Content uploaded by Pal Schmitt
Author content
All content in this area was uploaded by Pal Schmitt on Apr 18, 2016
Content may be subject to copyright.
Field Testing a Full-Scale Tidal Turbine
Part 3: Acoustic Characteristics
P´
al Schmitt∗, Bjoern Elsaesser∗, Matthew Coffin†, Joe Hood†and Ralf Starzmann‡
∗Marine Research Group
Queens University Belfast,Marine Laboratory
12-13 The Strand, BT22 1PF Portaferry
Northern Ireland
p.schmitt@qub.ac.uk
b.elsaesser@qub.ac.uk
†Akoostix/GeoSpectrum
10 Akerley Blvd, Unit 19
Dartmouth, NS B3B 1J4
Canada
mcoffin@akoostix.com
joe.hood@geospectrum.ca
‡SCHOTTEL HYDRO GmbH
Mainzer Strae 99, D - 56322 Spay
Germany
rstarzmann@schottel.de
Abstract—Like any new technology, tidal power converters are
being assessed for potential environmental impacts. Similar to
wind power, where noise emissions have led to some regulations
and limitations on consented installation sites, noise emissions of
these new tidal devices attract considerable attention, especially
due to the possible interaction with the marine fauna. However,
the effect of turbine noise cannot be assessed as a stand-alone
issue, but must be investigated in the context of the natural
background noise in high flow environments. Noise measurements
are also believed to be a useful tool for monitoring the operating
conditions and health of equipment. While underwater noise
measurements are not trivial to perform, this non-intrusive mon-
itoring method could prove to be very cost effective. This paper
presents sound measurements performed on the SCHOTTEL
Instream Turbine as part of the MaRINET testing campaign
at the QUB tidal test site in Portaferry during the summer of
2014. This paper demonstrates a comparison of the turbine noise
emissions with the normal background noise at the test site and
presents possible applications as a monitoring system.
Index Terms—noise, tidal turbines, full scale, environmental
impact
I. INTRODUCTION
The tidal industry, while still in its infancy, promises to
contribute to clean, carbon neutral electricity production in the
coming decades. As with any other new technology, potential
impacts on the environment during installation and operation
must be assessed carefully. One issue under consideration
is the noise emission of such machines [1, 2], although
quantifying the impact of underwater noise is a complex task.
Kastelein et al. (2008) [3] investigated the impact of noise
on different fish species while Ellison et al. 2012 [4] re-
searched the effect of acute and chronic noise on marine
mammals and recommended a new approach for ecosystem-
level management and spatial planning of off-shore work.
Generalisations about the effects of sound on marine fauna
should be made with care since the reactions depend on
various factors like location, temperature, physiological state,
age, body size, and school size. Other researchers already
indicate that the noise emitted by tidal turbines might interfere
with local fauna, although, due to the lack of existing turbine
noise data, previous studies are mainly based on noise sources
created from relatively vague assumptions [5].
A review by the Crown Estate [6] on the current knowledge
on sound emissions of wave and tidal power devices also
highlighted significant knowledge gaps, especially the lack of
fully publicly disclosed data from noise measurements in full
scale operating conditions.
Other areas of research on underwater noise pollution and
the effects on local fauna have been ongoing for decades.
Investigations range from ship traffic [7] to engineering works
like pile driving [8], to name just a few.
Typical installation sites for tidal turbines seem to be
subject to high levels of ambient noise. Strong currents can
mobilize sediments resulting in significant ambient noise at
frequencies (typically >1 kHz) related to the size distribution
of sediment present at the site [9, 10]. Turbulence advected
over hydrophones can result in flow noise, which, although
not a true ambient noise source, can dominate low frequency
noise measurements [11, 12]. The acoustic effects of new
anthropogenic noise sources must be considered in the context
of preexisting background noise levels.
This document presents both fixed and drifting acoustic
Fig. 1. Bathymetry of the test area. The red star shows the turbine position.
The coloured tracks indicate the sections used for averaging. Gridlines are
300m apart. Depth contours represent 2m.
Fig. 2. Hydrophone and data acquisition system.
measurements of a full-scale operating tidal turbine in Strang-
ford Lough, Northern Ireland. The turbine was installed on a
barge, moored in about 10 m water depth and located in a tidal
channel. A bathymetry of the test area can be seen in Figure
1.
