Andrew V. Newman, Jaime A. Convers, Hermann Fritz Georgia Institute of Technology, Atlanta, GA, USA Lujia Feng Earth Observatory of Singapore, NTU, Singapore.

Andrew V. Newman, Jaime A. Convers, Hermann FritzAndrew V. Newman, Jaime A. Convers, Hermann FritzGeorgia Institute of Technology, Atlanta, GA, USAGeorgia Institute of Technology, Atlanta, GA, USA

Lujia FengLujia FengEarth Observatory of Singapore, NTU, SingaporeEarth Observatory of Singapore, NTU, Singapore

Ting ChenTing ChenWuhan University, ChinaWuhan University, China

Gavin HayesGavin HayesNational Earthquake Information Center, USGS, USANational Earthquake Information Center, USGS, USA

Yong WeiYong WeiPacific Marine Environmental Lab, NOAA, USAPacific Marine Environmental Lab, NOAA, USA

The Character of Tsunami-genesis in Subduction zone earthquakes, with application to real-time seismic and geodetic warning

• Background on Tsunamigenic and Tsunami Earthquakes

• Identifying Earthquake Energy for real-time earthquake magnitudes and tsunami warnings

• Seismic and geodetic displacement in tsunamigenic earthquakes.

Examples from:– Sumatra (2004, 2005, 2010)

– Solomon Islands (2007, 2010)

– Chile (2010)

– Japan (2011)

• We are working to understand shallow trench rupture, its rupture potential, and imminent warning when it does.

A few key findings:– Still no clear answers for why megathrusts sometimes fail near the trench – The free surface allows for enhanced slip (breaking scaling laws)

– Splay faults are unnecessary

– Most shallow rupture is identifiably slow

Outline:

• Most megathrust events rupture between ~20-50 km depth

• Occasional tsunami earthquakes rupture almost exclusively shallower than 20 km [Polet and Kanamori; 2000].

– These events are problematic because they aren’t usually identified until after the fact

tsunami generation:

are “Earthquakes that create tsunamis much larger than expected given their initial magnitude” –Kanamori [1972].

Rupture in shallow increases tsunami-potential

• free-surface effect causing “unhinged slip” [Satake and Tanioka, 1999]

• Possible rupture on splay fault can enhance vertical displacement [Moore et al., 2008]

• Slowed rupture in Tsunami earthquakes region may enhance slip due to momentum conservation [Newman, unpublished] (e.g. end of a whip)

tsunami generation:

Further correcting for mechanism, depth and distance: Newman and Okal [1998]

E determinations: – All M0=1019 Nm

– 1997 – mid-2010

– Using full rupture duration

Results:– Megathrust ruptures have low

stress drop (reduced E/M0)

- ID new slow events (e.g. NB)

- Middle America Trench is E deficient

Radiated seismic energy

Using P-waves (25°≤ ≤ 80° )

[Convers and Newman, 2011]

Standard Earthquake

M~7.0

Slow-source Tsunami Earthquakemb ~5.8, MS ~7.0, MW~7.7

Energy Deficiency of Tsunami Earthquakes

Poorly consolidated and water rich sediments near the trench reduce rigidity

Normal Vs ~3.0 km/sNear-trench Vs ~.5-1.5 km/s

Shallow slownessGlobal survey of subduction zone events found shallow events have increased rupture duration (reduced rigidity, )

[Bilek & Lay, 1999]

An improved TsE Discriminant• E/M0 method requires a reliable Moment solution

which may take 30 min or more - too late for nearby environments where

• Instead, we aim to ID the slowness of these earthquakes

Theoretical observed rupture duration

Duration

Decay is somewhat dependent on locale

+ o

Energy Scattering

M 5

.5

Time ->

Theoretical Modeling

Total energy at site over time

Cumulative energy at site

Time ->

Theoretical Modeling

Constant Rupture Model gives shape of Energy Growth

X-over overestimates duration

However, Source-Time functions for large earthquakes are not usually box-car functions, and have significant slip decay before cessation.

