Introduction

El Niño Southern Oscillation (ENSO) is a large-scale climatic phenomenon related to variations in the sea surface temperature of the Pacific Ocean (Philander, 1983; Trenberth, 1997), which can lead to several meteorological changes, and it is classified in warm (El Niño) and cold (La Niña) phases. The strength of ENSO events can be measured with the Oceanic Niño Index (ONI), which allows inferences on the occurrence and intensity of El Niño or La Niña events. These events may result in quite different responses across regions and impact both marine and continental ecosystems (Caviedes, 2001; Garcia et al., 2004, Collins et al., 2010; Su et al., 2018; Pereira et al., 2020). They can cause changes in the global hydrological cycle, as well as patterns of precipitation resulting in extreme drought or flood in different parts of the planet (Ropelewski & Halpert, 1987; Philander, 1990; Grimm & Tedeschi, 2008; Henderson et al., 2018).

Environmental changes caused by ENSO events (i.e., El Niño and La Niña) may affect diverse parameters of ecological communities in different temporal and spatial scales (Holmgren et al., 2001; Glynn et al., 2017; Pineda et al., 2019). Regarding fish assemblages, changes in the composition and in patterns of distribution, recruitment, migration, and seasonality of several species have been directly related to these events (Godínez-Domínguez et al., 2000; Garcia et al., 2001; Garcia et al., 2004; Brander, 2007). These effects are well demonstrated for marine environments (Hollowed et al., 2001; Watters et al., 2003; Lehodey et al., 2005; Booth et al., 2011) and estuaries (Mol et al., 2000; Garcia et al., 2003; Possamai et al., 2018), where environmental changes can lead to a delayed response in the recruitment of some species (Moraes et al., 2012). However, few studies explored ENSO effects over freshwater fishes. In this context, historical drought events related to ENSO were associated with high mortality of fishes due to algal blooms in the Fly River system (Swales et al., 1999), as well as with the collapse of a migratory species in the Pilcomayo River floodplain, which depends on high water levels for their reproductive migration process (Smolders et al., 2000).

The hydrological cycle of diverse basins around the world are related to ENSO events, which may influence the frequency of floods, drought, water quality, discharge, and river flow (Chiew et al., 1998; Keener et al., 2010; Su et al., 2018). However, the natural conditions of many rivers – and consequently their biota – have been intensely modified by human activities, especially by the construction of dams (Dudgeon et al., 2006; Agostinho et al., 2008). For example, dams can reduce floods that would be caused by ENSO events, and may intensify drought events downstream (Ponton, 2001). Dams change rivers from lotic to lentic systems, affecting physical and chemical properties of water, nutrient cycling, habitat structure, species distribution and taxonomic and functional composition of communities (Tundisi & Straskraba, 1999; Fernandes et al., 2009; Oliveira et al., 2015). Fish diversity is intensely affected by dams, since these structures may affect the abundance, reproduction, development, recruitment of several species, and they often cause local or even regional extinctions (Terra et al., 2010; Cheng et al., 2015; Moran et al., 2018). These impacts are even more relevant considering migratory species, which depend on floods to fulfill their life cycles (Gubiani et al., 2007; Agostinho et al., 2007, 2016). Migratory species travel over long distances, depend on floods for reproduction, and they also use adjacent habitats to the main channel of rivers (e.g., floodplain lakes or oxbow lakes) as nursery and feeding sites (Baumgartner et al., 2004; Agostinho et al., 2004a, 2008).

In different stretches of the Paraná River (in Brazil, Paraguay and Argentina), it was demonstrated that the interannual variations in flow and water discharge are associated with ENSO events (Berri et al., 2002; Dai et al., 2009; Antico et al., 2018). El Niño events were also linked to intense floods in these regions in recent decades (Camilloni & Barros, 2000; Fernandes et al., 2009). However, in Brazil, there are about 150 reservoirs in the Paraná River and in its main tributaries, which have strong regulatory effects over their hydrological regime, changing historical patterns and, consequently, disturbing aquatic communities (Ward & Stanford, 1995; Agostinho et al., 2007; Stevaux et al., 2009; Agostinho et al., 2016). The Upper Paraná River region holds an extensive floodplain, composed of the Paraná River, its main channel, and regulated and unregulated tributaries (e.g., Paranapanema and Ivinhema rivers, respectively). The floodplain was intensively affected by the construction of the Sérgio Motta dam in its main channel, which halved the floodplain extension area, strongly affecting the water discharge and the water level oscillations (Agostinho et al., 2004a; Gubiani et al., 2007; Stevaux et al., 2009).