The turbine hub was fixed at a depth of 3.4 m below the wa-
ter surface. The seabed is characterized by small boulders on
bedrock base layer. The sea state was 1-0, and no precipitation
was falling. More details on the turbine setup can be found in
Jeffcoate et al. [13]. Measurements were performed in flood
tides when the mean flow direction in the channel is north-
west. During the measurements the tidal current velocities at
the turbine were 1-2 m/s, in the middle of the channel they
increase to up to 4 m/s.
The measurements obtained at the site are used to describe
the influence of changes in operating mode on the sound
signature. In a second section sources levels of the operating
turbine are estimated.
II. EQ UIPMENT
Underwater noise measurement equipment was provided
and operated by Akoostix Inc. Figure 2 shows the data acquisi-
tions system (DAQ), an Akoostix GuardBuoy. The GuardBuoy
uses a GeoSpectrum Technologies Inc. (GTI) omni-directional
M24 hydrophone, which can be seen on the left hand side of
the buoy in Figure 2 protruding from the watertight casing.
A metal cage protects the sensitive hydrophone head from
damage. The Guardbuoy samples at 96 kHz, providing mea-
surements from approximately 20 Hz to 40 kHz, and uses a
24-bit analogue-to-digital converter (ADC).
While GTI provided a sensitivity (calibration) curve for
the M24 hydrophone, technicians calibrated the assembled
Guardbuoy to account for other system gains such as analogue-
to-digital gain. This produced a calibration to convert the data
from from bits, as the data are recorded in .wav format, to
µPa. During calibration, the GuardBuoy was immersed in
a tank with an acoustic source and a calibrated reference
hydrophone. The acoustic source was positioned 0.74 m from
the GuardBuoy and the reference hydrophone. The source
transmitted a series of sinusoidal tones from 40 kHz down to
3 kHz in steps of 0.5kHz. Each pulse transmission was 0.98 ms
in duration, and the pulse repetition interval was 5 s. During
the test, the reference hydrophone voltage was recorded which,
in conjunction with a corresponding sensitivity curve, provided
a measure of the sound pressure level in the tank for each
tone. Once the test was complete, analysts downloaded the
GuardBuoy data, and measured the root mean square (RMS)
level of each tone, taking precaution to exclude signals re-
flected from the sides of the tank. The onset of reflected
signals corresponds to visible fluctuations in the amplitude
of the raw signal. Also, the measured signals were fit to a
sinusoidal curve to avoid underestimation of the RMS levels
due to under-sampling (with a sample rate of 96 kHz and test
frequencies up to 40 kHz, it’s unlikely that peak signal levels
were sampled, so curve fitting is required). The measured RMS
levels, which are in units of least significant bits (LSB), were
converted to dB re LSB/µPa as follows:
dB re LSB
µP a = 20 log RMSmeasured
Vref + Vrefsens (1)
where Vref is the voltage measured on the reference hy-
drophone, and Vrefsens is the reference hydrophone sensitivity
in dB re V/µPa. This calculation was repeated for each fre-
quency, resulting in a frequency dependent calibration curve.
For signals between 10 and 200 Hz, the test was repeated using
a carefully designed in-air calibration method, as the tank was
too small relative to the wavelength of the acoustic waves.
III. MET HO DS
Measurements were performed at the QUB tidal test site in
Portaferry, Northern Ireland, during the Marine Renewables
Infrastructure Network (MaRINET) testing campaign of the
SCHOTTEL STG turbine. Details of the turbine setup can be
found in Jeffcoate et al. [13].
Datasets used in this paper were recorded
at the times as shown in the following table:
Name Date Time
fixed 9m: Aug 25th 10:20:00 - 10:21:00
background 20min 26th Aug 08:29:00 - 08:49:00
background 60s 26th Aug 09:42:00 - 09:43:00
100m Aug 28th 10:00:41 - 10:01:41
250m Aug 28th 09:59:33 - 10:00:33
450m Aug 28th 09:58:05 - 09:59:05
Background noise was measured with the turbine turned off.
Two sets of recordings are discussed in this paper. In the first
set of recordings, the DAQ was fixed to the barge at a depth of
1.7 m. The range to the turbine remained constant at 9m during
these measurements. As a result, these recordings contain low-
frequency flow noise from the motion of the water relative
to the hydrophone. This presents a considerable experimental
challenge, since low-frequency signals from the turbine will
be masked by flow noise, making characterization of these
signals impossible [12, 11].
