I use the adjective ``teleseismic'' to denote events recorded at distances of more than 30 degrees for which the turning points of direct body-wave phases occur in the lower mantle. Green's functions for intermediate to long period body waves are quite simple at these distances and have long been used to determine source properties.
The Harvard CMT juggernaut rolls on [ Dziewonski et al., 1991-1994]. The Harvard group completed their 10,000th moment tensor solution during this quadrennium. With source mechanisms for approximately 650 earthquakes each year during the period 1977-1993, the CMT catalog represents the most complete catalog of global earthquake mechanisms. More recently the US Geological Survey has undertaken a similar effort [ Sipkin and Needham, 1991-1994]. These two global catalogs of source mechanisms provide first-order information on active tectonics as expressed by earthquakes worldwide. Preliminary results are disseminated via email over the internet in a service that continually keeps seismologists apprised of global seismic activity.
The tremendous success of these efforts in cataloging basic point-source parameters (location, origin time, moment tensor, seismic moment) of earthquakes at teleseismic distances contrasts sharply with attempts to estimate higher-order source parameters that involve spatial or temporal dimensions (slip duration, directivity, stress drop, moment-rate function). Methods to recover these parameters remain an area of active research.
A good example of the need for more work in this area is the Macquarie
Ridge earthquake of moment magnitude (
) 8.1. The point-source
parameters of this earthquake were well determined and consistent in
different studies; however, the higher-order source parameters were quite
different. Ekström and Romanowicz [1990] used
broadband data to determine that the moment-rate function was
concentrated in the first 30 s of the earthquake, although fault slip
continued for about 60 s after the origin time. They also noted a lack of
directivity in the S waves. In contrast to this reported lack of directivity,
Satake and Kanamori [1990] performed a deconvolution
of the P-waves and found evidence for a progression of 4 subevents from
south to north at high rupture velocity during the earthquake (i.e. strong,
unilateral directivity). More recently Tada et al.J[1993]
used array analysis of short period P waves on a small-aperture array in
Japan to argue for bilateral rupture during this earthquake.
Tichelaar and Ruff [1990] found that nearly all of the seismic moment
was released in the first 20 s and inferred a very large slip in the
earthquake of 36 m over a 50-km long zone. Their high inferred stress
drop of 37 MPa contrasts sharply with the apparent stress drop inferred
by Houston [1990] of 2-5 MPa. Anderson
and Zhang [1992] interpreted the 16-km deep centroid as evidence for
substantial co-seismic slip within the oceanic mantle.
Ihmlé and Jordan [1993] found evidence for a slow precursor from the
time shift spectrum of very-long-period normal modes, yet
Kedar et al. [1994] found none. The wide variation of inferred source
behavior found in these studies underscores the need for further research.
An interesting approach to determining the moment-rate function at
teleseismic and regional distances is the use of empirical Green's function
deconvolution. If a small event occurs near a much larger event and has
the same mechanism, then a deconvolution of the smaller event from the
larger event gives an apparent moment-rate function at that source-receiver
azimuth that is approximately valid at frequencies where the
finite size of the smaller event does not distort the waveform (i.e. below
the corner frequency). Ammon et al. [1993] applied this
technique to the 1992 Landers and 1992 Cape Mendocino earthquakes and
used the azimuthal dependence of the moment-rate function to resolve the
fault plane from the auxiliary plane. Kanamori et al. [1992] used a similar
method at regional distances in studying the Landers event.
Li and ToksözJ[1993] and Ammon et al.
[1994] also used a similar method to study the 1992 Colombia surface
wave magnitude(
) 7.3 earthquake. Jones et al. [1993]
used empirical Green's function deconvolution to demonstrate faulting on
conjugate planes during the 1992 Big Bear earthquake (
=6.4).
Di Bona and Boatwright [1991] applied an iterative
deconvolution of empirical Green's functions. Their method differed from
other empirical Green's function deconvolution methods because it
incorporated both tests of significance and positivity constraints; however,
they applied it to local earthquakes rather than teleseisms.
Other techniques for determining extended source parameters of earthquakes at teleseismic distances were developed and applied during this period. Kikuchi and Kanamori [1991] extended a previous deconvolution technique to include changes in both the moment-rate function and earthquake mechanism. Kikuchi et al. [1993] applied this technique to the Spitak, Armenia earthquake and found evidence for a slow afterslip event. Hartzell and Mendoza [1991], Hartzell et al. [1991], Hartzell and Langer [1993], Mendoza [1993], Mendoza et al. [1994] studied large teleseismic events by performing inversions for slip amplitude on a planar fault.