California sea lions interfere with striped marlin hunting behaviour in multi-species predator aggregations

Abstract

The open ocean offers a suite of ecological conditions promoting the occurrence of multi-species predator aggregations. These mixed predator aggregations typically hunt large groups of relatively small and highly cohesive prey. However, the mechanisms and functions of these mixed predator aggregations are largely unknown. Even basic knowledge of whether the predator species' interactions are mutualistic, commensal or parasitic is typically missing. Moreover, recordings of attack and capture rates of marine multi-species predator aggregations, which are critical in understanding how and why these interactions have evolved, are almost completely non-existent owing to logistical challenges. Using underwater video, we quantified the attack and capture rates of two high-trophic level marine predators, California sea lions (Zalophus californianus) and striped marlin (Kajikia audax) attacking schools of fishes in the Southern California Current System, offshore the Baja California Peninsula. Recording over 5000 individual attacks across 13 fish schools, which varied in species, size and predator composition, we found that sea lions kleptoparasitized striped marlin hunts and reduced the frequency of marlin attacks and captures via interference competition. We discuss our results in the context of the phenotypic differences between the predator species and implications for a better understanding of multi-species predator aggregations.

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

1. Introduction

Most prey species are predated upon by multiple predators [1], and the interactions between different predator species drive the spatial and temporal distribution of both prey and predators, playing a key role in structuring ecosystems [2–4]. In terrestrial systems, these between-species predator interactions commonly take the form of multi-species aggregations at a kill (a carcass). The interactions between predator species at these aggregations are predominantly kleptoparasitic, with one (or more) species receiving a benefit by exploiting the gains of another species without incurring the costs of, for example, capturing prey [5–9]. Feeding-related costs during such kleptoparasitic interactions primarily consist of competitive attempts to gain access to the carcass when it is being aggressively guarded by other predator species [7]. Kleptoparasitism in this context takes the form of scavenging, which is an important source of energy transfer between trophic levels and a key stabilizing force for food webs [8].

The abundance of carrion and kleptoparasitism in terrestrial systems is partly owing to the ecosystem's vegetation favouring the evolution of large herbivores and relatively small predator–prey size ratios (i.e. predators hunt relatively large prey which cannot be rapidly ingested) [10]. When turning our attention to open-ocean systems, we observe a different predominant form of predator–prey interaction which is characterized by relatively large predator–prey size ratios. Here, an abundance of single-celled primary producers has led to huge quantities of zooplankton being available for small-bodied carnivorous predators. In turn, this profusion of small, pelagic schooling fishes becomes a dominant prey resource for a wide variety of larger predators [10–13]. Extreme examples of this are marine upwelling regions, which often have a wasp-waist trophic structure owing to species-poor intermediate trophic levels, in which a few small pelagic fish species dominate their trophic level and are the main food source for species-rich assemblages of predators [14–17]. Note that carrion also exists in the open ocean—primarily in the form of whale carcasses [18–20]—but these form a minor part of the predator–prey interactions. Large pelagic fish schools differ substantially from animal carcasses, as there is not one unique moment of a kill, but an ongoing harvesting of prey out of a school, which continuously attempts to collectively evade predation. These unusual features of the predator–prey interactions make them an attractive subject for studying multi-species predatory groups.

Multi-species predator groups are common in the open ocean and have been the subject of numerous studies (e.g. [21–28]). Previous work has categorized species as either ‘catalysts’/‘producers’ or ‘opportunists’/‘scroungers’ [29]. Larger teleosts [21,25,26] and dolphins [30–32,34] are commonly identified as producers because they are presumed to corral prey schools from deeper waters to the surface and fragment smaller schools away from larger aggregations. Rorqual whales, Balaenopteridae, [27], sea lions, Otariidae, [30,31] and seabirds [24] are commonly identified as scroungers, perhaps because their morphology or physiology makes them less capable of isolating their own teleost prey schools [24,32,33].