S. Bilek, pers. Comm.

Method rapidly ID’s Tsunami Earthquakes

Discriminant: C=E/Tr3 (∝ E/M0)

[Newman and Convers, in revision]

For TsE, evaluating the ratio of energy to rupture duration (Tr) yields potentially more robust information given that TsE are:

- Low E (~10x) - Long Tr (~ 3x)

For events larger than 7 only TsE have E/Tr3 < 5e7

Particularly useful because both can be evaluated from just the P-wave energy growth.

Method rapidly ID’s Tsunami EarthquakesDiscriminant: C=E/Tr3 (∝ E/M0)

Real-time: http://geophysics.eas.gatech.edu/RTerg

Routinely and automatically calculate E and TR for all earthquakes with initial magnitude ≥ 5.5 (early 2009 to present)

Automatic webpage is generated and results are disseminated to the PI, students, USGS, and Pacific Tsunami Warning Center. (Feeds through email and text messages)

1st results often within 10 minutes of rupture, with automatic updates following.

Stand-alone version is now operational at the PTWC (GT system is research mode).

Earliest tsunami waves take 30 m or more…thus the methods can supply ample warning!

2010 Mw 7.8 Mentawai Earthquake• >400 fatalities (from tsunami)• many survivors thought M6• initial reports of 5-9 m runup • possible history of TsE here

with 1907 M7.6 recently identified [Kanamori et al., 2010]

Summary:• Event was a “Tsunami EQ” (TsE)

• Large tsunami given magnitude

• Shallow rupture• Slow TsE

• Low E/M0

• Long duration• Tsunami excitation can be

estimated from FFM w/ rupture velocity correction

Newman et al., GRL [2011]

Real-time detection1: Nucleation + 8:13 [m:s]2: Nucleation + 10:52 [m:s]3: Nucleation + 15:45 [m:s]4: Nucleation + 21:15 [m:s]5: Nucleation + 26:55 [m:s]

Newman et al., GRL 2011

Finite fault solution: 2010 Sumatran EQ

P, SH, Rayleigh and Love Waves modeled

Solution on 11.6° dip- max slip near 2 m updip of nucleation point- rupture duration, TR=125 s- M0=6x1020 Nm (MW=7.75)

Newman et al., GRL 2011

Slip scaling by VR

Shear velocity:

Upper/Lower plate params.

Slip

Rigidity

Density (constant)

Shear wave velocity

Rupture velocity ( ~0.8)

Upper plate slip:

For constant M0

Slip scaling by VR

€

Du = D lβ l

1.25VR

⎛

⎝ ⎜

⎞

⎠ ⎟

2

€

2 = μ ρ

Shear velocity:

€

u, l

D

μ

ρ

β

VR

Upper/Lower plate params.

Slip

Rigidity

Density (constant)

Shear wave velocity

Rupture velocity ( ~0.8)

Upper plate slip:

€

Duμ u = D lμ l

For constant M0

Tsunami models

Original Finite-Fault

Unscaled FFM max local runup < 3 m

Scaled FFM max local runup +10 m

Open-ocean wave heights well predicted well fit in period and timing but over-predict amplitude

Scaled solution

Tsunami models

Ocean-bottom pressure sensor is along highly attenuated path (along trench and near trend of strike)

Other inversion results:

P, SH waveform inversion of Lay et al., 2011

Case study: Solomon IslandsSolomon Islands earthquakes as a subduction end-

member

– Rapid convergence of very young crust allows normal megathrust seismogenesis in the very shallow subduction environment

– Near-trench land ideal for shallow seismogenic zone geodetic measurements

1 April 2007 MW 8.1 Solomon Islands megathrust earthquake and tsunami

1 April 2007 MW 8.1 Solomon Islands Earthquake

Taylor et al., Nat. Geosc. 2008Uniform 5m slip model with steep geometry (=38°)

1 April 2007 MW 8.1 Solomon Islands Earthquake

Data from Fritz and Kalligeris 2008

Flow-depth less affected by localized “splash-up”

