Diet and niche breadth and overlap in fish communities within the area affected by an Amazonian reservoir ( Amapá , Brazil )

We investigated the niche breadth and overlap of the fish species occurring in four environments affected by the Coaracy Nunes reservoir, in the Amapá Brazilian State. Seasonal samples of fishes were taken using a standard configuration of gillnets, as well as dragnets, lines, and cast-nets. Five hundred and forty stomach contents, representing 47 fish species were analyzed and quantified. Niche breadth and overlap were estimated using indexes of Levins and Pianka, respectively, while interspecific competition was evaluated using a null model (RA3). ANOVA and the kruskal-Wallis test were used, respectively, to evaluate differences in niche breadth and overlap between areas. The data indicate that the majority of the fish species belong to the piscivore, omnivore, and detritivore guilds. These species have likely colonized the environments due to the availability of suitable feeding resources, and the favorable physical conditions created by the river damming. Overall, few species have ample niches, but most of them are highly specialized. Resources seasonal variation had little effect on the feeding behavior of most species in the study areas. The null models indicated that competition was not a factor determining on community structure.


INTRODUCTION
Understanding the ecological mechanisms that support the coexistence of species in a given community and their partitioning of resources is one of the fundamental objectives of the ecological investigation of Neotropical fish assemblages (Cassemiro et al. 2008).Trophic resources partitioning is one of the principal factors that influence the structure of fish communities (Silva et al. 2008) and it may vary according to local characteristics, physical-chemical variables, the integrity of the environment, the composition of the fish fauna, and seasonality, as well as latitudinal gradients and other factors that may affect the dietary patterns and the feeding behavior of the fishes (Pianka 1969, Goulding 1980, Prejs and Prejs 1987, Hahn et al. 2004, Mérona and Mérona 2004, Lappalainen and Soininen 2006, Novakowski et al. 2008).
Niche breadth is an important parameter for the evaluation of the level of dietary specialization http://dx.doi.org/10.1590/0001-3765201420130053

383-405
Correspondence to: Júlio C. Sá-Oliveira E-mail: juliosa@unifap.brJÚLIO C. SÁ-OLIVEIRA, RONALDO ANGELINI and VICTORIA J. ISAAC-NAHUM in a given group of species (Segurado et al. 2011).Species with niches of reduced breadth are relatively specialized, whereas more ample niches are typical of generalist species.The analysis of niche overlap also provides an important approach for the evaluation of the structuring of communities in terms of the feeding niches of the different species that compose them (Corrêa et al. 2011).The degree of specialization for the exploitation of specific types of resources could be used to classify groups of species in feeding guilds (Winemiller and Pianka 1990).
Damming river causes changes to feeding behavior of species: herbivores can change their diets to invertivorous (Casatti et al. 2003), carnivores reduce predation on crustaceans and insects, making it essentially piscivorous (Santos 1995) and changing biotic interactions (competition and predation).Consequently, opportunistic strategy (feeding plasticity) is essential for species adaptation in the new environment (Araújo-Lima et al. 1995, Agostinho et al. 1999).In reservoir environments, the identification of the dietary resources that sustain populations and the understanding of feeding patterns are essential for the evaluation of the factors that are dominant on occurrence of the species in these environments and their distinctive regions (Esteves andGaletti 1994, Abelha et al. 2006).The Coaracy Nunes Dam was the first hydro-electric power station built in the Brazilian Amazon basin, with construction being started in 1967, and the reservoir being established in 1970 (ELETRONORTE 1997).The dam is located in the Ferreira Gomes municipality, in the state of Amapá.Despite its relatively long history, no research into the fish fauna of the area had been conducted prior to the present study.
This study compares the diets of the fish species in four regions influenced by Coaracy Nunes reservoir and it estimates niche breadth and overlap between the species taking into account the dry and wet seasons, in order to comprehend species resources partition.

