Complexity of the prey spectrum of Agaronia propatula (Caenogastropoda: Olividae), a dominant predator in sandy beach ecosystems of Pacific Central America

Taxonomic note

The taxonomy within the Olividae is notoriously problematic (Tursch & Greifeneder, 2001; Kantor et al., 2017). The widely distributed and very abundant O. semistriata (Gray 1839), for example, has consistently been confused in the recent literature with the equally common but more southerly distributed O. columellaris (Sowerby 1825), as reviewed by Troost et al. (2012).

In the case of the Central American Agaronia spp., we have often found the existing identification literature (Keen, 1971; López, Montoya & López, 1988; Sterba, 2004) to be ambiguous and not always helpful in the field. Following previous work (Rupert & Peters, 2011; Cyrus et al., 2012, 2015; Cyrus & Peters, 2014), we refer to the predatory Agaronia sp. observed in this study as A. propatula (Conrad 1849), albeit with hesitation.

Field observations and data analysis

Between November 2011 and April 2017, the intertidal zone of Playa Grande, Costa Rica (10°20′N, 85°51′W), was patrolled during opportunistic trips to search for A. propatula with filled metapodial pouches. Searches were performed during daylight, usually within the six hours around the time of low tide. Filled pouches almost always contained a prey item, indicating a successful predation attempt. Successful predators were picked up while burrowing into the sediment or sailing seaward in the backwash, and the prey was released by gently pressuring the metapodial pouch between fingertips (Fig. 1). The size of the predator and its prey were measured to the nearest 0.1 mm either using calipers or on images taken of predator and prey next to a ruler using ImageJ (https://imagej.nih.gov/ij/). Size was defined in gastropods as the shell length from apex to base, in bivalves as the longest axis across the shell, and in crustaceans as the anterior–posterior body length. All animals were released immediately after being measured or photographed at the same location where they had been collected. The handling period between capture and release was shorter than two minutes. A small number of predation events involving mole crabs, Emerita sp., which were observed on the beach of El Cuco, El Salvador (13°10′N, 88°06′W), in May 2011, were included in our analyzes for comparative purposes, as specified in the Results section.

The original data is available online as Table S1; Table 1 gives an overview. When the coefficient of determination (r²) suggested a possible correlation between predator size and prey size, the relationship was described by the geometric mean functional relationship (also known as standardized major axis) following Draper & Smith (1998). This was necessary since neither of the two parameters (predator size and prey size) could be considered independent variables (i.e., free of significant statistical error); in such cases conventional regression analysis does not yield meaningful descriptions of the relationship between the parameters. Binomial probability calculations and χ² tests were performed using online tools available at http://vassarstats.net/.

In a subset of the documented predation events, shelled prey items (21 O. semistriata, 14 Donax sp.) that had been removed from pouches were placed in small tanks containing sand and water immediately. Any activities of the released prey were recorded over the following 5 min.

Field studies were performed under the research permits ACT-OR-D-015 and ACT-OR-DR-064 to W.S.P. from the Ministerio de Ambiente y Energia de Costa Rica.

Modeling of expected rates of cannibalism

The expected distribution of successful cannibalization events between the different size classes of A. propatula was calculated numerically from the empirical size distribution of successful hunters under the assumption that individuals meet randomly in their habitat. Conditions for successful cannibalistic interactions empirically defined by Cyrus et al. (2015) were applied, as detailed in the Discussion section. The calculations are available online as Table S2.

Predator size distribution

The largest and smallest A. propatula we ever measured on Playa Grande had shell lengths of 60.1 and 13.2 mm, respectively, but such extreme sizes were rare. The subset of 327 A. propatula found with prey in their pouches, which is analyzed in the present study, ranged from 18.0 to 57.4 mm, with intermediate sizes from 25 to 37 mm representing a large majority (76%; Fig. 2). The size distribution of successful predators resembled a previously published distribution of A. propatula foraging in the intertidal zone at Playa Grande (Cyrus et al., 2015), although the peak at intermediate sizes seemed narrower and higher in the distribution of successful predators (Fig. 2).

