Content uploaded by Alexander O MacGillivray
Author content
All content in this area was uploaded by Alexander O MacGillivray on Aug 21, 2019
Content may be subject to copyright.
Slowing deep-sea commercial vessels reduces underwater
radiated noise
Alexander O. MacGillivray,
a)
Zizheng Li, and David E. Hannay
JASCO Applied Sciences, 4464 Markham Street, Victoria, British Columbia, V8Z 7X8, Canada
Krista B. Trounce and Orla M. Robinson
Vancouver Fraser Port Authority, 100 The Pointe, 999 Canada Place, Vancouver, British Columbia, V6C 3T4,
Canada
(Received 4 March 2019; revised 13 June 2019; accepted 20 June 2019; published online 23 July
2019)
During 2017, the Vancouver Fraser Port Authority’s Enhancing Cetacean Habitat and Observation
program carried out a two-month voluntary vessel slowdown trial to determine whether slowing to
11 knots was an effective method for reducing underwater radiated vessel noise. The trial was car-
ried out in Haro Strait, British Columbia, in critical habitat of endangered southern resident killer
whales. During the trial, vessel noise measurements were collected next to shipping lanes on two
hydrophones inside the Haro Strait slowdown zone, while a third hydrophone in Strait of Georgia
measured vessels noise outside the slowdown zone. Vessel movements were tracked using the auto-
mated identification system (AIS), and vessel pilots logged slowdown participation information for
each transit. An automated data processing system analyzed acoustical and AIS data from the three
hydrophone stations to calculate radiated noise levels and monopole source levels (SLs) of passing
vessels. Comparing measurements of vessels participating in the trial with measurements from con-
trol periods before and after the trial showed that slowing down was an effective method for reduc-
ing mean broadband SLs for five categories of piloted commercial vessels: containerships
(11.5 dB), cruise vessels (10.5 dB), vehicle carriers (9.3 dB), tankers (6.1 dB), and bulkers (5.9 dB).
V
C2019 Author(s). All article content, except where otherwise noted, is licensed under a Creative
Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
https://doi.org/10.1121/1.5116140
[JFL] Pages: 340–351
I. INTRODUCTION
Marine shipping has long been recognized as a major
source of underwater noise (Wenz, 1962) and is often the pre-
dominant source of man-made noise near major ports and ship-
ping routes (NRC, 2003). Chronic noise from marine shipping
has the potential to negatively affect marine animals that use
sound for performing critical life functions (Richardson et al.,
1995). A number of at-risk cetacean species inhabit the pro-
tected waters of southern British Columbia and northern
Washington State, referred to as the Salish Sea. In this region,
marine shipping is the dominant source of underwater ambient
noise (Bassett et al.,2012). Key among these species is the
endangered southern resident killer whale (SRKW), with a
population of only 74 individuals as of 2018 (Center for
Whale Research, 2018). This population was designated as
endangered under Canada’s Species at Risk Act in 2001,
which initiated the development of a recovery strategy (DFO,
2011,2016) and an action plan (DFO, 2017) to address the
current threats to northern resident killer whales and SRKWs
in Canadian Pacific waters. This recovery strategy designates
much of the Salish Sea as SRKW critical habitat—the habitat
necessary for the survival or recovery of the species. Under the
U.S. Endangered Species Act, critical habitat has also been
designated over much of the U.S. waters of the Salish Sea.
These designations offer the species legal protection of vital
habitat functions (e.g., ability to feed, socialize, and rest). As
with other odontocetes, killer whales (Orcinus orca) use sound
to navigate, communicate, and locate prey via echolocation.
Thus, underwater noise generated by vessels can degrade their
acoustical habitat and impede their life functions.
As part of its mandate to better understand and manage the
impact of shipping activities on at-risk whales, the Vancouver
Fraser Port Authority’s (VFPA’s) Enhancing Cetacean Habitat
and Observation (ECHO) program organized and managed a
voluntary vessel slowdown trial in Haro Strait. The trial investi-
gated noise reductions in SRKW habitat that could be obtained
by asking vessels to voluntarily reduce their speed through
water to 11 knots. Vessels are usually quieter when traveling
more slowly due to decreased propeller cavitation and machin-
ery vibration. The trial focused primarily on piloted commercial
vessels, but all types of motorized water craft were encouraged
to participate. The trial ran from 7 August to 6 October 2017,
during the time of year when SRKW density is historically high-
est in Haro Strait. Vessel source level (SL) measurements were
carried out during the slowdown trial by JASCO Applied
Sciences (JASCO), an international acoustical consulting and
applied research company with Canadian branches located in
Victoria, British Columbia, and Dartmouth, Nova Scotia.
This article describes the dedicated acoustic SL study
that was carried out before, during, and after the slowdown
trial to measure how vessel noise emissions were affected by
a)
Electronic mail: alex.macgillivray@jasco.com
340 J. Acoust. Soc. Am. 146 (1), July 2019 V
CAuthor(s) 2019.0001-4966/2019/146(1)/340/13
the slowdown protocol. Calibrated sound recordings were col-
lected on two hydrophone stations, situated directly adjacent
to the northbound and southbound Haro Strait traffic lanes, to
obtain high-quality SL measurements of individual vessels. A
land-based automated identification system (AIS) receiver
tracked vessels passing the Haro Strait hydrophone stations
during the trial. Additional measurements from a third, cabled
hydrophone station in the Strait of Georgia were used for
measuring noise from vessels transiting at normal speed after
leaving the slowdown zone.
Hydrophone data from these three stations were ana-
lyzed using an automated vessel noise measurement system
(PortListen
V
R
) developed by JASCO. This system tracked
passing vessels on AIS and automatically measured their
underwater acoustic SLs using calibrated hydrophone data.
To determine if slowdowns are an effective mitigation
method for reducing vessel noise emissions, SLs were mea-
sured during the slowdown trial and the pre-trail and post-
trial control periods.
II. METHODS
A. Slowdown trial overview
JASCO deployed two autonomous hydrophone stations
inside the slowdown zone (Fig. 1) to measure SLs of transit-
ing vessels. These Haro Strait hydrophones were installed one
month before the trial started (on 6 July 2017) and removed
three weeks after the trial ended (on 26 October 2017). The
purpose of collecting data outside the trial period was to mea-
sure baseline vessel noise emissions and to provide experi-
mental controls for the SL analysis. Additional vessel noise
measurements outside the slowdown zone were captured on a
cabled hydrophone array at the ECHO Strait of Georgia
Underwater Listening Station (ULS; Hannay et al., 2016).