MATERIALS AND METHODS
The Araguari is the main river of the Brazilian State of Amapá, with a total length of 498 km and a drainage basin of 38,000 km 2 .This river arises in the Tumucumaque range and discharges into the Atlantic Ocean, although it is strongly influenced by the Amazon River.The Coaracy Nunes reservoir is located approximately 200 km from the Atlantic Ocean, in the middle of Araguari River basin.The reservoir drains a total area of 23.5 km 2 , and has a mean discharge of 976 m 3 .s - , mean depth of 15 m, and a total volume of 138 Hm 3 .The local climate is typical of the Amazon basin, with a rainy season between January and June, and a dry season from July to December (Bezerra et al. 1990, IBGE 2010, ANA 2011).
The region's vegetation is characterized by elements of the typical lowland Amazon rainforest, savanna, and floodplain forest (várzea).For the present study, four areas influenced by the Coaracy Nunes reservoir (Fig. 1) were discriminated: 1 -Downriver Area (DWN): located downstream from the dam, this area presents lotic characteristics with the flow of water being influenced by the control of the dam's flood gates and the discharge of the turbines in the hydro-electric power station, which it could create areas with reduced flow; 2 -Reservoir (RES): main body of the reservoir, with semi-lotic characteristics intermediate between those of a river and a lake; 3 -Lacustrine (LAk): an area adjacent to the reservoir, with extremely lentic characteristics; 4 -Upriver Area (UPR): area upstream from the reservoir with lotic conditions.The effects of deforestation and permanent flooding from várzea are apparent throughout this area.A number of deforested areas and scattered settlements can be observed in the middle and upper reaches of the reservoir.The margins of the lower reservoir, lacustrine, and downriver areas are characterized by relatively well-preserved riparian vegetation.NICHE IN FISHES IN AMAZONIAN RESERVOIR Specimen collection was divided into eight bimonthly sampling campaigns between May 2009, and July 2010, with four samplings in flood season and four in dry season.Sampling sites were selected randomly within each area.The collection of specimens for the analysis of the composition of the community and the diet of the different species was conducted using a number of different fishing techniques, including cast nets, trawls, dragnets, harpoons, spears, hand-nets, hand-lines, and standardized samples with eight gillnets (mesh size of 10 -100 mm).
All captured specimens were identified to the lowest possible taxonomic level, measured (total length in mm), weighed (g), and photographed.Species identification was based on the available literature (Planquette et al. 1996, Santos et al. 2004, Buckup et al. 2007, PIATAM 2008) and was confirmed by specialists.Diet composition was based on analysis of 5 up to 10 stomach contents of the 47 most abundant species because the other species presented no content.Diet items were standardized in ten categories (Hahn and Delariva 2003, Mérona et al. 2005, Novakowki et al. 2007): 1 -plant material (unidentified remains of leaves, flowers and algae); 2 -insect; 3 -larva (terrestrial or aquatic); 4 -zooplankton; 5 -phytoplankton; 6crustacean (crab or shrimp); 7 -fish (whole animals or remains, including scales and fins); 8 -arthropod (other representatives of the phylum Arthropoda); 9 -detritus (organic detritus at different stages of decomposition); 10 -animal fraction (unidentified fraction of non-fish vertebrates).
Diet composition was measured by data volume, which was 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.
Diet composition was analyzed by volume (VO % ) and the frequency of occurrence (F O% ) using an optical microscope and a stereomicroscope (Hynes 1950, Hyslop 1980).These two parameters were combined to calculate the feeding index.IAi was calculated to characterize fish species diets (kawakami and Vazzoler 1980), which combines total volume (%) and frequency of occurrence (%) of each item (lowest taxonomic level):

5
*100 where IAi = alimentary index; FOi = occurrence frequency percentage and VOi = volumetric frequency percentage; i = 1, 2,.., n food item; Based on this analysis, the diet preferences and feeding specialization of the different species were evaluated on the basis of a FI ≥ 0.5 criterion for a given category or type of item.In some specific cases, where a number of different items were consumed in relatively reduced proportions, a criterion of FI ≥ 0.4 was adopted (Gaspar da Luz et al. 2001).Species that presented a co-dominance of plant and animal items, or a relatively balanced consumption (difference < 20%) of the two types of item, were considered to be omnivorous.
The niche breadth of each species was based on Levin's standardized index: where Bi = standardized index of niche breadth, pij = proportion of diet of predator i on prey j, and n = total number of item (resources).Bi values vary from 0 (species consume a single item) to 1 (species exploits available items in equal proportion).Values of Bi are considered high when higher than 0.6, moderate, when between 0.4 and 0.6 and low when below 0.4 (Novakowski et al. 2008).
Analysis of niche overlap between the most common species was based on classical Pianka's index (Pianka 1973), which is derived from the composition of the diet (percentages) of the different species: where O jk = Pianka's index of niche overlap between species j and k, pij = the proportion of the ith resource in the diet of species j, p ik = the proportion of the i the resource in the diet of species k, and n = the total number of items.Pianka index values were classified according to the scheme of Grossman (1986) and Novakowski et al. (2008) which follow the same boundaries as those of Levins index (see above).A basic assumption adopted here was that the different dietary resources were equally accessible to all species, given that no data was collected on the availability of resources within the study area (Abelha et al. 2006).
The data were initially analyzed for normality using the kolmogorov-Smirnov test and Levene's test for homogeneity of variance (Conover 1990, Sokal andRohlf 1995).Seasonal differences in the mean indices of niche breadth and overlap were evaluated using an Analysis of Variance (ANOVA) and the Kruskal-Wallis test respectively.A α < 0.05 significance level was considered for all tests.
In order to evaluate whether the pattern of niche overlap diverged significantly from a random distribution (absence of overlap), data on the abundance of diet resources by each species were randomized using null models based on 5000 iterations using the RA3 algorithm (randomization NICHE IN FISHES IN AMAZONIAN RESERVOIR algorithm) of the EcoSim program (Gotelli and Entsminger 2006), which runs a Monte Carlo resampling in order to create "pseudocommunities" (Joern andLawlor 1980, Winemiller andPianka 1990), and then compares the random communities statistically with the observed data set.The statistical significance of observed overlap with that indicated by the null model was evaluated considering α < 0.05.In this analysis, interspecific competition was suspected when the observed mean overlap was significantly lower than that expected by chance.When observed overlap is greater than that expected by chance, abiotic limitations could be provoking the homogenization of foraging patterns among species (Albrecht and Gotelli 2001).

