INTRODUCTION
Because saline coastal waters are a barrier to dispersal by freshwater-adapted organisms, streams of coastal watersheds tend to have fish assemblages that are insular with low richness compared with streams within large continental basins (Hugueny Reference HUGUENY1989). Certain species of marine crustacean and fish invade tropical coastal streams where they fill vacant freshwater niches and are integral components of food webs (Covich & McDowell Reference COVICH, MCDOWELL, Reagan and Waide1996, Winemiller Reference WINEMILLER1990). Migratory shrimp have been shown to play important roles in food webs of streams in Central America and the Caribbean (Covich & McDowell Reference COVICH, MCDOWELL, Reagan and Waide1996, March & Pringle Reference MARCH and PRINGLE2003, Pringle & Hamazaki Reference PRINGLE and HAMAZAKI1998), and amphidromous gobies (e.g. species of the genera Sicydium and Sicyopterus) are major periphyton grazers in streams of islands and coastal zones throughout the tropics (Barbee Reference BARBEE2005, Covich & McDowell Reference COVICH, MCDOWELL, Reagan and Waide1996, Inoue & Miyayoshi Reference INOUE and MIYAYOSHI2006, Winemiller Reference WINEMILLER1983). In the absence of native predatory species from freshwater fish families, transient marine fishes, such as species of Centropomidae, Lutjanidae and Pomadasyidae, may assume the role of top predator (Pusey et al. Reference PUSEY, KENNARD and ARTHINGTON2004, Winemiller Reference WINEMILLER1983).
This study examines food-web structure of coastal streams of the Caribbean and Pacific coasts of Costa Rica by comparing results from earlier dietary analyses performed by the first author with new results from stable isotope analyses. Stable isotope methods are useful for estimating production sources that support consumers and the relative trophic positions of food-web components (Peterson & Fry Reference PETERSON and FRY1987, Post Reference POST2002). Stable isotope analysis yields estimates based on the organism's recent history of assimilation of elements from food (in the case of animals) or nutrient pools (in the case of plants). Dietary analysis provides a more detailed estimation of feeding links, but large samples are required for reliable estimation of spatial, temporal and ontogenetic variation. Use of both methods not only allows for comparison of inferences, but also should yield a more accurate assessment of food-web ecology (Layman et al. Reference LAYMAN, WINEMILLER, ARRINGTON, de Ruiter, Wolters and Moore2005, Mantel et al. Reference MANTEL, SALAS and DUDGEON2004, Rayner et al. Reference RAYNER, PUSEY, PEARSON and GODFREY2010, Winemiller et al. Reference WINEMILLER, AKIN and ZEUG2007). Numerous studies analysing stable isotopes have identified algae as the principal source of organic carbon supporting food chains in tropical streams and wetlands (Brito et al. Reference BRITO, MOULTON, SOUZA and BUNN2006, Bunn & Boon Reference BUNN and BOON1993, Douglas et al. Reference DOUGLAS, BUNN and DAVIES2005, Hamilton et al. Reference HAMILTON, LEWIS and SIPPEL1992, Jepsen & Winemiller Reference JEPSEN and WINEMILLER2007, Lau et al. Reference LAU, LEUNG and DUDGEON2009). A few isotopic studies have shown allochthonous plant matter to be an important basal input for food webs in tropical streams (Lau et al. Reference LAU, LEUNG and DUDGEON2009), leading some to question whether tropical stream food webs are different from those of temperate streams (Dudgeon et al. Reference DUDGEON, CHEUNG and MANTEL2010).
Here we examine several hypotheses related to food-web structure of tropical coastal streams by comparing findings from two alternative methodologies for investigating trophic ecology. First, we hypothesize that stable isotope analysis will support findings from dietary analysis indicating that consumer biomass in forested tropical streams is supported mostly by aquatic and terrestrial macrophytes (leaves, fruits, plant detritus) with a lesser contribution from benthic algae (Winemiller Reference WINEMILLER1983, Reference WINEMILLER1990; Winemiller & Morales Reference WINEMILLER and MORALES1989). Second, dietary analysis based on stomach contents and stable isotope analysis of consumer tissues will yield consistent estimates of species trophic positions that define vertical food web structure. Finally, isotopic signatures of obligate freshwater species versus diadromous species that move between fresh and marine habitats will be different, with diadromous species having heavier C and N ratios indicative of marine production sources. Resolution of the last hypothesis would provide insight regarding the potential influence of diadromy on food-web patterns and dynamics in tropical coastal streams.
STUDY SITES
Food-web structure was investigated in two streams in Corcovado National Park (Pacific coast) and two streams in Tortuguero National Park (Caribbean coast) in Costa Rica, Central America (Figure 1). Corcovado sites were Quebrada Camaronal, a small, low-gradient stream draining low, densely forested terrain near the Sirena Station, and the Rio Claro, a larger stream draining forested ridges lying to the west of Sirena Station. Quebrada Camaronal empties into the lower estuary of the Rio Sirena, and the Rio Claro empties directly into the Pacific Ocean. The reaches of Quebrada Camaronal and Rio Claro surveyed for food-web components were located approximately 1 km upstream from the coast. More detailed descriptions of the two Corcovado survey sites and the surrounding region appear in Lyons & Schneider (Reference LYONS and SCHNEIDER1990), Winemiller (Reference WINEMILLER1983) and Winemiller & Morales (Reference WINEMILLER and MORALES1989). The fish communities of these streams contain a few freshwater species, such as the characid Astyanax fasciatus, the poeciliid Poecilia gilli and the catfish Rhamdia guatemalensis (Pimelodidae), plus several diadromous species (e.g. Centropomus nigrescens, Lutjanus argentiventris, Awaous transandeanus, Sicydium salvini).
