2011 Planning Workshop
- Organizers: Nico Luco, Paul Somerville
- Date: Monday, January 10, 2011 (10:00am-5:30pm)
- Location: SCEC Boardroom, University of Southern California, Los Angeles, CA
- Participants: Approximately 25
- 1 Objective
- 2 Agenda & Presentations
- 3 Participants
- 4 Summary of Outcomes
- 4.1 What ground motion simulations and/or simulation models should the SCEC GMSV TAG validate?
- 4.2 What validation methodologies should the TAG use?
- 4.3 For what engineering application should the TAG validate the simulations?
- 4.4 Should the TAG prioritize the archiving and distribution of simulations?
- 4.5 Where/how should the validation methodologies used by the TAG be implemented?
- 4.6 How should the TAG be organized, at least initially? In other words, how should the TAG operate?
- 5 Recommendations
- 6 References
SCEC is establishing a Technical Activity Group (TAG) focused on Ground Motion Simulation Validation (GMSV) in order to develop and implement testing/rating methodologies via collaboration between ground motion modelers and engineering users. The main purpose of the planning workshop will be to enumerate and prioritize work that should be conducted within the GSMV TAG. SCEC will only be able to fund relatively few projects per year that contribute to the GMSV TAG, so prioritization is particularly important. General ideas for work that could be conducted within the GMSV TAG include:
- Research on important ground motion or structural (e.g., building) response parameters and statistics that should be used in comparing simulated versus recorded seismograms.
- Comparisons of simulated ground motions with empirical ground motion prediction equations, in terms of both median predictions and the variability about them. Note that simulations for the CyberShake Project aim to accurately represent both the median and variability of ground motions.
- Compilation of representative nonlinear structural models of different types for which the responses to simulated versus recorded seismograms can be compared.
- Comprehensive analysis and documentation of the sensitivity of simulated ground motions to model input parameters and their interactions and uncertainties.
- Development of testing and/or rating metrics for simulated ground motions, perhaps considering testing concepts from the Collaboratory for the Study of Earthquake Predictability.
- Implementation of testing/rating methodologies into the SCEC Broadband Strong Motion Simulation Platform.
Agenda & Presentations
|10:00||Welcome and GMSV Technical Activity Group Background ||T. Jordan|
|10:05||Overview of Agenda ||N. Luco|
|What Ground Motion Simulations/Models to Validate?|
|10:15||SCEC Broadband Simulation Platform and Cybershake ||R. Graves|
|10:30||Puente Hills, ShakeOut, 1906, and Hayward Simulations ||B. Aagaard|
|10:45||Summary of Other Ground Motions Simulation Models ||P. Somerville / Y. Zeng|
|What Validation Methodologies? What Engineering Applications?|
|11:30||Goodness-of-Fit Criteria ||K. Olsen|
|11:45||Elastic and Inelastic Response Spectra Properties ||J. Baker|
|12:00||Spatial Correlation of Spectral Accelerations ||P. Bazzurro / N. Jayaram|
|13:15||Attenuation of Spectral Accelerations ||J. Stewart|
|13:30||Tall Building Response ||N. Jayaram|
|13:45||Summary of Other Validation Methodologies/Applications ||N. Luco|
|Initial Priorities for GMSV Technical Activity Group (TAG)?|
|14:30||What Engineering Applications? ||All|
|14:50||What Validation Methodologies?||All|
|15:10||What Ground Motion Simulations/Models to Validate?||All|
|15:50||Archiving/Distribution of GM Simulations?||All|
|16:10||Implementation of Validation Methodologies?||All|
|16:50||Summary of Priorities||N. Luco / P. Somerville|
Brad Aagaard (USGS), Norm Abrahamson (PG&E), Jack Baker (Stanford), Jeff Bayless (URS), Paolo Bazzurro (AIR Worldwide), Jacobo Bielak (CMU), Jorge Crempien (UCSB), CB Crouse (URS), Christine Goulet (PEER), Rob Graves (USGS), Tran Huynh (USC), Nirmal Jayaram (RMS), Tom Jordan (USC), Abbie Liel (Colorado), Pengcheng Liu (US Bureau of Reclamation), Qiming Liu (UCSB), Masha Liukis (USC), Nico Luco (USGS), Phil Maechling (USC), Kim Olsen (SDSU), Danijel Schorlemmer (USC), Paul Somerville (URS), Jonathan Stewart (UCLA), Farzin Zareian (UCI), Yuehua Zeng (USGS)
Could not attend: Ralph Archuleta (UCSB), Greg Beroza (Stanford), Steve Day (SDSU), Art Frankel (USGS), Steve Hartzell (USGS), Curt Haselton (CSU Chico), Swami Krishnan (Caltech), John McRaney (USC), Farzad Naeim (John A. Martin & Associates), Nilesh Shome (RMS)
Summary of Outcomes
What ground motion simulations and/or simulation models should the SCEC GMSV TAG validate?
