Migratory fishes from rivers to reservoirs: seasonal and longitudinal perspectives

ABSTRACT Migratory fishes have high ecological, social, and commercial value, and are strongly affected by river regulation. This study aimed to understand how migratory fishes use the longitudinal gradient, in an upstream to downstream direction of two free-flowing rivers and two reservoirs in a cascade within the Upper Grande River Basin, Brazil. The numeric abundance, biomass, richness, presence of fingerlings and juveniles, and the macroscopic gonadal maturation stage of migratory fishes were ascertained and evaluated bimonthly for two years. Recruitment in migratory fishes seems to rely completely on the free-flowing rivers upstream and their floodplains. Transition areas do not seem to have a significant role in recruitment. Therefore, we highlight the importance of maintaining the free flow of rivers and the integrity of their floodplains.


INTRODUCTION
Migratory fishes are ecologically, socially and commer cially important (Harvey and Carolsfeld 2003, Agostinho et al. 2007a, 2016, Pelicice and Agostinho 2008, Pelicice et al. 2018).Most species travel great distances during their life cycle.Their reproductive process, early development and recruitment depend on the natural flow regimen of the rivers they travel in (Gomes and Miranda 2001, Pelicice and Agostinho 2008, Pompeu et al. 2012, Agostinho et al. 2016, Winemiller et al. 2016).
The general reproductive migration pattern of Neo tropical freshwater fish includes an upstream displacement in rivers and their tributaries in the beginning of the wet season, followed by spawning.Spawning is triggered by a specific flow events and conditions such as water turbidity and temperature (Harvey and Carolsfeld 2003, Agostinho et al. 2007a, Pelicice and Agostinho 2008, Pompeu et al. 2012, Lopes et al. 2018).After spawning, the eggs develop and hatch as they are passively transported by the river currents to lateral depressions (floodplain lakes), which flood when the water overflows; although floodplains are considered the main nursery areas for fingerlings, they can also develop in the riverbed (Harvey and Carolsfeld 2003, Agostinho et al. 2007a, 2007b, Pelicice and Agostinho 2008, Pelicice et al. 2015, Lopes et al. 2019).The adult parents return to feeding areas where there are more resources and juvenile fishes are eventually recruited (Lucas and Baras 2001, Pompeu et al. 2012, Lopes et al. 2019).
Although the pattern described above has been observed in many river systems, Neotropical migratory fishes may use the habitat differently during reproduction, especially in their early development, depending on the characteristics of the river basin.In rivers without floodplain lakes and similar environments, transient environments between tributaries and the main river channel sometimes provide sufficient conditions and are used as nurseries (Zani boniFilho andSchulz 2003, Silva et al. 2020).In the case of rivers that are regulated by dams, the freeflowing stretches upstream of the reservoir determine the maintenance of the diversity of rheophilic fish, since they may allow migratory species to complete their life cycle (Marques et al. 2018, Car vajalQuintero et al. 2019, Lopes et al. 2019).In these cases, the riverreservoir transition may function as a recruitment area for migratory species.This transitional area provides conditions that are similar to those found in floodplain la goons, for instance less water turbidity and lentic areas with lateral vegetation, providing refuges and food resources for fingerlings and juveniles (ZaniboniFilho and Schulz 2003).
Some studies have documented the presence of eggs and larvae in the upper portion of reservoirs (Suzuki et al. 2011, Souza 2013), after which survival and recruitment have not been confirmed.The presence of fingerlings and juveniles in the riverreservoir transition are a sign that migratory fishes can complete their life cycles in the upper areas of reservoirs that have freeflowing stretches upstream.This information is crucial, especially in reservoir cascades, because these freeflowing stretches can intensify the impact on migratory species (Santos et al. 2017, Loures and Pompeu 2018, Santos et al. 2018).Reservoir cascades tend to decrease the numbers of migratory fishes towards downstream res ervoirs (Santos et al. 2017), while their richness tends to be higher in reservoirs that present lotic stretches upstream of the impounded area (Loures and Pompeu 2018).
This study aimed to evaluate how migratory fishes use the longitudinal gradient, in an upstream to downstream direction, along two freeflowing rivers and two reservoirs, in a cascade located in the Upper Grande River Basin, state of Minas Gerais, Brazil.We tested three hypotheses: 1) due to their rheophilic behavior and movements along critical habitats, the numbers, biomass, and richness of migratory fishes will decrease from the freeflowing rivers towards the reservoirs; 2) the riverreservoir transition and the floodplain lagoons along the freeflowing stretches of the Aiuruoca and Grande rivers allow the early development of migratory spe cies; and 3) migratory fishes, in mature and postspawning gonadal maturation stages, will be found predominantly in freeflowing rivers during the wet season.

