Hydroacoustic study of fin whales around the Southern Wake Island: Type, vocal behavior, and temporal evolution from 2010 to 2022

Fin whales (Balaenoptera physalus) are an endangered cetacean that populate global oceans with an estimated population of 10 000 (Perrin 2009). They produce a low-frequency call sequence with a source frequency spectrum centered around 20 Hz (20 Hz calls), which is thought to be a male song related to a courtship or reproductive acoustic display (e.g., Croll , 2002; Hatch and Clark, 2004; Watkins, 1981; Watkins , 1987). Those 20 Hz calls can be further divided into two different types based on the actual source frequency spectrum, with one named as type-A call having a source frequency range from 22 to 15 Hz and another termed as type-B call having a higher source frequency range from 30 to 18 Hz (e.g., Croll , 2002; Helble , 2020; Watkins , 1987; Weirathmueller , 2017). The two types of calls could coexist for several hours and appear alternately (e.g., Helble , 2020; Kuna and Nábělek, 2021; Weirathmueller , 2017). Using the active radio tagging data or passive acoustic monitoring data from the ocean bottom seismometers and hydrophones operated by the Comprehensive Nuclear Test Ban Treaty Organization (CTBTO) International Monitoring System (IMS) and some other organizations, researchers monitored those two types of calls to study the living habits of fin whale, including submerging, surfacing, blowing, moving, and migration in different oceans (e.g., Aulich , 2019; Bérubé , 2002; Fujino, 1960; Gedamke, 2009; Johnson and Tyack, 2003; Watkins, 1981; Watkins , 1987); the fin whale population density in the Pacific and Indian Oceans (e.g., Harris , 2018; McDonald and Fox, 1999; Miksis-Olds , 2019a; Sousa and Harris, 2015; Thomas , 2014); the fin whale call source levels in the tropical Pacific and Eastern North Atlantic Oceans (e.g., Harris , 2018; Miksis-Olds , 2019b); and the population structure and cultural transmission (defined by Boyd and Richerson, 1988, as the information or behavior acquired from conspecifics through some form of vocal learning in cetaceans) of fin whales in different oceans (e.g., Delarue , 2009; Hatch and Clark, 2004; Helble , 2020; Oleson , 2014; Širović , 2017; Thompson , 1992; Weirathmueller , 2017). Fin whale call acoustic signals were also used to constrain the shallow oceanic crustal velocity structure based on the arrival time of the seismic phases reflected and refracted from the oceanic crust (Kuna and Nábělek, 2021).

It is still under debate whether type-A and type-B calls come from a same fin whale with two distinguished vocal source frequencies or are made by two different individual fin whales. The two types of calls could generate a regular song pattern with approximately constant inter-pulse intervals (IPIs, which are defined as the time separation between two subsequent calls within a 60-s time window from Watkins (1987) or an irregular song pattern with widely distributed IPIs. When the two types of calls have irregular IPIs, researchers have observed that they were made by two fin whales separated by several kilometers to hundreds of meters, where each fin whale generates one type of call (e.g., McDonald and Fox, 1999; McDonald , 1995; Soule and Wilcock, 2013). However, the debate exists that these two types of calls often appeared alternately together and exhibited regular IPIs, making the source distinction of the two calls challenging. Some researchers proposed that those regularly alternated calls were generated by a single individual fin whale based on the high location coherence and close call proximity of the calls (e.g., Helble , 2020; Kuna and Nábělek, 2021), whereas other researchers proposed that they were made by two different fin whales with the type-B calls produced by a younger immature male and the type-A calls produced by a mature male (Soule and Wilcock, 2013).

The inability to distinguish the sources of the two types of calls also hindered the studies of population structure change and cultural transmission of fin whales. Thus far, the population structure and cultural transmission of fin whales were inferred based on the temporal trend of IPI values of the consecutive calls (e.g., Delarue , 2009; Hatch and Clark, 2004; Širović , 2017; Thompson , 1992; Weirathmueller , 2017). If those two types of calls indeed come from two different fin whales, the interaction between each type of call could provide important information on the vocal influence and cultural transmission between different fin whales and more accurate inference on the fin whale population change.

