Introduction

Most of the largest rivers and drainage basins in the world have been dammed (Nilsson et al. 2005; Winemiller et al. 2016). Impoundments, which are lentic bodies of water created by dams, alter the continuity in matter, energy, and nutrient transfer in rivers (Vannote et al. 1980; Ward and Stanford 1983), including modifications in the extent and timing of sediment and nutrient transport, and water discharge (Ward and Stanford 1995a; Friedl and Wüest 2002). Changes in water transparency, nutrient and sediment loads, channel morphology and granulometry are some long-term physico-chemical effects of dams on environments downstream (Ward and Stanford 1995b; Brandt 2000; Manyari and Carvalho Jr. 2007; Wohl et al. 2015), potentially leading to long-term oligotrophication (Ney 1996; Stockner et al. 2000). Changes in hydrology also alter connectivity between floodplains and channels, affecting many ecological processes (Bunn and Arthington 2002; Thomaz et al. 2007). Furthermore, reduction in nutrients typical of reservoir cascades (i.e. a series of impoundments in a river or/and its tributaries) can alter the structure of the primary producer community (Barbosa et al. 1999) and the abundance of consumers, including zooplankton, macroinvertebrates and fish (Sampaio et al. 2002; Hudon et al. 2012; Braghin et al. 2015; Lima et al. 2016; Santos et al. 2016).

In many large rivers, impoundments have changed hydrology by reducing the variability in annual discharge and flood pulses (e.g. Souza Filho et al. 2004; Agostinho et al. 2009), a major driver of ecological processes in associated floodplains (Junk et al. 1989; Junk and Wantzen 2004). Dams have also reduced sediment and nutrient transport leading to reduced turbidity and less dynamic bedload movements (Roberto et al. 2009; Stevaux et al. 2009). The fish community is especially affected by these changes because changing flooding patterns can alter resource availability (Mérona et al. 2005; Luz-Agostinho et al. 2008; Abujanra et al. 2009), reproductive success and recruitment (Agostinho et al. 2004b; Oliveira et al. 2014; Angulo-Valencia et al. 2016), and ultimately fish species composition and abundance.

As an important subset of the fish community, benthic invertivorous fish represent one of the links between primary producers and higher trophic levels in tropical aquatic ecosystems (Hahn et al. 2004; Kramer et al. 2013, 2015). These fish feed on benthic resources that include insect larvae, mollusks and other macro and microinvertebrates (Fugi et al. 2001), which are reportedly affected by changes in sediment granulometry, habitat composition and spatial and temporal heterogeneity (Quinn and Hickey 1990; Takeda and Fujita 2004). Studies have reported the effect of impoundments on detritivores (e.g. Mérona et al. 2001), which are fish that also exploit the bottom, but feed on different resources than invertivorous fish, especially particulates (Hahn et al. 2004).

Considering that the food resources and habitats of invertivorous fish are directly affected by changes caused by river impoundment, we investigated the long-term impact of impoundment on the structure of benthic invertivorous fish assemblages in the floodplain below a major dam. To this end, we examined changes in descriptors of water regime, water quality and assemblage structure of invertivorous fish pre and post-impoundment, in impounded and unimpounded rivers. We expected most descriptors of water regime, water quality and fish assemblages to differ pre and post impoundment in two rivers affected by the impoundment but not in the unimpounded river. Understanding changes in hydrology and biotic communities following river impoundment is not only important to inform assessment of anthropogenic impacts, but also to design conservation strategies to preserve native fish assemblages and floodplain ecosystem services, even in an altered state (Affonso et al. 2015).