Study area
This study was conducted in the Grande River, Upper Paraná River Basin.The Grande River has 12 hydropower dams.In its upper course, stretches of free flowing rivers are still found (Suzuki et al. 2011, Borges andAbjaudi 2016).
The longest ones, the Aiuruoca and Grande rivers, flow to the first reservoir, created by the Camargos Hydropower Plant.It started operating in 1960, with a power generation capacity of 45 MW and a reservoir with an accumulated area of 73.35 km².The Itutinga Reservoir, built in 1955, is located immediately downstream of the Camargos Dam and presents an installed capacity of 52 MW and a reservoir area of 1.72 km 2 (Cachapuz 2006).Both plants coordinate oper ations (Cachapuz 2006), with a 14day water residence time (Suzuki et al. 2011).The climate of the study area is classified as semihumid and mesothermal, presenting annual average temperatures ranging from 18 to 29 °C and annual mean precipitation from 1,450 to 1,600 mm, with four to five dry months (May to September) (Borges and Abjaudi 2016).
The Camargos Reservoir is a storage type and the water levels there can fluctuate up to 14 m, depending on the rain fall volume.The Itutinga Reservoir, which is a runofriver plant, is more stable in terms of water level fluctuations.

Sampling
We sampled 12 sites, separated by approximately 10 km from each other, distributed throughout the Aiuruoca and Grande rivers and Camargos and Itutinga reservoirs.Also, four marginal lagoons were sampled in Aiuruoca and four in Grande River.In Itutinga, there are no significant tributaries flowing to the reservoir (Fig. 1).Conversely, the Aiuruoca and Grande rivers merge to become the Camargos Reservoir.
We sampled fish every two months between March 2019 and December 2021 in Riv3,Riv4,Trans1,Trans2,Trans3,Cam1,Cam2,Cam3,Itu1 and Itu2,conducting over 12 surveys,six in the dry and six in the wet season.After the first year of sampling, we added two new sites upstream, to maximize the chance of sampling migratory fishes.There fore, we sampled Riv1 and Riv2 bimonthly with the other sites between September 2020 and December 2021.We captured fish using gill nets with different mesh sizes 3, 4, 5, 6, 7, 8, 10, 12, 14 and 16 cm, measured by opposite knots, with each gill net measuring 10 m long.The gill nets were set in littoral areas in the afternoon and removed the following morning, totaling 12 hours of exposure.We also used two semicircular hand nets (80 cm in diameter, 1 mm mesh size), 20 minutes per sampling site, and five hauls of beach seines per sampling site (5 m long, 2 m high, and 5 mm mesh size between opposite knots) to collect fingerling and juvenile migratory fishes.At sampling site Riv3, it was not possible to perform beach seine sampling in March due to depths of over 2 m in littoral areas.
We sampled within four lagoons in the Grande River and four in the Aiuruoca River in order to determine if these areas are used during the early development and further recruitment of migratory fishes.We employed gill nets (2.4, 3, 4, 5 and 6 cm, measured by opposite knots), semicircular hand nets, and beach seines, with the same standardized sampling effort as in the other sampling sites.
Fishes were anesthetized in Eugenol 50mg/l.We eval uated the gonadal maturation stages based of migratory species based on macroscopic characteristics.Specimens were dissected to determine sex and gonadal stage accord ing to Vazzoler (1996) (Table 1).Instituto Chico Mendes de Conservação da Biodiversidade authorized fish collection (License # 72534) and the Ethics Committee on Animal Use/ CEUA of the Universidade Federal de Lavras approved the project (protocol # 112/18).
After gonadal characterization, fishes were fixed in 10% formalin solution and transferred to a solution of 70% alcohol for conservation.In the laboratory, each individual was mea sured (total and standard body length in mm), weighed (g), classified (Nakatani et al. 2001, Ota et al. 2018, Ribeiro et al. 2019) and their migratory status was confirmed (Agostinho et Table 1.Classification of gonadal maturation stages of male and female based on macroscopic characteristics (Vazzoler 1996).

