Academia.eduAcademia.edu

Figure 25 – uploaded by Nathan Merchant

See full PDFdownloadDownload figure
Figure 4.6: Example of ship identification using AIS data*. Top: Range from hydrophone vs. time. Crosses denote individual AIS transmissions; lines connect transmissions from the same vessel; circles indicate closest points of approach, labelled with the MMSI number. Shaded area denotes 15-minute time window around SPL peak at 01:50; heavy line indicates track of vessel identified as source of peak. Note that the horizontal lines indicate AIS transmissions from stationary vessels. Middle: Power spectral density of concurrent acoustic data. Bottom: Broadband (0.01-1 kHz) SPL, showing ‘background’, ‘intermittent’, and ‘intermittent’ peaks. ship sources, and 68 (36%) unidentified.

Figure 4 6: Example of ship identification using AIS data*. Top: Range from hydrophone vs. time. Crosses denote individual AIS transmissions; lines connect transmissions from the same vessel; circles indicate closest points of approach, labelled with the MMSI number. Shaded area denotes 15-minute time window around SPL peak at 01:50; heavy line indicates track of vessel identified as source of peak. Note that the horizontal lines indicate AIS transmissions from stationary vessels. Middle: Power spectral density of concurrent acoustic data. Bottom: Broadband (0.01-1 kHz) SPL, showing ‘background’, ‘intermittent’, and ‘intermittent’ peaks. ship sources, and 68 (36%) unidentified.

Related Figures (46)

changeably given an appropriate conversion formula for the relevant latitude (e.g. Fotonott The sound speed in the ocean varies between around 1,450 and 1,550 ms_! (Lurton, 2010)
changeably given an appropriate conversion formula for the relevant latitude (e.g. Fotonott The sound speed in the ocean varies between around 1,450 and 1,550 ms_! (Lurton, 2010)
Figure 2.3: Spectral density levels of physical sounds from various studies (Medwin and Clay, 1998). The heavy solid line at the bottom is the minimum level at sea. A: seismic noise; B: ship noise; C: wave turbulence interactions; H: hail; W: sea surface levels at five different wind speeds; R1: drizzle (~1 mm/h) with 0.6 m/s wind over lake; R2: drizzle with 2.6 m/s wind over lake; R3: heavy rain (15 mm/h) at sea; R4: very heavy rain (100 mm/h) at sea; F: thermal noise.
Figure 2.3: Spectral density levels of physical sounds from various studies (Medwin and Clay, 1998). The heavy solid line at the bottom is the minimum level at sea. A: seismic noise; B: ship noise; C: wave turbulence interactions; H: hail; W: sea surface levels at five different wind speeds; R1: drizzle (~1 mm/h) with 0.6 m/s wind over lake; R2: drizzle with 2.6 m/s wind over lake; R3: heavy rain (15 mm/h) at sea; R4: very heavy rain (100 mm/h) at sea; F: thermal noise.
to have greater levels than those produced by stridulation, though neither type is intense. Figure 2.4: Snapping shrimp Alpheus euphrosyne*, approx. total length: 5 cm. Invertebrates
to have greater levels than those produced by stridulation, though neither type is intense. Figure 2.4: Snapping shrimp Alpheus euphrosyne*, approx. total length: 5 cm. Invertebrates
mong the facilities, with implications for the transferability of acoustic intelligence. facilities in each country (C. de Jong, pers. comm., 2013). Significant differences were found
mong the facilities, with implications for the transferability of acoustic intelligence. facilities in each country (C. de Jong, pers. comm., 2013). Significant differences were found
Figure 2.6: Radiated noise from R/V Oscar Dyson (Bahtiarian and Fischer, 2006).
Figure 2.6: Radiated noise from R/V Oscar Dyson (Bahtiarian and Fischer, 2006).
ergy devices are still at the development stage. Noise assessment of MREDs is likewise ir its infancy, especially for wave and tidal converters. Few in situ measurements of full-scak wave or tidal energy extractors have yet been published in the open literature [e.g. Haikoner et al., 2013; see also Bassett et al., 2011; Garrett et al., (In Press) ], and most work to date ha focused on modelling likely noise-generating mechanisms (e.g. Richards et al., 2007; Patricic et al., 2009; Lloyd et al., 2011). Offshore wind turbines do generate some noise during thei operational phase: underwater acoustic measurements of wind farms (Tougaard et al., 2009a found that turbine noise only exceeded background levels below 500 Hz, with a maximun third-octave level in the 25-Hz band of 126 dB re 1 Pa? at 14 m distance. Figure 2.7: Pelamis wave energy converter, length 150 m.* Perhaps the first events to raise public awareness of the acoustic impacts of underwater human activities were mass strandings of marine mammals associated with military sonar exercises (c.g. Frantzis, 1998; Evans and England, 2001). The specifications of military sonar devices are generally classified, so source characteristics are hard to ascertain. Active sonar systems are also used for civilian purposes, notably in fisheries (MacLennan and Simmonds, 1992) and seabed mapping (Blondel and Murton, 1997; Kenny et al., 2003), and also as simple echo- sounders and acoustic deterrent devices. Perhaps the first events to raise public awareness of the acoustic impacts of underwater human