Methods

Study area

The Upper Paraná River Basin drains approximately 891,000 km2 in the most populated and industrialized region in Brazil, and it is also the most highly impounded basin in South America. This basin includes a 230-km long floodplain in its southwestern section, between the Engineer Sérgio Motta Dam (which impounded the Porto Primavera Reservoir) and the Itaipu Dam, reaching up to 20 km wide in the western margin of the Paraná River. This floodplain stretch has a high diversity of habitats that includes the alluvial plain with numerous secondary channels, floodplain lakes and sections of the main channels of the Paraná River, Baía River and Ivinhema River (Fig. 1). The Paraná River main channel has three other large impoundments upstream of the study area, and most of its tributaries are heavily dammed (Agostinho et al. 2004c; Stevaux et al. 2009), constituting a reservoir cascade. The Baía River parallels the Paraná River and it is also bisected by Porto Primavera, but no other impoundments exist in its main channel. Unlike in the Paraná River, no water is discharged from Porto Primavera Reservoir into the Baía River, but several small unimpounded tributaries contribute to its flow below the reservoir. Water from the Paraná River spills into the Baía during floods or reaches its lower stretch through reverse flow (Comunello et al. 2003). Conversely, the Ivinhema River has its lower stretches inserted in a state conservation unit, and is one of the few remaining dam-free rivers in the Upper Paraná River Basin. In 2006–2010, the mean discharge of the Paraná River in the study area was 8998 m3/s, and the mean discharge in the Ivinhema River was 363 m3/s. No discharge data are available for the Baía River, but discharge is smaller than the Ivinhema River.

Fig. 1
figure 1

Location of sampling sites in the Ivinhema River (unimpounded), Baía River and Paraná River (both impounded), in relation to Porto Primavera Reservoir. Only the lower gauging station of the Ivinhema River is shown

Full size image

The flood dynamics in the upper Paraná River floodplain are complex. The wet season usually occurs from November to May, but more than one flood pulse can occur during this phase (Agostinho et al. 2009). The flood pulse is critical in terms of maintaining the high biodiversity of the floodplain (Junk et al. 1989; Agostinho et al. 2004c, 2009). The dry season usually occurs from June to October. The annual fluctuations in water level are on average about 2.5 m (Agostinho et al. 2000), but fluctuations as large as 7.5 m have been recorded as well as years with almost no flooding. Overall, the Paraná River exerts the largest influence on the water level in the floodplain; however, local precipitation both in the Baía River and in the Ivinhema River drainage basins also influence floodplain water levels (Souza Filho et al. 2004; Thomaz et al. 2004). Indeed, most floods (40%) occur due to rises in the Paraná River; however, often both the Paraná and Ivinhema rivers cause floods (30%), and the Ivinhema River alone (Comunello et al. 2003) causes some floods (28%).

The floodplain presents a great diversity of habitats. Lentic environments include floodplain lakes; some connect to the river all year, while others are isolated and connected indirectly to the river by groundwater or only during flooding events. Lakes differ in water quality depending on which river they connect to as well as connectivity dynamics (Rocha and Thomaz 2004; Thomaz et al. 2004). The main channels of the Ivinhema, Baía, and Paraná rivers represent lotic environments.

Porto Primavera Reservoir, the largest in the Paraná River and built to support hydropower, navigation and flood control, bisects the Paraná and Baía rivers. The dam was closed in 1998, but some water control had already begun in the early 1990s (BCD 2009). Filling of the reservoir was completed in March 2001. The reservoir has the fifth largest area in Brazil (approximately 2250 km2), extends 240 km upstream of the dam, is 19-m deep at the dam and operates as a run-of-the-river reservoir.

Water regime

Water level data were obtained through the HidroWeb online platform of the National Water Agency (ANA) at one gauging station in the Paraná River (Porto São José Jusante, code 64575003), and two stations in the Ivinhema River (UHE Itaipu Ivinhema, code 64617000, lower station—Fig. 1; Fazenda Ipacaraí, code 64614000, upper station). There are no agency gauging stations in the Baía River. A complete set of daily water levels was available for the Paraná River station from 1974 to 2015. In the Ivinhema River, data were available from 1974 to 2010 at the lower station and from 1984 to 2015 at the upper station. According to year, and according to wet and dry seasons, daily water levels were used to compute eight hydrological descriptors (see definitions in Table 1). Water level was represented by 7-day maximum (7max), 7-day minimum (7min), 30-day maximum (30max), 30-day minimum (30min), median, and seasonal range. The 7- and 30-day maxima and minima were calculated using a moving average procedure (TNC 2009). Median is the seasonal median water level in that year. Seasonal range is the difference between 7-day maximum and 7-day minimum in each season, and reflects the variability of water level within a season. Timing descriptors were represented by seasonal date of maximum level (season day max) and seasonal date of minimum level (season day min). Seasonal dates were computed by transforming Julian days to wet season day (i.e. 1–212, starting on November 1st) and dry season day (i.e. 1–153, starting on June 1st). These descriptors were computed with IHA version 7.1 (Indicators of Hydrological Alteration, Richter et al. 1998).

