Working Group: Ground Motion Prediction (GMP)
GROUP LEADERS

PAST RESEARCH RESULTS

MEETINGS & WORKSHOPS

Item 1
Item 2 Item 3 
RELATED RESEARCH

Item 1
Item 2 Item 3 
Research Objectives
The primary goal of the GroundMotion Prediction focus group is to develop and implement physicsbased simulation methodologies that can predict earthquake strongmotion waveforms over the frequency range 010 Hz. Source characterization plays a vital role in groundmotion prediction. At frequencies less than 1 Hz, the methodologies should deterministically predict the amplitude, phase and waveform of earthquake ground motions using fully threedimensional representations of Earth structure, as well as dynamic or dynamically compatible kinematic representations of fault rupture. At higher frequencies (110 Hz), the methodologies should predict the main character of the amplitude, phase and waveform of the motions using a combination of deterministic and stochastic representations of fault rupture and wave propagation. Note: the GMP focus group also shares interests with the GMSV TAG (Earthquake Engineering Implementation Interface, EEII) and CME (Special Project)  consult these sections for additional GMPrelated research priorities.
Research Priorities
 Developing and/or refining physicsbased simulation methodologies, with particular emphasis on high frequency (110 Hz and higher) approaches. This work could include implementation of simulation methodologies onto the Broadband Simulation Platform, or implementation of more efficient approaches in wave and rupture propagation schemes (in collaboration with CME), allowing accurate simulation of higher frequency ground motion in models with lower seismic wave speeds (e.g. in sedimentary basins). Determine spectral and spatial limits for simulating deterministic highfrequency wave propagation.
 Waveform modeling of past earthquakes to validate and/or refine the structure of the Community Velocity Models (CVMs) (in collaboration with USR). This includes exploration and validation of the effects of statistical models of structural and velocity heterogeneities on the ground motion, the significance of the lowest (Swave) velocities as frequencies increase, the significance of including geotechnical layers (GTLs) in the CVMs, and development and validation of improved (possibly frequencydependent) attenuation (intrinsic or scattering) models in physicsbased simulations (in collaboration with USR). Quantify uncertainty in the CVM structure and its impact on simulated ground motions. Note that the Central California Seismic Project (CCSP, see below) targets this goal specifically for Central California.
 Develop and implement simulation methods for the modeling of bending faults and multisegment ruptures. The highest priority need is for kinematic rupture generators for implementation on the Broadband Platform (BBP). Proposals are requested for 1) including the software modeling capability itself and 2) scientific research (e.g., analysis of dynamic rupture modeling on multisegmented faults) to inform input parameters such as the timing of the i^{th} segment rupture, moment distribution on segments and so on (see CME section on this RFP for related efforts).
 Investigate the importance of including 3D basin effects on ensemble averaged longperiod ground motions on the BroadBand Platform, e.g., by comparing ensemble averages of longperiod (<~1Hz) ground motions computed in 1D and 3D crustal models for events included in the GMSV. Develop and implement methods for computing and storing 3D Green's functions (GFs) for use in the Broadband Platform. Proposals for both source and sitebased GFs are solicited (see CME section on this RFP for related efforts).
 Develop and implement new models or implement existing models for frequencydependent site effects into the SCEC BroadBand Platform (site effects module). Because sitespecific profiles are rarely available for large scale simulations, the priority will be given to models that can work with generic site profiles or that use simplified site factors (e.g. empirical Vs30based factors for example). Models that require a site profile as input will also be considered. The site effects models are to be applied so as to produce time series that include site effects.
 Incorporate offfault plasticity into physicsbased ground motion simulation methodologies, quantify uncertainties, and validate the effects using observations from large earthquakes.