From 7 August through 6 October 2017 (60 days), vessels
were requested to voluntarily reduce their speeds to a target of
11 knots through water inside a designated slowdown zone in
Haro Strait. At the completion of each piloted transit, British
Columbia (BC) Coast Pilots reported to the Pacific Pilotage
Authority (PPA) on the vessel’s participation. A dataset that
amalgamated the PPA logs of vessel participation with AIS
information on vessel speeds (corrected for water currents)
over the Haro Strait slowdown area was provided to JASCO
for correlation with the noise measurements.
B. Hydrophone stations
The Haro Strait hydrophone stations consisted of two
calibrated JASCO AMAR-G3 (Autonomous Multichannel
Acoustic Recorders-Generation 3) units, deployed on sub-
sea moorings next to the northbound and southbound traffic
lanes. Mooring deployments and retrievals in Haro Strait
were conducted using the R/V Richardson Point, a 20 m
research vessel. After deploying the moorings, their precise
on-bottom locations were surveyed to an accuracy of 64m
using a surface-based transducer that measured the distance to
their acoustic releases (Teledyne Benthos 875-T, Falmouth,
FIG. 1. (Color online) Map of slowdown trial boundary and hydrophone station locations in Haro Strait and Strait of Georgia. Locations of northbound and
southbound vessel traffic routes were based on historical AIS ship tracking data.
J. Acoust. Soc. Am. 146 (1), July 2019 MacGillivray et al. 341
MA). Water depths were 250 m at the northbound-lane
AMAR and 210 m at the southbound-lane AMAR with hydro-
phones situated 3 m above the seabed.
Each AMAR used an M36 omnidirectional hydrophone
(GeoSpectrum Technologies Inc., Dartmouth, Nova Scotia,
Canada, 165 63dBre1V/lPa nominal sensitivity) for mea-
suring underwater sound pressure. The AMARs were pro-
grammed with a variable-bandwidth recording cycle to sample
acoustic data at 96000 Hz for 21 h per day [09:00–06:00 PDT
(Pacific Daylight Time)] and 128000 Hz for 3 h per day
(06:00–09:00 PDT). The recording channel had 24-bit resolu-
tion with a spectral noise floor of 20 dB re 1 lPa
2
/Hz and a
nominal ceiling of 168 dB re 1 lPa.EachAMARstoredthe
hydrophone data on 1792 GB of internal solid-state flash mem-
ory. The frequency-dependent laboratory calibration of each
AMAR and hydrophone was verified before and after deploy-
ment at 250 Hz using a Pistonphone Type 42AC precision
sound source (G.R.A.S. Sound and Vibration A/S, Holte,
Denmark) to ensure the sensitivity of the hydrophones did not
change over the deployment period.
Additional vessel noise measurements at normal transit
speeds were captured on the ECHO Strait of Georgia ULS,
which was a real-time hydrophone node installed on the
Victoria Experimental Network Under the Sea (VENUS)
Observatory operated by Ocean Networks Canada. The
Strait of Georgia ULS was situated on the seabed at 173 m
water depth in the northbound traffic lane, approximately
30 km southwest of Vancouver. It recorded hydrophone data
at a sampling rate of 64000 Hz with 24-bit resolution using a
digital streaming version of the same AMAR G3 electronics
used in the Haro Strait autonomous recorders.
C. AIS receiver
An AIS receiver was deployed atop Observatory Hill,
approximately 17 km west of the Haro Strait hydrophone sta-
tions (Fig. 1). The receiver, which was located near the hilltop
at the Herzberg Institute of Astrophysics, captured ship track-
ing data for the duration of the slowdown trial. The AIS
receiver consisted of an SR161 scanning VHF (very high fre-
quency) radio receiver and a 1.22 m whip antenna connected
to a notebook personal computer (PC). Logging software
(NMEA Router, Arundale) stored the raw AIS records on an
internal hard disk on the PC, and chart plotting software
(ShipPlotter, COAA, Portim~
ao, Portugal) displayed the
received ship tracks in real time. Data from the AIS receiver
were periodically backed up to JASCO’s servers in Victoria
over the cellular network via a mobile Wi-Fi stick. Raw AIS
data from the receiver were fed into the PortListen system to
analyze the vessel SLs. Vessel tracks from the AIS receiver
were matched to vessel transits recorded in the pilot logs
according to international maritime organization (IMO) num-
bers and transit times.
D. SL calculation
The PortListen system analyzes underwater radiated
noise measurements and AIS broadcasts from passing ves-
sels to calculate vessel SLs in terms of radiated noise level
(RNL) and monopole source level (MSL). While both RNL
and MSL measurements are collectively referred to here as
SLs, RNL is actually an affected SL (i.e., a SL measurement
affected by surface and seabed reflections). Only MSL
strictly corresponds to the ISO standard definition of a SL
(ISO, 2017). Both quantities are reported here so that mea-
surements from this study can be more easily compared with
those from other studies.
For time periods when a passing vessel was detected on
AIS, the system processed hydrophone data to obtain standard
decidecade (i.e., 1/3-octave) band sound pressure level (SPL)
inside a data window encompassing 630of the vessel’s
closest point of approach (CPA) to the hydrophone, according
to the methods specified in the ANSI S12.64 ship noise mea-
surement standard (ANSI, 2009). PortListen automatically
determined the measurement window and processed a single
channel of hydrophone data in 1-s periods stepped in 0.5-s
intervals (Fig. 2) using a Hanning-windowed fast Fourier
transform (FFT). It used the AIS speed and vessel position
together with a cepstral analysis of the Lloyd-mirror pattern
to determine the timing and location of the CPA of the vessel
to the hydrophone. The software calculated background noise
in each frequency band when the vessel was more than 2 km
away, and the measured vessel sound levels were corrected if
they exceeded background by 3–10 dB or rejected if they
were less than 3dB above background, per the ANSI S12.64
standard (ANSI, 2009).
RNL was calculated assuming spherical spreading prop-
agation loss [i.e., PL ¼20 log
10
(r)], per the ANSI standard,
whereas MSL was calculated using a frequency-dependent
PL model, based on the numerical solution of the acoustic
wave equation, which accounts for the effect of the environ-
ment on sound transmission. Since no single acoustic model
is applicable at all sampled ranges and frequencies, a hybrid
PL model was used to calculate MSL as follows:
(1) At frequencies less than 4 kHz and ranges less than
120 m, PL was calculated using a wavenumber integra-
tion model (Hannay et al., 2010;Jensen et al., 2011),
which computes reflection coefficients for layered elastic
media (Brekhovskikh, 1980).
FIG. 2. (Color online) Spectrogram of a single vessel measurement, show-
ing the CPA time (dashed line) and the measurement window (black box)
used for calculating vessel SLs. Acoustic data were processed using 1-s fast
Fourier transforms (FFTs; 50% overlap), shaded using a power-normalized
Hanning window.