Stage
Female Male Immature F1: Thin and transparent ovaries with low vascularity evident, very small in size.
Initial maturation F2: Bulky ovaries occupying about 1/3 to 2/3 of the celomatic cavity, oocytes are visible to naked eye.
M2: Bulky testicles with lobed shape, milky white in color.
Mature F3: Ovaries reach maximum volume, turgid and with numerous oocytes visible to the naked eye.

Data analysis
In order to understand the general patterns along the longitudinal gradient, data from the sites were grouped in four regions (River, Transition, Camargos and Itutinga) with similar hydrological conditions and characteristics (Table 2).To evaluate the possible differences in numbers, biomass, and richness among groups, we used the KruskalWallis test.Differences among dry and wet seasons were tested for each group using a Wilcox test.We quantified fingerlings and juveniles of migratory species per sampling site using the literature data on fingerlings' maximum sizes and the size of each species at first sexual maturation (Lopes et al. 2000, Nakatani et al. 2001, Esguícero and Arcifa 2010).To assess the second hypothesis and evaluate if the riverreservoir transition areas allow the development of migratory species and to determine how adults use the system, we plotted the number of individuals in each stage (fingerlings, juveniles, and adults) in both seasons.To assess the third hypothesis, gonadal maturation stages were visually compared between regions, in between the wet and dry seasons.

RESULTS
We sampled a total of 898 individuals belonging to seven migratory fish species, corresponding to six genera, four families and two orders (Table 3).
We observed significant differences in numeric abun dance (CPUEn: KW = 15.996;p < 0.05), richness (KW = 14.583; p < 0.05) and biomass (KW = 21.286;p < 0.05) among the regions (Fig. 2), with a general tendency for these numbers to decrease along the gradient.However, there were no season al differences in numeric abundance, richness and biomass.
Fingerlings, juveniles, and adults of migratory species were registered throughout the entire system.Although we failed to detect a unique pattern for each, we identified some trends.In general, migratory species tend to be found in greater numbers in the stretches of freeflowing rivers and transition zones.Juveniles of six migratory fish species and fingerlings of only one, Leporinus friderici (Bloch, 1794), were sampled.They were found in the lotic stretches, while Pimelodus maculatus Lacepède, 1803 was registered through out the entire sampled area (Fig. 3).The stages of gonadal maturation varied from one spe cies to another, and among regions and seasons.In general, the final stages (mature and postspawning) were prevalent during the wet season and more frequent in freeflowing rivers and transition zones (Fig. 4).There were juveniles of P. maculatus in almost every sampling location, but individuals in initial maturation and postspawning stages were also prevalent, being captured in freeflowing rivers and tran sition areas during the wet season.There were individuals of Brycon orbignyanus (Valenciennes, 1850) and Prochilodus lineatus (Valenciennes, 1837) in the postspawning stage in the Itutinga Reservoir, but with signs of follicular atresia.The numbers of individuals of L. friderici and other species in the postspawning stage increased in freeflowing rivers during the wet season.The majority of Megaleporinus piavussu (Britski, Birindelli & Garavello, 2012) adults and juveniles was found in transition areas, with more individuals in the postspawning stage.Adults of Megaleporinus obtusidens (Va lenciennes, 1837) were not found in freeflowing rivers and transition areas, despite the juveniles sampled in these areas during the wet season.Adults in mature and postspawning stages and juveniles of Salminus hilarii Valenciennes, 1850 were only found in freeflowing rivers and transition areas and mainly during the wet season.

