Whistle variation in Mediterranean common bottlenose dolphin: The role of geographical, anthropogenic, social, and behavioral factors

Abstract The studies on the variation of acoustic communication in different species have provided insight that genetics, geographic isolation, and adaptation to ecological and social conditions play important roles in the variability of acoustic signals. The dolphin whistles are communication signals that can vary significantly among and within populations. Although it is known that they are influenced by different environmental and social variables, the factors influencing the variation between populations have received scant attention. In the present study, we investigated the factors associated with the acoustic variability in the whistles of common bottlenose dolphin (Tursiops truncatus), inhabiting two Mediterranean areas (Sardinia and Croatia). We explored which factors, among (a) geographical isolation of populations, (b) different environments in terms of noise and boat presence, and (c) social factors (including group size, behavior, and presence of calves), were associated with whistle characteristics. We first applied a principal component analysis to reduce the number of collinear whistle frequency and temporal characteristics and then generalized linear mixed models on the first two principal components. The study revealed that both geographic distance/isolation and local environment are associated with whistle variations between localities. The prominent differences in the acoustic environments between the two areas, which contributed to the acoustic variability in the first principal component (PC1), were found. The calf's presence and foraging and social behavior were also found to be associated with dolphin whistle variation. The second principal component (PC2) was associated only with locality and group size, showing that longer and more complex tonal sound may facilitate individual recognition and cohesion in social groups. Thus, both social and behavioral context influenced significantly the structure of whistles, and they should be considered when investigating acoustic variability among distant dolphin populations to avoid confounding factors.

The acoustic variability (measured through changes in whistle frequencies and time characteristics) has been already described among Tursiops truncatus (hereafter referred to as bottlenose dolphin) populations of the western (Sicily and Sardinia, Italy) and eastern (Croatia) Mediterranean (La Manna et al., 2017). In the present work, the factors associated with bottlenose dolphin acoustic variability between two populations, located (2000 km apart) in Sardinia Island (western Mediterranean Sea) and Croatia (northern Adriatic Sea), were identified. To achieve this aim, the influence of anthropogenic conditions (noise levels and boat presence) and dolphin socio-behavioral context (dolphin behavior, group size, and the occurrence of calves within the group) on the whistle structure was considered separately. The identification of the factors associated with the acoustic variability between the two populations will draw the attention of researchers on what kind of features dominate the whistle structure of bottlenose dolphin-whether noise and boat presence can affect their communication more than the natural conditions. The results may contribute to providing relevant evidence for managing human activities in bottlenose dolphin areas.

| Study areas
The Sardinia and Croatia bottlenose dolphin (Figure 1 where at least 50% of them show a high level of site fidelity (La Manna & Ronchetti, 2018). Although the majority of these animals were sighted repeatedly every year and in different seasons, the population seems neither closed nor isolated. In fact, between 2012 and 2018, the discovery curve of the photo-identified dolphins never reached a plateau because of the regular entrance of new individuals (La Manna & Ronchetti, 2018).
The acoustic recordings in Croatia were undertaken in an area of about 2000 km 2 , that extends beyond the currently known home range of the population, around the islands of Cres and Lošinj ( Figure 2). These waters are characterized by numerous uninhabited small islands and islets, steep rocky shores, muddy sea bottoms, limestone reefs, and sea depths that do not exceed 90 m of depth. The islands attract a large number of tourists each year, particularly during summer when many of them reach these sites by leisure boats (Rako-Gospić et al., 2013;Rako-Gospić & Picciulin, 2016). This region is part of the home range of about 200 common bottlenose dolphins. Based on a photo-identification study, 184 individuals (95% CI = 152-250; CV = 0.17) were estimated by means of mark-recapture methods and the Mth estimator of Chao for closed populations (Pleslić et al., 2015;Rako-Gospić et al., 2017). The high sighting frequency and regular resightings of known individuals from year to year indicate their long-term fidelity to this specific region. Because of its importance as a habitat for this resident bottlenose dolphin population, the Cres-Lošinj area was designated as a site of community importance (SCI), part of the European Union NATURA 2000 ecological network (Cres and Lošinj SCI, HR3000161), in December 2014.