Metapodial pouch contents

Metapodial pouches never contained more than one prey item. About half of the prey found in pouches were O. semistriata (Table 1). Only in two cases was another heterospecific gastropod found, Mazatlania fulgurata (Philippi 1846). Over one-quarter of the prey were bivalves, with the genus Donax (burrowing clams) dominating. Crustaceans (mostly mole or sand crabs, Emerita sp.) and conspecifics contributed the rest of the prey spectrum (Table 1). Several common inhabitants of the study beach were not recorded as prey, including lunulate sand dollars (Mellita spp.), flatworms (Plathyhelminthes), and polychaetes.

Many shelled prey released from metapodial pouches were alive. We placed 21 O. semistriata and 14 Donax sp. in small tanks immediately after they were removed from pouches. After 5 min, 19 (90%) of the gastropods and 11 (79%) of the bivalves had resumed normal activity such as crawling or burrowing. In contrast, almost all of the crustaceans found in pouches were dead or so heavily injured that they had lost the ability for controlled locomotion. At least 12 of the 43 crustaceans removed from pouches had body parts missing or were visibly damaged.

Pouches often contained sand in addition to the prey item, especially when the prey was small (<10 mm). Occasionally pouches would be found that only contained sand with no identifiable prey. These cases were not regarded successful predation attempts and therefore are not listed in Table 1.

Although useful as an overview, the total prey spectrum presented in Table 1 failed to convey the complexity revealed by our data. Prey spectra for individual A. propatula size classes spanning 4 mm of shell length indicated significant size-dependent shifts (Fig. 3). The most common prey for small (<21 mm), intermediate (21–41 mm), and large (>41 mm) size classes of A. propatula were bivalves, hetrospecific gastropods, and conspecifics, respectively. Below, the relationship between predator size and prey size is analyzed for each major prey taxon.

Bivalve prey

Bivalves were captured by A. propatula of all sizes (Fig. 4A). The largest predated bivalves (Pitar sp., 39.4 mm; Donax sp., 38.8 mm) were similar in size to the largest non-predated bivalve (Donax sp., 42 mm) that we ever observed alive on Playa Grande (Fig. 4B). There was a trend for larger A. propatula to capture larger bivalves (r² = 0.336; Fig. 4B). In fact, the shell length of bivalve prey often was similar, and in one case even larger than that of the gastropods that had captured it (Fig. 4B). At the other end of the size spectrum, bivalves of under 3.7 mm shell length were notably absent (see the field marked “I” in Fig. 4B). This is interesting as bivalve prey immediately above this size limit were found frequently; one-fifth of all captured bivalves were between 3.7 and 4.7 mm.

Crustacean prey

Crustaceans were captured by most size classes of A. propatula (Fig. 5A). Similar to bivalve prey, there was a sharp lower limit of prey size at just below 4 mm (field I in Fig. 5B). There also was a lower limit of predator size at about 23 mm (note lack of data in field II in Fig. 5B). While most prey items ranged between 6 and 15 mm, two prey items exceeded 32 mm and were bigger than their prey. Intermediately sized prey was lacking (no data in field III; Fig. 5B). Consequently, there was no correlation between predator and prey size when the entire dataset was considered (r² = 0.016), while the analysis of the subset of events involving small prey only (<20 mm) hinted at a possible trend for larger predators to capture larger prey (r² = 0.178; Fig. 5B).

The two isolated cases where prey exceeded 30 mm involved mature Emerita sp.; one carried an egg mass under the abdomen. A similar pattern of size ratios between A. propatula and their Emerita sp. prey was found on the beach at El Cuco in El Salvador (black diamonds in Fig. 5B). Combined, the two datasets show an isolated group of six Emerita that exceeded the shell lengths of their predators (Fig. 5B). Two of the six large Emerita were dead, and the others proved incapable of burrowing into the sediment or any kind of locomotion after being removed from the metapodial pouches. All large Emerita had been captured by relatively small A. propatula (26.5–35.0 mm), whereas larger predators were found with small crustacean prey but never with mature Emerita (Fig. 5B; note lack of data in field IV).

Gastropod prey (excluding cannibalism)

Cannibalism excluded, O. semistriata was the only gastropod captured by A. propatula at a significant frequency (163 out of 165 observations involving gastropod prey; Table 1). The sizes of predators and of their gastropod prey were not correlated (r² = 0.081). Our data was in line with the threshold value of the predator-to-prey size ratio for predation on gastropods postulated by Cyrus et al. (2015): with the exception of a single outlier, the shell length of successful predators always was more than 1.45 times that of their prey (Fig. 6B). As with bivalve and crustacean prey, there was a sharp lower limit of gastropod prey size (field I, Fig. 6B).