342 J. Acoust. Soc. Am. 146 (1), July 2019 MacGillivray et al.
(2) At frequencies less than 4 kHz and ranges greater than
240 m, PL was calculated using a wide-angle parabolic
equation model (Collins, 1993), modified to treat reflec-
tion losses for an elastic seabed using a complex-density
equivalent fluid approximation (Zhang and Tindle, 1995).
(3) At frequencies less than 4 kHz and ranges between 120 m
and 240 m, PL was calculated from the average of the
parabolic equation and wavenumber integration models.
(4) At frequencies greater than 4 kHz, PL was calculated
using an image-method model (Brekhovskikh and
Lysanov, 2003), which accounts for surface and seabed
reflection coefficients and frequency-dependent absorp-
tion (Franc¸ois and Garrison, 1982).
Average PL in each decidecade band was based on the
mean propagation factor calculated at 50 frequencies, which
were spaced logarithmically between the minimum and max-
imum band limits. Mean source depth for the MSL calcula-
tion was taken to be half the vessel static draft reported on
AIS. The PL was smoothed by assuming the source depth
had a Gaussian distribution in a manner similar to Wales and
Heitmeyer (2002), in which the standard deviation was taken
to be 30% of the source depth. Additional details regarding
the automated SL measurement system are given in Hannay
et al. (2016). Seabed geoacoustic profiles for the Haro Strait
and Strait of Georgia sites were determined via inversion of
transmission loss measurements obtained using a controlled
sound source (Warner et al., 2014).
To ensure high data quality, only SL measurements with
CPA less than 1000 m from the hydrophones in Haro Strait
were accepted for subsequent analysis. In addition, an experi-
enced acoustic analyst performed a manual and systematic
quality review of every SL measurement. For each measure-
ment, the analyst inspected the vessel track, spectrogram, back-
ground noise levels, received levels, and SLs. Measurements
were rejected under the following circumstances:
(1) When other AIS vessels were present within six times
the measured CPA of the vessel of interest;
(2) When spectrograms visibly contained contaminating
noise from sources other than the vessel of interest
(including non-AIS vessels);
(3) When measurements had three or more decidecade
bands with signal-to-noise-ratio less than 3 dB in the
range 50–1000 Hz;
(4) When vessel tracks had an unsteady speed or heading in
the measurement window.
E. Noise reduction analysis
SL measurements of piloted vessels in Haro Strait (i.e.,
inside the trial boundary) were assigned to one of the follow-
ing trial groups based on the pilot participation logs:
(1) Control (measurements outside the trial period),
(2) Trial non-participants (vessels that did not slow down
for the trial), and
(3) Trial participants (vessels that slowed down for the trial).
Five categories of vessels were captured in sufficient
numbers in the pilot logs to be included in the trial groups:
(1) Containerships,
(2) Bulk carriers and general cargo (hereafter referred to as
“bulkers”),
(3) Tankers,
(4) Vehicle carriers, and
(5) Cruise ships (hereafter referred to as “cruise”).
Measurements were assigned to these categories based
on vessel types identified in the pilot logs. Other categories
of vessels were measured on the Haro Strait hydrophones
but were not represented in the pilot logs and were therefore
excluded from the trial groups. In rare instances, the vessel
participation status from the pilot dispatch did not match the
AIS speed. This resulted in participating vessels being iden-
tified as non-participants or vice versa. To account for these
discrepancies, a small number of measurements with outly-
ing speeds in the non-participant and participant groups were
discarded from the trial groups.
SL measurements were analyzed in the following three
frequency bands, which were recently identified by an expert
working group convened by the Coastal Ocean Research
Institute (CORI; Heise et al., 2017) as being best suited for
assessing the acoustical quality of SRKW habitat:
(1) Broadband (10–100 000 Hz) for evaluating behavioral or
physiological impacts,
(2) Communication masking (500–15 000 Hz) for evaluating
effects of noise on SRKW communication space, and
(3) Echolocation masking (15 000–100 000 Hz).
MSL measurements were evaluated for all three SRKW
frequency bands (broadband, 0.5–15 kHz, and >15 kHz),
whereas RNL measurements were evaluated only for broad-
band noise. MSL was the preferred metric for reporting SLs
in the SRKW communication and echolocation bands
because MSL better accounts for the effect of the environ-
ment on vessel SLs (e.g., from absorption, surface, and sea-
bed reflections) than RNL.
Trends of SL versus speed were analyzed based on speed
through water, as calculated from vessel speed over ground
(from AIS) with speed and direction data for surface currents.
Surface currents in the Strait of Georgia were obtained
directly from acoustic Doppler current profiler (ADCP) mea-
surements at the ECHO ULS. Time-dependent surface cur-
rents in Haro Strait were predicted using the Bedford Institute
of Oceanography’s WebTide model (Hannah et al.,2008).
Surface current measurements were not available for verify-
ing the model in Haro Strait, but the WebTide predictions
were found to be in good agreement with ADCP data at the
ECHO ULS (mean and standard deviation model-data resid-
uals were 0.03 60.11 m/s).
Where SL measurements were obtained at different
speeds, trend analysis was performed using Ross’s classical
power law model (Ross, 1976), which relates change in SL
to relative changes in speed,
SL v
ðÞ
SLref ¼Cv10 log10
v
vref
:(1)
In this equation, SL(v) is the SL at speed through water
v,SL
ref
is the SL at some reference speed v
ref
, and C
v
is a
J. Acoust. Soc. Am. 146 (1), July 2019 MacGillivray et al. 343
coefficient corresponding to the slope of the curve. Equation
(1) was assumed to apply both to RNL and MSL.
The effects of voluntary slowdowns on vessel noise
emissions were evaluated by comparing measurements in
the participant group with measurements in the non-
participant and control groups. The statistical significance of
differences between the three trial groups were tested using
a pairwise-ttest
1
to test the experimental hypothesis (e.g.,
that mean SLs in the participant group were significantly
lower than in the control group) against the null hypothesis
(e.g., that mean SLs in the participant group were the same
as in the control group).
III. RESULTS
A. Measurement summary
A total of 2765 vessel SL measurements were collected
during the control and trial periods on the two hydrophone
stations in Haro Strait and on the Strait of Georgia ULS. Of
these measurements, which included non-piloted vessels,
1930 were accepted (i.e., passed the manual quality review).
A total of 920 (out of 951) transits in the pilot logs were
matched to vessel measurements during the slowdown trial
period (many northbound transits matched two different
measurements, as vessels were measured once in Haro Strait
and once again in the Strait of Georgia). Summary statistics
of vessel design characteristics and drafts were calculated
from data broadcast over AIS during the study (Fig. 3).
