Acoustic similarity of flight calls corresponds with the composition and structure of mixed-species flocks of migrating birds: evidence from a three-dimensional microphone array

Abstract

Animals that migrate in mixed-species groups may communicate with both conspecific and heterospecific individuals, providing a low-cost mechanism for navigation whenever individuals share similar migratory routes or destinations. Many migratory birds produce calls while flying, but the function of these calls, and the forces contributing to their evolution, are poorly known. We studied flight calls in mixed-species groups of wood warblers (Parulidae), a biodiverse group of migratory songbirds. We used a spatial approach to examine whether acoustic similarity of flight calls varies with group composition, recording flight calls of mixed-species flocks with a wireless microphone array and triangulating the positions of birds in three dimensions. We found that the acoustic similarity of flight calls was correlated with spatial proximity: birds with similar calls fly closer together during migration. We also found relationships between acoustic similarity, flock size and mixed-species flock diversity: birds with similar calls fly in smaller flocks and in flocks with lower species diversity. Our results support the idea that migrating birds use flight calls to maintain contact with acoustically similar individuals in mixed-species flocks, with communication transcending species boundaries. These results suggest that acoustically similar flight calls are used as cues of group assembly for migratory animals.

This article is part of the theme issue ‘Mixed-species groups and aggregations: shaping ecological and behavioural patterns and processes’.

1. Introduction

Animals as diverse as whales and songbirds engage in long-distance migrations, and produce complex acoustic signals while they migrate. What are the functions of acoustic signals during migration? Many studies have investigated the ecology and behaviour of animals' migratory movements [1,2], including the effects of weather on migratory behaviour [3,4], but few studies have focused on the patterns and processes of animal communication in mixed-species groups during migration. Long distance migration may increase exposure to predation, require crossing inhospitable landscapes and require extreme energetic demands of long-distance movement [5–8]. Migration is a major source of mortality for migratory animals [9], and acoustic communication during migration may help animals mitigate migration-related mortality.

Migrating animals might use acoustic signals to maintain contact with nearby individuals, which could reduce the probability of disorientation and predation during their migratory journey [10,11]. One way migratory in which birds could maintain contact during migration is by uttering flight calls (short, high-frequency calls given during migration) [12,13]. One group that makes conspicuous use of flight calls are the wood warblers (family: Parulidae; hereafter ‘warblers’), a species-rich family of neotropical songbirds, at least 47 species of which migrate between temperate breeding grounds and south-temperate or tropical breeding grounds [14]. Warblers migrate nocturnally for at least four reasons: atmospheric conditions are more stable and less turbulent at night [15]; cooler night-time air reduces overheating [3]; many avian predators are not active at night; and stellar cues for migration are only visible at night [16]. During migration, warblers frequently utter flight calls that have a relatively high frequency range (6–11 kHz) and a short duration (50–250 ms) [13,17]. These flight calls, which are distinct from both contact calls and alarm calls, exhibit substantial variation in their acoustic structure [12,13,18] and show a strong phylogenetic signal [18,19]. A recent comparative study revealed that flight calls also exhibit acoustic convergence among species with similar migratory journeys, even after accounting for phylogenetic relatedness [19].

Despite growing interest in recording and understanding flight calls [13,17,20], the exact function of these calls remains unknown. Many hypotheses for the function of flight calls have been proposed. For example, some researchers have suggested that flight calls may play a role in stimulating migratory restlessness (zugunruhe) [21]. Many researchers have suggested that flight calls may also be important for maintaining group cohesion [13,22,23]. Some researchers suggest that flight calls may even be important beyond migratory periods [18,24]. How flight calls function to increase group contact, however, and whether flight calls are important in communication between members of the same species, or in cross-species communication, is unknown.

