Geology Reference
In-Depth Information
earthquake to the next on the Garlock Fault
(McGill and Sieh, 1991). Obviously, if the
displacement histograms display more than one
peak and are interpreted to record more than
one earthquake, some geomorphic evidence
should be sought to indicate that those features
showing greater displacements relate to older
earthquakes in comparison to features along the
same fault segment with smaller displacements.
Third, given the length of each ruptured fault
segment and the mean displacement along it, a
seismic moment can be estimated, if the depth of
the fault and a crustal rigidity are assumed or
known (Eqn 6.3). Often, modern seismicity
defines the local depth of the seismogenic layer
rather well, and, because most large earthquakes
will rupture through the entire seismogenic layer,
reasonable estimates of the seismic moment can
be made. Various rupture scenarios can be evalu-
ated along a segmented fault, ranging from rup-
ture of its entire length to rupture of the shortest
segment (McGill and Sieh, 1991). In each case,
the mean displacements for the relevant fault
segment are fed into the calculation, and the
moments for each segment are summed to yield
a total seismic moment and moment magnitude
for the rupture scenario. Such an approach clearly
provides a very useful data set for assessing seis-
mic hazards along such a fault. No consensus
exists, however, on when earthquakes tend to
rupture beyond segment boundaries. No histori-
cal strike-slip faults have been observed to rup-
ture across steps of more than 3-4 km between
segment terminations (Wesnousky, 2006),
whereas quite a few historical thrust faults have
ruptured multiple fault segments lying as much
as 15 km apart (Dong
et al.
, 2008; Rubin, 1996).
Hence, assessments of simultaneous ruptures of
multiple segments during prehistoric earthquakes
and their implications for future temblors will
require exceptionally tight dating control to dem-
onstrate rupture synchrony in the past.
One potentially controversial aspect of using
multiple offset features to characterize fault dis-
placements in several different earthquakes is
the fact that the age of most, if not all, of the
features is commonly unknown. At some sites, it
is possible to show that two or more sets of
features are displaced by distinctly different
amounts (Fu
et al.
, 2005; Liu
et al.
, 2004; Sieh
and Jahns, 1984) and can, therefore, confidently
be interpreted to represent two or more earth-
quakes. More commonly, individual displace-
ments all along the fault are amalgamated and
compared, with little knowledge of their relative
age. If they seem to fall into clusters, especially
ones with displacements that are multiples of
each other (Fig. 6.20), a natural tendency is to
interpret each cluster as representing the coseis-
mic slip that was added to the penultimate earth-
quake's cumulative slip. Whereas this strategy
seems reasonable, data from the Landers earth-
quake ( June 1992,
M
w
=
7.3) in southern California
suggest some pitfalls with this approach.
Although the total Landers rupture was
approximately 85 km in length, the map pattern
clearly indicates that the rupture comprises an
elongate zone of several known faults that were
linked together either by previously unrecog-
nized or by newly formed faults. Mapping of off-
sets along one of the previously known faults
(central Emerson Fault; Fig. 6.21A) showed
abrupt changes in the amount of offset along
its 5-km-long trace (McGill and Rubin, 1999).
Numerous ephemeral features, such as offset tire
tracks, indicate that the displacements all
occurred during the 1992 earthquake. When the
displacements were synthesized in a histogram,
however, they showed two discrete peaks
(Fig. 6.21B) at about 2 m and 4 m. In a paleoseis-
mic study, such peaks would probably be inter-
preted as being due to two different earthquakes,
rather than resulting from variations along a
short rupture during a single event. One approach
to try to determine if more than one event is
encompassed within the suite of offset features is
to subtract a running mean from each data point
(McGill and Rubin, 1999). For many data sets, the
remaining anomalies will define a single-peaked
probability distribution if the data derived from a
single event, whereas they may display multiple
peaks for more than one event. Nonetheless,
when the data are particularly noisy and vary
abruptly in space, this subtraction procedure can
still produce two peaks for a single earthquake
(as is the case with the Emerson Fault data),
Clearly, some caution is warranted when
interpreting histograms of offset features.