1 April 2007 Mw 8.1 Solomon Islands Earthquake

Data: Coral uplift, coastal subsidence( Fritz & Kalligaris, 2008; Taylor et al., 2008)

All data collected between 2 weeks and 1 month after event

Chen et al. GRL, 2009

Model Deformation:• We apply the analytic method of describing elastic

deformation do to displacement along a planar dislocation [Okada, 1992]

• Adapt method for a plane of discrete dislocation sub-fields (to model distributed rather than uniform slip)

Trade-off between smoothness and fitSmoothing minimizes ‘roughness’ of the slip surface [e.g. Harris & Segall, JGR, 1987]

S=slip

Identify a subjective trade-off between improved model fit and overly smoothed result (non-unique solution).

Testing with new stress minimization

[ Feng, Newman, and Chen, 2009] Chen et al., GRL, 2009

Coseismic Slip Distribution Comparison

Strain minimizationRoughness minimization

Model: = 29°, free-surface at trench; Maximum slip (100 m)

Feng, Newman, & Chen, 2009

Maximum slip stabilizes near 30 m

Solomon 2007 event results:

Interface slip Surface uplift

Two large patches of slip occur in the very shallow trench, though not a tsunami earthquake because tsunami affects were expected.

End-member: where subduction of very young crust allows for normal rupture in shallow region.

Chen et al., GRL, 2009

3 Jan 2010 Mw 7.1 Solomon Islands Earthquake and tsunami

News reports of tsunami devastating villages (strange for an M 7)

Trans-oceanic Tsunami identified by DART Buoys

NSF- RAPID funding for: postseismic geodetic tsunami surveys

(event occurred on Sunday; project funded on Friday; left on Saturday)

(Courtesy NOAA-PMEL)

X

Ndoro Point (7m runup)

Retavo Village (7.5 m runup)

Comparable tsunami inundationto 2007 Mw 8.1 earthquake

No postseismic deformation

Subsidence Everywhere

Subsidence along coast requires very large slip seaward (updip)

Predicted tsunami heights

Energy from SI event

- Not energy deficient (doesn’t look like other tsunami earthquakes)

Possibly not slow (fast rupture due to upward propagation of normal seismogenic interface)

or

May have been slow, but with very large slip

~33 s duration suggests 1-1.5 km/s rupture

a normal TsE, rupturing ¼ of its normal length?

• Most megathrust events rupture between ~20-50 km depth• Tsunami earthquakes rupture shallower than 20 km

Total megathrust rupture?:

are “Earthquakes that create tsunamis much larger than expected given their initial magnitude” –Kanamori [1972].

• Possible to rupture both? Difficult to ID since seismic models of interface slip are imprecise and little land exists near trench for geodetic study

Near-trench rupture

Sumatra 2004: total megathrust rupture.Sumatra 2005: normal megathrust rupture.

Chile 2010 : normal megathrust rupture.

Japan 2011: total megathrust rupture, similar to Sumatra 2004

E/M0

Energy rate

2.6e14 W

• Energy Rate for Tohoku 2011 event deficient• Rupture slower than Sumatra 05, or Chile 2010• Similar to 2004 event (perhaps smaller TsE area),

NEIC Finite-Fault ModelDmax ~30 m

=10.5°

FFM: Model 26Dmax ~50 m

= 12.36°

FFM: Model 26

GEOnet GPS processed by Caltech/JPL

GPS: Coseismic DisplacementsData are from model Aria 0.3 from CalTech/JPL Aria team (processed at 5:40 and 5:55 UTC, based on 5 min solutions)

Maximum displacements: 5.2 m horizontal, -1.1 m vertical

FFM: Model 26

GEOnet GPS processed by Caltech/JPL

Real-time PMEL source model

Source: NOAA/PMEL/Center for Tsunami Research Team

Real-time fault plane solutions are found using pre-determined Green’s functionsto estimate the far-field tsunami excitation

Real-time PMEL source model

Tsunami-inverted slip planes are used to predict seafloor displacement.