Historically, multi-species predator interactions have been difficult to quantify [35], especially in marine systems. Therefore, marine predators have typically been categorized based on when and how often they are observed associating with other species. However, the mechanisms and functions of these multi-species predator aggregations have not been thoroughly explored. For example, the size structuring of trophic layers in the open ocean and the abundance of meso-predators means many predators may form groups for the dual-strategy of increasing capture rate and decreasing predation risk [26,36,37]. A better understanding of the nature of these multi-species predator interactions requires detailed information on attack and capture rates of multiple species, ideally in the presence and absence of key competing predator species. Such information can shed light on the consequences of the capture efficiencies of different predator species in multi-predator groups. Thiebault et al. [38] measured the feeding success of cape gannets (Morus capensis) preying upon sardine schools that were simultaneously attacked by other predator species. Gannets nearly doubled their capture success when one or two attacks by other individuals occurred in the 15 s preceding the gannet's attack. This prior attack either came from long-beaked common dolphins (Delphinus delphis bairdii) or other cape gannets. Multiple hunters (con- or heterospecific) attacking a prey school (especially from multiple angles) can thus provide mutual benefits by fragmenting the prey school and reducing its ability to reform [38,39], thereby increasing feeding efficiency and driving the evolution of intraguild mutualisms.

Here we focus on the interactions between striped marlin (Kajikia audax) and California sea lions (Zalophus californianus) preying upon Pacific sardine (Sardinops sagax) and chub mackerel (Scomber japonicus) schools. These aggregations provide a multi-species interaction that permits a close assessment of the hunting performance (attack and capture rates) of both predator species, thus allowing an accurate classification of the relationship between the two predator species.

We conducted our observations in the Southern California Current System offshore the Bahía Magdalena lagoon complex (Baja California Sur, Mexico). Coastal upwelling in this area injects an abundance of nutrients into the surface layer [40,41] leading to high primary production. Consequently, striped marlin gather in this area during autumn and winter [42] where they attack schools of Pacific sardine, chub mackerel, thread herring (Opisthonema libertate) and Pacific anchovy (Cetengraulis mysticetus). Marlins are often joined by California sea lions, but occassionally also by other predators such as various dolphin species, sailfish (Istiophorus platypterus), mahi-mahi (Coryphaena hippurus), big-eye tuna (Thunnus obesus) and wahoo (Acanthocybium solandri). When the marlin corral the fish schools against the surface, marine birds such as magnificent frigatebirds (Fregata magnificens), brown pelican (Pelecanus occidentalis) and brown booby (Sula leucogaster) frequently attack as well. Smaller bird species such as storm petrels (Hydrobates sp.) and shearwaters (Puffinus sp.) also regularly join to feed on smaller pieces of fish tissue or scales which result from the mechanical damage that the marlin cause with their bills to their prey [43].

The striped marlin is a large predatory fish primarily living in the epipelagic zone of tropic and temperate waters of the Indian and Pacific Ocean [44]. They mainly feed on small to medium-sized pelagic fishes such as Pacific sardines and chub mackerel as well as cephalopods like Humboldt squid (Dosidicus gigas) [45]. When preying upon schooling fishes, striped marlin make rapid and often repeated dashes through the prey school, attempting to disperse the school and isolate a single sardine which the marlin then guide towards their mouths using the bill or, if the prey is sufficiently immobilized, directly ingest. When preying upon epipelagic fish schools, striped marlin predominantly hunt in groups and a characteristic of group-hunting striped marlin is that they corral schools of prey fishes against the ocean's surface which they then attack in an alternating fashion one at time, with individuals making repeated attacks in a row (dash sequence) while the other marlin swim close by, before interrupting the attacker's ‘dash sequence’ [43,46]. Their attack behaviour is similar to other group-hunting billfishes, such as sailfish, which also tend to access the prey resource one at a time. However, while striped marlin primarily accelerate through the school in a rapid ‘dash’, dispersing the prey fishes, sailfish more gently insert their bill into the school (without inducing prey evasion) and swipe or tap at prey fishes [43,47–50]. Both species injure more fishes than they consume in a single attack leading to the accumulation of injured and easier-to-catch prey fishes [49]. This accumulation of injuries, that impedes the locomotory capabilities of individual prey and thereby increases capture efficiency, has been identified to confer cooperative benefits in group-hunting billfishes [48].