A second set of measurements was collected with the DAQ
drifting past the turbine. For these tests the DAQ was attached
to a floating buoy so that the hydrophone was placed 4 m
below the water surface. The system was released upstream of
the turbine and recovered downstream. A Global Positioning
System (GPS) was placed in a watertight box, tethered to the
buoy, providing GPS position of the buoy throughout the drift.
Spikes in data were confirmed to be from tapping on or near
the buoy. The resulting acoustic data are less affected by flow
noise. However, this configuration may not be feasible for
longer-term monitoring since it requires personnel to deploy
and recover the buoy on a regular basis, and the risk of
damage to the buoy is markedly increased due to potential
collisions with other objects. Also, measurements are limited
to shorter durations and if source levels (SL)are desired, the
resulting data have to be corrected for changing position and
propagation conditions.
A hydrophone produces a voltage fluctuation that is propor-
tional to fluctuations in the ambient pressure. Sound is often
measured as the logarithm of the ratio of the root mean square
(RMS) pressure pto a reference value p0[14] as follows
SP L = 20 log10 p
p0(2)
The reference pressure for hydrophone measurements is 1 µPa,
while 20 µPa is used in air [15]. dB levels in water are thus
not comparable to dB levels in air due to different reference
pressures.
To assess the overall noise emissions of the turbine, 1/3oc-
tave spectra from 10 Hz to 20 kHz were computed as follows:
•the 96 kHz raw time-series data were processed using
16384 point FFT processing, resulting in 5.86 Hz spectral
resolution. Hann windowing with a 75 %overlap was
used, any frequencies less than 20 Hz and greater than
30 kHz were discarded later.
•Sensor and processing calibrations were applied to con-
vert to power spectral density (PSD)
•the PSD were averaged over 60s seconds, except for the
20-minute noise sample.
•1/3octave spectra were obtained by averaging over the
frequency bins that define the 1/3octave scale [14]
Using the range from the buoy to the turbine, computed
from GPS positions, noise data can be adjusted for trans-
mission loss (T L) due to spreading. To obtain SL from the
recorded levels RL the following equation was used:
SL =RL +T L (3)
This provides an estimated broadband source level at 1 m from
the source (turbine), which is a standard measure for noise
sources. T L can be estimated as a combination of spherical
spreading to the water depth D, and cylindrical spreading
beyond that distance as expressed in equation 4.
T L = 20 log10(D) + 10 log10(range)−10 log10 (D)(4)
It should be noted that equation 4 does not take into account
attenuation and losses due to interactions with the surface or
seabed and is thus a conservative estimate, yielding higher SL
than expected in reality.
Since the buoy is continuously drifting towards the turbine
during the measurement, the signal levels continuously in-
crease within the 60-second windows over which the averaged
spectra are computed. This increase will be more significant
for measurements closest the turbine where spherical spread-
ing likely contributes more to propagation loss, and the change
in source-to-receiver distance is comparable to the mean
distance over the averaging period. Therefore, the averages
will be biased higher by spectra in the latter portions of
the 60-second window. A more robust average would include
range-correction of the individual spectra prior to averaging,
or use shorter time periods where changes in range are not
as significant. These improvements are being considered for
future work.
IV. RES ULT S
A. Operating Modes
This section presents data taken during varying turbine
operation modes.
Figure 3 shows a spectrogram computed from a fixed
measurement, taken at the 25th Aug, around 10:30am. The
x-axis shows the frequency and the y-axis the time, while
the greyscale represents sound pressure levels, with brighter
shades indicating higher levels.
Figure 3 shows the three distinct operating modes of the
turbine. The top half shows the noise emissions of the turbine
in normal operating conditions, generating electricity. Several
discreet, narrowband sound sources can be observed as vertical
lines, varying only little in frequency and amplitude.
The centre part of Figure 3 shows the turbine in free-
wheeling mode, where the turbine blades are allowed to rotate
faster under the force of the current, after being disengaged
from the load of the internal generator. The narrowband sound
sources observed when the turbine is generating are absent
in free-wheeling mode. Instead, there are some narrowband
signals that have a more natural variable frequency instead of
controlled steps.
In the third operating mode the brake is applied. New sound
sources at higher frequencies appear for a short duration,
and several narrowband signals become prominent These
narrowband signals decrease in frequency, maintaining their
mechanical relationship (gear ratios) until the turbine stops
rotating completely (not shown in Figure 3).
