Spatial variations in fish assemblage structure in a southeastern Brazilian reservoir

We assessed the fish assemblage structure and composition of Nova Ponte Reservoir (Araguari River, Upper Paraná Basin, Brazil). We observed significant differences in abundance (p = 0.0003), richness (p = 0.0005) and diversity (p = 0.02) between lacustrine and riverine zones of the reservoir. Nine species were significantly more abundant in the riverine region: Astyanax altiparanae, Astyanax gr. fasciatus, Galeocharax knerii, Hoplias intermedius, Hypostomus sp., Leporinus friderici, Leporinus obtusidens, Pimelodus maculatus and Schizodon nasutus. The results indicated a longitudinal gradient in the composition and abundance of fishes in Nova Ponte Reservoir, reinforcing the importance of freely flowing riverine reaches for conserving native neotropical ichthyofauna and reflecting the strong adaptation of these species to riverine systems.


Introduction
In recent years, the construction of large dams, mainly for hydroelectric generation, has become one of the most frequent anthropogenic interventions in Brazilian rivers.Because of changes in flow regimes, sediment and nutrient transport, energy flow, and biota, a dam interrupts and changes important ecological processes (Benedito-Cecilio and Agostinho, 2000;Poff et al., 2007).The fish assemblages present in reservoirs are a result of the restructuring of those assemblages that occupied the river segment before damming (Araújo-Lima et al., 1995).This process is marked by the extirpation of some species, reduced abundances of many species (Gehrke and Harris, 2001;Gomes and Miranda, 2001;Gehrke et al., 2002), and proliferation of other species (Pelicice and Agostinho, 2009).It is expected that species with greater plasticity in their feeding and spawning requirements are among the dominant ones (Gehrke and Harris, 2001).Population depletions occur mainly among species that are migratory, long-lived, and have low reproductive potential (Agostinho and Gomes, 2006).
The longitudinal compartmentalization of reservoirs divides them into well-characterized zones, where the water mass is often divided into lacustrine (LZ), transitional, and riverine (RZ) zones (Thornton, 1990;Terra and Araujo, 2011).The RZ is characterized by greater water flow, shallower depths (Thornton, 1990), and more nutrient availability (Oliveira et al., 2003).Some authors emphasize that reservoir tributaries and riverine regions act as refuges for native species (Oliveira et al., 2005;Gao et al., 2010).This fact is reinforced by the tendency of native species to colonize sites with riverine features (Agostinho et al., 2007) and by the greater diversity and proportion of rare species found in tributaries and RZ (Oliveira et al., 2003).
Several studies have recognized the importance of remote sensing in identifying elements in the water column such as suspended sediment and chlorophyll (Harrington Junior et al., 1992;Hedger et al., 2002;Assireu et al., 2011).In reservoirs, those elements are good indicators of sediment and nutrient dynamics.In the RZ, the trend is for a high load of suspended sediment and photosynthetic algae (Wetzel, 2001).In this sense, data from satellite images combined with in situ data are very useful for assessing compartmentalization in seasonally variable tropical reservoirs (Assireu et al., 2011).
We assessed fish assemblage features and ecological parameters in a Brazilian reservoir to verify spatial patterns related to compartmentalization (division into LZ and RZ).Correlations between turbidity and pheophytin and biological parameters were examined to assess the influences of key limnological variables on fish assemblages.We hypothesized that there would be a nonrandom pattern in fish abundance and assemblage structure and composition between lacustrine and riverine zones.

Study area and site selection
Prior to artificial fragmentation by a series of dams along the hydrographic system, the Araguari River was an important migration route for native fish.Presently, the Araguari River main channel is modified by four large (installed capacity: 210 to 510 MW) and two small (installed capacity: 6.2 and 23 MW) hydropower plants.Its largest tributary, the Quebra-Anzol River, still features a freely flowing segment that is about 70 km long (Vono, 2002), which is one of the few remaining segments for recruitment of fish larvae and juveniles in this region of the basin.The Nova Ponte Hydroelectric Plant is located in the middle section of the Araguari River (Figure 1) and began operation in 1994.Most of Nova Ponte Reservoir (19° 07' 41.6" S, 47° 41' 22.8" W) is formed by the Quebra-Anzol River (Vono, 2002).The reservoir is 120 m deep near the dam, 115 km long, has a volume 12.8 billion m 3 , and is oligotrophic and stratified during most of the year (Vono, 2002).
We selected sampling sites by dividing the reservoir perimeter into 40 equal sections from a random first point defined according to the concept of a spatially balanced sample network (Stevens and Olsen, 2004) adapted for large reservoirs (Macedo et al., 2012).Each site was located at the beginning of each section and each site was 200 m long.We sampled the sites at the end of the rainy season in 2010.
We divided the reservoir into two zones in two steps.First, we used classified Landsat 5 TM images taken during our field sampling period (April 2010) to distinguish RZ and LZ visually.After acquisition and image geo-referencing, we performed atmospheric correction by the dark object subtraction method (Chavez Junior, 1988).We used the green (0.52 to 0.60 mm), red (from 0.63 to 0.69 mm), and near infrared spectral bands (0.76 to 0.90 mm), which aid proper detection of algae and suspended sediment concentrations (Jensen, 2006).We classified reservoir zones through use of the maximum likelihood method, using areas of the reservoir close to the dam (to characterize the LZ) and major tributary mouths (to characterize the RZ) as training areas.We expected that the RZ would have higher concentrations of suspended solids (Wetzel, 2001) and consequently higher turbidity than the LZ.Although turbidity was higher in the RZ of Nova Ponte reservoir, the decrease in light penetration did not preclude primary production.In addition, we expected abundant algae in this zone because light was sufficient and nutrients were abundant compared to the LZ (Wetzel, 2001).In our second step to classify reservoir zones, we used water samples to test whether there were significant differences in limnological parameters between those previously classified zones.We used in situ measurements of suspended solids (measured by turbidity) and algae (measured by pheophytin) concentrations.After finding the non-normality of these parameters using the Kolmogorov-Smirnov test, we employed a Mann-Whitney test corrected by Bonferroni criteria (Zar, 2009) to test differences between zones related to these parameters.At a smaller spatial scale, we also classified sites as tributary or reservoir.Tributary sites were located at or very near the mouths of small or large tributaries, and reservoir sites were distant from tributaries.This secondary site classification yielded tributary sites in the LZ and reservoir sites in the RZ (Figure 1).