RESULTS
A total of 108 species (Table A9-Appendix) and 1977 fish specimens were captured during the present study, of which 540 had stomach contents belonging to 47 species, which were included in the analysis of diet (Tables I and II).Half of the fish species (51%) consumed more than one item in all areas and seasons, except in the reservoir during the dry season, when 55% of the species consumed a single resource, reflecting a higher specialization.Fish was the item most consumed in all areas, followed by detritus and plant material (Tables I and II, Fig. 2).
In the downriver area, fish was the main item (23%) during the flood season, while detritus was the most consumed (18%) in the dry season.In the reservoir, fish was the main item (19%) during the dry season, whereas fish (21%), detritus (21%) and plant (21%) were the main items during the flood season.Two items -fish and insects -were the most consumed (25%) in the lacustrine area during the dry season, while insects and detritus were the main items (21%) during the flood season.In the upriver area, detritus was the most consumed item in both the dry (25%) and the flood (20%) seasons (Fig. 2).
In the downriver area, some species with a diverse diet presented a co-dominance of dietary items.These species included H. unimaculatus who consumed plant material and detritus in equal proportions.A similar pattern was observed in the reservoir and lake environments, in species such as G. proximus, H. unimaculatus, L. aff.parae, L. affinis, Leptodoras sp., and T. auritus who also presented relatively diversified diets, with a predominance of plant material.Triportheus angulatus consumed insects, arthropods, and plant material in roughly equal proportions, while the diet of L. affinis was based on three main itemsfish, insects, and plant material.In both the lake and upriver areas, equal proportions of detritus and plant material dominated the diet of H. unimaculatus.
Slight seasonal variation was observed in the diet of the majority of species (Tables I and II).Accordingly, Plagioscion squamosissimus, in the downriver area, changed its diet during dry season, feeding mostly on invertebrates (crustaceans, insects and arthropods).At reservoir area, A.      lacustrine area, where L. affi nis (and Pimelodus ornatus) fed on a greater diversity of items during the fl ood period, while C. gibbosus consumed more in the dry season.In the upriver area, A. ucayalensis and T. angulatus consumed more items in the dry season, while S. gibbus ingested more during the fl ood season (Tables I and II).
Niche breadth (Bi) values varied from 0.00 to 0.65.Most species presented relatively low values (Bi < 0.4) in all four areas (Tables I and II).In all areas and seasons, more than half the species returned Bi values of zero, although some species presented much higher values, such as Satanoperca acuticeps in the downriver area during the fl ood period (Bi = 0.65).The high frequency of Bi values lower than 0.4 in all the areas indicate that most species have relatively limited niches.However, increased variation (Bi > 0.4) was found in the lake and reservoir during the fl ood period, and in the downriver area during the dry season, indicating the presence of broader niches within these areas in comparison with the upriver area, where narrower niches were more typical (Fig. 3, Tables I and II).Nevertheless, no statistical difference was found in niche breadth (Fig. 3) among areas (ANOVA: F (3, 72) = 2.5301; p = 0.0639), seasons (F (1, 127) = 2.8002; p = 0.09671) or the season-area interface (F (3, 127) = 4.1386; p = 0.0776).
In general, niche overlap was greater in pairs of more specialized species, such as the piscivores:  diverse diet, feeding preferentially on fish, but also insects, zooplankton, and detritus, which reinforced the overlap of the feeding niche of this species with piscivores and omnivores.
The highest proportions of high overlap values (Oi > 0.6) were recorded in the upriver area during the flood period (37.5%), in the reservoir during both seasons (35.5%), and in downriver area, also during both seasons (31.5%).In the lacustrine area (dry = 28.57%;flood = 26.39%),high overlap values were less frequent, since they were in the upriver area during the dry season (26.5%).
These high values of Oi (> 0.6) could be indicative of the influence of interspecific competition between the pairs of species.However, observed overlap was significantly higher (p < 0.05) than expected according to the null (RA3) models (Table III), exposing that interspecific competition may not have constituted a major pressure in any of the seasons in any of the study areas.