ergy devices are still at the development stage. Noise assessment of MREDs is likewise ir its infancy, especially for wave and tidal converters. Few in situ measurements of full-scak wave or tidal energy extractors have yet been published in the open literature [e.g. Haikoner et al., 2013; see also Bassett et al., 2011; Garrett et al., (In Press) ], and most work to date ha focused on modelling likely noise-generating mechanisms (e.g. Richards et al., 2007; Patricic et al., 2009; Lloyd et al., 2011). Offshore wind turbines do generate some noise during thei operational phase: underwater acoustic measurements of wind farms (Tougaard et al., 2009a found that turbine noise only exceeded background levels below 500 Hz, with a maximun third-octave level in the 25-Hz band of 126 dB re 1 Pa? at 14 m distance. Figure 2.7: Pelamis wave energy converter, length 150 m.* Perhaps the first events to raise public awareness of the acoustic impacts of underwater human activities were mass strandings of marine mammals associated with military sonar exercises (c.g. Frantzis, 1998; Evans and England, 2001). The specifications of military sonar devices are generally classified, so source characteristics are hard to ascertain. Active sonar systems are also used for civilian purposes, notably in fisheries (MacLennan and Simmonds, 1992) and seabed mapping (Blondel and Murton, 1997; Kenny et al., 2003), and also as simple echo- sounders and acoustic deterrent devices. Perhaps the first events to raise public awareness of the acoustic impacts of underwater human
A relatively recent area of study has been the physiological effects that exposure to vessel noise may have on marine species. Evidence of physiological stress can be found in levels of the stress hormone hydrocortisone (also known as cortisol) in faecal samples of animals exposed to noise. Studies measuring cortisol levels with and without the presence of vessel noise have found significantly higher stress levels in fish (Wysocki et al., 2006) and whales (Rolland et al., 2012) exposed to ship noise. Other physiological responses include increased heart rate (in fish; Graham and Cooke, 2008) and respiration (in crabs; Wale et al., 2013a). These physiological responses have evolved to enhance animals’ ability to react to short-term threats in their natural environment, but may be maladaptive in the context of repeated or chronic elicitation by anthropogenic noise (Wright et al., 2007a). In the long term, height- ened metabolic rate and physiological stress induced by vessel noise may be detrimental to individual fitness and survival, with potential ramifications for population levels. Figure 2.8: Overview of PCAD concept, which relates behavioural responses to noise tc possible effects at the population level (National Research Council, 2005).
A relatively recent area of study has been the physiological effects that exposure to vessel noise may have on marine species. Evidence of physiological stress can be found in levels of the stress hormone hydrocortisone (also known as cortisol) in faecal samples of animals exposed to noise. Studies measuring cortisol levels with and without the presence of vessel noise have found significantly higher stress levels in fish (Wysocki et al., 2006) and whales (Rolland et al., 2012) exposed to ship noise. Other physiological responses include increased heart rate (in fish; Graham and Cooke, 2008) and respiration (in crabs; Wale et al., 2013a). These physiological responses have evolved to enhance animals’ ability to react to short-term threats in their natural environment, but may be maladaptive in the context of repeated or chronic elicitation by anthropogenic noise (Wright et al., 2007a). In the long term, height- ened metabolic rate and physiological stress induced by vessel noise may be detrimental to individual fitness and survival, with potential ramifications for population levels. Figure 2.8: Overview of PCAD concept, which relates behavioural responses to noise tc possible effects at the population level (National Research Council, 2005).
reference frame of fixed platforms, tidal flow can induce the same effects. One solution in high tidal flow environments such as proposed tidal energy sites has been to deploy freely drifting platforms (e.g. Wilson et al., 2011). These drifters consist of a hydrophone attached to a float containing a GPS tag. When the drifter is subsequently recovered, sound levels can be mapped through the deployment path. The first mobile PAM platforms were towed arrays - multiple hydrophones towed behind ; moving vessel - developed as early as the First World War (Gershman et al., 2000). They wer employed initially for anti-submarine warfare (ASW) applications, though these became les effective with the development of submarine quieting technology in the late 1960s. Extensiv towed arrays are also used in seismic surveys to detect the reflections of seismic sources fron the seabed. As platforms for ambient noise monitoring, they can be limited by noise fron the towing vessel if not towed far enough from the vessel (Urick, 1986), as well as flow nois: caused by turbulence around the hydrophones at low frequencies. drifting platforms (e.g. Wilson et al., 2011). These drifters consist of a hydrophone attached
reference frame of fixed platforms, tidal flow can induce the same effects. One solution in high tidal flow environments such as proposed tidal energy sites has been to deploy freely drifting platforms (e.g. Wilson et al., 2011). These drifters consist of a hydrophone attached to a float containing a GPS tag. When the drifter is subsequently recovered, sound levels can be mapped through the deployment path. The first mobile PAM platforms were towed arrays - multiple hydrophones towed behind ; moving vessel - developed as early as the First World War (Gershman et al., 2000). They wer employed initially for anti-submarine warfare (ASW) applications, though these became les effective with the development of submarine quieting technology in the late 1960s. Extensiv towed arrays are also used in seismic surveys to detect the reflections of seismic sources fron the seabed. As platforms for ambient noise monitoring, they can be limited by noise fron the towing vessel if not towed far enough from the vessel (Urick, 1986), as well as flow nois: caused by turbulence around the hydrophones at low frequencies. drifting platforms (e.g. Wilson et al., 2011). These drifters consist of a hydrophone attached