| Field methodology
The data were collected during dedicated boat surveys in spring and summer between 2015 and 2018; overall, 36 and 52 surveys were done in Croatia and Sardinia, respectively. The daily surveys were conducted using a 5.8 m RIB powered by a four-stroke 90 HP outboard engine in Croatia and a 9.7 m motor boat powered by a 270 HP inboard engine in Sardinia. With the aim to homogeneously cover the study area, routes were designed with a generally perpendicular direction with respect to the coast and depth contours. The experienced observers scanned the sea surface during daylight, in good sea conditions (Douglas sea state < 2 and Beaufort wind force < 2), with a visibility of over 3 nautical miles, and a boat speed between 10 and 30 km/h. The navigation routes were interrupted in case of sighting or when sea conditions deteriorated. A dolphin sighting was defined as an observation of one dolphin or a group of dolphins (defined as all individuals within visual range that were in apparent association, engaged in the same activity or moving in the same direction; Shane, 1990). During each sighting, data on dolphin group size and age class were recorded by two independent observers. The calves were defined as dolphins of no more than two-thirds the length of an adult (Shane, 1990). The individuals not belonging to the calf class were defined as adults/subadults. The adults were fully grown individuals, generally 2.5-3 m long, while subadults were not yet fully grown but were larger than the calves and did not travel in the typical calf position alongside an adult individual. The identity of individuals in the group was determined using standard photo-identification techniques (Würsig & Jefferson, 1990), using a Nikon D7000 and a Sony alpha 65 cameras equipped with 70-300 and 18-250 mm lenses in Sardinia and a Canon EOS 6D equipped with a 70-200 mm lens in Croatia. Each image was evaluated for photographic quality and grade of distinctiveness of the fin following Ingram (2000). Then, two researchers independently matched the photographic datasets to avoid misidentification (Pleslić et al., 2015;Pulcini, Pace, Manna, Triossi, & Fortuna, 2013). F I G U R E 1 The bottlenose dolphin (Tursiops truncatus) The behavior and acoustic data were collected 20 min following the first approach, allowing the animals to habituate to the research boat presence. The surface behavioral state was recorded by following continuous focal group sampling (Altmann, 1974;Mann, 2000). The observations lasted between 20 and 60 min. The behavioral states assigned included one of the following mutually exclusive categories (Lusseau, 2003;Shane, 1990): (a) foraging (animals usually dispersed, frequent direction changes, dive intervals longer than 3 min, fish chases at the surface, birds often in attendance); (b) traveling (consistent directional movement of dolphins, with regular surfacing, often much splashing); (c) socializing (interactive events observed, such as body contacts, pouncing and hitting with tail, chases, aerial events, no directed movements, and variable dive intervals); (d) milling (no net movement, individuals surface facing different directions, dive intervals variable but short); and (e) resting (slow movement of dolphins, no splashing, closely associated, short and synchronous dive intervals). If members of the group displayed more than one category, the predominant one (performed by more than 50% of the group members) was recorded (Mann, 1999). To avoid potential bias related to group composition and assigning an incorrect behavioral state to the recorded acoustic behaviors, the data collected when animals in the focal groups performed different behavior were not considered in the analysis. Concurrently to the collection of surface behavior data, the acoustic recordings were gained by means of a hydrophone. In Sardinia, a Sensor Technology SQ26-08 omnidirectional hydrophone (sensitivity −168.8 dB re 1 V/µPa; flat frequency response from 100 Hz to 30 kHz, ±3 dB), with a bandwidth between 20 Hz and 50 kHz, was lowered to a 5-10m depth and connected to an M-Audio MicroTrack II recorder or a ZOOM recorder (data format 24 bit WAV, sampling rate 96 kHz). Before each recording, the recording system was calibrated by applying a signal of 100 mV RMS at 2 kHz to the transducer input of the system by means of a signal generator.