- Most workshop participants agreed that the TAG should start with simulation models that are on the SCEC Broadband Simulation Platform, validating CyberShake simulations later. (Note: Validation of Cybershake simulations was further discussed at the January 27-28 CyberShake Meeting in Golden, Colorado organized by Ned Field.) The simulation models on the Broadband Platform are relatively reproducible and quality-assured.
- The choice to start with the Broadband Platform is in line with prevalent views amongst the workshop participants to prioritize broadband instead of low-frequency, deterministic simulations (like the CyberShake simulations).
- Some workshop participants felt that validating 1-D simulations is sufficient, whereas others felt that the TAG should focus on 3-D simulations. Most workshop participants agreed that comparisons of corresponding 1-D and 3-D simulations would be valuable. Such comparisons can be made with the Broadband Platform.
- As discussed at the workshop, even if the TAG finds that it needs to limit its geographic scope (e.g. to Southern California only), the expectation is that the results of the validations will be more broadly applicable.
What validation methodologies should the TAG use?
- Most workshop participants agreed that the TAG should start by validating against recorded ground motions from past earthquakes (e.g. the 1994 Northridge earthquake).
- It was proposed at the workshop that the TAG start with the approximately 8 magnitude 6.5 and larger earthquakes that have been recorded in California. Via collaboration with international colleagues, this list could be expanded to include, among others, the Kobe, Izmit, Chi-Chi, Totori, and Darfield earthquakes.
- A majority of workshop participants felt that the TAG should start by using elastic and inelastic response spectra for validation. A broad range of spectral frequencies, from 0.1 to 10 Hz, were of interest to workshop participants. Inelastic response spectra that account for degradation beyond that of bilinear single-degree-of-freedom oscillators were of interest as well. Other relatively simple goodness-of-fit measures, such as Arias duration, were also suggested.
- Some workshop participants, however, felt that the TAG should not wait to validate simulations for the response of at least a few multi-degree-of-freedom nonlinear building (and possibly site response) models. Such validation exercises may identify additional important metrics for validation (e.g. story drift correlations), and how/whether they relate to elastic and inelastic spectra.
- Whatever the goodness-of-fit measure, its spatial distribution should likely be mapped, as discussed by a few of the presenters at the workshop.
- Validation methodologies that compare against ground motion prediction models (GMPM’s) were discussed extensively at the workshop. Some participants did not consider such comparisons to be, strictly speaking, methodologies for validation; nevertheless, they considered them to be useful.
- The consensus of the workshop participants seemed to be that the TAG should make comparisons with GMPM’s, in parallel but with lower priority than the validation against recorded ground motions from past earthquakes.
- Whether via comparisons with GMPM’s or bins (e.g. by source-to-site distance) of recorded ground motions, the means and covariance (standard deviations and correlations) of goodness-of-fit measures are important parts of a validation methodology, as discussed by a few of the workshop presentations.
- Spatial correlation of goodness-of-fit measures (mainly elastic spectral acceleration) was also discussed as an important validation methodology. There was some concern, however, that the spatial coverage of data from past earthquakes may not be sufficient for validation against it.
For what engineering application should the TAG validate the simulations?
- As pointed out in the introductory presentation at the workshop, validation pertains to a “proposed domain of applicability” of the model. In particular, SCEC is interested in the use of ground motion simulations in engineering applications.
