Estimating fault slip rates, locking distribution, elastic/viscous properites of lithosphere/asthenosphere. Kaj M. Johnson Indiana University

Transcription

1 3D Viscoelastic Earthquake Cycle Models Estimating fault slip rates, locking distribution, elastic/viscous properites of lithosphere/asthenosphere Kaj M. Johnson Indiana University In collaboration with: Paul Segall, Stanford; Ray Chuang, Indiana; Junichi Fukuda, Indiana University (now at ERI)

2 Conceptual Model of Lithosphere Approach: develop forward model and inverse method to obtain joint estimation of (1) long-term fault slip rates (2) Interseismic creep rates (3) Location of locked patches (4) Lithosphere viscosity (5) Elastic thickness

3 Methodology/Philosophy Forward Model Inverse Method Need fast models Bayesian, probabilistic Semi-analytical methods large number of unknown model propagator matrix method parameters no optimization want boundary element methods for posterior probability distributions stress boundary conditions Monte Carlo sampling need to compute 100 K s of forward computations

4 Talk Outline Geologic/Geodetic fault slip rate discrepancy in southern California model dependence on slip rate estimate Illustration with 2D models (infinitely long faults) Inversions with 2D and 3D models -- reconcile discrepancy San Francisco Bay Area interseismic deformation inversion for: Fault slip rate Locking distribution Asthenosphere viscosity Earthquake repeat times

5 Slip Rate Discrepancies in Southern California Becker et al. (2004), Meade et al. (2005)

6 Standard Elastic Block Model e.g., Jim Savage, Rob McCaffrey, Brendan Meade Rigid Body Rotation (no fault locking) Elastic distortion due to fault locking

7 Viscoelastic Coupling Model Savage and Prescott (1978) T/t R = 4

8 2D Models of Interseismic Deformation Screw dislocation Savage and Burford (1978) 20 km 10 km Viscoelastic coupling model Savage age and Prescott (1983) T/tR = 5 Teq/T = 0.8

9 2D Models of Interseismic Deformation 20 km Viscoelastic coupling model Savage and Prescott (1983) T/tR = 5 Teq/T = 0.8

10 2D Models of Interseismic Deformation 20 km 10 km T/tR = 5 Teq/T = 0.8 Stress-driven creep model Johnson and Segall (2004)

11 2D Models of Interseismic Deformation teq/t = km T/tR = 1 Johnson et al (2007) T/tR = 5

12 Viscosity Varies with Depth Thatcher h and Pollitz (2008) T = 250 yrs Johnson et al. (2007)

13 Fault Slip Rates and Interseismic Deformation: Mojave Region Johnson et al. (2007)

14 Data Paleoseismic record GPS -- contemporary triangulation Landers postseismic i GPS

15 2D Viscoelastic Earthquake Cycle Model Mojave segment e of SAF system model Data: GPS Historical triangulation Paleoseismic s data Landers postseismic Inversion Estimate: slip rates viscosity Johnson, Hilley, and Bürgmann (2007)

16 2D Viscoelastic Earthquake Cycle Model Mojave segment e of SAF system model Data: GPS Historical triangulation Paleoseismic s data Landers postseismic Inversion Result: slip rate: mm/yr geology: mm/yr Elastic block models: mm/yr Johnson, Hilley, and Bürgmann (2007)

17 2D Viscoelastic Earthquake Cycle Model Inversion eso Results esuts

18 Lithosphere Viscosity Thatcher and Pollitz (2008)

19 model Faults are not 2D!

20 3D Earthquake Cycle Block Model paper in prep.

21 3D Earthquake Cycle Block Model

22 Construction of Steady State Velocity Field

23 3D Cycle Block Model for Southern California

24 Fault Geometry Based on SCEC Community Fault Model Rupture Segments

25 Interseismic Perturbation 15 km

26 Interseismic Perturbation 15 km tr = 25 yr

27 Interseismic Perturbation 15 km tr = 250 yr tr = 5 yr

28 Inversion Results: strike-slip rate (mm/yr) Geologic rates Elastic block model 15 km

29 Inversion Results: strike-slip rate (mm/yr) Geologic rates Viscoelastic Vscoeas c block model 15 km tr = 250 yr tr = 5 yr

30

31 San Francisco Bay Area GPS and geologic data

32 San Francisco Bay Area Triangulation data: post 1906 transient

33 Probabilistic Bayesian Inversion Posterior distribution Likelihood Prior distribution = k. x geodetic data & model geologic data p(, n, w GPS, w TRI d) = p(d, n, w GPS, w TRI ) p(, n, w GPS, w TRI ) Euler poles (linear) Relative weights on data data Nonlinear parameters

34 Algorithm Monte Carlo with Metropolis Step discrete samples marginal

35 Mixed Linear/Nonlinear Inversion p(, n, w GPS, w TRI d) = p( d, n, w GPS, w TRI ) x p( n, w GPS, w TRI d) Euler poles (linear) data Relative weights on data Nonlinear parameters obtain with least squares obtain analytical expression sample with Monte Carlo algorithm

36 3D Viscoelastic Cycle Model

37 Inversion Result: Fault slip rates

38 Inversion Results:

39 Inversion Result: Earthquake Recurrence Times

40 Fit to GPS data residual velocities

41 Conceptual Model of Lithosphere Approach: develop forward model and inverse method to obtain joint estimation of (1) long-term fault slip rates (2) Interseismic creep rates (3) Location of locked patches (4) Lithosphere viscosity (5) Elastic thickness

42 Simultaneous Estimation of Long-term Slip Rates and Locking Distribution

43 Simultaneous Estimation of Long-term Slip Rates and Locking Distribution

44 Simultaneous Estimation of Long-term Slip Rates and Locking Distribution assume steady asthenosphere flow fault stressing rate is constant

45 surface velocities Creep Rate Linearly Related to Euler Poles creep rate long-term slip rate boundary element creeping contribution tibti locked contribution tibti

46 Inversion Result Distribution of locked and creeping patches Interseismic creep rate

47 Distribution of locked and creeping patches locked creeping seismicity rate

48 Creep Rate Varies with Time 20 km 10 km Johnson and Segall (2004)

49 Creep Rate Below Locking Depth 20 km 10 km 20 km 10 km T/tR = 5 Teq/T = 0.8 different locking depths would be inferred

50 Incorporating Stress-driven Creep in Cycle Model

51 Incorporating Stress-driven Creep in Cycle Model

52 Incorporating Stress-driven Creep in Cycle Model

53

54 Stress-driven Interseismic Creep Inversion Result Distribution of locked and creeping patches approximate stress-drive creep model

55 Conclusion We nearly have a method to simultaneously estimate: (1) long-term fault slip rates (2) Interseismic creep rates (3) Location of locked patches (4) Lithosphere viscosity (5) Elastic thickness