In Croatia, a RESON TC 4,032 omnidirectional hydrophone (sensitivity − 170 dB re 1 V/µPa; flat frequency response from 10 Hz to 80 kHz, ± 2.5 dB), with a bandwidth between 5 Hz and 120 kHz, was lowered at approximately 5 m of depth and connected to a SOUNDDEVICES 702 high-resolution digital audio recorder (data format 24-bit WAV, sampling rate 192kHz), calibrated with a signal of 100 mV RMS at 2 kHz. In order to minimize mechanical noise, such as flow and cable strumming noise, which may be induced when low-frequency ambient noise is measured by cabled hydrophone, we recorded only in good sea and wind condition (Douglas sea state < 2 and Beaufort wind force < 2), in absence of wave motion. Furthermore, the engine and the instruments on board were switched off and the boat was still during all the recordings.
The behavioral categories have been assigned to each whistle according to surface behavior data collected as described above. To ensure the correct behavioral category assignment to each whistle, data collected in the presence of more than one group of dolphins within the visual range of the observers were discarded. Only whistles with the highest signal to noise ratio (strongly suggesting a close proximity of the recorded individuals) were analyzed (Heiler et al., 2016;Marley et al., 2017). During the behavioral and acoustic data collection, the research boat remained between 20 and 100 m away from the dolphins with the engine off. The number and type of boats present (except for those stationary with the engine off) within 500 m of the focal group were also recorded. The maximum distance of 500 m was chosen because it should be the approximate distance at which a boat noise is considerably higher than the background noise (following Sara ' et al., 2007), although this noise may change relevantly depending on the type and speed of the boat and the environmental conditions.
This strict definition is useful to distinguishe whistles from other narrow-band sounds produced by bottlenose dolphins (Simard et al., 2011;van der Woude, 2009). The whistles are distinguished as signature and nonsignature whistles (Caldwell, Caldwell, & Tyack, 1990;Janik, Dehnhardt, Todt, 1994;Cook et al., 2004;. The nonsignature whistles, also known as variant whistles, are not characterized by individually distinctive frequency-modulated patterns (Caldwell & Caldwell, 1979;Watwood et al., 2004;King et al., 2013), and their function remained poorly understood (Janik & Slater, 2000;Tyack, 1986). A signature whistle is defined as "a learned, individually distinctive whistle type in a dolphin's repertoire that broadcasts the identity of the whistle owner" . Thus, signature whistles are characterized by the same frequency modulation pattern (called contour), at frequencies ranging between 1 and 38 kHz (Boisseau, 2005;Hiley, Perry, Hartley, & King, 2016;May-Collado & Wartzok, 2008), with a duration between 0.1 and 4.0 s (Buckstaff, 2004). The signature whistles can be produced in loops (repetitions of the same elements), usually separated by intervals less than 250 ms (Esch, Sayigh, Blum, & Wells, 2009), and can also have an introductory and/or final loop  distinct from the central pattern. We considered any single-or multiple-loop whistle, connected or disconnected, as a unit of analysis (Esch et al., 2009).
Each whistle, recognized as a signature whistle by the observer following the SIGnature IDentification method (SIGID; see , was considered just once in the analysis. This rule was applied to reduce the risk of collecting whistles from the same in- (b) score 2 (whistle clearly visible from its start to its end); and (c) score 3 (prominent and dominant whistle). Only whistles scoring 2 or 3 were analyzed further. The duration; minimum, maximum, start, and end frequencies; frequency range; and the number of inflection points (Table 1; La Manna et al., , 2017Papale et al., 2013)  A selection of 5 s, immediately before each whistle of good quality (score 2 or 3), was analyzed by means of PAMGuide. The PAMGuide is a template code provided in R (Merchant et al., 2015), able to perform the signal processing steps required for the calibration procedure to obtain absolute sea ambient noise (SAN) levels.
For the purposes of this study, SAN is defined as "all sound (both natural and anthropogenic) except that resulting from the deployment or recovery of the recording equipment" (Robinson, Lepper, & Hazelwood, 2014). The choice of 5 s was a compromise between the supposed minimum time required to adjust a vocal emission to background noise level in mammals (2 s; Gillam, Ulanovsky, & McCracken, 2007;Hase, Miyamoto, Kobayasi, & Hiryu, 2016)