Map of Costa Rica showing locations of Corcovado National Park on the Osa Peninsula, Pacific coast, and Tortuguero National Park on the Caribbean coast.
The Tortuguero sites were Quebrada Estacion, a small stream at the park's northern limit on the barrier island, and the Rio Tortuguero in the park's interior. Both Tortuguero streams drain dense lowland rain forest; however, deforestation and human settlement have occurred along the northern bank in the lowest reaches of Quebrada Estacion over the past 15 y. Both streams drain into the Laguna Tortuguero, a deep narrow estuary that receives water from several major rivers including the Rio Tortuguero. This estuary normally has fresh water with salinity <1‰ near the surface and a saline wedge at the bottom that can extend several kilometres upstream depending on fresh-water discharge and tides. Reaches of the two streams that were surveyed for food-web components are located more than 5 km from the mouth of the Laguna Tortuguero. The Rio Tortuguero samples include samples taken from Caño Agua Fria Viejo, a side channel containing dense mats of floating aquatic macrophytes and surrounded by dense swamp forest. Detailed descriptions of the two Tortuguero survey sites and the surrounding region appear in Winemiller (Reference WINEMILLER1990) and Winemiller & Leslie (Reference WINEMILLER and LESLIE1992). Like the streams at Corcovado, the streams at Tortuguero contain both fresh-water adapted fish and macro-invertebrate species as well as diadromous species that, depending on species and life stage, spend variable periods of time in freshwater habitats (Gilbert & Kelso Reference GILBERT and KELSO1971, Winemiller & Leslie Reference WINEMILLER and LESLIE1992). Compared with the stream fish assemblages at Corcovado, those at Tortuguero are comprised of larger percentages, in terms of population abundance, of species from freshwater families such as Characidae, Poeciliidae and Cichlidae.
METHODS
Dietary analyses
Dietary data for fishes at each of the four study reaches were compiled from published studies by the first author. Volumetric estimates of diet components were derived from stomach contents analysis. Corcovado samples were for the wet season (June–July) during 1982, 1983 and 1986. For fish species that occurred at both Quebrada Camaronal and Rio Claro, specimens from both study sites were combined in the calculation of volumetric proportions of diet items (Winemiller Reference WINEMILLER1983, Winemiller & Morales Reference WINEMILLER and MORALES1989). For these nine species, the same dietary estimates were used for both sites. Tortuguero samples were taken monthly from February to December 1985, and data were aggregated into annual estimates of dietary composition (Winemiller Reference WINEMILLER1990). For species that occurred at both sites, dietary estimates for each site were independent based on specimens collected in situ.
Stable isotope analyses
During June 2004, samples of the most common plant and macrofaunal species were collected at all four of the stream survey sites for analysis of carbon and nitrogen isotope ratios. Submerged or floating leaves and seeds were collected by hand. Periphyton was collected from the surface of rocks by first rinsing to remove loose fine particulate matter, and then scraping to obtain attached algae and cyanobacteria. Fishes and macro-invertebrates were captured with a seine net, cast net, dip nets, or hook and line. Samples of periphyton, detritus, aquatic plants and terrestrial plants (including leaves and seeds) were crushed and preserved in salt. For fish and invertebrate specimens, a sample of muscle tissue (large specimens) or whole body minus viscera or shell (small specimens) was obtained and preserved in salt. Samples were subsequently stored in a freezer.
In the laboratory, tissue samples were thawed, soaked and rinsed in distilled water, then dried in an oven at 60 °C for 48 h. Dried samples were ground to a fine powder with pestle and mortar and stored in clean glass vials. Subsamples of each sample were weighed to 1 μg and pressed into Ultra-Pure tin capsules (Costech, Valencia, CA). Samples were analysed for stable isotope ratios (13C/12C and 15N/14N) at the Analytical Chemistry Laboratory, Institute of Ecology, University of Georgia, USA. Samples were dry combusted (micro Dumas technique) with a Carlo Erba CHN elemental analyser. Purified gases (CO2 and N2) were introduced into a Finnigan Delta C mass spectrometer, and isotopic composition was quantified relative to standard reference materials. Results are reported as parts per thousand (‰) differences from the corresponding standard:
![\begin{eqnarray}
\delta {\rm X} = [({\rm R}_{{\rm sample}}/{\rm R}_{{\rm standard}}) - 1] \times 10^3,\nonumber\\
{\rm where} \quad {\rm R} = \, ^{15}{\rm N}/^{14} {\rm N}\,{\rm or}\,^{13} {\rm C}/^{12} {\rm C}.\nonumber\end{eqnarray}](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20151024054926468-0506:S0266467411000277_eqnU1.gif?pub-status=live){kind=small}
Statistical analyses
Dietary data were used to estimate consumer taxon trophic position (TP) using the method presented in Winemiller (Reference WINEMILLER1990):
where TPj is the trophic position of food category j, pij is the volumetric proportion of food category j in the diet of species i, and n is the number of diet categories.