Large A. propatula seemed unsuccessful in preying on other gastropods, as documented in Fig. 6A and by the paucity of data-points in field II in Fig. 6B. If successful predation attempts on O. semistriata were randomly distributed across the observed size spectrum of successful A. propatula (Fig. 2), we would expect about 21 predation events to plot in field II in Fig. 6B, rather than just two. The probability for the observed result to occur by chance was negligible (p < 10^–6; binomial probability calculation). Finally, it was surprising that O. semistriata of over 19 mm shell length were hardly ever found in A. propatula’s pouches (field III in Fig. 6B), although large specimens of up to 22 mm were quite common at Playa Grande.

Conspecific prey (cannibalism)

Cannibalism was conducted mainly by A. propatula over 41 mm, and only 3 of 31 cannibalization events were conducted by predators smaller than this threshold (Fig. 7A). The prey in these three events were among the smallest A. propatula observed in this study and ranged from 13.4 to 18.7 mm. There was a weak correlation between predator and prey size (r² = 0.460; Fig. 7B), but this was strongly influenced by the three events involving small cannibals. When these three outliers were omitted from the analysis, no correlation was evident (r² = 0.041). As was the case for heterospecific gastropod prey (Fig. 6B), successful cannibals were at least 1.45 times larger than their conspecific prey (Fig. 7B). We never observed autotomy in the victims of cannibalism.

Bivalve prey

Bivalves under 15 mm shell length were most commonly captured by intermediately sized A. propatula between 25 and 35 mm (Fig. 4B), which likely reflects the large proportion of this size group in the population of the predator. Larger A. propatula also caught small bivalves, and the largest predator in our sample, an animal of 57.4 mm, was found with a bivalve of only 9.6 mm in its pouch (Fig. 4B). On the other hand, with the exception of a single predation that was below the 1:1 predator-to-prey size ratio (right of the dashed line in Fig. 4B), A. propatula only captured bivalves when it was larger than the potential prey (data-points left of the dashed line in Fig. 4B). In other words, the size of the largest bivalves captured increased with predator size. We interpret this pattern as a predator size-dependent widening of the “predation window” (Claessen et al., 2002), characterized by an increase of the variance of prey size as the predator grows larger (Peters, 1983). This phenomenon occurs when the growth of a predator enables it to subdue increasingly large prey. If the growing predator also retains a capacity for capturing small prey—that is, if the maximum size of accessible prey increases with predator size more rapidly than minimum prey size does—predator and prey size will be correlated and the variance of prey size will increase with predator size. Predation window widening with increasing predator size is not uncommon (Scharf, Juanes & Rountree, 2000; Dörner & Wagner, 2003; Chabot et al., 2008; Polidori et al., 2009). The interaction between A. propatula and its bivalve prey (Fig. 4B) adds another example.

Agaronia propatula appears to exploit the entire size range of bivalve prey available at the study beach, since the largest bivalves (∼40 mm) found in metapodial pouches were similar in size to the largest individuals found free-living at the site over the six-year observation period. At the lower end of the bivalve prey size distribution, the sharp limit at 3.7 mm is intriguing (Fig. 4B, field I). Prey size can be a limiting factor for predation efficiency not only if the prey is too large to be subdued by a predator, but also if potential prey items are too small to be detected or handled efficiently (Claessen, de Roos & Persson, 2000; Claessen et al., 2002; Brose et al., 2008; Costa, 2009). We frequently observed small bivalves in metapodial pouches together with significant amounts of sand. This suggests that A. propatula can handle small prey more successfully if the item is taken together with some of the surrounding sediment (compare Cyrus et al., 2015). If so, completed but unsuccessful predation attempts may result in predators carrying pouches with sediment only, being unaware of the lack of edible victims in their catch. This probably explains the occasional occurrence of metapodial pouches filled with sand but without prey. The observed lower size limit of bivalve prey may therefore suggest that A. propatula are unable to handle prey of less than 4 mm efficiently. A problem for this interpretation is that the lower limit of prey size did not seem to correlate with predator size (Fig. 4B), which we would expect if mechanical difficulties in manipulating small objects were responsible for the limit. A more plausible explanation may instead be that the lower limit of the bivalve prey spectrum reflects the lower limit of the size distribution of the available bivalve prey.

