next up previous contents
Home: Berkeley Seismological Laboratory
Next: Detection and Location of Up: Ongoing Research Projects Previous: Automated Moment Tensor Software   Contents

Waveform Constrained Seismic Velocity Structure in Northern California

Junkee Rhie and Douglas S. Dreger


1D and 2D shear wave velocity structure from Mammoth Lakes to Yreka is determined by SH waveform modeling and receiver-function analysis. Three local events are selected by considering epicentral distances from stations and the availability of reliable source parameters (e.g. Dreger et al., 1995). The six teleseismic events are also selected by considering preferred back azimuth and S/N ratio (Figure 30.1). Regional broadband waveforms from the 21 September 1993 Klamath Falls ($M_w$ 6.0), the 15 May 1999 Mammoth Lakes ($M_w$ 6.0), and the 10 August 2001 Portola ($M_w$ 5.2) events were well recorded by 4 to 5 BDSN stations that are also located nearly on the same NNW line (Figure 30.1). This naturally aligned configuration of three local earthquakes and stations provides an excellent opportunity to determine a waveform constrained velocity model along the profile. Before performing the waveform modeling, a receiver function technique is applied to constrain the depth of major discontinuities, especially Moho, at each station. 1D and 2D models are estimated by forward modeling of the broadband waveforms and the receiver-functions.


Receiver-function analysis is a very appropriate method to look at the discontinuities in the crust and upper-mantle. However, the velocity structure obtained from receiver-function analysis is not unique since receiver-functions, especially the radial components, are mostly sensitive to P to S velocity ratio rather than the absolute velocity at each depth (Ammon et al., 1990). Three component broadband waveforms from regional events are relatively more sensitive to the absolute velocity at each depth. These complementary properties give us a chance to obtain more constrained velocity structure than those obtained by only one or the other of the two methods. Langston (1994) showed that regional broadband waveform modeling based on an initial 1D model obtained by teleseismic receiver function inversion is very promising for the determination of source parameters and to infer details of crustal and upper mantle structure. But he also mentioned that lateral heterogeneity in velocity structure is still an obstacle to overcome. The basic idea of this work is the same as his, but the final goal is to determine 2D or 3D model, which can include the effect of lateral heterogeneities. Many 1D models obtained by waveform modeling of receiver functions and regional broadband waveforms can be used as a basis for more complex 2D and 3D models. The detailed process of waveform modeling scheme for 1D modeling is as follows; 1) The depth of major discontinuities, especially Moho, is constrained by receiver function inversion. 2) S-wave velocities in each layer are modified to fit the observed SH regional waveform. 3) P-wave velocity is obtained by perturbing P velocity to explain P-SV regional waveforms. 4) The P model can provide more constraint on the depths of major discontinuities identified by the receiver function analysis and then the above process is repeated to obtain a reliable 1D model, which can explain receiver function and regional waveforms simultaneously. The more complex 2D or 3D model will be estimated based on many 1D models.

Figure 30.1: Location map of stations and events for regional (left) and teleseismic (right) cases. The regional events and broad band stations are located on the NNW line. The teleseismic events of which back azimuth from BDSN is similar to the trend of NNW line was selected for receiver function analysis
\epsfig{file=rhie02_2_fig_01.epsi, width=8cm}\end{center}\end{figure}

Discussion and Results

A preliminary 2D shear wave velocity model for NNW profile in northern California was estimated. Three 1D models for the S structure under YBH, ORV and CMB stations are interpolated to get the 2D result. Stacked waveforms from teleseismic events are used as input waveforms for receiver function calculation. Tangential component of regional waveforms recorded at the three BDSN stations are modeled by forward waveform modeling approach with fixed depths of major discontinuities that were obtained from receiver function inversion. The variance reductions of each waveform fit are all about 60 % or larger, even for frequencies up to 0.5 Hz (Figure 30.2). Although we did not give more constraint on depths of major discontinuities by P-SV waveform modeling, the obtained 1D models seem to be able to explain regional tangential waveforms well. However, this 1D model cannot explain receiver functions simultaneously (Figure 30.2). The preliminary 2D model based on an interpolation of given 1D models from waveform modeling cannot explain the waveforms which propagated along the NNW line. However, this model roughly shows the reasonable variation of the velocity structure along the NNW line; deeper Moho and higher velocity layer under Sierra Nevada and shallower Moho and lower velocity layer under Klamath Mountains (Figure 30.3). The further direction of this study will be to estimate 1D models which can explain waveform of regional events and receiver functions simultaneously. P-SV waveforms will be incorporated and the results of the 1D analyses will be used to construct a 2D model for the path. In addition, other source-receiver pairs and receiver functions with wide variety of back azimuth will be used to estimate 3D velocity structure model.

Figure 30.2: Preferred receiver function models (grey) and refined models (black). The refined models were obtained by forward waveform modeling. Left: Radial component receiver function based on receiver function model (grey) and refined model (dotted). Right: Comparison between observed (black) and synthetic tangential waveforms (grey) for refined model
\epsfig{file=rhie02_2_fig_02.epsi, width=8cm}\end{center}\end{figure}

Figure 30.3: The preliminary 2D shear wave velocity model is constrained using three 1D models obtained from regional waveform modeling. Comparison between observed (black) and synthetic tangential waveforms (grey) for interpolated model
\epsfig{file=rhie02_2_fig_03.epsi, width=8cm}\end{center}\end{figure}


We thank C. Ammon for letting us apply his RFTN software package to all receiver-function related calculations.


Ammon, C. J., G. E. Randall, and G. Zandt, On the non-uniqueness of receiver function inversions, J. Geophys. Res., 95, 15,303-15,318, 1990.

Dreger, D., J. Ritsema, and M. Pasyanos, Broadband analysis of the 21 September, 1993 Klamath Falls earthquake sequence, Geophys. Res. Lett., 22, 997-1000, 1995.

Langston, C., An integrated study of crustal structure and regional wave propagation for southeastern Missouri, Bull. Seism. Soc. Am., 84, 105-118, 1994.

next up previous contents
Berkeley Seismological Laboratory
215 McCone Hall, UC Berkeley, Berkeley, CA 94 720-4760
Questions or comments? Send e-mail:
© 2002, The Regents of the University of California.