Effects of dams on downstream molluscan predator–prey interactions in the Colorado River estuary

Abstract

River systems worldwide have been modified for human use and the downstream ecological consequences are often poorly understood. In the Colorado River estuary, where upstream water diversions have limited freshwater input during the last century, mollusc remains from the last several hundred years suggest widespread ecological change. The once abundant clam Mulinia modesta has undergone population declines of approximately 94% and populations of predators relying on this species as a food source have probably declined, switched to alternative prey species or both. We distinguish between the first two hypotheses using a null model of predation preference to test whether M. modesta was preyed upon selectively by the naticid snail, Neverita reclusiana, along the estuary's past salinity gradient. To evaluate the third hypothesis, we estimate available prey biomass today and in the past, assuming prey were a limiting resource. Data on the frequency of drill holes—identifiable traces of naticid predation on prey shells—showed several species, including M. modesta, were preferred prey. Neverita reclusiana was probably able to switch prey. Available prey biomass also declined, suggesting the N. reclusiana population probably also declined. These results indicate a substantial change to the structure of the benthic food web. Given the global scale of water management, such changes have probably also occurred in many of the world's estuaries.

1. Introduction

Nearly two-thirds of the world's major rivers have been captured, diverted or otherwise modified for human use [1]. Given the world's growing population, the utilization of riverine resources (e.g. hydroelectric power, potable water) will remain high for the foreseeable future [2–4]. At the same time, the downstream ecological implications (i.e. trade-offs) of upstream water management decisions are not well understood, often for a lack of pre-management data [5–7]. That is, we have an incomplete accounting of how management decisions (e.g. dam construction, water diversions) affect downstream communities and species interactions, particularly in estuaries. In estuarine environments, where sedimentation rates are high and many organisms (e.g. molluscs) have hard skeletal components, pre-management data may still be within reach (e.g. [8–10]).

In the Colorado River estuary (CRE), for example, accumulations of molluscan remains have been used to better understand the community that existed prior to widespread damming and water diversion along the Colorado River during the twentieth century [10–12]. Much of this previous work focused on the once dominant clam Mulinia modesta (formerly M. coloradoensis [13]). Densities of M. modesta exceeded 50 individuals m⁻² during the pre-dam era—defined as the time period prior to the beginning of dam construction in the 1930s—and are scarce on the tidal flat today [9,10,12,14]. Consequently, M. modesta is no longer the most abundant clam in the CRE [9,10], and its role in estuarine carbon cycling has diminished proportionately [7]. Furthermore, upon examination of the shells of 600 pre-dam era M. modesta individuals for traces of predation at Las Isletas (figure 1), Cintra-Buenrostro et al. [11] found 23% had drill holes made by predatory snails (naticids and muricids; figure 1d), 27% had damage on the margins of their shells consistent with successful crab predation and 34% had repair scars from unsuccessful crab attacks. These results led Cintra-Buenrostro et al. [11] to propose three related hypotheses based on the importance of M. modesta as a food source for snails and crabs. They hypothesized M. modesta population declines probably resulted in corresponding declines in the populations of predatory species (hypothesis 1), switching by predators to alternative prey species (hypothesis 2) or a combination of both effects (hypothesis 3).

Figure 1.

(a) Location of the CRE with (b) an inset showing the study sites. (c,d) Specimens of the (c) drilling snail Neverita reclusiana and (d) once-common clam Mulinia modesta. Note the predatory drill hole near the umbo of M. modesta. IM, Isla Montague; LI, Las Isletas; CDA, Campo don Abel. (Online version in colour.)

Here we evaluate the preference—defined as deviation from a random diet—of the shell-boring naticid Neverita reclusiana (figure 1c) for M. modesta, with respect to potential alternative prey in the pre-dam era molluscan community, to differentiate between these hypotheses. As with many naticid species, N. reclusiana use a combination of acidic secretions and a scraping radula to bore characteristic drill holes into the shells of their prey (figure 1d) [15]. These trace fossils are readily preserved and record a reliable record of past predator–prey interactions [16]. Although naticid drill holes made by individuals of different species can be challenging to differentiate [17] and two naticid species are present in the CRE today, N. reclusiana was probably the only naticid present in the pre-dam era [18]. Thus, using the trace fossil record of N. reclusiana and a metric of prey preference by predators (Manly's α [19–21]), we evaluate the potential ecological effects of dam construction and water diversion on predator–prey interactions in the CRE. If N. reclusiana disproportionately preyed on M. modesta (i.e. a non-random diet composed mainly of M. modesta) then the N. reclusiana population will probably have declined alongside the M. modesta population (hypothesis 1). Alternatively, if N. reclusiana selectively consumed other prey, in addition to M. modesta, it is more likely that switching from M. modesta will have occurred, as N. reclusiana compensated for the population decline of M. modesta (hypothesis 2). Because support for hypothesis 2 cannot conclusively rule out the possibility of a N. reclusiana population decline, we subsequently use estimates of past and present prey densities on the CRE tidal flat to assess whether the potentially supportable population size of N. reclusiana has changed (hypothesis 3).