| Statistical analyses
The descriptive statistics (mean, standard deviation, CV = coefficient of variation and range) about the acoustic characteristics of all whistles in the dataset were used to describe the intra-and interpopulation variability. Before any analysis, data exploration was carried out following Zuur, Ieno, Walker, Saveliev, and Smith (2009).
We were interested in disentangling which variables influenced the geographic variation of whistles between Sardinia and Croatia.
Thus, we preliminarily tested if the sea ambient noise levels and group size were associated with the factor locality, to explore ecological and acoustical environment differences. We applied three generalized linear models (GLM) using a negative binomial distribution (to account for overdispersion), where SAN levels (in the three octave bands, 125 Hz, 2000 Hz, and 20 khz) were the dependent variables and locality was the predictive variable. The model diagnostics in R was used to verify the appropriateness and assumptions for each model, and the analysis of deviance (MASS package in R) was used to assess the significance of the predictors (Venables & Ripley, 2002).
Then, a principal component analysis (PCA) was used to reduce the seven acoustic whistle characteristics into two independent variables, after the assumptions (linear relation between variables, sampling adequacy, and presence of outliers) had been verified. Only the first two components (that together explained 71% of the total variance, see the Results Section) were retained because eigenvalues for the remaining five components were all < 1 (Kaiser's criterion). To perform PCA, the function prcomp of the R package Rstats (R Core Team, 2015) was used.
The association between the two PCs and the geographical (locality), anthropogenic (noise levels in the three bands and boat presence), social (group size and calf presence), and behavioral variables was tested using a generalized linear mixed model (GLMM-following Jansen, Plath, Brusquetti, & Ryan, 2016;Lee, Shaner, Lin, & Lin, 2016) with a gaussian distribution. The GLMMs are an extension of generalized linear models that allow for the inclusion of random effects, by modeling the covariance structure that is generated by the grouping of data (Zuur et al., 2009). They are very useful when the data are not independent, for example, when a variable is measured more than once from the same individuals. Because the whistles recorded within the same group are likely to be more related to each other than the whistles recorded

