CyberShake Workplan

CyberShake Goal:

PSHA-3DWP hazard curve anywhere in California and an expression of uncertainties for any hazard curve. Create basis for operational 3DWP ground motion forecasting system. We plan to develop a general approach usable anywhere in the world.

USGS CyberShake Meeting 26 Jan 2011

Meeting in Golden discussed scientific and technical issues about UCERF3.0 and CyberShake. Development priorities resulting from the meeting:

Proposed Development Phases

Due to the very large scale of the proposed 1Hz state-wide Hazard Map calculation, we plan to coordinate the first production CyberShake California 1.0Hz map on the official release of UCERF3.0 in June 2012.

1. CyberShake 1.0 production run competed in Oct 2009 used NCSA, TACC, USC HPCC. 3D wave propagation and other scientific software by Robert Graves with Globus, Condor DAGMan, and ISI Pegasus.
   1. Add historical validation events (Northridge, Landers, Chino Hills, Loma Prieta)
   2. Add precariously balanced rock sites
2. CyberShake 1.1 Graves and Pitarka (2010) rupture generator
   1. Can be re-run without regenerating SGTs.
3. CyberShake 1.2 recalculates map using CVM-H 11.2 velocity model
   1. Generate new SGT's for all seismic stations
   2. Use both old and new rupture variations
4. CyberShake 2.0 uses multiple CVM's, a new extended ERF generator, with state-wide capabilities, at 1Hz.
   1. 1.0 Hz AWP-ODC
   2. SRF
   3. CVM
   4. CVM Evaluation Capabilities
   5. Data Product Distribution
   6. Automated Testing Framework
5. CyberShake 3.0 is based on UCERF3.0, SCEC Extended ERF, and Unified California Velocity Model UCVM.

Essential Research Developments

1. Development of next generation extended Earthquake Rupture Forecast (EERF). Initial development using UCERF2.0 and transition to UCERF3.0 when available.
2. Development of state-wide 3D velocity model. Proposed approach is embedding high resolution background models into lower resolution background models.

• Extremes of UCERF 2.0 ruptures which need to be included in UCVM (kml)

1. Automated Testing Framework (ATF) to support comparative testing of CyberShake codes, regression testing of important modules, and CVM evaluation system.
2. Development of highly scalable SGT codes. Modifications to Graves code are required to achieve scale. Modifications to Olsen code are required to support SGT creation and seismogram synthesis

Software Development Milestones

1. Additional CyberShake 1.1 curves (ongoing)
2. Additional CyberShake 1.2 curves (ongoing)
3. Testing framework for CyberShake on new platforms (5/11)
   1. Determine testing framework to use
   2. Determine components to place under test
   3. Identify reference calculations
   4. Create tests for multiple systems
4. Olsen 0.5 Hz forward calculation which matches sample synthetics (5/11)
   1. Identify Graves SGT parameters
   2. Select sample synthetics for matching test (possibly the same ones as for the CyberShake tests)
   3. Determine if Olsen reciprocity (Po's code?) or Graves reciprocity should be used and integrate appropriately
5. Olsen 0.5 Hz hazard curves which match Graves 0.5 Hz hazard curves (6/11)
   1. Interface Olsen SGTs to SGT workflow
   2. Interface Olsen synthetics to post-processing workflow
6. 1D CyberShake curve functionality statewide (6/11)
   1. Integrate Rob's 1D codes
   2. Use statewide EERF
      1. Implement Cascadia solution
      2. Determine if UTM zones are used in calculation; if so, implement a fix
   3. Create 1D California CVM
   4. Perform verification using known 3D results and ARs
7. 3D parallelization of Graves SGT code (7/11)
   1. Provide reference solutions to Yifeng
8. Olsen 1.0 Hz hazard curves (8/11)
   1. Verify using known 0.5 Hz results
9. Graves 1.0 Hz hazard curves (8/11)
   1. Verify using known 0.5 Hz results
10. Olsen 1.0 Hz hazard curves which match Graves 1.0 Hz hazard curves (9/11)
11. Hazard curves using UCVM (10/11)
12. 1.0 Hz hazard curves on Blue Gene architecture (12/11)
    1. Port codes to BG/P
    2. Successfully run CyberShake tests on BG/P

Software Development Priorities

1. Define Reference Calculations for forward calcs, reciprocity, EEER generation, and hazard curves
2. Create state-wide Extended ERF (with UCERF2.0)
3. Create (1D or low res) California CVM
4. Show selection of EERF variations specific to a site for state-wide sites
5. Show merging of results in flat box, in spherical state-wide context
6. Decide on a database technology

CyberShake Data Management (Primary Calculation)