Fish sampling
We collected fish through use of gill nets with mesh sizes between 3 and 16 cm (distance between opposite knots), heights ranging between 1.6 and 1.8 meters, and lengths of 20 meters.At each sampling site, we placed a set of five pairs of nets comprising the following mesh sizes: 3-8, 4-10, 5-12, 6-14 and 7-16.Each pair was placed at a distance of 40 m from the other at an angle of approximately 45 degrees from the reservoir shore.The nets were set in the evening and removed the following morning, with a total exposure time of approximately 15 h.The specimens collected were fixed in 10% formalin and identified in the laboratory through use of a taxonomic key (Graça and Pavanelli, 2007).We considered as non-native species that did not originally belong to this part of the Upper Paraná Basin (according to Graça and Pavanelli, 2007).Individuals difficult to identify and voucher specimens were deposited in the ichthyological collection of the Universidade Estadual Paulista, São José do Rio Preto (São Paulo).

Data analyses
Because sampling effort was constant, we used the total number of collected specimens to compare sampling sites.We used the total number of species as an indicator of taxonomic richness, and estimated Shannon diversity (H') and evenness (Heip index) as proposed by Magurran (2011).
To assess differences between the LZ and RZ, we applied the Mann-Whitney test (Zar, 2009) on the medians of 1) the abundance, richness, diversity, and evenness and 2) the abundance of selected species (most abundant, migratory, rheophilic) in each site.We also calculated Spearman correlations between assemblage variables and turbidity and pheophytin to assess the biological effects of single environmental variables.Similarly, to verify the influence of tributaries on ecological parameters, we also applied the Mann-Whitney test (Zar, 2009) to assess differences in the medians of abundance, total richness, diversity, and evenness between tributary versus reservoir sites.
We also assessed the difference in assemblage composition between zones, through use of nonmetric multidimensional scaling (NMDS).We performed NMDS on a Bray-Curtis similarity matrix built from species relative abundance data transformed by log (x +1) after excluding species that occurred at fewer than 5% of the points (Legendre and Legendre, 1998;McCune et al., 2002).To validate the analysis, only stress values near 0.2 (Legendre and Legendre, 1998) were admitted.To test the significance of groups, we performed an ANOVA similarity (one way ANOSIM) test (Clarke and Warwick, 2001).
To evaluate the influence of site location (LZ or RZ) on abundance, we applied a Mantel test (Mantel, 1967).We compared two matrices: a similarity matrix, constructed from the captures from each site using the Bray-Curtis similarity index and a dissimilarity matrix, which had zeros for pairs of stations located in the same zone and 1 for pairs of sites located in distinct zones of the reservoir (Oliveira et al., 2004).
Abundance and richness were significantly higher at the tributary sites than at sites in the main body of the reservoir; however, there was no significant difference in diversity or evenness (Table 1).Spearman correlations between turbidity and richness (0.63), diversity (0.52) and abundance (0.55) were significant, but not between evenness and turbidity (0.02).Correlations between pheophytin and assemblage variables were not significant: richness (0.26), diversity (0.22), abundance (0.25), and evenness (0.003).
Despite the significant difference in abundance and species composition between zones, there was overlap between them as expected (ANOSIM: R = 0.27, p = 0.0003).There was a positive correlation between the similarity matrix constructed from the abundance data and the dissimilarity matrix based on site locations in different zones (Mantel: R = 0.18, p = 0.0004).This indicates non-random patterns of capture and the existence of a longitudinal gradient in abundance and species composition.