Taken together, our results appear to support the view that A. propatula indiscriminately exploits the bivalve population in its habitat, limited only by decreasing success rates of attacks when the predator-to-prey size ratio approaches unity.

Crustacean prey

There was no clear correlation between the size of A. propatula and their crustacean prey, and no unambiguous predation window widening (Fig. 5B). Most of the crustaceans found in metapodial pouches were dead or injured, and generally unable to move. This may explain the otherwise surprising fact that mole crabs and other crustaceans, which in a healthy state should be sufficiently agile to evade an approaching A. propatula, form a significant portion of the gastropod’s prey spectrum. In fact, small crustaceans frequently were seen scrambling from their hiding places under the sediment surface when an A. propatula drew close, moving rapidly out of the predator’s path and burrowing in again. Thus it seems that only dead, injured, or otherwise weakened crustaceans fall victim to A. propatula. This is in remarkable contrast to shelled prey, which in the majority of cases (gastropods: 90%; bivalves: 79%) resumed normal activities within 5 min of being released from metapodial pouches. Similarly, victims of cannibalistic attacks survived and retained locomotory capabilities when confined in pouches for many hours (Cyrus et al., 2015). We conclude that A. propatula, while being a predator of mollusks, acts as a facultative scavenger of crustacean prey.

Cyrus et al. (2012) put forward a similar argument regarding predation events involving very large Emerita sp. at El Cuco, El Salvador, suggesting that these mole crabs were immobile because they had recently molted. One may object that egg mass-carrying crustaceans—such as one of the Emerita sp. that we found in a pouch—are mature and therefore unlikely to just have undergone molting, but molt and reproductive cycles can alternate in short succession in the genus Emerita (Subramoniam & Gunamalai, 2003). Nonetheless, the argument by Cyrus et al. (2012) does not explain the lack of intermediately sized crustacean prey (field III in Fig. 5B) and the resulting isolated position of the cluster of events in which prey items were bigger than their predators (data-points right of the dashed line in Fig. 5B). We feel that the predatory interactions between A. propatula and Emerita sp. are too complex to be elucidated satisfactorily by data currently available.

Gastropod prey (excluding cannibalism)

There was neither a correlation between the sizes of predators and O. semistriata (r² = 0.081) nor an indication of predation window widening (Fig. 6B). The distribution of the data raised at least three questions. First, it remains unclear why O. semistriata of under 5 mm shell length never were found in pouches (field I in Fig. 6B). The developmental biology of the species including its life history as well as the size distribution of the population on the beach and its changes over the year will have to be determined to evaluate possible causes of this finding.

Second, the near complete absence of O. semistriata from the prey spectrum of A. propatula over 40 mm (field II in Fig. 6) cannot be explained by a lack of interest in this prey species from large A. propatula. In preliminary field experiments, we placed O. semistriata in the path of foraging A. propatula and found similar attack rates with predators of above and below 40 mm (S. D. Rupert & W. S. Peters, 2016–2018, unpublished data). We suggest an alternative explanation that is related to the tidal migrations both species perform by passive locomotion with the waves (Troost et al., 2012; Cyrus et al., 2012). Large A. propatula are most frequently found in the lowest part of the intertidal zone (A. Z. Cyrus, J. Swiggs & W. S. Peters, 2011–2018, unpublished observation), suggesting ontogenetic modifications of the species’ tidal migration behavior. If large predators remain confined to a zone around the low water line throughout the tidal cycle, they may rarely interact with O. semistriata, which need to follow the moving backwash zone to perform suspension feeding (Troost et al., 2012). If this is correct, we expect to find a dynamic separation zone between O. semistriata and large A. propatula that shifts constantly along the beach slope with the tides. Testing the hypothesis will require establishing size profiles of A. propatula along the beach slope at high temporal resolution.