Table 1 Upper and lower quartiles of water regime descriptors at two gauging stations in the Ivinhema River and one station in the Paraná River
Full size table

Water quality and fish

Data on water quality and fish densities were obtained in lotic and lentic floodplain habitats below Porto Primavera Reservoir pre (1986–1988) and post (2000–2015) impoundment, in the Paraná, Baía and Ivinhema river systems. The pre-impoundment sampling period included monthly samples totaling 24 months, whereas the post-impoundment period included quarterly samples totaling 62 months. Each month, water quality and fish samples were collected within the river channel and in floodplain lakes, except in the Paraná River where there was no sampling in lakes pre impoundment. Driven by fiscal cycles, the number of river channel sites sampled each sampling month over this 18-year effort varied from 3 to 4 (mean 3.2 sites/sampling month), and the number of floodplain lakes varied from 4 to 13 (mean 7.2 sites/sampling month).

We collected fish data at each site and recorded water quality simultaneously at most sites. Clusters of eight gill nets each with different mesh sizes (3, 4, 6, 8, 10, 12, 14 or 16 cm between opposite knots) were fished at each site (each net was 20 m long, 1.45–1.70 m deep). Nets were deployed for 24 h in each site, with fish removed every 8 h. Fish collected were identified to species and counted. Water conductivity and transparency (Secchi depth) were measured onsite, and total nitrogen, total phosphorous, and chlorophyll-a were measured in the laboratory from filtered water samples stored at − 20 °C, following the methodology in Mackereth et al. (1978).

The fish assemblage of the Upper Paraná River includes over 182 fish species (Agostinho et al. 2004a; Graça and Pavanelli 2007) We focused on six benthic invertivorous fish species (Fig. 2): Callichthys callichthys (Linnaeus, 1758), Hoplosternum littorale (Hancock, 1828), Iheringichthys labrosus (Lütken, 1874), Pimelodus maculatus Lacepède, 1803, Satanoperca pappaterra (Heckel, 1840) and Trachydoras paraguayensis (Eigenmann & Ward, 1907). These species comprised approximately 14% of the fish collections in 1986–1988 and 8% in 2000–2015. Most of these invertivorous fish have a specialized diet (Hahn et al. 2004). However, two species tend to be more generalists. Satanoperca pappaterra consumes detritus and plant material as well as benthic invertebrates (Hahn and Cunha 2005; Novakowski et al. 2016), and P. maculatus has benthic organisms as a major component of its diet, but also feeds on small fish and plant matter (Lobon-Cerviá and Bennemann 2000; Lima-Junior and Goitein 2003). Satanoperca pappaterra was probably introduced into the study area by aquaculture or fish farming, in the second half of the twentieth century before we began our study (Langeani et al. 2007; Ortega et al. 2015). Pimelodus maculatus migrates moderately long distances for reproduction and feeding (Makrakis et al. 2012), while I. labrosus and T. paraguaysensis migrate short-distances laterally (Bailly et al. 2008; Holzbach et al. 2009); the other study species are not reported to migrate (Mol 1996; Hostache and Mol 1998; Verba et al. 2011).

