Content uploaded by David R Barclay

Author content

All content in this area was uploaded by David R Barclay on Feb 04, 2019

Content may be subject to copyright.

Introduction

A series of experiments aimed at measuring the power spectrum

and vertical and horizontal coherence and directionality of

ambient noise in the deep ocean were carried out between 2009

– 2014 using the family of “Deep Sound” autonomous acoustic

recording platforms. In December 2014, an expedition was

mounted aboard the Schmidt Ocean Institute’s R/V Falkor to

the Challenger Deep in the Mariana Trench, with the objective

of proling the ambient noise in the deepest known spot in the

world’s oceans. e depth at the site is 10,916 m. When compared

with previous measurements made in the Philippine Sea to a

depth of 5,940 m (Barclay and Buckingham, 2013a), in the Tonga

Trench to a depth of 8,500 m (Barclay and Buckingham, 2014),

and in the Sirena Deep, Mariana Trench to a depth of 8,900 m,

the results show that below the critical conjugate depth (depth

at which the sound speed is higher than any sound speed above

it), the second-order noise eld statistics depend on the relative

contributions of locally generated surface noise and distantly

generated propagating noise. e depth dependence of the

sound eld varies with local surface conditions where near

constant noise power and coherence (directionality) is found

at moderate wind states and above, while a drop in noise level

below the critical depth is found during calm sea states. It is also

shown that the vertical noise coherence is well described by

the Cron & Sherman surface noise model (Cron and Sherman,

1962) with some depth-dependence explained by the local sound

speed variations.

Deep Sound

Deep Sound is a free-falling (untethered) acoustic recorder

designed to descend from the ocean’s surface to a pre-assigned

depth where it drops an iron weight and returns to the surface

under its own buoyancy with a speed of ~0.5 m/s in either

direction. ree variants of the instrument family have been

built, the Mk I, II and III. e data recorded at the Challenger

Deep was done by the Mk. II and as such this variant will be

briey described here. More complete descriptions of all three

are found in previous publications (Barclay et al., 2009; Barclay

and Buckingham, 2013b; 2014). e instrument comprises of a

3.6 cm thick Vitrovex glass sphere with a 43.2 cm outer diameter

and a depth rating of 9 km, four external hydrophones, a conduc-

tivity-depth-temperature (CTD) sensor, and recovery beacons: a

strobe, radio beacon, and satellite beacon.

Inside the sphere an embedded microprocessor monitors

and records the acoustic and oceanographic data, six degree of

freedom inertial motion and magnetometer (compass) data, and

system data such as internal temperature and remaining battery

charge. e ensemble of data is used to trigger a drop weight when

a set of pre-determined conditions are met. In practice, this is

typically when the instrument reaches a pre-programmed depth,

or when a certain deployment time limit has elapsed, but may also

be due to a low battery warning, overheating, or communications

failure with one of the external sensors.

Pressure time series are recorded on four High Tech Inc.

HTI-99-DY hydrophones arranged in an ‘L’ shaped array with

three vertical and two horizontal spacing. e hydrophones

are mounted on arms to places them outside the motion-in-

duced turbulence of the main instrument package. e acoustic

bandwidth of this hydrophone is 5 Hz – 30 kHz, and all four

channels are simultaneously sampled by the data acquisition

board at an over-sampled rate of 204.8 kHz and a dynamic range

of 24 bits. e CTD provides real time depth and vertical speed

data, and is used to calculate temperature, salinity, density, and

sound speed depth proles.

Battery recharge and data transfer are accomplished through

an external housing throughput connections and Wi-Fi connec-

tivity, allowing the system to be recovered and rapidly redeployed.

Sound speed profiles and the conjugate depth

e sound speed in seawater is driven by an empirically derived

equation of state linking temperature, pressure, salinity, and

density. In the deep ocean, the sound speed prole is typically

described by warm (fast) surface waters, with decreasing tempera-

ture driving sound speed down with depth until pressure becomes

the dominant term, now raising sound speed with depth. is

results in a sound speed minimum, typically around 700 m depth

in temperate Pacic waters, known as the sound channel axis.

Sound rays refract towards this axis and remain in this ocean

wave-guide, allowing sound to propagate over thousands of kilo-

meters with no attenuation due to interactions with the surface

and bottom (Munk et al., 1994).

