Numerical simulation of seismic cycles at a subduction zone with a laboratory-derived friction law

1
Numerical simulation of seismic cycles at a
subduction zone with a laboratory-derived
friction law

• Naoyuki Kato(1), Kazuro Hirahara(2), and Mikio
  Iizuka(3)
• (1) Earthquake Research Institute, University of
  Tokyo
• (2) Graduate School of Environmental Sciences,
  Nagoya University
• (3) Research Organization for Information Science
  and Technology

2
Numerical simulation of seismic cycles at a
subduction zone with a laboratory-derived
friction lawOutline

• 1. Composite rate- and state-dependent friction
  law
• 2. Forecast models for the Tokai earthquake
• 3. Mechanism of episodic slip events
• 4. Historical earthquake sequence along the
  Nankai trough
• - Future study -

3
Rate- and state-dependent friction law m m
a ln(V/V) b ln(Vq/L)
m friction coefficient V sliding velocity q
state variable a,b,L, m,V constants
dq/dt 1 - qV/L
slowness law
dq/dt - (qV/L) ln (qV/L)
slip law
dq/dt exp(-V/Vc) - (qV/L) ln (qV/L)
composite law Vc constant
Kato and Tullis (2001)
4
Simulation results of slide-hold-slide test
Simulation results of velocity stepping test
Symmetric responses to velocity increases and
decreases for the slip and composite laws explain
laboratory data
Stiffness independent higher healing rates for
the slowness and composite laws explain
laboratory data
Kato and Tullis (2001)
5
Simulation of stick-slip cycle of a spring-block
system
The recurrence interval and stress drop are the
largest for the composite law.
Kato and Tullis (2002)
6
Model for seismic cycle at the Tokai seismic gap
along the Suruga trough, central Japan
Seismic moment release of preseismic sliding
seismic gap
Back slip distribution estimated from GPS data by
Sagiya (1999)
2-D model for the Tokai earthquake Kato and
Hirasawa (1999)
Kato and Tullis (2002)
7
3-D model for the Tokai earthquake
Strain change (for one day before EQ)
Strain (1E-8)
Kuroki et al. (2002)
8
Hamamatsu
Episodic strain event near the hypothesized
source area of the Tokai earthquake detected by
GPS Data from Geographical Survey
Institute (http//www.gsi.go.jp)
Hamamatsu
9
Slow slip event beneath the Bungo channel,
southwest Japan Hirose et al. (1999)
Velocity changes for 3 years in southwest Japan
(c) Apr. 1, 1998 - Apr. 1, 1999
(b) Apr. 1, 1997 - Apr. 1, 1998
(a) Apr. 1, 1996 - Apr. 1, 1997
Observed and model horizontal displacement
Distribution of estimated fault slip over 10
months Moment magnitude 6.6
Saiki
Misho
Saiki
Misho
10
Possible mechanism of episodic slip event on a
plate interface For a spring-block model, the
ratio of spring stiffness to the critical
stiffness kc (B-A)/L controls sliding mode.
(B-A -dtss/dlnV, L characteristic slip
distance) k gtgt kc ? aseismic k kc ?
episodic k ltlt kc ? seismic For a finite
fault in a uniform elastic medium, the critical
fault dimension rc cGL/(B-A) may be defined,
and r/rc controls sliding mode. (c constant
1, G rigidity) r gtgt rc ? seismic r rc ?
episodic r ltlt rc ? aseismic
k
11
Simulation of slip on a flat fault in an infinite
uniform elastic medium A circular patch with
negative A-B value is embedded in a uniform fault
with positive A-B value.
The critical fault radius rc 4.12 km. The
radius r of the negative A-B patch is 3.0
km. r/rc 0.73
stable
Possibly unstable
Distribution of A-B (MPa)
12
Snapshot of distribution of slip velocity
ln(V/Vpl) (Vpl the assumed plate velocity 4
cm/yr)
The negative A-B patch is more strongly locked.
13
Snapshot of distribution of slip velocity
ln(V/Vpl) Time from the last slide 5.7 years
Episodic slip starts in the negative A-B patch.
14
Snapshot of distribution of slip velocity
ln(V/Vpl) Time from the last slide 13 hours
Episodic slip propagates in the negative A-B
patch, but the slip is not significantly
accelerated.
15
Snapshot of distribution of slip velocity
ln(V/Vpl) Time from the last slide 32 days
Slip is decelerated because the rupture front
enters the positive A-B region.
16
Snapshot of distribution of slip velocity
ln(V/Vpl) Time from the last slide 150 days
Very slow slip propagates in the positive A-B
region, while healing starts in the negative A-B
patch.
17
Snapshot of distribution of slip velocity
ln(V/Vpl) Time from the last slide 2.2 years
The negative A-B patch is strongly locked and the
next cycle starts.
18
Slip history during the entire cycle
r 3 km, rc 4.12 km A-B 0.2 MPa at 1-6 A-B
-0.2 MPa at 7-8
19
Slip history at the aseismic slip event
r 3 km, rc 4.12 km A-B 0.2 MPa at 1-6 A-B
-0.2 MPa at 7-8
TIME (days)
20
Summary of simulation for episodic slip
events Aseismic episodic slip event may be
simulated for a negative A-B patch with r/rc
0.73. Episodic events with various time duration
may be simulated by varying r/rc-value. The
duration of the event decreases with an increase
in r/rc. Episodic events may be simulated not
only by nonuniform distribution of A-B but also
by nonuniformity in the characteristic slip
distance L or in the effective normal stress. If
the source size and duration of an episodic event
are obtained, the value of (B-A)/L may be
estimated.
21
Earthquake Generation and Strong Motion
in 3-D
Heterogeneous Media
Quasi-Static Modeling of Earthquake Cycle
Interaction on and between Faults
Fault Constitutive Law
Wave Propagation in Heterogeneous Media
Dynamic Modeling of Earthquake Rupture
Simulation and Prediction of Strong Motion
Interplate Earthquake Fault
Inland Active Fault
Frictional Law
Plate Subduction
Viscoelastic Interaction
22
(No Transcript)
23
(No Transcript)
24
(No Transcript)
25
In the Earth Simulator Project, we take into
consideration heterogeneous viscoelastic
structure, a rate- and state-dependent
friction, and interactions of many segments
of plate boundary and inland active
faults. FEM Iizuka, Poster Solving
friction problem Prabhakar, Poster
26
Quasi-Static Earthquake Cycle Simulation with
GeoFEM
3D-FEM Mesh
750km
1150km
200km
No. of Nodes 11466 No. of Elements10000
1100km
900km
200km
Crust Plate Upper Mantle
No. of Nodes 59400 No. of Elements54752