Resolving locked asperities and slip deficit in unlocked regions: A new inversion method applied in the South America subduction zone
Matthew Herman, Rob Govers, Department of Earth Sciences, Utrecht University Pseudo-coupling Model
TrenchOrigin
400 km
25º
Length
+x +y
+z
Fixed
x-z motion only
1 meter Width
100 km
1000 km
}
}
Trench
1 m0 100 200 300 400 500 600 700 800 900 1000
Along-Strike (km)
0 100 200 300 400
Along-Dip (km)
0.0 0.2 0.4 0.6 0.8 1.0
Slip (m)
0.2 0.4 0.6 0.8 1.0
Slip (m)
Slip (m)
0.0 0.2 0.4 0.6 0.8 1.0
0.0
Plate Interface Slip
Locked
Pseudo-coupled
Locked
Near the asperity, the
pseudo-coupled interface accumulates high slip deficit.
Farther from the asperity, the
plates slide at the relative velocity.
Outside the asperity, the interface is free to slide, but sliding is restricted by the adjacent locked zone.
Model Setup
We displace the top and bottom of the subducting plate 1 meter while holding the backstop of the upper plate fixed. No slip is allowed in locked asperities, but the rest of the interface is unlocked and can thus slide freely.
In this study, we incorporate the physics of pseudo-coupling into an inter-seismic inversion so that we can determine:
The continuous nature of tectonic plates implies that inter-seismic slip deficit must be continuous on the plate boundary. As a result, areas outside mechanically locked asperities can accumulate slip deficit even in the absence of shear resistance. We call this “pseudo-coupling” to distinguish it from mechanical coupling. Previously, we quantified its effect conceptually (Herman et al., 2018).
Where and how much of the subduction plate interface is locked?
What is the corresponding slip deficit accumulation rate?
Inversion Approach
(C)
(D)
0 100 200 300 400 500 600 700 800 900 1000
Along−Strike Distance (km)
0 100 200 300 400
Down−Dip Distance (km)
(C)
(D)
0 100 200 300 400 500 600 700 800 900 1000
0 100 200 300 400
Down−Dip Distance (km)
0.0 0.2 0.4 0.6 0.8 1.0
Slip (m)
0.0 0.2 0.4 0.6 0.8 1.0
Fault Slip (m)
0 100 200 300 400
Down−Dip Distance (km)
0.0 0.2 0.4 0.6 0.8 1.0
Fault Slip (m)
0 100 200 300 400 500 600 700 800 900 1000
Along−Strike Distance (km)
FEMInversion
Direct
Calculation FEM
Along-strike
Down-dip
Input
Fault Slip
0 100 200 300 400 500 600 700 800 900 1000
Along−Strike Distance (km)
0 100 200 300 400 100 2030 4050 6070
Slip (mm/yr)
Best Fault Slip
0 100 200 300 400 500 600 700 800 900 1000
0 100 200 300 400
Horizontal Distance From Trench (km)
χ2 misfit: 8.31e−01
Surface
Displacements
0 100 200 300 400 500 600 700 800 900 1000
0 100 200 300 400
1 2 5 10 20 50 100
χ2
0 2000 4000 6000 8000 10000 12000 14000 16000 18000 20000
Iteration
Probability Locked
0 100 200 300 400 500 600 700 800 900 1000
0 100 200 300 400
Horizontal Distance From Trench (km)
0.0 0.2 0.4 0.6 0.8 1.0
Locked Prob.
Misfit vs. Iteration
Accepted
Rejected
Model
We search for distributions of locked and unlocked patches that produce good fits to inter-seismic surface velocities. Since there are potentially many good fitting solutions, we implement an algorithm (Metropolis-Hastings) to determine the probability that each patch is locked. Key to this approach is that the fault slip distribution and corresponding velocities are based on the physics of pseudo- coupling in every iteration of the search.
Initialize locked and unlocked
patches
Calculate pseudo-coupling
and surface velocities Calculate proposed model velocity misfit: χ
pro2Randomly flip
< 5% locked and unlocked patches
Lower χ
2than
current model? Calculate p=exp(χ
pro2-χ
curr2)
rand(0-1)<p?
Set proposed model as current
model
Keep current model
Yes
No
Yes
No Propose model
(Repeat many times)
Proof of Concept The search involves running
10,000+ models. Each requires calculating the slip distribution and computing the misfit with respect to the observed velocities.
To accelerate the search, we use an analytical solution for directly minimizing the shear stresses to calculate slip around locked patches instead of an FEM . This produces similar fault slip and horizontal surface motions as the FEM. Biases associated with vertical displacements preclude their use in this model.
Inversion Algorithm Workflow
0 100 200 300 400 500 600 700 800 900 1000
0 100 200 300 400 500 600
Surface
Displacements
1922
1906 1906
1877 1868
1835 1822
1819 1784
1751 1746
1730 1687
1604
Nazca Plate South America Plate
South America Subduction Zone
The oceanic Nazca plate subducts eastward beneath
the continental South America plate from Chile to
Colombia. This subduction zone has hosted 12 Mw 7.5+
earthquakes in the past 25 years (red symbols indicate these earthquake epicenters and red lines show 2, 5, and 10 meter slip contours).
