Long Period Regional Body Waves

Citation

Wallace, Terry Charles, Jr. (1983) Long Period Regional Body Waves. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/q7fe-4g63. https://resolver.caltech.edu/CaltechTHESIS:10282019-171911401

Abstract

The seismograms recorded at regional distances (2°-12°) are quite complicated due to the waveguide nature of the crust. The body wavetrains are essentially crustal reverberations. If these complicated waveforms are modeled with synthetic seismograms then significant information can learned about the seismic source and the structure along the travel path. With certain restrictions, the long-period regional body waves (P_nl) from shallow, continental earthquakes can be modeled with a layer (crust) over a halfspace (mantle). Generalized ray theory and the Cagniard-de Hoop technique can be streamlined for computing a synthetic seismogram in such a structure. We present an approximation to the travel time equation which results in an analytic inversion for the de Hoop contour. The simplicity of the individual rays requires that the displacement potential need only be evaluated at a small number of time points; small changes In structure are, to first order, expressed in terms of the timing of different arrivals. It is possible to "stretch" or "squeeze" the synthetic to simulate a change in structure. Therefore, a single Green's function can be used to Investigate a whole suite of structural models.

If the average crustal structure is known, the P_nl, waveforms are insensitive enough to structural details to allow the extraction of source parameters of moderate size earthquakes. The technique which Is used is an iterative least-squares waveform inversion which makes use of an error function determined by the cross-correlation of an observation and a synthetic. Since any synthetic seismogram is a combination of the three fundamental faults, the error functions can be written as a series of cross-correlations multiplied by constants corresponding to source orientation. Once the cross-correlations are computed the source orientation is determined iteratively, and only the constants have to be recalculated. The inversion procedure requires only a small data base. Several examples are presented to demonstrate its usefulness.

If the source orientation is known, then differences in the synthetic waveform and observed P_nl, can be parameterized in terms of the crustal thickness and Pn velocity. An inverson technique based on the error function previously described has been developed to determine crustal structure from P_nl. Once the structure is known for many paths a regionalized map can be produced. Such a map is presented for the western United States.

The ability to efficiently model P_nl makes it possible to use it as a routine tool. We present two example of this procedure, the first of which is the 1980 Mammoth Lakes earthquake sequence. The fault mechanisms which are determined at long-periods (> 5 seconds) differ significantly from those determined by the distribution and polarity of local short-period first motions. Although it is not possible to isolate the cause of the discrepancy, at least part of It appears to be structurally related. Local short-period arrivals which travel through the Long Valley Caldera could be systematically deflected. The second example Involves the signature of tectonic release on the long-period P waves from underground nuclear explosions. The distortion of explosion waveforms can be modeled as a double couple which has a strike-slip orientation. The modeling of the sP phases at upper mantle distances requires time functions which have short durations. The short duration can be interpreted in terms of very high stress drops if the tectonic release is triggered fault motion. For this reason we prefer a driven fault model.

Thesis Files