 Development of more realistic implementations of dynamic or kinematic representations of fault rupture, including simulation of higher frequencies (up to 10+ Hz). Possible topics include simulation of dynamic rupture on nonplanar faults and studying the effects of fault roughness on the resulting synthetic ground motion, and development of kinematic representations based on statistical models constrained by observed and/or dynamic ruptures. This research could also include the examination of current sourceinversion strategies and development of robust methods that allow imaging of kinematic and/or dynamic rupture parameters reliably and stably, along with a rigorous uncertainty assessment. Close collaboration with the Technical Activity Group (TAG) on Source Inversion Validation (SIV) is encouraged. Construct Equivalent Kinematic Source (EKS) models that approximate the effects of nearfault nonlinearities in a linear scheme and test the EKS model in CyberShake. Projects that involve dynamic earthquake rupture simulations should involve preliminary code testing using benchmarks developed by the Dynamic Rupture Code Verification Technical Activity Group (TAG).
 Verification (comparison against theoretical predictions) and validation (comparison against observations) of the simulation methodologies with the objective to develop robust and transparent simulation capabilities that incorporate consistent and accurate representations of the earthquake source and threedimensional velocity structure. Compare and assess engineering metrics in ground motion validation. Comparison of synthetic ground motions from deterministic and stochastic approaches to data for overlapping bandwidths. Close collaboration with the Technical Activity Group (TAG) on Ground Motion Simulation Validation (GMSV) is encouraged.
Recent Results
The primary goal of the GroundMotion Prediction focus group is to develop and implement physicsbased simulation methodologies that can predict earthquake strongmotion waveforms over the frequency range 010 Hz. Both media and source characterization play a vital role in groundmotion prediction and are important topics for GMP.
This past year’s accomplishments include:
Figure 1. Pwave velocity model in 3D view obtained from ambient seismic waveﬁelds. The color illustrates the fractional ﬂuctuation of Pwave velocities. The grey area shows the poorly resolved area according to the ray coverage. The black dots are the location of the stations projected at the depth of the horizontal slice. The red dot in the inset shows the location of the survey. From Nakata and Beroza, 2015. 
Figure 2. Fourier amplitude as a function of distance centered at 0.25 Hz for a simulation of the 2008 M5.4 Chino Hills, Ca, earthquake, with constant Q and frequencydependent Q. Dots depict values for individual stations and lines depict a 5point moving average. From Withers et al., 2015. 
Figure 3. Elastoplastic reduction of peak ground velocity with respect to a viscoelastic conditions from dynamic simulation (fmax = 2 Hz) of a M 7.8 earthquake rupturing the southern segment of the San Andreas fault. 
 Withers et al. incorporated frequencydependent Q into AWPODC as a powerlaw, and demonstrated the effects using realistic parameters for the Chino Hills earthquake.
 Lozos et al. simulated rupture on the northern San Jacinto fault using complex fault geometry with step overs, and a 3D velocity model. The results were combined with highfrequency scattering functions to generate broadband synthetics. The broadband synthetics were found to be in good agreement with the presence of precariously balanced rocks and leading GMPEs.
 Graves and Pitarka characterized kinematic ruptures for ground motion simulation of shallow crustal earthquakes, including shallow and deep 'weak' zones and mapping the effects of perturbations to the fault surface.
 Baker derived a predictive model for fling period and amplitude and compared to existing models. They found that ground motion simulations provided a rich and reliable data source for fling step, indicating an additional engineering use case for simulations. The work also validated the ability of simulations to predict fling in conditions not well captured by empirical data sources.
 Bradley et al. developed a new 3D seismic velocity model of Canterbury, New Zealand. The model explicitly represents the Canterbury sedimentary basin, and other significant geologic horizons, which are expected to have important implications on observed ground motions.
 Archuleta illustrated the undesired effects of rapid amplitude decay with distance of highfrequency (HF) synthetic ground based on 1D crustal velocity structures. He showed a simple solution to this problem by separating the wave propagation problem into a simplified single layer on top of a halfspace for the HF portion of ground motion and a more realistic 1D multilayer model for the lowfrequency portion of ground motion.
 Holden and Gerstenberger conducted broadband ground motion simulations using a suite of moderately sized aftershocks (M5.3+) from the Canterbury sequence. They used these simulations to investigate the sensitivity of near field ground motions to key engineering parameters including stress drop and rupture details such as velocity, directivity and slip distribution. Results show that adoption of parameters derived from spectral inversions of the strong motion dataset and method provides an improved and robust fit to the observed data, emphasizing the need for regionspecific considerations and the implications this has for GMPEs.
