Trophic guilds of fishes in sandbank habitats of a Neotropical river

The objective of this study was to characterize the trophic structure of the community of fishes exploiting riverine sandbank habitats. Collections were carried out during the period of October 1999 to December 2003, on six sand banks in the upper and middle portions of the Tocantins River drainage basin in central Brazil. The availability of food resources was evaluated based on the volume of the items present in the stomachs of all species. A total of 2,127 stomachs of fish belonging to 50 species were analyzed. Nine main trophic guilds grouped the local ichthyofauna according to diet. Aquatic-origin items were the preferred source for 55.5% of the groups analyzed, whereas terrestrial-origin items composed 44.4%. Items of undetermined origin (detritus and sediment), although present in 89% of the guilds, were the predominant food in only one trophic group. Terrestrial insects and fish were the food sources with the largest biomass available in the environment. Sandbank environments are homogeneous, with little shelter and food available; as a rule, the species that occupy these environments are generalists.


Introduction
From knowledge of the diet and abundance of the species of fishes in a community, it is possible to identify the different trophic categories, to make inferences about the trophic structure, to evaluate the degree of importance of the different trophic levels, and to understand the interrelations among the components of the community (Payne, 1986;Agostinho et al., 1997).
Sandbank environments are shallow and homogeneous, and because they are relatively unstructured, they offer little shelter or food compared to others, i.e., rocky-shore and vegetated-shore habitats.Lowe-McConnell (1999) described this kind of environment as extensive sandy areas found along riverbanks, which are covered with water during certain times of year.In the Araguaia-Tocantins Basin, the marked seasonality of rains leads to the formation of innumerable sandbank areas, especially during the dry season.The fish communities of these sand banks are little known, in spite of the growing loss of these environments resulting from the construction of hydroelectric reservoirs, which not only permanently floods many of the sand banks but also rearranges the annual river discharge pattern.
The assemblages of fishes on the sand banks are highly important in the dynamics of the environment, because they include a substantial number of benthic organisms, including forage fish, which serve as food for the predators (Lowe-McConnell, 1999).According to Arrington & Winemiller (2003), the fish that occur on river sand banks during the day generally use them for foraging; and fish that occur there during the night take advantage of the shallow water as a refuge from predation.
During this study, a trophic characterization of the species of fishes associated with the sand banks of the upper and middle parts of the drainage basin of the Tocantins River was carried out, with the objective of answering the following questions: (i) what is the food spectrum of the species that occupy the sand banks of the Tocantins River?(ii) how many trophic guilds can be identified in this kind of environment?
(iii) what is the importance of food of terrestrial and aquatic origin for the guilds present?(iv) what are the proportions of available food resources on the sand banks?
The fish were sampled by means of seines with a mesh size of 5 mm between opposing knots, 20 m long, 2 m high, and fitted with a center bag where the fish concentrated.During seine hauls, one end of the seine was pulled along the shoreline and the other end was pulled in a parallel direction offshore.At the offshore end of the seine the maximum depth was 1.5 m.For most of the year, water velocity ranged from zero in littoral areas to about 0.5 ms -1 in some sand banks in the river.During the low-water period (April through September), water transparency was total at the collection localities.During the high-water period (October through March), the mean transparency was 0.50 m.The substrates are dominated by alluvial deposits of fine white sand from the surrounding watershed.The specimens collected were fixed in 10% formalin and transported to the laboratory, where the total length (cm), standard length (cm), and total weight (g) were taken.
Subsamples of the species analyzed were deposited at the Laboratório de Ictiologia e Sistemática of the Universidade Federal do Tocantins, under the numbers listed in Table 2.
For the analysis of the stomach contents, the occurrence frequency and volumetric methods were used (Hynes, 1950;Hyslop, 1980).The data for volume were obtained either by compressing the material (food items) under a glass slide on a plate with a one-millimeter grid, to a known height (1 mm), and converting to milliliters based on the area covered; or by placing the items in a graduated cylinder and calculating the displacement of water.The volume of each item was converted to a percentage.We assumed that the results obtained using these two methods were similar.
The possibility of analysis of the data by grouping the sampling localities was evaluated using Kendalls' Coefficient of Concordance (W) (Siegel, 1975) applied to the volume of the food items (STATISTICA version 5.5;Statsoft, 2000).
In order to classify species into trophic groups, we first ordinated the data using Detrended Correspondence Analysis to remove the arch effect (DCA; Hill & Gauch, 1980;Gauch, 1994).DCA ordination allowed the analysis of the patterns of feeding similarity among the 50 species based on volume values, which were square-root transformed, before analysis, to reduce skewness.Second, we identified trophic groups based on the scores of the first two DCA axes with a non-hierarchical cluster analysis, the k-means method (STATISTICA 5.5;Statsoft, 2000).Non-hierarchical methods maximize intra-group homogeneity, without considering the hierarchy between groups (Valentin, 2000).The formation of the trophic groups was defined using the non-hierarchical grouping analysis (kmeans), because in certain cases, the hierarchical classification method does not adequately reflect the similarity relationships among the sampling units.Hierarchical techniques applied to a group of data produce a phenomenon called "chaining", which refers to the tendency for the method to incorporate sampling units into an already existing group, instead of forming a new one (Everitt & Dunn, 1991).Furthermore, when the number of sampling units is large, the dendrograms resulting from the hierarchical techniques are difficult to interpret because of the many groups that are formed.Thus, when the data are not adequately represented by a hierarchical structure and when the objective is to form a number of groups of previously fixed sampling units, the non-hierarchical k-means analysis can be used (Bishop, 1995).Similar methodology to separate trophic guilds was used by Luz-Agostinho et al. (2006) and Loureiro-Crippa & Hahn (2006).
The food resources consumed were classified in three groups: items of aquatic origin, items of terrestrial origin, and undetermined, a group that included those items for which it was impossible to determine the origin (e.g., detritus and sediment).
The availability of the food resources was evaluated based on the volume of the items present in the stomachs of all the species combined (Lawlor, 1980;Winemiller & Kelso-Winemiller, 1996).The number of stomachs analyzed per species was not proportional to their participation in the sample, and therefore the volumes were corrected using the equation proposed by Luz-Agostinho et al. (2006).