Paraná and Ivinhema rivers were recorded to have significant differences considering their fish assemblages, and one of the main factors influencing these differences is the respectively presence and absence of upstream dams (Fernandes et al., 2009; Granzotti et al., 2018; Oliveira et al., 2018). Since dam constructions have important regulatory effects on the water flow in downstream sites, it is possible that these impoundments interfere in the ENSO influence in the Paraná River. Based on this, the main goal of this paper was to evaluate the effects of ENSO (indexed by ONI) in the fish assemblages of the upper Paraná River floodplain. For this, we seek to answer the following questions: (i) Are ENSO events directly associated with water level of the rivers that compose the upper Paraná floodplain?; (ii) Do ENSO events influence fish assemblage attributes (total abundance, abundance of migratory species, species richness and assemblage structure) of Paraná and Ivinhema rivers?; and (iii) Are there different association patterns regarding ENSO events with water level and assemblage attributes in the dammed and undammed rivers?

Material and Methods

Study area

The upper Paraná River floodplain is located in the last lotic stretch of the Paraná River inside Brazil, between the Itaipu Reservoir (downstream) and the Sérgio Motta Dam (locally known as Porto Primavera; upstream). This stretch is approximately 250 km in length and covers an area of 5,268 km2 (Agostinho et al., 2004b). The main tributary in the East margin is the Paranapanema River, which is also intensely dammed (the last dam in this river is the Rosana Reservoir) and contributes to the regulation of the Paraná River water level.

The floodplain consists of three rivers: Ivinhema, Paraná, and Baía. For this study, sampling was carried out in six stations: three in the Paraná River (intensely impacted by the construction of dams) and three in the Ivinhema River (free of dams and in more pristine conditions), including the main channel and adjacent floodplain lakes, connected or not connected to the main channel of the rivers (Fig. 1). Sampling stations are sentinel stations of a Long Term Ecological Research (LTER), Site 6: Ventura Lake, Patos Lake, Ivinhema River channel (all located in the Ivinhema River), and Pau Véio Lake, Garças Lake and Paraná River channel (all located in the Paraná River).

Fig. 1
figure 1

Study area of the upper Paraná River floodplain, where: 1—Ventura Lake; 2—Patos Lake; 3—Ivinhema River; 4—Pau Véio Lake; 5—Paraná River; 6—Garças Lake

Full size image

Data collection

Monthly ENSO data from 1964 to 2018 were obtained from the National Oceanic and Atmospheric Administration (NOAA). ENSO data are given by the Oceanic Niño Index (ONI), for which values greater than or equal to 0.5 represent El Niño events and values less than or equal to -0.5 represent La Niña events (Fig. 2a). The values of ONI are calculated as quarterly means of ERSST.v5 SST anomaly for the Niño 3.4 region (i.e., 5° N–5° S, 120°–170° W; Huang et al., 2017). Values between − 0.5 and 0.5 represent Neutral periods when none of the events had strength to occur (i.e., absence of ENSO events).

Fig. 2
figure 2

Temporal series of the Oceanic Niño Index (ONI) and of the water level used. a Annual variation of ONI values (1964–2018); Dashed lines = thresholds of El Niño and La Niña events. b Daily variation in the water level of the Paraná River (1964–2018), which RD Rosana Dam; SMD Sergio Motta Dam; Dashed lines = period of 2000–2016 in the Paraná River. c Daily variation in the water level of the Ivinhema River (2000–2016)

Full size image

The hydrological variable used was the water level, from which we obtained two temporal series: 1964–2018 for the Paraná River and 2000–2016 for the Ivinhema River. The daily water level (m) of the Paraná River was obtained from Porto São José fluviometric station (code 64575000 – “Agência Nacional de Águas”; National Water Agency), located about 12 km upstream the study area. The data of Paraná River were grouped and analyzed in four time periods (Fig. 2b): (i) until the construction of Rosana Dam in the Paranapanema River (1964–1984); (ii) until the construction of Sérgio Motta Dam in the Paraná River (1964–1997); (iii) the complete temporal series (1964–2018), to verify the degree of alterations prompted by dams in the water level of the Paraná River; and (iv) the analogous temporal series for the Ivinhema River (2000–2016), to compare the two rivers in the same time scale. Daily water levels (m) of the Ivinhema River were obtained from Porto Sumeca fluviometric station (code 64617000 – “Agência Nacional de Águas”), located about 50 km upstream the study area (Fig. 2c).