In songbirds that use flight calls, species differences in calling behaviour may have evolved to facilitate interspecific communication between birds migrating together in mixed-species groups. There are at least four potential benefits to maintaining acoustic contact with heterospecific migrants through flight calls: (1) to reduce predation risk by increasing group contact [25]; (2) to provide navigational information that allows naive individuals to stay in contact with experienced individuals [26]; (3) to enhance orientation information about migratory direction [27]; and (4) to share information about the presence of high-quality habitats [20,28]. These benefits are likely to transcend species boundaries and may select for similarity in flight calls between different species, providing that the heterospecific animals share benefits of migrating together in mixed-species groups. The Migration Similarity Hypothesis holds that similarity in migratory journeys (overlap in breeding ranges, migration ranges, wintering ranges and timing of migration) is associated with acoustic similarity in flight calls [19]. Recent research, focusing on range overlap and acoustic similarity, demonstrated that multiple bird species with similar calls follow similar migratory journeys, consistent with this hypothesis [19]. An important, untested prediction of the Migration Similarity Hypothesis is that acoustic similarity in flight calls should influence the behaviour and composition of birds within actively migrating mixed-species flocks; testing this prediction was the motivation for the current study.

In this investigation, we studied actively migrating mixed-species flocks of migratory warblers, using a multi-channel microphone array to triangulate the position of free-flying migratory birds, to better understand the structure and composition of calling animals within mixed-species flocks. We used an eight-element microphone array [29] that exploited delays in sound arrival time at each of the microphones to triangulate the position of birds within actively migrating mixed-species flocks, in three dimensions. We isolated groups of warblers as they passed over the microphone array on the southern shore of Lake Superior. We calculated the time and position of each calling animal within each group, and we compared the spatial features of birds within these flocks to the acoustic similarity of their flight calls. We used these data to test a key prediction of the Migration Similarity Hypothesis [19]: that acoustic similarity of flight calls would influence the behaviour of flockmates in mixed-species flocks, where individuals would fly in closer proximity to flockmates that share acoustically similar calls (figure 1). At the flock level, as a consequence, we predicted that flocks made up of acoustically similar individuals would fly in closer proximity to each other (figure 1). These analyses stand to expand our understanding of interspecific communication in mixed-species groups, particularly the poorly known ecology of actively migrating animals, and to highlight the importance of acoustic signals during animal migration.

Figure 1.

Schematic representation of predictions of the Migration Similarity Hypothesis. We predicted that species with acoustically similar flight calls would fly closer together during active migration (left). We predicted that species with acoustically dissimilar flight calls would fly at greater distances (right).

2. Methods

(a) . Field methods

We collected data on the composition of mixed-species flocks of migratory warblers using an eight-element wireless microphone array [29]. We deployed the microphone array at a study site along the southern shore of Lake Superior in the Keweenaw Peninsula, Michigan, USA (figure 2). Situated along a major international flyway on the Great Lakes, this site regularly features the passage of large flights of warblers and other songbirds at night, and in the morning, during the migration season [30]. We positioned the array directly below a well-known flight path of migrating warblers, with microphones mounted on poles arranged in a square with 25 m sides, and microphones mounted on the top and near the bottom of each of the 7.1 m poles (figure 2). We recorded birds during nighttime and early morning movements. Full details of the microphone array and study site are given in Gayk & Mennill [29].

Figure 2.

(a) Site where mixed-species flocks of wood warblers (Parulidae) were recorded on the Keweenaw Peninsula, Michigan, USA. (b) Photograph of one of the eight microphones comprising a microphone array. (c) Schematic representation of the eight-element microphone array used to triangulate three-dimensional locations of wood warblers in active migration.

(b) . Flight call annotation and triangulation

We followed the laboratory-based methods for triangulating birds described in Gayk & Mennill [29]. This approach relied on recording the same bird's call in multiple spatially separated microphones, and then using the difference of arrival time to calculate the calling bird's estimated position. In a prior study, we established that microphone arrays can accurately triangulate birds with low error, including ground-truthed visual observations of migrating flocks [29]. Such a recording apparatus is capable of triangulating the position of acoustic signals across a variety of frequencies [29], at flight altitudes as great as 150 m [31]. Given that the site of our microphone array sometimes experienced wind and wave noise, we only selected flight calls for triangulation when the start and end of the call could clearly been seen above background noise in at least four microphone channels.

