Introduction

By 2050, the world population will reach 9.6 billion inhabitants. In order to meet the simultaneously increasing demand for food, world food production will have to raise by 70 %. Among the multiple alternatives addressing food supply, aquaculture is the animal farming sector that has mostly grown worldwide, coming to an average annual growth of 8.6 % between 1980 and 2012 (FAO 2014).

The world largest aquaculture producer is China with approximately 47.8 million tons. Although it holds the 17th position in the world ranking, with 479,399 tons in 2010 (MPA 2011), Brazilian aquaculture has grown expressively in the past few years, recording a 51 % increase in production between 2009 and 2011. Most aquaculture production in Brazil comes from continental aquaculture, especially from fish farming (87 % of the total production in the country), and Tilapia (Oreochromis niloticus L.) comes first as the most produced species (Ostrensky et al. 2008; MPA 2011) countrywide. Production has mainly increased because more fish farming companies are using net cages in the aquaculture areas of hydroelectric reservoirs, as is the case of the studied reservoir in the state of São Paulo (Mallasen et al. 2012).

Super-intensive fish farming systems, as is the case of net cages, present lower implementation costs when compared with semi-intensive systems, and high productivity. However, non-ingested food and fish excreta are directly released in the environment, raising principally nitrogen and phosphorus concentrations in the water (Guo and Li 2003; Guarino et al. 2005; Guo et al. 2009) and sediments (Boyd et al. 2007), which cause artificial eutrophication. Considered as one of the main environmental impacts related to aquaculture, artificial eutrophication interferes with water quality and causes modifications in the structure and dynamics of aquatic communities, especially phytoplankton communities, promoting cyanobacteria and algal blooms (Diaz et al. 2001; Borges et al. 2010; Kaggwa et al. 2011).

Blooms result in great social, economic, and environmental impacts, for example metabolite production (geosmina and 2-methylisoborneol-2-MIB) which lends undesirable odor and taste to both water and fish (Smith et al. 2008); bioaccumulation of cyanotoxins in the fish muscle tissues (Deblois et al. 2008; Eler et al. 2009; Dörr et al. 2010) that poses potential health risks to animals, birds and mammals as well as men whoever consume them; skin lesions and gill clogging result from excessive algae cells, which makes breathing difficult and even kills fishes and can also make the activity itself unfeasible (Seymour 1980; Xavier et al. 1991; Magalhães et al. 2003; Souza et al. 2012).

Despite the expressive growth of fish farming in net cages in Brazil, potential impacts of the activity in water bodies and the efficacy of the phytoplankton community as an indicator of the resulting changes, few studies have evaluated the influence of the involved activities on the structure of the phytoplankton community. Some of these studies were carried out by Borges et al. (2010) and Lins (2011) in tropical reservoirs and by Cavalcante (2010) and Bartozek et al. (2014) in subtropical reservoirs. High values of Cyanobacterial biomass and blooms mainly composed of Microcystis aeruginosa were recorded by Lins (2011), near net cages, indicating that intensive fish farming was an important anthropogenic disturbance in the reservoir. On the other hand, studies performed by Cavalcante (2010), Borges et al. (2010), and Bartozek et al. (2014) did not record the significant influence of fish farming on physical and chemical characteristics and on the phytoplankton community. Both Borges et al. (2010) and Bartozek et al. (2014) concluded that variations in the phytoplankton community and the analyzed physical and chemical parameters seem to have been mainly influenced by seasonality, by the small number of net cages, and by the hydrodynamics of the studied environments.

Therefore, the objective of this study was to analyze the phytoplankton community response and the spatial and temporal variation of abiotic variables due to installation of fish farming system for the production of tilapias in large volume net cages in a reservoir. We have tested two hypotheses: (1) there are higher nutrient concentrations in the site with tilapia farming than in the other two sites without farming; (2) net cages installed for tilapia farming will increase nutrient concentrations and consequently promote temporal changes in the Cyanobacteria richness and biomass.