Running
Free wheeling
Braking
Fig. 3. Spectrogram of the noise measured at a fixed position beside the
turbine. Grey values are given in dB re µPa2/Hz.
Running
Free wheeling
Braking
Fig. 4. Normalised spectrogram of the noise measured at a fixed position
beside the turbine.
The observed noise from 20-300 Hz is attributed to flow
noise.
Normalization of a spectrogram is performed to enhance
signals of interest, in this case narrowband signals. Normal-
ization is achieved by estimating a background noise level and
dividing the data by that level, though the method of noise
estimation varies with the type of signals to be enhanced. As
shown in Figure 4 this enables the identification of weaker
narrowband signals, particularly in the low frequency range.
Though this normalization mode optimizes narrowband sig-
nals, it also suppresses broadband noise and transient signals
that may otherwise show up as spikes or horizontal stripes
on the spectrogram. In this case, there is a broadband signal
between 4 and 6 kHz, which is visible in the unnormalized
image, but absent from the normalized image. In Figure 3
weak signals are not visible at higher frequencies, but they
can be detected in the normalized image, Figure 4.
B. Noise Levels
With the exception of one of the noise samples, spectra were
averaged over 60s. For the drifting measurements this equates
to approximately 100 m displacement with the current. One
noise sample is 60 s in duration and represents the minimum
noise levels recorded over the entire experiment. The second
noise sample is averaged over 20min and is a representative
measure of the typical background noise in the strait. In this
twenty-minute period, noise from a ferry and possible sounds
Fig. 5. Ranges used for distance correction. The red bands show data which
was averaged and then corrected for distance. Grey values are given in dB re
µPa2/Hz
of rocks moving on the sea floor due to the current could be
identified.
Figure 5 shows a spectrogram with 60 s samples highlighted
in red.
Figure 6 displays spectral density level over frequency for
both ambient noise measurements and three reference ambient
noise curves showing typical noise levels for sea state 1 and
6 along with the typical quietest ocean noise level as reported
by [16]. The noise levels related to sea states seem a useful
reference since they are widely accepted and many future
installation sites will be exposed to waves.
The noise measurement over 20 min is very similar to
the noise reported for sea state 6 for the frequency range
from 200-2000 Hz with noise levels decreasing with increasing
frequency from 72 dB re µPa2/Hz to 62 dB re µPa2/Hz. Above
2000 Hz the data for sea state 6 continues to drop while the
measured noise data remains almost constant. The 60 s noise
measurement includes a noise spike in the 200-300 Hz range.
The structure of the signals around 200 Hz do not exhibit the
discreet, ’stair-step’ appearance visible in the other turbine
signals, indicating that these signals do not respond to changes
in turbine speed, suggesting that they are not related to the
turbine.
The noise levels for sea state 1 are about 20 dB less than for
sea state 6 over the entire frequency range. Between 500 Hz
and 20000 Hz they are lower than for sea state 1. In the
following plots only the noise data averaged over 20 min and
the theoretical minimum is plotted for comparison with noise
samples taken with the turbine running.
Data below 100 Hz is probably affected by shallow water
effects, [15] estimates the cut-off frequency for a water depth
of about 10 m as 70 Hz.
Figure 7 shows ambient noise with data recorded while
the turbine was running. Turbine noise was measured both
with the hydrophone fixed 9 m from the turbine and with the
hydrophone drifting by the barge at varying distances.
As might be expected, the highest levels of around 100dB
re µPa2/Hz are measured when the hydrophone is only 9 m
10
20
30
40
50
60
70
80
90
100
100 1000 10000 100000
[dB re µP a2/Hz]
Frequency [Hz]
Background 20min
Background 60s
WenzM in
Seastate 1
Seastate 6
Fig. 6. Power spectral density over frequency for ambient noise as measured
below the barge and typical spectra for different seastates according to Wenz
(1962) [16].
from the turbine and fixed to the barge. The data from the test
runs performed at 100 m, 250 m and 450 m distance from the
turbine are all similar in shape but lower in level, as observed
sound levels typically reduce with distance from the source.
Floating measurements in 100 m and 250 m distance and
the fixed measurement all show elevated noise levels between
300 kHz and almost 10 kHz compared to the ambient noise.
The measurement taken at a distance of 450 m is overall
at levels close to the 20 min background noise, but, similar
to all other measurements performed when the turbine was
running, shows characteristic peaks at the same frequencies,
while background data results in a smooth line. These peaks
are consistent with the tones observed in the spectrograms
shown earlier.
