Heard off Heard: Acoustic detections of sperm whales (Physeter macrocephalus) and other cetaceans off Heard Island

Long-term passive acoustic monitoring (PAM) can provide an efficient and cost-effective means of detecting marine mammals, including sperm whales (Physeter macrocephalus) (André , 2017; Giorli , 2016; Giorli and Pinkerton, 2023; Mellinger , 2004; Miller and Miller, 2018). Sperm whales have been known to depredate toothfish from fishing vessels off Chile, Falkland, South Georgia, and Crozet and Kerguelen Islands (Hucke-Gaete , 2004; Tixier , 2019a; Towers , 2018). Recent efforts to better understand baseline behavior and interactions among sperm whales and fishing vessels in Southern Ocean waters have included an initial exploration of reported interactions proximal to the Sub-Antarctic islands of Heard Island and McDonald Islands (HIMIs), which represent Australia's most remote and southerly external territory (Tixier , 2019a).

Sperm whales are very loud (Møhl , 2000), highly vocal animals, and this makes PAM an excellent tool for studying their ecology. Sperm whales make loud echolocation clicks, approximately once per second for roughly 80% of their foraging dives, which average around 40 min in duration (Douglas , 2005). As the most commonly recorded sound of sperm whales, these long-range echolocation clicks are often called “usual clicks.” Sperm whales also make creak vocalizations that are associated with prey-capture attempts (Miller , 2004), and slow clicks that are believed to be used for communication (Oliveira , 2013). Furthermore, sperm whale clicks contain information about the size of the individual embedded within the interpulse interval (IPI) of the click waveform (Adler-Fenchel, 1980; Gordon, 1991; Møhl, 1999; Zimmer , 2005). Measurement of IPI enables estimation of the size of the whale's head, which has been demonstrated to be allometrically (Gordon, 1991; Growcott , 2011; Rhinelander and Dawson, 2004) and empirically (Dickson , 2021) related to the overall length of the whale. Thus, with recordings of sufficient duration, one can tell the behavior of sperm whales from the type of clicks, size from IPI estimates, and potentially the number of individuals in the recording from instantaneous click rates/interclick intervals (ICIs).

Here, we present results from a trial study of a moored PAM device deployed in deep water near toothfish fishing grounds off Heard Island to collect information on the seasonal presence of sperm whales in this region. We focus our analysis on describing the presence (temporal occupancy) of sperm whale echolocation clicks in the vicinity of the recorder, and the estimation of the IPI, and associated length, of recorded whales.

The exclusive economic zone (EEZ) surrounding HIMI is home to Australia's largest fishery for Patagonian toothfish (Dissostichus eleginoides). The HIMI toothfish fishery has been in operation since 1997 (Ziegler and Welsford, 2019). Commercial fishing activity at HIMI was initially restricted to trawling. However, both improvements in longline gear related to the mitigation of seabird bycatch and increases in catch rates of target species supported the introduction of longline gear in 2003 and then dominance of this fishing method within the waters of HIMI from 2011 to the present.

An Australian Antarctic Division (AAD) manufactured long-term acoustic recording device was moored off Heard Island in 2017 to listen for the presence of sperm whales and other marine mammals. The AAD Moored Acoustic Recorder (AAD-MAR), originally designed for year-long deployments in the Antarctic and described in detail by Miller and Miller (2018), was deployed by a long-line fishing vessel, FV Atlas Cove. The AAD-MAR included an HTI-90U hydrophone (HiTech, Inc., Long Beach, Mississippi) with a nominal sensitivity of −165 dB re 1 V/μPa. The AAD-MAR recording unit had a sample rate of 12 kHz and wrote 16-bit wav files. The recorder was AC-coupled via a high-pass filter with a −3 dB point of 6.6 Hz and contained an amplifier with a gain of 20 dB, and an anti-aliasing (lowpass) filter with a corner frequency of 4 kHz and roll-off of 120 dB/decade. The recording unit was contained within submersible glass sphere which was attached to a titanium frame at the top of a 33-m-long mooring. This was the first deployment and recovery of an AAD-MAR by a commercial fishing vessel.