The key to distinguish those two types of calls lies on the accurate determination of the source locations of the two calls. In this study, we use hydroacoustic data recorded from the Southern Wake Island of the Northwest Pacific Ocean to determine the source locations of the two types of calls in high precision and identify that the type-A and type-B calls come from individual type-A and type-B fin whales. With the identification of the two types of fin whales, we also study vocal influence and individual call behavior of type-A and type-B fin whales. We present fin whale call detection and location determination in Sec. II; identification of individual type-A and type-B fin whales in Sec. III; vocal influence, individual call behavior, and temporal call pattern of type-A and type-B fin whales in Sec. IV; and various inferences of this study in the context of previous work in Sec. V.

We use recordings from the H11S hydrophones to detect and locate the fin whale calls around the Southern Wake Island in the Northwest Pacific Ocean. In Secs. II A–II C, we present the hydroacoustic data, fin whale call detection, and examples of fin whale call location.

We use continuous hydroacoustic data from January 2010 to May 2022, which were recorded by the H11S hydrophone triplet located in the Southern Wake Island of the Northwest Pacific Ocean. The H11S hydrophone triplet is operated by the CTBTO IMS and consists of three hydrophones (H11S1, H11S2, and H11S3), which are situated in a triangle of a 2.0 km spacing. Hydrophones are moored in the sound fixing and ranging (SOFAR) channel axis at water depths of 750 m (H11S1), 742 m (H11S2), and 726 m (H11S3; Fig. 1). Hydrophone sensors have a flat frequency response from 10 to 100 Hz and a sampling rate of 250 Hz. In this study, hydroacoustic waveforms are filtered in a frequency range from 15 to 35 Hz.

Fin whale call detection includes two steps. First, we detect all possible events in the data and, second, classify the fin whale calls based on the spectral characteristics of the detected events.

We detect possible events using the short-time-average over long-time-average (STA/LTA) trigger algorithm (Trnkoczy, 2009). The algorithm detects an event based on the ratio of the average recording amplitudes in a short-time window (STA) and long-time window (LTA) as STA will capture signal of an event while LTA provides information about background noise in the recording. When the STA/LTA ratio exceeds a preset value in all hydrophones, an event detection is declared with the triggered time of the direct wave set to be the beginning time of the STA time window (Trnkoczy, 2009). In this fin whale call detection, we set the durations of STA and LTA to be 1 and 3 s, respectively, and the STA/LTA triggered and detriggered threshold levels to be 2.0 and 0.4, respectively. We only retain detections that have the left signal-to-noise ratio (SNR) and right SNR larger that 2.0 (left SNR is defined as the ratio of squared amplitude summation in the triggered time window to that of a time window of same duration proceeding the earliest triggered time, whereas right SNR is defined as the ratio of squared amplitude summation in the triggered time window to that of a time window of same duration starting at the latest detriggered time).

We identify fin whale calls from those detected events based on the spectral characteristics of the hydroacoustic recordings using the following three criteria: the central frequency is in the range of 15–35 Hz, the SNR in 6–10 Hz is less than two, and the spectral energy in 14–22 Hz is greater than two times that in either of the ranges 6–14 or 22–30 Hz, or there are at least two hydrophones satisfying that the spectral energy in 14–22 Hz is larger than that in 6–14 Hz and ten times larger than that in the range 22–30 Hz. All fin whale call detections are further validated through eye checking.

A total of 94 044 fin whale calls are detected around the Southern Wake Island from November to April (termed as a season) in 2010–2022. Two types of fin whale vocalizations exist based on the central frequency of the recording, with one type exhibiting a central frequency around 17–20 Hz and having most of the energy concentrating in the frequency range of 15–22 Hz, and the other type exhibiting a central frequency around 20–30 Hz and having most of the energy concentrating in the frequency range of 18–35 Hz (Fig. 2). We classify the fin whale calls as type-A and type-B based on the central frequency of the recording following the traditional naming in the literature, with type-A calls having a central frequency lower than 20 Hz and type-B calls having a central frequency higher than 20 Hz.