 Beroza and coworkers worked on using ambient seismic field data to explore amplification in urban Tokyo. They used a combination of 375 HiNet deep borehole seismometers across central Honshu as virtual sources and 296 seismic stations of the MeSONet work shallowborehole seismometers within the basin as receivers to map the basin impulse response. They found a linear relationship between vertical ground motion and basin depth at periods of 2 – 10 seconds that could be used to represent 3D basin effects in ground motion prediction equations. They also found that the strength of basin amplification depends strongly on the direction of illumination by seismic waves.
 Shaw and Jordan presented a statistical description of finescale velocity structure in the sedimentary basins of southern California that is intended to support high frequency ground motion simulations for earthquake hazards assessment. They defined the variability in both Vp and Vs, and established vertical and horizontal correlation lengths for finescale velocity structures using wells across the basin as well as in tightly clustered oil fields.
 Roten et al. continued to examine the effects of elastoplastic rheology on ground motions. Previous nonlinear simulations of the ShakeOut scenario based on a kinematic source have suggested that plastic yielding in the fault damage zone may reduce ground motion levels in the Los Angeles basin (LAB) by 30  70 % with respect to linear solutions. New simulations of spontaneous rupture were carried out on a planar, vertical fault roughly following the surface trace of the southern San Andreas fault (SAF) along ~250 km between Indio and Lake Hughes. Because the source in these simulations is fully dynamic, these simulations are comparable to benchmarks TPV26 and TPV27 of the SCEC/USGS dynamic rupture code verification exercise, which have been used to verify the AWPODC FD code with plasticity against several other FD and FE methods. These new simulations confirm that longperiod (< 1 Hz) peak ground velocities in the LAB would be reduced by up to 50% if sedimentary and crustal rocks are assumed to be nearly cohesionless. However, the dynamic simulations also show that PGVs in the LAB may still exceed 2 m/s if the strength of crustal rocks and sediments is very high (> 10 MPa). This result indicates that ground motions are more sensitive to the strength of crustal rocks than indicated by previous dynamic simulations, and highlight the need to better constrain the friction angles and cohesions used in such nonlinear simulations of dynamic rupture and wave propagation.
 Nakata and Beroza developed randomfield model representations of a 3D Pwave velocity model under Long Beach, CA, estimated from densearray recordings of the ambient seismic wavefield. They find that a von Karman model fits the imaged velocity model best, with horizontal and vertical correlation lengths of 0.51 km and 0.1 km, respectively, and a Hurst number of 0.040. They validate their results by showing that their model accurately predicts the observed decay of scattered waves in the coda of a nearby earthquake.
 Using noise correlation measurements from the Long Beach Array, processed to maintain relative amplitude information, Tsai et al. produced maps of surfacewave ground motion amplification over a range of frequencies from 0.67 Hz to 2.0 Hz. These maps show that ground motion site amplification can vary by a factor of 4 over distances as short as a few hundred meters, throughout the city of Long Beach, CA. The spatial amplification patterns are generally consistent with those that would be predicted from shallow velocity anomalies, but provide direct measures of amplification and are therefore more robust than amplification computed indirectly from velocity structure.
 Dunham et al. performed 2D dynamic rupture simulations on rough faults in heterogeneous media to determine the relative importance of source complexity and scattering in destroying coherence of the highfrequency seismic wave field. Their simulations demonstrate that random elastic heterogeneity of the offfault material, at levels representative of the crust, have only minor influence on the rupture process. Fluctuations in slip and rupture velocity are instead controlled by complex fault geometry. This conclusion is expected to carry over to 3D. An additional result of this study was that the effects of scattering became appreciable only beyond a few kilometers from the fault. At closer distances, incoherent highfrequency ground motion was dominated by source complexity. This result will likely change in 3D, and Dunham's group has developed a 3D version of their rupture dynamics code to address this problem.
