Sound Propagation Around Underwater Seamounts

1
Sound Propagation Around Underwater Seamounts

• THESIS DEFENSE
• by Joseph John Sikora III
• advisor Dr. Arthur B. Baggeroer

2
Motivation

• Seamounts def. underwater mountains
• Ubiquitous throughout the worlds oceans
• Believed to block acoustic energy in SOFAR
  channel
• Applications
• Acoustic tomography
• Source localization
• Underwater communications

3
Thesis

• Diffracted and reflected acoustic energy fills in
  the forward-scattered field of a seamount at
  long-range.
• Acoustic multipath in scattered field can be
  identified using parabolic equation approximation
  and ray trace modeling technique.
• Physically constrained adaptive array processing
  can improve acoustic ray identification in
  scattered field by improving angle/amplitude
  estimation.

4
Seamount Forward-Scattered Field

• Ocean waveguide containing a seamount
• Sound bends about minimum sound speed axis (about
  750 m)
• Seamount blocks acoustic energy
• Acoustic rays skip over seamount peak and form
  convergence zones in forward-scattered field

convergence zone
shadow zone
minimum sound speed axis
Signal Strength (dB)
seamount
Range from source (km)
RAM acoustic modeler, frequency250Hz, source
depth750m
5
Outline

• Basin Acoustic Seamount Scattering Experiment
  (BASSEX)
• Experiment setup
• Data/issues
• Adaptive array processing
• Robust, adaptive beamforming
• Structured covariance matrix maximum-likelihood
  estimation technique
• Acoustic scattered field of seamounts
• Convergence zone structure
• Resolvability of acoustic rays
• Conclusions/key contributions

6
Bathymetry and Source Locations
Kermit Roosevelt Seamount Complex

• Bathymetry def. sea floor depth
• ETOPO2 database
• LOAPEX
• ship-deployed source
• T1000
• m-sequence
• carrier 68.2Hz
• BW35Hz
• 41 periods
• 20 min length signal
• SPICEX moored source
• S1, S2
• m-sequence
• carrier 250Hz
• BW83Hz
• 11 periods
• 135 sec length signal

7
Raw Bathymetry Data from Swath Echosounder
50km
Kermit Roosevelt Seamount
900 m
Elvis Seamount
1300 m
8
Temperature and Sound Velocity Profile

• Temperature profile measured with Expendable
  Bathythermometers (XBT)
• Important for acoustic modeling i.e., Snells
  Law
• XBTs dropped every 4 hours along tow ship track

Munk profile
minimum sound speed axis
sound speed f (temperature,salinity,depth)
9
Tow ship track during BASSEX
Kermit Roosevelt Seamount Complex
moored acoustic sources
missing XBT casts between S1 and seamount complex
Hawaii
10
Towed Hydrophone Array - Issues

• Towed hydrophone array
• Measure at many locations in scattered field
• Measure acoustic ray arrival angle
• Signal to noise ratio (SNR) improvement
• Five Octave Research Array (FORA)
• Provided/operated by Pennsylvania State
  University
• 3 meter uniform sensor spacing i.e., cut for 250
  Hz
• Issues
• Random data dropouts
• Random start time error
• Sensor position

11
outline

• BASSEX experiment
• Experiment setup
• Data/issues
• Adaptive array processing
• Robust, adaptive beamforming
• Structured covariance matrix maximum-likelihood
  estimation technique
• Acoustic scattered field of seamounts
• Convergence zone structure
• Resolvability of acoustic rays
• Conclusions/key contributions

12
Data Processing
array processing
spatially separate acoustic rays
beam filtering
matched filtering correct for array orientation,
position, heading, depth
detection/estimation
acoustic rays amplitude, travel time, angle of
arrival
13
Adaptive Array Processing with MPDR

