Effects of prey refuge and predator cooperation on a predator–prey system

Introduction Cooperative hunting, i.e., the increase in successful prey capture observed when two or more individuals engage in a hunt, has been demonstrated in a wide variety of species [1–4]. In many cooperatively hunting species, hunts can best be described as opportunistic, simultaneous individual hunts [4], in which each animal tries to maximise the probability of catching the prey for itself. True coordination, as defined in [5], exists only if individuals play different roles during a hunt. Role differentiation implies that individuals will adopt roles that have a lower probability of personal success or a higher risk of injury than other roles would offer, e.g., hunts where some individuals act as chasers while others block the escape routes of prey. Such coordination is known for only a handful of species [5–8], all of which are mammals or birds. Individual role specialisation within coordinated hunts is even more rare and has only been observed in two studies to date [7,8]. Communication between group members to initiate a coordinated search for suitable prey (for which the term “intentional hunting” has been used) is known only from a single population of chimpanzees [5]. The same population of chimpanzees is also well known for respecting prey ownership, where the successful individual shares with cohunters [5]. While simultaneous feeding on a prey carcass may also occur in carnivores, access in these species is best predicted by individual rank and/or nepotistic toleration of related lower ranking individuals [4]. Here we describe interspecific and communicative hunting between the grouper, Plectropomus pessuliferus, and the giant moray eel, Gymnothorax javanicus, observed in the coral reefs of the Red Sea. Groupers are diurnal predators, whereas the morays are nocturnal hunters and usually rest in crevices during the day. The hunting strategies of the two predators are also very different. Groupers are semi-benthic piscivors, which hunt in open water. In order to avoid predatory groupers, reef fish hide in corals (apart from pelagic prey like fusiliers). Moray eels, in contrast, sneak through crevices in the reef and attempt to corner their prey in holes. Consequently, the best strategy for prey to adopt in order to avoid moray predation is to swim into open water. The hunting strategies of the two predators are therefore complementary, and a coordinated hunt between individuals of the two species confronts prey with a multipredator attack that is difficult to avoid [9]; prey are not safe in open water because of the grouper hunting strategy but cannot hide in crevices because of the moray's mode of attack. Here we first provide some descriptive information on the interactions between the two predators (i.e., frequency, duration, and distance between partners during a joint hunt) and use a simplified version of Waser's gas model [10] to show that associations are not due to random encounters. Second, we describe the signals produced by the groupers that serve to elicit joint hunting. Third, we present experimental evidence that the production of these signals is inhibited if the grouper is satiated. Finally, we present observational evidence that both partners increase their hunting success when they are in association. We then discuss the selective conditions that might promote such an unusual interspecific cooperation. Results Evidence That Associations Are Nonrandom Because the exact number of moray eels in our study area is unknown, we compared the distribution of observed durations of associations with the value predicted from a simplified version of Waser's gas model as an acceptable compromise [10]. The gas model yields mean association durations based on the assumption of independent movements (see Methods). Associations ranged from less than 1 min up to a value of 93 min. A proportion of observed durations of interactions fit the null hypothesis of random association, which predicts durations of 100 s. However, 56% of the 207 interactions lasted longer than predicted by the null hypothesis, among which 71% lasted at least three times longer (Figure 1). Figure 1 Many Associations between Groupers and Moray Eels Are Longer than Expected by Chance Observed frequency distributions of durations (min) of interactions between groupers and moray eels. The x-axis shows different time categories that were grouped in a non-linear fashion. The arrow almost above the 2 min time category indicates the average duration of associations (100 s) predicted by a null model, assuming independent movements of individuals of the two species. Figure 2 shows the average distance between a grouper and a moray eel per minute of joint hunting for a single observation, where video filming allowed detailed analysis of this variable. The two partners stayed together at a distance of between 1 and 3 times 70 cm (which is one approximate grouper body lengths) over a period of 38 min. Figure 2 The Two Predators Remain near Each Other During Joint Hunts Mean distance given as multiples of grouper body length (estimated 70 cm) between a grouper and a moray eel per minute association, analysed on screen from a 38-min film clip of a joint hunt that was already ongoing when the camera man joined. The gap in the data is due to the camera man focussing the lens on one individual such that nothing else would be seen on screen. Arrows indicate the timing of grouper signalling. Signals Produced by the Groupers Groupers actively visit moray eels at their resting places and make use of visual signals to engage morays in a joint hunt. This involves shaking the head at high frequency (3–6 shakes per second) directly in front of the moray eel, usually few centimetres away from the moray's head (Video S1). During head shakes, the soft part of the dorsal fin is erect while the bony part is flat, apart from short-term flickers of approximately 0.1–0.3 s duration. All 14 of the groupers in our study produced this head-shaking signal at least once during observations. In 58% (n = 120) of observations, the morays responded to head shaking by leaving their crevices, and the two fish then swam off through the reef (Video S2). Moray eels were never observed to signal to groupers. The joint activity of the two fish, measured from the