Based on the logs, a total of 1340 SL measurements
were assigned to the three experimental groups identified in
Sec. II E. Twenty-three of these measurements were dis-
carded because their measured speeds were inconsistent with
the pilot participation logs, resulting in 1317 valid SL mea-
surements for analysis. Analysis of the vessel tracks showed
that 94% of vessels transiting through the slowdown zone
passed within 1 km of the two hydrophones (mean and stan-
dard deviation horizontal CPA ¼445 6313 m).
To investigate how speeds in different vessel categories
were affected by slowdown participation, we calculated sta-
tistics of vessel speeds through water for the participant,
non-participant, and control groups (Fig. 4). Of the 951
piloted transits through Haro Strait during the slowdown
trial, 577 transits (or 61%) were reported as participating.
While not all participating vessels achieved the target slow-
down speed of 11 knots through water, 75% of participating
vessels traveled 12 knots or slower and 95% of participating
vessels traveled 13 knots or slower at the time of measure-
ment on the Haro Strait hydrophones. Pairwise t-tests
showed that mean speeds in the participant group were sig-
nificantly less than those in the non-participant and control
groups for all categories (Table I). Mean speed reductions
were greatest for containerships (7.7 knots) and smallest for
bulkers (2.1 knots).
The mean speeds of non-participant vessels were also
found to be slightly lower than the control group for bulkers
and vehicle carriers (cruise vessels had too few non-
participant transits to compare with the trial period). This
could indicate that non-participant vessels in these categories
also slowed down slightly during the trial period, or it could
be due to a small number of participating vessels being
recorded as non-participants in the pilot logs. Regardless of
the reason, the SL reduction for each category was therefore
calculated based only on the differences between the partici-
pant group and the control group.
Vessel speed and SL statistics were calculated sepa-
rately for measurements in seven additional vessel categories
that were not captured in the pilot logs and therefore lacked
information on slowdown participation (tug, fishing, govern-
ment/research, naval, small passenger, recreational, and
other vessels). In Haro Strait, 225 measurements in these
FIG. 3. (Color online) Box-and-whisker plots showing summary statistics for breadth, length, draft, and year built, versus vessel category, as broadcast over
AIS. The total number of samples is indicated to the left of each box. Missing values are not counted, and some obviously incorrect values have been manually
removed. Statistics of breadth, length, and year built are for unique vessels only (i.e., multiple measurements of the same vessel are counted only once). The
ends of the box show the upper and lower quartiles and the line inside the box shows the median. The whiskers and dots extend outside the box to the highest
and lowest observations, where the dots correspond to observations that fall more than 1.5 IQR (interquartile range) beyond the upper and lower quartiles.
344 J. Acoust. Soc. Am. 146 (1), July 2019 MacGillivray et al.
7 categories were assigned to 1 of 2 groups (control and
trial), depending on when they were collected. Of these cate-
gories, only naval vessels were found to have a statistically
significant reduction in mean speed (5.3 knots, p¼0.001)
between the control and trial periods. While they were not
captured in the pilot participation logs, vessels from the
Royal Canadian Navy (RCN) were anecdotally confirmed as
participating in the trial (all but three accepted measure-
ments in the naval category were RCN vessels). Vessels in
the remaining six categories did not appear to have reduced
their speeds in significant numbers during the trial.
B. Effect of trial participation on SLs
To investigate the effects of the voluntary slowdown on
vessel SLs, statistics of SLs across the participant, non-
participant, and control groups were calculated (Fig. 5). In
all five vessel categories, SLs (RNL and MSL) were lower in
the participant group than in the control group at the 5th,
25th, 50th, 75th, and 95th quantiles. Pairwise t-tests showed
strong evidence that mean SLs in the participant group were
lower than those in the non-participant and control groups
for all categories (Table II). Mean broadband SL (MSL)
reductions ranged from 5.9 dB for bulkers to 11.5 dB for con-
tainerships. The CORI-band analysis showed that the
reductions were frequency dependent with smallest mean
reductions (3.1–10.8 dB) in the SRKW communication
masking band (0.5–15 kHz) and greatest mean reductions
(5.1–17.8 dB) in the SRKW echolocation band (>15 kHz).
There was no evidence of significant differences in mean
SLs between the non-participant and control groups for any
vessel category, except for the broadband MSL of the bulker
category, which was 1 dB lower during the trial for non-
participants. This latter difference may be due to non-
participating bulkers traveling 0.4 knot slower, on average,
during the trial period than during the control period (see
Table I). Note, however, that RNL source levels and MSLs
above 500 Hz for bulkers did not show any significant differ-
ences between the non-participant and control groups.
Noise reductions from the Haro Strait measurements were
used to calculate equivalent speed scaling coefficients, accord-
ing to Eq. (1), for the Ross power-law model (Table III).
These speed coefficients (C
v
) reflect the mean decibel reduc-
tions in SLs that were associated with the mean speed reduc-
tion measured in each category. It is important to note that the
Ross model measures decibel changes in SL for relative
changes in speed. Thus, for example, while the MSL scaling
coefficient for bulkers (C
v
¼8.0) was greater than that for con-
tainerships (C
v
¼5.0), the overall broadband MSL reduction
for bulkers (5.5 dB) was still smaller than that for container-
ships (11.0 dB) because participant containerships reduced
their speed by a larger relative amount during the trial.
To investigate how slowdown participation affected
frequency-dependent noise emissions, mean decidecade-
band SLs (MSL) between the participant, non-participant,
and control groups were compared (Fig. 6). SL reductions
for participant vessels showed similar frequency dependence
for all five categories: the largest reductions were generally
below 100 Hz and above 1000 Hz, and the smallest reduc-
tions were in the intermediate-frequency range. This
frequency-dependence was likely due to the different noise
generating mechanisms that dominate different parts of the
radiated vessel noise spectrum—e.g., cavitation often domi-
nates at very low and very high frequencies, whereas
FIG. 4. (Color online) Box-and-whisker plots for five vessel categories, comparing speed through water in Haro Strait (at CPA) between the control, non-
participant, and participant groups (accepted measurements only, excluding outlier speeds). The total number of measurements is indicated below each box.
See Fig. 3for an explanation of the box-and-whisker plots.
TABLE I. Differences in mean vessel speeds (knots) between the partici-
pant, non-participant, and control groups (accepted measurements only).
Asterisks indicate the statistical significance of the differences as determined
from pairwise t-tests (* ¼p<0.05, ** ¼p<0.01, *** ¼p<0.001).
Boldface indicates a statistically significant difference (p<0.05). Dashes
indicate insufficient data.
