Meso and microhabitat analysis and feeding habits of small nektonic characins (Teleostei: Characiformes) in Neotropical streams

The habitat use and feeding habits of a set of nektonic fish species often found in small low-gradient streams in Brazil were investigated. The core species in the present study was Knodus moenkhausii (Eigenmann & Kennedy, 1903), the most abundant of five species in the nektonic guild. Records of the species in 22 streams indicate that K. moenkhausii is associated, on a mesohabitat scale, with runs, and on a microhabitat scale, with sandy bottoms, intermediate depth, and open sites without coverage or submerged vegetation. During snorkeling observations, two additional nektonic registered species showed spatial segregation from K. moenkhausii with respect to foraging microhabitats. Feeding habits of the observed nektonic species were significantly correlated with riparian vegetation. The six species studied appear to use different resources, as suggested by a non-metric ordination of diet, and sites with characteristic riparian vegetation. The studied species may be considered generalists with a tendency to insectivory, with K. moenkhausii feeding on the widest variety of resources. Such opportunism explains its ability to successfully occupy instream habitats with low complexity.


MATERIAL AND METHODS
The study site, situated in north-western São Paulo (Brazilian state), is characterized by a hot tropical climate (NIMER 1989) and is used primarily for livestock grazing. The rainy season is often registered from October to March, with maximum average temperature (31ºC) in January, followed by a dry season from April to September, with minimum average temperature (13ºC) registered in July (IPT 2000).
Samples were taken during the 2003-2004 dry seasons from tributary streams of the rivers São José dos Dourados, Turvo, and Grande (see Appendix 1). Standardized electrofishing (two passes of 25 minutes each) was conducted across all 95 collecting sites, in a 75 m stream section previously blocked by 5 mm mesh stop-nets located up and downstream. Specimens were fixed in 10% formalin, transferred to 70% EtOH, and deposited in the fish collection of the Departamento de Zoologia e Botânica, Universidade Estadual Paulista, São José do Rio Preto, São Paulo (DZSJRP).
The occurrence and mesohabitat use of K. moenkhausii and other nektonic species were evaluated in 22 (Appendix 1) of the 95 streams sampled. In this analysis, our goals were to determine whether the distribution and abundance of the set of nektonic species were associated with particular stream features, and whether their spatial distribution could be explained by habitat descriptors. In order to accomplish that, we performed a Canonical Correspondence Analysis (CCA) using the CANOCO 4.5 software (TER BRAAK & SMILAUER 2002) based on matrices with habitat variables and species abundance. Neither data transformation, nor interaction terms were defined, and no rescaling was selected among the CANOCO options. Nevertheless, rare species were down weighted.
Amongst 16 environmental descriptors used in the exploratory analysis, seven (those showing highest correlations with species abundance) were selected and categorized in an environmental matrix (Tab. I). Because habitat volume was not equal between sites, the relative abundance of each species was calculated by its abundance per habitat volume. The habitat volume, consisting of the water volume contained in the sampled stream stretch, was calculated (depth x width x length) and a matrix of species abundance per volume of habitat was constructed to minimize any bias.
In order to determine which microhabitats are predominantly exploited by K. moenkhausii, snorkeling sessions were conducted in a segment of an additional stream, the Ribeirão do Arara (20º09'28.3"S 50º32'17.6"W), a Rio Grande tributary located in the municipality of Dolcinópolis. Diving sessions were performed on three occasions (11 th September, 08 th October and 06 th November 2005), following a sampling protocol modified from ARANHA et al. (1998), in which a 10 m stream stretch was delimited and eight observation points were equally distributed ( Fig. 1). At each point, observations were conducted during 10 minutes (for a total of 240 minutes), and five habitat descriptors were recorded (Tab. I). Besides environmental characterization during observation sessions, all fish species present were recorded along with data on the abundance of each fish group or shoal, the vertical position of each specimen (bottom, middle, surface), the horizontal position of each specimen in relation to the banks (closer to left or right bank or in the middle), and feeding tactics.
To calculate the correlation of the microhabitat descriptors mentioned above and the abundance of K. moenkhausii a Mantel test was carried out, using Euclidean Distance coefficient of similarity for descriptors matrix and Morisita-Horn for the abundance matrix. This analysis was performed with the option of normalization and 5,000 permutations in NTSYSpc software (ROHLF 2000). From the resulting microhabitat similarity triangular matrix, a dendrogram was generated using Unweighted Pair Group Method with Arithmetic mean (UPGMA) cluster method. In this dendrogram, abundance categories for K. moenkhausii were subsequently plotted. Feeding analysis of K. moenkhausii and other nektonic species was performed on fishes sampled in 15 (Appendix 1) of the 95 streams. The 15 streams were chosen because they had an adequate number of specimens for gut content analysis. Gut contents of 10 adult specimens from each stream were examined under stereomicroscope and identified to the lowest recognizable taxon. Percent composition by number, frequency of occur-rence (GELWICK & MATTHEWS 1996) and dominance (HYNES 1950) were calculated for each item. The most important feeding items were determined using BENNEMANN'S et al. (2006) method, an adaptation of COSTELLO'S (1990) method, in which item dominance is employed instead of item weight. Items plotted on the upper right of the diagram are more frequent and dominant and, hence, are assumed to be more important in the diet of the species; by contrast, when several items are located on the lower portion of the diagram, a tendency to trophic opportunism is suggested.
An exploratory ordination technique (NMDS -Non-Metric Multidimensional Scaling Analysis) was employed to evaluate similarities between fish diets using a matrix with food items grouped into the following categories: algae, aquatic insects, crustaceans, fish scales, macrophytes, fishes, nematodes, debris of vascular plants, arachnids, terrestrial insects, eggs, insects fragments, and detritus, the latter corresponding to any highly digested feeding item. Values of percent composition by number were [log (x+1)] transformed and the Bray-Curtis similarity coefficient was calculated using PRIMER v6 software (CLARKE & GORLEY 2006), after specifying 25 random re-starts. Stress values were 0.1, indicating a good representation with no real prospect of a misleading interpretation (CLARKE & WARWICK 2001). In the same software, correlations between habitat descriptors and K. Table I. Environmental descriptors used in mesohabitat, microhabitat, and diet analysis.
Proportion of pools, runs, and riffles 1 Not categorized; evaluated as the percentage of each mesohabitat extension along each stretch.
ZOOLOGIA 27 (2): 191-200, April, 2010 moenkhausii and other species diet were evaluated by a Similarity Analysis (ANOSIM). A diet matrix containing the grouped items mentioned above was compared with another matrix containing environmental descriptors (Tab. I) which probably influence prey availabilities and capture success.