Materials and methods

The study was carried out in the fish farming area, located at the Ponte Pensa aquaculture area in the reservoir of Ilha Solteira power plant, SP, Brazil (Fig. 1). The cage culture is composed of two modules of eight large volume cages, 1200 m3 each (20 × 200 × 20 m), where 120 tons of tilapias are produced and, on average, 140 tons of food are used monthly.

Fig. 1
figure 1

a The arrow indicates the location of Ilha Solteira reservoir in São Paulo State, Brazil. b The rectangle indicates the location of the lateral arm of the Ponte Pensa River in Ilha Solteira reservoir. c Geneseas fish farming area (20°17′S and 50°58′O) indicated by the dark line and arm area of the Ponte Pensa River by the light line. d Location of sampling sites: S1 (750 m downstream from the farming site, 20°16′134″S and 50°59′107″O; average depth of 30 m), S2 (at the farming site, 20°16′452″S and 50°58′812″O; average depth of 27 m), and S3 (650 m upstream from the farming site, 20°16′853″S and 50°58′980″O; average depth of 26 m) (Mallasen et al. 2012—modified)

Full size image

Ilha Solteira reservoir has a cumulative hydraulic operation system that controls the Parana River flow. The reservoir has a flooded area of 1195 km2, volume of 210.6 × 108 m3, and a water average residence time of 47.6 days and is classified as oligotrophic (CESP 2006; Mallasen et al. 2012). The climate in the region is tropical humid, dry in the winter, and the average annual rainfall ranges between 1100 and 1500 mm (Mallasen et al. 2012).

The Ponte Pensa aquaculture area, located in Ilha Solteira reservoir, 20°16′34,96″S and 50°59′02,75″W, was created for fish farming using net cages and it is the first aquaculture area to operate in Brazil. The aquaculture area has a delimited area of 30.88 km2, an average depth of 10.4 m, an average output of 172 m3 s, a residence time of 21.6 days, and a supporting capacity of 4599 tons of phosphorous per year (David et al. 2015).

Samples were collected every month for 2 years (August 2011–July 2013) from three collection sites: S1 (750 m downstream from the farming site, 20°16′134″S and 50°59′107″O; average depth of 30 m), S2 (at the farming site, 20°16′452″S and 50°58′812″O; average depth of 27 m), and S3 (650 m upstream from the farming site, 20°16′853″S and 50°58′980″O; average depth of 26 m) (Fig. 1). The sites sampling were defined with GPSMAP 76CS/Garmin. In order to evaluate changes in the phytoplankton community caused by the activities involved in tilapia farming with net cages, collections started five months prior to the installation of the net cages. Thus, sampling began in August 2011 and the cages were installed in January 2012. Water temperature (°C), pH, electrical conductivity (µS cm−1), and dissolved oxygen (mg L−1) were measured in situ on the surface with the YSI Professional Plus (ProPlus) multiparameter meter. Water transparency (m) was estimated with a Secchi disk (m). Water samples for other abiotic variables were collected in Van Dorn sampling bottles. Ammonia, nitrogen, and total phosphorus were analyzed according to the methodology described in APHA (2005).

Phytoplankton samples for taxonomic analysis were collected with a 20 μm-mesh plankton net. The collected material was preserved in 4–5 % formol. Taxonomic analysis was carried out with a Zeiss Axioplan 2 imaging microscope. Samples for the phytoplankton quantitative analysis were collected from the subsurface in collecting bottles and preserved in 1 % acetic Lugol’s solution. Quantitative analyses were performed according to the Utermöhl’s (1958) method, on a Zeiss Axiovert 25 inverted microscope. A 50-mL sedimentation chamber was employed. The sample sedimentation time was 3 h for each centimeter of the bucket height (Lund et al. 1958). Counting limit was established through the species-rarefying curve and until reaching 100 individuals of the most common species. Each one of the cells, colonies, coenobia, and filaments was considered as an individual.

Richness (R) was considered as the total number of taxa found in each quantitative sample. Biovolume was estimated by multiplying the density of each taxon by the average volume of 30 individuals, depending on the size of the population. Cell volumes were calculated based on geometric models according to Hillebrand et al. (1999) and Fonseca et al. (2014).