1. Amplitudes, Waveforms, for any event in catalog
2. Maps
3. UCERF3.0 (ERF)
4. Fault geometries for all ruptures in ERF
5. UCEERF3.0 (Extended ERF)
6. UCVM (Unified California Velocity Model)
7. Geotechnical data state-wide
8. Location of every sites.
9. Closest curve to any geographical site
10. GTL profile every site
11. Vertical profile for every site
12. All velocity meshes in simulation regions
13. All ruptures in simulation region
14. PSHA Maps (intensity measures include PGV, PSA2.0s, SA3,0,SA5,0,SA10.0)
15. PSHA Hazard Curves
16. Single component hazard curve
17. Disaggregrated Curves to identify ruptures based on any parameters
18. Rupture variations
19. Fault Geometries
20. Fault Maps
21. Elevation
22. Parameter plots (initial stress, hypocenter, final slip, slip rate, Supershear), and animations, of every SRF.
23. Hazard curve to rupture variation set to sorted amplitudes (by peak intensity measure, by distance) to amplitude values to seismograms for amplitudes to rupture variation to slip on rupture to velocity model used.
24. Ask questions about a site. What was simulation region, what was velocity model, what was rupture variation selection, what was peak amplitudes, what was different complements, what are smallest amplitudes
25. SGT established as authorized data set for use in Operational Earthquake Forecasting

Value-Added CyberShake Data Products:

1. CyberShake 1Hz in 1D.
2. PAGER for any CyberShake event
3. Building response (18 story steel frame building) for full suite of rupture variations for specific events. Try for all Northridge variations.
4. Short-term ERF probabilities Adjustor (PA)
5. ShakeMap for any Rupture
6. Rupture Library:
7. Seismogram Library:
8. Basin entrainment study
9. Geotechnical layer (GTL) evaluation through comparison to borehole recordings
10. For each site for each event:
    1. 1Hz seismograms
    2. Broadband seismograms
    3. Peak Amplitudes
    4. Duration of Shaking
    5. EEW analysis for network
    6. Reference building response
    7. Scenario ShakeMap
    8. Ground Motion animation
    9. Rupture animation

Software Infrastructure Issues:

1. Multiple wave propagation and post-processing codes
2. Check AWP-ODC implementation with a 1Hz version of Rob Graves code.
3. Show ability of both codes to get same answer at 0.5Hz and 1.0 Hz

EERF Issues

1. Automated conversion into and out of SRF format for Olsen code.
2. Background seismicity
3. Establish sufficient variability
4. Introduction of rise-time variability
5. Introduction of linked events

Site specific information

1. Sites in background regions

Validation

1. Comparison of CyberShake amplitudes to empirical amplitudes. Does cybershake reproduce the empirical distribution?
2. Validation of UCEERF
3. Validation of UCVM
4. Amplitude (and distribution) of CyberShake amps by distance compared against empirical attenuation relationship.
5. Spectral content of seismograms compared to observed seismograms
6. GOF for seismograms from rupture variations of historical earthquakes
7. Frequency content of rupture sources
8. Comparison of hazard curves to precarious rocks sites
9. Comparison of seismograms to “unusual records” including I-10/I-215, Northridge garden center, and Santa monica and I-14 fallen freeways

Calculation Optimizations:

1. Create EERF:
2. Create set of statewide rupture variations
3. Create set sites
4. Map rupture variations to sites
5. Define minimum set of velocity meshes
6. Associate velocity mesh for each site
7. Combine seismogram extraction with peak amplitude calculation to output seismograms and peak amplitudes

Visualization Issues:

1. Need maps of ruptures and rupture variations.
2. Need maps showing simulation volume and ruptures in the simulation volume.

Data Management Issues

1. Base large-scale data management and computation on virtual data management concepts within Pegasus including separation between Logifical File Name (LFN) and Physical File Name (PFN).
2. Integrate a digital object identifier (DOI) scheme to identify specific data products. DOI's link to LFN's and then to PFN's.
3. Track DOI's through computational data product dependency chart identifying types of data and a data hierarchy showing data products that are derived.

Query Capabilities Needed

1. Create a Simulation Description Request Format:
2. Data Request Format:
3. Name:
4. User:
   1. Data Builder
   2. Parser
   3. Verifier
   4. Reducer
   5. Annotator
   6. Packer
   7. Notifier

Science Risks:

1. Need to deliver a seismic hazard curve and an expression of its uncertainty
2. Never done 1hz curve
3. Agreeing on state-wide cvm
4. Handling background seismicity in ucerf3
5. Handling fault to fault ruptures allowed in ucerf3
6. Agreeing on variation in extended erf
7. Agreement on final ucerf3.0 extended erf

Technical Risks:

1. Extensive data management issues including presentation, querying, and access
2. Need for both relational and no-relational data management and retrieval.
3. Workflows on biggest hpc system
4. Formalized software testing environment
5. State-wide cvm development
6. State-wide erf development

Organizational Risks:

1. Assurances that we will get computer time and data management to support the work in 2012 since we are asking for 700MSU/3000MSU total available.
2. Review processes within SCEC for giving computational green light
3. All CME groups work should contribute to or build on CyberShake results
4. Maintaining CME staff at USC and other organizations.