Pilot whale stranding: acoustic analysis report

5. Results

5.1 Cetacean detections

5.1.1 Tolsta site

Broadband data

Cetacean vocalisations (echolocation clicks and whistles) were detected on the broadband recorders from the Tolsta site regularly throughout the analysis period. The mean duration of identified cetacean vocalisations per 10 minute (600 seconds) recording was 52.4 seconds (range= 0 – 599.8 seconds). A total of 75.9% (155 out of a total of 204) of recording windows contained inferred cetacean activity. Example detections are provided in Figure 6.

Figure 6. Spectrograms of cetacean detections captured in the Tolsta dataset; a (top) shows a cetacean whistle sequence, and b (bottom) shows an echolocation click sequence. The x-axis displays time (in minutes); the y-axis displays the sound frequency; this is shown in Hz where, for example, 1 x 10⁴ or 10,000 Hz is equivalent to 10 kHz. The colour bar, shown on the right, represents the amplitude (sound pressure level) in dB re 1 μPa, from blue (low) to yellow (high). Device sampling frequency set up at 128 kHz. The spectrogram was generated through MATLAB software with a window length of 2048 and 50% overlap, accounting for the gain applied during recording (15 dB) and sensitivity of the device (-170.0 dB re 1V/µPa).

In total 1,330 unique cetacean detection events were recorded in the broadband data throughout the Tolsta deployments. A slight diurnal pattern was observed, with more cetacean detections occurring overnight (Figure 7). The mean number of detections per 10 minute recording window was 17.5 at night compared to 14.1 during daylight hours. However, mean total time of recordings of cetaceans during daylight hours was 387 seconds per 10 minute recording window compared to 260 seconds at night, meaning although fewer unique detections, individual vocalisation events lasted longer during daylight hours. In most cases cetacean presence was inferred from the presence of whistles used for communication, although some echolocation sequences were also detected.

Figure 7. Inferred cetacean detections between 10^th and 16^th of July, from the broadband data. Bars represent individual 10 minute recording windows from the first deployment (767-Tolsta, left) and the second deployment (868-Tolsta, right). The height of the bars indicates (a) the total time (in seconds) of inferred cetacean vocal behaviour recorded during that window and (b) total number of unique cetacean detection events during that window. The colour of the bar indicates whether the recording occurred during predominantly daylight hours (between local astronomical sunrise and sunset; 05:00-22:00 UTC; brown) or during night time hours (between local astronomical sunset and sunrise; 22:00 – 05:00 UTC; dark blue).

F-POD data

F-POD data at the Tolsta site suggest a reasonably steady detection rate of delphinid species (which may include pilot whales) and harbour porpoises, generally peaking between 18:00 UTC and 04:00 UTC and decreasing as time of day approached 12:00 UTC (Figure 8). Additionally, the F-POD classifier detected vessel sonar on two occasions during the reporting period, both on 16^th July, from 04:00 to 08:00 UTC and from 10:00 UTC to 12:00 UTC, ranging from 1-5 minutes of sonar logged per hour. The sonar encountered, according to the F-POD, had a frequency that ranged from ~27 kHz to ~59 kHz. Porpoises and delphinids were also detected during this period.

5.1.2 Garenin site

Broadband data

Of the 169 10-minute recordings from the deployment at 747-Garenin, only a total of 20 had any cetacean detections. Of these, five contained click detections but no whistles, 12 had whistles only and no clicks, and three had both clicks and whistles. However, detections could not be identified to species level. Only four of the 20 marine mammal detections occurred during day hours, with three of them consistent with click only detections.

There were only two 10-minute periods (out of a total of 167) with cetacean detections at the 864-Garenin deployment, both on 16^th July, one at 10:01 UTC and the other one at 23:30 UTC. However, the ability to detect cetaceans in this second deployment period was severely compromised due to the presence of self-noise in the recorder.

F-POD data

F-POD data from the Garenin site showed delphinid and porpoise detections only, but at a much lower rate than the Tolsta site (Figure 9). No sonar was detected at this site.

5.2 Anthropogenic and non-cetacean sounds

5.2.1 Tolsta site

Throughout the seven days of data analysed, shipping noise was regularly encountered (Figure 10) at the Tolsta site. Shipping noise consisted of continuous narrowband signals with several harmonics extending up to 40 kHz.