Vessel category
Control versus
Participant
Control versus
Non-participant
Non-participant
versus Participant
Bulker 2.09*** 0.40** 1.69***
Containership 7.67*** 0.19 7.48***
Cruise 6.15*** — —
Tanker 2.30*** 0.07 2.23***
Vehicle Carrier 5.89*** 1.03* 4.87***
J. Acoust. Soc. Am. 146 (1), July 2019 MacGillivray et al. 345
machinery noise dominates at middle frequencies (Kipple
and Gabriele, 2007). A peak in the 25-kHz band for cruise
vessels was observed during at least six cruise vessel mea-
surements from the post-trial control period. Discussions
with the cruise ship operators indicated that these narrow-
band noise emissions originated from depth sounders used
for navigational safety. We found no evidence of similar
noise emissions during the slowdown trial period. As the
sample size for cruise vessels was small (14 control period,
16 participating, and 2 non-participating), the peak at 25 kHz
resulting from the use of depth sounders by some vessels
limited the comparative analysis between control and trial
periods in this decidecade band (and in the SRKW echoloca-
tion band >15 kHz).
For the seven categories of vessels not captured in the
pilot logs, only naval vessels showed clear evidence of
reduced noise emissions during the trial period. The mean
broadband MSL of RCN vessels was 6.3 dB lower during the
trial than during the control period, but this value should not
be considered a high-confidence estimate because it was
based on a small number of measurements (n¼19).
Furthermore, SLs of naval vessels were already quite low,
which resulted in relatively high numbers of rejections due
to insufficient signal-to-noise ratios.
C. Effect of speed on SLs for individual vessels
Several factors other than speed, such as draft, size, and
loading, may influence the underwater noise emissions of
individual vessels. Furthermore, different vessels in the
same category may normally transit at different service
speeds and have different baseline noise emissions (due to
differences in vessel design). To control for these effects,
reduced-speed SLs measured on the Haro Strait hydrophones
were directly compared with service-speed SLs of the same
vessels measured on the Strait of Georgia ULS. To ensure
consistency of vessel operating conditions (e.g., loading and
draft), measurements were compared only if they were col-
lected on the same northbound transit under direction of the
same pilot. A total of 107 matched pairs of measurements
met these criteria (Fig. 7). Data from these repeated mea-
surements were then fit to the Ross model [Eq. (1)] to deter-
mine the trend of vessel SLs with changes in speed (Table
IV). Measurements from all five categories of vessels were
pooled to fit a common trend line to all the data. In addition,
separate trend lines were fit to the three vessel categories
(bulker, containership, and tanker) that contained enough
measurements to permit a regression analysis.
The best-fit speed scaling coefficients for all the data
were highly significant and strongly positive (3.4 <C
v
<6.5) for all four SL metrics (a higher C
v
value corresponds
to a greater reduction of SL with decreasing speed). This
analysis showed that reductions in vessel SLs associated
with slower speeds were greatest above 15 kHz in the
SRKW echolocation band and lowest between 0.5–15 kHz in
the SRKW communication band. The slope of the MSL
trend was greater than that of the RNL trend, which indicates
frequencies below 50 Hz were more strongly affected by
changes in speed (MSL is more heavily weighted toward
low frequencies). These results are consistent with the find-
ings of the decidecade-band analysis (Sec. IIIB), which
showed SL reductions were greatest at low and high frequen-
cies (see Fig. 6). The category-specific trends were largely
consistent with the trends of the pooled measurements to
within the estimated uncertainty bounds of the best-fit coeffi-
cients. Therefore, the speed scaling coefficients derived from
measurements of individual vessels were broadly consistent
with speed scaling coefficients derived from analysis of the
slowdown trial groups (i.e., Table III).
FIG. 5. (Color online) Box-and-whisker plots for the five vessel categories, comparing SLs in Haro Strait between the control, non-participant, and participant
groups: broadband RNL, broadband MSL, MSL (0.5–15 kHz; SRKW communication masking), and MSL (15þkHz; SRKW echolocation masking). The total
number of measurements is indicated below each box. See Fig. 3for an explanation of the box-and-whisker plots.
346 J. Acoust. Soc. Am. 146 (1), July 2019 MacGillivray et al.
IV. DISCUSSION
A. Comparison with past studies
Previous studies of underwater radiated noise from
marine shipping generally indicate that SLs are proportional
to vessel speed (in accordance with intuition) but reported
trends were not necessarily consistent between studies.
Measurements of post-World-War-II shipping suggested a
strong power-law relationship between SLs and vessel speed
(C
v
¼5–6), and that this trend carried across different vessel
types (Ross, 1976). Measurements of individual cargo
vessels (Arveson and Vendittis, 2000) and cruise vessels
(Kipple and Gabriele, 2007) also showed a strong positive
relationship between SL and speed that was broadly consis-
tent with the Ross model. On the other hand, statistical anal-
yses based on measurements of large numbers of vessels in a
seaway reported insignificant or relatively weak speed
trends: two studies found no significant relationship with
speed (Wales and Heitmeyer, 2002;McKenna et al., 2012)
and three other studies found speed trends in the range
0.8–1.1 dB/knot (McKenna et al., 2013;Simard et al., 2016;
Veirs et al., 2016), which were weaker than the trends
reported here (Table V). The ability of these latter studies to
determine speed trends was likely limited because vessels in
a seaway typically do not deviate substantially from their
design speeds.
The present study overcame this limitation by imple-
menting a voluntary slowdown protocol, which provided the
necessary experimental controls for a high-confidence deter-
mination of source-level-versus-speed trends for different
categories of vessels. Gray and Greeley (1980) showed that
changes in a vessel’s SL depend on its design speed, not nec-
essarily on its absolute speed. Thus, grouping similar vessels
was also important for determining speed trends (e.g., as rec-
ognized by McKenna et al., 2013) because different types of
vessels have different design speeds. The large number of
measurements collected in the present study permitted
grouping measurements of similar vessels into distinct cate-
gories, while maintaining sufficient numbers of measure-
ments in each category for statistical confidence. The larger
speed variations within each category gave the present study
greater power than previous studies to accurately character-
ize speed trends of noise emissions. These two features are
likely the reasons for higher source-level-versus-speed
trends reported here than reported by past studies that ana-
lyzed measurements of large numbers of vessels in a
seaway.
It is interesting that the broadband trends reported here
for modern vessels (mean and standard deviation year built
was 2009 66.3) were broadly consistent with those origi-
nally reported by Ross for post-World-War-II vessels. The
broadband speed trends reported by Ross (1976) were attrib-
uted, in large part, to noise originating from propeller cavita-
tion. For fixed-pitch propellers (which comprise most of the
present-day merchant fleet), the intensity of cavitation noise
is related directly to propeller rotation rate and follows a
power law (Ross, 1976). This suggests that the broadband
noise reductions observed during the slowdown trial may
TABLE III. Equivalent power-law speed scaling coefficients (C
v
) calculated from the mean SL reductions measured during the slowdown trial. Speeds corre-
spond to mean speed through water recorded at the time of measurement for each category. SL reductions for each vessel category were taken to be the differ-
ence between the participant and control groups in Table II.