If the frequency response of either the hydrophone or the pre-amplifier varies significantly within the frequency range of interest, then a frequency-dependent correction should be ap- plied. The operations in Eq. 2.8 and Eq. 2.9 are effected using a frequency-dependent pream- plifier gain and a hydrophone sensitivity curve, respectively. These corrections are carried out in the frequency domain, i.e. after a discrete Fourier transform (DFT) has been applied to the signal. In practice, these frequency-dependent steps are integrated with the subsequent stage: calculation of the power spectral density (PSD), detailed below. Figure 2.10: Signal path and calibration sequence for a typical passive acoustic monitoring system. Values in square brackets indicate signal units at each processing stage. To assess the frequency characteristics of the sound pressure signal, x ,p,, the signal is trans- uPa? posed into the frequency domain using a DFT. The time-series signal is segmented, multi- plied by a window function, and then the DFT is applied to each time segment. This yields a frequency spectrum for each segment, which can be plotted as a spectrogram (sometimes termed periodogram), showing the variation in frequency content through time. For many
If the frequency response of either the hydrophone or the pre-amplifier varies significantly within the frequency range of interest, then a frequency-dependent correction should be ap- plied. The operations in Eq. 2.8 and Eq. 2.9 are effected using a frequency-dependent pream- plifier gain and a hydrophone sensitivity curve, respectively. These corrections are carried out in the frequency domain, i.e. after a discrete Fourier transform (DFT) has been applied to the signal. In practice, these frequency-dependent steps are integrated with the subsequent stage: calculation of the power spectral density (PSD), detailed below. Figure 2.10: Signal path and calibration sequence for a typical passive acoustic monitoring system. Values in square brackets indicate signal units at each processing stage. To assess the frequency characteristics of the sound pressure signal, x ,p,, the signal is trans- uPa? posed into the frequency domain using a DFT. The time-series signal is segmented, multi- plied by a window function, and then the DFT is applied to each time segment. This yields a frequency spectrum for each segment, which can be plotted as a spectrogram (sometimes termed periodogram), showing the variation in frequency content through time. For many
Figure 2.11: Seismic airgun array recorded at ~6 km distance in Mayumba National Park, Gabon, October 30 2012. Sampling frequency: 96 kHz. (a) Power spectral density: 0.05-s Hann window, 99% overlap. (b) Pressure waveform illustrating peak-to-peak and zero-to- peak metrics. (c) Cumulative energy of pulse showing 90% energy envelope.
Figure 2.11: Seismic airgun array recorded at ~6 km distance in Mayumba National Park, Gabon, October 30 2012. Sampling frequency: 96 kHz. (a) Power spectral density: 0.05-s Hann window, 99% overlap. (b) Pressure waveform illustrating peak-to-peak and zero-to- peak metrics. (c) Cumulative energy of pulse showing 90% energy envelope.
the University of Victoria) and the University of Aberdeen. The first (Section 3.1; Merchant Figure 3.1: Map of the Strait of Georgia, British Columbia, Canada, showing the location of the VENUS hydrophone (black cross) used in this chapter. The colour scale indicates the density of shipping in the Strait based on AIS data for May 2012 (1 x 1-km grid).
the University of Victoria) and the University of Aberdeen. The first (Section 3.1; Merchant Figure 3.1: Map of the Strait of Georgia, British Columbia, Canada, showing the location of the VENUS hydrophone (black cross) used in this chapter. The colour scale indicates the density of shipping in the Strait based on AIS data for May 2012 (1 x 1-km grid).
To examine the effect of varying averaging time, the temporal resolution of the 1/3 octave spectra was reduced from 1 s using the standard Welch (1967) method (i.e. using the mean of each frequency band in linear space) for averaging times of up to ~10 s. To limit the influence of transients in the spectrum, Parks et al. (2009) proposed an alternative to the Welch method using the median instead of the mean, and subsequently to use the mode of these median values as the average level. This approach was also implemented for comparison to the standard mode. In the discussion below, ‘integration time’ refers to the time over which Pons Was calculated (1 s), and ‘averaging time’ is the length of the averaged power spectrum windows (from 1 to ~10’ s).
To examine the effect of varying averaging time, the temporal resolution of the 1/3 octave spectra was reduced from 1 s using the standard Welch (1967) method (i.e. using the mean of each frequency band in linear space) for averaging times of up to ~10 s. To limit the influence of transients in the spectrum, Parks et al. (2009) proposed an alternative to the Welch method using the median instead of the mean, and subsequently to use the mode of these median values as the average level. This approach was also implemented for comparison to the standard mode. In the discussion below, ‘integration time’ refers to the time over which Pons Was calculated (1 s), and ‘averaging time’ is the length of the averaged power spectrum windows (from 1 to ~10’ s).
Figure 3.3: (a) Estimated probability densities (PDs) for each month in the 125-Hz band. Bin width: 0.1 dB, integration time: 1s. (b) PDs of octave-separated 1/3 octave bands over observed frequency range of shipping noise. (c) 24-h SEL‘calculated from PDs in (b).
Figure 3.3: (a) Estimated probability densities (PDs) for each month in the 125-Hz band. Bin width: 0.1 dB, integration time: 1s. (b) PDs of octave-separated 1/3 octave bands over observed frequency range of shipping noise. (c) 24-h SEL‘calculated from PDs in (b).
