Aquatic insects as the main food resource of fish the community in a Neotropical reservoir

We evaluated the feeding of fish species of the Nova Avanhandava Reservoir, low Tietê River, São Paulo State, Brazil. Fishes were collected in two stretches of the reservoir: Santa Bárbara (14 samples) and Bonito (two samples) between September 2002 and March 2004, using gill and seining nets. The results of stomach contents analysis were expressed with the frequency of occurrence and gravimetric method, combined in the Alimentary Index (AI). The 20 species studied consumed 52 food items, grouped in 10 food categories: aquatic insects, terrestrial insects, crustaceans, fish, macroinvertebrates, microcrustaceans, algae, vegetal matter, detritus/sediment and scales. The aquatic insects (mainly Chironomidae, Odonata and Ephemeroptera) were the most common food resources, consumed by 18 species. The diet composition of the community (species grouped) indicated that the dominant food category in the diet of fishes was aquatic insects (AI = 77.6%), followed by crustaceans (AI = 7.1%). Four trophic guilds were identified according a cluster analysis (Pearson distance): insectivorous (10 species), omnivorous (4 species), detritivorous (3 species) and piscivorous/carcinophagous (3 species). Despite the highest number of species, the insectivorous guild was responsible for more than 80% in captures in number and biomass (CPUEn and CPUEb). The low values of niche breadth presented by all species, along with the low values of diet overlap between species pairs indicate a high degree of food resources partitioning among species. The aquatic insects, despite being the main food resource of insectivorous fishes, also complemented the diet of other species, which demonstrate the importance of this food resource for the fish community, sustaining a high diversity, abundance and biomass of fishes.


Introduction
Studies of trophic structure and resource partitioning of fish assemblages are important for the understanding of ecological relationships among fish species and with other aquatic organisms, which can support management of natural populations (Agostinho et al., 1995;Woodward & Hildrew, 2002).In fish communities, the abundance of trophic guilds is greatly dependent on environment conditions.Some can present a high abundance of piscivorous fishes, especially in the first years of closure of reservoirs, since there is a high abundance of prey (Mérona et al., 2001(Mérona et al., , 2003)).Omnivorous fish may be abundant in environments with seasonal variations, such as in the Pantanal floodplain (Resende, 2000).Detritivorous fish are abundant in some reservoirs (Carvalho et al., 1998;Alvim & Peret, 2004) and floodplains (Peretti & Andrian, 2004), where the detritus and sediment are important in the maintenance of fish fauna, leading to increases in the energetic efficiency of the community, and promoting the cycling of the matter and nutrients.
In reservoirs, the main food resources consumed by the fish fauna are originated in the aquatic system, such as aquatic insects, other invertebrates, zooplankton, detritus and fish (Agostinho et al., 2007), sustaining a great diversity and abundance of fishes (Araújo-Lima et al., 1995).Among aquatic insects, several groups are abundant in Neotropical reservoirs, especially dipterans of the family Chironomidae, ephemeropterans and odonats (Higuti et al., 2005;Hahn & Fugi, 2007).These organisms have an important role in the metabolism of aquatic systems (Pereira & De Luca, 2003), and as food resource for fishes in streams (Pouilly et al., 2006;Pinto & Uieda, 2007) and higher order rivers (Horeau et al., 1998).They are also significant food resource for fishes in reservoirs (Arcifa & Meschiatti, 1993;Callisto et al., 2002;Luz-Agostinho et al., 2006).
The aim of this study was to characterize the feeding ecology of the fish community of a Neotropical reservoir (Nova Avanhandava Reservoir, low Tietê River, São Paulo State, Brazil), and to access the importance of aquatic insects for the trophic structure of this fish community.

