Remote sensing using noisePeter Gerstoft, Scripps Institution of Oceanography

Paradigm shift: we are turning noise into useful data, from which structure information can be extracted.

Noise gives similar information as using a source. Environmentally friendly!

Noise Interferometry (NI) has seen remarkable growth in the last 5 years

Origin of seismic/acoustic noise

Super Typhoon Ioke (in 2006)

Tracking Tropical Cyclones

• Evidencing nonlinear wave-wave interactions in the deep ocean (Longuet-Higgins, 1950)

• Tracking wave-wave interactions rather than a storm itself

Zhang, Gerstoft, and Bromirski (under review)

Track of storm from microseisms (0.2 Hz)

Mechanism involves ocean acoustics!

Ocean waves

Deep ocean bottom

Classic seismic P-wave propagation

1 2

*

Sources yielding constant time-delay τ lay on same hyperbola

τ=0

τ=L/c

-L/c +L/cτ

0

2→1 1→ 2

*

Free space noise correlation

C12(τ)

C12 ( ) P(r1, t)Pr t d t.

dC12 ( )

d G(t) G( t)

Ambient noise EGFs (20-100 Hz)

Dis

tan

ce (

m)

Time (s)

EGF envelopes (dB) with modeled travel times (dotted) between hydrophones

Amplitude (dB) Wd=80 m

230 m long array Brooks and Gerstoft (JASA 2009a 2009b); Fried at al (JASAEL 2008)

Green’s functions estimate

Wd=70 m

230 m long array

(a) Vertical lowered source

(b) Towed source

(c) Ship noise

(d) Ambient noise

• HLA elements parameterized by distance and azimuth: model vector :

• Travel times from peak of empirical Green’s function: observed data vector:

• A priori array is straight

1

2 3 20 21origin

Noise array localization• Methodology adapted from Sabra [2005]

• Objective function minimize difference between observed traveltimes and computed traveltimes from model vector, whilst ensuring “smooth” fit

• Objective function minimized using MATLAB’s nonlinear least-squares function

• Six largest travel-time difference rejected for each computation

• Lower and upper limits set to half and twice a priori distances

• Variation of the smoothness ‘weighting’ seen to have negligible effect

A priori vs a posteriori geometry

A priori geometry A posteriori geometry

Siderius et al., JASA 2006,Gerstoft et al., JASA 2008, Harrison, JASA 2008Harrison, JASA 2009, Traer et al., JASA 2009,Siderius et al., JASA 2010

B1

B2

Using ambient noise on a drifting array we can map the bottom properties

Passive fathometer

Fathometer comparison to seismic

South of Sicily (NURC: 32 phones spaced at 0.5 m)

Dabob Bay, Wa(16 phones spaced at 0.5 m)

Gerstoft et al., JASA 2008,

Background

•Passive fathometerSiderius. JASA, 2010

Active source (Uniboom)

MVDR Passive fathometer

Retrieving temporal velocity variations

A temporal change in velocity along the path between two stations is revealed as an increase in dt with propagation distance, when comparing the cross-correlations from two different time periods.

dt

Measured velocity change associated with damage from earthquakes and volcanic precursors.

Brenguier et al, Science 2008

Velocity change across a fault

Conclusion

• Noise provides useful signal• We can obtain a partial Greens function

Applications:• Locating noise sources• Used for obtaining Earth structure (many applications)• Fathometer• Structural health monitoring• Human body monitoring

Downward beam (MAPEX2000bis)

fdesign c

2d, where d 0.5m

Above design frequency, downgoing noise appears as upgoing

Up

Down

Ang

le

Frequency (kHz)

32 element NURC array

SW06: d=4m => No fathometry!We need dense arrays to get sufficient resolution.

New Siderius arrays (d~0.5m) makes fathometry feasible.

ErnestoSeismic Beamforming: a seismic array in California detected low frequency signals on Sep 2 from a direction consistent with the SW06 site.

=> Stay ashore

Ernesto provided ideal conditions for noise cross-correlation

Future fathometer work

Experimental data shows array is subject to wave driven motion, preventing coherent averaging

• Model for amplitudes• Coherent averaging• Averaging time• Is the array moving up and down?

Humphreys & Clayton

(JGR, 1990) Polet (G3, 2007)?

Storms

Teleseismic body-wave tomography (regional)

P Waves Imaging Earth Structure

Storms (seismic sources in open ocean) can fill azimuth gaps

History of seismic/acoustic interferometry

• 1968 Claerbout• 1980’s experiment at Stanford• 1990’s helioseismology• 2001 Weaver and Lobkis• 2004 first papers in seismology, & ocean acoustics (Roux

and Kuperman)• 2008 book “Seismic interferometry: History and present

status”• 2009 book “Seismic interferometry”• 2009 ~100 papers/year; 3 in Science or Nature /year

Progress due to better computer resources, instrumentation and theory.

Still lots of low hanging fruits!

Ocean noise interferometry publications www.mpl.ucsd.edu/people/pgerstoft

• Traer, Gerstoft and Hodgkiss (2010), Ocean bottom profiling with ambient noise: a model for the passive fathometer, submitted JASA.

• Siderius, Song, Gerstoft, Hodgkiss, Hursky, Harrison (2010), Adaptive passive fathometer processing, JASA.

• Brooks, Gerstoft (2009), Green’s function approximation from cross-correlation of active sources in the ocean, JASA.

• Brooks, Gerstoft (2009), Green's function approximation from cross-correlations of 20–100 Hz noise during a tropical storm, JASA.

• Traer, Gerstoft, Song. Hodgkiss (2009), On the sign of the adaptive passive fathometer impulse response, JASA.

• Gerstoft, Hodgkiss, Siderius, Huang, Harrison (2008), Passive fathometer processing, JASA.• Brooks, Gerstoft, Knobles (2008), Multichannel array diagnosis using noise cross-correlation,

JASA EL.• Traer, Gerstoft, Bromirski, Hodgkiss, Brooks (2008), Shallow-water seismo-acoustic noise

generation by Tropical Storms Ernesto and Florence, JASA EL.• Brooks and Gerstoft (2007), Ocean acoustic interferometry, JASA,

• Tomorrow:

Bill Hodgkiss nearfield geoacoustic inversion

Caglar Yardim PF and objective functions