Figure 3.4: (a) Average SPLs in 125 Hz band for each month; averages with February ship signature [see Fig. 3.2(b)] omitted are plotted in gray. Integration time: 1 s. (b) Total average SPL for 125 Hz band vs. averaging time. ‘Parks mode’ refers to the averaging method in Parks et al. (2009)
Figure 3.4: (a) Average SPLs in 125 Hz band for each month; averages with February ship signature [see Fig. 3.2(b)] omitted are plotted in gray. Integration time: 1 s. (b) Total average SPL for 125 Hz band vs. averaging time. ‘Parks mode’ refers to the averaging method in Parks et al. (2009)
high noise levels as an oil rig was towed into the area by two vessels operating with dynamic positioning (see Fig. 3.6(a) from June 16 onwards). The received vessel noise was concentrated below 1 kHz, and exhibited a tide-dependent Lloyd’s mirror effect. In the SPD [Fig. 3.5(a)], this sustained period of vessel noise appears as a secondary modal ridge't ~40 dB greater than the primary mode in the range 0.1-1 kHz. In contrast, this underlying bimodality is concealed by the linear mean, SPL, and percentiles, which could be misleading if used as the sole method of analysis.
high noise levels as an oil rig was towed into the area by two vessels operating with dynamic positioning (see Fig. 3.6(a) from June 16 onwards). The received vessel noise was concentrated below 1 kHz, and exhibited a tide-dependent Lloyd’s mirror effect. In the SPD [Fig. 3.5(a)], this sustained period of vessel noise appears as a secondary modal ridge't ~40 dB greater than the primary mode in the range 0.1-1 kHz. In contrast, this underlying bimodality is concealed by the linear mean, SPL, and percentiles, which could be misleading if used as the sole method of analysis.
Figure 3.6: Power spectral densities of Moray Firth deployments over the range 25 Hz-100 kHz: (a) 13 June to 7 July 2012 (b) 7-27 September 2012. (c) VENUS deployment over range 10-1600 Hz from 14 December 2011 to 1 August 2012.
Figure 3.6: Power spectral densities of Moray Firth deployments over the range 25 Hz-100 kHz: (a) 13 June to 7 July 2012 (b) 7-27 September 2012. (c) VENUS deployment over range 10-1600 Hz from 14 December 2011 to 1 August 2012.
Figure 4.1: Deployment location: Falmouth Bay, UK. 4.1.2 Materials and methods
Figure 4.1: Deployment location: Falmouth Bay, UK. 4.1.2 Materials and methods
Figure 4.2: M-weightings for low-, medium- and high-frequency cetaceans and pinnipeds (in water) (Southall et al., 2007). The shaded areas indicate the frequency bandwidth of the recordings (‘Full Bandwidth’; 5 Hz - 8 kHz) and the nominal frequency bandwidth of ship- ping noise (‘Shipping’; 10 Hz - 1 kHz) used in this study for the calculation of SPL and SEL.
Figure 4.2: M-weightings for low-, medium- and high-frequency cetaceans and pinnipeds (in water) (Southall et al., 2007). The shaded areas indicate the frequency bandwidth of the recordings (‘Full Bandwidth’; 5 Hz - 8 kHz) and the nominal frequency bandwidth of ship- ping noise (‘Shipping’; 10 Hz - 1 kHz) used in this study for the calculation of SPL and SEL.
Mean hourly wind speeds at the Culdrose weather station ranged from 2 to 17 knots (1.0-8.. ms~'), with a maximum hourly increase of 6 knots (3.1 m s~!). Wind speeds in this rang have been associated with variations of up to around 20 dB in shallow water ambient nois levels (Urick, 1983). Spectra characteristic of wind noise did not feature in the frequenc spectrum of the intermittent component, which was reviewed visually. This implies tha either the wind-generated noise was below the adaptive threshold, meaning that the rate o increase in broadband (0.01-1 kHz) SPL due to wind did not exceed 6 dB per 1.5 hours ( dB per hour), or that any rapid increases in wind speed were masked by local vessel activity Rainfall was recorded at Culdrose in 12 hours of data over the 9 day period, with a maximun rate of 0.8 mm per hour. Since rain generates noise at frequencies above 1 kHz (Nystuen 2001), it was not considered to contribute to the broadband (0.01-1 kHz) levels used for nois classification.
Mean hourly wind speeds at the Culdrose weather station ranged from 2 to 17 knots (1.0-8.. ms~'), with a maximum hourly increase of 6 knots (3.1 m s~!). Wind speeds in this rang have been associated with variations of up to around 20 dB in shallow water ambient nois levels (Urick, 1983). Spectra characteristic of wind noise did not feature in the frequenc spectrum of the intermittent component, which was reviewed visually. This implies tha either the wind-generated noise was below the adaptive threshold, meaning that the rate o increase in broadband (0.01-1 kHz) SPL due to wind did not exceed 6 dB per 1.5 hours ( dB per hour), or that any rapid increases in wind speed were masked by local vessel activity Rainfall was recorded at Culdrose in 12 hours of data over the 9 day period, with a maximun rate of 0.8 mm per hour. Since rain generates noise at frequencies above 1 kHz (Nystuen 2001), it was not considered to contribute to the broadband (0.01-1 kHz) levels used for nois classification.
Figure 4.4: Broadband (0.01-1 kHz) SPL for a representative 24-h period (28 July) showing classification of ‘intermittent’ and ‘background’ data. Integration time: 300 s. The solid line is the adaptive threshold level. 2001), it was not considered to contribute to the broadband (0.01-1 kHz) levels used for nois«
Figure 4.4: Broadband (0.01-1 kHz) SPL for a representative 24-h period (28 July) showing classification of ‘intermittent’ and ‘background’ data. Integration time: 300 s. The solid line is the adaptive threshold level. 2001), it was not considered to contribute to the broadband (0.01-1 kHz) levels used for nois«
of 6.4 dB. In total, there were 314 peaks in the intermittent SPL data (mean: 34.9 per day).
of 6.4 dB. In total, there were 314 peaks in the intermittent SPL data (mean: 34.9 per day).