77 -vertical ‘measurements’ are subset at 0.5°x0.5° grid

Solutions weighted 3:1 to GPS components ( 3 components x 545 station = 1635 values)

Joint GPS-tsunamimeter inversion

Model: Comb3

Dmax ~50 mUzmax ~8 m

GPS: Coseismic DisplacementsData are from model Aria 0.3 from CalTech/JPL Aria team (processed at 5:40 and 5:55 UTC, based on 5 min solutions)

Maximum displacements: 5.2 m horizontal, -1.1 m vertical

GPS: Coseismic DisplacementsModel is solution developed from joint GPS-’tsunami’ inversion

GPS: Coseismic DisplacementsData and Model

GPS: Coseismic DisplacementsResiduals are shown to scale from before

RMS - H= 0.028 m RMS -V= 0.039 m

N=545 stations (bounded by plotted region)

Joint inversion results:

(Newman, Nature 2011)(Sato et al., Science, 2011)

Summary (Japan Trench):Tohoku 2011 earthquake was a Total Megathrust rupture

Joint GPS-tsunami, and teleseismic FFM models agreeDmax~50 m, in 2 lobes very near trench

adjacent to Sanriku 1896 tsunami earthquake

Massive slip in TsE regionrequired for observed tsunami

Summary:• Improvements in seismic tsunami warning for giant and slow tsunami

EQs:

• TsE from Mentawai, Java (2x), Nicaragua, Peru (not Solomons)

• Giant EQs in Japan, Sumatra and Chile

• In the Solomon Islands:

• Shallow slip is the norm, and tsunamis are frequent

• up to 30 m in 2007 Mw 8.1

• up to 7.6 m in 2010 Mw 7.1 (TsE, but possibly not slow)

• Some giant earthquake exhibit total megathrust rupture, enhancing their tsunamigenesis

• Evaluating tsunami hazards going forward…

Seafloor geodesy for shallow locking studies:

Onland geodesy cannot constrain offshore deformation

http://www.kaiho.mlit.go.jp/info/kouhou/h23/k20110406/k110406-2.pdf

Sato et al., Science, 2011

Geodetic Survey Division Office Marine Navigation and Oceanographic Department

Dense seafloor geodetic networks can:

• Image locking building for shallow rupture– (need long-term stable measurements)

• Provide direct warning of rapid seafloor displacement– Rather than focus on seismicity which is a symptom of slip,

geodetic tools can directly observe deformation driving tsunami.

– (Need cable or buoyed systems to relay data)

• This will cost $$ (or €€, ¢¢,¥¥, ££)– If instrument costs go down to $50k, can densly instrument

the Japan trench for ~$20M (EarthScope PBO is ~$100M)– About 0.01% of the projected cleanup cost for Japan.

DART – Real-time ocean bottom pressure sensors

measuring the height of the water column (long wave-length) above them

NOAA and PTWC

50 m seafloor displacement

http://www.jamstec.go.jp/j/about/press_release/20110428/

From repeat seafloor bathymetry

Near trench locking off of Peru

In 1996 a destructive tsunami earthquake occurred near trench about 100 km north

Large Subduction Earthquakes Moment Magnitude ∝ log10 (Area x Slip)

Slip by magnitude:1+ m in M710+ m in M8

Recurrence controlled by: Convergence rate (1-10 cm/yr)

and locking efficiency 5 -100%

Area increased by angle/thick crust/reduced T.

~20 km

Tsunami Generation:

Tsunami amplitude is controlled by vertical seafloor displacement

Impact on land controlled by:• distance from source• “fetch” (lateral extent) of seafloor displacement• path (focusing/ blocking features)• coastal amplification features (e.g. harbors)

• Timing controlled by distance/depth of water column (a few min to 10+ hours)

Very Large and Great Earthquakes:

Subduction zones:

•> 90 % of all great earthquakes (M8+)

•All giant earthquakes (M9+) •Most explosive volcanism (not discussed today)

Heavy population density in coastal environment

Newman, Peng, & Hoyos (in revision)

Sea floor pressure sensors