We determine the locations of some fin whale calls that have clear arrivals of the direct wave and multiples that bounce between the sea surface and ocean bottom. The location is determined based on the best fitting between the observed and predicted arrival times of the direct wave and multiples and the assumption that the handpicked first arrival of fin whale calls is the observed arrival time of the direct wave (see text S1 for the verification of the handpicked first arrival as the arrival time of the direct wave). In practice, we first estimate fin whale call backazimuth based on the handpicked first arrival of direct waves in the data (see text S2 for the fin whale call backazimuth determination); we then grid-search the best-fitting location of the call in the neighboring locations centered at 20 points progressing away from the center of the hydrophone array along the inferred backazimuth direction with a step of 500 m. In grid-searching the best-fitting location of the call, first, we grid-search the possible horizontal locations with various assumed call depths by minimizing the root mean square (RMS) difference between the observed direct wave arrival times and predictions. The best-fitting depth and horizontal locations are those that have a good match between the observed and predicted arrival times of multiples. The horizontal searching region is 4 km (N–S direction) × 4 km (E–W direction), centered at each point along the inferred backazimuth, and the searching grid size is 5 m (N–S direction) × 5 m (E–W direction). The search depth range is 0–300 m with a search depth interval of 5 m. Water sound velocity can be calculated based on the local salinity and water temperature, but the calculated velocity does not generate a good fit between the observed arrival times and predicted arrival times of the multiples (text S3). The best-fitting water sound velocity that is appropriate for the ocean water in the region is, thus, also searched in the grid-search process by performing grid-search of best-fitting location through various assumed water sound velocities (Fig. S5). The best-fitting location and water sound velocity are those that generate the minimal RMS difference between the observed and predicted direct wave arrival times and a good match between the observed and predicted arrival times of the multiples. The combination of the arrivals of the direct wave and multiples places a tight constraint on fin whale horizontal location and depth as the arrival time of the direct wave is sensitive to the horizontal position of the fin whale call and the relative arrival times of upgoing multiples, and the direct wave is sensitive to the depth of the fin whale call. The predicted arrival times are calculated through ray tracing of the direct waves and multiples between the ocean surface and seafloor using the BELLHOP beam tracing program (Porter, 2011; Porter and Bucker, 1987) with the sea surface and seafloor elevations extracted from the topography model of SRTM15+. The location uncertainties in the longitudinal and latitudinal directions are defined as the standard deviations of 500 location results determined based on 500 travel time datasets that have a random perturbation of $±0.05$ s (travel time picking error defined as a time period of fin whale call waveforms) added to the handpicked arrival times of the direct wave. The depth uncertainty is defined as the maximal deviation from the best-fitting source depth due to the arrival time pick error of $0.05$ s of the multiple phases (see text S4 for the details of the determination of source depth uncertainty). We show an example of location determination using an event occurring at 02:40:32 UTC on February 13, 2020 (event 1; Fig. 2). We first obtain the inferred backazimuth of $−7.68°$ for that event, based on the handpicked arrivals of the direct wave. We then perform source location searches along the inferred backazimuth direction and obtain the best-fitting horizontal location of $(166.6954°E,18.5076°N)$ and source depth of 15 m with a minimum RMS residual of 0.1 ms [Fig. 3(a)]. The uncertainties of the event source locations are estimated to be 64 m in longitude, 74 m in latitude, and 15 m upward and 25 m downward in depth [Figs. 3(a) and S6]. The horizontal location of the event source is well constrained as evidenced from a small minimal RMS arrival time residual of 0.1 ms [Figs. 3(b) and 3(c)]. The source depth is also well constrained as the observed arrival times of the multiples can only be matched by the predictions with a source depth of 15 m and not with other assumed source depths [Figs. 3(b) and 4].

When the fin whale call recordings do not exhibit clear arrival of multiples, the call depths cannot be well determined. In those cases, we fix the source depth at 15 m during the determination of event location. Given a possible range of call depth in 15–150 m, the fixed depth assumption of 15 m would introduce a maximum possible deviation of 30 m in longitude and 52 m in latitude, which are less than the location uncertainties of 45 m in longitude and 60 m in latitude for the horizontal location determination (see an example in Fig. S7).

High-precision determination of the call locations could provide a mean to distinguish whether the type-A and type-B calls are made by two separate fin whales or a single whale with two distinct frequencies. We show that type-A and type-B calls around the Southern Wake Island are made by two individual fin whales, based on the large horizontal and depth separations between the alternate type-A and type-B calls inferred from the hydroacoustic data. We present the location separations of alternated type-A and type-B calls from 11 continuous time segments in 2010–2022. The large depth separations are observed in seven time segments between the alternate type-A and type-B calls, and large horizontal source separations are observed in two time segments between the alternate type-A and type-B calls. In the other two segments, the depth and horizontal positions of the alternate type-A and type-B calls are close to each other and not distinguishable within the location uncertainties. We show three examples of location separation of the two types of calls with two examples from the time periods when the alternate type-A and type-B calls are regular and another example from a time period when the alternate type-A and type-B calls are irregular (see text S5 for the determination of IPI regularity of fin whale calls). Examples of other time periods are presented in text S6.