There is also a centuries-long historical record of great earthquakes (Kelleher, 1972) (rupture extents of these events are indicated by red bars west of the trench; Mw 8.0+ events are labeled with their dates).
The upper plate is densely instrumented by continuous and campaign GNSS stations measuring surface motions. We use the inter-seismic velocity field (measured before the Mw 7.5+ earthquakes) as the constraints on the plate interface locking distribution. We also test how much of the velocities can be explained by forearc sliver motions.
Seismotectonics
Geodetic Observations i ii
iii iv
Locking Rigid Total Pre.
Observed
20 mm/yr
95%
−76˚ −74˚ −72˚ −70˚ −68˚ −66˚ −64˚ −62˚
−38˚
−36˚
−34˚
−32˚
−30˚
−28˚
−26˚
−24˚
−22˚
−20˚
0 20 40 60 80
Back−slip (mm/yr)
i
21 mm/yrMedian:0 20 40 60 80
Back−slip (mm/yr)
ii
36 mm/yrMedian:0 20 40 60 80
Back−slip (mm/yr)
iii
27 mm/yrMedian:0 20 40 60 80
Back−slip (mm/yr)
iv
26 mm/yrMedian:i ii
iii
iv
v
−82˚ −80˚ −78˚
−18˚
−16˚
−14˚
−12˚
−10˚
−8˚
−6˚
−4˚
−2˚
0˚
2˚
4˚
0 10 20 30 40 50 60 70
Back−Slip (mm/yr)
0 20 40 60 80
Back−slip (mm/yr)
i
26 mm/yrMedian:0 20 40 60 80
Back−slip (mm/yr)
ii
22 mm/yrMedian:0 20 40 60 80
Back−slip (mm/yr)
iii
3 mm/yrMedian:0 20 40 60 80
Back−slip (mm/yr)
iv
25 mm/yrMedian:0 20 40 60 80
Back−slip (mm/yr)
v
34 mm/yrMedian:−76˚ −74˚ −72˚ −70˚ −68˚ −66˚ −64˚ −62˚
−38˚
−36˚
−34˚
−32˚
−30˚
−28˚
−26˚
−24˚
−22˚
−20˚
0.0 0.2 0.4 0.6 0.8 1.0
Fraction Locked
Median: 0.24
0.0 0.2 0.4 0.6 0.8 1.0
Fraction Locked
Median: 0.10
0.0 0.2 0.4 0.6 0.8 1.0
Fraction Locked
Median: 0.21
−82˚ −80˚ −78˚
−18˚
−16˚
−14˚
−12˚
−10˚
−8˚
−6˚
−4˚
−2˚
0˚
2˚
4˚
0.0 0.2 0.4 0.6 0.8 1.0
Probability Locked
0.0 0.2 0.4 0.6 0.8 1.0
Fraction Locked
Median: 0.08
0.0 0.2 0.4 0.6 0.8 1.0
Fraction Locked
Median: 0.01
0.0 0.2 0.4 0.6 0.8 1.0
Fraction Locked
Median: 0.16
Locking Probability Slip Deficit Rate
Frequency
Frequency
0.0 0.1 0.2 0.3 0.4
Angular Velocity (°/Ma)
Chile Sliver
Frequency
Frequency
0.0 0.1 0.2 0.3 0.4
Angular Velocity (°/Ma)
Peru Sliver
Frequency
Frequency
0.0 0.1 0.2 0.3 0.4
Angular Velocity (°/Ma)
Ecuador−Colombia Sliver
Colombia-Ecuador
Northern Peru
Southern Peru
Northern Chile
Central Chile
Southern Chile 5-10% of area locked
Slip deficit rate: 10-60 mm/yr Sliver translation: 1-2 mm/yr
1-2% of area locked
Slip deficit rate: < 5 mm/yr Sliver translation: 2-3 mm/yr
14-18% of area locked
Slip deficit rate: 25-70 mm/yr 2001 Mw 8.4
2007 Mw 8.0
20-30% of area locked
Slip deficit rate: 20-70 mm/yr 2014 Mw 8.1
10% of area locked
Slip deficit rate: 15-75 mm/yr 2015 Mw 8.3
Sliver translation: 3-4 mm/yr
20-30% of area locked
Slip deficit rate: 30-75 mm/yr 2010 Mw 8.8
There are systematic
variations in the locking probability along strike.
Locking correlates well with locations of great megathrust events.
Recent earthquakes
nucleate in and around the edges of
high-probability locked zones. They rupture
into regions with slip deficit rates of 50%
or more of the
convergence rate.
The largest events rupture through
multiple asperities.
High
pseudo-coupling
Medium pseudo-coupling Low pseudo-coupling
Upper plate Trench
Subducting Plate
Inter-seismic
Slip Deficit
Co-seismic
Slip Magnitude
Fully locked asperities accumulate slip deficit at the convergence rate.
Earthquakes begin in and around these regions.
Seismicity No seismicity
Trench