SCEC Broadband Platform Validation Exercise and SRL Focus Section
SCEC has completed phase 1 of its Broadband Platform (BBP) groundmotion simulation exercise, evaluating the potential applications for engineering of the resulting 0.01–10 s pseudospectral accelerations (PSAs) generated by five different methods. The exercise included part A, in which the methods were evaluated based on the bias of simulation results to observations for 12 wellrecorded historical earthquakes: 7 in the western United States, 2 in Japan, and 3 in the eastern United States/Canada. In addition, part B evaluated simulation results for Mw 5.5, 6.2, and 6.6 scenarios at 20 and 50 km from the fault. The methods were assessed based on the bias of the median PSA for the 12 events in part A and on a specified acceptance criterion compared with Next Generation AttenuationWest (NGAWest) groundmotion prediction equations (GMPEs) in part B. The results were evaluated by the bias of mean PSA from simulations using 1D velocity models with average shearwave velocity in the upper 30 m of 863 m/s with respect to recorded data corrected for site effects. Nine articles describing the scientific and technical accomplishments were published in a focus section of the January/February 2015 issue of Seismological Research Letters.
SCEC HighFrequency Ground Motion Validation Exercise
As part of SCEC’s HighF research initiatives, verification and validation of deterministic ground motion prediction for the 2014 M5.1 La Habra, CA, earthquake is underway. Three codes currently participate in the comparisons, namely AWPODC and AWPRWG (4thorder finite difference, FD) and Hercules (2ndorder finite elements, FE). The exercise uses a point source with mechanism derived from strongmotion data and a sliptime history obtained from a dynamic roughfault model with frequency content up to 5 Hz. The areal extent of the simulation region is 180 km x 135 km, with a target depth of 62 km. The model covers the entire greater Los Angeles basin and other structural features in its vicinity. The verification has progressed in incremental steps from a simple halfspace model via a smooth 1D crustal model, to ongoing efforts involving 3D crustal variation and a minimum Swave velocity of 500 m/s. Comparisons between codes have been made with lossless and frequencyindependent anelastic attenuation, with tests exploring the significance of frequencydependent Q. Results from the verification exercise at the various complexity levels have allowed to identify the numerical parameters necessary for the codes to yield synthetics with a satisfactory level of agreement. Current efforts include verification and validation in a 3D volume of the CVMS4.26, where strong motion data is available at 350+ stations within the model region. The simulations have primarily been carried using parallel processing on NCSA Blue Waters.
Exploring Basin Amplification Sensitivity Using the Ambient Seismic Field
Denolle et al. have studied Kanto Basin amplification by exploiting the information carried by the ambient seismic field. They used 375 HiNet deep borehole seismometers across central Honshu as virtual sources and 296 seismic stations of the MeSONet work shallowborehole seismometers within the basin as receivers to map the basin impulse response. They find a linear relationship between vertical ground motion and basin depth at periods of 2 – 10 seconds that could be used to represent 3D basin effects in ground motion prediction equations. Moreover, they find that the strength of basin amplification depends strongly on the direction of illumination by seismic waves.
Select Publications
 Anderson, J.G. (2015). The Composite Source Model for Broadband Simulations of Strong Ground Motions, Seismological Research Letters, January/February 2015, v. 86, p. 6874, First published on December 17, 2014, doi:10.1785/0220140098. SCEC Contribution 1997
 Atkinson, G.M., and K. Assatourians (2015). Implementation and Validation of EXSIM (A Stochastic Finite‐Fault Ground‐Motion Simulation Algorithm) on the SCEC Broadband Platform, Seismological Research Letters, January/February 2015, v. 86, p. 4860, First published on December 17, 2014, doi:10.1785/0220140097. SCEC Contribution 6067