Fig. 2
figure 2

Illustration of the six invertivorous fish species included in this study. Sizes are drawn to scale of the mean standard lengths recorded during this study

Full size image

Data analysis

We used a two-way permutation MANOVA (PERMANOVA) to test if water regime indexed by the eight hydrological descriptors changed between periods (i.e. pre and post impoundment) while accounting for possible differences in season (i.e. wet and dry). Temporal correlation in the data does not affect PERMANOVA because randomly permuting the years removes any inherent temporal correlation (Anderson 2001). Each gauging station was analyzed separately. We used all available water years after 1999 to represent post impoundment, while pre impoundment was represented by 1974–1990 water years in the Paraná River and Ivinhema lower station, and 1983–1990 in Ivinhema upper station. The period between 1991 and 1998 was not considered due to possible influences of dam construction. We used a greater number of pre-impoundment years for hydrological analysis than for fish assemblage analysis because more data were available, allowing for a more robust comparison. All hydrological descriptors, except those associated with timing and seasonal range, were log-transformed to reduce skewness, and all variables were normalized (zero mean and unit variance) before computing the resemblance matrix with Euclidean distances.

We assessed changes in water quality and fish using three-way PERMANOVA analyses that tested for differences between time periods in the three rivers, and accounted for any effects of season (i.e. wet and dry) and habitat type (i.e. lentic and lotic). Total nitrogen, total phosphorous, Secchi depth and chlorophyll-a were log-transformed to reduce skewness and all five water quality variables were normalized before the analysis. A resemblance matrix using Euclidean distances was assembled to perform the water quality PERMANOVA.

For fish, we first conducted a permutational ANOVA, with the same factors described above, for total abundance (total catch rate) of all invertivorous fish combined. Next, we performed a PERMANOVA to test for shifts in the six species assemblage. Analyses were applied to catches standardized to 1000 m2 of the gill net cluster by season, water year, river, and habitat type. All catch rates were log-transformed to reduce skewness and the resemblance matrix computed using the Bray-Curtis dissimilarity metric. In the PERMANOVA, a dummy variable was used to include samples containing no invertivorous fish species (11 out of 218 samples) (Clarke et al. 2006) and we used type III sum of squares to account for the unbalanced design. If the PERMANOVA identified a significant (p ≤ 0.05) period × river effect, we performed Principal Coordinate Analysis (PCoA) to visualize differences, using the same resemblance matrices of the respective PERMANOVA. All analyses were performed in the software PRIMER 6.0 with the add-on PERMANOVA, and we report pseudo-F (F) values for the main test and pseudo-t (t) values for pairwise comparisons; 999 unrestricted permutations of raw data were used in the main test as well as in the post-hoc tests for factors and interactions.

Results

Water regime

Water regime changed in the Paraná River pre and post impoundment (F = 5.8, p < 0.01), but not in the upper (F = 0.6, p = 0.59) or lower (F = 2.1, p = 0.08) Ivinhema River. These three analyses also showed significant effects of season (p ≤ 0.01) but no significant (p ≥ 0.41) period × season interactions, suggesting that seasonality did not change post impoundment. Every description of water level decreased in the Paraná River after dam construction (Table 1); however, the timing of seasonal maxima and minima did not change (Fig. 3a). The greatest decreases were observed with median water levels, maximum values of wet season and seasonal range of the wet season (Table 1). However, high overlap of data (Fig. 3a) and dispersion of data (Fig. 3b, c) suggest that discharge varies greatly among years in the same period. Although reductions in water level are apparent, water regimes post dam may resemble those pre dam in some years.

Fig. 3
figure 3

a Ordination plot of water regime descriptors in the Paraná River pre impoundment (1974–1990) and post impoundment (1999–2015) in the wet and dry seasons. Vectors represent correlations of water regime descriptors with PCoA axis 1 and 2. b PCoA 1 axis values for samples pre and post impoundment in wet and dry seasons. c PCoA 2 axis values for samples pre and post impoundment in wet and dry seasons. In b, c squares/circles represent means and whiskers represent 95% confidence intervals

Full size image

Water quality

Water quality changed with period (F = 3.7, p = 0.01), river (F = 12.9, p < 0.01) and habitat type (F = 12.3, p < 0.01), but not season (F = 2.5, p = 0.07). The river × habitat type interaction was statistically significant (F = 5.5, p < 0.01) indicating that the magnitude of the difference between lentic and lotic environments differed in all three rivers (p < 0.01 for all combinations). The river × period interaction was not statistically significant (F = 1.6, p = 0.14) indicating overall differences between periods were similar among rivers. Secchi depth increased in the Paraná River post impoundment but not in the Baía and Ivinhema rivers (Table 2). Conductivity increased and total phosphorous decreased in all three rivers, while chlorophyll-a and total nitrogen did not change.

