Open AccessAutomated multimode inversion of surface and S waveforms
Author(s) -
Lebedev Sergei,
Nolet Guust,
Meier Thomas,
Van Der Hilst Rob D.
Publication year - 2005
Publication title -
geophysical journal international
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.302
H-Index - 168
eISSN - 1365-246X
pISSN - 0956-540X
DOI - 10.1111/j.1365-246x.2005.02708.x
Subject(s) - seismogram , waveform , geology , amplitude , inversion (geology) , algorithm , signal processing , seismic wave , computer science , geophysics , seismology , optics , physics , telecommunications , radar , tectonics
SUMMARY Inversion of the surface, S , and multiple‐ S waveforms is an effective means of constraining the structure of the upper mantle, including the transition zone. Exploiting the resolving power of the enormous volume of presently available data requires efficiency of data processing and waveform modelling. An established method for rapid generation of synthetic seismograms is the summation of surface‐wave modes under an assumption that the effect of seismic‐wave scattering is negligible. This assumption is valid at best for portions of a broad‐band signal, the portions being generally different for different seismograms. Interactive selection of such parts of the signal is impractical for large data sets of tens or hundreds of thousand seismograms. Here we present a fully automated waveform inversion technique with selection of signal uncontaminated by scattered waves implemented as its integral element. Waveforms are inverted using non‐linear optimization and the results are put in the form of orthogonalized linear constraints on average elastic structure along the source–station paths. Structural information from waves of different amplitudes and different types is balanced by means of time‐ and frequency‐dependent weighting, also used to suppress the propagation of errors in the data. The equations obtained from processing thousands of seismograms can be inverted together for high‐resolution upper‐mantle models. The technique has been applied to a large Western Pacific data set. Analysis of the results suggests that it has been effective and, in particular, confirms that ‘chance’ fits of scattered waves or noise do not pass the automated selection procedure. Results of the processing also provide an empirical mapping of the field of the JWKB‐approximation validity for modelling the propagation of surface waves. While there is no sharp boundary of the JWKB regime, the probability of the approximation validity decreases with increasing distance and frequency.