Category
Mean control
speed (knots)
Mean slowdown
speed (knots)
C
v
(Broadband RNL)
C
v
(Broadband MSL)
C
v
[MSL (0.5–15 kHz)]
C
v
[MSL (15þkHz)]
Bulker 13.47 11.38 7.59 8.07 4.17 7.01
Containership 18.88 11.21 4.93 5.08 4.10 7.85
Cruise 16.76 10.62 5.41 5.30 5.44 —
Tanker 13.68 11.39 7.24 7.63 4.51 9.88
Vehicle carrier 17.26 11.37 5.09 5.10 4.09 7.66
TABLE II. Differences in mean SLs (dB) between the participant, non-
participant, and control groups (accepted measurements only). The uncer-
tainty (6) indicates the 90% confidence interval of the difference (control
versus participant only). Asterisks indicate the statistical significance of the
differences, as determined from pairwise t-tests (* ¼p<0.05, ** ¼p
<0.01, *** ¼p<0.001). Boldface indicates a statistically significant dif-
ference (p<0.05). Dashes indicate insufficient or missing data (the MSL
[15þkHz] band for cruise ships in the control group was removed due to
noise from depth sounders—see the text).
Vessel category
Control versus
participant
Control versus
non-participant
Non-participant
versus participant
Broadband RNL
Bulker 5.56 60.75*** 0.26 5.30***
Containership 11.17 60.79*** 0.71 10.46***
Cruise 10.74 64.54*** ——
Tanker 5.78 61.70*** 1.08 4.70***
Vehicle carrier 9.24 61.14*** 0.08 9.16***
Broadband MSL
Bulker 5.91 60.74*** 1.02** 4.89***
Containership 11.52 60.88*** 1.08 10.44***
Cruise 10.52 65.20*** ——
Tanker 6.08 61.82*** 0.77 5.31***
Vehicle carrier 9.25 61.60*** 0.75 8.50***
MSL (0.5–15 kHz; SRKW communication masking)
Bulker 3.05 60.90*** 0.55 2.50***
Containership 9.29 60.98*** 0.20 9.09***
Cruise 10.78 64.15*** ——
Tanker 3.59 62.21*** 1.94 5.53***
Vehicle Carrier 7.42 61.24*** 0.02 7.40***
MSL (15þkHz; SRKW echolocation masking)
Bulker 5.18 61.32*** 0.00 5.18***
Containership 17.77 62.12*** 0.63 18.40***
Cruise — — —
Tanker 7.89 63.79*** 0.07 7.82***
Vehicle carrier 13.87 62.48*** 2.00 15.86***
J. Acoust. Soc. Am. 146 (1), July 2019 MacGillivray et al. 347
have been due, in large part, to reductions in cavitation noise,
although additional research is needed to rigorously identify
the importance of different noise generating mechanisms in
this dataset. There is also evidence that variable-pitch propel-
lers do not share the same noise-versus-rotation-rate charac-
teristics as fixed-pitch propellers (Traverso et al., 2015), and
therefore slowdowns may not provide similar noise savings
for vessels that use controllable-pitch propulsion (e.g., such as
many of the ferries in the Salish Sea).
B. Benefit of slowdowns for SRKW
One possible drawback of slowdowns as a noise mitiga-
tion approach is that they prolong the overall time of noise
exposure. This may result, e.g., in shorter durations of quiet
periods between vessel transits and an increase in the mini-
mum ambient noise level. Nonetheless, the overall noise
sound exposure level (SEL) in any band, measured at a sta-
tionary receiver and assuming other noise sources do not
FIG. 6. (Color online) Mean MSL in decidecade frequency bands for the control, non-participant, and participant groups (Haro Strait measurements only).
The total number of measurements in each group is indicated at the bottom of each panel. Missing data points below 30 Hz for non-participant cruise ships cor-
respond to frequency bands with no valid measurements (i.e., where background noise exceeded signal level).
348 J. Acoust. Soc. Am. 146 (1), July 2019 MacGillivray et al.
contribute, will be reduced for any single vessel transit pro-
vided that the speed scaling coefficient, C
v
, exceeds 1
(Frankel and Gabriele, 2017). This criterion was amply met
for all vessel categories and frequency bands analyzed in the
present study. Thus, slowdowns are expected to provide a
net reduction in noise exposure and average ambient noise
level in SRKW critical habitat.
It is nonetheless important to consider the frequency
dependence of specific mitigation measures when evaluating
their benefits since broadband measures do not necessarily
reflect the way that mammals use their acoustical environ-
ment. Analysis of the frequency dependence of the noise
reductions, using the CORI bands, showed that the speed-
related SL reductions were lower in the SRKW communica-
tion masking band (0.5–15 kHz) and greater in the SRKW
echolocation masking band (MSL >15 kHz). Furthermore,
sound propagation and baseline ambient noise is highly
frequency-dependent, so changes in SLs do not necessarily
reflect sound levels received by the animals themselves.
Additional studies carried out through the ECHO program
have used data presented here along with detailed modeling
of vessel noise propagation and SRKW distribution to quan-
tify the potential benefits of slowdowns on SRKW in their
critical habitat (Joy et al., 2019).
V. CONCLUSIONS
Analysis of 1317 SL measurements collected during the
Haro Strait slowdown trial, and during the pre- and post-trial
control periods, showed that slowing speed is an effective
method for reducing underwater radiated noise from com-
mercial vessels. We found statistically significant reductions
FIG. 7. (Color online) Change in SL versus speed ratio for vessels during the same northbound transit over the Haro Strait hydrophones and Strait of Georgia
ULS (i.e., inside versus outside the slowdown zone). The lines are the best-fit trendlines, based on Eq. (1). Each point on the plots represents the difference in
measured SLs for a pair of measurements (i.e., equal to the dB difference between the low-speed and high-speed measurement).
J. Acoust. Soc. Am. 146 (1), July 2019 MacGillivray et al. 349
in SLs for five categories of piloted vessels recorded on two
hydrophone stations during the control and trial periods:
containerships, bulkers, tankers, vehicle carriers, and cruise
vessels. In all cases, reductions in noise emissions were pro-
portional to changes in speed. Those categories with the fast-
est vessels (containerships, cruise vessels, and vehicle
carriers) exhibited the greatest reductions, whereas catego-
ries with slower vessels (tankers and bulkers) exhibited more
modest reductions. Mean reductions in broadband MSL
were 11.5 dB for containerships, 10.5 dB for cruise vessels,
9.3 dB for vehicle carriers, 6.1 dB for tankers, and 5.9 dB for
bulkers. Limited information on trial participation was avail-
able for seven other categories of non-piloted vessels.
Except for naval vessels, there was no strong evidence that
non-piloted vessels changed their speeds or reduced their
noise emissions during the trial, relative to the control
period.