Biplots of δ15N and δ13C values of fishes, macro-invertebrates, periphyton and living and dead plant material were used to compare patterns of isotopic variation within and between sites. Given that δ13C values of diet items are usually conserved within 1‰ in consumer tissues (usually with consumer tissues shifting up to +1‰ δ13C in relation to their food, McCutchan et al. Reference MCCUTCHAN, LEWIS, KENDALL and MCGRATH2003), the relative importance of alternative sources of organic carbon assimilated is indicated by the relative positions of the consumer and the potential sources on the x-axis of the biplot. We did not attempt to perform quantitative estimates of the per cent assimilation of alternative primary production sources by consumers using a mixing model, because we were not able to obtain sufficient samples of either periphyton or the diverse terrestrial macrophytes in the riparian zone of these tropical wet forests. Therefore, our graphical analysis provides a general and preliminary assessment of source contributions to consumers in the food webs.
In contrast to carbon isotopes, nitrogen isotope ratios of consumer tissues typically are 2.5–3.4‰ higher than tissues of their food items (Vanderklift & Ponsard Reference VANDERKLIFT and PONSARD2003), and therefore δ15N can serve as an indicator of trophic position in addition to refining estimates of source contributions based on δ13C values. The trophic position of each consumer taxon of each stream was estimated based on fractionation of δ15N using the method of Vander Zanden & Rasmussen (Reference VANDER ZANDEN and RASMUSSEN1999):
where δ15Nreference is the mean δ15N value for a representative primary consumer, and 3.4‰ is the mean trophic fractionation. In lake studies, molluscs have been used as the representative primary consumers – bivalves as consumers of pelagic primary production sources, and gastropods as consumers of benthic primary production sources. Because the four Costa Rican streams are in regions with high rainfall and run-off as well as dense forests that shade streams, we assumed relatively little availability of pelagic production sources. We, therefore, based our calculation of consumer TPs on gastropods (the neritinid snail Neritina latissima Broderip, 1883 common at Corcovado, and ampullariid snails of the genus Pomacea common at Tortuguero) as reference primary consumers of periphyton and macrophytes. To the extent that these herbivores consume profitable sources of primary production available within their habitats, their δ15N signatures should provide a reliable index of the average trophic fractionation between the first and second trophic levels. The advantage of using primary consumers rather than primary producers as the trophic fractionation index is that primary consumers integrate the isotopic signal based on their proportional consumption of alternative production sources among the many encountered in the habitat (Anderson & Cabana Reference ANDERSON and CABANA2007, Vander Zanden & Rasmussen Reference VANDER ZANDEN and RASMUSSEN1999). In contrast, when the mean for primary production sources is employed as the reference for trophic fractionation, it is assumed that all sources are consumed by herbivores in equal proportions. To evaluate the influence of the choice of isotopic baseline on estimates of consumer TPs, we also calculated TP by using the mean δ15N of the primary producer samples obtained from each site and adding 1 instead of 2 to the right-hand side of the equation that calculates consumer TP.
To compare trophic position estimates based on isotopic and dietary methods, biplots were constructed with the mean TP of fish species based on isotopic data on x-axis and TP based on dietary data on the y-axis. To test for statistical significance of the linear relationship, we performed Pearson's correlation on untransformed TP values. To compare TP of obligate versus facultative (migratory) freshwater consumer taxa, we categorized each macro-invertebrate taxon and fish species as either freshwater or diadromous. All insect taxa, snails of the genus Pomacea, and fishes from freshwater families (e.g. Characidae, Cichlidae, Poeciliidae) were classified as freshwater, and all decapod crustaceans and fishes of estuarine and marine families (e.g. Gobiidae, Lutjanidae, Tetraodontidae) were classified as diadromous. Mean between-group differences of stable isotope ratios (C, N) and TP were tested with a t-test (two-tailed, significance at α = 0.05 for each stream separately.