Area -Season
Mean observed

TABLE III Probability test of null models (RA3) between the mean observed and expected trophic niche overlap for fish assemblages in the areas of influence of Coaracy Nunes Dam (State of Amapá, Brazil). (p-value = pobs averages observed; pesq = p-value of expected average).
During the present study, large quantities of Macrobrachium shrimp were captured as bycatch during trawls, especially in the impounded areas, which indicates the availability of this resource as a local complementary item used by many species and as the main item for M. chrysargyrea and G. proximus.JÚLIO C. SÁ-OLIVEIRA, RONALDO ANGELINI and VICTORIA J. ISAAC-NAHUM

DISCUSSION
Small seasonal variation between consumed items in the different areas caused insignificant change on the breadth values of the seasons and areas.The smallest seasonal variation in diets of piscivores, herbivores and detritivores may reflect the abundances of the resources exploited by these guilds throughout the study period.While discreet, the variation observed in the diets of the species of the remaining guilds was probably related to seasonal fluctuations in the availability of preferred items.Niche overlap did not change either, and despite high values concerning some guilds, especially detritivores and piscivores, the interspecific competition does not control the community development.Despite the high diversity of fish species in the reservoir (108), 47 of them remain in the same guilds regardless of season or region.Following 40 years of damming, fish species in Coaracy Nunes reservoir apparently reached trophic homogeneity.
Even though in some environments the temporal dynamics influence the carbon source and consequently the diet composition of many fish species (zeug and Winemiller 2008) in different habitats in Pantanal, a Brazilian floodplain, there is no pattern on the use of seasonal food resources (Novakowski et al. 2008, Angelini et al. 2013).In our study, few generalist species such as L. affinis and C. gibbosus and the piscivore-omnivores A. falcirostris and P. ornatus showed opportunistic feeding behavior, which was probably related to the seasonal resource abundance (Araújo-Lima et al. 1995).Otherwise the highly specialized feeding behavior of some species, such as the piscivores H. aimara, B. cuvieri, and S. gibbus, can be accounted for by the relative abundance of prey species (e.g., H. unimaculatus) within all the areas studied.
The relatively high frequencies of fish, crustaceans, insects, plant material, and detritus in the diets of the species analyzed in the present study are similar to the pattern recorded in other reservoirs in Brazil (Ferreira 1984, Hahn et al. 1998) reflecting the typical pattern expected for a fish community in South America, in particular in artificial reservoirs (Mérona et al. 2001, Loureiro 2000, Angelini et al. 2006) where the opportunistic behavior does not mean that the species are able to use the full diversity of available feeding resources, but that they may shift from one resource to another, according to their needs (Gerking 1994).
In the same way, low levels of consumption of plankton in the study areas, as well as the absence of fish species specialized for the exploitation of this resource were consistent with the reduced abundance of planktivores in reservoirs, as recorded at a number of locations (e.g., Agostinho et al. 1994, Hahn et al. 1998, Delariva 2002).Species in downriver area, consumed more plankton than in other areas, but even so, in small amounts.This area is mainly characterized by a more heterogeneous environment, with marginal lakes rich in nutrients that may support the production of phytoplankton, which are transferred to the river during the ebb period, but other resources seem more abundant since they were more prevalent in stomachs contents.
The presence of plant material in the diets of a number of species in all four study areas could be related to the availability of this resource, derived from the riparian vegetation, which occurs throughout the study area, and contributes with fruits, seeds, flowers, and filamentous algae to the resource base.This indicates that the colonization of environments such as reservoirs by herbivorous species is related to both the composition of the original fish fauna of the river prior to impoundment (Agostinho et al. 1999, Silva et al. 2008) and the availability of this resource, derived primarily from the riparian vegetation (Barthem and Goulding 1997).In the present study, despite the small number of herbivorous species recorded overall, plant material was an important complement of the diets of many species, in particular, omnivores.

