Abstract:
This thesis presents elastic finite difference simulations of two
dimensional wavefield propagation for line-9, recorded during the 1993 part
of the multi-institutional, multi-year Mendocino Triple Junction (MTJ)
Seismic Experiment. Line-9 is a high-resolution, large-aperture seismic
profile oriented in north-south direction, extending from the subduction
regime north of the MTJ to the transform regime south of the MTJ. Snapshots
of the wavefield are recorded during the simulation to create a movie which
assists with the analysis of the synthetic seismograms. The simulations are
computed on a 64 processor CM-5 parallel supercomputer.
First a velocity model obtained by the Mendocino Working Group is
examined. The model is based on information from depth-migrated
single-fold reflection profiles and tomographic inversion of first arrivals. The
synthetic seismic sections show acceptable first arrival traveltime fit but fail to
correctly model secondary arrivals and amplitude variations of the wavefield.
Based on the shortcomings of the initial model an alternative model for
the north-south line is proposed. The new model introduces several distinct
structures, like a subducted Gorda plate, a slabless window, and a wedge
shaped low velocity zone between Franciscan material and Gorda plate. The
big difference between the two models is that the initial model is too smooth
to contain any discontinuities and cannot account for any reflection arrivals. In
addition all structures in the lower crust of the alternative models are located
about 4 km deeper than the equivalent velocities in the initial model.
The seismic sections of the alternative model show much improved
traveltime and amplitude fit. Also, some characteristics of the wavefield
observed in the field data (e.g. traveltime oscillations, amplitude variations,
and discontinuous arrivals) are also observed in the synthetic seismograms..
The position of the southern end of the Gorda plate (SEDGE) is at 80 km in the
model (39.7°N). Because of the limited resolution it is not possible to resolve
whether the Gorda plate just north of the SEDGE is fragmented or continuous.
To add attenuation by scattering or coda waves generation to the model,
the effects of a stochastic velocity description for Franciscan rocks on the
wavefield are explored. The best velocity contrast between sandstones and
melange units in a bimodal, seifaffine, sinuous connected model is
determined as 0.5 km/s. Stochastic models of this kind successfully model
background noise and coda waves and to some degree discontinuous phases,
amplitude variations, and traveltime oscillations. They cannot, however,
account for the full range of characteristics observed in the field data. For this,
additional large scale velocity variations must be added to the model.
For the final model, the (deterministic) alternative model and the
statistical description of Franciscan rocks are combined. The resulting
wavefield shows good amplitude and traveltime fit and the observed
wavefield characteristics are similar to the characteristics of the field data.