RESULTS
Sources of primary production supporting aquatic food webs
In all four of the coastal streams, many fish and invertebrate consumer taxa had C isotope ratios heavier (higher δ13C values) than the averages of basal source samples obtained during this study (Appendix 1, Figure 2). In Quebrada Camaronal (Corcovado Park, Pacific coast), δ13C of four primary producer samples ranged from −32.1‰ (periphyton) to −27.1‰ (seeds of Virola tree), and mean consumer values ranged from −32‰ (poeciliid fish Poecilia gilli) to −25‰ (prawn Macrobrachium sp. 1) (Figure 2). In the Rio Claro (Pacific coast), δ13C of four primary producer samples ranged from −30.6‰ (seeds) to −15.5‰ (periphyton), and mean consumer values ranged from −30‰ (goby fish Sicydium salvini) to −20.5‰ (juvenile black snook Centropomus nigrescens) (Figure 2). In Quebrada Estacion (Tortuguero Park, Caribbean coast), δ13C of six primary-producer samples ranged from −41.3‰ (water lily Nymphaea sp.) to −29.9‰ (submerged wood), and mean consumer values ranged from −37‰ (poeciliid fish Phallichthys amates) to −24‰ (pipefish Oostethus lineatus) (Figure 2). In the Rio Tortuguero (Caribbean coast), δ13C of 12 primary producer samples ranged from −33.9‰ (unidentified submerged aquatic macrophyte) to −29.9‰ (water lettuce Pistia stratiotes), and mean consumer values ranged from −30‰ (juvenile fat snook Centropomus parallelus) to −21.5‰ (silverside Atherinella sp.) (Figure 2). A comparison of the periphyton samples from the four streams reveals that δ13C values were similar for Quebrada Camaronal (−32.1‰, −29.1‰), Quebrada Estacion (−35.7‰) and Rio Tortuguero (−35.7‰), but the single periphyton sample scraped from rocks in the Rio Claro had a much heavier signature (−15.5‰). Several fish species that were identified as benthic grazers based on observations of foraging behaviour and presence of algae and fine particulate detritus in stomachs (Winemiller Reference WINEMILLER1983, Reference WINEMILLER1990; Winemiller & Morales Reference WINEMILLER and MORALES1989) plotted in positions of two-dimensional isotope space that reflected periphyton as their likely principal food resource (aligned ± 1‰ on the x-axis and +3–3.5‰ on the y-axis); these grazers included Poecilia gilli and Sicydium salvini at Quebrada Camaronal, Phallichthys amates and P. gilli at Quebrada Estacion, and P. gilli at Rio Tortuguero (Figure 2). Poecilia at Quebrada Estacion revealed large between-individual variation in carbon isotope ratios (−39.0‰, −37.8‰, −27.6‰). All of these individuals were subadults between 28–31 mm standard length (SL). Poecilia from the other streams varied less, and were a mixture of sizes (75–90 mm SL at Quebrada Camaronal, 25–53 mm SL at Rio Tortuguero). Another species at Quebrada Estacion showing large between-individual variation in carbon signatures was the omnivorous sleeper, Dormitator maculatus: 52 mm SL (−27.6‰), 54 mm (−22.4‰), and 110 mm (−36.6‰).
Biplots of δ13C and δ15N (mean ± 1 SE) of plants (circles), invertebrates (triangles), and fishes (squares) sampled in Corcovado National Park streams: Quebrada Camaronal (a) and Rio Claro (b); and Tortuguero National Park streams: Quebrada Estacion (c) and Rio Tortuguero (d). Full taxon names appear in the Appendix 1.
Two basal production sources at Quebrada Estacion (Nymphaea and an unidentified submerged macrophyte) and one at Rio Tortuguero (an unidentified submerged macrophyte) had δ13C values significantly lower than any of the consumers sampled at the same site. Tissue from living water hyacinth (Eichhornia crassipes) at Quebrada Estacion had a nitrogen signature higher than most consumers having a similar range of carbon values. Similarly, at Rio Tortuguero, the two samples of terrestrial plants had δ15N values as high or higher than taxa known to be herbivores. At Quebrada Estacion, half of the consumer taxa that were sampled had δ13C values significantly higher than any of the six in situ production sources. Likewise, about half of the 31 consumer taxa sampled at the Rio Tortuguero had carbon signatures that were significantly higher than any of the 12 in situ production source samples.
Vertical food-web structure and consumer trophic positions
Correlations between diet-based estimates of consumer trophic positions (TPs) and isotope-based estimates of TPs were as follows: Rio Claro: r = 0.53, df = 11, P > 0.05, Quebrada Camaronal: r = 0.18, df = 10, P > 0.05, Rio Tortuguero, r = 0.55, df = 20, P < 0.01; Quebrada: r = 0.73, df = 11, P < 0.01 (Figure 3). For the Tortuguero streams, isotope-based TP values tended to be higher than diet-based values, whereas there was no systematic pattern for the two Corcovado streams. Due to the manner by which TP is calculated from δ15N data, estimates based on snails as the reference and those based on the mean of primary producers as the reference were perfectly correlated (Appendix 1). TP values based on δ15N data and using primary consumers (snails) as the reference yielded estimates that were higher than those calculated using the mean of primary-producer samples as the reference (mean ± SD difference = 0.87 ± 0.29).
Comparison of consumer trophic position values estimated using dietary data with trophic position values estimated using stable isotope ratios in Corcovado National Park streams Quebrada Camaronal (a) and Rio Claro (b) and Tortuguero National Park streams Quebrada Estacion (c) and Rio Tortuguero (d). The diagonal line represents perfect one-to-one correspondence of the two estimates.