We focused our analyses on flocks of birds, defining a ‘flock’ as a group of triangulated warblers that were separated in time by less than 2.0 s and separated in space by less than 400 m. When we recorded a gap of greater than 2.0 s or a distance of more than 400 m, we considered such birds to belong to separate flocks. For each flock we measured features of the entire flock and features of the individuals within the flock. At the flock level, we quantified the number of birds in each flock, the number of species in each flock, and the species diversity of each flock using the Shannon index (= −∑ p_i ln p_i; for each flock of warblers this represented the summed total proportion represented by each species divided by the total flock size (p_i), multiplied by the natural log of each flock proportion (p_i)) [32]. At the individual level, we calculated the distance, in metres, between each bird within the flock, the time delay between each call for every pair of birds in the flock, the nearest neighbour of each triangulated bird and each bird's altitude. Also at the individual level, we calculated the position of each bird within the flock (i.e. the first, second, third, etc. bird to be detected by the microphone array, reflecting their physical position within the flock) divided by the total number of birds in the group; this allowed us to calculate where each warbler was in the group in relation to the total flock size.

Our microphone array depended on birds calling in order to detect them and triangulate their position. It is possible that some individuals flew over the microphone array without calling, although our field observations suggest that this is uncommon. If some birds passed over the array in silence, our acoustic estimates of flock sizes are underestimates, and our acoustic diversity measurements vary from the actual flock diversity.

(c) . Acoustic analyses

For each species of warbler that we detected, we calculated the acoustic similarity of their flight call to all other species’ flight calls using an established protocol [19]. This analysis relied on collecting 14 acoustic measurements of 36 species of wood warblers (864 flight calls in total), which were used to perform a Principal Components Analysis. We constructed an index of acoustic similarity by plotting the Euclidean distance between the centroids of each species in n-dimensional space (a dimensionless scale). We used this as the measure of acoustic similarity between pairs of triangulated wood warbler species’ flight calls. In all cases, a low score for a pair of warblers represented a high acoustic similarity in their flight calls. For each flock, we also calculated average acoustic similarities for all birds in the flock to each other. This allowed us to calculate one average acoustic similarity value for each flock.

(d) . Statistical analyses

To understand the relationships between acoustic similarity of flight calls and the behaviour of birds and the features of flocks, our analyses relied on the three features of individual birds, and two features of flocks. Individual features consisted of the proximity of each bird to the nearest neighbour (defined as the spatial distance between triangulated birds within each flock in three dimensions), the altitude of birds and the position of each bird within the flock. Flock features consisted of the Shannon Diversity index value of the entire migrating flock (described above; hereafter ‘flock species diversity’) and the flock size. We used these data to test a prediction of the Migration Similarity Hypothesis [19]: that acoustic similarity of flight calls would predict the position of flockmates, where birds would fly in closer proximity to flockmates with acoustically similar calls. As a direct consequence of our first prediction, at the flock level we predicted that flocks made up of acoustically similar individuals would fly in denser formation. We also used these data to test the prediction that acoustic similarity of warbler flight calls varies with the proximity of flockmates, and the flock size, flock species diversity and position of individuals within mixed-species flocks.

To test the prediction that warblers would migrate in closer proximity when flocks were made up of acoustically similar individuals, we calculated distances between each triangulated individual. For these relationships, data did not meet the assumptions for linear regression (assumptions of homoscedasticity; Goldfeld–Quandt test: GQ_58,57 = 104.5, p < 0.001). Six of the acoustic similarity values were outliers (greater than two standard deviations beyond the mean) and were removed from analysis. To account for nonlinearity in these data, we used Generalized Additive Mixed Models (GAMs) with acoustic similarity of each nearest neighbour within groups as a predictor variable of proximity in metres and in time (seconds) in the R package MGCV [33]. We also used GAMs to examine the relationship between proximity, acoustic similarity, flock species diversity, flock size and position of individuals within flocks. We also explored the relationship of the 18 individual wood-warbler species using species as a random effect. This approach allowed us to address whether flight-calling behaviour varies with proximity of warblers to each other within flocks, flock size, flock species diversity, position of individuals within flocks, and whether such calls function primarily in interspecific or intraspecific communication. We graphed the relationships by plotting the average acoustic similarity of warbler flight calls in flocks versus the proximity of birds within flocks in space, the species diversity of flocks, the position of individuals within flocks, and the size of warbler flocks. For each comparison we include a plot with a line of best fit with the 95% confidence interval.