Results
A total of 2,473 stomachs from 124 fish species contained some type of food.However, diet characterization was carried out only for species for which we found more than ten stomachs with identifiable food items, resulting in a data set of 2,127 stomachs representing 50 species.
Kendall's coefficient of concordance (W = 0.7481 and p < 0.05) showed that there was a correlation between the food items at the six sampling points, and therefore they were treated together.
Analysis of the DCA results showed that axis 1 accumulated the greatest variability of species, with an eigenvalue of 0.3229, followed by axis 2 with an eigenvalue of 0.0356 (Fig. 1).
Nine trophic groups with similar diets could be discriminated through the K-means analysis of the scores derived from DCA axes 1 and 2. The R statistic showed that adding new groups would not significantly reduce the value of the sum of squares within the groups, indicating that nine appeared to be the most parsimonious number of groups for the data set (Table 1).The nine trophic groups are shown in Table 2.The species scores along DCA axes 1 and 2 (Fig. 1) are circled, so as to visualize more easily the nine groups formed by the K-means analysis.
The nine trophic groups were characterized as follows: Group 1: seven species (14% of the species analyzed) preferentially consumed terrestrial and aquatic insects, to-gether with fish, sediment, plant matter, and algae.The exceptions were Myleus cf.torquatus and Pimelodus blochii, which ate mainly fish.
Group 3: two species (4%) fed mainly on terrestrial insects.They also ate significant amounts of other invertebrates and detritus.
Group 4: three species (6%) consumed a large amount of terrestrial insects and fish.Exodon paradoxus consumed mainly fish (scales), whereas the other species consumed terrestrial insects and fish in equal proportions.
Group 5: seven species (14%) basically exploited the bottom, consuming a large amount of detritus and sediment, together with algae.
Group 6: eight species (16%) fed on a wide variety of items, such as algae, detritus, aquatic insects, terrestrial insects, microcrustaceans, other invertebrates, fish, sediment, and plant matter.Terrestrial insects predominated in the diet of most of the species, except for Hemiodus unimaculatus for which the most prominent item was sediment.
Group 7: ten species (20%) had diets with a strong predominance of terrestrial and aquatic insects.However, Hyphessobrycon sp.B and Moenkhausia sp.E consumed a high proportion of plant matter and fish respectively.
Group 9: two species (4%) with food habits ranging from algae to plant matter, or from invertebrates to fish.However, these species preferentially ingested microcrustaceans or detritus, respectively.
Terrestrial and aquatic insects, plant matter, and sediment were the food items consumed most often by the fish assemblage (Table 2).Aquatic-origin items were present in 55.5% of the groups analyzed.These consisted mainly of aquatic insects such as larvae and pupae, followed by fish and microcrustaceans.Terrestrial resources, consisting mainly of terrestrial insects and plants, were the preferred source for 44.4% of the groups.The items of undetermined origin, detritus and sediment, were present in 89% of the guilds; however, they were the predominant food only in guild 5 (Fig. 2).   1. Results used to calculate the R-statistic (R), to assess whether the increase from k to k+1 groups was significant.Sum of SQ-within = sum of sum of squares within; n-k-1 = degrees of freedom; P = probability of Type 1 error.Table 2. Food resources used by the fish species, estimated by volume.N = number of stomachs examined; 1 = Algae; 2 = Detritus; 3 = Aquatic Insects; 4 = Terrestrial Insects; 5 = Microcrustaceans; 6 = Other Invertebrates; 7 = Fishes; 8 = Sediment; 9 = Plant Matter; -= absence of specimens.
The analysis of the proportions of food items suggested that terrestrial insects and fish were the main food resources available to fish in the environment (Fig. 3).