Primary consumer taxa other than snails, with TP values (based on δ15N with snails as reference) between 1.5 and 2.5, were as follows: Quebrada Camaronal: Odonata, unidentified aquatic insect, atyid prawn and Macrobrachium sp. 1 and sp. 2; Rio Claro: Odonata, crab, atyid shrimp, Sicydium, Trinectes and Poeciliopsis; Quebrada Estacion: Hemiptera and crab; Rio Tortuguero: no taxon other than snails with TP < 2.5 (Appendix 1). Based on δ15N with snail references, top predators in Quebrada Camaronal were Pomadasys bayanus (TP = 4.0), Agonostomus monticola (3.8), Awaous banana (3.3), Gobiomorus latifrons (3.3) and Astyanax fasciatus (3.3). TP values of top predators in the Rio Claro, a larger stream than Q. Camaronal, were similar: Centropomus nigrescens (4.0), Lutjanus argentiventris (3.9), L. novemfasciatus (3.85), Pomadasys bayanus (3.5), Sphoeroides annularis (3.4) and Gobiomorus latifrons (3.3). TP values of top predators in the two Caribbean streams were as follows: Quebrada Estacion: Sphoeroides testudineus (4.5), Gobiomorus dormitor (4.2), Guavina guavina (3.9), Rhamdia guatemalensis (3.8) and Parachromis loisellei (3.6); R. Tortuguero: Atractosteus tropics (4.5), Belonesox belizanus (4.1), Lutjanus jocu (3.8), Centropomus ensiferus (3.8), Astatheros rostratus (3.8), Parachromis loisellei (3.7), Eucinostomus sp. (3.7), Parachromis dovii (3.6), Alfaro cultratus (3.6), Amphilophus citrinellus (3.6), Theraps maculicauda (3.6), Archocentrus centrarchus (3.6), Centropomus undecimalis (3.6) and Pomadasys crocro (3.6). The mean TP of obligate freshwater taxa was not significantly different than the mean TP of diadromous taxa from the same local community for any of the four streams (two-tailed t test, P > 0.05).
Mean δ13C of freshwater versus diadromous taxa was significantly different only for Quebrada Estacion at Tortuguero (t = 2.11, df = 37, P < 0.05), with diadromous taxa having heavier signatures on average (Figure 4). Among the between-group comparisons of mean δ15N, only Rio Claro had a significant difference (t = 3.04, df = 14, P < 0.01).
Comparison of mean (±1 SD) standardized isotopic ratios of diadromous (D) and freshwater (F) consumer taxa from the four tropical streams: δ13C (a) and δ15N (b). Sample sizes appear in parentheses.
DISCUSSION
Basal production sources
Based on the positions of consumers relative to periphyton samples in biplots (Figure 2), periphyton could have been a significant source of carbon supporting many and perhaps most primary consumers in every stream. This inference for the Rio Claro is based on a single periphyton sample with a high δ13C value that greatly exceeded values for periphyton from the other three streams in this study. Mean carbon isotope signatures of benthic scrapers in the Rio Claro (e.g. Neritina snails and Sicydium gobies) were much lower (−30‰) than this periphyton sample, which suggests there could have been error associated with this periphyton sample. Standing biomass of periphyton was extremely low in the Rio Claro, and periphyton scraped from rocks could have been contaminated with calcium carbonate, which could have elevated the carbon isotopic signature. Even so, high δ13C values for periphyton samples (−10‰ to −18‰) have been recorded by other studies of tropical streams (March & Pringle Reference MARCH and PRINGLE2003, Mendoza-Carranza et al. Reference MENDOZA-CARRANZA, HOEINGHAUS, GARCIA and ROMERO-RODRIGUEZ2010), although lower values (−30‰ to −25‰) seem to be more commonly reported (Kilham & Pringle Reference KILHAM and PRINGLE2000, Mantel et al. Reference MANTEL, SALAS and DUDGEON2004, Verburg et al. Reference VERBURG, KILHAM, PRINGLE, LIPS and DRAKE2007).
Based on relative positions in the isotopic biplots, leaf litter, fruits, seeds and certain aquatic macrophytes are potentially important sources of carbon supporting macro-invertebrates and fishes in all four streams. For example, atyid shrimp in Quebrada Camaronal plotted approximately +2‰ above Virola fruits and seeds on the y-axis, and aligned with them on the x-axis. Nitrogen isotope signatures of periphyton were 1.5–3‰ higher than that of atyid shrimp, making it an unlikely resource used by these grazers. Using C and N stable isotope data, Yam & Dudgeon (Reference YAM and DUDGEON2005) and Lau et al. (Reference LAU, LEUNG and DUDGEON2009) estimated that periphyton, fine particulate organic matter, and lesser fractions of leaf litter supported atyid shrimp biomass in both shaded and unshaded streams in Hong Kong. In forest streams in Puerto Rico, shrimp (Atya, Macrobrachium and Xiphocaris spp.) were estimated to have assimilated most of their carbon from periphyton (δ13C: c. −18‰) and biofilm (δ13C: −25‰ to −20‰), and leaves (δ13C: −30‰ to −29‰) seemed to contribute little (March & Pringle Reference MARCH and PRINGLE2003). Lau et al. (Reference LAU, LEUNG and DUDGEON2009) reported that leaf litter had generally lower C and N isotopic signatures compared with periphyton in Hong Kong streams, and isotopic ratios of both basal production sources and consumers differed with regard to season and stream shading. Robust conclusions about the importance of local production sources for the food web cannot be made, because our samples were collected over a period of a few days with little replication in space. Aquatic production sources, periphyton (microalgae) in particular, may undergo temporal shifts in isotopic ratios in response to changing environmental conditions, or may vary depending on physical conditions in specific microhabitats (Boon & Bunn Reference BOON and BUNN1994, Finlay Reference FINLAY2004, Finlay et al. Reference FINLAY, POWER and CABANA1999). Both Corcovado and Tortuguero national parks have variable monthly rainfall but lack strong wet/dry seasonality, and the 2004 samples were collected during a relatively dry period with stable stream discharges at both locations.