The highest turbine noise levels occur below 1kHz in all
tests. In that range the fixed test reveals sound levels of
about 100 dB re µPa2/Hz, which is about 30 dB more than the
20-minute ambient noise sample. The one-third octave spectral
levels drop about 10 dB per 150m in the range between
300 Hz to 10 kHz.
Sediment generated noise has been shown in other studies
of tidal currents to be relevant in frequencies above 1 kHz for
gravel beds [10]. Evidence from divers suggests that the sea
floor at the test site and large areas of the Strangford Narrows
consists of small boulder on bedrock.
All floating measurements, except the one at 450 m range,
show a spike around 200 Hz , similar to the one observed in
the 60 s ambient noise measurement shown in Figure 6. This
spike may not be caused by turbine sound and since the spectra
below 100 Hz are comparable to background noise levels,
projection of these frequencies to source level estimates is
not appropriate. Therefore, these frequencies are also excluded
40
50
60
70
80
90
100
110
120
130
140
100 1000 10000 100000
[dB re µP a2/Hz]
Frequency [Hz]
Background
WenzM in
fixed 9m
100m
250m
450m
Fig. 7. Power spectral density over frequency for ambient and turbine noise
measured at a fixed location 9mfrom the turbine and floating tests at various
distances.
from the following plots.
Figure 5 shows a spectrogram highlighting the data that are
averaged to form the noise curves for the 450, 250, and 150
metre ranges shown in Figures 7 and 8. The track followed by
the buoy in each range are also shown in the map in Figure
1.
Figure 8 shows the previous data projected back to a
standardized distance of 1 m from the turbine, resulting in
broadband source level SL estimates. After the correction, all
drifting measurements align fairly well, indicating broadband
source levels of 120 dB re 1 µPa2/Hz @ 1m around 300 Hz,
decreasing slightly to 100 dB re µPa2/Hz @ 1m at 10 kHz.
It should be reiterated that the values shown in Figure 8 are
projected levels at 1 metre from the turbine, and the actual
signal level depends on range from the source, as shown in
Figure 7.
V. CONCLUSIONS
This paper presented results from an underwater noise mea-
surement campaign of a full scale tidal turbine. It was shown
that fixing the hydrophone relative to the flow introduces
low frequency noises. Allowing the hydrophone to drift with
the current significantly reduces the flow noise, but requires
considerable effort.
Noise characteristics of a turbine vary due to mode of
operation. Constant running, free spinning and braking can
readily be identified for the SCHOTTEL SIT turbine.
Considering the environmental background noise, third-
octave spectral levels indicate that turbine sound is relevant
from 100 Hz.
Floating measurements require correction for signal loss due
to varying range. A simple TL model appeared sufficient in
60
70
80
90
100
110
120
130
1000 10000
[dB re µP a2/Hz]
Frequency [Hz]
Background
WenzM in
fixed 9m
100m
250m
450m
Fig. 8. Third-octave spectral levels for turbine noise adjusted to a standardized
distance of 1mfrom the source, compared to the 20-minute ambient noise
sample.
this case because corrected results for different ranges agree
well.
Overall sound levels of the SCHOTTEL SIT turbine are on
the same order as natural and other anthropogenic background
noise measured at the test site. However, further analysis is
required to determine the range of ambient noise conditions
experienced at the test site.
During these tests, the current speed was estimated at
approximately 1-2 m
/s. However, higher current speeds occur at
the test site, and the turbine is expected to operate in higher
flow conditions. Therefore acoustic measurements in higher
flow conditions are required to understand the full range of
acoustic signal levels emitted by the turbine.
REF ER EN CE S
[1] G. Cada, J. Ahlgrimm, M. Bahleda, T. Bigford,
S. D. Stavrakas, D. Hall, R. Moursund, and
M. Sale, “Potential impacts of hydrokinetic and
wave energy conversion technologies on aquatic
environments,” Fisheries, vol. 32, no. 4, pp. 174–181,
2007. [Online]. Available: http://dx.doi.org/10.1577/
1548-8446(2007)32[174:PIOHAW]2.0.CO;2
[2] B. Polagye, B. V. Cleve, A. Copping, and K. Kirkendall,
Eds., Environmental Effects of Tidal Energy Develop-
ment. U.S. Dept. Commerce, NOAA Tech. Memo.
NMFS F/SPO-116, March 2010.