RESULTS
According to CCA results (Fig. 2, Tab. II), the runs proportion and turbidity were the most significant descriptors explaining species-environment relationships (p = 0.042 and 0.020 respectively). The runs proportion was the unique environmental descriptor associated with K. moenkhausii. By contrast, branches and logs proportion, substrate type, and riffles proportion explained the abundances of A. altiparanae and B. stramineus. Hemigrammus marginatus is clearly associated with proportion of pools in the system, being the only species associated with Axis 2 (Fig. 2).
Based on data obtained during snorkeling sessions, specimens of K. moenkhausii were particularly associated with microhabitats 45 cm deep on average; without submerged vegetation; without coverage; and with sandy bottom covered by a fine detritus layer (Tab. III, Fig. 3). Like K. moenkhausii, A. altiparanae and A. fasciatus individuals were also registered in sandy bottoms without marginal vegetation or coverage, though they were less abundant under these conditions. Specimens of K. moenkhausii were registered in 16 of the 24 observation sessions, and co-occurred with A. altiparanae and/or A. fasciatus in eight of them. On 12 occasions, interactions among individuals were observed (only at the points 4, 5, 6 and 7; Fig. 1), either among K. moenkhausii individuals (n = 9) or between K. moenkhausii and A. altiparanae or A. fasciatus individuals (n = 3); of these, 10 interactions involved some kind of dispute or confrontation.
Adults of K. moenkhausii were only occasionally found in areas near the margin; whenever an adult K. moenkhausii moved to a marginal area previously occupied by individuals of A. altiparanae, the latter would immediately relocate to the middle of the channel, and then rapidly return to the margins, in the    GRANT & NOAKES 1987), close to the bottom, on the water surface (i.e. surface picking, see SAZIMA 1986) or close to the margins. Spatial segregation during feeding activities was recorded with K. moenkhausii exploring mainly the middle of the water column (61% of the registers), A. altiparanae foraging mostly closer to the margins (78%) and A. fasciatus foraging closer to the bottom (46%) (Fig. 4).
Percent composition by number revealed predominance of allochthonous items (46% of the total items for all species), dominated by terrestrial insects (34%), with respect to autochthonous items (42%), dominated by aquatic insects (29%). Undetermined items, represented by eggs, detritus and unidentified debris, accounted for 12%. The most important items were those with the highest frequencies and dominances (Fig.  5). Riparian vegetation was the exclusive descriptor that showed significant correlation with fish species diet (Table IV). Based on this result, the three conditions that represent the descriptor physiognomic states (absent, shrubs, or trees) were further plotted into an NMDS ordination diagram (Figs 6-7). Either for the set of nektonic species evaluated together (Fig. 6) or for K. moenkhausii (Fig. 7) alone, the ordination indicated that diet differed according to conditions of the riparian vegetation.