2. Methods

(a). Sampling

Samples were taken from the molluscan death assemblages at three sites in the CRE—Isla Montague, Las Isletas and Campo don Abel—following the north–south salinity gradient that existed prior to widespread damming and water diversions in the Colorado River basin, which began in the 1930s [22] (figure 1). Shells in the death assemblages are time-averaged, but dating via amino acid racemization of Chionista fluctifraga has shown that more than three-quarters of shells originated 100–300 years ago, during the pre-dam era [23]. The death assemblage samples were collected from cheniers—sedimentary deposits that are formed through shoreward tidal movement and accumulation of large clasts (e.g. shells) and the removal of smaller clasts (e.g. clays, silts) by longshore currents [24]—at each site. Three samples (approx. 10 l) were collected at Isla Montague and Las Isletas, and five (approx. 2 l) were collected at Campo don Abel. Samples were taken at random at a spacing of approximately 30 m from the top 10 cm of chenier surfaces (approx. 0.25–0.50 m²), which are well mixed with the subsurface [23]. All samples were wet sieved using a 5 mm mesh in the laboratory, all individuals were identified to the species level, and all predatory drill holes were tallied. Only specimens judged to be at least 85% complete were included and clam count totals were halved to estimate their abundance to account for each individual specimen having two elements (i.e. left and right valves of the shell [25]). For each sample, species with more than 150 specimens were randomly subsampled (see electronic supplementary material, S1). Because drilling predation by species of the snail family Muricidae can be readily confused with naticid drill holes in thin-shelled clam prey [26], all drill holes were measured across their outer diameter and their position on the prey shell recorded to distinguish between predators. Drill holes that were small (less than 1.0 mm in outer diameter) or did not have the characteristic countersunk appearance of a naticid drill hole were not included in the analysis as there is greater likelihood that they were made by a muricid rather than a naticid snail [25,26]. If after this filtering there was still ambiguity as to the origin of the drill hole, the location of the drill hole on the shell was evaluated because naticid predation is highly stereotyped in site selection (e.g. [27])—particularly compared with the most common drilling muricid, Eupleura limata, in the CRE (J.A.S. & G.P.D. 2014, personal observation) [26]—such that drill holes are commonly found near the umbo in clam prey (figure 1d).

(b). Analysis of preference

We applied the preference metric proposed by Manly et al. [19] and further developed by Chesson [20,21] (see also [28])—hereafter referred to as Manly's α—to distinguish between the first two hypotheses. Manly's α incorporates drilling frequency and normalizes species-specific drilling frequencies for the entire community:

2.1

where r_i is the number of prey type i that are drilled, n_i is the number of individuals from species i in the community and m is the number of prey types [29]. To calculate a normalized α, species-specific r_i/n_i (i.e. drilling frequency) is divided by the sum of all species' drilling frequencies [21]. If the predator only consumes prey type i, the value of α_i will be 1, whereas if prey type i is always avoided the value of α_i will be 0. If a predator is not selective, α_i will be equivalent to 1/no. prey types in the community. We applied this non-selective scenario as a null model for predator preference to test the hypothesis that M. modesta was the preferred prey of N. reclusiana during the pre-dam era in the CRE.

At each site, specimen and drill hole counts for those species that were drilled at least once were used to calculate species-specific Manly's α values according to equation (2.1). As such, the results discussed below provide a conservative estimate of N. reclusiana preference because the removal of species not likely to be in the diet of N. reclusiana increases the null threshold () used to evaluate the preference for each species. Independently at each site, species-specific alphas were compared to the null α value () using Bayesian posterior distributions to evaluate the probabilities that naticids exhibited preference for the respective prey species. Additionally, using a Bayesian formulation, each species's α was compared with the M. modesta α value to give the probability that M. modesta was more greatly preferred than the second species. Low species-specific sample sizes (e.g. n < 25), which commonly occur in palaeo-communities and in the CRE dataset, can make interpretations difficult due to the high degree of uncertainty associated with small samples [25]. Although it does not completely alleviate the issue, the Bayesian formulation employed here provides more information than a traditional, frequentist approach, including true probability statements derived from posterior distributions and credibility intervals [30]. Analyses were conducted in R (see [29] for R code and a discussion of practical considerations when applying Manly's α to palaeo-ecological data).