California sea lions are large group-living marine mammals that are apex predators of fishes and cephalopods [51,52]. They have been the focus of a number of studies of foraging behaviour, many of which were individual-based and focused on prey choice and diving behaviour [53–63]. However, only a few studies have examined attack behaviour or group hunting in the pinnipeds. While the tactics of Galapagos sea lions (Zalophus wollebaeki) hunting schools of pelagic fishes are partially described in studies investigating sea lion pods driving schooling prey into shallow waters for capture [64,65], to our knowledge, no such information is available for the closely related and much more widespread California sea lion. It is known, however, that California sea lions exhibit a large degree of plasticity and opportunism in their foraging behaviour. For instance, they are able to take advantage of a human-made structure (the Bonneville Dam), which facilitates prey access to migrating salmonid fishes [66]. Furthermore, they exploit the detection capabilities of dolphins to locate food sources and have been characterized as parasites during hunting interactions [30]. Importantly, their prey spectrum overlaps with that of striped marlin, suggesting that their interactions with striped marlin may be competitive [67].

In contrast with most previous studies on marine multi-species predator groups, our observations of the hunting behaviour of striped marlin and California sea lions provide data on attack and capture rates of both species. By making comparisons of the attack and capture rates of striped marlin, with and without the presence of sea lions, assessments can be made about the effect of the presence of sea lions on the hunting success of striped marlin. This helps to clarify if the two species are in a mutualistic, commensal or parasitic relationship with each other. If the relationship of the two species is mutualistic or commensal, we would expect that the capture rate of striped marlin to increase in the presence of sea lions. However, based on previous observations [43], we hypothesized that the presence of sea lions is kleptoparasitic and will decrease the attack and capture rate of striped marlin and decrease their chance of capture per attack.

2. Methods

We collected underwater video footage of striped marlin and California sea lions hunting schools of either Pacific sardine or chub mackerel in 2018 (8–12 November) and 2019 (7–13 November), 15–30 km offshore the Mexican state Baja California Sur in the Pacific Ocean (24° 54.52–48.5′ N, 112° 34.46–23.51′ W) (electronic supplementary material, video S1).

At this location, striped marlin, either in the presence or absence of sea lions, could reliably be encountered when actively hunting prey fish schools, which we opportunistically located by following seabird aggregations. Upon arriving at these prey schools, we gently entered the water and started to record the animals using GoPro HERO4 Cameras (120 fps), keeping a 5–10 m distance. In 18 (out of 31) active hunts we encountered, the prey school and predators were moving too fast to maintain contact and we could not collect useable footage. For the remaining 13 hunts, we were able to maintain contact and record attacks. We recorded continuously until either all fishes in the prey school were consumed, the prey and predators moved away, or the cameras' batteries were empty. In four of these hunts, striped marlin were the only predators attacking the school with sea lions absent from the immediate area. In seven hunts both striped marlin and sea lions were continuously present, and in two hunts striped marlin were present first before being joined by sea lions.

In 2021, we collected additional video footage of the hunting behaviour with an unmanned aerial vehicle (UAV, Phantom 4 Pro V2.0, DJI) from 10 to 30 m height (2.7 K, 60 fps). We flew the UAV 22 times across 2 days (12–13 November) resulting in approximately 5 h of video. We recorded 12 unique hunts of prey schools across these 22 flights. Striped marlin were present at all prey schools while sea lions were present at a third of all schools. The main advantage of using drones was the ability to follow fast-moving fish schools that were under attack.

(a) . Underwater video analysis

In total, we analysed 415 min of underwater video footage (mean ± s.e. per hunt: 32 ± 7 min, range: 5–89 min). For each video sequence, we started at t = 0 and continuously recorded the timing of each attack (either by a striped marlin or sea lion) and whether or not the attack resulted in a successful capture (yes/no), using VirtualDub (https://www.virtualdub.org/), BORIS (http://www.boris.unito.it/) and VLC (https://www.videolan.org/). We also recorded whether subsequent attacks were performed by the same or a different individual. Multiple consecutive attacks by the same individual (i.e. without another individual interrupting the sequence of attacks) were defined as a ‘dash sequence’.