DISCUSSION
The numbers, biomass, and richness of migratory fishes differed among regions, and slightly decreased along the gradient surveyed.Migratory fishes were found in greater numbers of individuals and species in rivers and in the transition zone.The biomass decreased from the uppermost to the downmost region, except for Itutinga, possibly due to fish stocking.Juveniles and fingerlings were captured in low numbers of individuals and species richness was also low, indicating that recruitment does not occur in the riv erreservoir transition areas.On the other hand, marginal lagoons seem to present an important environment for the early development of migratory fishes, since four of the seven species were captured in these areas.Mature individuals, which were more frequent in rivers and transition sites, were found during the wet season.The system is used differently by the different migratory species.Most species were found in low numbers in lentic sites and seemed to depend on stretches of freeflowing rivers to complete their life cycles.
The species richness of migratory fishes recorded in our study represents 78% of the total number of native mi gratory species from the Upper Grande River (Santos 2010).Only Zungaru jahu (Ihering, 1898) and Salminus brasiliensis (Cuvier, 1816) were not sampled.These species have large body sizes and specific adaptations for migration and repro duction (Agostinho et al. 2003, ZaniboniFilho et al. 2017) Table 3. Numeric abundance of migratory fish species collected from the Upper Grande River and CI-UFLA voucher number.We observed a greater biomass in river and transition zones, most likely because these migratory species prefer flowing waters to lentic areas (Pompeu et al. 2012, Agostinho et al. 2016).In addition, reservoir cascade systems tend to reduce local connectivity, leading to a loss of habitats for  Groups formed in the system by gonadal maturation stage: 1) immature, 2) initial maturation, 3) mature; 4) post-spawning.migratory species and, consequently, a loss of richness and abundance (Cheng et al. 2015, Loures andPompeu 2018).Such pattern was not evident in our study, possibly because this system is composed of only two mediumsized reservoirs, and migratory fish stocking (mainly B. orbignyanus and P. lineatus) happened in the Itutinga (Alves et al. 1998) and Camargos reservoirs.Therefore, since there is not a selfsus tainable population of migratory fishes in the lowermost reservoir (Itutinga), our data suggest that fish stocking may eventually mask important ecological gradients.
The eventual displacements of migratory species could result in seasonal differences in the structure of their assemblages along the studied gradient, especially in regions connected to the lotic remnant.However, none of the regions showed seasonal differences in richness, numeric abundance and biomass.For the transition region and Camargos, such a pattern indicates that these areas do not play a relevant role in the life cycle of these species, although they are continuously occupied by them.The absence of seasonal differences in the sampled riverine stretches suggests that breeding sites must be located further up the sampled sites.In the case of selfsustainable populations, the spawning site is located far enough to enable the development of ichthyo plankton, preventing larvae from drifting to inappropriate places, such as reservoirs (Olden 2016).Furthermore, the great abundance of P. maculatus, whose migratory nature is controversial, may have masked possible patterns.
Fingerling and juvenile stages were not abundant with in the riverreservoir transition sites, indicating that they did not contribute to migratory fish recruitment.The extreme water level fluctuations due to the operational requirements of the Camargos Dam decreases the habitats and refuge areas available for them (Nobile et al. 2019).Therefore, the migratory species found in the studied systems rely on the abundant floodplain lagoons along the Aiuruoca (Lima et al. 2010) and the Grande rivers to complete their life cycle.In addition, Suzuki et al. (2011) sampled considerable amounts of eggs at the freeflowing stretches upstream of the Camar gos Dam, at the Grande and Aiuruoca rivers, indicating that the upper stretches of both rivers are used for spawning and floodplains for their early development.Presumably, after that, when the water level drops, the juveniles recruited at the floodplains return to the river in search of their feeding areas (Agostinho et al. 2003, Silva et al. 2015).
