Sailfish generate foraging opportunities for seabirds in multi-species predator aggregations

1. Introduction

The open ocean harbours the largest predator aggregations on the planet [1,2]. Thousands of predators of different species can gather in the same location and simultaneously predate upon large schools of prey [2]. Despite the ecological relevance of these multi-species predator aggregations and extensive literature documenting the co-occurrence of particular combinations of predator species, their synergistic or antagonistic nature is rarely assessed (but see [3,4]). This is likely due to difficulties quantifying predator behaviours such as attack and capture rates as well as the spatio-temporal positions of predators and prey in the wild. This means that the mechanisms that underlie the antagonistic or synergistic nature of predator interactions are also poorly understood [4].

Various marine predators are known to associate with each other, including for example: dolphins and tuna [5–7], seals and billfishes [4], humans and dolphins [8], sharks and trevally [9,10], dolphins and baleen whales [11,12]. However, most commonly studied are associations of large subsurface predators (e.g. delphinids, tuna, pinnipeds) with seabirds (e.g. terns, gulls, shearwaters) [3,5,13–19]. The prevalence of these particular associations is usually explained by three proposed benefits that subsurface predators provide for seabirds: (1) subsurface predators make fish schools accessible to birds by driving them to the surface; (2) injuries inflicted on prey fish by subsurface predators make tissue and scales available to bird species that do not prey on the whole fish; (3) predator attacks on the prey school reduce its cohesion and thus reduce the collective defensive mechanisms of the fish. The first two hypotheses have been supported by direct evidence [16,20–22], however, a direct measure of prey cohesion in the wild in the context of multi-species aggregations (and thus support for the third hypothesis) is currently lacking.

The benefit that has received the most attention in the existing literature is that subsurface predators take part in collective herding; driving prey animals from the lower layers in the water column to the surface where they become accessible to birds [5,13,15,19,22–26]. Driving schooling prey to the surface is a common predator strategy in the ocean and likely benefits subsurface predators by restricting the movement and potential escape routes of prey schools. There are numerous anecdotal accounts of this behaviour [6,24,27], however, due to logistical difficulties observing predator–prey interactions at depth, quantification of predators driving prey schools to the surface has only been achieved via hydro-acoustics and animal-borne underwater cameras [20–22]. When orcas collectively herded schools of herring (Clupea harengus) to the surface the fish were subsequently observed to be preyed upon by common gulls (Larus canus) and white-tailed eagle (Haliaeetus albicilla) [20]. In the absence of such direct evidence, an alternative scenario could be that both predator types, avian and subsurface, encounter prey that is already at the surface and hence available to avian predators.

The second benefit concerns smaller species of petrels and shearwaters (Procellariidae) which may not be able to target certain prey fish due to their relatively small body size and therefore opportunistically feed on floating scales and tissue [16,23,28]. Larger predators can injure more fish within a school than they capture per attack [29], leading to excised tissue being available for consumption by the birds. This hypothesis has been proposed to explain the association of Parkison's petrel (Procellaria parkinsoni) with melon-headed whales (Peponocephala electra) and false killer whales (Pseudorca crassidens), with the whales dismembering large prey at the surface and leaving food scraps in the process [16].

The third potential benefit that could facilitate these associations is related to a disruption of the collective defence of fish schools by the subsurface predators. Many defensive benefits for individuals in a group such as the confusion effect [30,31], rely on cohesive organization [32]. Predators interfere with these defensive mechanisms by reducing cohesion through successive attacks [33]. In the context of multi-species predator aggregations, Cape gannets (Morus capensis) were more than twice as successful at catching fish when attacks by other predators occurred in the 15 s prior to their attack [3]. An associated individual-based model showed that school cohesion (measured in terms of changes in the schools radius—the distance of each fish to the school's centroid over time) was reduced by successive predator attacks (attacks from either subsurface or avian predators [3,34]. Although this explanation is likely, direct measures of prey cohesion in relation to predator attacks are needed to demonstrate the causal effect of reduced cohesion. Otherwise, an alternative explanation could be that both subsurface and avian predators exploit favourable conditions (i.e. states of reduced prey cohesion) that are not caused by attacks, but rather, for example, via natural variation in prey speed [35]. Currently, there are no studies that directly quantified prey cohesion in the context of multi-species predator aggregations.