Study site and samplings
The Nova Avanhandava Reservoir, located at 358 m of altitude, is the fifth reservoir of the Tietê River cascade, built in 1982 for hydroelectricity generation.The reservoir has a surface area of 210 km 2 , total water volume of 2.720 x 10 6 m 3 , mean discharge rate of 688 m 3 .s - , maximum depth of 30 m and water permanence time of 46 days (Torloni et al., 1993;Rodgher et al., 2002).According to the energy company AES Tietê, variation in the level of the water was almost insignificant in the period studied (357.9 to 358.3 m).
The fish were collected in two stretches of the reservoir, the Santa Bárbara (21º05'25"S 50º07'18"W) and the Bonito (21º17'50"S 50º08'17"W) (Fig. 1).The fish samplings were Fish were caught in six sites of each stretch, using gillnets with mesh sizes from 3 to 14 cm (opposite knot lengths).Gill nets were placed in the littoral and deeper zones in the afternoon and removed in the following morning (14 hours of exposure).We also used a seining net (10 x 1.5 m, 5 mm mesh) occasionally in some months to capture young fish of some species that occurred in low frequency in gill nets.Nets were handled by two individuals and dragged three times into aquatic vegetation and margin of the reservoir.Fish collected were identified to species, weighed (0.01 g approximation) and measured (total and standard length), preserved in 10% formalin and conserved in 70º GL alcohol.Voucher specimens were deposited in the Museu de Zoologia -Universidade Estadual de Londrina (MZUEL 4729 to MZUEL 4760), Londrina, Paraná.
A sub-sampling of all sizes of each species was performed (about 10 individuals of each species per sampling), and stomach with contents were removed and preserved in 10% formalin.In the laboratory, these stomachs were transferred to 70° GL alcohol.

Diet analysis
Stomach contents were analyzed under a stereoscopic microscope, identified to the most detailed taxonomic level possible and weighed (wet weight) on a scale with 0.0001 g approximation; when this procedure was not possible (for small items), we established a percentage of total content weight (Hynes, 1950).The results were expressed as the frequency of occurrence and gravimetric method (Hyslop, 1980), combined in the Alimentary Index (Kawakami & Vazzoler, 1980).
Fish were grouped into trophic guilds according to a cluster analysis performed with the data of relative weight of food categories, using the Pearson distance and UPGMA amalgamation method.The analysis was computed with the software Statistica 5.1 (StatSoft, Inc., 1996).
The contribution of trophic guilds to the community was calculated as the percentage of CPUE in number of individuals (CPUEn) and biomass (CPUEb), expressed as number or kg of fishes by 1000 m 2 of gill nets, during 14 hours of exposure (Agostinho & Gomes, 1997).Fish caught with seining nets were not included in the CPUE, due to the differences in the catch effort.
The relative level of diet specialization of the species (niche breadth) was calculated by the Levin's index (Hurlbert, 1978): where B i is the niche breath standardized index, P ij is the proportion of the food item j in the diet of i species and n is the number of food items.The B value ranges from 0 (species consumed mainly one food resource) to 1 (species consumed all food resources in similar proportions).The values of niche breadth were arbitrarily considered high (> 0.6), intermediate (0.4 -0.6) or low (< 0.4) (Novakowski et al., 2008).
The diet overlap between species was calculated with the relative weight of food items, using the Pianka Index (1973): where: θ jk = Pianka's measure of diet overlap index between j and k species; P ij = proportion resource i of the total resources used by species j; P ik = proportion resource i of the total resources used by species k; n = total number of resource states.This index varies from 0 (no overlap) to 1 (total overlap).The overlap values were arbitrarily considered high (> 0.6), intermediate (0.4 -0.6) or low (< 0.4) (Novakowski et al., 2008).
A null model was used to evaluate the significance of the food overlap between species (Gotelli & Entsminger, 2007).The matrix was randomized 10,000 times, using the following options: niche breadth retained and zeroes reshuffled (RA3, Winemiller & Pianka, 1990).The mean overlap observed is compared to the mean overlap calculated in the null distribution (p < 0.05).The feeding overlap and null model analysis were computed using EcoSim 7.0 (Gotelli & Entsminger, 2007).