3. Results

We analysed 46 flocks of warblers recorded with an eight-channel microphone array, with an average of 9.2 ± 1.1 individual birds per flock (mean ± s.e.; range of 2–23 birds per flock; see electronic supplementary material, table S1). The individuals within these flocks represented 18 different species of warblers including species belonging to the bioacoustic categories of Upsweep calls, Downsweep calls, and Zeep calls [12,34,35]. Thirty-three of the 46 flocks were mixed-species flocks, and 13 were single-species flocks (single-species flocks were comprised of 6 different species, including 5 flocks comprised solely of Nashville Warbler, Leiothlypis ruficapilla; 3 flocks solely of Cape May Warbler, Setophaga tigrina; 2 flocks comprised solely of Black-throated Green Warbler, Setophaga virens; 1 flock each of Tennessee Warbler, Leiothlypis peregrina; American Redstart, Setophaga ruticilla; Magnolia Warbler, Setophaga magnolia). The average flock species diversity was 2.37 (electronic supplementary material, table S1), and ranged from groups with just one species (flock species diversity of 0) to flocks with 7 species. Altitudes of triangulated warblers ranged from 1.1 m to 107.8 m above ground level, with a mean altitude of 10.3 ± 0.9 m (mean ± s.e.; figure 3).

Figure 3.

Histogram of wood warbler altitudes triangulated with a three-dimensional microphone array in the Keweenaw Peninsula, MI, USA. Data are drawn from 222 triangulated individuals of 18 wood warbler species in 46 different flocks.

We measured the physical distance between nearest flockmates in three dimensions for 44 different species pairs (electronic supplementary material, table S2). The most numerous species pair was Nashville Warblers(Leiothlypis ruficapila) migrating in close proximity with other Nashville Warblers (n = 61 pairs; average distance: 9.8 m; electronic supplementary material, table S2).

In general, one predictor—acoustic similarity of flight calls—was associated with the distance between flockmates, flock species diversity and flock size. Acoustic similarity of flight calls did not predict the temporal separation between flockmates, nor the altitude of birds. Warblers varied in the acoustic similarity of their nearest flockmates and the proximity of their nearest flockmates (see electronic supplementary material, figure S1). Warbler species, included as a random effect in all models, however, was not associated with any of the variables we measured; therefore we present results for the fixed effects only.

In detail, acoustic similarity of warbler flight calls corresponded with the spatial distance between migrating birds: among pairs of nearest individual warblers, birds that were more acoustically similar flew closer together (R² = 0.10; F = 24.64; p < 0.0001; figure 4a). Acoustic similarity also showed a relationship with the species diversity of flocks; most warblers flew in flocks of medium diversity composed of species with moderate acoustic similarity (R² = 0.58; F = 150.60; p < 0.0001; figure 4b).

Figure 4.

Relationships between the acoustic similarity of flight calls (the average across all pair-wise comparisons for each flock) and (a) proximity between individuals, (b) flock species diversity (a Shannon index; see §2), and (c) flock size during active migration of wood warbler flocks, as well as (d) the relationship between flock species diversity and proximity between individuals, for flocks of wood warblers (Parulidae) recorded with a microphone array. Lines-of-fit are shown in blue with 95% confidence interval in grey.

Acoustic similarity of flight calls also showed a relationship with the size of warbler flocks: migrating warblers were more likely to fly in medium-sized flocks that were acoustically similar (mean flock size = 9.20; R² = 0.16; F = 41.46; p < 0.0001; figure 4c). Flock species diversity also corresponded with the proximity of warblers within flocks: as flock species diversity increased, the proximity of birds within flocks decreased. Flocks with fewer species showed closer proximity between flockmates (R² = 0.20; f = 25.91; p < 0.0001; figure 4d). Both the relationships of acoustic similarity with flock size and flock species diversity are related to the fact that groups comprising fewer species will, by definition, exhibit higher acoustic similarity.