Attacks were defined differently for the two predator species. Marlin attack using two different strategies: (i) swimming directly at and usually through the school at high speed, often incidentally contacting the prey fish with their bill, and (ii) using their bill more intentionally to swipe or tap their prey swimming past or through the school [39] (electronic supplementary material, video S2). Both strategies were classified as attacks. Sea lions display very characteristic attack movements, generally attacking their prey from below in a type of roll that we have called a ‘loop attack’ (electronic supplementary material, video S3). In these attacks, the prey school is approached in a horizontally straight line, but 0.5–1 m deeper than the school. Underneath the prey school, the sea lion rapidly moves its head upwards (dorsiflexion) and propels itself towards the school by rapidly adducting its front flippers in a clap-like manner. This leads to a characteristic loop-like trajectory. Prey contact in such a stereotypical ‘loop attack’ typically occurs after a half turn when the sea lion's chest is directed to the surface. Deviations from this general attack pattern were also registered as attacks, under the condition that the sea lion's behaviour was targeted at prey. Video footage was analysed in slow motion (0.20–0.50x the normal speed), using replay or zoom function when required. Sometimes, the prey school was temporarily obscured in the video (e.g. because it was too far away from the camera for accurate measurements of hunting behaviour). These periods were excluded from all analyses (mean ± s.e. per hunt: 7 ± 5 min, range: 0–62 min). It was not possible to record the total number of striped marlin present at each prey school because they hunt in large groups and swim around the prey school, making it challenging to quantify their total number. Sea lions hunted in smaller groups, and we used differences in size and characteristic markings to identify individual sea lions to determine their group size at each prey school (mode = 2, range = 1–5) (electronic supplementary material, figure S1).

As we encountered striped marlin hunting in the absence and presence of sea lions (but not sea lions hunting by themselves, see below), we focused the analysis of the hunting behaviour on the comparison of striped marlin hunting with and without sea lions. From our continuous data recordings, we extracted the following variables for striped marlin: (i) the time interval in between two consecutive striped marlin attacks (which could either be the same or a different marlin); (ii) the time interval in between two successful prey captures (which could either be the same or a different striped marlin); and (iii) whether the individual striped marlin reapproached after an attack (yes/no). For the two former variables, we only included events for which there was no missing footage in between two consecutive attacks/captures (since missing footage could potentially contain an unobserved attack/capture). Regarding the latter variable, it is important to note that individual striped marlin reapproach the prey school multiple times in one attack sequence (dash sequence) [43,46]. As described above, for each attack (by either a striped marlin or sea lion), we scored whether it resulted in a successful prey capture (yes/no).

We additionally scored the prey species and prey school size for each hunting event. Prey school size was estimated using counts from still frames of the video and sorted into three categories: small (less than 150 fishes), medium (150–700) and large (greater than 700).

We re-coded a subset of the data (12 min (242 attacks) over 12 different fish schools) to assess inter-observer reliability for the variables recorded above. The most variation between observers occurred in classifying whether attacks occurred or not (inter-observer reliability rating (IRR) of 96.7%). After accounting for this the inter-observer reliability of other statistics was high: time between attacks (IRR 98%), success (yes/no) (IRR 99.2%); whether the subsequent attack was by the same or different individual (IRR 98.3%).

(b) . Statistical analysis

For all statistical analyses, we used R (version 4.1.2). We used Bayesian hierarchical generalized linear models using the brm function from the brms package [51] and its default priors. For each model, we ran three chains in parallel with 3000 iterations each, of which the first 1500 were discarded as burn-in. Visual inspection of the Markov chains and the Gelman-Rubin statistic (Rhat) of all final models indicated that all Markov chains converged.

To test whether striped marlin and sea lions differed in capture success when placing an attack in each other's presence, we fitted ‘attack successful’ (yes/no) as response variable, ‘predator species’ (marlin/sea lion), ‘prey species’ (mackerel/sardine) and ‘prey school size’ (small/medium/large) as population-level effects and ‘prey school’ as group-level effect, using a binomial regression. We only included attacks when both species were present, that is, we excluded observations when only striped marlin were present.

