Data Sets (Task A)
- The Database Working Group met in person and through numerous conference calls to address the reprocessing and QA/QC of the NGA-East CENA dataset. We have established rules for reprocessing to address issues with components timing, windowing and reprocessing that are specific to CENA data. New code improvements were also included to facilitate the processing. The codes and the process were tested with subsets of data, which led to further refinements. The final protocols are being reviewed by the group and the TI leads before the reprocessing can start.
Reference rock and site amplification models (Task B)
- All the tasks below are being led by the Geotechnical WG (GWG). The group continues to be very active with their tasks and have monthly meetings attended by the TI Lead Dr. Goulet.
- Reports on reference rock velocity and site kappa are under review by the TI team and PPRP.
- The receiver function approach proposed by Ni and Somerville (presentation made at GWG meeting, 2010) has been implemented for sites with recorded ground motions. The group developed relationships between VSZ and VS30 using the profile database for CENA.
- The group is developing a VS30 proxy method based topographic slope conditioned by regional geology (glaciated and non-glaciated regions) in CENA.
- The group has developed a site correction/amplification function for recorded motions. The group continues to update this site correction/amplification function.
- The group continued their development of protocols of the large scale site response simulations that will be conducted using input ground motions and typical soil profiles. The GWG is finalizing the development of idealized soil profiles for the simulations of site response. This includes velocity profiles and dynamic soil properties. The output of these simulations will be used for development of nonlinear site amplification functions or models.
Velocity structure database and regionalization
- Dr. Mooney (USGS) and his team have completed the database building phase of their work and submitted their report to the NRC in January 2013. The GIS-enabled database consists in a detailed compilation of published interpretations numbering in excess of 2,000 measurements. The database contains depth-dependent seismic velocity profiles for CENA.
- A series of maps were updated based on the final database information. Figure 1 show an example of such a map for the crustal thickness of CENA. The map was generated using a larger number of data points than for any previous map. The coastal regions have a crustal thickness of about 30 km and the continental interior about 40 km, with some significant lateral variations.
- A preliminary regionalization based on the velocity database was developed. Figure 2 shows the main geologic divisions of the North American crust: the central platform, the north-central craton, and the eastern regions, which inclose the Appalachians and coastal provinces.
Figure 2. Generalized geologic setting of North America. Key provinces for this study are: (1) the central Platform and cratonic regions; (2) the Gulf Coastal Province; (3) The Appalachians; and (4) the eastern Coastal Plains.
- The next task to be completed by Dr. Mooney and his team is the evaluation of the impact of different regions on ground motions using ratios of response spectra. Several NGA-East project members participated in a conference call on March 7, 2013, to discuss the calculation of response spectral acceleration for various regions in CENA. This conference call succeeded in developing a work plan for this sub-task, involving the selection of appropriate representative profiles and Q for each regions and the selection of input parameters for the simulations to be conducted using the FK computer code, a full waveform Green’s functions propagation code (Zhu and Rivera, 2002).
Geometrical Spreading and Attenuation for Eastern North America
- Drs. Atkinson (U. of Western Ontario) and Boore (USGS) are working on using data in eastern North America to put constraints in the geometrical spreading and whole-path attenuation models that will be used in stochastic method, point-source simulations of ground motions. The idea is to derive event-specific apparent anelastic attenuation from Fourier spectra computed from data recorded in the distance range of 150 to 500 km, assuming a geometrical spreading of 1/R^0.5 beyond 50 km, and then use those path-dependent parameters to extrapolate the motions back to a distance of 1 km, assuming various rates of geometrical spreading at distances within 50 km. The rate that best agrees with the seismic moments of the events is then chosen, and residuals at all distances using this “best” bilinear geometrical spreading model are studied. These residuals showed some trends at lower frequencies that resulted in a frequency-and-distance dependent modification to the bilinear model. They are working on a paper describing their research, and Dr. Boore has modified his SMSIM programs to allow for the new geometrical spreading model (Boore, 2009).
