| Summary: | We have developed a method to measure finite-frequency amplitude and traveltime anomalies of teleseismic P waves. We use a matched filtering approach that models the first 25 s of a seismogram after the P arrival, which includes the depth phases pP and sP. Given a set of broad-band seismograms from a teleseismic event, we compute synthetic Green's functions using published moment tensor solutions. We jointly deconvolve global or regional sets of seismograms with their Green's functions to obtain the broad-band source time function. The matched filter of a seismogram is the convolution of the Green's function with the source time function. Traveltimes are computed by cross-correlating each seismogram with its matched filter. Amplitude anomalies are defined as the multiplicative factors that minimize the RMS misfit between matched filters and data. The procedure is implemented in an iterative fashion, which allows for joint inversion for the source time function, amplitudes, and a correction to the moment tensor. Cluster analysis is used to identify azimuthally distinct groups of seismograms when source effects with azimuthal dependence are prominent. We then invert for one source time function per group. We implement this inversion for a range of source depths to determine the most likely depth, as indicated by the overall RMS misfit, and by the non-negativity and compactness of the source time function. Finite-frequency measurements are obtained by filtering broad-band data and matched filters through a bank of passband filters. The method is validated on a set of 15 events of magnitude 5.8 to 6.9. Our focus is on the densely instrumented Western US. Quasi-duplet events ('quplets') are used to estimate measurement uncertainty on real data. Robust results are achieved for wave periods between 24 and 2 s. Traveltime dispersion is on the order of 0.5 s. Amplitude anomalies are on the order of 1 db in the lowest bands and 3 db in the highest bands, corresponding to amplification factors of 1.2 and 2.0, respectively. Measurement uncertainties for amplitudes and traveltimes depend mostly on station coverage, accuracy of the moment tensor estimate, and frequency band. We investigate the influence of those parameters in tests on synthetic data. Along the RISTRA array in the Western US, we observe amplitude and traveltime patterns that are coherent on scales of hundreds of kilometres. Below two sections of the array, we observe a combination of frequency-dependent amplitude and traveltime patterns that strongly suggest wavefront healing effects. © 2006 The Authors Journal compilation © 2006 RAS.
|
|---|