Sound speed proles are shown in Fig. 1, collected by

Deep Sound at the ve deployments sites visited: Challenger

Deep (11°21.599’ N, 142°27.249’ E), Challenger Deep East

Basin (11°22.130’ N, 142°35.172’ E), Sirena Deep (12°43.680’ N,

145°47.820’ E), Tonga Trench (16°31.8760’ S, 172°12.0930’ W),

and Philippine Sea (22°13.572’ N, 126°13.807’ E). Despite their

disparate locations, the proles are remarkably similar with

the sound channel axis between 700 - 850 m. Additionally, the

conjugate depth, the depth at which the sound speed is equal to

the sound speed at the surface, is at ~ 5 km for all deployments,

with the exception of the Philippine Sea, which has a critical

depth just below 4 km due to its higher latitude and cooler surface

waters. e conjugate depth is signicant, since no surface-gen-

erated sound propagating in the deep sound channel may

penetrate below it and, according to Weston (Weston, 1980),

the ambient noise from surface sources is thought to decay to a

Figure 1. The ensemble of sound speed profiles collected by the Deep Sound

instruments 2009 – 2014 (deployment location indicated by colour), with the

solid black line depicting the sound channel axes and the dashed coloured

lines showing the conjugate depths.

The depth dependence of

ambient noise coherence in

the Challenger Deep

By David Barclay, Michael Buckingham and Dieter Bevans

Technical Contributions

Acoustics Bulletin July/August 2017

36

negligible level. Prior to the development of Deep Sound, meas-

urements of ambient noise below the critical depth to conrm

Weston’s prediction were sparse due to the relative diculty of

deploying moored arrays in such deep waters (Morris, 1978; Gaul

et al., 2007).

Power spectrum, vertical coherence,

and directionality

As Deep Sound descends and ascends through the water column,

each hydrophone records a pressure time series. From the Fourier

transform, X

i

(w), of the time series at the i

th

hydrophone, the

power spectrum

(1)

is formed, where T is the record length of each transform, the

brackets denote an ensemble average, w is angular frequency and

the asterisk indicates a complex conjugate. e spatial coherence

function between hydrophones i and j, dened as the cross-spec-

tral density normalized to the geometric mean of the power

spectra at the two sensors, is also constructed as a measure of the

directionality of the noise eld:

(2)

(i) Cox’s equation for vertical noise directionality

In a plane wave noise eld, such as ambient ocean noise, the

vertical noise coherence can be related to the vertical noise

directionality through an integral expression derived by Cox

(Cox, 1973)

(3)

where i=√(-1), q is the polar angle measured from the zenith,

and t

d

=d/c is the acoustic travel time between sensors separated

by a distance d with a local sound speed of c, and F(q) is the

noise directional density function, representing the noise power

incident on the sensor per unit polar angle. e integral relation-

ship is linear, thus the noise directional density function may

represent a superposition of source spatial distributions.

(ii) Cron and Sherman model of wind driven surface noise

A theoretical model of wind driven surface noise in an innitely

deep ocean with an iso-velocity sound speed prole was derived

by Cron and Sherman (Cron and Sherman, 1962), where a

random uniform distribution of monopole sound sources is

placed on a plane immediately beneath the at, pressure-release

sea surface. In eect, each monopole and its negative image in the

sea surface acts as a vertical dipole with power directivity function

cos

2

q

. From this geometry, the depth-independent vertical noise

coherence on a pair of sensors has the closed form solution

(4)

representing a directional density function for downward

traveling noise, consistent with the lack of an ocean bottom reec-

tion, of the form

(5)

where u(.) is the Heaviside unit step function.

The depth-dependence of noise across the

conjugate depth

Philippine Sea, Sirena Deep, and Tonga Trench

Measurements made during the PhilSea experiment in May 2009

P38

Technical Contributions Technical Contributions

Acoustics Bulletin July/August 2017

36 Acoustics Bulletin July/August 2017 37

in the Philippine Basin found that the Cron and Sherman model

described in Eq. (4) and the experimentally measured vertical

coherence function agree remarkably, both above and below the

critical depth, with some ltering need at low-frequencies due to

turbulent ow-noise generated on the surface of the hydrophones

by their motion descending and ascending through the water

column, and at high frequencies due to system noise (see (Barclay

and Buckingham, 2013a) for more detail). e agreement between

the data and model at all depths, with both amplitudes and

zero-crossings well matched in the real and imaginary compo-

nents of the coherence, is shown in Fig. 2, taken from (Barclay

and Buckingham, 2013a). In computing the theoretical curves

in Fig. 2, the locally measured CTD sound speed was used in Eq.

(4). It follows from this result that the vertical directionality of the

wind noise in the band 1 – 10 kHz must be independent of depth

and well approximated by the cosine law in Eq. (5). e sound

speed prole has a negligible eect on the vertical directionality

of surface-generated wind driven noise since it is predomi-

nantly all downward traveling at steep angles, even at the sound

channel axis. Similarly, any bottom reections make a negligible

contribution to the noise eld, even at depths near the seabed.