moment the moray left the crevice until the moray re-entered a crevice and did not re-emerge, ranged from a few seconds up to 44 min. Joint movement was often interrupted, because moray eels could remain in a crevice for several minutes before moving on. Groupers often repeated the head-shaking signal in these situations (Figure 2, arrows). Groupers did not always signal to morays when they visited them; groupers could simply pass nearby and/or lay down on the sand next to a moray. However, joint hunting was significantly more likely to occur if a grouper signalled (70 out of 120 observations) than if it did not signal (11 out of 38 observations) (χ2 = 8.4, n=158, degrees of freedom = 1, p 50 h of observations focussing on moray eels with either HF or Jürgen Schauer (Leibniz Institut für Meereswissenschaften [GEOMAR], Kiel, Germany) sitting in front of moray eels for filming purposes. With one exception, morays never moved during these film sessions and were not visited by groupers. We do not know whether these results reflect natural encounter rates with groupers from a moray perspective or whether the data are biased due to the presence of humans. Application of Waser's gas model to the grouper–moray associations. In brief, the model uses the average velocities of two groups or individuals in a 2-D space, plus a criterion for maximal distance, to calculate mean durations of associations [10]. Because we noted the movements of groupers only relative to the coastline, our calculation simplifies to a 1-D space, which yields lower swimming speeds and hence increases association durations predicted by the null hypothesis. The calculations further simplify because moray eels rarely, if ever, moved and can be assumed to be stationary objects. The crucial determinant of association durations predicted by the null hypothesis is therefore the swimming speed of groupers relative to the coastline and our association criterion of 10 m. We constructed a map of the coastline and measured distances between key landscape points. We noted the position of a grouper relative to the coastline every 15 min and, in addition, we noted each time a grouper changed direction. This information combined with the duration of each protocol allowed us to calculate the average swimming speed of each grouper relative to the coast line. The average speed used for the null model is the average speed of all groupers (one value per grouper, n = 14 individuals, unpublished data). The predicted mean association duration of the simplified model is the time it takes a grouper on average to swim 20 m, i.e., 100 s. Signalling by groupers and average distance between partners during joint hunts. To test the effects of feeding on grouper signalling, we allowed six individuals to eat a fish (purchased at the local market) at the onset of an observation session and then followed them for 120 min and recorded their behaviour following our standard observational protocol. We then compared these data with our recordings of the same individuals during observation sessions where we did not know their hunger state (matched pair design). To analyse the signals in more detail and to document the behaviours of groupers and moray eels, we filmed interactions between groupers and moray eels (all successful sequences were made following a grouper, see above). Illustrations are provided in Videos S1 and S3. In addition, we observed one 38-min sequence of a joint hunt, from which we were able to analyse on-screen the distance between the two partners during the hunt, measured as multiples of grouper body length. To calculate hunting success of groupers in association and alone, we only considered protocols where groupers were observed to hunt at least once. In addition, once a grouper had caught a fish during an observation session, the rest of the protocol was discarded, because hunting success stopped any further hunting efforts. A total of 286 h of observations were available. The groupers spent approximately 11% of that time in association with moray eels (31.5 h out of 286 h). The small sample size of observed successful hunts (n = 16 successful hunts performed by 10 different individuals) precluded statistical analysis at the individual level. This is because any analysis where the null hypothesis deviates from a 50:50 distribution (in our case it is roughly 10:90) is only useful if the number of observations is sufficiently high that predicted values for the rare situation is at least one unit. In our case, an individual with one observed successful hunt would be most likely to have achieved that success when on its own, and the most likely theoretical outcome for hunting success in association is zero. We therefore analysed all data together using a binomial test with a truncated expectation (10% of all successful hunts observed in association). The hunting success of moray eels could not be analysed in the same way as the grouper data, because solitary morays were obviously not hunting. To calculate their hunting success rate in association, we took into account the fact that morays often did not react to grouper visits by increasing their activity or they stopped searching for prey while the groupers were still motivated. Thus, groupers spend a large proportion of their time in association signalling to or waiting next to moray eels that did not leave their crevice. This time must be included to calculate grouper hunting efficiency in association, but not in the calculations for moray hunting efficiency. For the morays, we summed the number of times from first moray movement to the last moray movement during each joint hunting event. The total number of prey caught by morays during association was divided by this value to calculate the morays' hunting success rate. Supporting Information Video S1 A Grouper Signalling to a Moray Eel Resting in a Cave (1.1 MB WMV) Click here for additional data file. Video S2 Grouper and Moray Eel Swimming off Together after the Grouper Signalled (2.0 MB WMV) Click here for additional data file. Video S3 A Grouper Performing a Headstand Shaking of Its Head above the Hiding Place of a Prey That Escaped the Hunt (2.1 MB WMV) Click here for additional data file. Video S4 A Moray Approaches the Place Where the Grouper Performs Its Headstand Shaking (2.1 MB WMV) Click here for additional data file.