Discussion
The dominance of so few species in Nova Ponte Reservoir seemed remarkable; however, this has been observed in other reservoirs, which are new environments and have unstable assemblages, often greatly simplified  Table 1.Results of the Mann-Whitney U-test (values in bold are significant), for differences in abundance, richness, evenness, and diversity (Shannon) between tributary (TR) MB) sites and between lacustrine (LA) and riverine (RI) zones (and main body of the reservoir (see Figure 1 for site and zone locations).(Oliveira and Goulart, 2000).The dominance of C. piquiti and P. nattereri at most sites and their distribution in both reservoir zones demonstrate their relatively rapid colonization success in Nova Ponte Reservoir.Both species were also found downstream in Itumbiara (Vono, 2002) and Capim Branco (Rêgo, 2008) Reservoirs.The presence of alien species is widely reported in other reservoirs located in the Upper Paraná Basin (Gomes and Miranda, 2001;Alves et al., 2007;Carvalho and Silva, 2007).Introductions and establishment of Cichla species were also mentioned (Oliveira et al., 2003;Luiz, 2010) as was the suppression of the local fish fauna (Pelicice and Agostinho, 2009).The establishment of alien fish species is a key biological mechanism for transforming native fish assemblages (Leprieur et al., 2008;Hughes and Herlihy, 2012).These changes are even more drastic in reservoirs, which are biologically simplified environments (Luiz, 2010).

Ranks
The nonrandom pattern in total abundance and the greater abundance of some species in sites that were distant from the dam are consistent with the expectation that the dam alters several characteristics of fish assemblages along the river-reservoir axis (Prchalová et al., 2009;Terra et al., 2010).Irz et al. (2002) suggested that the processes that determine the fish distribution in reservoirs are a combination of what occurs in lakes and rivers, partly influenced by river continuity, and partly the result of the confinement of species in reservoirs.
Except for H. intermedius, all species that had significantly different catches between reservoir zones, occurred in smaller numbers in the LZ.According to Agostinho et al. (2008), environmental conditions that occur in the LZ impose a severe environmental filter.Therefore, it is expected that species adapted to lacustrine environments may colonize and become dominant, while others are locally eliminated (Regier et al., 2013).Rheophilic and migratory species, for example, tend to strongly decrease over time after the construction of dams (e.g.Oberdorff and Hughes 1992;Hughes et al., 2005).
Lotic characteristics determine the distribution of native species in reservoirs (Irz et al., 2002).In the RZ, the presence of conditions more similar to the original environment can explain the occurrence of a greater number of species, particularly those with specific requirements for migration and spawning such as L. obtusidens.Migratory species generally avoid lacustrine conditions and concentrate in riverine and transitional zones (Agostinho et al., 2007).Several authors also emphasized the importance of tributaries for the maintenance of remaining native fish populations in reservoirs (e.g.Agostinho et al., 2004;Oliveira et al., 2004).Greater abundances of fish at the confluence of tributaries can also be related to the tendency of many fish species to colonize sites with fluvial features (Agostinho et al., 2007).
The positive correlation found between turbidity and biological parameters (richness, diversity, abundance) indicates that many species select reservoir habitats most resembling the original riverine environment.Species associated with riverine systems usually require specific environmental characteristics to complete their life histories, particularly the reproduction, rearing, and refuge components, which vary seasonally with the hydrological regime, temperature, and turbidity (Agostinho et al., 2008).
The lacustrine regions of reservoirs severely restrict fish occupation because of thermal and chemical stratification, greater depth, and fewer microhabitats (Oliveira et al., 2003;Agostinho et al., 2004).Other features of reservoirs may limit species richness and influence the structure of assemblages in these environments, among them turbidity (Gido et al., 2009).Nearer to the dam the water mass has greater residence time and lower dissolved nutrient concentrations (Agostinho et al., 2008).Some of these features exist in the LZ of Nova Ponte Reservoir, which is oligotrophic and thermally stratified, has steep slopes, and great depth (120 m at the dam; Vono, 2002).Some authors stated that great depth is a key factor determining the spatial distribution of species (Irz et al., 2002) and Prchalová et al. (2008) reported greater abundance, biomass, and richness in shallower reservoir locations.
In the Upper Paraná Basin, the lack of species adapted to lacustrine environments could also explain the low colonization success of the LZ by most fish species (Agostinho et al., 2008;Terra et al., 2010).Likewise, a large portion of the fish fauna of the Araguari River Basin is represented by migratory or rheophilic species (Vono, 2002), which depend on riverine environments to complete their life histories.
In conclusion, our hypothesis of a nonrandom pattern in fish abundance and assemblage structure and composition was corroborated.The correlation between turbidity and richness, abundance, and diversity was also confirmed.These results reinforce the importance of riverine river segments for preserving the native Neotropical ichthyofauna and reflect the strong adaptation of these species to riverine systems.

Figure 1 .
Figure 1.Nova Ponte Reservoir lacustrine and riverine zones and locations of tributary and reservoir sites.

Table 2 .
Results of the Mann-Whitney U-test (values in bold are significant) for differences in abundances of frequently captured species between the lacustrine and riverine zones (see Figure1for zone locations).Migratory species. 2 Rheophilic species. 3Non-native species. 4Most abundant species.