Parameter
Unit Description

Start frequency Hz
The frequency measurement at the start of the whistle.

End frequency Hz
The frequency measurement at the end of the whistle.

Min frequency Hz
The lower frequency limit of the selection box.

Max frequency Hz
The upper frequency limit of the selection box.
Frequency range Hz Total bandwidth, calculated by max frequency minus min frequency.

Duration
Sec Total duration, calculated by end time minus start time.

Number of inflection points -
The number of inflection points defined as the change from positive to negative or negative to positive slope in the contour. and calf presence and behavioral states as factors). Since resting and milling had a negligible sample size, the only behavioral states included in the models were feeding, traveling, and socializing. Because one of the main aims of the analysis was to assess the effect of locality, in both models we included the interaction between locality and the fixed terms. We followed a forward selection procedure to select the best models, based on the Akaike's information criterion (AIC) and likelihood ratio tests (Zuur et al., 2009

| Ethical note
The study was entirely observational, and no special permit was necessary. Special care was taken when approaching the animals to reduce any disturbance and alteration of the natural behavior.

| RE SULTS
Based on the photo-identification data, we identified 72 individuals in Sardinia and 256 individuals in Croatia, differently associated within the groups. Particularly, in Sardinia 30% of the individuals were sighted only once, 44% from 2 to 6 times, 15% from 7 to 11 times, and 11% from 12 to 17 times, while in Croatia 48% of the individuals were sighted only once, 49% from 2 to 6 times, and 3% from 7 to 8 times. The group size in Sardinia ranged between 2 and 16 animals, with a mean of 7.6 (SD ± 3.2) and a median of 8, while in Croatia it ranged between 2 and 46, with a mean of 22.1 (SD ± 12.3) and a median of 23. GLM associated group size to locality (Table 4). In Sardinia, 65% of whistles were recorded in absence of calves; 53% of them were recorded while dolphins were socializing, 30% while they were traveling, and 17% while they were foraging. In Croatia, 87% of whistles were recorded in presence of the calves; 65% of them were recorded while dolphins were traveling, 20% while they were foraging, and 15% while they were socializing. Among all the whistles, 56% and 66% were recorded in presence of boats, in Sardinia and Croatia, respectively.

| Acoustic recordings
In Sardinia, a total of 27 hr 49 min of recordings were collected over 60 days, from which we extracted a total of 1980 whistles belonging to 60 different groups. In Croatia, a total of 28 hr 29 min of recordings were collected over 68 days, during which 89 dolphin groups were sighted, producing in total 927 whistles. From the whole dataset, stereotyped whistles were considered only once, and only whistles scored as 2 or 3 were analyzed further, reducing the initial database as follows: 456 whistles from 52 groups and 361 whistles from 36 groups in Sardinia and Croatia, respectively, with a general mean values of whistles higher in Croatia. Moreover, CVs were higher for duration and number of inflection points across all areas, and higher for all characteristics in Sardinia compared to Croatia, with the exception of number of inflection points and frequency range ( Table 2).
The SAN levels were higher in Sardinia compared to Croatia, both in the 125 Hz and 20 kHz bands (Table 3). Accordingly, GLM results showed that SAN levels were associated with the factor locality, both in the 125 Hz and in the 20 kHz bands, while no influence was found on the SAN level in the 2 kHz band (Table 4). Note  Table 5). We used GLMMs on PC1 and PC2 to investigate the effect of SAN and boat presence and socio-behavioral variables with separated models. On the basis of the graphic validation of the models, no problems were found (Appendix S1: ES1, ES2, ES3, and ES4).

| Influence of noise and boat presence on whistle structure
The PC1 was significantly associated with the interaction between SPL in the 125 Hz band and locality and with the interaction between boat presence and locality (Table 6). In particular, PC1 decreased with the increasing noise in the 125 Hz band mainly in Croatia and decreased in the presence of boats in Sardinia (Figure 4). Thus, because the PC1 and min, max, start, and end frequencies were negatively correlated, the latter characteristics tend to increase with the increasing noise in Croatia and in presence of boats in Sardinia. The PC2 correlated only to locality (Table 7) and was higher in Sardinia compared to Croatia ( Figure 5). In particular, duration and frequency range were lower in Sardinia.

| Influence of socio-behavioral variables on whistle structure
The socio-behavioral variables, such as behavior, group size, and calf presence, have also had important influences on bottlenose dolphins' whistle. This was highlighted by the association of PC1 with behavioral state and with the interaction between calf presence and locality (Table 8). In particular, PC1: (a) in the presence of calves, decreased in Croatia while increased in Sardinia and (b) increased during foraging and socializing compared to travel ( Figure 6). Thus, min, max, start, and end frequencies (the negatively correlated characteristics to PC1) in Croatia tend to increase in the presence of calves, while they tend to decrease in Sardinia and were lower during foraging and socializing compared to travel. The PC2 was associated only with group size (Table 9) and decreased with the increasing group size (Figure 7). Thus duration, number of inflection points, and frequency range tend to increase, as the group size increases in both localities.