Table 2 Upper and lower quartiles of water quality variables and invertivorous fish species catch rates (fish/1,000 m2), pre and post dam construction in the three river systems (Cond conductivity, Chloro Chlorophyll-a)
Full size table

Invertivorous fish assemblage

In all, we collected 12,878 invertivorous fish. The average total catch rate of invertivorous fish differed between periods, and type of habitat (Table 3), while the effects of river and season were not statistically significant. The interaction terms period × river and river × habitat (Table 3) were statistically significant. Post-hoc comparisons showed that average catch rate of all invertivorous fish combined did not change pre and post dam in the Ivinhema River (t = 0.1, p = 0.90) nor in the Baía River (t = 1.6, p = 0.10) but it decreased from 57.5 to 16.5 fish/1000 m2 of gillnet in the Paraná River (t = 4.1, p < 0.01).

Table 3 Results of permutational analysis of variance for total catch rate of benthic invertivorous fish and multivariate, benthic invertivorous fish assemblages
Full size table

Considering both periods, T. paraguayensis represented 31.3% of the total invertivorous fish catch, H. littorale represented 23.3%, I. labrosus 20.7%, P. maculatus 18.3%, S. pappaterra 5.9% and C. callichthys 0.5%. Invertivorous fish assemblages differed between periods, rivers (Fig. 4a), and habitat types but not between seasons (Table 3). The statistically significant interaction period × river indicated that the invertivorous fish assemblage changed differently across rivers after dam construction. Post-hoc multiple comparisons showed that assemblage differed pre and post impoundment in the Paraná (t = 4.9, p < 0.01, Fig. 4d) and Baía rivers (t = 1.9, p = 0.02, Fig. 4c) but not in the Ivinhema River (t = 1.4, p = 0.10, Fig. 4b). The river × habitat type interaction was statistically significant, suggesting that the difference between lentic and lotic environments differed among rivers. The most marked changes occurred for T. paraguayensis, which decreased drastically in the Baía and Paraná rivers, and S. pappaterra, which increased in the Paraná River (Table 2). Iheringichthys labrosus and P. maculatus also decreased in the impounded rivers (Fig. 4c, d; Table 2). The lotic environment in the Paraná River post dam had the smallest catch rate for five of the six species (Table 2).

Fig. 4
figure 4

Ordination plots of the invertivorous fish assemblage a in all three river systems, not separated according to pre and post periods; b in the unimpounded Ivinhema River separated by period; c in the impounded Baía River separated by period; and d in the impounded Paraná River separated by period. Vectors represent Pearson correlations of fish species densities with PCoA axis 1 and 2, and were moved from the center of the graph to improve visualization of text

Full size image

Discussion

We identified changes to the water regime and invertivorous fish assemblages below the Porto Primavera Reservoir. Changes were evident in the Paraná River, which is distinctly impacted by this impoundment, less obvious in the Baía River which is only partially influenced by the dam, but largely indistinguishable in the unimpounded Ivinhema River. Changes in water regime were reflected by variables descriptive of water-level magnitude, whereas variables descriptive of water-level timing showed little modification. Changes in fish assemblages were reflected by a decrease in catch rate of invertivorous fish in the Paraná River, and a shift in species relative abundances in the Baía and Paraná rivers, but not in the unimpounded Ivinhema River.