DISCUSSION
On a mesohabitat scale, K. moenkhausii is particularly associated with runs, which are streams or river reaches with slow current, slightly deeper than riffles, with a smaller gradient and with non-turbulent waters (RINCÓN 1999). Most runs investigated in this study have sandy bottoms without branches or logs, or macrophytes, representing habitats with very low complexity, which offer limited microhabitat diversity. Several studies have pointed out that a higher diversity of microhabitats may support a more diversified fish fauna (GREGORY 1992, UIEDA & BARRETO 1999, BÜHRNHEIM & Cox FERNANDES 2003, BRAGA 2004, mainly by offering more shelter possibilities and foraging sites (DELARIVA et al. 1994, CASATTI et al. 2003. By contrast, few species can benefit from homogeneous habitats, contributing to higher species dominance in these communities (SILVA 1992). In fact, K. moenkhausii was one of the most dominant species in streams with low complexity, whereas in more heterogeneous streams, the relative abundance of other nektonic species increased, as observed for A. altiparanae and B. stramineus. These results suggest that K. moenkhausii has great ability to occupy disturbed sites, a fact that may partially explain its noticeable abundance in the streams of the study region.
The contribution of autochthonous and allochthonous items in the diet of the six species studied was similar, with autochthonous items being slightly more represented. According to LOWE-MCCONNELL (1987), allochthonous input is very important in the diet of stream fishes, a hypothesis that has been corroborated by several authors (SABINO & CASTRO 1990, SABINO & ZUANON 1998, CLARO JR et al. 2004, MELO et al. 2004; by contrast, autochthonous items have been found to be more relevant by some studies (COSTA 1987, UIEDA et al. 1997, CASATTI 2002. Even in very shaded streams, autochthonous microalgae productivity can be large enough to support all organisms in upper trophic levels (BRITO et al. 2006).
The terms "generalist" and "opportunistic" have been widely used in stream fish studies particularly to characterize members of the Characidae (ABELHA et al. 2001, FERREIRA et al. 2002, REZENDE & MAZZONI 2003, BENNEMANN et al. 2005, CENEVIVA-BASTOS & CASATTI 2007. These terms are used to classify species that feed on several taxa, have a wide trophic niche, and therefore belong to more than one trophic group (GOLDSTEIN & SIMON 1999). These terms indicate the trophic plasticity, or feeding versatility, displayed by generalist and opportunist species, which can exploit certain alternative resources in the absence of their preferred items (GOLDSTEIN & SIMON 1999, and see examples in ARCIFA & MESCHIATTI 1993, ESTEVES & GALETTI JR 1995, ESTEVES 1996. Trophic plasticity is likely to be an important attribute of K. moenkhausii, since feeding habits can change according to resource availability (CENEVIVA-BASTOS & CASATTI 2007). Within this context, this species is likely to be little affected by habitat degradation.
Based on the discussion above, all species studied here can be considered generalists that consume a great amount of insects, an abundant resource in streams (ANGERMEIER 1985, KIKUCHI & UIEDA 1998, BUENO et al. 2003, COSTA et al. 2006. However, despite the fact that these species feed on similar items, they vary in their feeding tactics and foraging microhabitats. For example, B. stramineus frequently captures insects on the water surface (CASATTI & CASTRO 1998), which explains the predominance of terrestrial insects in the diet of this species. Individuals of H. marginatus are also insectivores, but fed mainly on aquatic insects, which may be explained by their position in the middle of the water column (CASATTI et al. 2003), enhancing their opportunities to capture drifted autochthonous items.
Differences in feeding habits among sites were influenced by the presence of riparian vegetation. It is known that processes that occur in riparian zones influence the aquatic biota (GREGORY 1992, PUSEY & ARTHINGTON 2003 and lead to changes 6 Figures 6-7. Ordination diagram representing two first axis of a Non-Metric Multidimensional Scaling Analysis of the six nektonic species and K. moenkhausii diet (6), and on the last species separately (7). Symbols represent sites according to riparian vegetation: absent (open circles), shrubs (star), and trees (dark circles).

7
ZOOLOGIA 27 (2) : 191-200, April, 2010 in the diet of different fish species (see WAITE & CARPENTER 2000, FERREIRA et al. 2002, reinforcing its importance as a buffer zone (BOJSEN & BARRIGA 2002, SWEENEY et al. 2004, POULET et al. 2005. It is important to mention that the presence of arboreal riparian vegetation, even in poor state of conservation, provided distinct feeding resources to the fish fauna, reinforcing the findings of a previous study carried out by the authors of this paper (CENEVIVA-BASTOS & CASATTI 2007).
Spatial segregation of feeding sites was noticeable among three of the species studied. According to previous studies (UIEDA et al. 1997, ARANHA et al. 1998, CASATTI et al. 2003, differences in spatial distribution and foraging microhabitats suggest partitioning of feeding resources, minimizing the likelihood of competition among species. Competition occurs when a set of species, under natural conditions, has identical niches and resources are in short supply (WIENS 1977, PIANKA 2000. In order to identify an instance of competition, one or more of the species involved must show decreased fitness. Therefore, a dispute that one species wins and another species loses, depending on each species differential competitive abilities and their ability to deal with natural enemies (e.g. MORRIS et al. 2004), is implied.
According to niche and competition theories, the synchronized dislodgement among K. moenkhausii and Astyanax spp., observed here, could be interpreted as a resource partitioning mechanism that prevents competition (SCHOENER 1974).
We believe, however, that rather than being a mechanism to avoid competition, the observed synchronized dislodgements correspond to individual adjustments to an environment that includes individuals of other species (Jorge Jim pers. comm.).
Thus, the spatial segregation observed between the species K. moenkhausii and Astyanax spp. might not be an arrangement to avoid competition, but a dynamic adjustment (J. Jim pers. com) among co-occurring species in relation to habitat (sensu MITCHELL's 2005 habitat concept) conditions.