Fish samplings were carried out quarterly (as the values of the ONI) in the surveys conducted in the Paraná and Ivinhema rivers under the Long Term Ecological Research, from 2000 and 2018. Fishes were collected using 11 gillnets (20 m long) with different mesh sizes (2.4 until 16.0 cm between opposite knots), which were operated near the margins and set for 24 h in each sampling location. Sampled individuals were anesthetized with eugenol (Resolution n° 1000/12, “Conselho Federal de Medicina Veterinaria”), euthanized, packed in thermal boxes, and transported to the field laboratory. After, fish were identified (Graça & Pavanelli, 2007; Ota et al., 2018). For each individual captured, we recorded the date and sampling station. As the fishing effort were the same for all samplings, the number of individuals caught was considered the abundance. Fish data were evaluated separately according to each river (Paraná and Ivinhema rivers). As we were not interested in the spatial variability inside each river, sampling locations (lakes and main channel) were not considered in the analysis.

Data analysis

To evaluate if ENSO events are directly associated with water level (first question), we applied cross-correlation function (CCF) analyses (Chatfeild & Xing, 2019) considering ONI as the time series of influence and the water level time series as the affected time-series (see Probst et al., 2012). For the water level data, CCFs were performed with each period of Paraná and Ivinhema rivers described in the section data collection, to evaluate the possible influence of dams on the floodplain responses to ENSO events.

To determine if ENSO events influence fish assemblage attributes (second question), we calculated the total abundance, abundance of migratory species, and species richness for each monthly sample. To assess the variation in the structure of fish assemblages for each river, a Principal Coordinate Analysis was applied (PCoA; Anderson et al., 2006) on the abundance data using the function “cmdscale” from the package “stats” in the R Environment (R Core Team, 2020). The PCoA was performed on the Bray-Curtis resemblance matrix and the first two axes of PCoA, which represented most of the variability in the original data matrix, were retained for interpretation. Thus, we obtained five dependent variables for each river (total abundance, abundance of migratory species, species richness, and structure of fish assemblage represented by PCoA1 and PCoA2), referred as fish assemblage attributes. After, cross-correlation functions were performed for each river separately, with ONI as the time series of influence and each assemblage attribute as the affected time series.

Cross-correlation function analysis allowed us to quantify the association and to identify the lags between the two-time series as a function of the displacement of one of the series relative to the other (third question). Thus, we were able to evaluate the nature of the association (positive or negative) of ONI with water level and assemblage attributes, the lag which presented the higher correlation value in these relationships (i.e., the time in which the association was stronger), and the extent of the response of the affected temporal series (i.e., how long the correlation was significant).

As the hydrological cycle lasts approximately a year in the upper Paraná River floodplain, CFFs were performed with a maximum lag of one year for water level. On the other hand, CFFs were performed with a maximum lag of two years for fish assemblage attributes since some fish species might present a delayed response to environmental changes (Moraes et al., 2012; Baumgartner et al., 2018). Juveniles of migratory species, for example, will be recruited to fishery stocks about one or two years after flood events (Agostinho et al., 1993; Gomes & Agostinho, 1997; Agostinho et al., 2004a; Oliveira et al., 2015). According to our sampling design, each lag in CCFs output for water level corresponded to one day, while for fish assemblage attributes time series corresponded to one month.

In some cases, it might be difficult to interpret which variable is leading and which variable is lagging in temporal analyses (Lingard et al., 2017). Here, we assume that ONI leads water level and fish assemblage attributes since the opposite was ecologically impossible. Therefore, only the left side of the CCFs graphical results was interpreted, which despite presenting negative values on the horizontal axis, it represents positive time lags of water level and assemblage attributes relative to the ONI time series. As it was difficult to visualize the stronger correlations in the graphical results of the CCFs regarding to water level data, we applied a LOESS Curve Fitting (Local Polynomial Regression) to find the higher lag-correlation between ONI and water level of each river. Therefore, water level results of CFFs were also summarized in a table. Significant values of CCFs were obtained by \(\left( {\alpha = 0.05} \right)\) at \(\pm 2/\surd n\), where n is the length of the time series (Berryman & Turchin, 2001; Chatfeild & Xing, 2019). The analyses were performed using the functions “ccf” from the package “forecast” (Hyndman et al., 2020), “cmdscale” and “loess” from the package “stats” in R Environment (R Core Team, 2020).