Anthropogenic sounds of potential interest (i.e. qualitatively different to the typical ambient conditions and particularly any pulsed sounds of any frequency) were catalogued for the seven days analysed. Impulsive noises were found on 16^th July 2023 at 01:00 UTC and 05:00 UTC on the morning of the day the stranding was reported; further details are below. There were no other pulsed repeating sounds found during the seven days analysed apart from the echosounder on the vessel which retrieved the first device. Due to the amount of self-noise on the recordings there is a risk some anthropogenic sounds may have been missed or misclassified if their frequencies overlapped with the self-noise. It should also be highlighted that the detection range of broadband recorders is affected by additional noise sources, along with the signal strength and range to source. Furthermore, the duty cycling on the broadband recorders means that data were only collected for 10 minutes, with a 20 minute gap until the next recording. Any sounds occurring during the off period would not have been recorded.

Impulsive sound was detected throughout the 10 minute recording on 16^th July 2023 at 01:00 UTC (Figure 11). This sound consisted of three to four pulses emitted every one second with a pulse length of ~0.15 seconds. The spacing between pulses followed a pattern, with pulses one to two at 0.01 seconds and pulses two to three at 0.02 seconds. Most of the energy of the pulses was concentrated at ~12 kHz. The amplitude of the sound decreased over time, suggesting that the source was either on a moving vessel, (although vessel noise was not discernible) or was static and the amplitude was being reduced over time. Cetacean whistles were also present throughout the recording. The self-noise from the recorder was also audible as clicking.

Figure 11. Spectrograms of some of the encountered impulsive sounds on 16^th July at 01:00 UTC, showed as a sequence (a) and zoomed in for more detail (b). The x-axis displays time (in minutes); the y-axis displays the sound frequency; this is shown in Hz where, for example 10,000 Hz is equivalent to 10 kHz. The colour bar, shown on the right, represents the amplitude (sound level) in dB re 1 μPa, from blue (low) to yellow (high). Device sampling frequency set up at 128 kHz. The spectrogram was generated through MATLAB software with a window length of 2048 and 50% overlap, accounting for the gain applied during recording (15 dB) and sensitivity of the device (-170.0 dB re 1V/µPa). Note that in Figure 11.b the vertical lines are the electrical self-noise from the faulty broadband sound recorders.

Impulsive sounds were also found throughout the 10 minute recording on 16^th July 2023 at 05:00 UTC (Figure 12). This sound consisted of a pulse emitted every 0.8 seconds and a pulse length of ~0.3 seconds. Most of the energy of the pulses was concentrated at ~50 kHz. The amplitude of the noise decreased over the 10 minute recording. There was also low frequency ~1 kHz shipping noise detectable on the recording which also decreased over time. This suggests that the sound source was on the vessel producing the shipping noise but this cannot be stated with any certainty.

Figure 12. Spectrograms of some of the encountered impulsive sounds at ~50 kHz at 05:00 on 16^th July, showed as a sequence (a) and zoomed in for more detail (b) and c). The x-axis displays time (in minutes); the y-axis displays the sound frequency; this is shown in Hz where, for example 5x10⁴ or 50,000 Hz is equivalent to 50 kHz. The colour bar, shown on the right, represents the amplitude (sound pressure level) in dB re 1 μPa, from blue (low) to yellow (high). Device sampling frequency set up at 128kHz. The spectrogram was generated through MATLAB software with a window length of 2048 and 50% overlap, accounting for the gain applied during recording (15dB) and sensitivity of the device (-170.0 dB re 1V/µPa).

The sonar that was detected by the FPOD between 04:00 UTC to 08:00 UTC and 10:00 UTC to 12:00 UTC (Figure 8), was not found on the broadband recordings apart from the impulsive sound found at 05:00 UTC (Figure 12) given the coinciding time stamp. It is not always expected that adjacent recorders pick up the same sounds. The FPODs and broadband devices operate in different ways, and sounds collected by both are classified using different processes.