In this paper, we investigate the question of whether Indo-pacific sailfish (Istiophorus platypterus) create foraging opportunities for terns (Onychoprion sp.) when attacking schools of anchovies (Encrasicholina sp.) and sardinellas (Amblygaster sp.). We observed that sailfish attacks on fish schools frequently preceded attacks by terns which led to the hypothesis that sailfish created a by-product benefit for terns by reducing school cohesion and increasing prey vulnerability. First, we measured how often sailfish attacks directly preceded tern attacks, then we quantified the polygon outline of the prey school (as a proxy for cohesion) using high-resolution drone footage at regular time intervals before tern attacks. This was done to test whether bird attacks were preceded by sudden decreases in cohesion. Subsequently, we measured the polygon outline of the prey school before, during, and after a sailfish attack to test whether sailfish attacks resulted in sudden decreases in prey school cohesion. Finally, we directly related the polygon area of the prey school at the moment of the tern attack to the timing of the sailfish attacks, to test whether sailfish attacks were likely to create a by-product benefit for terns.

2. Methods

We collected high resolution drone video footage (DJI P4P pro v. 2.0, 2.7 k, 60 fps, 5–30 m) of sailfish and terns hunting schools of small prey fish, most likely anchovies (Encrasicholina sp.) and sardinellas (Amblygaster sp.) [36] in 2023 (21–27 August), off the eastern coast of the Malaysian peninsula in the South China Sea (2°56'15.4″ N 103°42'14.1″ E) (electronic supplementary material, video S1). Both predator species can reliably be encountered during this time off the east coast of the Malaysian peninsula. Sailfish were located opportunistically, by approaching seabird aggregations that preyed on the same schooling fish. The only seabirds that were observed feeding were of the genus Onychoprion, either sooty (O. fuscatus) or bridled (O. anaethetus) terns. Other seabirds (i.e. lesser frigate birds, Fregata ariel) occasionally accompanied sailfish, although they were never observed feeding. Five different fish schools were filmed in total where sailfish and terns were both attacking the same prey school. From this footage, we selected approximately 18 min where conditions were optimal (i.e. little to no wave or wind action, and minimal glare) in order to accurately measure school cohesion (five videos between 92 and 327 s).

Within the videos, we determined the time of every attack (made by sailfish (N = 126) or terns (N = 60)) that occurred. Sailfish attacks were identified by rapid slashes of the rostrum at prey schools [37]. Due to the small size of prey groups that were encountered, sailfish only attacked one at a time. Bird attacks were identified by swoops towards the ocean's surface and subsequent contact of their bill with the water. Capture success could only be assessed for seabird attacks, as captured prey was clearly visible in their beaks after an attack. To test our question related to the interaction between sailfish and seabirds we analysed the 32 attacks by terns which were preceded by sailfish attacks as well as the corresponding 32 sailfish attacks. These attacks were spread across the five fish schools, with up to 17 terns within a frame, however, although this is likely only a subset of total flock size in the area, we cannot exclude the possibility of pseudo-replication, as birds were not individually identifiable. To quantify prey behaviour, we exported still-frames out of the video, using VirtualDub2 (v. 1.10.5), and measured the area of the fish school using a convex polygon in ImageJ (polygon size) [38]. A convex polygon was used given the top-down, two-dimensional images of the drone. This measure was used as a proxy for school cohesion (figure 1; electronic supplementary material, image 1) [32]. This was done at the point of each bird attack, as well as each half second prior to this (up until 3.5 s before the attack). The same method was used to measure the impact of sailfish attacks on the prey school. Here we measured polygon size at the point of attack and 2 s prior and after the attack. Due to the small size of the prey fish (< 15 cm) as well as waves and light reflections, which occasionally restricted our view into the water, polygon size was likely underestimated, as some individual fish after sailfish slashes may have remained undetected.