Discussion
Although in tropical environments there are species of fishes with marked trophic specializations, the majority of species show broad flexibility in feeding (Agostinho et al., 1995;Araújo-Lima et al., 1995;Lowe-McConnell, 1999).Feeding flexibility is a reflection of the interaction between the quality or quantity of the available food, and the degree of morphological and behavioral limitations exhibited by the species, the latter with ontogenetic variations (Luz et al., 2001).Therefore, determination of the trophic guilds of tropical fishes is often hampered by the broad feeding overlap of the species.Separation of the guilds by visual analysis of the dominant items in the stomachs often results in a high degree of subjectivity.
The 50 species of fishes caught along sand banks of the Tocantins system were grouped in nine trophic guilds.Melo et al. (2004), studying 71 species of the Araguaia River basin, and Luz-Agostinho et al. (2006), analyzing 64 species in the basin of the Paraná River, both using clustering methods and work-ing in different habitats, also found nine trophic guilds.Mérona & Rankin- de-Mérona (2004) analyzed 74 species in a floodplain lake in central Amazonia, and separated 11 trophic guilds based on the dominant food items.Pouilly et al. (2003) analyzed the diets of 48 species from lakes in the Bolivian Amazon, and separated eight trophic guilds using hierachical methods.However, the number of guilds established for different environments in the Paraná River basin varied from 7 to 10 (Hahn et al., 1997(Hahn et al., , 2004;;Peretti & Andrian, 2004;Loureiro-Crippa & Hahn, 2006;Luz-Agostinho, 2006).Understanding of the ecological meaning of the variation in the number of trophic guilds is impeded by the use of different methods to separate the guilds.Therefore, studies are needed to compare the accuracy of these methods and eventually to standardize the methodology.
The number of trophic groups and the similarity in the diet of the species of different groups, a result of the range of diet, made it difficult to assign names to the guilds.The dietary similarity results in part from the hydrological instability and homogeneity of the sandbank environment.According to Poff & Allan (1995), trophic generalism is greater in hydrologically variable environments.The shallowness of the sand banks makes them susceptible to random environmental variations caused by the winds and daily fluctuations of the water level, especially in areas influenced by reservoirs.
The sandbank environments are shallow and homogeneous; the sediment is composed of sand and a smaller quantity of organic detritus carried in by the water.It is expected that in environments with these characteristics, terrestrialorigin food sources will be important food sources for the fish assemblage.Analysis of the origin of food items revealed that aquatic-origin resources, composed mainly of aquatic insects (larvae and pupae) were the main food sources for 55.5% of the trophic guilds; whereas allochthonous items, mainly terrestrial insects and plant matter, were the main constituents of the diet for 44.4% of the guilds.However, the importance of terrestrial-origin items to the sandbank fishes can be seen in the large volume of terrestrial insects, which represented more than 40% of the total volume of the items used by the fish.The relative importance of terrestrial-and aquatic-origin food varies according to the composition of the fish assemblage, and also the habitat (Casatti, 2002;Gurgel et al., 2002;Melo et al., 2004;Dias et al., 2005).
Terrestrial and aquatic insects, plants, and sediments were present in the diet of more than 80% of the fish species analyzed.Luz et al. (2001), working with ponds on the Upper Paraná River floodplain, observed that insects were the most available resource in two of the three ponds in the study.Insects are the principal invertebrates in the food of fish.In general, we can state that nearly all species of fishes consume insects at some stage of their lives (Goulding et al., 1988).These arthropods function trophically as a bridge between the inaccessible or indigestible forms of primary production and the fish communities (Goulding et al., 1988).Luz-Agostinho et al. (2006), studying fishes of the Corumbá Reservoir, established that the availability of resources varied in the affluent streams, and that terrestrial insects were the most important resource.There are few tropical species with specialized feeding habits (Abelha et al., 2001).Most species are generalists, using available foods, although with a certain degree of preference.Riverine sandbank environments are markedly homogeneous and afford little food or shelter, leading to increased sharing of the limited variety of available resources.Nine main trophic guilds grouped the local ichthyofauna in respect to diet.However, the wide overlap in consumption of these resources makes these guilds very heterogeneous, even considering that this number has been recorded in samples from other kinds of habitat, as previously mentioned.The food resources responsible for sustaining the fish community in the environment were principally terrestrial and aquatic insects.

Fig. 1 .
Fig. 1.Scores for the fish species (a) and their food items (b) along Axes 1 and 2, derived from detrended correspondence analysis (DCA).The circles in (a) indicate the groups determined by the k-means analysis.

Fig. 3 .
Fig. 3. Food resource availability, inferred from the volume of items in the stomachs analyzed for all species combined.

Fig. 2 .
Fig. 2. Proportions of terrestrial, aquatic, and undetermined resources used by the fish trophic groups.