The next series of models examined whether the presence of sea lions affected striped marlin's hunting behaviour. To test whether the presence of sea lions affected the time interval in between two consecutive striped marlin attacks, we fitted ‘attack interval’ (in seconds) as response variable, and ‘sea lion presence’ (yes/no), ‘prey species’, and ‘prey school size’ as population-level effects and ‘prey school’ as group-level effect. These count data were best approximated by a Poisson distribution. Therefore, we used a Poisson regression. In all models described below, we fitted the same population- and group-level effects unless stated otherwise. To test whether the number of sea lions present affected the time interval in between two consecutive striped marlin attacks, we ran a similar model, replacing ‘sea lion presence’ by ‘number of sea lions’ (one or two sea lions (n = 6 groups)/five sea lions (n = 3 groups)). For this analysis, we only included striped marlin attacks in the presence of sea lions. The full model did not converge. Therefore, we removed ‘prey species’ and ‘prey school size’ from the population-level effects.

To test whether the presence of sea lions affected the capture interval in between two consecutive striped marlin captures, we fitted ‘capture interval’ (in seconds) as response variable using a Poisson regression. Comparing the striped marlin's capture interval in the presence of one or two versus five sea lions was not possible because there were too few observations of uninterrupted striped marlin capture intervals in the presence of five sea lions.

Next, we investigated whether sea lion presence affected the length of the striped marlin's dash sequence (i.e. the number of attacks in one sequence by the same individual) and their attack success when placing an attack. For the former, we fitted whether the same striped marlin reapproached after an attack (yes/no) as response variable in a binomial regression. For the latter, we fitted ‘attack successful’ (yes/no) as response variable in a binomial regression.

Finally, we determined whether sea lions or marlin dominate access to the prey school. This was done by calculating the attack transition probabilities of individuals within and between species. When an individual finished its dash sequence (which could be one or multiple attacks), we scored whether the next individual belonged to the same species or not. We only included transitions in which there was no missing footage during the transition. We only did this for groups with five sea lions. Although we do not know the exact number of marlins in these groups, the absolute minimum is 10 (based on counts in single video frames), though the true numbers are most likely substantially higher.

3. Results

(a) . Producer role of striped marlin

Observations over the 3 years of data collection strongly suggest that the striped marlin is the species that isolates small prey groups from larger schools and that sea lions join after this has been accomplished. Striped marlin were present at all prey schools we recorded (25 out of 25); including very large prey schools which the striped marlin broke up into smaller fragments (figure 1a,b). By contrast, sea lions were only present in just over half of the recorded hunts (13 out of 25) and never by themselves as the single predator species (figure 1c,d,f). Moreover, sea lions twice joined striped marlin after the beginning of the video recording (figure 1e). When both species were present, sea lion attacks were more successful than striped marlin attacks (brms: β [confidence interval (CI)] = 0.44 [0.23–0.64]; figure 1g; electronic supplementary material, table S1). Sea lions were successful in 19.3% of their attacks and striped marlin in 14.1% of their attacks. There was no effect of prey species (β [CI] = −0.13 [−1.27–0.97]) or prey school size (medium: β [CI] = −0.16 [−1.16–0.78]; large β [CI] = 1.04 [−0.24–2.24]) on capture success.

Figure 1.

(a) Striped marlin (top left) attacking a large fish school; (b) striped marlin fragmenting a fish school into smaller parts; (c,d) striped marlin and sea lions attacking a small fish school. Image taken from the (a–c) UAV footage and (d) underwater video; (e) distribution of attacks over time coloured by species for the 13 prey schools recorded underwater; (f) proportion of recorded hunts (combined underwater and UAV recordings) where striped marlin or sea lion were the only species present (yellow: marlin 12 out of 25, sea lion 0 out of 25) or part of a mixed group (green: marlin 13 out of 25, sea lion 13 out of 13); (g) success probability of striped marlin and sea lions for an individual attack (images F. Dhellemmes and M. Hansen). (Online version in colour.)

