Annual Meeting - SCEC.org

Poster Abstracts
the experiment is for each model to generate a forecast that specifies the expected number of earthquakes for the next 24 hours
in bins specified by ranges of latitude, longitude, and magnitude. The spatial cells are 0.1 degrees by 0.1 degrees, and the
models target all earthquakes with magnitude greater than or equal to 3.95 in bins of 0.1 units. We examined the first three
years of results; during this time, 270 target earthquakes occurred in the testing region, including a swarm in Baja California
throughout February 2008 and the April 2010 M 7.2 El Mayor-Cucapah earthquake.
Through the testing component of this experiment, we discovered a flaw in the STEP implementation; it was corrected, and
we evaluated forecasts based on the revised implementation. To evaluate the models, we conducted investigations of
consistency (each model relative to the observation) and comparison (each model relative to the other). Based on joint loglikelihood
measures, we found that both models are reasonably consistent with the observations of interest: the number of
earthquakes, their spatial distribution, their magnitude distribution, and the joint distribution. To compare the models, we
computed the rate-corrected average information gain per earthquake; this is the natural logarithm of the probability gain.
STEP had an average information gain of 0.43 relative to ETAS; in other words, STEP had, on average, a 54% higher forecast
rate than ETAS in bins where target earthquakes occurred, indicating that STEP was superior in this experiment.
FOCAL MECHANISM WITH ERROR ESTIMATION USING REGIONAL WAVEFORMS: A CASE STUDY
WITH THE 2008 CHINO HILLS EARTHQUAKE (B-072)
Z. Zhan, S. Wei, and D. Helmberger
Earthquake focal mechanisms are now routinely reported using regional records. However, due to difficulties in fitting
seismograms with inadequate velocity model, and the usually non-linear inverse methodology, error estimation of focal
mechanism is not a trivial problem. Here we explore the potential of using the bootstrapping method to estimate the error of
focal mechanism from Cut-And-Paste (CAP). CAP first cuts the regional seismograms into Pnl wave and surface wave
segments, then fits them allowing different time shifts, period bands and weights. CAP finds the best waveform-fitting focal
mechanism by grid-searching the whole model space so we have the full information about each station's waveform misfit for
every possible focal mechanism. This information naturally goes into the bootstrapping method. We validate this technique
with the 2008 Chino Hills earthquake, which is well recorded and studied with the dense regional data from the TriNet. We
also test the possibility to correct Amplitude Amplification Factors (AAFs) for each station using the information from the
large number of bootstrapping solutions. Tests with both synthetic test and real data show largely reduced errors of focal
mechanism parameters.
STRAIN LOCALIZATION AND CRUSTAL THICKNESS VARIATION ACROSS MAJOR STRIKE SLIP
FAULTS IN SOUTHERN CALIFORNIA REVEALED BY RECEIVER FUNCTION STUDIES (B-140)
P. Zhang, M.S. Miller, J.F. Dolan, I.W. Bailey, and D.A. Okaya
The degree to which faults are localized or distributed within the continental lithosphere has been a controversial subject for
decades. Reflection and refraction studies indicate that the San Andreas fault (SAF) extends into the lower crust as a narrow
fault zone in northern California and along the Mojave segment of the fault. However, the discretness of lithospheric
deformation along the southernmost segments of the SAF system (e.g., the Elsinore, the San Jacinto and the Coachella segment
of the San Andreas) remains unclear. We have begun a detailed study of how the distribution of strain at depth is in southern
California using P receiver functions to image crustal and lithospheric thickness variations. We specifically focus on the
depths of the Moho and lithosphere-asthenosphere boundary (LAB) underneath the major faults. Our results for the geometry
of crustal and lithospheric interfaces provide insight into the strain localization in this region. Our preliminary results using
events recorded by Southern California Seismic Network and USarray reveal complex crustal and lithospheric structure in this
region. A SW-NE profile across the Elsinore, San Jacinto, and and San Andreas faults indicates that the Moho dips to the
southwest. Moreover, receiver gathers at certain stations near the faults show a strong back-azimuthal variation. Using simple
velocity models to project the Ps conversions to depth, we find that some of the back-azimuth variation can be related to
different sides of the faults. These back-azimuthal variations of the Moho signal indicate three-dimensional complexity
beneath the central San Jacinto fault that may suggest strain localization at these depths.
AWP-ODC-SGT: WAVE PROPAGATION AND DYNAMIC RUPTURE SIMULATION SOFTWARE WITH
STRAIN GREEN’S TENSOR (SGT) CAPABILITIES FOR CYBERSHAKE 2.0 (A-031)
J. Zhou, S. Callaghan, S. Song, P. Chen, K. Olsen, R. Graves, P. Small, P. Maechling, Y. Cui, and T. Jordan
CyberShake (Graves, et al., 2008/2011) is a SCEC research project to develop a physics-based computational approach to
create a probabilistic seismic hazard analysis (PSHA). CyberShake uses fully 3D wave propagation simulations to forecast
260 | Southern California Earthquake Center