 Bowden, D.C., V.C. Tsai, and F.C. Lin (2015), Site amplification, attenuation, and scattering from noise correlation amplitudes across a dense array in Long Beach, CA. Geophys. Res. Lett., 42, 13601367. SCEC Contribution 6068
 Bydlon, S. A., and E. M. Dunham (2015), Rupture dynamics and ground motions from earthquakes in 2D heterogeneous media, Geophysical Research Letters, 42(6), 17011709, doi:10.1002/2014GL062982. SCEC Contribution 2064
 Crempien, J.G.F., and R.J. Archuleta (2015). UCSB Method for Simulation of Broadband Ground Motion from Kinematic Earthquake Sources, Seismological Research Letters, January/February 2015, v. 86, p. 6167, First published on December 17, 2014, doi:10.1785/0220140103. SCEC Contribution 2085
 Dreger, D.S., and T.H. Jordan (2015). Introduction to the Focus Section on Validation of the SCEC Broadband Platform V14.3 Simulation Methods, Seism. Res. Lett. v. 86, p. 1516, First published on December 17, 2014, doi:10.1785/0220140233. SCEC Contribution 6069
 Dreger, D.S, G.C. Beroza, S.M. Day, C.A. Goulet, T.H. Jordan, P.A. Spudich, and J.P. Stewart (2015). Validation of the SCEC Broadband Platform V14.3 Simulation Methods Using Pseudospectral Acceleration Data, Seismological Research Letters, January/February 2015, v. 86, p. 3947, First published on December 17, 2014, doi:10.1785/0220140118. SCEC Contribution 1983
 Goulet, C. A., N.A. Abrahamson, P.G. Somerville, and K.E. Wooddell (2015). The SCEC Broadband Platform Validation Exercise: Methodology for Code Validation in the Context of Seismic‐Hazard Analyses, Seism. Res. Lett. v. 86, p. 1726, First published on December 17, 2014, doi:10.1785/0220140104. SCEC Contribution 6070
 Graves, R.W., and A.Pitarka (2015). Refinements to the Graves and Pitarka (2010) Broadband Ground‐Motion Simulation Method, Seismological Research Letters, January/February 2015, v. 86, p. 7580, First published on December 17, 2014, doi:10.1785/0220140101.
 Isbiliroglu, Y., R Taborda, J Bielak (2015). Coupled soilstructure interaction effects of building clusters during earthquakes, Earthquake Spectra 31 (1), 463500. SCEC Contribution 1947
 Lozos, J., D.D. Oglesby, J. Brune, K.B. Olsen (2015). Rupture and Ground‐Motion Models on the Northern San Jacinto Fault, Incorporating Realistic Complexity, Bull, Seis. Soc. Am., Vol 105., 19311946, doi: 10.1785/0120140327. SCEC Contribution 6071
 Lozos, J., K.B. Olsen, J. Brune, R. Takedatsu, R. Brune, and D.D. Oglesby (2015). Broadband ground motions from dynamic models of rupture on the northern San Jacinto fault, and comparison with precariously balanced rocks, Bull, Seis. Soc. Am., Vol 105., 19471960, doi: 10.1785/0120140328. SCEC Contribution 6072
 Maechling, P.J., F. Silva, S. Callaghan, and T.H. Jordan (2015). SCEC Broadband Platform: System Architecture and Software Implementation (2015). Seism. Res. Lett. v. 86, p. 1726, First published on December 17, 2014, doi:10.1785/0220140104. SCEC Contribution 1981
 Nakata, N., and G. Beroza (2015). Stochastic characterization of mesoscale seismic velocity heterogeneity in Long Beach, California, Geophysical Journal International, 203(3), 20492054. doi: 10.1093/gji/ggv421. SCEC Contribution 6012
 Olsen, K.B, R. Takedatsu (2015). The SDSU Broadband GroundMotion Generation Module BBtoolbox Version 1.5, Seism. Res. Lett. 86,1, 8188. SCEC Contribution 1985
 RamirezGuzman, L., R.W. Graves, K.B. Olsen, O.S. Boyd, C. Cramer, S. Hartzell, S. Ni, P. Somerville, R.A. Williams, and J. Zhong (2015). Ground‐Motion Simulations of 1811–1812 New Madrid Earthquakes, Central United States, Bull, Seis. Soc. Am., Vol 105., 19611988, doi: 10.1785/0120140330. SCEC Contribution 6073
 Taborda, R., and J. Bielak (2014). Ground‐motion simulation and validation of the 2008 Chino Hills, California, earthquake using different velocity models, Bull. Seis. Soc. Am. 104, doi: 10.1785/0120130266. SCEC Contribution 1798
 Withers, K.B., K.B. Olsen, S.M Day (2015). Memoryefficient simulation of frequency dependent Q, Bull. Seis. Soc. Am., accepted for publication. SCEC Contribution 6002