(b) . Attack and capture frequencies

Comparing the interval between two subsequent attacks of striped marlin (either by the same or a different striped marlin), in the presence and absence of sea lions, we found that sea lion presence increased the striped marlin's between-attack interval (mean ± s.d. attack interval in the absence of sea lions: 3.24 ± 2.73 s, in the presence of sea lions: 6.32 ± 9.36 s; brms: β [CI] = 0.40 [0.32–0.47]; figure 2a; electronic supplementary material, table S2). There was no effect of prey species on between-attack interval (β [CI] = −0.13 [−1.27–0.97]). Between-attack intervals at medium- or large-sized prey schools did not differ from those at small-sized prey schools (medium: β [CI] = −0.16 [−1.16–0.78]; large: β [CI] = 1.04 [−0.24–2.24]). Striped marlin's between-attack intervals were longer in the presence of five sea lions as compared to one or two sea lions (β [CI] = 1.01 [−0.16–2.16]); electronic supplementary material, figure S2). Comparing the interval between two subsequent successful prey captures of striped marlin (either by the same or a different striped marlin), in the presence and absence of sea lions, we found that sea lion presence increased striped marlin's capture interval (mean ± s.d. capture interval in the absence of sea lions: 15.2 ± 15.1 s; in the presence of sea lions: 41.2 ± 73.8 s, β [CI] = 1.08 [0.98–1.17]); figure 2b; electronic supplementary material, table S3). There was no effect of prey species on capture interval (β [CI] = 0.38 [−1.30–2.06]). Capture intervals at medium- or large-sized prey schools did not differ from those at small-sized prey schools (medium: β [CI] = −0.31 [−1.74–1.11]; large β [CI] = 0.22 [−1.67–2.16]).

Figure 2.

The effect of sea lion presence on striped marlin's attack interval, capture interval and attack sequence length. (a,b) The normalized probability density function (PDF) of the time in between two consecutive striped marlin (a) attacks and (b) captures. (c) The normalized PDF of the dash sequence length of striped marlin. PDFs are shown for striped marlin in the absence (blue) and presence (red) of sea lions. Inset shows mean of the posterior distributions, and error bars show 95% credible intervals of Bayesian regression models. (Online version in colour.)

(c) . Dash sequence length and capture success

Next, we investigated whether sea lions actively interfere in the dash sequence of single striped marlin and their capture success when placing an attack. Individual striped marlin are known to regularly make repeated attacks in one dash sequence. Comparing an individual striped marlin's likelihood to reapproach the prey school in the presence and absence of sea lions, we found that sea lion presence did not alter striped marlin's likelihood to reapproach (β [CI] = 0.13 [−0.19–0.45]); figure 2c; electronic supplementary material, table S4), hence the length of striped marlin's dash sequence did not differ in the presence versus absence of sea lions. Furthermore, prey species did not affect the likelihood to reapproach (β [CI] = −0.50 [−1.38–0.43]), nor did prey school size (medium: β [CI] = −0.13 [−0.86–0.60]; large β [CI] = 0.40 [−0.58–1.49]).

When a striped marlin carried out an attack, the likelihood that the attack resulted in a successful prey capture was not affected by the presence of sea lions (β [CI] = −0.13 [−0.48–0.23]); electronic supplementary material, figure S4 and table S5). Capture success probability was not affected by prey species β [CI] = −0.40 [−1.63–0.92]); or prey school size (medium: β [CI] = 0.19 [−0.82–1.35]; large β [CI] = −0.15 [−1.65–1.38]). Sea lions thus did not directly interfere with the dash sequence of single striped marlin, nor did their presence lower the capture probability when striped marlin attacked.

(d) . Dominance over access to the prey school

When an individual sea lion finished an attack (in mixed-species groups with five sea lions present), it was much more likely to be followed by another sea lion (82.0%), than by a striped marlin (18.0%). Whenever a striped marlin made an attack, it was more likely to be followed by a sea lion (71.1%) than by another striped marlin (28.9%) (electronic supplementary material, figure S3). This information has to be seen in the context of the relative abundance of both species near fish schools. Despite the fact that we do not know the exact number of marlin at each fish school, it was clear that they considerably out-numbered sea lions at every fish school. This suggests that sea lions dominated the access to prey schools and that sea lion presence made it more difficult for the striped marlin to regain access to the prey school between dash sequences.