Acoustic similarity of warbler flight calls did not predict two characteristics of warbler flocks. Acoustic similarity did not predict the proximity of warbler flockmates in time (R² = 0.0019; F = 0.75; p = 0.39) and acoustic similarity did not predict the birds’ flight altitude (R² = 0.57, F = 0.9; p = 0.30). Birds' positions within the flocks did not show a relationship with spatial proximity between flockmates (R² = 0.23; F = 0.005; p = 0.94).

4. Discussion

Microphone-array analyses of the composition and behaviour of migrating flocks of warblers, and detailed analyses of the acoustic structure of warbler flight calls, reveal that acoustic communication is commonplace in mixed-species flocks comprising species with small to medium levels of diversity and that acoustic similarity varies with flocking behaviour. Migratory animals face many challenges to navigation and survival during migration, and communication with conspecific and heterospecific migrants may provide direct benefits, thereby leading to acoustic convergence of their flight calls. Warbler species with acoustically similar flight calls migrated in closer association with other birds. These results provide direct behavioural support for the Migration Similarity Hypothesis, at least within this group of migrants that exhibit moderate to high levels of acoustic similarity. We also found relationships between acoustic similarity, flock size and mixed-species flock diversity, where birds with acoustically similar flight calls were found flying in smaller flocks and flocks with lower species diversity. These patterns suggest that within mixed-species groups of migratory animals, acoustic signals may provide a low-cost mechanism for communication with both conspecific and heterospecific individuals.

The role that communication might play during the migrations of mixed-species flocks of birds has received scant attention. Numerous studies have suggested that avian flight calls play a functional role in maintaining acoustic contact between migrants [13,21–23], however, we have a poor understanding of how flight calls function to increase group contact among active migrants. The Migration Similarity Hypothesis suggests that species with similar migratory journeys (i.e. overlapping breeding ranges, migration ranges, wintering ranges and timing of migration) have evolved acoustically similar calls [19]. Acoustic convergence in flight calls would benefit migrants by increasing flock contact with individuals sharing similar migratory journeys, thereby reducing the likelihood of individuals being predated upon (e.g. the selfish herd theory [25]), by reducing the probability of disorientation when crossing dangerous or unfamiliar terrain, or by increasing the likelihood of finding high-quality stopover habitat [19]. We recently investigated a key prediction of the Migration Similarity Hypothesis by comparing acoustic similarity of flight calls to similarity in migration routes, migratory destinations and migration timing, and found support for the idea that warblers with overlapping migratory journeys have evolved acoustically similar flight calls [19]. Yet the Migration Similarity Hypothesis had never been explored from the perspective of direct studies of migrating warblers in the field, in part because visualizing the position of free-living birds during migration is fraught with challenges. We met these challenges in the current study by using a microphone array to record and triangulate the position of birds in three dimensions and found that acoustic similarity of flight calls corresponds with the structure of mixed-species flocks of warblers and the behaviour of birds within those flocks. Our results suggest that there may be an advantage to migrating with acoustically similar individuals because these individuals follow the same route and have the same timing of migration. More research is needed to examine why a few acoustically dissimilar individuals also migrate in close proximity. In other groups of organisms, including fish and ungulates, species have been demonstrated to sort into passive groups based on speed and size [36,37]. Little is known about variation in flight speed in warblers, and this is a worthy area of future investigation, possibly involving measurements in flight tunnels. The species we studied share similar body sizes and morphology, and consequently we predict that interspecific variation in flight speed does not explain the flock associations we describe here.

The warblers that we studied showed variation in their tendency to associate with other species, with some species (e.g. Nashville Warbler, Leiothlypis ruficapilla) showing a striking tendency to only associate with highly acoustically similar species. By contrast, some commonly detected species in our study (e.g. Cape May Warbler, Setophaga tigrina; Black-throated Blue Warbler, S. caerulescens; American Redstart, S. ruticilla; Golden-winged Warbler, Verminvora chrysoptera) had a low to medium acoustic similarity to most other warblers. This could indicate that species with acoustically distinctive calls migrate with diverse species of flockmates, and face different selection pressures.