Species that presented total relative biovolume above 2 % and together totaled 80 % of the total biovolume were considered as descriptor species.

Information about rainfall was provided by the São Paulo state agrometeorological center CIIAGRO (Centro Integrado de Informações Agrometeorológicas da Secretaria de Agricultura e Abastecimento do Estado de São Paulo).

The joint evaluation of abiotic data was based on the principal component analysis (PCA). The canonical correspondence analysis (CCA) was used to correlate abiotic and biotic variables (descriptor species), using PC-ORD 6.0 for Windows (McCune and Mefford 2011). Data were converted applying [log (x + 1)].

Results

Total annual rainfall was 1343 mm in 2011, 964 mm in 2012, and 1089.7 mm in 2013. The lowest rainfall values, below 10 mm, were recorded in winter—August/September of 2011, July/August of 2012, and July of 2013—whereas the highest, above 214 mm, were recorded in summer—January and March of 2013 (Fig. 2a). Rainfall data confirmed the seasonality, humid summer, and dry winter. The average, minimum, and maximum values of physical and chemical variables are shown in Table 1.

Fig. 2
figure 2

Monthly rainfall (mm) (a) in Santa Fé do Sul, SP region, and air and water temperatures (°C) (b) and water transparency (m) (c) in sampling sites S1, S2, and S3 during the studied period (source CIIAGRO, Centro Integrado de Informações Agrometeorológicas da Secretaria de Agricultura e Abastecimento do Estado de São Paulo)

Full size image
Table 1 Physical and chemical variables related to the water in the Ponte Pensa River in Ilha Solteira reservoir and sampling stations (S1 = 1, S2 = 2, and S3 = 3), between August 2011 and July 2013
Full size table

The principal component analysis (PCA) explained 56 % of the joint variability of the data on the two first components, indicating the seasonality as the main factor that coordinated the abiotic changes between August/2011 and December/2012. All sampling units of 2011 and most of 2012 are on the negative side of axis 1, associated with the highest values of dissolved oxygen (DO) (r = 0.7) and water transparency (Trans) (r = 0.6) (Fig. 3). On the other hand, the sampling units of 2013 are clustered on the positive side of axis 1 along with some sampling units of 2012 associated with increased concentrations of ammonium ion (NH 4) (r = 0.8) (Fig. 3). The sampling units of October, November, and December 2012 are grouped on axis 2 positive side, associated with increased concentrations of nitrogen and total phosphorus (TN and TP) (r = 0.7 and 0.5) and with high water temperatures (r = 0.5) (Fig. 3). PCA separated sample units in time scales; the three seasons presented the same standards and the same abiotic responses over the studied period.

Fig. 3
figure 3

Principal component analysis (PCA) based on nine limnological variables of 72 sample units (temporal and spatial). Abiotic variables: temperature (Temp), dissolved oxygen (DO), electrical conductivity (Cond), turbidity (Turb), transparency (Trans), pH, total phosphorus (TP), ammonia (NH4), and total nitrogen (TN). Sample units are identified according to the year they were collected as shown: months (Ja January, F February, Mr March, A April, Mi May, Jn June, Jl July, Ag August, S September, O October, N November, D December) and sampling sites (site 1 = 1, site 2 = 2, and site 3 = 3)

Full size image

Two time phases were identified: Phase 1—with samples taken before the net cages were installed, between August and December 2011 and from January to November 2012, characterized by high concentrations of dissolved oxygen and higher pH and water transparency values (negative side of axis 1, Fig. 3), and Phase 2—corresponding to samples from January to July 2013, 1 year after the installation of the net cages, characterized by higher concentrations of total phosphorus and ammonium ion and high conductivity values (positive side of axis 1, Fig. 3 PCA).