4. Discussion

We found converging evidence that sea lions negatively interfere with marlin hunting behaviour when the two species interact at prey schools. Striped marlin and sea lions were not observed together in the absence of prey schools in a non-hunting context, as has been observed with sea lions and cetaceans in the Southern California Bight [53]. Thus, it is unlikely that both species are socially attracted to each other for anti-predator benefits [33,54,55].

It is likely that the two species aggregate on the same resources. Local spatio-temporal positioning of prey fishes such as sardine, herring and mackerel are often coupled with the tidal cycle (e.g. [58]) or other abiotic variables, providing a predictable resource for different predator species to discover. This should promote the occurrence of multi-species predator groups in the open ocean. However, sea lions were only observed hunting in the presence of striped marlin, while striped marlin were recorded hunting alone in almost half of the recorded hunts. Moreover, we observed sea lions joining hunts where striped marlin were already present. Altogether this suggests sea lions benefit from the effort of striped marlin in finding prey schools and corralling these up to the surface, thereby making them easier to find and access. Based on these observations the relationship between the two species might thus be characterized as kleptoparasitic. An intriguing question is how sea lions find these prey schools after they have been corralled to the surface by the striped marlin. Possibly, sea lions use diving seabirds to locate prey schools. Striped marlin typically push the fish schools to the ocean's surface where they come under attack by flocks of frigate birds that are large enough to be seen from a great distance. Potentially, sea lions searching for prey use ‘spy-hopping’ behaviour (brief vertical elevation of the body to raise the head over the water surface and facilitate above-surface observations), like they do when following dolphins [30]. Alternatively, they might use acoustic or chemical cues. Sensory structures of sea lions, such as vibrissae, which have the ability of detecting small hydrodynamic trails that remain in the water column after several minutes [68] may also play a role. Future work studying mechanisms underlying the formation of multi-species predator groups could integrate abiotic data with predator sensory modalities to study what cues different predator species use for locating prey.

Sea lions herd schools of fishes [64,65] and also hunt during shallow epipelagic dives [53,55,69]. Here, however, sea lions kleptoparasitized the efforts of striped marlin when corralling and fragmenting the prey schools at the ocean's surface [30]. Jourdain & Vongraven [27] reported a similar multi-species interaction between orcas (Orcinus orca) and humpback whales (Megaptera novaeangliae) predating on herring schools (Clupea harengus). Orcas initiated the hunt of herring schools in 94% of the cases, herding the fish and making them accessible to scrounging humpback whales.

Once sea lions were hunting alongside striped marlin, they directly interfered with the latter's attack behaviour. The presence of sea lions doubled the mean time interval between striped marlin attacks and almost tripled the mean time interval between striped marlin captures. This is clear evidence of interference competition caused by the inability of striped marlin to gain access to the resource while sea lions were attacking. Once marlin managed to gain access to the prey resource, their stereotypical dash sequence behaviour and likelihood of capture per attack was not affected by the presence of sea lions. In many multi-species predator groups, multiple predators can attack the resource at the same time (e.g. [32,38,70]). Such high frequency of attacks can create the potential for all attackers, regardless of species, to benefit from the increase in prey disturbance and corralling. This effect is thought to increase if the species are phenotypically diverse, such that they attack with very different strategies or from different angles (e.g. birds attacking from above using vertical dives and teleosts attacking from below). Striped marlin and sea lions, however, attacked the prey resource one at a time, with the school of prey fishes reforming and regaining cohesion between consecutive attacks. This most likely explains why we did not find evidence for mutual benefits from the attacks of different species [38,39].