[3] R. A. Kastelein, S. van der Heul, W. C. Verboom,
N. Jennings, J. van der Veen, and D. de Haan, “Startle
response of captive north sea fish species to underwater
tones between 0.1 and 64khz,” Marine Environmental
Research, vol. 65, no. 5, pp. 369 – 377, 2008.
[4] W. T. Ellison, B. L. Southall, C. W. Clark, and
A. S. Frankel, “A new context-based approach to assess
marine mammal behavioral responses to anthropogenic
sounds,” Conservation Biology, vol. 26, no. 1, pp.
21–28, 2012. [Online]. Available: http://dx.doi.org/10.
1111/j.1523-1739.2011.01803.x
[5] M. K. Pine, A. G. Jeffs, and C. A. Radford, “Turbine
sound may influence the metamorphosis behaviour
of estuarine crab megalopae,” PLoS ONE, vol. 7,
no. 12, p. e51790, 12 2012. [Online]. Available:
http://dx.doi.org/10.1371%2Fjournal.pone.0051790
[6] The Crown Estate, “Review of current knowledge of
underwater noise emissions from wave and tidal stream
energy devices,” August 2013.
[7] N. D. Merchant, E. Pirotta, T. R. Barton, and P. M.
Thompson, “Monitoring ship noise to assess the
impact of coastal developments on marine mammals,”
Marine Pollution Bulletin, vol. 78, no. 12, pp. 85 –
95, 2014. [Online]. Available: http://www.sciencedirect.
com/science/article/pii/S0025326X13006802
[8] H. Bailey, B. Senior, D. Simmons, J. Rusin, G. Picken,
and P. M. Thompson, “Assessing underwater noise levels
during pile-driving at an offshore windfarm and its
potential effects on marine mammals,” Marine Pollution
Bulletin, vol. 60, no. 6, pp. 888 – 897, 2010. [Online].
Available: http://www.sciencedirect.com/science/article/
pii/S0025326X10000044
[9] P. Thorne, “Laboratory and marine measurements
on the acoustic detection of sediment transport.”
Journal of the Acoustical Society of America,
vol. 80, no. 3 , Sep. 1986, p.899-910., 1986, cited
By 15. [Online]. Available: http://www.scopus.com/
inward/record.url?eid=2-s2.0-0022547156&partnerID=
40&md5=eb14e7a6b63a0520140b95adf74516ec
[10] C. Bassett, J. Thomson, and B. Polagye, “Sediment-
generated noise and bed stress in a tidal channel,”
Journal of Geophysical Research: Oceans, vol.
118, no. 4, pp. 2249–2265, 2013, cited By 4.
[Online]. Available: http://www.scopus.com/inward/
record.url?eid=2-s2.0-84878028623&partnerID=40&
md5=875d7f5f966e8402fece098bbbb5d342
[11] M. Strasberg, “Nonacoustic noise interference in
measurements of infrasonic ambient noise,” The Journal
of the Acoustical Society of America, vol. 66, no. 5, pp.
1487–1493, 1979. [Online]. Available: http://scitation.
aip.org/content/asa/journal/jasa/66/5/10.1121/1.383543
[12] C. Bassett, J. Thomson, P. H. Dahl, and
B. Polagye, “Flow-noise and turbulence in two tidal
channels,” The Journal of the Acoustical Society of
America, vol. 135, no. 4, pp. 1764–1774, 2014.
[Online]. Available: http://scitation.aip.org/content/asa/
journal/jasa/135/4/10.1121/1.4867360
[13] P. Jeffcoate, R. Starzmann, B. Elsaesser, S. Scholl,
and S. Bischoff, “Field measurements of a
full scale tidal turbine,” International Journal of
Marine Energy, no. 0, pp. –, 2015. [Online].
Available: http://www.sciencedirect.com/science/article/
pii/S221416691500017X
[14] W. Richardson, J. Charles Greene, C. Malme, and
D. Thomson, Marine Mammals and Noise, 1st ed. Aca-
demic Press, 1995.
[15] Underwater Noise Measurement Good Practice Guide
No. 133, National Physical Laboratory, 2014, crown
Copyright.
[16] G. M. Wenz, “Acoustic ambient noise in the ocean:
Spectra and sources,” The Journal of the Acoustical
Society of America, vol. 34, no. 12, pp. 1936–1956,
1962. [Online]. Available: http://scitation.aip.org/content/
asa/journal/jasa/34/12/10.1121/1.1909155