Results

The four analyzed periods of the water level in the Paraná River showed a declining pattern over time regarding the association with ONI, and an increasing pattern regarding time lag, after the construction of each dam (Rosana Dam and Sergio Mota Dam; Table 1). A higher correlation value between ONI and Paraná River time series was found before the construction of the Rosana Dam. The period before the construction of the Sérgio Mota Dam presented a similar correlation value with the entire time series of the Paraná River. The CCF between ONI and water level of the Ivinhema River presented the highest correlation value when compared with all Paraná River temporal series, mainly considering the same period (2000–2016; Table 1). In addition, the time lag of the Ivinhema River demonstrated a similar trend to the period of 1964–1984 in the Paraná River. In general, the extent of the water level time series regarding ONI lasts the entire period (maximum lag of one year) for both rivers (Fig. 3).

Table 1 Summarized results of Cross-Correlation Functions (CCFs) between the Oceanic Niño Index (ONI) and the water levels of Paraná and Ivinhema rivers after applied a LOESS Curve Fitting (Local Polynomial Regression). r = higher correlation value at a respective lag time. All values were significant
Full size table
Fig. 3
figure 3

Cross-Correlation Functions (CCFs) between the Oceanic Niño Index (ONI) and the water level of Paraná and Ivinhema rivers. a period of 1964–1984 of Paraná River; b period of 1964–1997 of Paraná River; c the entire time-series of Paraná River (1964–2018); d period of 2000–2016 of Paraná River; e Ivinhema River time-series (2000–2016). The dashed lines indicate the thresholds of significance

Full size image

The CCFs performed with the fish assemblage attributes showed different patterns between Paraná and Ivinhema rivers (Fig. 4). The Paraná River presented a significant and negative association between the total abundance and ONI (r = -0.193; Fig. 4a), while a significant and positive associations were found for the abundance of migratory species (r = 0.251; Fig. 4c) and PCoA 2 (r = 0.308; Fig. 4i) with ONI. Species richness and PCoA 1 did not present significant correlation values with ONI in the Paraná River (Fig.4e and 4g, respectively). Besides, for the Paraná River, the variables which presented the higher extent were total abundance and PCoA 2 (16 months) while the variable which presented the lower extent was abundance of migratory species (11 months). Considering the time with higher correlation values, Paraná River presented higher time lags correlation in 14 months for total abundance, 8 months for the abundance of migratory species, and 12 months for PCoA2.

Fig. 4
figure 4

Cross-Correlation Functions (CCFs) between the Oceanic Niño Index (ONI) and each fish assemblage attribute of Paraná (left side graphs) and Ivinhema (right side graphs) river. The dashed lines indicate the thresholds of significance. E = extent of the association (considering the last significant value of lag)

Full size image

In the Ivinhema River, all fish assemblage attributes presented significant and positive associations with ONI (Correlation values: Total abundance = 0.279, Abundance of migratory species = 0.216, Species Richness = 0.256, PCoA1 = 0.219; Fig. 4b, d, f, h, respectively), except PCoA 2 (Fig. 4j). In addition, in the Ivinhema River, the variable which presented the higher extent was species richness (16 months) while the variables which presented the lower extent were abundance of migratory species and PCoA 1 (12 months). Considering the time with higher correlation values, Ivinhema River presented higher correlation lags in 8 months for total abundance and for abundance of migratory species, in addition to 6 months for species richness and for fish assemblage structure, represented in PCoA1.

Discussion

Answering our three questions, the results of this study showed that: (i) ENSO events are directly associated with the water level variation of the studied rivers; (ii) ENSO events influenced fish assemblage attributes of Paraná and Ivinhema rivers; and (iii) water level and fish assemblage attributes in the dammed and undammed river presented different association patterns regarding to ENSO events.

There was a positive and significant association between ONI (ENSO index) and water level of Paraná and Ivinhema rivers. However, this association decreased in the Paraná River as the river became more regulated, suggesting that upstream dams may interfere in the influence of ENSO on the water level of the Paraná River. This can be sustained mainly by the contrasting results in the period of 2000–2016 of Paraná and Ivinhema rivers, which represented the period that the water level time-series of the Paraná River was most intensely affected by upstream dams. Moreover, the increasing values of the time lag response of water level to ONI also corroborated our argument that dams may influence the ENSO effects in this river. It is also important to highlight that, in the Paraná River, the water level variation was greatly changed after the construction of dams, decreasing the frequency and intensity of floods, and also decreasing maximum and minimum water levels, as demonstrated by our historical series, and previously by Gubiani et al. (2007). Meanwhile, the Ivinhema River is free of dams and presented a higher correlation value than all periods in the Paraná River.