Competitive interference is stronger if one of the two species dominates access to the prey resource. Dominating prey access is potentially difficult to achieve for a mobile, highly manoeuvrable resource such as a school of prey fishes. Nonetheless, when sea lions were in groups, they appeared to dominate prey access (over 70% of marlin attacks were followed by a sea lion attack). This occurs despite marlin significantly out-numbering sea lions. Sea lions' ability to dominate prey access may be the result of their flexible and manoeuvrable body plan [71] compared to the striped marlin's limited dorsoventral flexibility. Striped marlin's attack strategy of high-speed, straight-line dashing [43] is only feasible when there are few obstacles between them and the prey resource, while sea lions can manoeuvre themselves with precision near their prey. Sea lions also had higher capture efficiencies than striped marlin. This more efficient depletion of available resources further compounds their negative effect on striped marlin hunting success.

In terrestrial systems, alleviating the threat of aggressive kleptoparasites at carcasses can be a major selective pressure promoting the formation of same-species groups among terrestrial predators [72]. This could be a potential hypothesis for why striped marlin form groups in this system—to aggressively defend the resource from kleptoparasitic sea lions. We, however, did not observe aggression behaviour from striped marlin to sea lions and propose that striped marlin group formation relates to either predator-risk dilution and/or increased capture efficiencies [48].

Despite the existence of a high number of trophic ecology studies of sea lions using scats and stable isotope analysis [51,71], little is known about direct behavioural interactions with other predator species when feeding on fish schools. Given that we never observed fish schools under attack by sea lions in the absence of marlin we could not directly compare the foraging performance of sea lions in the presence and absence of striped marlin, which could have produced additional insights. We predict that it might be difficult for sea lions to keep fish schools near the surface (where they are particularly easily caught) in the absence of striped marlin that keep pushing them up.

Another topic for future research involves the different attack strategies of striped marlin and sea lions. The attacks of marlin on fish schools are usually made by dashing through the schools from the side [43] whereas sea lions usually perform a vertical loop upwards or downwards through the fish schools. Collective defences of fish schools against vertically attacking predators have received little attention to date [73], and it is unknown whether they are more vulnerable to this type of attack strategy.

This study investigated the behavioural interactions occurring within mixed-species groups of predators, comprised of striped marlin and sea lions. The prey fish species that striped marlin and sea lions were hunting are ecological keystone species in the Southern California Current System (offshore Baja California Peninsula) [74]. The striped marlin population may be critically important for linking this resource to other predators, such as sea birds, other teleost species, potentially rorqual whales, and in this study case, sea lions. Sea lion rookeries off the Baja California Peninsula are declining owing to oceanographic warm anomalies in recent years, which poses a potential negative impact on their trophic ecology [52,75,76]. Thus, the results of this study provide important background information on the trophic ecology of sea lions from this region. Finally, the presence of many other predator species that feed on the fish schools isolated and pushed against the surface by marlin opens the possibility to further unravel the cooperative or kleptoparasitic nature of these multi-species predator associations.

Ethics

All research was conducted under permits SGPA/DGVS/02460/18, SGPA/DGVS/01643/19 and SGPA/DGVS/08074/21, and we followed the ASAB ethics (Guidelines for the treatment of animals in behavioural research and Teaching 2020, Animal Behaviour 159, i-xi) recommendations for fieldwork.

Data accessibility

The data supporting this article has been uploaded as part of the electronic supplementary material [77].

Authors' contributions

M.J.H.: conceptualization, data curation, funding acquisition, investigation, methodology, project administration, resources, visualization and writing—original draft; R.H.J.M.K.: formal analysis, visualization, writing—review and editing; M.L.: data curation, writing—review and editing; J.H.: data curation, investigation, writing—review and editing; K.P.: data curation, writing—review and editing; F.D.: data curation, investigation, methodology, writing—review and editing; F.T.: writing—review and editing; F.R.E.-V.: conceptualization, investigation, writing—review and editing; J.K.: conceptualization, data curation, funding acquisition, investigation, methodology, project administration, resources, supervision, 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

We acknowledge funding by a Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) grant awarded to (M.J.H.) GZ:HA 9403/1-1 AOBJ: 675842f, further funding by the DFG under Germany's Excellence Strategy – EXC 2002/1 ‘Science of Intelligence’ – project number 390523135 (J.K.) and funding from the Leibniz-Institute of Freshwater Ecology and Inland Fisheries to (M.J.H., F.D. and J.K.).