- Dr. Boatwright (USGS) has been working on a final report on the attenuation structure of northeastern North America. He has significantly revised his map of Lg attenuation in Northeastern North America: the region to the north of the St Lawrence River from Ottawa to the Lower St Lawrence is characterized by Q ~ 580f^0.46, similar to the attenuation that he previously determined for western Ontario. Dr. Boatwright has verified the revised attenuation structure using recordings of the 1990 & 1993 Mt Laurier earthquakes, the 1997 La Malbaie earthquake, the 1999 St Anne des Monts earthquake, the 2000 Kipawa earthquake, the 2002 Ausable Forks earthquake, the 1991 M3.7 Cobleskill earthquake, the 2000 M3.5 Newcomb earthquake, the 2010 Val-des-Bois earthquake, and the recent 2012 M3.6 Beloeil earthquake.
- Dr. Frankel (USGS) continued his analysis of earthquakes in the Charlevoix region and also investigated the decay of distance of S-waves from an aftershock of the Mineral, Virginia earthquake, using data supplied by Martin Chapman of Virginia Tech. Coda normalized S-wave amplitudes for the Virginia aftershock at 14 Hz show a steep decay of R^(-1.4±0.13), without correction for Q. This is steeper than the average decay of coda-normalized S-waves found from seven earthquakes in the Charlevoix area R^(-1.1±0.16), also without correcting for Q. Frankel is working on additional analyses to understand the cause of this difference. Dr. Frankel examined the hypocentral-depth dependence of the amplitude decay of the Charlevoix events. No clear dependence on the hypocentral depth could be resolved with the current set of earthquakes. Frankel determined coda envelopes for additional, recent earthquakes in the area, but found that they did not have sufficient signal at the longer lapse times in the coda used in the previous analysis. He is continuing the writing of a paper on the amplitude decay of the coda-normalized S-waves. He has also continued to run finite-fault stochastic simulations using 1/R geometrical spreading (up to 80 km distance), as the next step for producing broadband synthetics to develop new GMPE’s.
- The large ground motion simulations validation is progressing in collaboration with various stakeholders. The group, including the TI Leads, continues to have productive weekly calls. The first seven events for validation have been defined (Part A. Validation against recordedevents) and three generic scenarios have been defined for the comparison with GMPEs (Part B. Validation against GMPEs). This first round of validation is focused on ATR events and GMPEs, for which the models have been tested against in the past. CENA events are to be run this summer.
- EXSIM was fully implemented on the SCEC broadband platform. Drs. Atkinson Assatourians (U. of Western Ontario) have been working on optimizing their input parameters for four WUS events, two Japanese event as well as for a comparison with NGA-West1 GMPEs.
- The Irikura Recipe and the Reno/Anderson models are being implemented by the SCEC software development team in collaboration with Dr. Somerville and Mr. Bayless (URS) and Dr. Anderson (UN Reno).
- New post-processing plots have been implemented by Mr. Silva (SCEC) in collaboration with Dr. Goulet (PEER). Two new sets of plots involving GMPEs (2008 NGA-West1 models) are now generated on the platform:
1) a plot that produces a goodness-of-fit (GOF) with spectral period where the GMPE predictions are compared with the observed data. Figure (Figure 3b) shows an example of the ne plot. The comparison of the fit against data for both the simulation method and the GMPEs allows to visualize the event-specific features of an earthquake (event term in a GMPE). Although not used formally for the evaluation of the methods, this type of plot allows to perform a reality check on the observed GOF.
2) a plot that compares all the simulations from a method to GMPEs and a predefined acceptance criterion. Figure (Figure 3a) shows the goodness of fit across all stations for a single realization of the source. Figure (Figure 4) shows the comparison of simulations for generic scenarios described in (2) above. The box plots show the median, +/- standard deviation and the extrema from the simulations.