Figure 4.5: (a) Median 24-h SEL in third-octave bands, calculated from 9 consecutive 24-h periods. (b)-(d): total SEL and SEL due to intermittent and background components. The centre lines of the boxes denote the median and the box limits indicate the first quartile. The whiskers are the maximum and minimum values recorded. The shaded areas indicate the nominal bandwidth of shipping noise (0.01-1 kHz) background component was comparatively stable at all frequencies | Fig. 4.5(d) | The median 24-h SEL was concentrated above ~100 Hz, with a broad peak at 315 Hz [Fig. 4.5(a)]. The intermittent component was most dominant between around 30 and 2,000 Hz [Fig. 4.5(a)]. The variability of the intermittent data [ Fig. 4.5(c)] appears to account for the variability of the total 24-h SEL [Fig. 4.5(b)] above ~30 Hz. In contrast, the 24-h SEL of the background component was comparatively stable at all frequencies [Fig. 4.5(d)].
Figure 4.5: (a) Median 24-h SEL in third-octave bands, calculated from 9 consecutive 24-h periods. (b)-(d): total SEL and SEL due to intermittent and background components. The centre lines of the boxes denote the median and the box limits indicate the first quartile. The whiskers are the maximum and minimum values recorded. The shaded areas indicate the nominal bandwidth of shipping noise (0.01-1 kHz) background component was comparatively stable at all frequencies | Fig. 4.5(d) | The median 24-h SEL was concentrated above ~100 Hz, with a broad peak at 315 Hz [Fig. 4.5(a)]. The intermittent component was most dominant between around 30 and 2,000 Hz [Fig. 4.5(a)]. The variability of the intermittent data [ Fig. 4.5(c)] appears to account for the variability of the total 24-h SEL [Fig. 4.5(b)] above ~30 Hz. In contrast, the 24-h SEL of the background component was comparatively stable at all frequencies [Fig. 4.5(d)].
Figure 4.7: Positions of uniquely identified CPAs categorised by vessel type. The size of each circle corresponds to the magnitude of the associated SPL peak, ranging from 92.8 to 148.6 dB re 1 yzPa?. Cross denotes the location of the deployment. Lines indicate paths of AIS transmissions during the deployment period. speed and direct trajectories between AIS transmissions. The coordinates of uniquely identified CPAs were distributed within Falmouth Bay and fur- ther south into the English Channel (Fig. 4.7). The largest cluster of CPAs to the east of the deployment corresponds to the paths of vessels entering and leaving Falmouth Harbour and the Bay. A second cluster ~15 km south of the deployment site corresponds to paths of vessels navigating along the coast past the headland at Lizard Point. Small vessels were distributed within the Bay close to the deployment site, while the main shipping routes were populated by tankers and cargo ships. The tanker furthest west in the English Channel appears to have been falsely identified as the coast obscures the line of sight to the hydrophone. Error in the position of the CPA could also be the cause, since these were calculated assuming constant speed and direct trajectories between AIS transmissions.
Figure 4.7: Positions of uniquely identified CPAs categorised by vessel type. The size of each circle corresponds to the magnitude of the associated SPL peak, ranging from 92.8 to 148.6 dB re 1 yzPa?. Cross denotes the location of the deployment. Lines indicate paths of AIS transmissions during the deployment period. speed and direct trajectories between AIS transmissions. The coordinates of uniquely identified CPAs were distributed within Falmouth Bay and fur- ther south into the English Channel (Fig. 4.7). The largest cluster of CPAs to the east of the deployment corresponds to the paths of vessels entering and leaving Falmouth Harbour and the Bay. A second cluster ~15 km south of the deployment site corresponds to paths of vessels navigating along the coast past the headland at Lizard Point. Small vessels were distributed within the Bay close to the deployment site, while the main shipping routes were populated by tankers and cargo ships. The tanker furthest west in the English Channel appears to have been falsely identified as the coast obscures the line of sight to the hydrophone. Error in the position of the CPA could also be the cause, since these were calculated assuming constant speed and direct trajectories between AIS transmissions.
Figure 4.8: Map of study area. PAM units were deployed at The Sutors and Chanonry. Me- teorological data for Chanonry were acquired from a weather station at Ardersier; time-lapse footage for The Sutors was recorded from Cromarty (see text).
Figure 4.8: Map of study area. PAM units were deployed at The Sutors and Chanonry. Me- teorological data for Chanonry were acquired from a weather station at Ardersier; time-lapse footage for The Sutors was recorded from Cromarty (see text).
detailed analysis of illustrative events. The PAM units were independently calibrated using a pistonphone in the frequency range 25- 315 Hz. This calibration agreed with the manufacturer’s declared sensitivity to within +1 dB, and so the manufacturer’s data were used for the entire frequency range (25 Hz - 192 kHz). Acoustic data were processed in MATLAB using custom-written scripts. The power spectral density was computed using a 1-s Hann window, and the spectra were then averaged to 60-s resolution using the standard Welch method (Welch, 1967), producing a single spectrum for each 1-minute recording. These were then concatenated to form a master file for subsequent analysis. Spectral analysis revealed low-amplitude tonal noise from the recording system at various frequencies above 1 kHz (Merchant et al., 2013). This system noise contaminated a small proportion of the frequency spectrum (<0.1%) and was omitted from the analysis. The analysis also showed that the noise floor of the PAM units was ~47 dB re 1 uPa’, exceeding background noise levels above ~1.5 kHz. Although anthropogenic, biotic and abiotic sounds could still be detected and measured at these high frequencies, background noise levels above ~1.5 kHz could not be determined.
