Earth Resources Laboratory :: Abstracts

Crustal structure from teleseismic bodywave data

John Edward Foley

Submitted to the Department of Earth, Atmospheric, and Planetary Sciences on May 25, 1990 in partial fulfillment of the requirements for the degree of Doctor of Science

Abstract

In this thesis we have developed and implemented a series of techniques to determine information about the crustal velocity structure of the earth beneath a network of seismic stations from the analysis of teleseismic P-waveforms. We examined the usefulness of methods which utilize vertical component teleseismic P-seismograms recorded on 2 seismic monitoring networks. The first is located in the northeastern United States and is utilized as a test area for the new methods, and the second is in Larderello, Italy, the site of one of the world's largest geothermal energy production facilities which is currently being explored with a variety of geological and geophysical methods.

The main general conclusion of this study is that the analysis of vertical component teleseismic P-waveforms can provide very useful information about the crustal velocity structure of the earth. It has long been recognized that the delays in travel time of direct P-waves can image the broad lateral extent of the low velocity zone in Larderello. To improve this model of the earth structure we examined the waveforms for primary reflections from deep velocity discontinuities which either have regional extent or are isolated to the vicinity of individual receivers. The measure of the travel times from these phases (although much more difficult to make than the direct arrival) hold valuable information about the crust. We developed two methods to extract this information from the vertical component teleseismic P-waveforms. The first is the application of a simulated annealing technique to the problem of relative travel time determination and works on the premise that within a window in each waveform a wavelet is common to all stations recording the same event. We use this optimization method to locate the Moho in New England and to determine accurate measures of direct arrivals in the Larderello data. The second method relies upon an important data transformation which simplifies and regularizes the waveforms. This transformation is a two-step process, where we first determine the source wavelet common to all receivers for each recorded event and then convert each source into a simple and repeatable zero-phase wavelet. Once transformed, we take advantage of the wide variety of event incidence angles present in the New England and Larderello data sets. Each primary reflection two-way travel time is dependent on the event incidence angle (or ray parameter), and we exploit this dependency to determine the relative travel times and average velocities to major discontinuities in the crust by using a ray parameter trajectory stacking scheme (called the rpt method).

To extract all of the available information about the crustal velocity structure out of teleseismic waveforms, one must incorporate the entire waveform into the analysis. To this end, we have developed and applied a waveform inversion method to refine the details of the velocity model sketched by the previous techniques. This method is based on the calculation of sensitivity functions, or partial derivatives, of the predicted seismogram to changes in each of the parameters which are used in the calculation of the synthetic waveform. This waveform matching scheme uses the misfit to the data and the Frechet kernel to update the model, and with this process we can resolve important velocity features in the crust.

In addition to these general conclusions we have determined a number of specific important and interesting details about the velocity structure of the Larderello geothermal area. The travel time residual inversion yielded information about the size and extent of the low velocity feature in the crust. This intrusive body is about 20 km by 20 km in lateral extent and exists from depths of about 6 km to below 40 km. The strong travel time residual in the area (about 1 second over about 30 km) indicated a region of intense reduced velocity to by at least 20% (melts of igneous rocks are reduced in velocity by 30 to 40%). The rpt method was applied to the Larderello data to help clarify this picture of the crust, and we found that beneath most stations in the region, strong velocity discontinuities exist at depths of 20 to 25 km. This regional feature is interrupted in the central portion of the area where a negative gravity anomaly is strongest and where temperatures are most elevated. This area has a number of more isolated velocity contrasts.

Our waveform inversion technique confirms many of the findings of the previous applications to teleseismic data and supplements them with detailed information about the crustal velocity structure (particularly in the upper 3 to 10 km). This part of the crust is difficult to image with conventional reflection techniques utilizing vertical component teleseismic waveform data (direct arrivals, primary reflections and full P-waveforms) and two data enhancement techniques (simulated annealing and source equalization) can reveal some of the fine details of the velocity structure of the crust.