Two hundred and six taxa were identified and distributed in 11 phytoplankton groups. Chlorophyceae (73 taxa) and Cyanobacteria (34 taxa) contributed the most to species richness. In October, November, and December of 2011 and 2012, which correspond to rainy months and high temperatures, the highest richness values were recorded (Fig. 4). In November 2011, the highest richness value, 39 taxa, was recorded in S2 and the lowest values, in August 2012, in S2 and S3, eight and ten taxa, respectively.

Fig. 4
figure 4

Monthly variation of taxa richness in the sampling sites during the studied period (S1, S2, and S3)

Full size image

The total biovolume of the phytoplankton community did not supersede 2.7 mm3 L−1 and the highest values were recorded in February 2012 and June 2013 (Fig. 5). The increase of Chlorophyceae biovolume (January, February, and March of 2012 in the three sites) was determined by greater contribution of Coelastrum sp. (Figs. 5, 6). The increase of Cyanobacteria biomass (May, June, and July of 2013) occurred due to the dominance of M. aeruginosa (Kützing) Kützing and Dolichospermum circinalis (Rabenhorst ex Bornet & Flahault) Wacklin, Hoffmann & Komárek (Figs. 5, 6). M. aeruginosa dominance was also recorded in November 2012 in S3 (Fig. 6c).

Fig. 5
figure 5

Monthly variation of total biovolume (mm3 L−1) of phytoplankton classes in the three sampling sites (S1 = site 1, S2 = site 2, and S3 = site 3)

Full size image
Fig. 6
figure 6figure 6

Temporal variation of descriptor species in the three sampling sites: S1 (a), S2 (b), and S3 (c). Scales with different magnitudes

Full size image

Total biovolume was mostly composed of Cryptophyceae due to high density of Rhodomonas lacustris Pascher & Ruttner and Cryptomonas brasiliensis Castro, Bicudo & Bicudo in nearly all the studied period (Figs. 5, 6). Chlorophyceae contributed the most in hotter months (November, December 2011 and 2012). Cyanobacteria contributed to most of the biovolume, especially in the last three months of the studied period. However, Cyanobacteria were also also significantly found in November 2012 in the S3 station (Fig. 5).

Among the identified taxa, ten species were classified as descriptors, contributing to 88 % of the total biovolume: Coenocystis quadriguloides Fott, Coelastrum sp., Thorakochloris nygaardii Fott, C. brasiliensis Castro, Bicudo & Bicudo, C. curvata Ehrenberg, R. lacustris Pascher & Ruttner, Anathece sp., D. circinalis (Rabenhorst ex Bornet & Flahault) Wacklin, Hoffmann & Komárek, M. aeruginosa (Kützing) Kützing, and Ceratium furcoides (Levander) Langhans.

The canonical correspondence analysis (CCA), using six descriptor species of the system and six environmental variables, explained 47 % of the data joint variability in the first two components, presenting statistically significant (P = 0.001) Eigenvalue for axis 1 (λ1 = 0.662) and axis 2 (λ2 = 0.251) according to Monte Carlo test. The species–environment correlation was high and significant for both axes of the CCA (P = 0.001). The intra-set correlation and the canonical coefficient indicated that the temperature (Temp) (r = 0.8) and ammonium (NH4) (r = 0.5) were the most important variables in axis 1 ordination, whereas conductivity (Cond) (r = 0.9) stood out in axis 2 ordination (Fig. 7). Sample units from Phase 2 and Phase 1 were ordinated on the positive side of axis 1, associated with higher temperatures and more abundance of Coelastrum sp. (Coela, r = 0.6) and T. nygaardii (Thora, r = 0.5). Sampling units related to 2013 were ordinated on the negative side of axis 1, associated with higher concentrations of ammonium and total phosphorus and larger biomass of M. aeruginosa (Micro, r = 0.4) and D. circinalis (Dolic, r = 0.3) (Fig. 7).