detailed analysis of illustrative events. The PAM units were independently calibrated using a pistonphone in the frequency range 25- 315 Hz. This calibration agreed with the manufacturer’s declared sensitivity to within +1 dB, and so the manufacturer’s data were used for the entire frequency range (25 Hz - 192 kHz). Acoustic data were processed in MATLAB using custom-written scripts. The power spectral density was computed using a 1-s Hann window, and the spectra were then averaged to 60-s resolution using the standard Welch method (Welch, 1967), producing a single spectrum for each 1-minute recording. These were then concatenated to form a master file for subsequent analysis. Spectral analysis revealed low-amplitude tonal noise from the recording system at various frequencies above 1 kHz (Merchant et al., 2013). This system noise contaminated a small proportion of the frequency spectrum (<0.1%) and was omitted from the analysis. The analysis also showed that the noise floor of the PAM units was ~47 dB re 1 uPa’, exceeding background noise levels above ~1.5 kHz. Although anthropogenic, biotic and abiotic sounds could still be detected and measured at these high frequencies, background noise levels above ~1.5 kHz could not be determined.
An autonomous underwater acoustic logger (C-POD, Chelonia Ltd., www.chelonia.co.uk was independently deployed at each of the two sites as part of the bottlenose dolphin SAC monitoring programme (Cheney et al., 2013). C-PODs use digital waveform characterizatio: to detect cetacean echolocation clicks. The time of detection is logged together with othe click features, which are then used by the click-train classifier (within the dedicated analysi software) to identify bottlenose dolphin clicks. Here, the data from the C-PODs were use only to confirm dolphin occurrence at the two sites throughout the deployment periods More detailed analysis is ongoing and will be reported elsewhere.
An autonomous underwater acoustic logger (C-POD, Chelonia Ltd., www.chelonia.co.uk was independently deployed at each of the two sites as part of the bottlenose dolphin SAC monitoring programme (Cheney et al., 2013). C-PODs use digital waveform characterizatio: to detect cetacean echolocation clicks. The time of detection is logged together with othe click features, which are then used by the click-train classifier (within the dedicated analysi software) to identify bottlenose dolphin clicks. Here, the data from the C-PODs were use only to confirm dolphin occurrence at the two sites throughout the deployment periods More detailed analysis is ongoing and will be reported elsewhere.
Figure 4.11: Effect of weather and tides on ambient noise in Chanonry. (a) 1/3 octave band spectrum from 26-31 August, 60-s resolution; (b) Rainfall and mean wind speed recorded at Ardersier; (c) Tide level and speed predicted by POLPRED model (d) Spearman ranked correlation coefficient of each process across frequency range for entire dataset.
Figure 4.11: Effect of weather and tides on ambient noise in Chanonry. (a) 1/3 octave band spectrum from 26-31 August, 60-s resolution; (b) Rainfall and mean wind speed recorded at Ardersier; (c) Tide level and speed predicted by POLPRED model (d) Spearman ranked correlation coefficient of each process across frequency range for entire dataset.
absolute values. the day, corresponding to increased vessel noise. Mean levels [Fig. 4.12(b)] are largely de- termined by the loudest events (Merchant et al., 2012a), in this case particularly loud vessel passages, which were both louder [Fig. 4.12(b)] and more variable [Fig. 4.12(c)] at The Su- tors. The week-long presence of rig-towing vessels evident in Fig. 4.10(a) was omitted from The Sutors data as this high-amplitude event entirely dominated the mean levels for The Su- tors in Fig. 4.12(b). Note that the median levels [Fig. 4.12(a)] are likely to be raised by the noise floor of the PAM device above ~10 kHz (Merchant et al., 2013), and do not represent absolute values. Figure 4.12: Hourly variability in noise levels at both sites in 1/3 octave bands. Left column: Chanonry; Right column: The Sutors. a) Median b) RMS Mean c) Broadband (0.1-1 kHz) level.
absolute values. the day, corresponding to increased vessel noise. Mean levels [Fig. 4.12(b)] are largely de- termined by the loudest events (Merchant et al., 2012a), in this case particularly loud vessel passages, which were both louder [Fig. 4.12(b)] and more variable [Fig. 4.12(c)] at The Su- tors. The week-long presence of rig-towing vessels evident in Fig. 4.10(a) was omitted from The Sutors data as this high-amplitude event entirely dominated the mean levels for The Su- tors in Fig. 4.12(b). Note that the median levels [Fig. 4.12(a)] are likely to be raised by the noise floor of the PAM device above ~10 kHz (Merchant et al., 2013), and do not represent absolute values. Figure 4.12: Hourly variability in noise levels at both sites in 1/3 octave bands. Left column: Chanonry; Right column: The Sutors. a) Median b) RMS Mean c) Broadband (0.1-1 kHz) level.