5.2.2 Garenin site

Anthropogenic continuous sound below 1 kHz was regularly encountered throughout the Garenin site deployments, and was identified as being most likely shipping noise. From the spectrum of noise, some of the activities were very close to the recorder. Additionally, <1 kHz “knocking sounds” were identified in both deployments, although these were only further investigated in 747-Garenin deployment as the self-noise present in 867-Garenin impeded further analysis.

Timing patterns were extracted for these “knocking sounds”. A total of 192 events of pulse series (knocking bouts) were recorded, over the three days of data that were analysed (Figure 13). The number of pulses per bout ranged from 2 to 102, with a mean of 18.2 pulses (±17.8 SD). The duration of bouts ranged from 4 to 6 s, with a mean of 6.8 s and a median of 4.44 s. The pulses were grouped into three categories: (1) quiet, (2) medium, and (3) loud events, indicating the relative signal-to-noise ratio and suggesting variations in distance from the hydrophone, though still relatively close. The mean amplitude, measured in sound pressure level (dB re 1 μPa), of the quiet bouts was 120 dB (SD = 3.7 dB), of the medium bouts was 124 dB (SD = 4.8 dB), and of the loud bouts was 129 dB (SD = 5.9 dB). The mean signal-to-noise ratio (SNR), measured in dB RMS, of the quiet bouts was 10 dB (SD = 3.8 dB), of the medium bouts was 15 dB (SD = 5.9 dB), and of the loud bouts was 27 dB (SD = 5.9 dB) (Figure 14).

The “knocking sounds” were quiet, short sounds, suggesting they originated from a low-intensity source close to the recorder. The appearance of the pulses in bouts with relatively constant timing between them (inter-pulse interval) and a timing pattern (stable inter-pulse interval with a slower inter-pulse interval in some cases at the start or end of a bout) suggested a possible biological source, most likely a species of fish. The pulses were detected between 04:00 UTC and 21:00 UTC throughout the three days of recordings, with no pulses detected at any other time of day. After reviewing fish sound samples from fishsounds.net, and consulting a fish acoustics expert, the “<1 kHz knocking sounds” were determined to most likely come from haddock (Melanogrammus aeglefinus). Male haddock are known to produce a variety of knocking sounds during the spawning season (Hawkins & Amorim 2000), although they can produce sounds all year-round.

5.3 Other information sources

Sounds detected on the Tolsta site broadband recorders that were consistent with vessel based sources (i.e. echosounder transducers) were visually compared in time and space with Marine Directorate held Automated Identification System (AIS) and Vessel Monitoring System (VMS) data. No vessels in the study area were identified that could have produced the sounds detected.

Scottish Government requested information from the MOD on military activity out to 100 nautical miles from the stranding, from 9^th to 16^th July 2023. Below is a record of the activities that were reported:

• There was one frigate and two mine hunters in the area. Further location detail cannot be provided but there was no use of active sonar (except standard navigational (echosounder) types) by these vessels
• There was no live firing of weapons in the area.
• There was no MOD activity on the MOD Hebrides Range (some commercial vessel traffic noted) or at the Cape Wrath Firing Range.
• There was a trial of a circa 5 m Autonomous Underwater Vehicle occurring at BUTEC for two days (11^th and 13^th) which was carrying a low powered MBES (Multi Beam Echo Sounder) as well as its own navigational echosounders.
• There was a low level (circa 220 ft) sortie by a fixed wing maritime patrol aircraft for four hours on 10^th July centred to the West of Lewis. Again no use of active sonar by the aircraft or from deployed sonar buoys.

OPRED reported that there were no licensed oil and gas activities in the region, in the week preceding the stranding.

MD-LOT reported that dredging works took place at Stornoway Deep Water Port in the week preceding the stranding.

MD-LOT also reported that geophysical surveys were undertaken at the Spiorad Na Mara offshore wind farm lease area, to the west of the Isle of Lewis. Two vessels were involved in the surveys. Many of the pieces of equipment that were in use produced sounds at very high frequencies, or at low source levels, but on the 10^th, 11^th, 12^th and 13^th of July, a sub bottom profiler operating at around 100 kHz (85 kHz to 155 kHz), with a beam width of ~2° and a source level around 247 dB re 1 µPa @ 1 m was used on one vessel, and on 11^th July a sub bottom profiler operating at 8 kHz and 100 kHz (source level not provided) was used on the other vessel.