SI2-SSI: A Sustainable Community Software Framework for Petascale Earthquake ModelingThomas H. Jordan (USC), Jacobo Bielak (CMU), Yifeng Cui (UCSD), Kim B. Olsen (SDSU)
S C E Can NSF+USGS center
ScientificSoftwareElements
Users:
Reinsurance
USGS
SCEC
EERCsNEES
RiskManagement
EarthquakeEngineering
Insurance
EmergencyPreparedness
FEMA
EarthScope
Public GTLTera3DPseudo-
dynamic
UCVM
CyberShake
Broadband
SEISM-IO Work-flows
Hercules
AWP-ODC
SORD
SWP
OpenSHA
OpenSees
EarthquakeEngineeringGeoscience
ComputerScience
SEISMSimulation Framework Valid
ation
Fra
mew
ork
Validation Exercises
Horizontal Integration
Vertical Integration
The main goal of this project is to advance and inte-grate scientific software elements (SSEs) devel-oped by SCEC into a sustainable software environ-ment for integrated seismic modeling (SEISM) as a software ecosystem for physics-based seismic hazard analysis (SHA).
SEISM will support the use of HPC by earthquake scientists, engineers, and risk managers to create the simulation-based products needed to improve SHA, and will provide the earthquake modeling community with tools that include model formula-tion, verification, prediction and validation, with a sustainable software engineering framework.
SEISM will accommodate a new generation of time-dependent earthquake rupture models for earth-quake forecasting that will interface with other soft-ware such as OpenSHA and OpenSees to produce more accurate hazard maps and to simulate the seismic performance of structural and geotechnical systems, thus facilitating a “rupture-to-rafters” mod-eling approach to earthquake risk management.
SEISM will incorporate Cyber Science and Engi-neering tools and practices into seismic hazard analysis, and will engage, train and expand a di-verse STEM workforce of geoscientists, computer scientists, and earthquake engineers.
Goals
Impact
SEISM is a collaboration among disciplinary groups in geoscience, computer science, and earthquake engineering. Our work plan is therefore structured around synergistic activities among these three groups. Integration of the simulation framework is being coordinated primarily between the geosci-ence and computer science groups. The validation exercises will be jointly managed by the geoscien-tists and earthuake engineers using a validation framework developed by the latter in cooperation with our software development team.
Work Plan
SEISM will develop and test two main computational platforms (CyberShake and BroadBand) and multiple SSEs (SWP, SORD, AWP-ODC, Hercules, GTL, Tera3D). The two main platforms and their products will be made available across domains. An additional platform, UCVM, will be developed to support the seismic veloc-ity models and meshes/grids necessary for earthquake simulation software integration.
Computational Platforms and SSEs
The CyberShake Platform
The prototype CyberShake hazard model has indi-cated that the standard at-tenuation relations used in probabilistic SHA based on empirical attenuation rela-tions severely underesti-mate the hazard probabili-ties. SEISM will advance the CyberShake platform as a simulation-based ap-proach to seismic hazard analysis. This platform will allow users to access the different layers of hazard information: (1) hazard maps, (2) hazard cirves, (3) disaggregation of hazard data in terms of rupture magnitude and dis-tance, (4) rupture models of earthquakes with the highest hazard to site, and (5) simulated seismograms for particular locations.
The Broadband Platform
The Broadband platform generates 0-10 Hz ground motions using deterministic low-frequency and stochastic high-frequency simulations. The Broadband platform integrates different scientific software elements includ-ing rupture generation, low-frequency deterministic seismogram synthesis, high-frequency stochastic seis-mogram synthesis, and non-linear site effects. These complex scientific codes have been integrated into a system that supports easy on-demand computation of broadband seismograms. The Broadband platform is designed to be used by both geoscientists and engineers with some experience interpreting ground motion simulations.
High-Fidelity Forward Simulation Tools and Results
SEISM includes various HPC applications for high-fidelity forward simulation of rupture dynamics and wave propaga-tion (SORD, AWP-ODC, Hercules). These codes are developed to be highly efficient and scalable and have been tested in multiple verification and validation exercises.
The SEISM Project is supported by the National Science Foundation, award No. 1148493. Additional SCEC team members include: J.W. Baker, G.C. Beroza (Stanford); P. Chen (U. of Wyoming); E.M. Dunham (Stanford); S.M. Day (SDSU); R.W. Graves (USGS); I. Iervolino (U. Naples, Italy); N. Luco (USGS); P.J. Maechling (USC); J.P Stewart (UCLA); R. Taborda (CMU); F. Zareian (UCI). Private and international partners include: Pacific Gas & Electric and the REAKT European project. Scientific publications and additional information about the various software elements and platforms can be found at http://scec.usc.edu/scecpedia and at www.scec.org.
Peak ground horizontal ground velocities (hot colors) and seismograms (white lines) from the M8 simulation of a “wall-to-wall” rupture of the southern San Andreas fault.
Goodness-of-fit validation from a 4 Hz simulation of the Mw 5.4 2008 Chino Hills earthquake in the Greater Los Angeles Region and comparison of the seismograms at a particular location, re-corded (white) and simulated (green).
The GreaterLos Angeles
Region
Pacif ic Ocean
2008Shino HillsEarthquake
Foodness-of-Fit Score4 100 2 6 8
RuptureGenerator(optional)
Low-FrequencySynthesis
High-FrequencySynthesis
Non-LinearSite Response
ExtractResponseSpectra
Verificationand Validation
(optional)
RuptureDescription
(SRF)
DeterministicSimulation
ObservedSeismograms(or alternative
simulation)
StochasticMethod
Longitude
Latitude−118.70−118.65
−118.60−118.55
−118.50−118.45
34.20
34.25
34.30
34.35
−20
−16
−12
−8
−4
0
0
50
100
150
200
250
300Slip (cm)
Cou
nt
0
20
40
60
0 1 2 3 4 5 6 7 8 9 10Score
0.1 1 100
100
200
300
0.1 1 100
100
200
0.1 1 100
50
100
150
0.1 1 100
10
20
30
Period (s) Period (s)
Sa (c
m/s
2 )Sa
(cm
/s2 )
Time (s)
CyberSHakePlatform
100
10-1
10-2
10-3
10-4
10-5
10-6
10-2 10-1 100
Prob
abilit
y R
ate
(1/y
r)
PGA (g)
1. Hazard Map
2. Hazard Curves
3. Hazard Disaggregation
4. Rupture Model
5. Seismograms
%C
ontri
butio
n
Rupture Distance (km)
40 50 60 70
0 10 20 30
80 90 100 110
20
16
12
8
4
0
Magnitud
e
5
6
7
8
9
Refinementand Partitioning
Meshingand Gridding
SourceGeneration
Solving
ForwardSimulation
1 00 1
Output Data
4D WavefieldStations, Planes, Volume
Model Formulation
SimulationParameters
SeismicVelocity Models
Top Related