Figure 4.13: Sample spectra recorded at Chanonry. (a) Vocalisations and echolocation clicks of bottlenose dolphins on 12 Aug at 17:50. Spectra have the same frequency range but (a) has a finer amplitude range; (b) Oil tanker with closest point of approach (CPA) at 04:30 on 18 Aug. ottlenose dolphin vocalisations were also recorded on the PAM units [Fig. 4.13(a)]. There as considerable overlap between the frequency and amplitude ranges of vocalisations anc ip noise observed, indicating the potential for communication masking. Sample spectr: om Chanonry of a passing oil tanker [Fig. 4.13(b)] and bottlenose dolphin sounds [Fig 13(a)] clearly illustrate that observed vocalisations in the range ~0.4-10 kHz coincide ir 1e frequency domain with ship noise levels of higher amplitude during the vessel passage Ithough underwater noise radiated by the vessel in Fig. 4.13(b) extends as high as the 5( Hz echosounder, masking at high frequencies is likely to be localised due to the increasing sorption of sound by water as frequency increases. This is apparent in the form of the ‘oustic signature: the highest frequencies are only visible at the closest point of approact SPA), while low-frequency tonals are evident more than 30 min before the vessel transit: ast the hydrophone, when AIS data indicates it was 9 km away. Note also the upsurge ir roadband (rather than tonal) noise following the CPA, as cavitation noise from the propelle: ecomes more prominent in the wake of the vessel. These effects can be observed more tuitively in the time-lapse footage (paired with acoustic and AIS data) documenting thi: assage included in the Supplementary Material. passage included in the Supplementary Material.
Figure 4.13: Sample spectra recorded at Chanonry. (a) Vocalisations and echolocation clicks of bottlenose dolphins on 12 Aug at 17:50. Spectra have the same frequency range but (a) has a finer amplitude range; (b) Oil tanker with closest point of approach (CPA) at 04:30 on 18 Aug. ottlenose dolphin vocalisations were also recorded on the PAM units [Fig. 4.13(a)]. There as considerable overlap between the frequency and amplitude ranges of vocalisations anc ip noise observed, indicating the potential for communication masking. Sample spectr: om Chanonry of a passing oil tanker [Fig. 4.13(b)] and bottlenose dolphin sounds [Fig 13(a)] clearly illustrate that observed vocalisations in the range ~0.4-10 kHz coincide ir 1e frequency domain with ship noise levels of higher amplitude during the vessel passage Ithough underwater noise radiated by the vessel in Fig. 4.13(b) extends as high as the 5( Hz echosounder, masking at high frequencies is likely to be localised due to the increasing sorption of sound by water as frequency increases. This is apparent in the form of the ‘oustic signature: the highest frequencies are only visible at the closest point of approact SPA), while low-frequency tonals are evident more than 30 min before the vessel transit: ast the hydrophone, when AIS data indicates it was 9 km away. Note also the upsurge ir roadband (rather than tonal) noise following the CPA, as cavitation noise from the propelle: ecomes more prominent in the wake of the vessel. These effects can be observed more tuitively in the time-lapse footage (paired with acoustic and AIS data) documenting thi: assage included in the Supplementary Material. passage included in the Supplementary Material.
Figure 4.14: AIS analysis example with time-lapse footage. (a) Still of time lapse footage show- ing vessel whose CPA occurred at 09:00 on July 4; (b) Map of AIS movements in 6-hour period centred on CPA. Black cross denotes location of PAM unit in The Sutors, circles indi- cate CPAs labelled with Maritime Mobile Service Identity (MMSI) number; (c) Range of AIS transmissions from PAM unit versus time; (d) 1/3 octave spectrum of concurrent acoustic data; (e) Broadband level in frequency range 0.1-1 kHz, showing peak identification using adaptive threshold.
Figure 4.14: AIS analysis example with time-lapse footage. (a) Still of time lapse footage show- ing vessel whose CPA occurred at 09:00 on July 4; (b) Map of AIS movements in 6-hour period centred on CPA. Black cross denotes location of PAM unit in The Sutors, circles indi- cate CPAs labelled with Maritime Mobile Service Identity (MMSI) number; (c) Range of AIS transmissions from PAM unit versus time; (d) 1/3 octave spectrum of concurrent acoustic data; (e) Broadband level in frequency range 0.1-1 kHz, showing peak identification using adaptive threshold.
Figure 4.15: Broadband SEL per day for days with uninterrupted AIS coverage of The Sutors. (a) 0.1-1kHz (b) 1-10 kHz. Noise exceeding the adaptive threshold was attributed to AIS vessel movements or classed as unidentified. ‘Rig towed using DP’: this data did not exceed the adaptive threshold, but was attributable to AIS vessels (see text). (c) Mean SPL per day for four 1/3 octave frequency bands, including those proposed for use in the MSFD (63 and 125 Hz).
Figure 4.15: Broadband SEL per day for days with uninterrupted AIS coverage of The Sutors. (a) 0.1-1kHz (b) 1-10 kHz. Noise exceeding the adaptive threshold was attributed to AIS vessel movements or classed as unidentified. ‘Rig towed using DP’: this data did not exceed the adaptive threshold, but was attributable to AIS vessels (see text). (c) Mean SPL per day for four 1/3 octave frequency bands, including those proposed for use in the MSFD (63 and 125 Hz).
Figure 4.16: Relationships between broadband SEL (0.05-1 kHz) per day and mean SPL per day at The Sutors for four 1/3 octave frequency bands, including those proposed for use in the MSFD (63 and 125 Hz).
Figure 4.16: Relationships between broadband SEL (0.05-1 kHz) per day and mean SPL per day at The Sutors for four 1/3 octave frequency bands, including those proposed for use in the MSFD (63 and 125 Hz).