As a result, the noise below the critical depth shows no change

in directionality.

Additionally, measurements of the power spectral density

across the wind driven noise band show no depth dependence

across the critical depth, as illustrated in Fig. 3. e wind speed

at 10 metres above the sea surface, U

10

, had a steady mean of 10

m/s during the deployment. e remarkable consistency of the

second-order noise eld statistics across all depths is consistent

with the noise originating from local, wind driven surface sources,

with a negligible contribution from distantly located noise sources

propagating in the sound channel.

Figure 4. (a) Real and (b) imaginary parts of the vertical noise coherence in

the Challenger Deep, as a function of depth over the wind noise band and

plotted in dimensionless frequency, wtd=wd/c, where d is the sensor spacing

and c is the local sound speed. The black line indicates the conjugate depth

and the black dashed line is sound channel axis.

Figure 3. The depth-dependence of ambient noise power in the wind driven

noise band in the Philippine Sea, with centre frequencies at 1, 3, 5, and 8 kHz

with a 5% bandwidth. The solid line indicates the sound channel axis depth

while the dashed line shows the conjugate depth.

Figure 2. (a) Real and (b) imaginary parts of the filtered coherence function

versus frequency at 500 m depth intervals on the ascent in the Philippine Sea.

The jagged lines represent data and the smooth curves are the filtered version

of Cron and Sherman’s expression in Eq. (4). To separate the curves, 1 has

been added (subtracted) for each depth increment above (below) 3000 m.

P37

Technical Contributions

Acoustics Bulletin July/August 2017

38

ese results are true for the typical deep-water sound

speed prole recorded in the Philippine Sea, as well as the

other deep-water locations where Deep Sound was deployed,

such as the Sirena Deep, and the Tonga Trench (Barclay and

Buckingham, 2014).

Challenger Deep

In December 2014, Deep Sound was deployed in the Challenger

Deep, Mariana Trench, as part of a multi-disciplinary research

cruise aboard the R/V Falkor that included the deployment of

demersal and benthic animal and microbe collecting landers, and

a video camera lander.

Deep Sound returned the depth-dependent vertical noise

coherence from the surface to its maximum depth rating of 9 km,

as illustrated in Fig. 4. In contrast to previously collected data,

a clear dierence between the coherence above and below the

conjugate depth is apparent. Below the conjugate depth, at 7.5

km, the vertical coherence ts a modied Cron and Sherman

model, illustrated in Fig. 5, conrming that the noise is due to

locally generated surface sources, and predominantly downward

propagating with no apparent reection from the seabed. Note

that below 1 kHz, the wind noise begins to roll o in strength

due to the nature of the sources (Wenz, 1962), while inco-

herent turbulent ow noise dominates the signal, causing the

real component of the coherence to drop to zero instead of the

predicted unity.

e model has been modied to t the data by the addition

of a frequency independent scaling factor to account for the

broadband mixture of turbulent ow noise with characteristic

frequency slope f-5/3 (Lighthill, 1954) and the ambient noise on

the trailing hydrophone. Neither of these model modications

has relocated the zero crossings of the coherence curves, and

a more complete discussion is given in previous publications

(Barclay and Buckingham, 2013a). Lastly, the model has been

given a small DC oset corresponding to zero time delay noise

across the entire band, which may be due to either electronic

system noise, or to noise arriving broadside to the array. e

contribution is small overall, and based on previous observations

with this instrument, most likely due to system noise.

However, above the conjugate depth, the vertical noise

coherence changes in character, as illustrated in Fig. 4. At the

sound channel axis, the vertical coherence appears to be nearly

Figure 5. (a) Real and (b) imaginary parts of the vertical noise coherence

below the conjugate depth, at 7.5 km in the Challenger Deep. The blue

line is the data and the black line is the modified Cron and Sherman model

calculated with the local sound speed.

P40

With an impressive portfolio of projects for some of the UK’s most prestigious buildings as well as 25 years’

experience in acoustics, we are the obvious choice for everything acoustic.

Photographs showcase some of our recent projects where all acoustic products were

supplied and installed by Acoustic GRG Products Ltd.