Based on positions in the isotope biplots, most consumer taxa of the two coastal streams at Tortuguero National Park (Caribbean drainage) probably are supported by variable fractions of material derived from periphyton, terrestrial leaf litter, and certain aquatic plants, including water hyacinth (Eichhornia crassipes), water lettuce (Pistia stratiotes) and water primrose (Ludwigia sp.). Seston, including phytoplankton, also could be a significant basal production source supporting consumers in the Rio Tortuguero. The lower reaches of both the Rio Tortuguero and Quebrada Estacion have very low gradients with slow velocities during periods of low discharge. During the driest periods, these lower reaches actually reverse flow during high tides, and these lentic conditions would be conducive for phytoplankton growth and retention. Two of the aquatic macrophytes sampled at Tortuguero (water lily, Nymphaea sp., and an unidentified submerged plant) had extreme negative δ13C values and apparently contributed little or no carbon to consumer taxa, however it should be noted that, in a manner similar to benthic microalgae, aquatic macrophytes can undergo temporal shifts in isotopic signatures (Boon & Bunn Reference BOON and BUNN1994).
In all four of the coastal streams we examined, many consumers had heavier carbon isotopic signatures than any of the local primary production sources that were sampled. Because stable isotopes of carbon usually do not fractionate more than approximately 1‰ during transfer between trophic levels (McCutchan et al. Reference MCCUTCHAN, LEWIS, KENDALL and MCGRATH2003), the carbon isotope ratio of a consumer should reflect the proportional assimilation of carbon from primary production sources supporting food chains leading to a given consumer. Limited samples of the most conspicuous fishes, aquatic macro-invertebrates, periphyton, aquatic plants and terrestrial vegetation, including seeds and fruits that had fallen into or near the water, were collected from each of the study streams. At least four non-mutually exclusive hypotheses could account for the disparity between consumers with carbon isotopic signatures that were heavier than those of all basal production sources collected from the habitat. First, our study could have failed to sample at each location one or more important plants that were enriched in 13C. Given the limited number of periphyton, aquatic macrophyte and terrestrial plant material sampled during the study, this hypothesis cannot be falsified without analysis of additional samples. It seems unlikely that more samples of the same basal resource categories, including primary producer taxa that were rare in these habitats, would produce very many, or perhaps any, heavy δ13C values. Tropical C4 grasses typically have heavy δ13C (−20‰ to −10‰, Jepsen & Winemiller Reference JEPSEN and WINEMILLER2007), but these grasses were essentially absent from the riparian habitats of the wet forests of Corcovado and Tortuguero. Shifts in the abundance of primary production sources associated with hydrological variation was documented in a North American temperate river (Zeug & Winemiller Reference ZEUG and WINEMILLER2008), and this possibility cannot be ruled out given the very limited time interval in which our tissue samples were collected.
A second possible explanation for the presence of many consumers with heavier δ13C values than any of the local sources that were collected was already mentioned. Benthic microalgae, aquatic macrophytes and/or terrestrial vegetation entering the streams could have undergone temporal shifts in their isotopic ratios. Studies have documented such shifts in algae and aquatic plants in both temperate and tropical lotic ecosystems (Boon & Bunn Reference BOON and BUNN1994, Finlay Reference FINLAY2001, Reference FINLAY2004; Finlay et al. Reference FINLAY, POWER and CABANA1999). These isotopic shifts apparently are caused by physico-chemical environmental factors, including water velocity and pressure, changes in relative availability of organic and inorganic carbon molecules, and other factors. In addition, bulk periphyton isotopic ratios could be influenced by dynamics of cell growth and community dynamics in response to changing nutrients, light or grazing pressure. Our samples were collected during a relatively dry period with relatively stable stream discharges at both locations.
A third hypothesis is that migratory invertebrates and fishes in each coastal stream could have assimilated carbon derived from primary producers in the adjacent marine ecosystem before entering the freshwater habitat. Marine ecosystems tend to be associated with more 13C- and 15N-enriched plants and animals when compared with freshwater ecosystems (Garcia et al. Reference GARCIA, HOEINGHAUS, VIEIRA and WINEMILLER2007, Winemiller et al. Reference WINEMILLER, AKIN and ZEUG2007, Reference WINEMILLER, HOEINGHAUS, PEASE, ESSELMAN, HONEYCUTT, GBANAADOR, CARRERA and PAYNE2011). Some of the predatory fishes of these streams, such as juvenile and adult snooks (Centropomus spp.) and snappers (Lutjanus spp.), migrate between freshwater streams and coastal brackish or marine waters. Thus, their relatively heavy isotopic signatures could reflect recent histories of feeding on a mixture of food resources from marine and freshwater ecosystems. According to Lyons & Schneider (Reference LYONS and SCHNEIDER1990), all but three or four of the fish species in the Rio Claro spend a portion of their lives in the estuary or Pacific Ocean, and fish community structure was correlated more with distance from the coast than with local habitat variables. Snails of the genus Neritina migrate seasonally for several kilometres in Rio Claro (Schneider & Frost Reference SCHNEIDER and FROST1986), and presumably the Quebrada Camaronal as well. Palaemonid shrimp of the genus Macrobrachium are amphidromous, with juveniles migrating upstream, sometimes for hundreds of kilometres and crawling over numerous natural obstructions (Benstead et al. Reference BENSTEAD, MARCH, PRINGLE and SCATENA1999).