Higher values of ONI, which represent El Niño events, were associated with higher values of water level and most of the fish assemblage attributes in the Paraná and Ivinhema rivers. Thus, it is possible that El Niño events are associated to floods in the studied rivers as suggested by Fernandes et al. (2009), indicating that these events affect the structure of the fish assemblages and their attributes, mainly in the Ivinhema River due to the absence of dams regulating its water level. Previous studies showed that ENSO events may present a strong effect on fish assemblages that inhabit floodplains, regarding its effects on the hydrological regime (Swalles, 1999; Smolders et al., 2000; Camacho Guerreiro et al., 2020). Besides, Ponton (2001) also evidenced that dams may amplify droughts caused by these events on the Sinnamary River, in the Amazon region, directly affecting the fish assemblage downstream. Thus, our results corroborate that ENSO events may affect fish assemblages in floodplain ecosystems, but the presence of dams upstream can interfere in this association.

The differences found between fish assemblage attributes in the Paraná and Ivinhema rivers are probably related to the different environmental conditions of each river. In the Paraná River, dams cause the retention of sediments and flow regulation which leads to several impacts downstream, mainly considering the fish assemblage (Gubiani et al., 2007; Granzotti et al., 2018). Dams interrupt migration routes of migratory species and reduce the recruitment of fish species by trapping eggs and larvae in upstream sites, besides changing the food-web structure and favoring visual piscivorous fish (Suzuki et al., 2011; Turgeon et al., 2019; Rodrigues et al., 2020). In contrast, the Ivinhema River presents more pristine conditions, with higher values of total abundance and species richness than the Paraná River (Fernandes et al., 2009; Granzotti et al., 2018; Oliveira et al., 2018). Thus, it is possible that the differences in fish abundance and species richness between Paraná and Ivinhema rivers lead to different responses regarding ENSO events, as evidenced by our results.

Total abundance of migratory species was positively associated to ONI in the Paraná and Ivinhema rivers, despite the negative effect of dams over these species in the Paraná River (Agostinho et al., 2004a Makrakis et al., 2019). This result demonstrated that higher water levels triggered by El Niño events may favor the maintenance of migratory species populations, which take advantage of floods in the Paraná River and migrate to unregulated tributaries to spawn (Agostinho et al., 2003; Baumgartner et al., 2004). Besides, regarding the extent of the response to ONI, we found that the abundance of migratory species presented a significant association until 11 months and 12 months in the Paraná and Ivinhema rivers, respectively. Recruitment and abundance responses of migratory species may be slower when compared with non-migratory species due to their long life cycles (Agostinho et al., 1993; Gomes & Agostinho; 1997; Bailly et al., 2008; Suzuki et al., 2009; Oliveira et al., 2015), occurring a year after the flood (Agostinho et al., 2004a). Thus, the extent of the association in the two rivers may be explained by the delayed recruitment of migratory species.

The structure of the fish assemblages (indexed by the PCoA axes) was also positively associated to ONI (PCoA 2 for the Paraná River and PCoA 1 for the Ivinhema River). This was expected since life history traits of fishes of the upper Paraná River floodplain are closely related to the hydrological cycle (Agostinho et al., 2004c). Drought periods can facilitate the predatory activities of piscivorous fish, by increasing the confinement and density of prey (Rodríguez & Lewis, 1997; Turesson & Bronmark, 2007). On the other hand, flood periods can increase the input of allochthonous feeding resources for fish (e.g., detritus), provide nursery areas and shelter by flooding terrestrial vegetation, and generate bottom-up effects by affecting other organisms of the floodplain (e.g., phytoplankton, zooplankton and macrophytes) (Bailly et al., 2008; Aspin et al., 2018; Granzotti et al., 2018; Quirino et al., 2018).

Final considerations

The Oceanic Niño Index (ONI) proved to be a good proxy to assess the relationship of ENSO events with the water level variation and fish assemblages. The relationship between ENSO events and hydrological variables of the Paraná River has already been demonstrated in previous studies (Amarasekera et al., 1997; Camilloni & Barros, 2000; Berri et al. 2002; Antico et al., 2018), but for other regions of the basin. Here, beyond the association between ONI and the water level of the rivers that compose the upper Paraná floodplain, we demonstrated how dams may interfere in the effects of ENSO events in the freshwater environments of the region. Besides, previous studies suggested that ENSO effects were associated with fish assemblage attributes (e.g., Fernandes et al., 2009), and our findings corroborated this association, which might also be true for other tropical floodplains. El Niño events were associated with migratory species in both rivers and this response was evidenced until nearly one year after the events. It has also been shown that the construction of dams seems to have an even greater impact on the environments and organisms that live downstream, once dams can interfere in the effects of major climatic events over the ecosystems. Further studies should consider the temporal dynamics of changes caused by ENSO events in different hydrological attributes (e.g., seasonality, duration and intensity of floods), and consequently these effects over biological communities.