Assimilation of Continuous GPS Data into Kinematic Fault Models

1
Assimilation of Continuous GPS Data into
Kinematic Fault Models
Jeff McGuire1, Paul Segall2, and Shinichi
Miyazaki3
1 Woods Hole Oceanographic Institution 2 Stanford
University 3 ERI, University of Tokyo
2
1996-1998 Bungo Channel Aseismic Slip
Transient
Time (yr)
3
Observation Equation
x Antenna Position Vs Secular Velocity s
Transient Fault slip
L Benchmark Wobble f(t) Reference Frame
error e estimation error
4
State Vector at Measurement Epoch K
Fault Slip
Benchmark Wobble
Reference Frame
Smoothing Parameters
5
Observation equation

• Separates spatially coherent signals from local
  ones
• Enforces positivity of slip-rate
• Enforces a spatially smooth solution

State Transition Equation
Current Fault Slip Model

• Encorporates random walk model of benchmark
  motion and slip-rate

6
Extended Kalman Filter
A Priori Estimate
Backsmoothed Estimate
Prediction
Residual
Update
7
(No Transcript)
8
Bungo Channel Event Summary

• Up to .6 m of slip over 1 year between depths
  of 30-60 km
• The Bungo channel event started about 1 month
  after the 2nd Hyuganada earthquake
• There was no propagation between the Hyuga-nada
  earthquake afterslip and the Bungo channel event
• The average stress-drop was about .15 M Pa
  (1/10th of a ordinary earthquake)
• A previously unidentified, Mw 6.5 event
  occurred in
• mid 1997 in the Hyuga-nada area

9
(No Transcript)
10
Displacement (mm)
Time from 1999.5 to 1999.9
11
1999 Cascadia Event Summary

• Slip-Pulse like propagation with a centroid
  velocity of 5 km/day
• Mw 6.9
• Maximum slip of 6 cm
• Slip confined between about 25 and 40 km depth
• Average Stress Drop of about .015 M Pa
• Stress Changes on the order of 10-5
  lithostatic drive rupture propagation over
  distances of 200 km.

12
The Network Inversion Filter and ACES

• Monitor the bottom boundary of the seismogenic
  zone
• Test constituitive laws for the aseismic region
  of faults just below the seismogenic zone.
• Encorporate a true dynamic model for
  xk1t(xk)
• -At what spatial/temporal scales does
  information about aseismic slip-rate become
  useful?