1. Edward (Ned) Field USGS, Pasadena plus Tom Jordan, Nitin Gupta, Vipin Gupta, Phil Maechling, Allin Cornell, Ken Campbell, Sid Hellman, & Steve Rock OpenSHA Community Tools for Seismic-Hazard Analysis

2. (1)SHA needs a more physics-based approach to modeling. (2)Lack of consensus on how to construct more physics- based models means we’ll have multiple options. (3) All viable models need to be considered for “proper” SHA (SSHAC Report, 1995). (4)SHA therefore needs a computational infrastructure capable of handling a potentially great number of arbitrarily complex models (a “Community Modeling Environment”). Status of Seismic Hazard Analysis (SHA)

3. What is SHA? Goal of Seismic Hazard Analysis: The probability that some “Intensity-Measure Type” (e.g., PGA, SA) will exceed a specified “Intensity-Measure Level” (e.g., 0.5 g)

4. What is SHA? (1)Earthquake-Rupture Forecast (ERF) Probability of all possible fault-rupture events (M≥~5) for region & time period (2)Intensity-Measure Relationship (IMR) Gives IM exceedance probability at a site for a given fault-rupture event Attenuation Relationships (traditional) (no physics) Full-Waveform Modeling (developmental) (more physics) Two main model components:

5. Why SHA needs more physics: (1)Earthquake-Rupture Forecast Probability of all possible fault-rupture events (M≥~5) for region & time period Two main model components: But, others see time-dependent effects and interactions: (Stein & Others) However, there is no consensus on how to build these types of models. Thus, SCEC has a working group (RELM) that is developing a variety of viable physics-based models (see www.RELM.org). The model used in our National Hazard Maps assumes that each earthquake rupture is completely independent of all others (including the previous one at the exact same location).

6. Why SHA needs more physics: (2)Intensity-Measure Relationship Gives IM exceedance probability at a site for a given fault-rupture event Attenuation Relationships (traditional) (no physics) Two main model components: Lack of physics can lead to non-physical results (e.g., PGA increasing indefinitely with exposure time). Inherent limits with respect to accuracy (SCEC Phase III report).

7. Why SHA needs more physics: (2)Intensity-Measure Relationship Gives IM exceedance probability at a site for a given fault-rupture event Full-Waveform Modeling (developmental) (more physics) Two main model components: Potentially more accurate. However, still developmental due to limited structural knowledge and computational demands. SCEC is working on this (e.g., Pathway II of ITR collaboration).

8. Why SHA needs more physics: (2)Intensity-Measure Relationship Gives IM exceedance probability at a site for a given fault-rupture event Attenuation Relationships (traditional) (no physics) Full-Waveform Modeling (developmental) (more physics) Two main model components: For now, NGA is developing a variety of these (no consensus here either), including constraints from simulations (hybrids).

9. (1)SHA needs a more physics-based approach to modeling. (2)Lack of consensus on how to construct more physics based models means we’ll have multiple options. Status of Seismic Hazard Analysis (SHA) RELM NGA e.g., SCEC ITR “Pathway II”

10. (1)SHA needs a more physics-based approach to modeling. (2)Lack of consensus on how to construct more physics based models means we’ll have multiple options. (3) All viable models need to be considered for “proper” SHA (SSHAC Report, 1995). Status of Seismic Hazard Analysis (SHA)

11. What’s the problem getting multiple, physics-based models into the analysis (why hasn’t it happened)? 1) Geophysicists are comfortable building models and making predictions, but uncomfortable recommending use of them until they’re properly tested and validated. an exception: (Stein & Others) critics say this is premature (we don’t yet know enough)

12. What’s the problem getting multiple, physics-based models into the analysis (why hasn’t it happened)? 1) Geophysicists are comfortable building models and making predictions, but uncomfortable recommending use of them until they’re properly tested and validated. But practitioners can’t wait. They have to make informed decisions today, and need to use whatever models are available.

13. What’s the problem getting multiple, physics-based models into the analysis (why hasn’t it happened)? 1) Geophysicists are comfortable building models and making predictions, but uncomfortable recommending use of them until they’re properly tested and validated. an exception: (Stein & Others) If you don’t like this, produce a competing alternative. (RELM)

14. Another problem in getting multiple models into the analysis … Geophysicists generally don’t have the time to learn and conduct end-to-end SHA calculations (e.g., they’d need to combine their rupture forecast with some Intensity-Measure Relationship), & … SHA practitioners generally don’t have time to learn and implement physics based models. an example: (Stein & Others)

15. So, to improve SHA we need to: allow/encourage scientists to create their own physics-based models without making additional demands on their time, and enable practitioners to easily use these models without having to understand or implement every detail of the model.