Two examples of location separation of the regularly alternated type-A and type-B calls include one from 15:15:46 to 15:50:00 UTC on February 14, 2019 (Fig. 5) and the other from 02:00:00 to 02:40:00 UTC on February 13, 2020 (Fig. 6). In both examples, the horizontal locations and depths of the consecutive fin whale calls are well constrained by the observed arrival times of the direct wave and multiples (Figs. S10 and S11). In the first example of a 550-s time segment, the alternate type-A and type-B calls are separated in horizontal location and depth. The separations of the horizontal positions between the alternate type-A and type-B calls are 165 m, which are much larger than the location uncertainty of 40 m [left panel of Fig. 5(b)]. This is the largest horizontal distance separation that we find in all the time segments of the regularly alternated type-A and type-B calls, presenting a reference of the largest horizontal separation observed between the two types of calls made when they are regularly alternated. The two types of calls are also made at different depths, where the type-A calls are made in shallow depths migrating from a depth of 15 m to larger depths up to 45 m and back to the depth of 15 m, whereas type-B calls are made at larger depths migrating from a depth of 105 m to a shallower depth of 60 m and back to the depth range of 90–105 m [right panel of Fig. 5(b)]. The large horizonal separations between the alternate type-A and type-B calls are well resolved, and there is little trade-off between horizontal position and depth in the determination of call locations. Note that the differences of the inferred horizontal locations are 13 m for the type-A and type-B calls for all permissible depths of the calls within the resolutions (text S7 and Fig. S12). In the second example of a 656-s time segment, the alternate type-A and type-B calls are made in close horizontal positions that are not distinguishable within the location uncertainties [left panel of Fig. 6(b)]. However, they are made at different depths [right panel of Fig. 6(b)]. Type-A calls in that segment are made first at a depth of 15 m and then migrate to larger depths up to 105 m, whereas type-B calls stay at depths between 90 and 135 m in the same time period [right panel of Fig. 6(b)]. The large depth separations of the alternate type-A and type-B calls in that segment are well resolved with the combined constraints of the arrival times of the direct wave and multiples. We illustrate the tight constraints on the call depths using the waveform segments observed for a pair of alternated type-A and type-B calls in the time series of Fig. 6 (Fig. 7 for the observations at hydrophone H11S1 and Figs. S13 and S14 for the observations at hydrophones H11S2 and H11S3). Note that the observed arrival times of the direct waves can be well matched by the predictions under any assumed depths of 15–195 m (Fig. 7). However, the observed arrival times of the multiples can only be matched by the predictions with the inferred source depth of 135 m for the type-B call and 15 m for the type-A call, not with other assumed source depths [Figs. 7(a) and 7(b)]. The little trade-off between call depth and horizontal location can also be demonstrated by considering the difference of determined call depths for all the permission horizontal locations of the calls. Note that the differences of the inferred depths are 25 m for the type-A calls and 45 m for the type-B calls for all permissible horizontal positions of the calls within the resolutions (text S8 and Fig. S15). The inferred depths of the fin whale calls in the above two examples are similar to the range of fin whale call depths reported by Stimpert (2015). The large horizontal and depth position separations of source locations between the regularly alternated type-A and type-B calls indicate that they are made by two different fin whales.

Location determinations also show that type-A and type-B calls are in separated locations when they are made alternately with an irregular IPI. We present an example of horizontal location separation of alternate type-A and type-B calls derived from a continuous time segment of hydroacoustic data from 20:30:00 to 21:12:00 UTC on January 15, 2020. Note that in the example of a 220-s segment, the alternate type-A and type-B calls show irregular IPI values in the time window [Fig. 8(a)]. The horizontal locations of the 25 consecutive fin whale calls in the time window are well constrained by the observed arrival times of the direct wave (Fig. S16). The horizontal separations of the alternate type-A and type-B calls are larger than 5.0 km, far exceeding the location error of 2.0 km [Fig. 8(b)]. The large horizontal separations of the alternate type-A and type-B calls with irregular IPI values further support the inference that the two types of calls are made by two different fin whales.

With the large source separation confirming that alternate type-A and type-B calls come from different individual fin whales, we term the fin whales as type-A and type-B fin whales in association with the type-A and type-B calls that they make.

In Secs. IV A–IV C, we present vocal influence of type-A fin whale calls on type-B fin whale calls and call behaviors of type-A and type-B fin whales.