| Geographic variability
The variability in acoustic communication among dolphin populations can be affected by geographic distance and isolation, genetic differentiation, adaptation to the local acoustic and social environment, or the concomitant action of these different factors (Ey & Fischer, 2009;Parks et al., 2009;Sun et al., 2013;Wilczynski & Ryan, 1999). Based on the GLMM models run with the geographical and anthropogenic variables (noise and boat presence), only PC2 was influenced solely by locality. Thus, duration, number of inflection points, and frequency range would seem to be characteristics of the whistles influenced mainly by the specific population. This result is consistent with the fact that duration and inflection points may reflect information about the identity of the individuals (Steiner, 1981).
However, this result must be taken with caution because other factors, which were not considered in the present study, could have affected the whistle characteristics expressed as PC2. Moreover, as we have recorded some of the photo-identified individuals several F I G U R E 4 Interaction effect between (a) locality and SPL 125 Hz (dB re 1 µPa) and (b) locality and boat on the spectral property of the whistles (PC1) as predicted by the GLMM (elaborated with the package "nlme" in R) times, in both locality, this could have influenced our finding, despite the analytical methods adopted to contain this aspect.
The geographically distant and genetically distinct populations may exhibit acoustic variability (Funk et al., 2005;Irwin et al., 2008).
Although genetic characterization of the dolphins inhabiting the two study areas is lacking, the acoustic variability here (see also  (Hoffmann et al., 2012). However, where acoustic communication is mediated by learning processes (as in cetaceans), and therefore by cultural transmission (Bain, 1986;Ford, 1991;Janik & Slater, 2000), populations that are geographically but not genetically distant may also manifest acoustic variability (Camargo, Rollo, Giampaoli, & Bellini, 2007;Rendell & Whitehead, 2005;Rossi-Santos & Podos, 2006). Regardless of the mechanism responsible for such variability whether geographic distance and isolation, genetic variability, and/ or cultural drift (Wilkins et al., 2012), our results provide evidence for variability in whistles between the two populations.

| Influence of noise and boat presence on whistle structure
The variation in acoustic communication can manifest itself as an adaptation to local noise and boat presence conditions, even in the absence of geographic isolation, genetic causes, or cultural drift (Ansmann et al., 2007;Baron et al., 2008;van Ginkel et al., 2017;La Manna et al., 2013;Leao et al., 2016;Luís, Couchinho, & Santos, 2014;Morisaka et al., 2005;Papale et al., 2015;Rako-Gospić & Picciulin, 2016). According to the acoustic adaptation hypothesis, the signals emitted by an animal are adapted to the environment in which it lives, to minimize degradation, maximize signal transmission, and ensure long-range communication (Ey & Fischer, 2009).
Moreover, sound transmission in the ocean can change as a function of pressure, temperature, salinity, depth, bathymetric contour, among a range of other factors, and thus varies geographically and over time (Richardson, Greene, Malme, & Thomson, 1995 Note dolphin produces whistles with energy mostly between 2 and 20 kHz (Richardson et al., 1995). However, the lowest frequencies of the whistles produced by the dolphins in both study areas were lower, and dolphins' ability to hear sound at 100 Hz and to produce sounds at 200 Hz is known (Herzing, 1996;Johnson, 1968;Turl, 1993). Even if with our approach we cannot prove causation, other studies found a similar association between low-frequency noise and whistle characteristics. For example, Marley et al. (2017) found that noise in the 1 kHz octave band was more strongly associated with dolphin whistle characteristics than noise in the 16 kHz and 32 kHz octave bands. Moreover, the association between low-frequency noise and the spectral property of the whistles (min, max, start, and end frequencies) found mainly in Croatia is consistent with similar frequency shift related to low-frequency noise in other studies (Fouda et al., 2018;van Ginkel et al., 2017;Luís et al., 2014;Marley et al., 2017;May-Collado & Wartzok, 2008;Rako-Gospić & Picciulin, 2016;Wang et al., 1995). The frequency shift is recognized as one among the several mechanisms that dolphins use to cope with noise, likely aimed at increasing Note: The upper section shows the significant effects of the assessed explanatory variables on PC1. Value, standard errors (SE), t-values, and significance level (p-value) for variables retained in the best model are provided for fixed effects (explanatory variables), while estimates of the standard deviations (SD) are reported for random effects (group). The lower section presents the results of the model selection and significance of dropping the nonsignificant variables from the full model to obtain the best model. ":" = interaction.