We do not know whether flock composition is stable and consistent in membership across the period of migration, or whether flock composition ebbs and flows [13]. If migratory flocks are stable, acoustically similar flight calls should be advantageous for species that have a high likelihood of migrating together across part or all of their migratory journey. This could vary by species and could be particularly important for long-distance migrants such as Blackpoll Warbler (Setophaga striata) and Connecticut Warbler (Oporornis agilis), which face longer periods of migratory challenges, a lower yearly population turnover and potentially a higher inter-year fidelity to the same migratory route [8,38,39].

In our dataset, mixed-species flocks of warblers were composed of relatively few species per flock. As expected, acoustic similarity of flocks increased as flock diversity declined. Flock size and acoustic similarity also showed a relationship, with increases in flock size leading to declines in average acoustic similarity within flocks. As flock species diversity increased, birds flew with greater distances of separation. This suggests that warblers are more likely to fly at greater distances from flockmates as more species are present, likely because flocks are composed of less acoustically similar individuals. Future work could examine relationships between the optimum sizes of flocks and acoustic similarity of flockmates.

Finally, we found no relationship between acoustic similarity and the altitude of migration of mixed-species flocks. We also found no relationship between acoustic similarity and how closely warblers were spaced in time. The former result may represent a constraint on detections with our wireless microphone array (i.e. flocks flying at high altitudes would not be detected by our microphones), or may reflect the relatively low altitudes that are commonplace at our study site. The in-flight behaviour of migratory songbirds during migration is not well known and little prior work has quantified flock sizes and the behaviour of active migrants. Past studies have suggested that birds are widely spaced and not actively associating during nocturnal migration [13,40,41]. The spacing of migrants moving at higher altitudes during nocturnal migration, however, may be different from lower-altitude flights, or dawn ascents or descents, but no available technology exists for sampling the composition and behaviour of high-altitude flocks at the species level.

Our ability to quantify the composition and behaviour of migratory birds was limited by the accuracy of our microphone array. As we discuss in Gayk & Mennill [29], our estimates of error for our microphone array were comparable to high-quality triangulations of ground-based source sounds [29,42–44] and synthetic signals, which are known for being easy to localize [43]. These estimates of triangulation accuracy suggest that the triangulation estimates produced by our microphone array for flying warblers are highly accurate [19]. We were careful in this study to include only triangulated groups of warblers recorded under optimum conditions and to exclude calls recorded under conditions of rain, wind, or with wave interference.

Future research should focus on species-level differences in flight-calling behaviour, and additional functions of flight calls. Bird species may migrate at different altitudes and densities and this could have an effect on the transmission properties of calls. The airspace surrounding migratory birds is expected to impose constraints on the transmission of calls, and may drive selection on call structure [31], which is not mutually exclusive from the Migration Similarity Hypothesis and likely works in concert to act on flight call structure. Additionally, the degree to which rare versus common species migrate with heterospecific animals is unknown, as is the importance of interspecific communication for such animals. Rare species are likely to be members of groups in which they are acoustically dissimilar unless they have evolved acoustically convergent flight calls with more common species. Future work could use microphone arrays to examine the frequency with which individual species associate at the species level, whether different species fly at different densities, how this affects call structure and whether acoustically similar groups associate during migration (e.g. Zeep, Upsweep and Downsweep groups) [12,28]. Much remains to be learned about the behaviour and ecology of mixed-species groups of migratory animals. By focusing on the ecology and behaviour of mixed-species groups of actively migrating animals, we expect to gain important new insights into migrant communities, the evolutionary benefits of migrating in mixed-species groups, the evolution of migratory routes and the conservation of declining migrants.

Data accessibility

The data are provided in electronic supplementary material [45].

Authors' contributions

Z.G.G.: conceptualization, formal analysis, investigation, methodology, project administration, visualization, writing—original draft, writing—review and editing; D.J.M.: conceptualization, data curation, funding acquisition, methodology, project administration, supervision, validation, writing—original draft, writing—review and editing.

Both authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declarations

The authors declare no conflict of interest.

Funding

This research was funded through scholarship support from the Government of Ontario to Z.G.G., and grants from the Natural Sciences and Engineering Research Council of Canada (NSERC), the Government of Ontario, the University of Windsor, and the Canada Foundation for Innovation to D.J.M.