(c). Molluscan biomass in the Colorado River estuary

Estimates of N. reclusiana densities have not previously been made for the pre- or post-dam era. Therefore, in order to evaluate hypothesis 3, that N. reclusiana switched prey and its population declined, we used estimates of prey density on the CRE tidal flat and naticid energetic needs to estimate the maximum sizes of N. reclusiana populations that could potentially be supported during the pre- and post-dam eras. We restrict prey here to clams, as data on snails are limited. After accounting for breakage of shells in the cheniers, Kowalewski et al. [10] estimated densities of large (greater than 12.5 mm) M. modesta in excess of 50 individuals m⁻² in the pre-dam era. By contrast, today, the density of large clams—predominantly C. fluctifraga and M. modesta [9]—is approximately 3 individuals m⁻². Given that the CRE tidal flat is approximately 1.2 × 10⁸ m² [10], a density of 50 clams m⁻² is equivalent to a population of 6.0 × 10⁹ clams. Comparatively, a density of 3 clams m⁻² amounts to a population of 3.6 × 10⁸ clams (see electronic supplementary material S2 for associated assumptions).

Using estimates of naticid energetic requirements from the literature for a phylogenetically closely related species, Neverita duplicata, it is possible to estimate the number of N. reclusiana that could be sustained on those prey populations, assuming prey were a limiting resource and that all other variables (e.g. pathogens; predation on N. reclusiana) remained constant. Studying the western Atlantic species N. duplicata, Edwards & Huebner [31] estimated annual energetic requirements of approximately 385 kJ for a large individual (approx. 39 mm in maximum diameter) or 218 kJ for a small individual (approx. 25 mm in maximum diameter). Using a prey size of 25 mm in shell length—a prey size both ‘small' or ‘large' naticids could probably consume—we apply an energetic value of 3.09 kJ per individual for M. modesta and 3.31 kJ per individual for C. fluctifraga (see electronic supplementary material S2 for calculations and conversion factors) to estimate the size of the N. reclusiana population that could be supported during the pre-dam era and today.

3. Results and discussion

(a). Predator preference in the Colorado River estuary

Neverita reclusiana exhibited preferences for multiple prey species at each of the three sites in the pre-dam era CRE and those preferences were variable among sites (figure 2). These results (see electronic supplementary material S3, table S1 for species-specific data) were confirmed when considering all species in the analysis regardless of whether they were drilled (electronic supplementary material S3) and only the subset of species shared between sites (electronic supplementary material S4).

Figure 2.

Neverita reclusiana preferences (primary axis) for molluscan prey species at (a) Isla Montague, (b) Las Isletas and (c) Campo don Abel, and probabilities that Mulinia modesta, indicated by the red arrow in each panel, was more preferred than the other prey species (secondary axis, red triangles). Green shading indicates prey species were likely to be more preferred than predicted by the null condition (). Yellow shading indicates preference was indistinguishable from the null condition. Red shading indicates prey species were likely to be less preferred than predicted by the null condition. The dashed line in each panel represents the site-specific null α value and the black bars around the estimates of Manly's α give the 95% credibility interval.

At Isla Montague, which had the lowest average salinities due to its northern position near the mouth of the Colorado River [22], nine species were included in the analysis and M. modesta was a preferred prey (α_M.m. = 0.205; probability α_M.m. > α₀ = 0.827; α₀ = 0.111). Additionally, the Bayesian formulation of α values found that Eupleura limata (α_E.l. = 0.209; probability α_E.l. > α₀ = 0.830) and Felaniella sericata (α_F.s. = 0.386; probability α_F.s. > α₀ = 0.872) were preferred (figure 2a). The probabilities that M. modesta had a higher α value than the latter two species were relatively low (probability α_M.m. > α_E.l. = 0.477; probability α_M.m. > α_F.s. = 0.302), supporting the conclusion that alternative prey species were preferred along with M. modesta.