NICHE IN FISHES IN AMAZONIAN RESERVOIR
The coexistence of species with the same feeding habits depends on the breadth and overlap of their niches (Pielou 1972, Evans 1983).While we did not detect any seasonal variation in niche breadth in any of the study areas, the broader niches recorded in the reservoir and lake during the flood period could indicate that a more ample resource base, in terms of both diversity and abundance, was available during this period.Mérona et al. (2003) concluded that the reduced niche breadth generally found in reservoir fishes -as observed in the present study -indicates that the populations of generalist species could become reduced in size or even extinct as the environment becomes increasingly stable.
It seems reasonable to conclude that the adoption of a more specialized feeding strategy may be advantageous in older reservoirs (Silva et al. 2008), and that this tendency may at least partly account for the predominance of the narrow niches recorded in the present study, given that Coaracy Nunes reservoir is now more than 40 years old.These results also suggest that niche breadth is not an important factor regulating the diversity of species in the reservoir or lake area, and that this conclusion also applies to other reservoirs (Agostinho et al. 2005), given that, theoretically, a reduction in niche breadth would be expected with an increase in the number of species in the community (Schoener 1974).
In the downriver area, the broader niches recorded during the dry season could have been related to the increase in environmental heterogeneity and the abundance of resources caused by decrease of level of water which lead to the creation of many habitats and increased the density of fish fauna.These factors tend to reinforce competition, which would force the less competitive and/or more specialized species to modify their diets or include additional items in order to coexist (Pielou 1972).
The high degree of niche overlap recorded in the present study for piscivores and detritivores may reflect the relatively ample categories adopted for the classification of the resources exploited by these guilds.This could have resulted in an overestimate of overlap (Uieda 1983, Sabino andCastro 1990), given that the specific details that differentiate the diets could have been overlooked.
Overlap in the detritivores was related primarily to the marked abundance of this resource throughout the study area, without necessarily being reflected in competitive processes given that, in addition to the relative abundance of resources, the species tended to segregate over time and space during foraging, especially in complex habitats (Matthews 1998, Schoener 1974, Colwell and Futuyma 1971).
The null model analysis indicated that mean niche overlap was significantly higher than that expected, which suggested that interspecific competition was not a significant mechanism of niche partitioning in any of the fish communities within the study area, and that the species tended to share the most abundant resources.Under these conditions, the absence of competition would be expected (Pianka 2000), and variation in population parameters would be determined by fluctuations in abiotic factors, such as the unpredictable daily variation in the level of the reservoir, which creates a permanent state of instability on the study area.This affects, not only the feeding behavior, but also the reproductive patterns and the predator defense of the different species (Agostinho et al. 1999, Oliveira et al. 2005).

CONCLUSION
The results of the present study indicated that the species that composed the fish communities of the area influenced by the Coaracy Nunes reservoir were able to share preferred resources with small variations among seasons and areas, reflecting feeding resources abundance.Niche overlap did not change either, and despite high values concerning some guilds, espe-cially detritivores and piscivores, whose resources were abundant, and interspecific competition does not control the community development.

ACKOWLEDGMENTS
This study is part of Julio C. Sá-Oliveira's Ph.D. dissertation for his Post Graduate in Ecology at the Federal University of Pará, Brazil, under guidance of V. J. Isaac-Nahum.We would like to thank Centrais Elétricas do Norte -ELETRONORTE and Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) for scholarships.Reviewer #1 made valuable suggestions on the manuscript.
ucayalensis and H. unimaculatus ingested a more ample diversity of items during the flood period while L. affinis consumed more items during the dry season, as opposed to what occurred in the JÚLIO C. SÁ-OLIVEIRA, RONALDO ANGELINI and VICTORIA J. ISAAC-Plant material zooplankton Phytoplankton Detritus Guild Downriver Dry Flood Dry Flood Dry Flood Dry Flood Dry Flood Dry Flood Dry Flood Dry Flood Dry Flood Dry Flood

TABLE I Alimentary index (AI) values and trophic classification of species analyzed in the areas under influence of Coaracy Nunes Dam (Amapá State, Brazil; n: number of stomachs analyzed
). NICHE IN FISHES IN AMAZONIAN RESERVOIR

TABLE I (
continuation) NICHE IN FISHES IN AMAZONIAN RESERVOIR

TABLE A3 Niche overlap between pairs of species analyzed from Reservoir area of Coaracy Nunes Dam, Ferreira Gomes -Amapá (AM-Brazil). Reservoir -dry
Af Au Bc Sg Ee Cg Po Gp Hu Lap Laf Lsp Tan Tau Paf Psp Cin Hyp Mly Ttri