Fourth, many of the resident freshwater macrocrustaceans and fishes could have preyed upon invertebrates and juvenile fishes that immigrate into streams from coastal marine waters. For example, the ‘tismische’ at Tortuguero is a mass immigration of post-larval shrimp and fishes (dominated by Eleotris spp. and other Gobiiformes) from marine coastal waters that occurs several times per year (Gilbert & Kelso Reference GILBERT and KELSO1971, Winemiller & Ponwith Reference WINEMILLER and PONWITH1998), thus providing a massive influx of food resources for resident freshwater consumers. The tismische is an example of a spatial food-web subsidy involving transfer of biomass originating from the marine ecosystem into coastal streams, but other examples can be cited. Upstream migrations of Neritina snails in the Rio Claro provide a marine-derived subsidy for predators in the freshwater food web. Although they do not typically migrate in large numbers in synchrony, many juvenile marine fishes enter freshwater streams where some of them ultimately fall prey to larger fishes.
The isotopic data currently available cannot directly test these four alternative hypotheses for consumers with carbon isotope signatures heavier than in situ primary production sources, but the second and third seem most likely. Nonetheless, insight may be gained from the comparison isotopic ratios of freshwater versus diadromous taxa (Figure 4). As noted previously, studies of coastal zones have consistently documented heavier carbon and nitrogen ratios of both plants and consumers from marine habitats compared to nearby freshwater ecosystems (Garcia et al. Reference GARCIA, HOEINGHAUS, VIEIRA and WINEMILLER2007, Mendoza-Carranza et al. Reference MENDOZA-CARRANZA, HOEINGHAUS, GARCIA and ROMERO-RODRIGUEZ2010, Winemiller et al. Reference WINEMILLER, AKIN and ZEUG2007, Reference WINEMILLER, HOEINGHAUS, PEASE, ESSELMAN, HONEYCUTT, GBANAADOR, CARRERA and PAYNE2011). Diadromous species tended to have heavier carbon and nitrogen signatures than their freshwater taxa at every stream except for Rio Tortuguero, however only two of these six cases were statistically significant (Quebrada Estacion for δ13C; Rio Claro for δ15N). Quebrada Estacion is a small stream that empties into the Laguna Tortuguero (estuary) and thus was easily colonized by crustaceans and estuarine fishes such as pufferfish (Sphoeroides testudineus). The Rio Claro empties directly into the Pacific Ocean, and its community was dominated by decapod crustaceans and fishes from marine families, including pufferfish, snappers (Lutjanidae), snooks (Centropomidae) and pipefish (Syngnathidae).
Trophic position
Differences between dietary and isotopic estimates of consumer trophic positions were likely due to three factors: (1) biases associated with small sample sizes for the diet estimates of several species (affecting TP estimates from stomach-contents data), (2) error in estimating contributions of basal sources supporting consumers (affecting TP estimates based on isotopic data) and (3) error in the assumed trophic fractionation of δ15N (Caut et al. Reference CAUT, ANGULO and COURCHAMP2009). Nonetheless, there were moderate to high correlations between TP values derived from the two methods for three of the four streams; the only low correlation was for Quebrada Camaronal. There are obvious strengths and limitations of each method. Dietary analysis provides a short-term feeding history of the individual organism (minutes to hours), and its precision and accuracy are highly dependent on resolution of food items identified within gut contents, sample size (number of guts examined), and heterogeneity of the sample (e.g. temporal, spatial and body-size variation within the consumer population). Stable isotope analysis provides a time-integrated estimate of feeding history (weeks to months), and thus may not reflect recent feeding history, especially for organisms that move between ecosystems. Using the isotopic method, Lau et al. (Reference LAU, LEUNG and DUDGEON2009) found that maximum TP estimated for shaded streams in Hong Kong remained fairly constant between the dry and wet seasons, whereas TP values for unshaded streams were higher and showed seasonal fluctuations. Clearly, the temporal scale of sampling for both methods influences TP estimates. In the present study, all Corcovado samples were obtained during June–July. The Tortuguero diet samples were integrated over an entire annual cycle and isotope samples were obtained during June, and yet Tortuguero TP estimates had highest correlations for the two methods.
Isotopic-based TP values of certain species were clearly underestimated, e.g. atyid shrimp, odonate nymph and pseudothelphusid crab all with values < 2 (strict herbivore = 2.0). Other isotope-based TP values appear to be overestimates, e.g. herbivorous Poecilia gilli, Phallichthys amates and Brycon guatemalensis with TP ~ 3; the dietary method estimated TP ~ 2 for these species. The dietary method probably underestimated TP for Awaous banana (2.75), a goby that engulfs and winnows sand and other fine substrates to sift out aquatic invertebrates. These fish have relatively short guts consistent with carnivory, which was reflected by the isotope-based TP (3.34). Gut-contents analysis of the Awaous diet included a minor fraction of detritus, but this material apparently is not assimilated. Similarly, the cichlids Archocentrus centrarchus and Theraps maculicauda had large disparities between isotopic and diet-based TP estimates, and this appears to have been due to ingestion of large fractions of detritus but with low assimilation.