Third, there was a threshold of the predator-to-prey size ratio for successful predation on gastropods, as suggested by Cyrus et al. (2015): A. propatula had to be at least 1.45 times larger (regarding shell length) than its gastropod prey for successful capture (Figs. 6B and 7B). However, our data indicate that the trophic relationship between A. propatula and O. semistriata is influenced by an additional size-dependent factor. With a size ratio threshold of 1.45, more than half of the A. propatula in our sample of successful predators (Fig. 2) should be capable of subduing O. semistriata of 19–22 mm shell length, a common size class at the study site. Therefore we should expect numerous data-points in field III in Fig. 6B. Intriguingly, this field is almost empty. This surprising finding is unlikely to be caused by a differential distribution of large O. semistriata (>19 mm) and foraging, appropriately sized A. propatula (>30 mm) on the beach, as we routinely observed them together in close vicinity. It rather seems that O. semistriata can outgrow the predation pressure exerted by even the largest A. propatula, independently of the predator-to-prey size ratio. This could be explained if the efficacy of flight responses in O. semistriata depended on body size in a non-linear manner. The validity of this hypothesis will have to be tested in the field by measurements of the velocities of differently sized animals in various predation-related situations.

Conspecific prey (cannibalism)

Cyrus et al. (2015) concluded that cannibalism is a significant factor in the population ecology of A. propatula. Our enlarged dataset even points to a higher proportion of cannibalism in the prey spectrum (9.5% of all documented predation events; Table 1) than that reported by Cyrus et al. (5.6%).

In many but not all cannibalistic taxa, cannibalism is size-dependent with larger individuals preying on smaller ones (Polis, 1981). Among molluscs, this holds true for cephalopods (Ibáñez & Keyl, 2010) and shelled but not necessarily unshelled gastropods (Baur, 1992). Results from field experiments in which the size-dependence of the success rate of cannibalistic aggression in A. propatula was characterized, suggested three simple rules (Cyrus et al., 2015). First, cannibalistic attacks will be successful if the size ratio, expressed as shell length of the predator divided by the shell length of the victim, is at least 1.46. Second, the attacks will be unsuccessful if the ratio is less than 1.18. Third, if the ratio is intermediate (i.e., at least 1.18 but below 1.46), half of the attacks will be successful. Applying these rules, Cyrus et al. (2015) estimated the decline of intraspecific predation pressure a snail experiences as it is growing, demonstrating a selective pressure favoring rapid development.

It is tempting to interpret the dominance of large individuals among the successful cannibals (particularly conspicuous in Fig. 7A) as an effect of this size-dependence: larger snails are more efficient cannibals after all, so they must be over-represented in the group of successful cannibals. Our data, however, indicate an influence of additional factor(s). If we consider a group of A. propatula with the size distribution of our sample of 327 successful predators (Fig. 2; see original data in Table S1), and further assume that these animals meet randomly on a beach (i.e., the probabilities for a given snail to encounter one of its 326 conspecifics are identical for all possible pairings), then we can apply the three rules of Cyrus et al. (2015) to generate the expected size distribution of cannibals in our sample (for details, see Table S2). This expected distribution exhibits a bimodal shape with two peaks (Fig. 8, gray bars). The peak in the large size classes is caused by the high success rate of big animals in cannibalistic interactions—large specimens are a small minority in our sample (compare Fig. 2), so conspecifics they encounter are almost always much smaller and therefore subdued easily. The peak at intermediate sizes (above 30 mm shell length) is caused by the large proportion of intermediately sized animals in our sample. While their success rate in cannibalism is low, their number is high, so that they appear prominently in the expected size distribution of successful cannibals.

The expected size distribution differs remarkably from the observed distribution, which for comparison is also shown in Fig. 8 (white bars). Compared to expectations, individuals of less than 40 mm shell length are under-represented while those over 40 mm are over-represented among the observed cannibals. The difference between the expected and the observed distributions (Fig. 8) is statistically significant at p < 0.01 (χ² = 22.191, df = 10). This finding is unlikely to result from an over-estimation of the intraspecific aggressiveness of small (<40 mm) A. propatula in our model calculations, as the behavioral field experiments by Cyrus et al. (2015) compellingly demonstrated that animals of all sizes will be cannibals if they can. The proportional mismatch between the expected and the observed size distributions (Fig. 8) could be explained more convincingly if our assumption of random encounters between the individuals were incorrect. Above we had suggested that certain features of the predator–prey size relationship of A. propatula and O. semistriata could be explained by ontogenetic modifications of the tidal migration behavior of A. propatula. We now hypothesize that the resulting differential distribution of the size classes of A. propatula along the beach slope also affects the encounter rates between the size classes, and thus the frequency of cannibalism in the different size classes. An accurate characterization of the localization of animals of different sizes throughout the tidal cycle will be required to develop this idea further.