Inversion of Body-Wave Seismograms for Upper Mantle Structure

Citation

Given, Jeffrey Wayne (1984) Inversion of Body-Wave Seismograms for Upper Mantle Structure. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/cgdh-0e19. https://resolver.caltech.edu/CaltechTHESIS:10242018-091909982

Abstract

We invert observed long- and short-period body-wave seismograms, travel times, and apparent velocity data to further constrain the compressional velocity structure in the upper mantle beneath northwestern Eurasia and the shear-wave velocity structure beneath western North America.

Long- and short-period WWSSN seismograms from nuclear explosions in the Union of Soviet Socialist Republics are incorporated with apparent velocity observations to derive an upper mantle model for northwestern Eurasia. The compressional waves from these explosions have several distinctive features that provide important new information about the character of the upper mantle in the region. The seismograms from 9° to 13° exhibit impulsive first arrivals, P_n, implying a smooth positive velocity gradient between depths of 60 and 150 km. There is a consistent pulse arriving about 2s after P_n at the distances of 13° to 17°, and at larger ranges there are distinct reflections from two major discontinuities in the mantle. Synthetic seismograms displaying these features indicate a velocity model that correlates with other models from around the world, with a distinctive lid and low-velocity zone. The arrival following P_n is modeled by positioning the low-velocity zone between 150 and 200 km. The model is relatively smooth from a depth of 200 km down to 420 km, where a 5% jump in velocity produces a triplication in the travel time curve from 15° to 23°. The observations from 21° to 26° clearly show another discontinuity at a depth of 675 km with a 4% change in velocity. These results suggest that stable continental regions may have a shadow zone that extends beyond 17°. Below 250 km there is no distinguishable difference between the model proposed for northwest Eurasia and models derived for the United States.

A systematic inversion technique is proposed to extract the maximum amount of information from these data. We use the WKBJ method to compute approximate synthetic seismograms in a radially heterogeneous earth. Where the WKBJ method breaks down, in low-velocity zones and near discontinuities, a generalized ray expansion is used in a layered model approximation to the velocity structure to isolate the energy that has reflected from these regions. Synthetic seismograms computed using these approximations compare very well to those computed by the more accurate method of summing primary reflections in a generalized ray sum yet require 1/20 the computation time. With this efficiency it is feasible to compute the differential seismograms necessary to pose an inverse problem.

With a fast means of computing synthetic seismograms, an inverse problem can be posed to relate the differences between observed and synthetic seismograms to perturbations in the velocity structure. The problem is nonlinear, especially at high frequencies, but at long periods an iterative technique based on a linearized relation between perturbations in the velocity structure and the seismograms is effective if a reasonable initial model is assumed. Some simple tests of the method indicate that convergence to a satisfactory final model is possible even when starting with a model that predicts substantially different seismograms than those observed.

We invert long-period SH waves recorded on WWSSN seismographs at distances from 15° to 31° in the western United States and East Pacific Rise to determine the upper mantle shear velocity structure beneath these regions. A high velocity gradient near 400 km produces clear later arrivals from 15° to 17°. We interpret large later phases observed al distances from 23° to 27° as another large velocity gradient at between 600 and 720 km depth. Inversion of these seismograms suggests that the velocity gradient in the upper 200 km of the mantle is small; there is an increase in the velocity gradient around 250 km resulting in a 4% velocity increase by 360 km. The large velocity gradient near 400 km results in a velocity increase of around 8½% between 360 km and 420 km depth. The velocity gradient becomes smaller between 420 and 600 km with a cumulative increase of 5% over these depths. The total increase in velocity from 600 to 750 km is about 14%. Below 750 km the velocity gradient is assumed to be similar to those predicted by global studies of travel times.

There are differences in published travel time data and models that have been derived to fit the SS phases and SS-S differential times observed in this region. The discrepancies amount to about 5s in the direct S-wave travel time at distances of 15° to 18°. The discrepancy appears to be on the order of 3 s from 19° to 23° and is not resolvable beyond. These disagreements are probably the manifestation of large velocity heterogeneities in the uppermost mantle; either assumption concerning absolute travel times can be fit by models that are virtually identical below 270 km. Absolute travel times can constrain absolute velocities and, thus, are necessary to constrain the depth to discontinuities. Waveform data can constrain the structural details better. A joint waveform and travel time inversion method is a very useful tool for interpreting seismograms for earth structure.

Thesis Files