The maximum trophic position (TP of the top predator) can be interpreted as an indicator of the mean food-chain length of the food web (Hoeinghaus et al. Reference HOEINGHAUS, WINEMILLER and AGOSTINHO2008, Post et al. Reference POST, PACE and HAIRSTON2000). Comparing the four Costa Rican streams, the maximum TPs were: Q. Camaronal 4.04 – Pomadasys bayanus, R. Claro 4.01 – Centropomus nigrescens, Q. Estacion 4.46 – Sphoeroides testudineus and R. Tortuguero 4.52 – Atractosteus tropicus. Thus, there was no difference in maximum TP between the small and large stream at each location, but the Tortuguero streams had higher maximum TP (+0.5). Ecosystem size was not correlated with maximum food-chain length as was found for temperate lakes (Post et al. Reference POST, PACE and HAIRSTON2000), because stream sizes were fairly evenly matched between locations (Q. Camaronal = Q. Estacion, R. Claro = R. Tortuguero). Another potential determinant of maximum TP is species richness, but this relationship is weak: Q. Camaronal = 16 fish spp. and R. Claro = 18 fish spp. (Winemiller & Morales Reference WINEMILLER and MORALES1989); Q. Estacion = 22 fish spp., R. Tortuguero = 59 fish spp. (Winemiller Reference WINEMILLER1990). Aquatic ecosystem productivity (i.e. net aquatic productivity) was not measured in these streams, but it seems likely that they would rank as R. Tortuguero > Q. Estacion > R. Claro > Q. Camaronal. This ranking is inferred based on hydrology, nutrient inputs and degrees of shading. Both the Rio Tortuguero and Quebrada Estacion receive water from both local watershed and the estuary via tidal fluxes, and therefore should receive seston inputs from relatively lentic habitats downstream. Whereas Quebrada Estacion is heavily shaded and supports few aquatic macrophytes, the Rio Tortuguero has a broader channel with more exposure to sunlight that supports dense beds of floating and submerged aquatic macrophytes. No aquatic macrophytes are found in either of the Corcovado streams, but the wider channel of the Rio Claro is more exposed to sunlight than the narrow channel of Quebrada Camaronal that is almost entirely covered by rain-forest canopy. Thus, maximum TP in these coastal streams is probably influenced most strongly by ecosystem productivity, with species richness and other factors having secondary influences, and ecosystem size having little apparent correlation.
Conclusions
Overall, the two methods revealed similar patterns of resource utilization and vertical trophic structure in the four coastal streams, and differences could be attributed to data deficiencies and/or potential sources of bias associated with either method. These limitations notwithstanding, it appears probable that both periphyton and terrestrial vegetation, including fruits, are important basal production sources supporting biomass of aquatic organisms in all four streams. For three of the four streams, isotopic signatures of substrate grazing fishes were consistent with observed diets that were dominated by periphyton and fine particulate matter that likely is derived from a combination of periphyton and terrestrial vegetation.
Marine fishes and invertebrates enter coastal streams where they take up residence for variable periods of time (Kinzie Reference KINZIE1988, Nordlie Reference NORDLIE1981, Pender & Griffin Reference PENDER and GRIFFIN1996) and become integral parts of the stream food web. Certain decapod crustaceans and fishes have life cycles whereby adults spawn within freshwater streams, and their eggs and larvae drift to the estuary (McDowall Reference MCDOWALL1988), providing food resources for consumers along the route (March et al. Reference MARCH, BENSTEAD, PRINGLE and SCATENA1998). Post-larval and juvenile decapods and fishes actively migrate from coastal waters into freshwater streams, sometimes en masse as with the tismische phenomenon at Tortuguero. Several studies have shown marine primary producers with heavier C and N isotopic ratios than producers from regional freshwater ecosystems, and diadromous species in the four Costa Rican streams tended to have heavier isotopic signatures than obligate freshwater species, although distributions were broadly overlapping in most cases. Thus, evidence for a trophic contribution of migratory organisms was weak with the exception of two comparisons that yielded statistically significant differences. It is important to note that diadromous fishes can influence coastal stream food webs in ways besides influencing predator–prey interactions and providing subsidies. Marine immigrants sometimes influence ecosystem processes, such as sediment accumulation and benthic net primary production (Pringle Reference PRINGLE1996, Pringle & Hamazaki Reference PRINGLE and HAMAZAKI1998, Pringle et al. Reference PRINGLE, BLAKE, COVICH, BUZBY and FINLEY1993), in tropical coastal streams.
ACKNOWLEDGEMENTS
Institutional support and scientific permits were provided by the Ministry of Agriculture and Natural Resources and the National Park Service of Costa Rica. We thank C. Saenz, L. Kelso-Winemiller, M. Winemiller and G. Stunz for field assistance, and R. Plowes and A. Vega for assistance in arranging permits and field logistics. Funding for this project was provided by the NSF Undergraduate Mentoring in Environmental Biology program (grant #0203992).
δ13C and δ15N values, estimated trophic positions (TP) based on stable isotope data using either snails or plants as the baseline, trophic positions based on dietary data, and sample sizes for dietary and isotopic analyses of organisms (F = obligate freshwater; D = diadromous) comprising the food webs of four tropical streams. Fish species names follow Bussing (1998).