Fig. 7
figure 7

Ordination according to CCA (axes 1 and 2) of 72 sampling units generated from the six descriptor species in the system and six abiotic variables: temperature (Temp), electrical conductivity (Cond), transparency (Trans), total phosphorous (TP), ammonium (NH4), and total nitrogen (TN). Species: Rhodomonas lacustris (Rhodo), Cryptomonas brasiliensis (Crypto), Coelastrum sp. (Coela), Thorakochloris nygaardii (Thora), Microcystis aeruginosa (Micro), and Dolichospermum circinalis (Dolic). Sampling units are identified according to collection year as shown: months (Ja January, F February, Mr March, A April, Mi May, Jn June, Jl July, Ag August, S September, O October, N November, D December) and sampling stations (S1 = 1, S2 = 2, and S3 = 3)

Full size image

Discussion

Results of the analyzed abiotic variables suggest the influence of tilapia farming in net cages in relation to time and not in relation to the sampling sites.

The fish farming system with net cages caused temporal changes in the phytoplankton community structure, especially detected as of May 2013, as shown by PCA and CCA. The hydrodynamic characteristics in the farming area, such as short residence time (21.6 days) and water flow (172 m3 s−1) (David et al. 2015) in addition to farming management, probably interfered with the biota and chemical compound distribution, thus explaining the similarity among the sampling sites and attenuation of possible changes in the phytoplankton community structure.

Mallasen et al. (2012) also detected spatial similarities among the sampling sites in the area of fish farming with net cages and related them to the hydrodynamic characteristics of the arm. Borges et al. (2010) and Bartozek et al. (2014), while analyzing the influence of the activities of fish farming with net cages on the phytoplankton community, reported that the effects of the activities on the water physical and chemical variables depend mainly on the intensity of the farming system, of the depth of the environment, and of the retention time of water.

The high richness recorded during the study was probably favored by rainfall, corroborating some studies on tropical and subtropical reservoirs that showed strong influence of rain on the composition of phytoplankton species (Giani and Figueiredo 1999; Bittencourt-Oliveira 2002; Kaggwa et al. 2011). Bittencourt-Oliveira (2002) attributed the increase in phytoplankton richness in rainy seasons to the highest nutrient concentrations from terrestrial runoff.

In this study, the influence of rainfall on the increased phytoplankton community richness may be associated with detachment of periphytic algae growing on the net of the cages, caused by winds and stronger rains as well as more turbulence in the water. Tucci et al. (2004) described the occurrence of the same event in Salto Grande reservoir, which influenced phytoplankton richness.

The high richness recorded in Chlorophyceae (73) and Cyanobacteria (34) corroborate works that confirmed the contribution in the number of species of these groups in continental waters: Huszar et al. (2000), Tucci et al. (2004), Deng et al. (2007), Dellamano-Oliveira et al. (2008), Fonseca and Bicudo (2008), Romanov and Kirillov (2012), and Zhang et al. (2012), indicating, in general, that they are widely distributed especially in tropical reservoirs.

The contribution of the Cryptophyceae to biovolume alternated with Chlorophyceae and Cyanobacteria along the studied period also corroborate Klaveness (1988, 1989) and Bicudo et al. (2005, 2009), who stated that the increase of Cryptophyceae occurs when other algae groups decrease, characterizing them as “opportunistic,” and Borges et al. (2010), who recorded increased Cryptophyceae density when Cyanobacteria density decreased.

The high frequency of Cryptophyceae in the entire study period can be attributed to hydrodynamics in the area of Ponte Pensa arm in the reservoir (Mallasen et al. 2012) as well as light availability (5 m of water transparency) and lower nutrient availability. According to Reynolds et al. (2002), Cryptophyceae species are able to develop in countless types of habitat due to the high surface-to-volume (S/V) ratio and quick growth rate.

In periods of water column mixing with high turbulence, Cryptophyceae can reach their maximum population, as documented by Klaveness (1988) and also as observed in tropical reservoirs by Santos and Calijuri (1998), Silva et al. (2005), Train et al. (2005), Pivato et al. (2006), and Rodrigues et al. (2007). The predominance of Cryptophyceae at Ponte Pensa arm over virtually the entire period may be related to mixing and turbulence conditions as mentioned in the studies referred above. We also considered that the current should be a major force function, since the three sampling sites were similar in the dynamics of environmental variables (David et al. 2015).