Detailed analysis of hydroacoustic data indicates that type-A fin whales exert a vocal influence on type-B fin whales. We illustrate the influence from the following interaction behaviors of the two types of fin whale: (1) the change of type-B calls in source central frequency and IPI value in the presence and absence of type-A calls, (2) the initiation of call sequences by type-A calls after each call break, and (3) the pairing of the two types of calls in the alternate type-A and type-B call sequences.

Type-B calls exhibit different source characteristics in the time windows when they simultaneously appear with type-A calls (85.33%) in comparison with those in the time windows when they appear alone (14.67%). We analyze the distribution of standard deviation in central frequency and IPI for the type-B calls from 94 continuous time sequences in 2010–2022, including 76 time sequences with alternate type-A and type-B calls and 18 time sequences with only type-B calls (Fig. 9). In the time sequences of alternate type-A and type-B calls, the standard deviation of the central frequency for the type-B calls is less than 0.5 Hz, and the standard deviation of the IPI between consecutive type-A and type-B calls is less than 2.0 s (Fig. 9). However, in the time sequences of type-B calls alone, the standard deviation of the central frequency for the type-B calls is distributed broadly in 0.7–2.2 Hz, and standard deviation of the IPI values between consecutive type-B calls is distributed widely in 3.0–15.0 s (Fig. 9). Type-B calls also exhibit an immediate change of source characteristics as soon as type-A calls appear. We present an example of type-B calls exhibiting different vocal characteristics between the absence and presence of type-A calls in a time sequence starting at 21:25:00 UTC on February 12, 2020. In that sequence, type-B calls emerge alone in the early part of the sequence and vocalize with irregular IPI values varying from 2.0 to 40.0 s, followed by the presence of alternate type-A and type-B calls in the later part of the sequence with type-B calls exhibiting regular IPIs with the alternate type-A calls (top panel of Fig. 10). Location determination indicates that the early and later type-B calls are made in the nearby region with clear tracks (Fig. S17). Type-B calls exhibit broad central source frequencies varying from 20 to 30 Hz when appearing alone in the early part of the time sequence, but they show a narrow central frequency around 22 Hz when appearing alternately with type-A calls in the later part of the time sequence (bottom panel of Fig. 10).

The initiation of type-A calls is consistently observed in the 76 continuous time sequences of 2010–2022 with regularly alternated type-A and type-B calls. In those sequences, the type-A call initiates the first call after each call break. We present an example of such phenomenon in a time sequence from 00:54:42 to 01:47:00 UTC on November 6, 2021. In that sequence, alternate type-A and type-B calls exhibit regular IPI values (top panel of Fig. 11) and narrow source frequencies (middle panel of Fig. 11). There are several breaks of calls between the short call sequences. The type-A call is always the one that initiates the first call after each call break, leading the call sequences (bottom panel of Fig. 11).

IPI values suggest existence of A-B pairing but nonexistence of B-A pairing. We use IPI values to identify possible pairings of type-A and type-B fin whales (A-B or B-A). When the IPI values of a fin whale call pair of type-A and type-B (e.g., A-B with A leading B) are regular in a time sequence but those of the alternate pairing (e.g., B-A with B leading A) are not, we consider the regular call pair constitutes a pairing with the former call, exhibiting vocal influence on the later call. In the 76 continuous time sequences of 2010–2022 with regularly alternated type-A and type-B calls, 26 sequences show regular IPI values between the alternate type-A and type-B calls and irregular IPI values between the alternate type-B and type-A calls, indicating A-B pairing. In the other 50 sequences, regular IPI values are observed between the alternate type-A and type-B calls and between the alternate type-B and type-A calls. In these cases, it becomes ambiguous to classify the sequences as A-B or B-A pairing based on the regularities of the IPI values as the regular IPI values could be a result of A-B pairing with type-A calls exhibiting a regular IPI value of their own or B-A pairing with type-B calls exhibiting a regular IPI value of their own. We classify the pairing based on the observed IPI values of two types of calls when they appear alone. Type-A calls exhibit a regular IPI value between their own consecutive calls in 87.57% of the time when they appear alone, but type-B calls never exhibit a regular IPI value between their own consecutive calls when they appear alone. We, thus, hypothesize that the type-B fin whales do not exhibit vocal influence on another type-B fin whale nor do they have a regular pattern of vocalization of their own. We, therefore, classify the calls in those 50 ambiguous sequences as A-B pairing, where type-A calls exhibit a regular IPI value of their own and influence alternate type-B calls with a regular IPI in A-B pairing.

In summary, (1) type-B calls exhibit a broad source frequency and an irregular IPI value when they appear alone but a narrow central frequency and a regular IPI with type-A calls when they appear alternately with type-A calls, (2) type-A calls initiate the first call after each call break during the time sequences of regularly alternated type-A and type-B calls, and (3) only A-B pairings and no B-A pairings exist in the alternate type-A and type-B calls. We conclude that type-A fin whales exert a vocal influence on type-B fin whales.

Detailed analysis of hydroacoustic data suggests that the regularly alternated type-A calls reflect a regular vocal pattern of a same type-A fin whale rather than communication of two type-A fin whales. In fact, one type-A fin whale stops calling when another type-A fin whale approaches it in close distance. We present such type-A fin whale call behaviors in a time sequence from 04:00:00 to 07:30:00 UTC on February 23, 2013. In that sequence, IPIs of type-A call exhibit a regular value of 30 s in the time window of 04:00:00–05:14:50 UTC and two regular values of 28 s and 33 s in the time window of 06:17:50–07:30:00 UTC, whereas irregular IPI values are observed in the middle part of the time sequence with values varying from 0.6 to 30 s [left top panel of Fig. 12(a)]. One single time varying trend of amplitude and backazimuth of the calls is observed in the early and late time windows of regular IPIs while two branches of time varying amplitude and backazimuth are observed in the middle time window of irregular IPIs with backazimuths separated by 100 $°$ [left middle to left bottom panels of Fig. 12(a)]. The separations of amplitude and backazimuth indicate that these calls are made by two fin whales. We separate the calls of two type-A fin whales (named A1 and A2) based on backazimuth of the calls [right panel of Fig. 12(a)]. Type-A1 fin whale appears in 04:00:00–06:17:50 UTC, whereas type-A2 fin whale appears in 05:14:50-07:30:00 UTC [right panel of Fig. 12(a)]. Location of the calls further indicates that those two type-A calls are at least 4.0 km apart with their own swimming trajectories [Fig. 12(b)]. Those two fin whales vocalize regularly in their own vocalization pattern [right panel of Fig. 12(a)]. The location determination further indicates that those two type-A fin whales seem to have a negative effect on each other's vocalization because when the type-A2 fin whale swims close to the type-A1 fin whale at a distance of about 1.6 km, the type-A1 fin whale stops making sounds while the type-A2 fin whale maintains its previous pattern of regular vocalization [right panels of Figs. 12(a) and 12(b)].

We analyze temporal patterns of fin whale call behaviors from January 2010 to May 2022. A total of 59 436 type-A calls and 34 608 type-B calls are detected in a span of 1017 days. A change of call types is evident in December 2018. Type-A calls occur in every year from 2010 to 2022 [Fig. 13(a)], whereas type-B calls occur rarely in 2010, 2012–2014, 2017, spring up in February and December 2011, January 2015, and become prevalent after November 2018 [Fig. 13(b)].

For type-A calls, 24 210 calls appear with type-B calls with 98.40% in the A-B pairing [magenta bars in Fig. 13(a)] and 1.60% not in A-B pairing [gold bars in Fig. 13(a)], and 35 226 calls appear alone with 87.57% exhibiting a regular IPI between consecutive A-A calls [pacific blue bars in Fig. 13(a)] and 12.43% exhibiting an irregular IPI between consecutive A-A calls [sky blue bars in Fig. 13(a)]. For type-B calls, other than those 24 210 calls appearing with type-A calls (magenta and gold bars in Fig. 13), 10 398 calls appear alone [dark red bars in Fig. 13(b)]. Type-A and type-B calls in the A-B pairing occur in 2011–2022 but only spring up in February 2011 and after November 2018 (magenta bars in Fig. 13). Type-A calls that appear with type-B calls but not in A-B pairing occur rarely in December 2011, January and February 2015, December 2016, and January 2020 [gold bars in Fig. 13(a)]. Type-A calls that appear alone with a regular IPI value between consecutive A-A calls spring up in 2010–2017 and hardly occur after 2020 [pacific blue bars in Fig. 13(a)]. Type-A calls that appear alone with an irregular IPI value between consecutive calls occur scattered with small numbers in 2010–2021 [sky blue bars in Fig. 13(a)]. Type-B calls that appear with type-A calls but not in A-B pairing occur rarely in December 2011, January and February 2015, December 2016, and January 2020 [gold bars in Fig. 13(b)]. Type-B calls that appear alone occur in 2011–2022 but spring up in December 2011, January 2015, December 2018, and February and December 2020 [dark red bars in Fig. 13(b)].

Type-B calls are delayed for 7 yr to become prevalent around the Southern Wake Island (Fig. 13), and the majority of them are influenced by type-A calls when they appear with type-A calls [magenta bars in Fig. 13(b)]. We suggest two possible scenarios to explain such a temporal fin whale call behavior. The first scenario is that type-B fin whales started to vocalize frequently and exhibited close vocal influence by nearby type-A fin whales after 2018, after some vocal learning in the learning period between 2010 and 2017. The second scenario is that many new type-B fin whales came to the region after 2018, i.e., the change of the number of type-B fin whale calls in the study period reflects change of type-B fin whale population in the Southern Wake Island. It is impossible to distinguish these scenarios based on their call behaviors alone. Other means of fin whale monitoring, such as photo tracking and radio/satellite tagging, could be employed to distinguish these scenarios.

In the section, we discuss the following issues in the context of the results from this study and past studies: the debate about whether type-A and type-B calls come from a single fin whale or two different fin whale individuals, possible relationship between the two types of the fin whales, the influence of type-A fin whales themselves, culture transmission between the two types of fin whales, and change of fin whale population structure.

Our results do not support the conclusion made based on the location coherence and close call proximity that the alternate type-A and type-B calls in the regular song patterns are generated by a single individual fin whale (e.g., Helble , 2020; Kuna and Nábělek, 2021). Our results show that although these two types of calls are usually in close proximity, they are made in separated locations that are distinguishable based on the hydroacoustic data. Our incorporation of the multiple data places tight constraints on the call depths and provides further clarification that the two types of calls are separated in depth. Our results support the conclusion that type-A and type-B calls are separated by several kilometers when they exhibit an irregular IPI value (e.g., McDonald and Fox, 1999; McDonald , 1995; Soule and Wilcock, 2013). However, our results reveal that these two types of calls are also separated when they exhibit regular IPIs, further strengthening the conclusion that they are made by two individual fin whales.

Type-B calls were found to exhibit regular IPIs when they appear alternately with type-A calls and irregular IPIs when they appear alone, and such behavior was used to infer that the type-A and type-B calls were produced by male fin whales and immature male fin whales, respectively (Soule and Wilcock, 2013). We observe a similar behavior, but we attribute it to be a part of the inferencing behaviors of the type-A calls on the type-B calls and/or a result of type-B fin whale learning from type-A fin whales. Our interpretation is based on the other observed broad inference of the type-A calls and more detailed behaviors of the type-B calls in the presence and absence of the type-A calls. The observed behaviors of type-A fin whale initiation of the call sequences after each call break and the existence of only A-B pairings, coupled with the observed transitions of source spectrum and IPI values of the type-B calls during the emergence of type-A calls, support our inference that the changing IPI pattern of the type-B calls is part of the inferencing behaviors of type-A calls on the type-B calls and/or a result of type-B fin whale learning from type-A fin whales. Considering this together with the other observed behaviors and interactions of type-A and type-B fin whales observed in this study, we agree with the assessment that the type-A and type-B calls could be produced by male fin whales and immature male fin whales, respectively (Soule and Wilcock, 2013).

The negative effect of a type-A call on another type-A call that we observe is consistent with the previous results from the radio tagging and passive acoustic monitoring (e.g., Watkins, 1981; Watkins , 1987). In those studies, they found no type-A calls when multiple type-A fin whales are separated by or within 1.0 km from each other (e.g., Watkins, 1981; Watkins , 1987). Our results suggest that the no type-A calls observed in those studies likely represent the negative inference between different type-A fin whales. Our results also reveal another behavior of the negative inference that a type-A fin whale stops calling when another type-A fin whale approaches it within a distance of about 1.6 km, whereas the other fin whale maintains its previous pattern of regular vocalization.

Our observed pattern of increasing numbers of type-B fin whale calls after November 2018 is similar to those reported in the Northeast Pacific Ocean (e.g., Weirathmueller , 2017), Hawaii (e.g., Helble , 2020), and Southern California (e.g., Širović , 2017). In those studies, the increase in numbers of type-B fin whale calls was also observed to be accompanied by a long-term IPI pattern shift from A-A to A-B patterns and a gradual change of IPI value between different types of consecutive calls (e.g., Helble , 2020; Širović , 2017; Weirathmueller , 2017). Researchers used those IPI pattern changes to propose a culture transmission based on the assumption that the IPI patterns could change with the environment (e.g., Weirathmueller , 2017), and a population structure change based on the assumption that the IPI pattern from each type of consecutive calls is unique to a specific population and/or stock (e.g., Delarue , 2009; Hatch and Clark, 2004; Širović , 2017; Thompson , 1992). Note that these two interpretations did not consider the type-A and type-B calls coming from two different fin whales. In this study, with the identifications of the type-A and type-B calls from different fin whale individuals and the type-A calls exhibiting vocal influence on type-B calls, we identify a clear culture transmission from type-A fin whales to type-B fin whales and a clarified picture of population change of type-B fin whales in the Southern Wake Island if the number change of type-B calls reflects the population change of the region.

We analyze the type, vocal behavior, and temporal evolution of the fin whale calls around the Southern Wake Island in the Northwest Pacific Ocean from January 2010 to May 2022, using the hydroacoustic data from the H11S hydrophones in the Southern Wake Island. We identify that (1) the type-A calls (calls with the central source frequencies varying from 17 to 20 Hz) and type-B calls (calls with the central source frequencies varying from 20 to 30 Hz) are made by two individual fin whales, based on the large source location separations determined in high precision using arrival times of various types of hydroacoustic waves from the fin whale calls; (2) type-A fin whales exhibit vocal influence on type-B fin whales in several ways: type-B fin whales become paired with type-A calls and vocalize regularly when they appear with type-A fin whales, and type-A fin whales always lead the calling of type-B fin whales; and (3) some type-A fin whales stop calling when another type-A fin whale approaches within a distance of about 1.6 km. With the identifications of the type-A and type-B calls from different fin whale individuals and the type-A calls exhibiting vocal influence on type-B calls, we study the temporal call evaluation of these two types of fin whales in the Southern Wake Island. A total of 94 044 fin whale calls are detected during the period of 2010–2022, including 59 436 type-A calls and 34 608 type-B calls. Type-A calls occur every year, whereas type-B calls are delayed for 7 yr to become prevalent after November 2018. We suggest two possible scenarios to explain such a temporal fin whale call behavior. The first scenario is that type-B fin whales started to vocalize frequently and exhibited close vocal influence by nearby type-A fin whales after November 2018 after some vocal learning in the learning period between 2010 and 2017. The second scenario is that many new type-B fin whales came to the region after November 2018.

See the supplementary material for supplementary figures and more details on the verification of the handpicked first arrival as the arrival time of the direct wave, fin whale call backazimuth determination, inappropriateness of theoretical water sound velocity for call location determination, simultaneous determination of water sound velocity and spatial location of fin whale calls, determination of source depth uncertainty of fin whale calls, determination of IPI regularity of fin whale calls, robustness of the determination of horizontal position and depth separations of alternate type-A and type-B calls, fin whale call location determinations, and predicted arrivals of various phases marked on the waveforms of the calls from more tracks.

We thank associate editor Aaron M. Thode and two anonymous reviewers for their constructive reviews, which have significantly improved the paper. We thank the Preparatory Commission for the CTBTO, Peter Polzer, Jolanta Kusmierczyk‐Michulec, Peter Nielsen, Mario Zampolli, Ilia Mishenin, and Jiayuan Yao for their help in extracting the waveform data, and Xiao Xiao for his help in analyzing the vocal behaviors of fin whale calls. The views expressed herein are those of the authors and do not necessarily reflect the views of the CTBTO Preparatory Commission. This work is supported by the National Natural Science Foundation of China under Grant Nos. NSFC42130301 and NSFC42250201.

The authors have no conflicts to disclose.

Hydroacoustic data are available from the CTBTO International Data Centre, Vienna, through the virtual Data Exploitation Centre (vDEC, available at https://www.ctbto.org/specials/vdec/). The BELLHOP beam tracing program (Porter, 2011; available at http://oalib.hlsresearch.com/AcousticsToolbox/) is implemented based on the Python package arlpy (available at https://pypi.org/project/arlpy/). The topographical data is from the SRTM15+ (Tozer , 2019; version 2.1, available online at https://doi.org/10.5069/G92R3PT9). The figures are plotted using the generic mapping tools (Wessel , 2013) and ObsPy (Beyreuther , 2010).