HR: 15:10h
AN: T42C-07 [PDF]
TI: 3-D Numerical Modeling of Rupture Sequences of Large Shallow Subduction Earthquakes
AU: * Liu, Y
EM: liu@esag.harvard.edu
AF: Department of Earth and Planetary Sciences,
Harvard University, 327 Pierce Hall,
29 Oxford Street, Cambridge, MA 02138 United States
AU: Rice, J R
EM: rice@esag.harvard.edu
AF: Department of Earth and Planetary Sciences,
Harvard University, 327 Pierce Hall,
29 Oxford Street, Cambridge, MA 02138 United States
AU: Rice, J R
EM: rice@esag.harvard.edu
AF: Division of Engineering and Applied Sciences,
Harvard University, 224 Pierce Hall,
29 Oxford Street, Cambridge, MA 02138 United States
AB:
We study the rupture behavior of large earthquakes on a 3-D shallow subduction fault governed by a rate and state friction
law, and loaded by imposed slip at rate $V_{pl}$ far downdip along the thrust interface. Friction properties are temperature,
and hence depth, dependent, so that
sliding is stable ($ a - b > 0$) at depths below about 30 km. To perturb the system into a nonuniform slip mode, if such a
solution exists, we introduce small along-strike variations in either the constitutive parameters $a$ and $(a - b)$, or the
effective normal stress, or the
initial conditions. Our results do show complex, nonuniform slip behavior over the thousands of simulation years. Large
events of multiple magnitudes occur at various along-strike locations, with different recurrence intervals, like those of
natural interplate earthquakes. In the model, a large event usually nucleates in a less well locked gap
region (slipping at order of 0.1 to 1 times the plate convergence rate $V_{pl}$) between more firmly locked regions (slipping
at 10$^{-4}$ to 10$^{-2}$ $V_{pl}$) which coincide with the rupture zones of previous
large events. It then propagates in both the dip and strike directions. Along-strike propagation slows down as the rupture
front encounters neighboring locked zones, whose sizes and locking extents affect further
propagation. Different propagation speeds at two fronts results in an asymmetric coseismic slip distribution, as is
consistent with the slip inversion results of some large subduction earthquakes [e.g., Chlieh et al., 2003].
Current grid resolution is dictated by limitations of available computers and algorithms, and forces us to use constitutive
length scales that are much larger than realistic lab values; that causes nucleation sizes to be in the several kilometers
(rather than several
meters) range. Thus there is a tentativeness to present conclusions. But with current resolution, we observe that the
heterogeneous slip at seismogenic depths (i.e., where $a - b < 0$ ) is sometimes accompanied
by events that have clearly aseismic slip rates (10$^{2}$ to 10$^{3}$ $V_{pl}$), and seem to migrate in the strike direction
at depths slightly downdip from the seismogenic zone. Sometimes they reach a gap like described above in the shallow locking
region and rupture the updip
seismogenic zone. The frequency-size distribution of simulated large events on the thrust interface shows an increase in
slope around $M_w = 7.5$, like has been found for actual earthquakes [Pacheco and Sykes, 1992]. We are also
investigating the relation of hydrolytic reactions to the hydrologic state of the thrust interface, and their possible
relation to observed non-volcanic tremors and aseismic transients downdip of the seismogenic zone [Obara, 2002; Dragert et
al., 2001], by coupling the fluid motion induced pore pressure changes into the system.
DE: 7209 Earthquake dynamics and mechanics
DE: 7260 Theory and modeling
DE: 8010 Fractures and faults
DE: 8150 Plate boundary--general (3040)
SC: Tectonophysics [T]
MN: 2003 Fall Meeting