Thirteen species were included in the analysis of N. reclusiana preference at the middle site, Las Isletas. Several species were preferred (α₀ = 0.077)—Chionopsis gnidia (α_C.g. = 0.239; probability α_C.g. > α₀ > 0.999), E. limata (α_E.l. = 0.123; probability α_E.l. > α₀ = 0.957), Lamelliconcha concinnus (α_L.c. = 0.137; probability α_L.c. > α₀ = 0.794) and Cosmioconcha palmeri (α_C.p. = 0.106; probability α_C.p. > α₀ = 0.780)—however, M. modesta was not among them (α_M.m. = 0.081; probability α_M.m. > α₀ = 0.571; figure 2b). Furthermore, all four species were probably more preferred than M. modesta by N. reclusiana, as probabilities of M. modesta having a greater α value were less than 0.001, 0.038, 0.242 and 0.257, respectively. The results from Las Isletas support the second hypothesis that N. reclusiana preferred prey species other than M. modesta.

Mulinia modesta (α_M.m. = 0.035; probability α_M.m. > α₀ = 0.017; α₀ = 0.056) was not a preferred prey species of N. reclusiana at the southernmost site, Campo don Abel, where salinities during the pre-dam era were approximately normal marine (i.e. 34 psu) during much of the year [22]. Of the remaining 17 prey species, four had α values indicating a greater than 70% probability of being preferred and an additional five species were likely to be preferred over M. modesta (figure 2c). With respect to the hypothesis being tested here, that M. modesta is more preferred than other potential prey species in the community (hypothesis 1), the evidence from these prey species (e.g. C. gnidia: α_C.g. = 0.099, Probability α_C.g. > α₀ = 0.969, probability α_M.m. > α_C.g. < 0.001; L. concinnus: α_L.c. = 0.085, probability α_L.c. > α₀ = 0.737, probability α_M.m. > α_C.p. = 0.078) suggests that many other species were preferred over M. modesta.

In the pre-dam era CRE, M. modesta was one of the only preferred prey species in the north, but was not a preferred prey species at the middle or southern sites (figure 2). The increase in salinity from north to south during the pre-dam era [22] probably explains the observed differences in N. reclusiana preference for prey (see electronic supplementary material S4 for discussion and dismissal of alternative explanations of the preference trend).

As described by the estuarine quality paradox [32–34], brackish environments often exclude species because the natural conditions in those environments are physiologically unfavourable [35,36]. Likewise, naticid predation intensity has been shown to decrease as salinities approach brackish conditions (e.g. [37,38]). Accordingly, community-wide drilling frequency and prey species richness in the CRE increased southwards, away from the brackish salinities (figure 3). Drilling frequency on M. modesta followed a similar trend but α values for M. modesta did not (figure 3). Mulinia modesta, like the congeneric Atlantic M. lateralis [39], is an opportunistic species that thrives in disturbed habitats, such as the brackish CRE [12], setting it apart from many of the species in the CRE metacommunity that were absent in the north. With the southwardly increasing prey richness and predation intensity, a more complete picture of N. reclusiana prey preference presents itself. Mulinia modesta was probably preferred in the north due to the scarcity of more highly preferred prey species (i.e. n < 10; see electronic supplementary material S3, table S1 [29]). Moving to the southernmost site, M. modesta was still abundant in the community (27% [12]), suggesting that it should have been commonly encountered by N. reclusiana. Yet, more favourable prey (e.g. more energetically favourable; sensu [40]) became available to N. reclusiana as the environmental conditions became more amenable to a wider range of prey species. Indeed, the M. modesta α value (0.035) indicates that, if all species present in the south were equally abundant, M. modesta would only comprise 3.5% of the N. reclusiana diet. By contrast, when N. reclusiana had fewer species to select from in the north, M. modesta would have contributed 20.5% to the N. reclusiana diet (figure 3). Based on these results for the preference of N. reclusiana, it is likely that this predatory species was able to switch to alternative prey species as the M. modesta population declined during the post-dam construction era, supporting hypothesis 2.

Figure 3.

Summary of Neverita reclusiana predation on M. modesta (α, drilling frequency) compared with community-wide drilling frequency and richness of prey species along north–south salinity gradient in the CRE. Vertical black bars indicate the null condition for α values at each site. (Online version in colour.)