For more information visit acousticgrg.co.uk, email matt@acousticgrg.co.uk or call 01303 230 994

//

everything acoustic

Technical Contributions Technical Contributions

Acoustics Bulletin July/August 2017

38 Acoustics Bulletin July/August 2017 39

constant across the entire wind noise band, with the character-

istic roll o below 1 kHz due to ow noise, shown in Fig. 6a. On

closer inspection, illustrated in Fig. 6b, the coherence is the linear

superposition of a frequency independent constant and a scaled

Cron and Sherman coherence function. is combination can be

related to directionality by considering Eq. (3),

(6)

where is the scaled Cron and Sherman contribution,

is a scaled version of Eq. (5), Γ

0

is the frequency independent

constant coherence contribution and F

0

(

q

) is its corresponding

vertical noise directionality. Eq. (6) is linear, so we can solve for

the unknown noise directionality, as it is clear that that Γ

0

and

F

0

(

q

) are Fourier transform pairs and thus unique. erefore F

0

(

q

)

must be due to noise arriving broadside, or exactly horizontally on

the vertically separated sensors, and the total noise directionality

at the sound channel axis must be given by

(7)

where A and B are scaling constant that ensure the normaliza-

tion condition,

(8)

is satised, plus the requirement that the coherence must be unity

at zero frequency (or zero separation).

Since this horizontal contribution to the noise eld is inde-

pendent of frequency over the wind driven band of 1 – 10 kHz,

we can conclude that wind generated surface noise, not distant

shipping noise or any other narrower band source, is propagating

in the sound channel axis. Although wave induced surface sound

sources have dipole radiation patterns oriented in the downward

direction, and are therefore very inecient at coupling into the

sound channel, bathymetric features such as the continental shelf

slope, sea mounts, and islands are known to contribute surface

noise to the channel via scattering (Carey and Evans, 2011). e

Challenger Deep is 300 km from the island of Guam, providing the

appropriate bathymetry for such a scattering mechanism.

e U

10

wind conditions during the Challenger Deep deploy-

ment were light, with a mean of 4 m/s at the beginning of the

deployment, slowly rising to 8 m/s by the end of the recording

period. e contribution of locally generated surface noise was

therefore small, relative to the distantly generated noise in the

sound channel, giving rise to a noise directionality that more

heavily favors the second term in Eq. (7).

From this conclusion, can divide Fig. 4 into two regions.

Above the conjugate depth, where the noise eld is comprised

of a relatively large component due to distantly generated noise

propagating in the sound channel and a small locally generated

surface noise component, and below the conjugate depth, where

the ambient noise is solely due to locally generated surface

wave action.

Conclusions

e importance of local surface conditions in determining the

noise coherence in a deep-water environment is demonstrated

by the collection of these data sets. When winds are 10 m/s or

greater, the noise eld is dominated by downward propagating

locally generated noise, as was found in the Philippine Sea, Sirena

Deep, and Tonga Trench data. For winds lighter than 8 m/s, the

noise eld above the conjugate depth comprises mostly of a hori-

zontally propagating component of distantly generated surface

wave noise with a small downward propagating locally generated

component found both above and below the conjugate depth, as

found in the Challenger Deep. In both cases, reections from the

seaoor do not contribute signicantly to the noise eld.

e Cron and Sherman model describes the locally generated

component of the ambient noise eld which contributes with

constant power at all depths, while the distantly generated noise

can be described as a purely horizontal contribution to the noise

directional density function, or as a broadband constant value to

the vertical noise coherence.

Acknowledgements

e Schmidt Ocean Institute, UC Ship Grants, and the Oce of

Naval Research North Pacic Acoustic Laboratory (NPAL) group

provided ship time. is research was supported by the Oce of

Naval Research, Ocean Acoustics Code 322OA, under Grant Nos.

N00014-10-1-0092 and N00014-10-1-0286, sponsor Dr. Robert

Headrick, and N00014-15-1-2629, sponsor Dr. Kyle Becker.

Dr David Barclay is the Canada Research Chair in Ocean

Technology Systems in the Department of Oceanography,

Dalhousie University. He was an Oce of Naval Research post-

doctoral fellow in Ocean Acoustics at Woods Hole Oceanographic

Institution, and received a PhD. in 2011 from the Scripps Institution

of Oceanography, UC San Diego. Michael J Buckingham is a

professor of ocean acoustics in the Marine Physical Laboratory at

Scripps Institution of Oceanography, University of California, San

Diego. His research focuses on ocean acoustic propagation, ambient

noise in the marine environment, and acoustic imaging. Dieter

Bevans is a PhD student at the Scripps Institution.

References

A full list of references can be obtained from the Editor,

CharlesEllis at charles.ellis@ioa.org.uk .

Figure 6. Real part of the vertical noise coherence in the sound channel axis,

at 1 km depth in the Challenger Deep over (a) the full range of coherence,

showing the frequency independent constant coherence due to horizontally

propagating noise, and (b) over a limited, zoomed range to show the details

of the locally generated noise coherence. The data are plotted in blue and

the modelled coherence corresponding to the directionality given by Eq. (7) is

plotted in black.

P39

Technical Contributions

Acoustics Bulletin July/August 2017

40