F I G U R E 6
Effect of (a) the interaction between locality and calf and (b) behavior on the spectral property of the whistles (PC1) as predicted by the GLMM (elaborated with the package "nlme" in R) transmission efficiency and detectability of their acoustic signals (Rako-Gospić & Picciulin, 2016). Another mechanism that could explain the increased whistle frequencies is the Lombard effect, where the vocal production changes in response to an increase in amplitude of background noise (Brumm & Zollinger, 2011).
Even though in Sardinia average levels of noise (recorded at times that whistles were recorded) were higher than in Croatia, whis-  whistles, and the attempt of the calves to replicate the adult whistles can be related to such difference. These aspects need further investigation to be disentangled. Therefore, this study supports the importance of considering the social and behavioral context of the animals when investigating the acoustic structure of communication signals (Guerra, Dawson, Brough, & Rayment, 2014;Heiler et al., 2016;Marley et al., 2017), and its variability between populations to avoid misleading conclusions. Note: The upper section shows the significant effects of the assessed explanatory variables on PC2. Value, standard errors (SE), t-values, and significance level (p-value) for variables retained in the best model are provided for fixed effects (explanatory variables), while estimates of the standard deviations (SD) are reported for random effects (group). The lower section presents the results of the model selection and significance of dropping the nonsignificant variables from the full model to obtain the best model.

F I G U R E 7
Effect of group size on the temporal property and frequency modulation of the whistles (PC2) as predicted by the GLMM (elaborated with the package "nlme" in R) The results of the study should be interpreted considering some methodological limitations. The approach used does not allow identifying the emitter of the whistle; therefore, the answer found must be generically attributed to the group. Therefore, it is not possible to highlight the individual contribution of dolphins reactions to the presence of boats or to noise levels, and on the other hand, the response of same individuals may be oversampled. Furthermore, only some of the potential factors associated with acoustic variability were considered in the study. Other important sources of variability, either related to characteristics of vessel traffic (in terms of boat speed, behavior, and exact distance to dolphins) or to the environment such as features affecting sound propagation (i.e., water temperature, depth, and type of substrate), have not been considered here and deserve attention in future studies.

| CON CLUS ION
Geographic variation of the bottlenose dolphin whistles has been found in several oceans worldwide (Azevedo et al., 2007;La Manna et al., 2017;May-Collado & Wartzok, 2008;Morisaka et al., 2005;Papale et al., 2014;Wang et al., 1995), including the Mediterranean Sea, but to our knowledge, this is the first attempt to identify the factors associated with such variability. The results provide evidence supporting the hypothesis that the geographically isolated populations of Sardinia and Croatia may have developed their communication, in response to both the acoustic environment and vessel traffic condition in which they live, and to other factors, such as cultural/ genetic drift.
The anthropogenic noise in the marine environment is generally increasing (Hildebrand, 2009), and it can have a wide range of impacts on wildlife, including individual behavior, physiology (Shannon, McKenna, Angeloni, & Wittemyer, 2016), and metabolic cost (Holt, Noren, Dunkin, & Williams, 2015), with unknown consequences in the long term and at the population level. While our results are not conclusive, the variability identified in this study either provides evidence of isolated populations or supports for changes in acoustic behavior in response to differing environments, in which human activities may be among the cause of such difference. Additional research is required to improve our understanding of the causes of acoustic variability in dolphin populations and to improve management of anthropogenic noise and boat traffic in both areas.

ACK N OWLED G M ENTS
We want to thank Progetto Natura, Nauta rcs, and Blue World Institute for the financial and logistic support during this study.
Part of this research was financed by National Foundation for Civil Society, Croatia. We wish to thank all the students and volunteers for their assistance in the fieldwork, particularly Fabio Ronchetti, Francesco Perretti, Katharina Leeb, Enrica Garau, Claudia Pitoni, Tihana Vučur, and Marko Radulović. Special thanks are due to Marco Bonato and Lisa Granziol who provided logistic support for data collection.

CO N FLI C T O F I NTE R E S T
None Declared.