Analysis of decidecade-band SLs for piloted vessels
showed that noise reductions associated with slowdown par-
ticipation were frequency dependent with the largest reduc-
tions measured at the low and high ends of the frequency
range and a minimum reduction in the 100–1000 Hz range.
This was also borne out in the CORI-band analysis, which
showed that MSL reductions were largest in the high-
frequency SRKW echolocation masking band (>15 kHz)
and smallest in the mid-frequency SRKW communication
masking band (0.5–15 kHz). The observed frequency depen-
dence may have been related to differences in how slowing
down affects the various noise generating mechanisms that
contribute to the radiated noise spectrum of marine vessels.
It is speculated that cavitation noise increases with speed at
all frequencies, but that machinery noise dominates at mid-
frequencies and has a weaker speed dependence. This would
explain larger radiated noise variations at low and high fre-
quencies than at mid-frequencies.
Results of the trial group analysis were cross-checked
by directly analyzing repeat SL measurements of the same
vessel transit at different speeds. A trend analysis of SLs ver-
sus change in speed based on 107 repeated measurements
between Haro Strait and the Strait of Georgia resulted in
overall frequency-dependent trends that were consistent with
the analysis of the trial groups in Haro Strait. Thus, source-
level-versus-speed trends for individual vessels were found
to be consistent with the reductions in noise emissions
achieved by slowing down the vessel population as a whole.
The speed trends identified in this study were broadly
consistent with those reported by Ross (1976) but greater
than those reported by several previous studies (Wales and
Heitmeyer, 2002;McKenna et al., 2013;Simard et al., 2016;
Veirs et al., 2016). It is believed that this is due to the volun-
tary slowdown protocol, which was unique to the present
study and provided an important experimental control with
greater power for characterizing the effects of speed on ves-
sel SLs.
ACKNOWLEDGMENTS
A large team of highly skilled technical experts
contributed to the success of this study. JASCO’s hardware
engineering team designed, assembled, and shipped the
AMAR systems and moorings. JASCO’s software
engineering team provided expert support for the PortListen
software. H
elo€
ıse Frouin-Mouy, Caitlin O’Neill, Graham
Warner, and Jennifer Wladichuk (JASCO) carried out the
TABLE IV. Best-fit power-law speed scaling coefficients (C
v
) calculated from SL measurements of the same vessels transiting northbound at different speeds
over the Haro Strait hydrophones and Strait of Georgia ULS. The r
2
value indicates the fraction of data variance explained by the fitted model (i.e., the coeffi-
cient of determination). The uncertainty (6) indicates the 90% confidence interval of the fit. Asterisks indicate the statistical significance of the trend (* ¼p
<0.05, ** ¼p<0.01, *** ¼p<0.001).
Fit parameter Broadband RNL Broadband MSL MSL (0.5–15 kHz) MSL (15þkHz)
All data (n¼107)
a
C
v
3.75 60.50*** 5.10 60.56*** 3.35 60.47*** 6.52 60.75***
r
2
0.677 0.755 0.653 0.752
Bulker (n¼63)
a
C
v
3.10 61.00*** 6.05 61.15*** 1.91 60.89*** 4.78 61.52***
r
2
0.385 0.646 0.232 0.405
Containership (n¼23)
a
C
v
4.17 60.78*** 4.96 60.71*** 3.97 60.55*** 6.93 60.94***
r
2
0.848 0.905 0.912 0.930
Tanker (n¼13)
C
v
3.78 62.74* 8.01 63.25*** 1.65 62.45 5.50 63.53**
r
2
0.429 0.706 0.151 0.490
a
The nvalues were slightly lower in the MSL (15þkHz) band for bulkers (n¼59) and containerships (n¼19) due to the lower signal-to-noise ratio in this
frequency range.
TABLE V. Equivalent decibel-per-knot SL reductions for each vessel cate-
gory. Values were calculated by dividing the difference in mean SLs by the
difference in mean speeds between the participant and control groups in
Table III.
Category
dB/knot
(broadband
RNL)
dB/knot
(broadband
MSL)
dB/knot
(MSL
0.5–15 kHz)
dB/knot
(MSL
15þkHz)
Bulker 2.66 2.83 1.46 2.46
Containership 1.46 1.50 1.21 2.32
Cruise 1.75 1.71 1.75 —
Tanker 2.52 2.65 1.57 3.43
Vehicle carrier 1.57 1.57 1.26 2.36
350 J. Acoust. Soc. Am. 146 (1), July 2019 MacGillivray et al.
mooring deployments and retrievals of the hydrophone
stations in Haro Strait. Mitch Herron and the crew of R/V
Richardson Point (Seaward Engineering) provided vessel
support during the fieldwork. Edward Chapin and his
colleagues at the Herzberg Institute of Astrophysics (NRC)
permitted JASCO’s AIS receiver station to be installed at
Observatory Hill. Xavier Mouy (JASCO) decoded the AIS
data and provided additional software support for PortListen.
Jason Wood (SMRU Consulting North America) provided
additional AIS data from Lime Kiln to fill in gaps in
coverage from the Observatory Hill receiver. Karen Hiltz
(JASCO) performed editorial review of this manuscript. The
ECHO Program’s Advisory Working Group worked with the
ECHO team to establish the concept and parameters of the
slowdown trial with further refinements by the ECHO
Program’s Vessel Operators Committee. The ECHO team
worked together with the PPA and BC Coast Pilots to
establish a system that allowed pilot data to be collected
during the trial. The ECHO Strait of Georgia ULS was
installed on the VENUS Observatory, which was operated
by Ocean Networks Canada and supported by Transport
Canada. This work was funded by the VFPA with support
from the Government of Canada.
1
Using the pairwise.t.test function in the Rstatistical analysis pack-
age (version 3.4.2).
ANSI/ASA (2009). S12.64/Part 1, American National Standard Quantities
and Procedures for Description and Measurement of Underwater Sound
from Ships Part 1: General Requirements (American National Standards
Institute and Acoustical Society of America, New York).
Arveson, P. T., and Vendittis, D. J. (2000). “Radiated noise characteristics
of a modern cargo ship,” J. Acoust. Soc. Am. 107, 118–129.
Bassett, C., Polagye, B., Holt, M., and Thomson, J. (2012). “A vessel noise
budget for Admiralty Inlet, Puget Sound, Washington (USA),” J. Acoust.
Soc. Am. 132, 3706–3719.
Brekhovskikh, L. M. (1980). Waves in Layered Media (Academic, New York).
Brekhovskikh, L. M., and Lysanov, Y. P. (2003). Fundamentals of Ocean
Acoustics (Springer-Verlag, New York).
Center for Whale Research (2018). “Orca population as of December 31,
2017” (Center for Whale Research, Friday Harbor, WA).