According to Bicudo et al. (2009), one particularity of Cryptophyceae is their continuous presence in aquatic ecosystems, i.e., they occur along the entire year, but hardly present as the dominant species. Differently from what was discussed by Bicudo et al. (2009), our results confirm the dominance of Cryptophyceae. Rhodomonas lacustris was dominant in over 59 % of the analyzed samples. R. lacustris and C. brasiliensis are flagellated and nanoplanktonic (6–10 µm) that require high light incidence. Minute species are characterized by relatively easy dispersion, due to their size (v < 103 μm3), fast absorption and assimilation of nutrients, and high rates of reproduction (r 20 > 10 × 10−6 s−1), facilitated by the high sv −1 ratio (>0.5 μm−1) (Brasil and Huszar 2011). We have attributed the species dominance to these characteristics which allow these organisms to thrive in turbulent and oligotrophic environments with high incidence of light where concentrations of essential nutrients like nitrogen and phosphorus are low as in the studied environment (Lopes et al. 2005; Oliveira and Calheiros 2000).

The ecological importance of Cyanobacteria is related to their capacity to develop blooms of toxic species especially in tropical eutrophic environments (Huszar et al. 2000; Komárek et al. 2002; Oliver and Ganf 2000; Tucci and Sant’Anna 2003; Honda et al. 2006; Deng et al. 2007; Sant’Anna et al. 2008; Kosten et al. 2012). Despite high concentrations of nitrogen and phosphorus in the food, Cyanobacterial blooms were not recorded. However, in November of 2012 and over the last three months of the study period, higher biovolumes of Cyanobacteria were recorded, due to the dominance of M. aeruginosa and D. circinalis. Microcystis and Dolichospermum have the highest number of toxic species in Brazil. However, toxic events of these genera have a different distribution. Blooms of Microcystis toxic species occur both in tropical and subtropical regions, whereas data on Dolichospermum toxic bloom occurrence are restricted to subtropical regions. Therefore, Microcystis species seem to have greater tolerance to different climates and environmental conditions than do the species of Dolichospermum (Sant’Anna et al. 2008).

Despite being widely distributed in continental waters, Microcystis aeruginosa reaches high densities in eutrophic conditions, characterized by low chemical and thermal stability, low turbulence, and irradiance (Michard et al. 1996; Oliver and Ganf 2000), whereas D. circinalis can reach high densities both in stratification conditions in the water column (Reynolds et al. 2002; Westwood and Ganf 2004) and in the water column mixing (Becker et al. 2004; Bovo-Scomparin and Train 2008).

The studied environment does not present favorable trophic conditions for the ample development of both species. However, in the last six months of the study period, i.e., 1 year after the beginning of the tilapia farming, increased concentrations of ammonia and total phosphorous were recorded, conditions that might have been favorable to the increase of biomass of these species. This statement can be confirmed through CCA which indicates strong association of these species with increased concentrations of total phosphorus and ammonium ion.

Our results indicate that the activities related to tilapia farming using the net cage system promoted increased concentrations of nutrients in the water, especially ammonium ion and total phosphorus, and consequently changes in the structure of the phytoplankton community attested by the increase in the Cyanobacterial biomass in the last three months of the study period.

We have concluded that in time the structure and dynamics of the phytoplankton community were influenced by the production of tilapia in net cages. However, changes in sampling seasons were not recorded. Changes were observed in chemical variables such as increased concentrations of ammonia and total phosphorous. In time, changes in abiotic variables were reflected, although subtly, in the community structure and dynamics. One year after the net cages were installed, in 2013, the greatest biovolume of M. aeruginosa and D. circinalis was recorded compared with the other identified species, and these two cyanobacteria species were associated with higher concentrations of ammonia and total phosphorous. The hydrodynamic characteristics of the analyzed system, such as short residence time (21.6 days) and outflow (172 m3 s−1), probably facilitated the assimilation of disturbances in the water quality caused by the employed production process and, therefore, serve as determinant in attenuating the impact of the organic load (from feed and fish metabolism) in the studied aquatic system.