Fig. 4). The sites are labeled with respect to those near the Cascadia subduction zone on Vancouver Island (the highest hazard); in the active crustal region in southwestern BC; and in the stable continental interior of eastern BC. BC Hydro as PSHA User PSHA Results Frequency of Exceedance per Year 1.E+00 1.E-01 1.E-02 1.E-03 1.E-04 1.E-05 1.E-06 Site in Stable Continental Interior Site on Vancouver Island Sites in Southwestern BC As an owner of a large portfolio of assets (buildings, 1.E-07 transmission system, generation 0.01 0.1 1 10 facilities, dams), it was BC Hydro s Peak Ground Acceleration (g) intent at the start of the project to establish an in-house capability to implement the PSHA model. In Figure 4. Illustration of the variation in the ground keeping with its role in the past, motion hazard in BC, reflecting the significant such an objective is unique for most differences in the rate of seismicity and tectonics. PSHA project sponsors. To develop this capability, BC Hydro acquired the rights to the PSHA

software that was used in the project and internal staff was trained in the use of the software and the model. Acknowledgments The authors thank BC Hydro for initiating and sponsoring the PSHA study, and thank the Participatory Peer Review Panel (Drs. Carl Stepp, Kenneth Campbell and Kevin Coppersmith) for its guidance. BC Hydro also thanks the late Dr. Allin Cornell for his initial direction and guidance as its Advisory Board Member. The contributions provided by the Geological Survey of Canada, the US Geological Survey and other resource agencies are greatly appreciated. References 1. Cornell CA. Engineering Seismic Risk Analysis. Bulletin of the Seismological Society of America. 1968; 58: 1583-1606. 2. Budnitz RJ, Apostolakis G, Boore DM, Cluff LS, Coppersmith KJ, Cornell CA, Morris PA. Recommendations for Probabilistic Seismic Hazard Analysis: Guidance on Uncertainty and Use of Experts, US Nuclear Regulatory Commission Report NUREG/CR-6372, 2 Volumes. 1997. 3. Kammerer AM, Ake JP. Practical Implementation Guidelines for SSHAC Level 3 and 4 Hazard Studies, US Nuclear Regulatory Agency NUREG-2117, Office of the Nuclear Regulatory Research, Washington, DC. 2012. 4. Lawrence M, McCann M, Ostenaa D, Wong I, Unruh J, Hanson K, Olig S, Clague J, LaForge R, Lettis W, Swan B, Zachariasen J, Youngs R, Addo, K. The BC Hydro SSHAC Level 3 Seismic Source Model. Proceedings of the 10th National Conference on Earthquake Engineering, Earthquake Engineering Research Institute, Anchorage, AK, 2014. 5. Ostenaa D, McCann M, Lawrence M, Youngs R. Embedded Faults: A New Alternative for Incorporating Geologic Data in Seismic Source Zone Modelling. Proceedings of the 10 th National Conference in Earthquake Engineering, Earthquake Engineering Research Institute, Anchorage, AK, 2014. 6. LaForge R, Youngs R, Lawrence M. The Alternative Characteristic (AC) Model as Implemented for the BC Hydro SSHAC 3 Source Model. Proceedings of the 10 th National Conference in Earthquake Engineering, Earthquake Engineering Research Institute, Anchorage, AK, 2014. 7. Atkinson GM. Single-Station Sigma. Bulletin of the Seismological Society of America. 2006; 96: 446-455.