The main considerations for the deployment location were (1) at a depth between 1 and 3.5 km in order to be consistent with Antarctic deployments and not exceed the maximum rated depth of the mooring; and (2) a location well out of the way of fishing operations to reduce chances of fouling of long-line gear, but close enough to provide information on sperm whale presence near fishing grounds. The resulting location for deployment was 53°S, 76° 8.8'E at a depth of approximately 1050 m (see Fig. 1). The AAD-MAR recorded continuously at 12 kHz sample rate from September 16, 2017 and recorded until batteries were depleted on March 31, 2018, collecting 4703 h (6.5 months) of underwater sound recordings.

To obtain an overview of the acoustic data, long-term spectral averages (LTSAs) were created in PAMGuard. LTSAs were used to generate a visual representation of the acoustic data where each column represented 1 min of audio and each row represented a band of acoustic frequencies. Two different LTSAs were used: one to view mid-frequency sounds between 0.1 and 6 kHz, and the other to view low frequency sounds between 0.01 and 250 Hz. LTSAs were inspected by an expert (B.S.M.) to quickly assess the overall quality of recording and where possible to identify the presence of marine mammals, anthropogenic, and environmental sources of noise. Additionally, a very high-level synoptic overview was also calculated from by taking mean of all 1-min spectra for each hour.

An automated detection algorithm was used to detect hours of recording that contained echolocation clicks of sperm whales. The algorithm used is described in Miller and Miller (2018), and the principles of operation are described here. First, the click detection algorithm from the software program PAMGuard (version 2.01.05) was used to detect clicks within the recording (Pamguard Development Team, 2021). All parameters for the PAMGuard click detector and other modules are included in the supplemental material. The PAMGuard click detector is not specific to sperm whales, so the next stage of analysis involved classification of all clicks within an hour to determine whether they were produced by sperm whales. To achieve this, the ICI was calculated for all clicks within an hour, and a histogram of ICIs was created for each hour. Then, histogram bins known to correspond to ICIs of sperm whales (between 0.1 and 1.5 s) were summed and divided by the ICI of bins known to correspond to environmental noise (between 3 and 3600 s) to obtain a classification score. Last, thresholds of 0.5 for classification score and a minimum number of 300 clicks per hour were applied. Hours were considered positive classifications, and sperm whales considered present when the classification score and number of clicks exceeded these thresholds. Hours with positive classifications were inspected manually, and all false positives were removed. Finally, the total number of (false-positive-free) hours with sperm whale presence were calculated for each day (24 h starting from 00:00 UTC).

The parameters chosen for this detector and classifier were identical to those from Miller and Miller (2018), which were found to have true positive rate of 96% (i.e., it missed only 4 of 100 h with detections) and a false positive rate of 1% (i.e., 1/100 of the hours without sperm whales may be classified as a detection) on a stratified random sample of recordings from the Antarctic. True and false positive rates will be data-dependent, but these rates derived from Antarctic data can be considered indicative if sperm whale sounds and noise conditions are not substantially different from those in this study (see Sec. 4).

Analysis of sperm whale IPI was undertaken to investigate the size of detected sperm whales. Temporally contiguous blocks of hours with detections were combined into a recording session and were processed further using PAMGuard's IPI module (Miller , 2013) to obtain a single estimate of IPI of the acoustically dominant whale for that session. IPIs, I, were converted to total lengths in meters, L, based on the equation of Dickson (2021): L = 0.9016(I) + 8.593.

In addition to the time series of detections throughout the year, we also investigated presence over the diel cycle. Hours with detections of sperm whales were assigned to a light regimen based on solar altitude (day: altitude ≥0; night: altitude <0). The odds ratio test, conditioned on days with detections, was used to determine whether there was a difference between presence and absence by light regimens.

Inspection of both synoptic (1 h) and detailed (1 min) LTSAs for quality control of recordings revealed sounds not only from sperm whales, but also Antarctic and pygmy blue whales (with call types from both the Southwest Indian Ocean and Southeast Indian Ocean), fin whales, minke whales, killer, and/or pilot whales, and occasionally noise from nearby ships. Due to the non-systematic nature of inspection of LTSAs (i.e., only intense events were inspected in further detail with no attempt made to systematically document every occurrence of every species), occurrences of these detections were not quantified in time, but exemplars of these detections, and full duration LTSA plots are included as Supplementary materials.

