I define ``near-source'' to be a distance within several fault dimensions of the earthquake. At these distances near-field and direct body-wave contribution to ground motion are often the dominant arrivals [ Spudich and Frazer, 1984]. A distinguishing characteristic of the near-source region is that Green's functions vary strongly with position on the fault, which provides an opportunity for detailed imaging of the spatial and temporal distribution of slip. During the time of this report, there were several earthquakes in North America that were studied intensively using near-source data.
The October 18, 1989 Loma Prieta earthquake (
=
6.9) occurred before
the period of this report; however, much of the seismic source modeling of
this earthquake was published in 1991 (see the special issue of the Bulletin
of the Seismological Society of America, 1991). Near-source modeling of
this event to determine the spatial distribution of slip resulted in models
that were grossly similar, but with important differences [
Beroza, 1991; Steidl et al., 1991; Wald et
al., 1991]. Each of these studies found that slip was concentrated in two
zones about 15 km apart, one to the northwest and one to the southeast of
the hypocenter. Beroza [1991] found that slip was highest
to the southeast of the hypocenter and that the rake varied strongly from
predominantly strike slip to the southeast to predominantly reverse slip to
the northwest. Steidl et al.J[1991] found nearly equal
amounts of slip on both patches, with a more muted variation of rake.
Wald et al. [1991] found that slip was greatest to the
northwest of the hypocenter with nearly constant oblique rake over both
the high slip regions.
The June 28, 1992 Landers earthquake (
=7.2) was also the subject of
several near-source modeling efforts (see the special issue of the Bulletin
of the Seismological Society of America, 1994). Campillo
and Archuleta [1993] found that the broadband data could be explained
with a relatively simple rupture consisting of uniform slip on each of three
fault segments. They noted that the mainshock began rather weakly, as
did the Loma Prieta earthquake, and that although the mainshock
ruptured over several fault segments, there was some delay as it
transferred from one segment to the next. Cohee and
Beroza, [1994a] and Wald and Heaton [1994] both
published finite-source models for this earthquake based on
approximately 20 strong-motion records. Cohee and
Beroza [1994a] modeled the strong-motion data alone, whereas
Wald and Heaton [1994] modeled both the strong-motion
data and the geodetic data using a more complicated model. The slip
models in these studies were mostly similar; however, there were
differences, particularly in the rupture propagation models.
Cohee and BerozaJ[1994b] demonstrated that these
differences are attributable to the non-uniqueness inherent in the problem
and that the details of the rupture front propagation are difficult to discern
from such data sets.
Other earthquakes in North America that were studied using near-field modeling techniques during the last four years include the 1987 Whittier Narrows earthquake [ Hartzell and Iida, 1990; Zeng et al., 1993], the Sierra Madre earthquake [ Wald, 1992], and the 1985 Michoacan, Mexico earthquake [ Mendez and Anderson, 1991; Ruppert and Yomogida, 1992].
The frequencies of interest for earthquake engineering are typically greater than 1 Hz. This is a higher frequency band than can be modeled using the deterministic models described in the previous paragraphs. Zeng et al. [1993] and Cocco and Boatwright [1993] both used the envelope of high frequency accelerograms to map the source of high frequency accelerations on the fault. Zeng et al. [1993] applied the method to the 1989 Loma Prieta earthquake and found that sources of high frequency radiation tended to be located on the edges of sources of radiation at lower frequency. This is consistent with the idea that sudden changes in rupture velocity or slip amplitude are a primary cause of high frequency accelerations [ Madariaga, 1983].