Data processing

Regional travel-time tomography relies on the observations of travel-time variations of P and S waves from distant earthquakes across an array of seismograph stations densely deployed in the area of interest. Therefore, I searched all the available broadband stations located in the study region, which is centered at the RGR and ranges from 28°N to 42°N in latitude and 100°W to 114°W in longitude. The waveform data recorded globally at teleseismic distances, i.e., epicentral distance greater than 30^o, with magnitudes >= 5.5 were then retrieved from Data Management Center (DMC) of the Incorporated Research Institutions for Seismology (IRIS) by means of a handy JAVA-based SOD (Standard Order for Data) program. The SOD program is a very useful tool in seismology developed by the University of South Carolina and designed to automatically select, download, and routinely process massive seismic data which match the user-defined criteria specified by earthquake parameter, station distribution or a combination of both [Owens et al., 2004]. For the events that occurred prior to the year of 2001, the origin times and hypocenter locations were based on the relocated ISC (International Seismological Centre) catalog by [Engdahl et al., 1998]. For those recent events occurring in 2007 and 2008, the earthquake information was obtained from the Monthly Hypocenter Data File (MHDF) and Weekly Hypocenter Data File (WHDF) distributed by National Earthquake Information Center (NEIC) at the US Geological Survey (USGS).

Table 1 compiles a number of permanent and temporary seismic networks spanning over the last three decades which provide broadband stations and data used for the

tomographic study. It is noted that the station coverage is much denser on the western half of the study region due to the timely schedule of the USArray transporting to the Colorado Plateau during the years of 2007 and 2008 (figure 1-1).

Network Description Time span #station

TA USArray 2007-2008 170

XM La Ristra 1999-2001 58

XK CDROM 1999-2000 34

XL Deep Probe 1997 32

XG Rocky Mountain Front 1992 26

US United States National Seismic Network 1991 ~ 9

IW Intermountain West 2004 ~ 3

IM TXAR Array 1970~ 2

NR NARS Array 1983 ~ 2

IU Global Seismograph Network 1988 ~ 2

UU University of Utah Regional Network 1981 ~ 1

Table.1 List of seismic networks and numbers of available broadband stations used in the study.

Because the instruments of earlier seismic experiments before 2001 were equipped with various types of broadband sensors (STS-2, CMG-3T, CMG-3ESP, and CMG-40T), they would yield considerably different responses for a velocity impulse input, particularly in the pulse tails that accordingly had influences on estimated travel time residuals by waveform cross correlation. Therefore, prior to cross-correlation travel time measurements, the instrument responses of the raw seismograms were equalized by removing their individual instrument responses and convolving the resulting waveforms with the CMG-40T sensor response.

In order to measure relative travel-time anomalies of various body-wave phases between stations, including P, PcP, PKPdf, S and ScS waves used to constrain the variations of P and S wave speeds in the upper mantle beneath the RGR, I first subtracted the arrival times of individual phases predicted from a radially-symmetric Earth model,

AK135 for the continental velocity structure [Kennett et al., 1995]. The seismograms were then windowed 80 s before to 200 s after the expected phase arrivals and applied with a fourth-order, zero-phase Butterworth band-pass filter.

The travel time of an actually band-limited seismic wave measured by a waveform cross correlation method is explicitly frequency dependent, because the lower frequency wave ‘feels’ the structure farther off the ray path and experiences larger degree of wavefront healing while the higher frequency wave is more sensitive to small-scale structures close to the ray [Hung et al., 2001; Yang and Hung, 2005]. Such frequency-dependent characteristics are well represented by 3-D finite-frequency sensitivity kernels which make it tenable to jointly invert travel-time data measured at multiple frequency bands, each of which contributes complementary knowledge to the heterogeneous velocity structure of the earth’s interior. I thus chose to measure the travel-time anomalies of P- and S-wave phase arrivals in both high- and low-frequency ranges using the first complete cycles of the band-pass filtered waveforms on the vertical and transverse component, respectively. As the P- and S-wave signals usually comprise different frequency contents, the bandwidths selected for P and S waves are not the same in order to suppress the noise level and highlight the signals suitable for cross correlation travel-time measurements. The cutoff frequencies of the high- and low-frequency band-pass filters applied to P and S phase arrivals are listed in Table 2.

Moreover, to avoid the potential contamination of multiple phases arriving closely within the cross correlation time windows, the events used for measuring various types of phase arrivals are limited to certain ranges of epicentral distances within which only the desired single phase is expected to emerge according to the travel-time curves predicted by the 1-D reference Earth model (see Table 3).

Fig. 2-1 displays the distributions of azimuths and epicentral distances of 1196

earthquakes in high-frequency data and 1045 earthquakes in low-frequency data that provide the travel-time data for the tomographic imaging of P and S velocity structures.

The majority of the useful events occurring along the circum-Pacific seismogenic belt are concentrated in the azimuth range counterclockwise from northwest to southeast, while the events in the Eurasian and African continents toward the opposite azimuths are rather scarce. The uneven geographical distribution of data is commonly a natural defect for the passive tomographic imaging such that the regularization will come into play in the inversion to solve for the physically-reliable velocity models intrinsically with spatially-varying resolutions.

Fig. 2-1. Geographical distribution of all the events used in the study. Large circles denote the events for the measurement of low-frequency travel time data; small circles denote those for high-frequency data.

Phase High frequency (Hz) Low frequency (Hz)

P 0.3 – 2.0 0.03 – 0.125

S 0.1 – 0.5 0.02 – 0.2

Table. 2 Frequency bands used for P- and S-wave travel-time measurements.

P wave S wave

P 30 – 99.3 S 30 – 80

PKPdf 120 - 180 ScS 50 - 75

PcP 30 - 70

Table. 3 Ranges of epicentral distances of the events used for measurements of various types of phase arrivals.