- Rather than identifying particular engineering applications, however, the workshop participants focused on the more broadly applicable validation methodologies described above. Those methodologies were considered to be applicable to, among other engineering applications, physics-based probabilistic seismic hazard analysis (via GMPM’s), spatial correlation models (and coherency?), site response analysis (and slope displacements), and nonlinear response history (structural) analysis.
Should the TAG prioritize the archiving and distribution of simulations?
- The consensus of the workshop participants seemed to be that, initially, archiving and distribution of simulations should be prioritized only to the extent that it is needed for the purposes of the TAG. For example, it was thought to be premature to focus on distribution to practicing engineers, although early communication with them on this topic may be important for planning purposes.
- A key question here is what metadata must/should be stored. USGS Data Series 413 serves as an example of metadata for simulations. The PEER NGA Flatfile may also serve as a template for what metadata to store.
- Simulations could be archived and distributed via Websims (Olsen & Ely, 2009) or data centers like the Southern California Earthquake Data Center and the IRIS Data Management Center, which apparently are already interested in the archiving and distribution of validated ground motion simulations. The PEER Strong Motion Database is another option that could be engaged by the TAG.
- Workshop participants also discussed the importance of documenting which inputs to the simulations models are random (aleatory), uncertain (epistemic), or deterministic.
Where/how should the validation methodologies used by the TAG be implemented?
- Like the one validation module that is already part of the Broadband Platform, workshop participants seemed to agree that the validation methodologies used (or developed) by the TAG should also be incorporated into the Broadband Platform.
- Since the Broadband Platform already encompasses computer code written in several different programming languages (e.g. Fortran, C, Matlab), the TAG can use and/or develop validation methodologies in any language.
- The goodness-of-fit measure(s) proposed by Olsen & Mayhew (2010) are ready for implementation of the Broadband Platform (and have been implemented since the workshop).
How should the TAG be organized, at least initially? In other words, how should the TAG operate?
- A specific and detailed organizational/operational plan was not developed by the workshop, but several ideas on how the TAG should proceed were discussed, as summarized below.
- The next step for the TAG is to go to the SCEC Planning Committee with either a joint proposal or individual but coordinated proposals, through the first SCEC4 Collaboration Plan. A second GSMV workshop may be necessary in order to develop this(these) proposal(s).
- Like the composition of the workshop participants, the members of the TAG must include both producers of simulations (ground motion modelers) and users (seismic hazard, building response, and risk modelers).
- One way of engaging the requisite simulation users is via the PEER Ground Motion Selection and Modification (GMSM) Program, which is a group of earthquake engineers and seismologists that conducts coordinated research on appropriate GMSM methods for nonlinear dynamic response simulations. Several of the workshop participants are already members of the GMSM group. Currently they are only funded through their own individual research projects, but have continued to coordinate their work via roughly quarterly meetings.
- Particularly since the Broadband Platform will be used to manage simulations for it, the PEER NGA East project may be interested in supporting the TAG. At the least, the TAG should be coordinated with the NGA-East Simulations Working Group.
- The TAG could link with OpenSees (Open System for Earthquake Engineering Simulation), which is a software framework for simulating the seismic response of structural and geotechnical systems. As an example, inelastic (and elastic) response spectra for validation could be computed using OpenSees.
- Pacific Gas and Electric has provided some funding for development of the Broadband Platform, and hence it may be a source of support of the TAG which workshop participants felt should start by validating simulation models that are on the Broadband Platform.
- Equipped with the priorities identified by the workshop and summarized above, Tom Jordan offered to find resources to support operation of the TAG. NSF Geosciences and NSF Engineering may be a source of support, perhaps via a multi-year joint SCEC-PEER proposal.