Development of Updated Stochastic Models (Task F)
- Dr. Boore (USGS) worked on the magnitude scaling of ground motions, as predicted by standard point-source stochastic models, compared to the scaling shown by observed data. Although the data are from active tectonic regions, they are the only data available to test predictions of magnitude scaling from the stochastic model, and the predicted motions from point-source simulations will form the basis of ground-motion prediction equations derived for eastern North America. The comparisons show that the stochastic point source model using closest distance to the source does not predict the observed tendency for saturation of short-period ground motions. Using an effective distance, as advocated by several authors, to try to capture finite-fault effects helps in reducing the disagreement between observations and simulations, but more careful and thorough work needs to be done to derive these finite-fault adjustment factors. Dr. Boore is doing this in collaboration with Dr. Atkinson (U. of Western Ontario) and Dr. Alessandro (GeoPentech), using both simulations from finite-fault stochastic simulations (using program EXSIM) and using the empirically based NGA-West2 ground-motion prediction equations as a convenient summary of observed ground motions.
- Dr. Boore derived generalized two-corner source spectral models for which the high-frequency spectral levels are equal to those of the standard single-corner frequency source model for a specified stress parameter. The new model is a generalization of one used in several papers by Gail Atkinson and colleagues (e.g., Atkinson and Boore, 1995, and Atkinson and Silva, 2000)., but in the earlier model, there was no flexibility in the shape and amplitude of the source spectra. For example, the high frequency spectral level was not a free parameter. The generalized models should be useful in modeling both observed and simulated data. After deriving the new models, Dr. Boore revised his suite of stochastic method simulation programs (SMSIM) to allow use of the new models.
Sigma (Standard Deviation) (Task J)
Uncertainty from simulated ground motions (J.1):
- Dr. Spudich is working with Dr. Cirella of the Istituto Nazionale di Geofisica e Vulcanologia (INGV) in Rome, Italy, on the quantitative inclusion of error bounds in ground motion inversions to determine source rupture behavior. The goal is ultimately to place bounds on the correlation of rupture parameters (eg peak slip velocity and rupture speed) based on inversion of real earthquake data. These bounds will be compared with the correlations of rupture parameters found in dynamic rupture models. Drs. Spudich and Cirella have been working to implement the theory of Yagi and Fukahata (2011) (YF) in the simulated annealing inversion algorithm of Piatanesi et al. (2007). The YF theory shows how to include Greens function errors into a ground motion inversion for rupture behavior. YF theory works in the time domain and parameterizes the earthquake rupture using constant-slip subfaults that rupture instantaneously. The Piatanesi inversion works in the frequency domain and uses a bilinear function to represent the peak slip velocity, risetime, rupture time, and skewness of the source time function. Drs. Spudich and Cirella have been able to perform the theoretical work to adapt the YF theory to the inversion method of Piatanesi. Dr. Tinti (INGV) is working on characterizing the error in theoretical Greens functions. Drs. Piatanesi and Cirella are modifying the inversion code. Dr. Cirella is selecting good candidate earthquakes, which will probably be Chuetsu, Tottori, and L’Aquila.
Develop models for standard deviation (within-event and single-station) (J.2):
- The Sigma Working Group is analyzing the recently compiled NG-West2 dataset, especially for small magnitude data, to quantify single-station sigma. This task will be repeated with the NGA-East dataset in Fall 2013.
Boore, D. M. (2009). Comparing stochastic point-source and finite-source ground-motion simulations: SMSIM and EXSIM, Bull. Seism. Soc. Am. 99, 3202-3216.
Piatanesi, A., A. Cirella , P. Spudich, and M. Cocco (2007). A global search inversion for earthquake kinematic rupture history: Application to the 2000 western Tottori, Japan earthquake, J. Geophys. Res., 112, B07314, doi:10.1029/2006JB004821.
Yagi, Y. and Fukahata, Y. (2011), Introduction of uncertainty of Green’s function into waveform inversion for seismic source processes. Geophys. J. Int., 186: 711–720.
Zhu, L. and Rivera, L. A. (2002), A note on the dynamic and static displacements from a point source in multilayered media. Geophys. J. Int., 148: 619–627.