Kinematic Modeling of the 2004 Parkfield Earthquake

Ahyi Kim and Douglas Dreger


The September 28, 2004 Mw6.0 Parkfield earthquake (epcenter 35.815N, 129.374W and depth 8.1km) is probably the best recorded moderate sized earthquake to date. The regional broadband geodetic monitoring arrays, InSAR, as well as densely spaced local-distance strong motion instruments provide unprecedented coverage enabling detailed analysis of the kinematic rupture process. In-depth examination of kinematic rupture models, their resolution, and reconcilation with peak ground motion distribution is necessary for better understanding earthquake rupture physics as well as applications of such models for rapid ground motion hazard reporting.


Figure 9.1: Location of the 2004 Parkfield earthquake. The near-fault strong motion stations are plotted as the triangles. Red star indicates the hypocenter and Black dots represents the aftershocks.
\begin{figure}\epsfig{file=ahyikim06_1_1.eps, width=8cm}\end{figure}

To obtain the kinematic model for 2004 Parkfield earthquake, we carried out a joint inversion combining the near-fault strong motion records and geodetic data sets. We used 60 near-fault horizontal-componet strong motion records from 30 stations (Figure 1). Each record was integrated to velocity, band-pass filtered using a Butterworth filter with a corner frequency of 0.5 Hz, and resampled to 10 samples per second. The distance range between the fault and the stations is 0.4-7.9km.

Coseismic GPS deformation from 13 sites was obtained from 1 sample per second GPS time histories. To obtain the coseismic deformation, the GPS time histories were averaged from 15 minutes before the event to 2-10 minutes afterward.

The coseismic deformation from ENVISAT InSAR data was obtained from a pre-event scene on 7/3/2003 and post event scene on 9/30/2004. We used the interferogram which was processed by Johanson et al. (in press). Johanson et al. removed the part of the interferogram affected by non-tectonic effects, especially atmospheric changes.

Inverse Method

For the inversion, we used the method of Harzell and Heaton (1983), which is a linear, multiple time window approach to invert the data for the spatio-temporal distribution of fault slip, the average rupture velocity and its variation, and the slip rise time. Green's functions for the seismic data were calculated using the FK integration method (Saikia, 1994) for frequencies up to 0.5 Hz with the two 1D velocity models reported in Liu et al. (2006), representing the velocity contrast across the SAF.

To generate the Geodetic green's function, we followed the method of Wang et al. (2003), assuming the GIL7 layered elastic structure model (Dreger and Romanowicz, 1994).

We assumed the fault plane dimension is 44kmx18km, which is divided into 1kmx1km subfaults. The rise time was allowed to vary from 0.6 sec to 3.6 sec, which is based on the preliminary regional distance results (Dreger et al., 2005). The strike ($140^o$) and dip ($89^o$) were obtained from the aftershock distribution and some trial and error wavefrom fitting.

Inversion results

Figure 9.2: Slip models obtained independently from the a) seismic waveforms, b) GPS, and c) InSAR data sets. The model shown in d) compares the result when the three data sets are inverted simultaneously. These models are view NE and the left side is to the NW and the right to SE. The hypocenter is plotted as a white star. In the preferred model most of the slip is around hypocenter and between 10-20km NW of the hypocenter. After shocks are plotted as black dots. The x and y axes show along strike distance and depth in km. The slip located 10km to the SE is needed to fit the large amplitudes observed at the Cholame array.
\begin{figure}\epsfig{file=ahyikim06_1_2.eps, width=8cm}\end{figure}

Figure 2a), 2b), and 2c) show slip models obtained from seismic, GPS, and InSAR data sets independently. They show some common features: (1) rupture propagated mainly to northwestward from the hypocenter, and (2) high slip value 10-20km northwest of the hypocenter. In detail, however, these three models are quite different. Basically, the inversion results from the geodetic data are dominated by the shallow slip northwest of hypocenter and some deep slip in southern part of the fault. The kinematic model obtained from the seismic only data shows much more complexity compared to those of geodetic data sets. The results using near-fault strong motion data show a slip distribution that is consistent with what we obtained previously with the rgional seismic waveform and near-fault GPS data in that there is a high slip patch at hypocenter and 10-20km north terminating near Middle Mountain ( Dreger et al., 2005). This model also has some low levels of slip to the south that are needed to explain the waveforms and the high peak amplitudes at stations located south of the epicenter. Due to the non-uniqueness of this type of source inversion, it is important to obtain model that simultaneously fit different types of data. The joint inversion results are seen to show similarity with each of the independent inversions and the level of fit to each data set is acceptable.

The joint inversion and preferred model (Figure 2d) indicates a slight bilateral rupture, which is dominated by northward rupture. This model has two primary asperities: one near the hypocenter, and the other 10-20km northward of the hypocenter, which is consistent with our preliminary model with local and regional seismic and GPS data ( Dreger et al., 2005). The best rupture velocity was found to be 2.6km/s, and the average rise time was 1.6 sec. The relatively higher slip in the InSAR only model might be due to included postseismic deformation, since the post-event scene is 2 days after the mainshock. In the preferred model, the scalar seismic moment was found to be 1.10e25dyne-cm(Mw6.03), while the peak slip was found to be 45.03cm. Slip model from the joint inversion showed us that the use of the geodetic data in the inversion places considerable additional constraint on the slip distribution.


Dreger, D. S., and B. Romanowicz, Source characteristics of events in the San Francisco Bay region, U. S. Geol. Surv. Open-File Rept., 94-176, 301-309, 1994

Dreger, D., L. Gee, P. Lombard, M. Murray, and B. Romanowicz, Rapid Finite-source Analysis and Near-fault Strong Ground Motions: Application to the 2003 Mw 6.5 San Simeon and 2004 Mw 6.0 Parkfield Earthquakes, Seis. Res. Lett., V.76, 40-48 2005

ns: Application to the 2003 Mw 6.5 San Simeon and 2004 Mw 6.0 Parkfield Earthquakes, Seis. Res. Lett., V.76, 40-48 2005 . Dreger D., Murray M., Nadeau R., and Kim A., Kinematic Modeling of the 2004 Parkfield Earthquake, SSA 2005 annual meetin g abstract 2005.

Harzell, S. H. and T.H. Heaton, Inversion of strong ground motion and teleseismic waveform data for the fault rupture history of the 1979 imperial Valley, California earthquake,Bull. Seusm. Soc. Am., 73, 1553-1583 1983.

Johanson, I., E. Fielding, F Rolandone, and R. Burgmann, Coseismic and Postseismic Slip of the 2004 Parkfield Earthquake from Space-Geodetic data, Bull. Seusm. Soc. Am., (in press)

Liu, P., S. Cautodio, and R. J. Archuleta, Kinematic Inversion of the 2004 Mw6.0 Parkfield Earthquake Including site effects, Bull. Seusm. Soc. Am., (in press)

Saikia, C. K., Modified frequency-wave-number algorithm for regional seismograms using Filon's quadrature - modeling of L(g) waves in eastern North America, Geophys. J. Int., 118, 142-158, 1994.

Wang, R.,F.Martin, and F. Roth, Computation of deformation induced by earthquakes in multi-layered elastic crust - FORTRAN pr ograms EDGRN/EDCMP, comp. Geosci., 29, 195-207 2003.

Berkeley Seismological Laboratory
215 McCone Hall, UC Berkeley, Berkeley, CA 94720-4760
Questions or comments? Send e-mail:
© 2006, The Regents of the University of California