• Minimum Power Distortionless Response (MPDR)
  Beamformer
• Minimize output power
• Preserve signal in target direction
• Sensor weights
• wH(?,k) (vH(?,k)S-1(?)v(?,k))-1S-1(?)v(?,k)
• Function of covariance matrix inverse
• Maximum-likelihood spectral covariance matrix
  estimation
• Without physical constraints
• xj(?) - snapshot
• S(?) Ejxj(?)xjH(?)
• snap-shotslt2N
• Small/zero eignevalues
• High sensitivity to sensor perturbations

w weight vector S spectral covariance
matrix v array manifold vector N sensors H
Hermitian transpose E expectation ?
frequency k wavenumber 2p/? cos?
14
Robust Adaptive Array Processing

• Robust MPDR Beamforming
• Diagonal loading
• S(?) ? S(?) s2LI
• Quadratic constraint wHwTogt1/N
• Reduces sensitivity
• Equivalent to adding sensor/white noise
• BASSEX
• MPDR used to process SPICEX receptions
• snap-shots lt 2N
• Moving noise sources
• Changing sensor position/array orientation
• White noise gain constraint i.e., wHw lt 3/N
• Time-varying implementation

To sensitivity I identity matrix N
sensors S spectral covariance matrix s2L
diagonal loading ? frequency
15
PCML Covariance Matrix Estimation

• Maximum-likelihood spectral covariance matrix
  estimation
• With physical constraints
• Array geometry - uniform line array
• Physically realizable model for the covariance
  matrix
• Hermitian Toeplitz
• Sensor noise
• Uncorrelated propagating plane wave interference
• Determine maximum-likelihood covariance matrix in
  constraint set
• Motivation
• Robust to snap-shot deficiency
• Reduced sensitivity to sensor perturbations
• Improved resolution/SNRout

To sensitivity I identity matrix N
sensors S spectral covariance matrix s2L
diagonal loading ? frequency
16
PCML Covariance Model and Constraints
Covariance matrix model
sensor noise
propagating plane waves
Constraints
sensor variance
plane wave power
I identity matrix N sensors S spectral
covariance matrix ? frequency ? wavelength k
wavenumber 2p/? cos?
wavenumber
17
PCML Estimating pn and s2
Global maximum-likelihood solution

• s2 determined from brute force estimation
• Orthogonal basis fn
• Uniform sampling of wavenumber space
• Spacing ?cos(?) 2/N
• Issue wavenumber domain is undersampled

N sensors S spectral covariance matrix s2
diagonal loading ? angle off-endfire fn
basis vector
18
PCML Wavenumber-Power Sampling

• Question where should we sample
  wavenumber-power?
• cxxm - autocorrelation sequence, length 2N

19
PCML Wavenumber-Power Sampling

• cxxm - autocorrelation function, length 2N
• ?cos(?)1/N i.e., 2N basis vectors required
• Avoids aliasing cxxm
• Equal weighting for all propagating plane waves

20
PCML Performance with Simulated Data
example performance with simulated data
SNRout
20 sensor standard array steered to cos(?) 0.4
interference at cos(?i) 0.5, 10 dB 5 dB sensor
noise
2N
Reed et al.
PCML
21
PCML Performance with Real Data
example BASSEX data record LOAPEX m-sequence
signal
half power beam width
m-sequence signal power
measure of array resolution estimated from
bearing-time response lower HPBW ? higher
resolution
wavenumber-power estimate P(?,k) Bartlett
unbiased EP(?,k)P(?,k) MPDR biased
EP(?,k)P(?,k) (L-N1)/L (Lsnap-shots)
22
Outline

• BASSEX experiment
• Experiment setup
• Data/issues
• Adaptive array processing
• Robust, adaptive beamforming
• Structured covariance matrix maximum-likelihood
  estimation technique
• Acoustic scattered field of seamounts
• Convergence zone structure
• Resolvability of acoustic rays
• Conclusions/key contributions