Changes in water quality were more ambiguous. A distinct change to water quality directly attributable to impoundment was water transparency, which increased in the Paraná River post dam (also reported by Roberto et al. 2009) but did not change in the Baía and Ivinhema rivers. The other water quality metrics did not change at all or also changed in the unimpounded river. It is unclear whether changes to the unimpounded river were induced by long-term land use changes in the Ivinhema River catchment (Arai et al. 2012) or simply by sampling error attributable to the high variability associated with sampling rather ephemeral water quality metrics. We suggest that changes in water regime and water transparency after the impoundment of Porto Primavera directly or indirectly affected invertivorous fish assemblages in the impounded rivers and have consequences through the entire aquatic ecosystem.

Water regime in the Paraná River changed after the impoundment of Porto Primavera Reservoir, shifting towards lower 7- and 30-day maxima and lower 7- and 30-day minima. Souza Filho et al. (2004), Souza Filho (2009), and Rocha (2010) have reported additional changes through examination of other water metrics, including decreased flood duration and modified discharge-water level relationships. The observed reductions in water levels can lead to loss of aquatic habitat through various mechanisms including (1) decreased connectivity among diverse floodplain habitats (Ward and Stanford 1995b; Agostinho et al. 2004c; Thomaz et al. 2007); and (2) loss of habitat volume through shallower depths (Brandt 2000) and shrinking lateral span. Such changes can cause long-term shifts in invertivorous fish assemblages as well as the entire fish community (Agostinho et al. 2001, 2008).

Differences in the pre and post impoundment structure of benthic invertivorous fish assemblages in the impounded and unimpounded rivers suggest an effect attributable to the change in water regime. The species whose catch rate decreased (namely T. paraguayensis, I. labrosus and P. maculatus) undertake either small or moderate reproductive migrations cued by flooding, stage lateral migrations to feed, and their larvae and juveniles depend on increasing water levels to access nursery areas in backwaters and lakes (Vazzoler 1996; Agostinho et al. 2000, 2001, 2004a; Bailly et al. 2008). Conversely, the only species that increased, particularly in the Paraná River, S. pappaterra, is an equilibrium strategist (i.e. low fecundity, brood protection, non-seasonal spawning; Winemiller 1989; Vazzoler 1996; Verba et al. 2011), thus is virtually independent from flood patterns to reproduce and possibly benefits from flood absence and dry years (Agostinho et al. 2004a, b).

Water quality changes below dams are linked to water storage upstream, discharge regime, and their effects on sediment and nutrient loads (Brandt 2000; Wohl et al. 2015; Agostinho et al. 2016). Reservoir cascades, such as the one in the Paraná River, are common in many large river systems and can have cumulative and synergic effects on the trophic status of downstream reaches (Barbosa et al. 1999; Miranda et al. 2008), because these cascades exacerbate the sediment and nutrient trapping capacities of single reservoirs (Barbosa et al. 1999; Kummu and Varis 2007). The increase in water transparency caused by sediment trapping can increase the abundance of submerged macrophytes (Sousa et al. 2010), which in turn may maintain the clear water state of the environments in the absence of benthic fish (Mormul et al. 2012). Clear water also increases predation pressure on eggs, juveniles, and small-size species (Agostinho et al. 2004a; Turesson and Brönmark 2007; Dodrill et al. 2016). In fact, larval densities have decreased in our study area post closure of Porto Primavera, including those of some invertivorous fish (Sanches et al. 2006), thus decreasing recruitment success and population sizes, especially of species that do not provide parental care (such as T. paraguayensis, I. labrosus and P. maculatus, the invertivorous fish that decreased in our study). The two invertivorous species whose densities were apparently not affected by construction of the Porto Primavera Reservoir (H. littorale and C. callichtys) also depend on flooding to reproduce, but take care of their offspring in nests (Mol 1996; Hostache and Mol 1998), possibly offsetting the effects of increased predation induced by water transparency on reproduction success.