Juveniles and adults of Pimelodus maculatus were found in almost all groups and seasons, although individuals in mature and postspawning gonadal maturation stages had been found mainly in freeflowing rivers and riverreservoir transition areas during the wet season.There are questions as to whether P. maculatus can be considered a migratory species.(Santos et al. 2012, ZaniboniFilho and Schulz 2003Oldani et al. 2007, Arcifa and Esguícero 2012).According to Agostinho et al. (2003), it probably needs fewer free stretches of river to complete its life cycle.In the Paraná River Basin, P. maculatus is considered widely distributed (Agostinho et al. 2003), even in reservoirs (Agostinho et al. 2007a), as demonstrated by its dominance in five cascade reservoirs in the Araguari River (Loures and Pompeu 2018).Although juveniles have been previously collected from the Itutinga Reservoir, which has no freeflowing stretches or important tributaries, most of the adults presented the immature gonadal stage and no fingerlings were captured, indicating that they were found there because individuals are passing through the spillway or turbines of the Camargos Dam (Alves et al. 1998).
Samples of Brycon orbignyanus and P. lineatus in the postspawning stage were found within the Itutinga Res ervoir, although no juvenile or fingerlings were collected in this area.Juveniles of B. orbignyanus were sampled in the Camargos Reservoir.The first is categorized as an en dangered species (Akama et al. 2018).The decline in the numbers of B. orbignyanus has been linked to the changes in the original flood regimen, especially changes caused by dam constructions, and removal of riparian vegetation (Tonella et al. 2019).P. lineatus travels hundreds of kilome ters to reproduce (Makrakis et al. 2012) and the decline in the number of individuals of this species is also associated with an interruption in the connectivity between habitats caused by dams, and also overfishing (Baigun et al. 2013).Both species depend on specific environmental conditions such as freeflowing rivers and nursery areas to complete their life cycles (Agostinho et al. 2003, ZaniboniFilho andSchulz 2003).Their presence in the study area is probably associated with fish stocking, implemented by the local power company (CEMIG) at Itutinga (Alves et al. 1998) and Camargos reservoirs until 2019.
As expected for a species depending on freeflowing rivers for reproduction (Resende 2003, Agostinho 2007a), the frequency of capture of L. friderici increased in the freeflowing rivers during the wet season.An increase in the number of individuals in the postspawning stage in these areas was observed at the same period.L. friderici individuals, which inhabit reservoirs, depend on lotic environments to reproduce (Lopes et al. 2000).This species was the only one for which we collected fingerlings and juveniles, suggesting that its recruitment might occur at the riverreservoir tran sition areas.However, it is important to confirm whether these areas are often used for recruitment or weather the occurrence of L. friderici there is sporadic.
In the cascade reservoir system with important freeflowing river stretches, upstream of the Camargos res ervoir, we verified that migratory fish reproduction seems to depend on the freeflowing rivers and their associated floodplains.On the other hand, riverreservoir transition areas do not seem to play a significant role in recruitment.Therefore, the lotic stretches of the Grande and Aiuruoca rivers seem to maintain selfsustainable populations of the registered migratory species, acting as source areas for the individuals that use the reservoirs.In this sense, we highlight the importance of maintaining these freeflowing rivers and the conservation of their floodplains.Finally, we encourage further research aiming to investigate migratory fishes and icthyoplankton in the upper stretches of the Grande and Aiuruoca rivers, including their headwaters, to confirm the observed patterns.The maintenance of freeflowing stretches upstream of reservoirs and tributaries should be considered when planning new hydropower plants.This could help attenuate the regional impacts of these plants, especially on migratory fishes.

Figure 4 .
Figure 4. Gonadal maturation stage of migratory species during the wet (right side) and dry (left side) seasons in each group.Groups formed in the system by gonadal maturation stage: 1) immature, 2) initial maturation, 3) mature; 4) post-spawning.

Table 2 .
Sampling sites and groups formed considering similar hydrological conditions and characteristics.Marginal lagoons were also represented only to characterize the area.
MLG2Grande marginal lagoon About 6 km from Riv1.Native riparian forest nearby and human influence.Used by cattle.During high flood, it can connect with other smaller lagoons and get a lot wider.-21.5228°,-44.1674°MLG3GrandemarginallagoonAbout 2.5 km from Riv1.Presence of native riparian forest and human influence.Used by cattle.During high flood, it can get large and deep.-21.486786°,-44.1870°MLG4Grandemarginal lagoon About 6 km from Riv1.Large lagoon with native riparian forest.During high flood, it can get large and deep.Little human influence.-21.4872°,-44.1786°A