16. So, to improve SHA we need to: allow/encourage scientists to create their own physics-based models without making additional demands on their time, and enable practitioners to easily use these models without having to understand or implement every detail of the model. RELM OpenSHA SCEC ITR Collaboration

17. (1)needs more physics. (2)we’ll have multiple options. (3) “proper” SHA requires all options. (4)SHA therefore needs a computational infrastructure capable of handling a potentially great number of arbitrarily complex models (a “Community Modeling Environment”). Status of Seismic Hazard Analysis (SHA)

18. A framework where any arbitrarily complex (e.g., physics based) SHA component can “plug in” for end-to-end SHA calculations. Hazard Calculation IntensityMeasure Type & Level (IMT & IML) Intensity- Measure Relationship List of Supported Intensity-Measure Types List of Site-Related Independent Parameters Earthquake-RuptureForecast List of Adjustable Parameters Site Location List of Site- Related Parameters Prob(IMT≥IML) Time Span OpenSHA: object oriented free, open source platform ind. web/GUI enabled Java (or wrapped code) Evaluated & Validated Any IMR or ERF can be plugged in

19. IntensityMeasure Type & Level (IMT & IML) Rupture n,i Magnitude Probability Ave. Rake Rup. Surface Hypocenter Param. List Intensity-Measure Relationship List of Supported Intensity-Measure Types List of Site-Related Independent Parameters Earthquake- Rupture Forecast Source 1 Source 2 Source i Source I Time Span Source i Rupture 1,i Rupture 2,i Rupture n,i Rupture N,i Site Location List of Site- Related Parameters OpenSHA … … … … Conditional probability of exceedance Rupture probability

20. Intensity-Measure Relationship List of Supported IMTs List of Site-Related Ind. Params IMT, IML(s) Site(s) Rupture Attenuation Relationships Simulation IMRs exceed. prob. computed using a suite of synthetic seismograms Vector IMRs compute joint prob. of exceeding multiple IMTs (Bazzurro & Cornell, 2002) Multi-Site IMRs compute joint prob. of exceeding IML(s) at multiple sites (e.g., Wesson & Perkins, 2002) Various IMR types (subclasses) Gaussian dist. is assumed; mean and std. from various parameters

21. SHA Models Implemented: OpenSHA: Intensity-Measure Relationships (Attenuation Relationships) Boore et al. (1997) Abrahamson & Silva (1997) Campbell (1997) Sadigh et al. (1997) Field (2000) Abrahamson (2000) Campbell & Bazorgnia (2003) ShakeMap (2003) Earthquake Rupture Forecasts PEER Area PEER Non-Planar Fault PEER Multi-Source PEER Logic Tree Poisson Fault ERF USGS/CGS (1996) STEP So. Cal. (real time) (2003) STEP Alaska Pipeline (2003) WGCEP (2002)

22. Applications Available: OpenSHA: 3) Hazard Map Data Calculator decoupled because (3) typically takes hours or days 2) Scenario ShakeMap Calculator 4) Hazard Map Plotter 1) Hazard Curve Calculator

23. Issue 1: Who is going to host the multiple (perhaps physics- based) models as well as the databases they depend upon, especially since revisions/updates are inevitable? Answer: Developers will host their own models and data resources on their own computers as “Web Services” (enabling run-time access over the internet). (the SCEC ITR Collaboration is helping here)

24. Demos

25. How Has the ITR Collaboration Helped? Issue 1: Who is going to host the multiple, perhaps physics- based models, as well as the databases they depend upon (especially since revisions/updates are inevitable)? Answer: Developers will host their own models and data resources on their own computers as “Web Services” (enabling run-time access over the internet).

26. How Has the ITR Collaboration Helped? Presently Implemented Web Services: 1.SCEC Community Velocity Model (for setting site types) 2.GMT Mapping Service (for making hazard maps) 3.Earthquake Rupture Forecasts: USGS/CGS 1996 Forecast STEP So. Cal. Forecast WGCEP-2002 Forecast (as wrapped Fortran Code)

27. How Has the ITR Collaboration Helped? Web Services make our applications lightweight & platform independent (e.g., users don’t have to install GMT) Web Services put the maintenance onus directly on the host

28. How Has the ITR Collaboration Helped? Issue 2: Hazard map calculations typically take several hours for just one set of models, but “proper” SHA requires that we explore all possible model combinations (or all parameter settings within a model). Answer: GRID computing (1st test: 7.5 hrs --> 30 minutes).

29. The End