Of the 4703 h of recording, 727 were above the automated detection threshold for sperm whales and were inspected for false positives: 719 h inspected were confirmed to contain sperm whale echolocation clicks; 8 h inspected were found to be false positives and were removed from further analysis. Sperm whales were seldom present from September through January (n = 35 h) but were present on one or more hours of nearly every day of February and March (n = 684 h) (Fig. 2). These 719 h yielded 119 recording sessions that were used to attempt to estimate IPI.

IPI estimation yielded precise estimates (within ±1 ms) for 34 recording sessions. IPIs ranged 7.133 to 8.133 ms but were heavily skewed toward the higher values. These IPIs corresponded to a median length of 15.8 m with a minimum of 15.1 m and max of 16.00 m (Fig. 2).

Sperm whale clicks were detected during both day and night but showed a bias toward daylight hours with a peak near solar noon (Fig. 3). An odds ratio test revealed detections were significantly more likely during daylight hours than night (p = 0.0024).

This work represents the first long-term acoustic survey for cetaceans in the vicinity of Australia's Sub-Antarctic Heard and McDonald Islands. Preliminary detections of marine mammal sounds included a diverse range of baleen whales, killer, and/or pilot whales (supplementary material Fig. S2), and revealed insights about the temporal occupancy and demographics of sperm whales in the vicinity of the recording site (Figs. 2 and 3).

Inspection of LTSAs revealed the presence of sounds known to be produced by two different populations of pygmy blue whales (Southeast Indian Ocean and Southwest Indian Ocean), Antarctic blue whales, fin whales, minke whales, and killer whales. This list should not be considered exhaustive for two reasons: (1) inspection of LTSAs was intended only for the purposes of a qualitative check of the data; and (2) the LTSA parameters that were inspected are very useful for finding highly repetitive and long-duration narrowband sounds but are not suitable for finding short duration sounds that are not repeated regularly. Thus, sounds, such as blue whale D-calls (Rankin , 2005), fin whale 40 Hz downsweeps (Miller , 2021; Širović , 2013; Wiggins and Hildebrand, 2020), sei whale calls (Calderan , 2014), right whale calls (Webster , 2016), humpback whale social sounds (Dunlop , 2008), and other non-stereotyped non-repetitive sounds, are unlikely to be noticed when inspecting LTSAs, but could still be present in the data.

All three populations of blue whales that were detected are known to inhabit the central Indian Ocean (Samaran , 2013). However, our detections of pygmy blue whale sounds represent the furthest south that these species have been detected on long-term moored recordings in the central Indian region, and potentially the first published detection of the SWI call type in Australian waters. The recordings of fin whales and minke whales may be a first for this region but appear consistent with what is known about fin (Aulich , 2022) and, to a lesser extent, minke whale distribution outside of the Antarctic (Ropert-Coudert , 2014).

While baleen whale calls are known to propagate long distances in the Southern Ocean, the intensity of the calls presented in Fig. 2 suggests that the animals producing these calls in these instances are likely within a few tens of kilometers from the recorder, rather than hundreds or thousands of kilometers away. However, further quantitative modelling of the detection range would be required to confirm this.