(b). Molluscan biomass in the Colorado River estuary

The analysis performed here clearly demonstrates N. reclusiana had preferences for a variety of prey species in the pre-dam era and it was probably capable of switching to alternative prey (hypothesis 2). Preference for alternative prey is not, however, sufficient evidence to dismiss the possibility of a concurrent reduction in the N. reclusiana population. Using estimates from the literature on clam abundance and energetics (electronic supplementary material S2), we found that a naticid must consume between 66 (all C. fluctifraga, small naticid) and 125 (all M. modesta, large naticid) prey individuals each year to meet its minimum energetic requirements. Assuming a constant M. modesta population (50 individual m⁻²) and applying these values to the total tidal flat area (1.2 × 10⁸ m² [10]) yields a maximum pre-dam era density of 0.40–0.71 naticids m⁻², or a standing population of 4.82 × 10⁷–8.51 × 10⁸ individuals. Comparatively, the clam population on the tidal flat today (3 individual m⁻²)—predominantly C. fluctifraga and M. modesta [9,10]—can only support a maximum density of 0.02–0.05 naticids m⁻², or a population of 2.89 × 10⁶–5.47 × 10⁶ naticids. These estimates are only a conservative approximation of the naticid population, as they do not incorporate snail prey species and the parameters used in the calculations were not derived from CRE species. Nonetheless, these estimates demonstrate the potential implications of the decline in clam density in the CRE if prey were a limiting resource. Specifically, it is highly likely that the N. reclusiana population has also declined in response to the reduction in clam biomass on the CRE tidal flat, supporting hypothesis 3.

(c). Ecological change in the Colorado River estuary

Our analyses suggest that N. reclusiana is very likely to have switched to alternative prey species in the absence of M. modesta and its population has probably also declined due to a reduction in biomass of potential clam prey on the CRE tidal flat. With respect to the hypotheses proposed by Cintra-Buenrostro et al. [11], it would seem that their third option, ‘both effects', has the most support. We have, of course, only evaluated the hypothesis for one of the three groups of predators considered by Cintra-Buenrostro et al. [11], with muricid snails and crabs remaining unstudied. The response of the shell-boring E. limata is probably most similar to N. reclusiana given the mode of predation. The response of other predators—including the shell-grinding muricid, Hexaplex nigritus, and shell-crushing or shell-peeling crabs—to the reduction in clam biomass may differ due to differences in predatory behaviours, perhaps with greater effect depending on their prey preferences. As the thick-shelled C. fluctifraga has become the most abundant clam in the CRE [9], predatory behaviours that were sufficient for thin-shelled M. modesta prey may no longer be effective. Indeed, Smith & Dietl [18] reported the human-induced range expansion of the naticid snail, Notocochlis chemnitzii, into the CRE and its novel utilization of edge-drilling behaviour to more efficiently drill thick clam prey at their thinner shell margin rather than through the relatively thick umbonal region. Although none of the species discussed here have become locally extinct since the pre-dam era, their relative abundances and strengths of interactions in the CRE have undeniably changed. Given the importance of snail and crab predators in the benthos [38,41–44], it is highly likely that the entire food web has been affected.

4. A tangled web of altered estuarine interactions for the world's major river systems

Just as upstream water management decisions along the Colorado River have altered species interactions and food web dynamics in the downstream CRE, there have probably been substantial ecological consequences in the estuaries of other major rivers that have been altered for human use [6]. As with the CRE, pre-impact ecological data were not recorded in most estuaries but are probably attainable through the utilization of the geohistorical data recoverable from molluscan remains [8,9]. Given that estuaries tend to be highly productive ecosystems and are consequently of economic importance [45], understanding shifts in estuarine food web dynamics resulting from past, present and future water management decisions may have profound implications for the people relying on estuarine ecosystems for goods (e.g. shellfisheries for food) and services (e.g. nutrient cycling). If, in the future, society chooses to attempt the restoration of these estuarine ecosystems, or elects to alter them further, species interactions, not just species abundances and distributions, must be considered.

Data accessibility

The datasets supporting this article have been uploaded as part of the electronic supplementary material.

Authors' contributions

J.A.S. collected field data, carried out the sample processing, participated in the conception and design of the study, and drafted the manuscript; J.C.H. carried out the statistical analyses; G.P.D. collected field data; participated in the conception and design of the study, and helped draft the manuscript. All authors gave final approval for publication.

Competing interests

We have no competing interests.

Funding

This work was supported by funding to G.P.D. (National Science Foundation EAR 1420978) and J.A.S. (Cornell University's Atkinson Center for a Sustainable Future; Geological Society of America; Paleontological Society; Sigma Xi).