Results
A total of 544 stomachs from 20 species had the contents analyzed (Table 1).The most widely consumed food resources were aquatic insects (mainly Chironomidae, Libelullidae and Polymitarcyidae; 18 fish species; 90% of the total species); vegetal matter (17 species), detritus/sediment and macroinvertebrates (15 species) and microcrustaceans, terrestrial insects and fish (12 species).
The diet composition of the community (all species grouped) indicated that the dominant food category in the diet of species was aquatic insects, with AI = 77.6%,followed by crustaceans (AI = 7.1%) and vegetal matter (5.2%) (Fig. 2).The other food resources represented about 10% of the diet of the community.
Detritivorous: Steindachnerina insculpta, Oreochromis niloticus and Leporinus friderici, which feed exclusively on detritus or complement their diet with aquatic insects, fish and vegetal matter.
Within the insectivorous guild, the majority of species were observed to be aquatic insect feeders, except M. intermedia, which consumed most terrestrial insects (particularly Hymenoptera).Despite occupying the same trophic guild, species differ in the items consumed.Gymnotus carapo, R. quelen, P. squamosissimus and S. macrurus   1.
consumed mostly Odonata nymphs (Libellulidae and Gomphidae and non-identified fragments), while S. maculatus and T. nematurus consumed mainly Ephemeroptera Polymitarcyidae nymphs and sub-imagos.Diptera Chironomidae larvae and pupae were important items in the diet of A. altiparanae, M. intermedia, R. quelen and S. pappaterra.In addition, many species from other trophic guilds, such as A. affinis and T. rendalli complemented their diet with this food item.Aquatic insects that presented a high occurrence in the diet of species were Chironomidae (17 species; preferential item in 7 species), followed by Libellulidae (11 species, preferential item in 5) and Polymitarcyidae (6 species, preferential item in 4).Odonata Gomphidae was eaten by eleven species, and three species demonstrated preference for this item.Niche breadth presented low values for all species (mean = 0.20), varying from 0 (S.insculpta) to 0.37 (G.proximus), indicating that species has a restrict diet, consuming preferentially a small number of food items (Table 1).
The feeding overlap were greater than expected by chance (p < 0.00001), indicating that the values did not occur at random.The values of diet overlap were low (mean = 0.17) for 85% of species pairs (Table 2).Species that presented intermediate or high diet overlap were mainly from the same trophic guild.Among insectivorous fishes, the intermediatehigh values occurred in 22 % of species pairs (10 interactions), while for all species pairs of herbivorous, detritivorous and piscivorous/carcinophagous intermediate or high values of overlap were observed.Other eight values of intermediate or high overlap happened among fishes from different guilds, but that share some resources, such as detritus, vegetal matter and shrimp (Table 2).