Pursuing the workshop suggestions above, four initial efforts are proposed to advance the validation of simulated ground motions:
1) Validation of simulated ground motions for past earthquakes using elastic and inelastic response spectra: It is proposed to compute and compare response spectra (elastic and inelastic) from ground motions recorded during the past earthquake listed in Table 1 below with those simulated for the same recording stations. In addition to means and standard deviations of these response spectra for each spectral frequency, the validations will include statistics related to the shapes of the elastic response spectra and the correlations between elastic response spectral values at multiple periods, both of which are features that have been shown to be important for structural response to ground motions (Baker & Cornell, 2006). The extent to which these validations in terms of elastic response spectra translate to validation in terms of inelastic response spectra will be assessed. Furthermore, any systematic spatial variations of the simulated-versus-recorded comparisons will be investigated, e.g. differences near the earthquake source.
2) Automation of the validation tests developed (above): In order to facilitate validation of other/future simulations, it is proposed to develop software modules for the validation of simulated ground motions for past earthquakes in terms of elastic and inelastic response spectra. An important aspect of this effort will be automated use of OpenSees (Open System for Earthquake Engineering Simulation), which is a software framework for simulating the seismic response of structural and geotechnical systems. By using OpenSees to compute the requisite response spectra, in the future validation in terms of more realistic multi-degree-of-freedom (MDoF) nonlinear systems like those considered below can more seamlessly be added to the proposed validation modules. To demonstrate this, it is also proposed to develop software modules for automated computation of the generalized interstory drift spectrum (Miranda & Akkar, 2006), which employs a continuum model of a multistory building that will be used in the validation tests developed below.
3) Validation of simulated ground motions for past earthquakes for geotechnical systems: While validation in terms of elastic and inelastic response spectra may translate to valid simulated ground motions for response history analysis of some nonlinear geotechnical systems (e.g. site response), it is important to demonstrate any such connections and understand their limitations. Hence, it is proposed to compute and compare the responses of selected geotechnical systems to the same simulated and recorded ground motions for past earthquakes used above for validation in terms of response spectra. Both nonlinear site response and slope displacement analyses will be performed. In comparing the results of these analyses for simulated versus recorded ground motions, the aim is to isolate any differences that are not simply due to differences in the elastic and inelastic response spectra for the two sets of ground motions. One potential source of such differences is ground motion duration, which selected geotechnical systems can be particularly sensitive to (e.g. Bray & Rathje, 1998).
4) Validation of simulated ground motions for past earthquakes for multi-degree-of-freedom (MDoF) nonlinear building systems: Like for the geotechnical systems described above, it is important to verify any connections between validation in terms of elastic and inelastic response spectra and validation for the response of more realistic MDoF models of nonlinear building systems. Hence, it is proposed to compute and compare both generalized interstory drift spectra (described above) and the responses of selected reinforced concrete frame buildings (also computed via OpenSees), again for the aforementioned simulated and recorded ground motions for past earthquakes. Importantly, in the latter case any degradation of stiffness and strength with cycles of motion will be modeled, and hence the building response is likely to be sensitive to more than just the elastic and inelastic response spectra that will initially be used for validation. Moreover, in addition to the more standard drift/displacement and acceleration metrics, energy metrics for building response will be computed. In some cases, more significant differences between simulated (or stochastic “artificial”) and recorded ground motions have been observed for such metrics (Iervolino et al, 2010; Jones & Zareian, 2011).
Table 1. Ground-motion records available for 13 historic California earthquakes of M>=6.
|Year||Location||Mw||# records||R20 (km)*|
|1986||North Palm Springs||6.1||32||40|
* R20 = distance from rupture that contains 20 records
Baker J.W., Cornell C.A. (2006). Correlation of response spectral values for multi-component ground motions. Bulletin of the Seismological Society of America, 96(1):215-227.
Bray, J.D., Rathje, E.M. (1998). Earthquake-induced displacements of solid-waste landfills. Journal of Geotechnical and Geoenvironmental Engineering, 124(3): 242-253.
Iervolino I., De Luca F., Cosenza E. (2010). Spectral shape-based assessment of SDOF nonlinear response to real, adjusted and artificial accelerograms. Engineering Structures, 32(9): 2776-2792.
Jones P., Zareian F. (2011). Seismic response of a 40-storey buckling-restrained braced frame designed for the Los Angeles region. The Structural Design of Tall and Special Buildings. doi: 10.1002/tal.687.