Besides the observed changes in water regime and water quality, damming reportedly causes modifications in channel substrate downstream (Brandt 2000), such as the coarsening of bed material, decreased sediment heterogeneity, and increased bank erosion (Stevaux et al. 2009). These changes directly affect the benthic macroinvertebrate community (Stevaux and Takeda 2002; Bredenhand and Samways 2009), whose composition and abundance are influenced by sediment type and grain-size (Quinn and Hickey 1990; Takeda and Fujita 2004). In turn, these substrate changes affect invertivorous fish populations through food availability and diversity (e.g. Paragamian 2002). Nevertheless, invertivorous fish assemblage structure, as well as that of other fish guilds, is probably indirectly influenced by factors and interactions that we were unable to assess in this study. For example, substrates can be altered by (1) degradation of the littoral zone caused by increased daily variability in water level required by energy production demands (Souza Filho 2009; Poff and Schmidt 2016); and (2) alterations in riparian vegetation caused by longer dry periods and subsequent changes in land use (e.g. expansion of cattle grazing and agriculture in the floodplain). These changes rework the input of allochthonous resources into substrates (Agostinho and Zalewski 1995; Thomaz et al. 2004), changing bottom up regulation of populations and aquatic community production.

Freshwater ecosystems provide many services including biomass production and nutrient cycling (Dobson et al. 2006), which are directly related to energy flow. The relative abundance of organisms in various trophic levels is an important factor in energy flow through the ecosystem (Depczynski et al. 2007). For example, decreased abundance of intermediate predators (such as invertivorous fish) could shorten food chain length (Diehl and Feissel 2001; Post 2002; Hoeinghaus et al. 2008). In fact, the Paraná River was found to have a shorter food chain when compared to the Ivinhema River in our study area (Hoeinghaus et al. 2008), which was attributed to changes in relative importance of different trophic pathways (i.e. based on plant detritus, phytoplankton or periphyton) in river ecosystems after impoundment (Hoeinghaus et al. 2007, 2008). Because invertivorous fish are directly linked to both the detritus and periphyton pathways (Hahn et al. 2004; Lopes et al. 2009), changing their assemblage structure or representation in the fish community could decrease the diversity of mechanisms for energy flow in the ecosystem. These ecosystem changes are expected to influence other fish guilds as well (e.g. Mérona et al. 2005), through either direct impacts on populations or changing biological interactions. For example, in the Upper Paraná River floodplain, detritus processing might have declined due to diminishing populations of the detrivore Prochilodus lineatus caused by decreased access to backwaters (Gubiani et al. 2007), and diet shifts of piscivores have been reported following changing prey assemblages, including decreased consumption of T. paraguayensis (Luz-Agostinho et al. 2008).

The damming of the Upper Paraná River Basin is a big-scale disturbance that has produced big changes in hydrology and fish assemblages, and may require big solutions if fish assemblages are to be preserved until the societal services provided by dams are no longer needed. Various climate change models and scenarios predict moderate increases in precipitation and runoff in the Paraná Basin (Nohara et al. 2006; Marengo et al. 2009, 2012), but likely insufficient to ease the impacts of impoundments on the water regime. Thus, to preserve native fish assemblages further intervention may be needed. A number of big-scale management actions may be applied separately or jointly to address the conservation objective, although practicality and effectiveness may vary. First, keeping existing unimpounded tributaries free of dams can prevent further changes in the natural water regime and biotic assemblages (Affonso et al. 2015; Braghin et al. 2015). Second, engineering artificial floods (e.g. Acreman et al. 2000; Patten et al. 2001) by coordinated water releases from reservoirs throughout the upper Paraná Basin may preserve connectivity in the floodplain (Agostinho et al. 2009; Oliveira et al. 2014). This action may not be feasible if discharge volumes required to engineer sufficiently large and lengthy floods represent a sizeable fraction of the water impounded in the basin, even if the floods are not created annually. Nevertheless, this is an alternative worth exploring with modelling tools (e.g. Rauschenbach 2016) and in collaboration with water resource engineers. Third, intensive management of fish habitat within the floodplain (e.g. Buijse et al. 2002; Gregory et al. 2008; Ollero 2010) to maintain flow and suspended sediment, to retain lateral connectivity between channels and lakes, and to continue access to other habitat needs, even at the reduced water levels dictated by the new water regime, may preserve functionality of the floodplain. Coordinated action among public and private sectors is essential to successfully implement these and other conservation activities.