Discussion
In the Nova Avanhandava Reservoir, ten species used aquatic insects (mainly Chironomidae, Odonata and Ephemeroptera) and terrestrial insects as their main food resource, and all the other species, in some extent, complemented their diet with these food resources, indicating that aquatic insects are the main food resource in this reservoir.Despite aquatic insects are an important food resource for fishes in floodplains (Hahn et al., 1997b) and in recently impounded rivers (Luz-Agostinho et al., 2006), the insectivorous fishes are not the dominant trophic guild in the majority of Brazilian reservoirs (Agostinho et al., 2007).
In this reservoir, the highest captures in number and biomass of fish occur in the littoral zones (Vidotto & Carvalho, 2007), which can explain the high consumption of aquatic insects by many species.The littoral zones presents high spatial heterogeneity due to the presence of macrophytes  and woody debris that allow the colonization of a diversified associated fauna, such as crustaceans and aquatic insects (Oliveira et al., 2005).They are abundant food resources for species of diverse trophic guilds (Hahn et al., 1997b), sustaining great fish diversity in these regions of reservoirs (Gido et al., 2002).
In the littoral regions, aquatic insects have a great ability to colonize different types of substratum, such as sediment (Covich et al., 1999), leaves, bulks, stones and macrophytes (Fulan & Henry, 2006).This wide distribution in the water body allows the capture by different Neotropical fish that present diverse feeding tactics: the bottom-feeders, such as S. pappaterra and G. proximus, since they also consumed detritus and sediment; the surface feeders, such as T. nematurus and S. maculatus, which capture large amounts of sub-imagos of Ephemeroptera, probably at the moment of their migration to emerge (Hahn et al., 1997b); and species that feed near macrophyte stands, where the Odonata nymphs and Chironomidae are abundant (Fulan & Henry, 2007), such as A. altiparanae, P. squamosissimus, G. carapo, S. macrurus and R. quelen.
The insectivorous species A. altiparanae, P. squamosissimus and S. maculatus are dominant in terms of CPUEn and CPUEb in this reservoir (Vidotto & Carvalho, 2009), and widespread in several Brazilian reservoirs (Agostinho et al., 1995).Astyanax altiparanae is widely recognized for its great trophic plasticity and ecological valence (Agostinho et al., 1999), and ability in colonizing recently-impounded reservoirs (Dias et al., 2005) and other environments (Orsi et al., 2004).In this reservoir, this species presented a diet based on Chironomidae larvae and Polymitarcyidae nymphs, complemented with terrestrial Hymenoptera, characterizing a strict insectivorous diet.Casatti et al. (2003) observed the preference for terrestrial insects in the Rosana Reservoir (Paranapanema River).However, under certain conditions, A. altiparanae seem to prefer vegetal matter and seeds (Cassemiro et al., 2002); under other conditions the species is omnivorous, consuming both vegetal and animal matter (Bennemann et al., 2005), demonstrating the high trophic plasticity of this species in different environments.
Plagioscion squamosissimus, an introduced species, was first registered in the commercial catch around 1980 in this reservoir (Torloni et al., 1993).This species, widely described as piscivorous (Almeida et al., 1997;Hahn et al., 1997a), presented Odonata nymphs and shrimp as their main food resources in Nova Avanhandava Reservoir.In stretches under the influence of the Capivara Reservoir (Paranapanema River), Bennemann et al. (2006) observed that, since the introduction of corvina in 1992 to 2002, this species switched their diet from fish (mainly A. altiparanae) to shrimps, due to the reduction in the abundance of prey species, and to the fact that this species is not a selective predator, eating all prey that is available.
Serrasalmus maculatus, another species generally known as piscivorous (Hahn et al., 1998;Agostinho et al., 2003), based its diet mainly on Polymitarcyidae sub-imagos and Gomphidae nymphs in the Nova Avanhandava Reservoir.The aquatic insects comprised a great part of the diet of small specimens (Ls < 5 cm) of S. spilopleura (= S. maculatus) analyzed by Casatti et al. (2003).Costa et al. (2005) observed a trophic ontogeny in the diet of S. spilopleura (= S. maculatus), where individuals with Ls from 6 to 12 cm consumed aquatic insects, but individuals larger than 12 cm preferred fish, becoming exclusively piscivorous at 22 cm.In the present study, this feeding habit was observed in individuals with Ls from 4.5 to 23.5 cm, indicating that even adult fish prefer aquatic insects in the Nova Avanhandava Reservoir.
Despite the high number of food items explored, all species presented low values of niche breadth, indicating the use of a limited number of resources.The high dominance by few resources suggest their abundance in the environment (Novakowski et al., 2008), and is not related to morphological limitations to the consumption of specific resources.
Low diet overlap values were observed in almost all species pairs, indicating that species present differences in the diet composition.These results, combined with the low niche breadth of all species suggest a high degree of food partitioning among species, and has been observed in other fish communities of the Neotropical region (Uieda et al., 1997;Esteves et al., 2008;Novakowski et al., 2008;Mérona & Rankin-de-Mérona, 2004).However, when overlap was analyzed independently within each trophic guild, some different patterns were observed.Among insectivorous fishes, the trophic guild with the greatest number of species, few species pair presented overlap values higher than 0.4, indicating a spatial segregation in the use of the food resources, that may be due to differences in the strategies of foraging, as discussed above.Within the detritivorous and herbivorous only intermediate or high values of overlap were observed, but these species share abundant food resources, such as detritus and vegetal matter, and may use these resources opportunistically (Mathews, 1998).For fishes from the piscivorous/carcinophagous guild, the difficulty in the identification of many preys may have interfered in the results, due to the high stage of digestion of the fishes.Nevertheless, even when species present high niche overlap other factors can allow coexistence, such as spatial heterogeneity and habitat complexity, combined with environmental, temporal, population and behavioral aspects (Schoener, 1974;May, 1986).
In conclusion, we observed that aquatic insects play an important role in the maintenance of the fish community of the Nova Avanhandava Reservoir, sustaining a high diversity of species and abundance of fish.In addition, species present different feeding styles, independent of the trophic position, leading to a plentiful partitioning of the food resources.
carried out monthly between September 2002 and August 2003.In the Santa Bárbara stretch it also occurred in October and December 2003, and March 2004 (14 samples); and in July 2003 and March 2004 in the Bonito stretch.

Fig. 3 .
Fig. 3. Cluster analysis of diet similarity among species, showing the trophic guilds of the fish community from the Nova Avanhandava Reservoir.

Fig. 4 .
Fig. 4. Relative catch per unit effort (CPUE) in number (CPUEn) and biomass (CPUEb) of trophic guilds of the fish community of the Nova Avanhandava Reservoir.