As the height of the drone, distance to the fish school, as well as angle of the camera, varied between schools, we used the relative polygon size, enabling us to compare prey behaviour throughout a recording. The polygon size at the first measurement point of each series (either −3.5 or −2 s) was set as 100%, as the fish school was still unaffected by the approaching predators.

For all statistical analyses we used R (v. 4.2.2) [39]. To test at which time points polygon size significantly increased we used a Friedman test with Bonferroni post-hoc tests. We performed a linear regression to investigate the interaction between polygon size at the time point of tern attack and the time after a sailfish attack.

3. Results

Out of 32 bird attacks that were directly preceded by sailfish attacks, 56% of these attacks occurred within 1.5 s after the sailfish attack. Polygon size increased significantly before a tern attack (Friedman χ² = 71.603, d.f. = 7, p-value = < 0.0001, figure 2a). Shortly before the tern attack (at times −1.5, −1, −0.5 s and during the attack at time 0 s) the polygon size was significantly larger than the school's initial polygon size (at time −3.5 s) (Bonferroni corrected post-hoc pairwise comparisons, all p < 0.001, figure 2a).

Sailfish attacks significantly increased polygon size (Friedman χ² = 81.61, d.f. = 4, p-value = < 0.0001, figure 2b). Polygon size at the time of a sailfish attack, and up to 2 s afterwards, were significantly larger than the initial size (at −2 s) (Bonferroni corrected post-hoc pairwise comparisons, all p ≤ 0.0001). In fact, polygon size at the moment of sailfish attack was on average six times larger than its initial size (figure 2b).

The sooner a tern attack took place after a sailfish attack, the larger the polygon size (figure 2c; F_{1, 29} = 6.015, p = 0.02044). One tern attack was excluded from the analysis as a sailfish attacked 0.2 s after the tern. Tern capture success was not analysed in relation to relative polygon size because there were only six successful attacks among a total of 32 attacks.

4. Discussion

5. Conclusion

Building on the pioneering work of Thiebault et al. [3] we provide direct evidence for the effects of predator attacks on the cohesion of prey groups and how this in turn could facilitate the association of avian predators with large teleost predators in the open ocean. Reports of associations of seabirds with large teleost predators and dolphins are very common [3,5,13–19] and provide many opportunities for further studies of the mechanisms underlying the generation of foraging benefits between species.

Ethics

All research was conducted in line with the laws and legislation of National Committee for Research Ethics in Science and Technology (NENT), Malaysia and complied with the guidelines regarding the treatment of animals in behavioural research (ASAB, 2022).

Data accessibility

The data are provided in the electronic supplementary material [52].

Declaration of AI use

We have not used AI-assisted technologies in creating this article.

Authors' contributions

M.L.: conceptualization, formal analysis, investigation, writing—original draft; A.L.B.: formal analysis, investigation, writing—review and editing; K.P.: investigation, writing—review and editing; P.B.: investigation, writing—review and editing; S.K.: investigation, writing—review and editing; P.R.: conceptualization, writing—review and editing; M.J.H.: investigation, writing—review and editing; A.Y.T.: conceptualization, writing—review and editing; J.K.: conceptualization, investigation, writing—review and editing.

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

Conflict of interest declaration

We declare we have no competing interests.

Funding

This research was supported by funding from the DFG under Germany's Excellence Strategy – EXC 2002/1 ‘Science of Intelligence’, project no. 390523135 (A.L.B., J.K., P.B. and P.R.), and from the Leibniz-Institute of Freshwater Ecology and Inland Fisheries. M.J.H. was supported by a Deutsche Forschungsgemeinschaft (GZ:HA 9403/1-1 AOBJ: 675842f).