Cetacean sightings across the Kerguelen Plateau have primarily been collected aboard research vessels of opportunity or commercial fishing vessels. A geophysical survey conducted in waters to the southeast of Heard Island during early 2020 undertook systematic visual surveys en route and reported long-finned pilot whales (Globicephala melas), fin whales (Balaenoptera physalus), sei whales (Balaenoptera borealis), hourglass dolphins (Lagenorhynchus cruciger), southern bottlenose whales (Hyperoodon planifrons), humpback whales (Megaptera novaeangliae), and a mixed pod of Commerson's (Cephalorhynchus commersonii) and dusky dolphins (Lagenorhynchus obscurus) (Todd and Williamson, 2021). Satellite tracking data from humpback whales have been used to suggest the presence of a focal foraging area for this species in the BANZARE Bank area located in the southern Kerguelen Plateau (Bestley , 2019). Historical surveys conducted in a similar region highlighted the presence of sperm whales, minke whales, and humpback whales in association with oceanographic features (Tynan, 1997). The region has also been found to provide important habitat for a resident and presumed endemic population of a sub-species of Commerson's dolphin (C. commersonii kerguelenensis) in the northern areas of the Kerguelen Plateau within the French EEZ (IUCN-Marine Mammal Protected Areas Taskforce, 2021). Depredation by both sperm whales and orca has been documented in both the Kerguelen and HIMI fisheries (Tixier 2020). Numerous beaked whale fossils have been recovered by longline fisheries operating across the Plateau (Lambert , 2019). Furthermore, a similar suite of cetacean species documented in the various research studies listed above has been reported by fisheries observers aboard commercial fishing vessels within the French EEZ of Kerguelen (Gasco , 2019).

Despite the limited duration of our study, our acoustic detections confirm that two acoustic populations of pygmy blue whales and blackfish (killer or pilot whales) appear to make use of this area. Our observations of Antarctic blue whales and minke whales add to this list of species and populations that use this area. Based on our findings of continuous daily presence of sperm whales from late January through March, it seems likely that the area is important to sperm whales too — at least during some months or seasons.

The automated sperm whale detector worked very well on this dataset despite being developed for detection of sperm whales further south in Antarctic waters. The performance of automated detectors is often site-specific. The false positive rate was much lower in this study off Heard Island, 8/3984 = 0.2%, than for Antarctic recordings where average false positive rate was 1.1% (Miller and Miller, 2018). Additional work would be required to confirm the true positive rate of detections off Heard Island. This would involve manual inspection of a sufficiently representative subsample of hours to determine the number of missed detections, and this was not done here because it is time-intensive and somewhat subjective.

The detections of sperm whales revealed a very clear seasonal temporal trend, with hardly any detections from September until the end of January. From the end of January through the end of recording in March, sperm whales were present most days, often for many of the hours of that day. Studies of sperm whale demographics and seasonality have been conducted to the northwest of HIMI within the Kerguelen longline fisheries. These studies have noted that depredation by sperm whales was seasonal, and the highest numbers of interactions per fishing unit occurred between October and March with peaks in January (Labadie , 2018). The seasonality observed at these adjacent lower latitudes appears to be different than what we observed off HIMI. Seasonality of passive acoustic recordings within the Southern Ocean yet at more southerly latitudes appeared more consistent with the results off HIMI. For example, in the Ross Sea, i.e., (i) ∼73S and (ii) ∼6340S, suggested that detection rates were lower with increasing latitude and that there were also some gaps in presence in December to January (Giorli and Pinkerton, 2023). In the high-latitude waters of the Southern Kerguelen Plateau adjacent to HIMI, sperm whales have been found to be present when sea ice was low to absent (predominantly from December to February, but occasionally in November, and March). The seasonality of sperm whales at HIMI is consistent with northward migration of animals from the Southern Kerguelen Plateau. However, further years of data, and/or tracking of individual animals (e.g., with tags or capture-recapture methods) would be required to confirm whether the observed timing is annually consistent and a stable part of the life history of these animals, or just coincidental.

Diel trends in detections were clear as well. While sperm whales near our HIMI recording site were detected at all hours around the clock, they were significantly more likely to be detected during daylight hours than at night. This higher likelihood of daylight detection was in accord with similar studies of sperm whales at higher latitudes in the deep waters of the Mediterranean (André , 2017) and Antarctic (Miller and Miller, 2018), though different from diel patterns found in other studies off Hawaii and the Ligurian Sea (Au , 2013; Giorli , 2016). A PAM study conducted in the Ross Sea that partially overlapped the temporal coverage of the current study showed a preference for daylight occurrence in one of two recorders (Giorli and Pinkerton, 2023). It is been hypothesized that diel patterns in echolocation are linked to diel differences in diving behavior, sleep, and potentially to prey availability, though evidence to test these hypotheses remains scant (Davis , 2007; Miller , 2008; Teloni , 2008).