Figure A-2: False Negative 2/5. The magnitude of the SPL peak (148.6 dB re 1 Pa”) and the trajectory of the closer vessel, a 77-m cargo ship (518432000), suggest that the closer vessel was the source of this peak. [ Top: Range from hydrophone vs. time. Crosses denote individual AIS transmissions; lines connect transmissions from the same vessel; circles indicate closest points of approach, labelled with the MMSI number. Shaded area denotes 15-minute time window around current SPL peak. Horizontal lines indicate AIS transmissions from station- ary vessels. Middle: Power spectral density of concurrent acoustic data. Bottom: Broadband (0.01-1 kHz) SPL, showing ‘background’ (green), ‘intermittent’ (red), and ‘intermittent’ peaks (black)].
Figure A-2: False Negative 2/5. The magnitude of the SPL peak (148.6 dB re 1 Pa”) and the trajectory of the closer vessel, a 77-m cargo ship (518432000), suggest that the closer vessel was the source of this peak. [ Top: Range from hydrophone vs. time. Crosses denote individual AIS transmissions; lines connect transmissions from the same vessel; circles indicate closest points of approach, labelled with the MMSI number. Shaded area denotes 15-minute time window around current SPL peak. Horizontal lines indicate AIS transmissions from station- ary vessels. Middle: Power spectral density of concurrent acoustic data. Bottom: Broadband (0.01-1 kHz) SPL, showing ‘background’ (green), ‘intermittent’ (red), and ‘intermittent’ peaks (black)].
Figure A-4: False Negative 4/5. The further CPA at ~40 km is a 13-m catamaran (235075634), while the nearer CPA is a 190-m bulk carrier (240271000). [ Top: Range from hydrophone vs. time. Crosses denote individual AIS transmissions; lines connect transmissions from the same vessel; circles indicate closest points of approach, labelled with the MMSI number. Shaded area denotes 15-minute time window around current SPL peak. Horizontal lines indicate AIS transmissions from stationary vessels. Middle: Power spectral density of concurrent acoustic data. Bottom: Broadband (0.01-1 kHz) SPL, showing ‘background’ (green), ‘intermittent’ (red), and ‘intermittent’ peaks (black) ].
Figure A-4: False Negative 4/5. The further CPA at ~40 km is a 13-m catamaran (235075634), while the nearer CPA is a 190-m bulk carrier (240271000). [ Top: Range from hydrophone vs. time. Crosses denote individual AIS transmissions; lines connect transmissions from the same vessel; circles indicate closest points of approach, labelled with the MMSI number. Shaded area denotes 15-minute time window around current SPL peak. Horizontal lines indicate AIS transmissions from stationary vessels. Middle: Power spectral density of concurrent acoustic data. Bottom: Broadband (0.01-1 kHz) SPL, showing ‘background’ (green), ‘intermittent’ (red), and ‘intermittent’ peaks (black) ].
Figure A-7: False Positive 2/3. The CPA at ~40 km is a 19-m fishing vessel (227346000). Due to the speed of the transit, it is probable that this peak was due to a small vessel close to the deployment not transmitting AIS data. [ Top: Range from hydrophone vs. time. Crosses de- note individual AIS transmissions; lines connect transmissions from the same vessel; circles indicate closest points of approach, labelled with the MMSI number. Shaded area denotes 15- minute time window around current SPL peak. Horizontal lines indicate AIS transmissions from stationary vessels. Middle: Power spectral density of concurrent acoustic data. Bottom: Broadband (0.01-1 kHz) SPL, showing ‘background’ (green), ‘intermittent’ (red), and ‘inter- mittent’ peaks (black)].
Figure A-7: False Positive 2/3. The CPA at ~40 km is a 19-m fishing vessel (227346000). Due to the speed of the transit, it is probable that this peak was due to a small vessel close to the deployment not transmitting AIS data. [ Top: Range from hydrophone vs. time. Crosses de- note individual AIS transmissions; lines connect transmissions from the same vessel; circles indicate closest points of approach, labelled with the MMSI number. Shaded area denotes 15- minute time window around current SPL peak. Horizontal lines indicate AIS transmissions from stationary vessels. Middle: Power spectral density of concurrent acoustic data. Bottom: Broadband (0.01-1 kHz) SPL, showing ‘background’ (green), ‘intermittent’ (red), and ‘inter- mittent’ peaks (black)].
Figure B-2: Decelerating vessel example. First noise peak [see part (e)] occurs as vessel decel- erates to a halt at 00:50, while second peak corresponds to CPA at 07:50 as vessel transits into the Cromarty Firth. (a) Still of time lapse footage showing vessel; (b) Map of AIS movements in period centred on CPA. Black cross denotes location of PAM unit in The Sutors, circles indicate CPAs labelled with Maritime Mobile Service Identity (MMSI) number; (c) Range of AIS transmissions from PAM unit versus time; (d) 1/3 octave spectrum of concurrent acous- tic data; (e) Broadband level in frequency range 0.1-1 kHz, showing peak identification using adaptive threshold.
Figure B-2: Decelerating vessel example. First noise peak [see part (e)] occurs as vessel decel- erates to a halt at 00:50, while second peak corresponds to CPA at 07:50 as vessel transits into the Cromarty Firth. (a) Still of time lapse footage showing vessel; (b) Map of AIS movements in period centred on CPA. Black cross denotes location of PAM unit in The Sutors, circles indicate CPAs labelled with Maritime Mobile Service Identity (MMSI) number; (c) Range of AIS transmissions from PAM unit versus time; (d) 1/3 octave spectrum of concurrent acous- tic data; (e) Broadband level in frequency range 0.1-1 kHz, showing peak identification using adaptive threshold.
Related topics:
AcousticsUnderwater AcousticsNoise PollutionImpact of Underwater NoiseUnderwater noise
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved