Cultural turnover among Galápagos sperm whales

While populations may wax and wane, it is rare for an entire population to be replaced by a completely different set of individuals. We document the large-scale relocation of cultural groups of sperm whale off the Galápagos Islands, in which two sympatric vocal clans were entirely replaced by two different ones. Between 1985 and 1999, whales from two clans (called Regular and Plus-One) defined by cultural dialects in coda vocalizations were repeatedly photo-identified off Galápagos. Their occurrence in the area declined through the 1990s; by 2000, none remained. We reassessed Galápagos sperm whales in 2013–2014, identifying 463 new females. However, re-sighting rates were low, with no matches with the Galápagos 1985–1999 population, suggesting an eastward shift to coastal areas. Their vocal repertoires matched those of two other clans (called Short and Four-Plus) found across the Pacific but previously rare or absent around Galápagos. The mechanisms behind this cultural turnover may include large-scale environmental regime shifts favouring clan-specific foraging strategies, and a response to heavy whaling in the region involving redistribution of surviving whales into high-quality habitats. The fall and rise of sperm whale cultures off Galápagos reflect the structuring of the Pacific population into large, enduring clans with dynamic ranges. Long-lasting clan membership illustrates how culture can be bound up in the structure and dynamics of animal populations and so how tracking cultural traits can reveal large-scale population shifts.


Introduction
Behavioural repertoires change over time. Changes may result from adaptive genetic evolution and genetic drift, phenotypic plasticity, and individual or social learning (e.g. [1,2]). Changes can take many generations (typical for genetic evolution) or occur over a small part of the life cycle (e.g. phenotypic plasticity in the face of rapid environmental change). When behaviour is socially learned and shared, hence culture, evolutionary processes 2016 The Authors. Published by the Royal Society under the terms of the Creative Commons Attribution License http://creativecommons.org/licenses/by/4.0/, which permits unrestricted use, provided the original author and source are credited.

Field methods, photo-identification and acoustic recordings
Sperm whales were tracked visually and acoustically in deep waters (more than 1000 m) across the Tropical Pacific, day and night during two-to four-week surveys between 1985 and 2014 (figure 1; electronic supplementary material, tables S1-S3). Given the logistical challenges of offshore surveys, sampling was unevenly distributed; the Galápagos archipelago was the main study area (electronic supplementary material, tables S1-S3). Annual encounter rates off the Galápagos were calculated as number of groups of female and immature whales encountered divided by total hours of acoustic and visual search (i.e. total effort minus time following whales) [22]. Individuals were identified from photographs based on patterns of natural marks on the trailing edge of their tails, assisted by a semi-automated photo-identification protocol [23]. We rated each photograph from poor to very high quality (Q = 1-5) based on focus, exposure, orientation, per cent cover and tilt of the fluke [24]. Distinctively small animals (of about less than 2 years) were considered calves; distinctively large animals as mature males; the others were considered females and immatures [24]. We analysed only Q ≥ 3 photographs of females and immatures. From a total of 14 286 photographs, we identified 4468 individuals (electronic supplementary material, table S2).
Sperm whale codas were recorded using various hydrophone arrays and recording devices over the duration of the study (electronic supplementary material, method S1) [14]. All recordings were analysed using RAINBOW CLICK software [25] in which individual clicks were manually marked and designated as part of codas. From acoustic recordings from across the Pacific, we sampled 17 045 codas (electronic supplementary material, table S3).

Assigning coda repertoires to photo-identified groups
We assigned coda recordings to groups of individual whales photo-identified together. We considered all recordings made on the same day when continuously following a group of sperm whales to be of the same group [14]. Codas recorded on two different days were considered to be from the same group if at least 25% of the photo-identified individuals were re-sighted [26]: m ab > 0.25 × min{n a , n b }, where m ab is the number of individuals photo-identified on both days, n a is the number of individuals identified on the first day, and n b on the second day. We discarded groups whose recorded repertoires contained less than 25 codas [14].
To account for any potential autocorrelation in coda production during the same day, all coda recordings on a given day from a given group represented a single repertoire. Under the assumption that coda production of a given group on a given day is independent of its production on a subsequent day, repertoires from different days were treated as replicates of a group's repertoire and were considered independent samples of a group's coda production [27]. We used permutations to test differences between group repertoires (electronic supplementary material, method S2).

Continuous and categorical similarity between coda repertoires
We compared group repertoires using the absolute inter-click intervals (ICIs, i.e. the time between the onset of one click to another in a coda sequence) to represent the temporal structure (rhythm and tempo) of their codas [27,28]. To quantify similarity between coda repertoires, we applied continuous and categorical metrics to this multivariate dataset. The former was used to define the vocal clan partitions; the latter was used to define coda types that illustrate the differences in the patterning theme of the codas between clans (analyses pathway: electronic supplementary material, figure S1).
For the continuous approach, we calculated the multivariate similarity of two codas of the same click length (i.e. same number of clicks) using the Euclidean distance between their ICI vectors (electronic supplementary material, method S3) [14]. With the categorical approach, we classified codas into discrete types based on their rhythm and tempo using OPTICSxi hierarchical clustering [29] (in this context: [27]). We ran OPTICSxi on the absolute ICI independently for each set of codas of the same click length, performing a sensitivity analysis a priori to define the algorithm initial parameters (electronic supplementary material, method S4). We labelled the coda types according to number of clicks and rhythm, based on previous nomenclature [14,26].

Assigning photo-identified groups to vocal clans
The original partition of vocal repertoires into clans [14] (electronic supplementary material, method S5) used hierarchical clustering analyses based on the continuous multivariate similarities of standardized ICIs of codas, and the k-means algorithm to categorize codas into types [14]. Here, we used the updated methods for comparing repertoires described above to re-analyse this data set together with the repertoires recorded off Galápagos in 2013 and 2014 (electronic supplementary material, table S3). To assign the 2013 and 2014 groups to clans, we first built an average-linkage clustering dendrogram using the continuous multivariate similarity matrix for the combined dataset; then we identified whether these groups clustered together into a distinctive branch (indicating a new clan) or whether they clustered with previous clans. We measured the accuracy of the dendrogram representation using the cophenetic correlation coefficient (CCC), and considered CCC > 0.8 to indicate a reliable representation. The dendrogram robustness was measured by bootstrap resampling [14]. All groups' coda repertoires were randomly sampled with replacement (100 replicates), their similarities were recalculated and the proportion of times a given branch was replicated used to indicate the robustness of that branch.

Acoustic repertoires
From 1985 to 1999, coda repertoires of 64 groups of female and immature were recorded across the Tropical Pacific [14]; in 2013-2014, we recorded 15 new groups off the Galápagos (electronic supplementary material, table S3). Mantel tests confirmed that repertoire similarity between acoustic recordings from the same group was greater than between different groups indicating that groups had significantly different repertoires (Galápagos 2013-2014: r = 0.413, p < 0.001; Pacific 1985-1999: r = 0.170, p < 0.001; combined: r = 0.176, p < 0.001).
The categorical analysis of the full dataset identified 27 distinct coda types containing from 3 to 12 clicks varying in rhythm and tempo (electronic supplementary material, figure S2). Although the OPTICSxi algorithm classified only the most stereotyped codas (4091/17 045 codas; 24%), discovery curves of classified coda types were nearly asymptotic (electronic supplementary material, figure S3), suggesting that most coda types made by the sampled groups were represented. Coda types were robust  . Asterisks indicate years with no dedicated surveys off Galápagos (the larger gap in the 2000s was mainly motivated by a lack of opportunistic sightings in the area), but in some of these years there were surveys in surrounding areas (electronic supplementary material, tables S1-S3).
to variation across the OPTICSxi input parameter space (electronic supplementary material, figure S4). Coda type classification described the thematic differences in coda patterning driving clan partitioning, defined by the continuous analysis including all codas (figure 3).

Clan structure
The original partitioning of clans in the Tropical Pacific [14] was preserved in our analysis, with Regular, Four-Plus and Short clans depicted in our dendrogram as largely similar to the original analysis (figure 3a). There were some minor changes: the groups recorded off Tonga and in the western Caribbean (branches with lower bootstrap support in the original analysis) clustered with groups belonging to the Plus-One clan; and four groups (two designated as Short, and two as Regular) clustered with different clan branches (figure 3a). We expected some changes because our new analysis was different from the original [14] in two ways. First, we included the newer groups recorded off Galápagos (electronic supplementary material, figure S5). Second, we used absolute instead of the relative ICI used in the original clan partition, as recent studies [27,28] suggest that tempo, in addition to rhythm, is an important element of coda diversity. Nevertheless, the dendrogram in figure 3 is an appropriate depiction of the coda repertoire similarity among groups of whales (CCC = 0.896), with good support from the bootstrap analysis.

Discussion
Our study demonstrates cultural turnover in the sperm whale dialects off the Galápagos Islands over the last 30 years. We attribute these changes to a turnover in the clans utilizing these waters, as the shift in the acoustic repertoires matches the complete replacement of individual sperm whales off the Galápagos. These findings confirm previous suggestions that clans are stable over time (at least in repertoire, and almost certainly in membership) but dynamic over space [15]. Our long-term analysis indicates that the coda repertoires remain little changed across three decades in the Pacific, highlighting that Pacific sperm whales roam over very wide geographical areas as members of large, long-lasting cultural clans [14,30].
None of the individual sperm whales using the waters off Galápagos in 2013 and 2014 were seen in the area during the previous three decades. Our photo-identification findings increased the number of individual sperm whales catalogued across the Tropical Pacific, but the rate of photographic recaptures off the Galápagos between 2013 and 2014 was low despite our long-term and large-scale sampling. Offshore surveys, however, impose several logistical challenges making our sampling effort patchy in time and space (electronic supplementary material, table S1). We acknowledge the consequent uncertainty regarding presence of whales in waters near the Galápagos, as well as in unsampled years. Yet, our photo-identification data provide strong evidence for large-scale movements between discrete study areas and across years. We propose, therefore, that the drastic demographic change we report was driven by emigration of groups of whales from different clans.
There are four lines of evidence that support emigration out of Galápagos, rather than changes in the composition of the clans themselves, as the most likely mechanism for the local decline in sperm whale sightings. First, there were several re-identifications of Galápagos groups and clans off northern Chile and the Gulf of California (figure 1a, see also [20,31]), evidencing that sperm whales do move long distances. Second, sperm whales seem not change their clan membership, or if so, only very rarely [24]. Third, sperm whales are slow-reproducing, long-living animals [24] and the last three decades is a relatively short window in their lifespan during which no high mortality was evident [22]; all of these make death and birth very unlikely to be the drivers of the replacement of individuals off Galápagos. Finally, errors in individual identification cannot be a major factor because marks used to photo-identify animals rarely change [32]; indeed, Atlantic sperm whale individuals have been re-identified across 30 years within a single study area much smaller than ours [12]. Combining these facts with our findings on coda repertoires, we suggest that the same clans from three decades ago still populate the Tropical Pacific, but what seems to have occurred is a large shift in the habitat used by each of the clans.
Sperm whales are nomadic. In the Pacific, social units have wide ranges, performing long-distance movements (mostly spanning about 2000 km, some over 4000 km) within relatively short temporal scales [20], emphasizing the magnitude of the spatial scale relevant for sperm whales. The long-distance movements are made by individuals travelling together, because sperm whales live in nearly permanent social units [11]. These units belong to large clans with dynamic ranges [10,15], which are stable emergent social structures [30] within which coda usage is conserved over time [15]. Therefore, the radical cultural turnover in sperm whale dialects off Galápagos reflected a clan replacement, i.e. a local turnover in whales using the area as a consequence of their natural movements over large spatio-temporal scales. This contrasts with the cultural revolutions among humpback whales where songs changed dramatically but with little turnover of individuals [4][5][6].
The sperm whales recently identified off Galápagos are not members from the Regular and Plus-One clans once common in the area. Instead, they are members of two existing clans (Four-Plus and Short) previously heard across the Pacific but very rare or absent in Galápagos waters. The Four-Plus clan was consistently heard off northern Chile, while the Short clan spread over the Tropical Pacific and only few of its members (a single social unit) had been previously identified off Galápagos [14]. The new whales immigrated from neighbouring waters in the wider Pacific. Our photo-identification data may give some indication of origin: there were some matches with Four-Plus clan members seen previously in the Gulf of California, where both Short and Four-Plus clans may be present [33].

Why were the clans replaced?
The Galápagos Islands, and more broadly the eastern Pacific, were historically important grounds for sperm whales [34]. Although there were numerous whales when our Galápagos studies started in 1985, emigration drastically reduced their numbers between 1990 and 2000. Following this exodus, members of different clans have been slowly repopulating the Galápagos. The fall and rise of sperm whale clans off Galápagos lead to two questions. Why did members of the original clans leave? Why are the new whales from other clans rather than return of the original clans? While our data show a clear shift in Galápagos sperm whale dialects, the underlying mechanisms for the large-scale displacement of clans are necessarily speculative. In what follows, we describe two non-exclusive hypotheses.
The first scenario involves large-scale environmental shifts. Drastic environmental changes force cultural groups to adapt their strategies or move (e.g. [8]). Like many other predators as well as some herbivores, food availability is a major driver of movement for sperm whales. They tend to go where the prey is, moving from areas of low to high feeding success [35]. The abundance of preferred prey, for instance jumbo squid (Dosidicus gigas [21]), may fluctuate naturally across the Pacific and in response to environmental changes such as the El Niño Southern Oscillation (ENSO). ENSO events impose massive changes in the Pacific [36], including anomalous sea surface warming, large influx of deep warm waters, and fluctuations in primary productivity and nutrient cycling [36,37]. In particular, the extreme ENSO events in the early 1980s and late 1990s represented remarkable warming in the equatorial Pacific, devastating marine fauna [36,38], including marine communities in the Galápagos region [39]. The reduced productivity of tropical and equatorial Pacific waters considerably decreased the feeding success of sperm whales off Galápagos [40]. ENSO events are becoming more frequent and intense [36,41]; due to cetaceans' high and adaptive mobility, leaving affected areas is their immediate response [42]. We know sperm whales from different clans tend to move and forage differently [17]. In years of normal temperatures, the foraging strategy of the Regular clan outperforms the Plus-One; whereas in the warmer, less productive ENSO years the foraging successes of both clans is reduced considerably but the Plus-One's strategy becomes more efficient than the Regular's [17]. Clans may conserve their foraging strategies even during remarkable environmental changes [17], thus living in this large-scale dynamic habitat, groups of whales from particular clans may relocate, moving to areas where their foraging strategies are likely to maximize their food intake. This assumed cultural inertia of foraging strategy-not uncommon in marine mammals [43,44]-implies that large-scale movement is favoured over remaining in a changing habitat and adapting to the new conditions. This may explain both why the original clans left and why the new immigrants are from different clans, but implies that changes to the ecosystem around Galápagos [37][38][39] are perceived differently by sperm whales from different clans.
The second scenario involves lagged responses to the population decline caused by modern whaling [22]. Sperm whales in the general vicinity of the Galápagos were heavily hit by nearly unregulated, as well as pirate, whaling between 1957 and 1981 [45,46]. The extreme depletion of sperm whales of the eastern Pacific in those years focused on the relatively inshore waters of the Humboldt Current off Peru and Chile as the legal whaling used catcher-boats operating from mainland ports [34,46]. The whaling may have left a surplus of sperm whale prey, re-opening a niche in the rich Humboldt Current waters. In the case of density-dependent habitat selection [47], whales would redistribute themselves according to habitat quality. Therefore, low whale density in productive coastal waters may have stimulated the eastward migration out of the Galápagos in the 1990s [22]. If the population slowly recovers, it would redistribute to first occupy high-quality coastal waters then adjacent areas [47], which may explain the modest and recent return to Galápagos waters documented here. In this scenario, the turnover of clans off the Galápagos would result from a general eastward movement: first of the Regular and Plus-One clans from the Galápagos to more coastal waters, and then of the Four-Plus and Short clans from oceanic and northern waters to the Galápagos. The underlying assumption is that different cultural foraging strategies characteristic of each clan [17] perform similarly in different areas.
In both scenarios, the turnover of clans using the Galápagos indicates that there may be social dynamics driving movement decisions. Group displacement implies a compromise between individual decision and group conformity (e.g. [48]). Thus, once some members of one clan have decided to leave a particular habitat, other members may choose to move with clan-mates rather than remain within that habitat-a within-clan gregariousness that may be mediated by specific codas that identify clan membership [27]. As associations with familiar conspecifics can facilitate acclimation to novel habitat [49], the benefits of foraging and associating with behaviourally similar clan members may outweigh the cost of displacement to a new habitat. This assumes that clan membership is important for the success of the individuals and social units that comprise them, which fits well with recent evidence that sperm whale movement decisions are shared [50] and that individuals conform to the predominant behaviour of clan members [30]. Overall, these findings show that tracking cultural traits can reveal large-scale population shifts, which further illustrates the key role culture can play in the structure and dynamics of animal populations and their communication systems.

Conclusion
Learned communication repertoires can be either stable across or change within generations in response to cultural selection and drift [51][52][53]; yet a population's repertoire is rarely completely replaced. We found an influx of immigrants from different cultural clans replacing those that used to be in the area decades ago. This local cultural turnover was an epiphenomenon of large-scale displacement of sperm whales organized by vocal clan, suggesting that clan structure is temporally stable but spatially flexible. The changes in sperm whale acoustic repertoires off Galápagos are clear, but the ultimate causes of this cultural turnover remain unclear. Unravelling the drivers of large-scale relocation of cultural groups will allow us to better understand animals' response to the changing ocean, the dynamics of depressed populations and the importance of culture in animal societies. Ethics. Appropriate ethical approval and licences were obtained. Field protocols for approaching, photographing and recording sperm whales were approved by the Committee on Laboratory Animals of Dalhousie University. Permits to carry out research in Ecuadorean waters were issued by Parque Nacional Galápagos (PC-10-13; PC-12-14), Ministerio de la Defensa Nacional (005-2013; 002-14) and Ministerio de Ambiente Ecuador (IC-FAU-DNB/MA002-12; 009-13).