The detection range of echolocation clicks of sperm whales is much less variable than that of baleen whales. Previous studies have found maximum detection ranges of sperm whales to be approximately 10–20 km (Barlow and Taylor, 2005; Leaper , 2000; von Benda-Beckmann , 2018), and it is reasonable to assume similar maximum ranges for our recorder. Within a site, detections are known to vary over time depending on background noise, distance, depth, and orientation of the sperm whales, so the effective range in this study may vary from 2 to 20 km throughout the recording.

The results from our analysis of IPI suggest that sperm whales off Heard Island are large males. The IPIs detected in our study correspond to a total length between 15 and 16 m with a median of 15.8 m. These results are in accord with hypotheses and whaling data that suggest sperm whales at latitudes higher than 50° are almost exclusively adult males (Gaskin, 1970). Studies into the demographics of depredating sperm whales in the Kerguelen fishery used photographs and visual observations to support the assumption that all of the depredating sperm whale individual observed were male (Labadie , 2018). In the context of the toothfish fishery, differences in rates of sperm whale depredation have been found between HIMI and nearby, but lower in latitude, Kerguelen Island (Tixier , 2019b). If future acoustic studies were conducted off Kerguelen, these could test whether such differences are correlated with IPIs/lengths of depredating whales.

Future data collection off HIMI could help fill knowledge gaps on sperm whale presence, behavior, and size from March to September. Future work on the HIMI and Antarctic acoustics datasets could focus on more detailed investigation of the hours with sperm whale presence. This could include estimation of the number of animals detected in each hour (Douglas , 2005), creak (feeding) rates, as well as investigation of other types of sounds, such as slow clicks. This would provide additional information on natural on-site behaviors and group sizes. Although the rates of cetacean depredation at HIMI are considered to be relatively low (Tixier , 2020), the coupling of acoustic data with visual observations and the documentation of sperm whale depredation from fishing vessels might provide useful insights into sperm whale foraging patterns within HIMI. Insights into such data would be enhanced with consistent documentation of presence-absence of individuals, estimation of group sizes, and the collection of photographic images to allow identification of individuals (Gasco , 2016). Given the involvement of killer whales in depredation events, there would be value in mimicking these data collection protocols for this species also. Increasing the capacity to undertake long-term monitoring on all cetacean species within the HIMI EEZ would also address monitoring objectives outlined within the Marine Reserve Management Plan for HIMI (Commonwealth of Australia, 2014) and provide data to underpin the recent recognition of Heard and Kerguelen islands as an important area for marine mammals (IUCN-Marine Mammal Protected Areas Taskforce, 2021).

The results of our work comprise the first long-term acoustic survey for marine mammals in the very remote waters off Heard Island. Though our detections of baleen whales are mostly ad hoc and descriptive, they nevertheless shed light on the distribution of several species and populations, including Antarctic and pygmy blue whales, fin whales, and Antarctic minke whales. Though our recordings did not span a full calendar year, our analysis was nevertheless able to provide some baseline systematic information on the seasonal presence, temporal occupancy, diel behavior, and size structure of sperm whales in this area. Our results provide yet another example of the utility of fixed PAM for surveying the distribution and occupancy of marine mammals in remote and difficult to access parts of the ocean.

The authors thank the officers and crew of FV Atlas Cove and AFMA observer Henry Oak for deploying and recovering the AAD-MAR; as well as Rhys Arangio from Austral Fisheries for coordination between AAD and vessels.

This science was conducted independently and impartially by scientists and technicians at the AAD. Deployment and recovery of the AAD moored acoustic recorder was conducted with the assistance of fishing vessels provided by Austral Fisheries.

No animals were approached during this study, and equipment used in this study does not emit any sound or light, nor is there any other means by which it could impact animals.

Data and metadata used in this study are freely available from the Australian Antarctic Data Centre via the following link: https://data.aad.gov.au/metadata/records/AAS_4102_longTermAcousticRecordings and the following DOI: http://dx.doi.org/10.26179/19e3-2z24.