23
Ambient Noise Environment
3.0
-3.0
Cp1.48
inf
1.48 km/s
vibration
24
Convergence Zones in Forward-Scattered Field
Kermit Roosevelt Seamount
Convergence zones
Elvis Seamount
Shadow zones
direction from moored SPICEX source S2
test station Delaunay triangulation
interpolation
25
Forward-Scattered Field Width/Shape
acoustic pressure time series
(km)
normalized acoustic pressure averaged 0-20 deg
off-endfire
15 dB-complete acoustic shadowing
26
Elvis Seamount Forward-Scattered Field
refracted
reflected
refracted
RAM simulated data
shadow
BASSEX results
50 km
27
RAM Simulated Results
Ray Trace Results
BASSEX Results
MPDR beamformed
Bartlett beamformed
engine noise
Reduced Time (sec) 1485 m/sec
21
20-
19
18
17-
10
0
10
20
0
0
10
20
20
arrive angle (deg)
arrive angle (deg)
arrive angle (deg)
28
Side-Scattered Acoustic Field
peak pressure level (dB)

• LOAPEX source
• glitchy data
• broadside
• PCML beamformed
• (cubic spline interpolation)

test station
direct sound path
Kermit Roosevelt Seamount peak
29
Side-Scattered Field Hour 1, Period 41
reflected
reflected
sidelobes
refracted
reflected

• Reflected acoustic rays appear above 95 degrees
• Accurate angle/amplitude measurement
• Important for verifying future 3-D
  range-dependent acoustic models

sidelobes
refracted
reflected
sidelobes
refracted
30
conclusion

• The formation of convergence zones was observed
  in the forward-scattered field of the Kermit
  Roosevelt Seamounts, confirming theoretical
  models.
• RAM (parabolic equation approximation) and RAY
  (ray trace) numeric models used to reconcile
  acoustic arrival patterns
• The forward-scattered field was shown to span
  projected width of the seamounts down to 3-4km,
  with 15dB/complete shadowing observed.
• Predicting reflected acoustic rays in the
  forward-scattered field was shown to require very
  accurate sound velocity profile measurement.
• Detailed observation of Kermit Roosevelt Seamount
  side-scattered field observed using PCML
  beamformer and ship-deployed LOAPEX source.
  Future work required to investigate accuracy of
  3-D acoustic model.

PCML performance approx. 2X resolution vs.
Bartlett reduced sidelobe vs. Bartlett robust to
snap-shot deficiency/sensor position
error side-scattered acoustic energy observed in
Kermit Roosevelt scattered field
31
Key Contributions

• Seamounts
• Ubiquitous throughout the worlds oceans
• Believed to block acoustic energy
• Previous work
• Explosive sources
• Inadequate bathymetry/sound velocity data
• BASSEX experiment
• Identified diffracted/reflected acoustic energy
  in the forward-scattered field of seamounts
• Applications
• Source localization
• Comprehensive Nuclear Test Ban Treaty
• Naval vessel detection/concealment
• Predict travel time across ocean basins
• Acoustic tomography (ATOC experiment)
• Underwater communications

32
Key Contributions

• Physically Constrained Maximum-Likelihood
  Beamforming (PCML)
• Estimate range/angle/amplitude of acoustic rays
  in seamount scattered field
• Robustness to snap-shot deficiency
• Reduced beamformer sensitivity
• Reduced wavenumber-power estimate bias/variance
• Improved computational efficiency
• Improved array resolution
• Applications
• Array processing in rapidly changing noise
  environments
• Method could be extended to planar/vector sensor
  arrays

33
Future Work

• physical experimentation
• use fixed receiver arrays to measure acoustic ray
  amplitude/phase statistics
• when using towed hydrophone receiver array,
  measure target signals behind/away from tow ship
• numeric simulation
• use side-scattered field results to verify 3-D
  acoustic modelers
• acoustic tomography

34
Acknowledgments

• advisor
• Arthur Baggeroer
• committee members
• Jim Lynch
• David Staelin
• Henrik Schmidt
• chair Mark Grosenbaugh
• graduate students
• Hyun Joe Kim
• John-Paul Kitchens
• Digital Signal Processing Group
• researchers (WHOI OASIS Inc.)
• Edward Scheer
• Keith Von Der Heydt
• Kevin Heaney
• family
• Joseph Sikora II
• Elaine Marks
• Robin Masi