Collins, M. D. (1993). “A split-step Pad
e solution for the parabolic equation
method,” J. Acoust. Soc. Am. 93, 1736–1742.
Fisheries and Oceans Canada (DFO) (2011). “Recovery strategy for the
northern and southern resident killer whales (Orcinus orca) in Canada,” in
Species at Risk Act Recovery Strategy Series (Fisheries and Oceans
Canada, Ottawa, ON, Canada), p. 80.
Fisheries and Oceans Canada (DFO) (2016). “Report on the progress of
recovery strategy implementation for the northern and southern resident
killer whales (Orcinus orca) in Canada for the period 2009–2014,” in
Species at Risk Act Recovery Strategy Report Series (Fisheries and Oceans
Canada, Ottawa, ON, Canada), ii þ51 pp.
Fisheries and Oceans Canada (DFO) (2017). “Action plan for the northern
and southern resident killer whale (Orcinus orca) in Canada,” in Species
at Risk Act Action Plan Series (Fisheries and Oceans Canada, Ottawa, ON,
Canada), v þ33 pp.
Franc¸ois, R. E., and Garrison, G. R. (1982). “Sound absorption based on
ocean measurements: Part II: Boric acid contribution and equation for total
absorption,” J. Acoust. Soc. Am. 72, 1879–1890.
Frankel, A. S., and Gabriele, C. M. (2017). “Predicting the acoustic expo-
sure of humpback whales from cruise and tour vessel noise in Glacier
Bay, Alaska, under different management strategies,” Endanger. Spec.
Res. 34, 397–415.
Gray, L. M., and Greeley, D. S. (1980). “Source level model for propeller
blade rate radiation for the world’s merchant fleet,” J. Acoust. Soc. Am.
67, 516–522.
Hannah, C. G., Dupont, F., Collins, A. K., Dunphy, M., and Greenberg, D.
(2008). “Revisions to a modelling system for tides in the Canadian Arctic
Archipelago,” Can. Tech. Rep. Hydrogr. Ocean Sci. 259, 62.
Hannay, D., Racca, R., Carr, S., Jenkerson, M., Melton, R., Hedgeland, D.,
and Tsoflias, S. L. (2010). “A model based assessment of underwater
sound levels during a seismic source soft-start operation,” in SPE
International Conference on Health, Safety and Environment in Oil and
Gas Exploration and Production (Society of Petroleum Engineers, Rio de
Janeiro, Brazil), p. 7.
Hannay, D. E., Mouy, X., and Li, Z. (2016). “An automated real-time vessel
sound measurement system for calculating monopole source levels using a
modified version of ANSI/ASA S12.64-2009,” Can. Acoust. 44, 166–167.
Heise, K., Barrett-Lennard, L., Chapman, R., Dakin, T., Erbe, C., Hannay, D.
E., Merchant, N., Pilkington, J., Thornton, S., Tollit, D. J., Vagle, S., Veirs,
V., Vergara, V., Wood, J., Wright, B., and Yurk, H. (2017). “Proposed met-
rics for the management of underwater noise for southern resident killer
whales,” in Coastal Ocean Report Series, Report for the Coastal Ocean
Research Institute (Ocean Wise, Vancouver, Canada), p. 31.
ISO (2017). 18405.2:2017, Underwater Acoustics—Terminology (International
Organization for Standardization, Geneva, Switzerland), p. 62.
Jensen, F. B., Kuperman, W. A., Porter, M. B., and Schmidt, H. (2011).
Computational Ocean Acoustics (AIP–Springer, New York).
Joy, R., Tollit, D., Wood, J., MacGillivray, A., Li, Z., Trounce, K., and
Robinson, O. (2019). “Potential benefits of vessel slowdowns on endan-
gered southern resident killer whales,” Front. Mar. Sci. 6, 344.
Kipple, B., and Gabriele, C. (2007). “Underwater noise from skiffs to ships,”
in Fourth Glacier Bay Science Symposium, pp. 172–175.
McKenna, M. F., Ross, D., Wiggins, S. M., and Hildebrand, J. A. (2012).
“Underwater radiated noise from modern commercial ships,” J. Acoust.
Soc. Am. 131, 92–103.
McKenna, M. F., Wiggins, S. M., and Hildebrand, J. A. (2013).
“Relationship between container ship underwater noise levels and ship
design, operational and oceanographic conditions,” Sci. Rep. 3, 1760.
National Research Council (U.S.) (NRC) (2003). Ocean Noise and Marine
Mammals (National Research Council, Ocean Studies Board, Committee
on Potential Impacts of Ambient Noise in the Ocean on Marine Mammals,
The National Academies Press, Washington, DC).
Richardson, W. J., Greene, C. R., Jr., Malme, C. I., and Thomson, D. H.
(1995). Marine Mammals and Noise (Academic, San Diego, CA).
Ross, D. (1976). Mechanics of Underwater Noise (Pergamon, New York).
Simard, Y., Roy, N., Gervaise, C., and Giard, S. (2016). “Analysis and
modeling of 255 source levels of merchant ships from an acoustic observa-
tory along St. Lawrence Seaway,” J. Acoust. Soc. Am. 140, 2002–2018.
Traverso, F., Gaggero, T., Rizzuto, E., and Trucco, A. (2015). “Spectral
analysis of the underwater acoustic noise radiated by ships with controlla-
ble pitch propellers,” in OCEANS 2015, Genova, Italy, pp. 1–6.
Veirs, S., Veirs, V., and Wood, J. D. (2016). “Ship noise extends to frequen-
cies used for echolocation by endangered killer whales,” PeerJ 4, 1657.
Wales, S. C., and Heitmeyer, R. M. (2002). “An ensemble source spectra
model for merchant ship-radiated noise,” J. Acoust. Soc. Am. 111,
1211–1231.
Warner, G. A., O’Neill, C., McCrodan, A., Frouin-Mouy, H., Izett, J., and
MacGillivray, A. O. (2014). “Underwater acoustic measurements in Haro
Strait and Strait of Georgia. Appendix A,” in Hemmera, SMRU Canada,
and JASCO Applied Sciences. 2014 (b) Roberts Bank Terminal 2
Technical Data Report-Underwater Noise. Ship Sound Signature Analysis
Study (Technical Report for Port Metro, Vancouver).
Wenz, G. M. (1962). “Acoustic ambient noise in the ocean: Spectra and
sources,” J. Acoust. Soc. Am. 34, 1936–1956.
Zhang, Z. Y., and Tindle, C. T. (1995). “Improved equivalent fluid approxi-
mations for a low shear speed ocean bottom,” J. Acoust. Soc. Am. 98,
3391–3396.
J. Acoust. Soc. Am. 146 (1), July 2019 MacGillivray et al. 351
