Antonio Cesr Gooy, Lus Ulisses Rovigtti Chivelli, Jrre Hugh Oxfor,Rˆomulo Btist Rorigues, Igor e Oliveir Ferreir, Arypes Suteri Mrones,Clui Aprei Honorto Silv, Dley Neu,＊＊

aUniversidade Estadual de Maring´a, Av. Colombo, 5790, Centro de Ciˆencias Exatas, Departamento de Química, Maring´a, PR, Brazil

bUniversity of Georgia, Poultry Science Department, Athens, GA, USA

cUniversidade Federal do Rio Grande do Sul PPG em Zootecnia, Av, Bento Gon[start]ç[end]alves, 7712, Porto Alegre, RS, Brazil

dUniversidade Federal da Grande Dourados, Faculdade de Ciˆencias Agr´arias, Rodovia Dourados - Itahum, km 12, Cidade Universit´aria, Dourados, MS, Brazil

Keywords:

ABSTRACT

1.Introduction

With the increasing world population, estimated at 8.24 billion people for the year 2030 (Taagepera, 2014), the supply of food will continue to be a challenge for the nations. In this sense, aquaculture can contribute to the supplementation of a good part of the protein consumption worldwide. In 2016 a total of 80 million tons of aquatic animals were produced, with 67% of those being fish (FAO, 2018). Fish farming has high developmental potential, contributing to the social and economic development of different regions of the world through producing protein sources of high nutritional value (Sabbag et al., 2018,pp. 307–315).

One of the most relevant and complex criteria of fish farming involves maintaining suitable water quality for successful aquaculture farming. As the quality of the water becomes a limiting factor, continuous recycling of the water is used to reduce the metabolic and food residues of animals, by doing this, the water flow becomes a means of diluting chemicals and biological components present in the water tank system (Pereira & Mercante, 2018).

Aquatic environments are dynamic and can undergo great variations in their physical and chemical characteristics throughout the year,seasons, production cycles, or even during intervals of one day (Wetzel,2001). Knowledge about water quality variables, such as: dissolved oxygen, pH, alkalinity, transparency, temperature, nitrogen compounds,and phosphorus in water can contribute to decision making (Neu et al.,2014) when these values are outside the appropriate range for the fish(Cs´abr´agi et al., 2019; Godoy et al., 2018).

Spatial variations can cause significant heterogeneity in the distribution of nutrients and metabolites in water bodies, caused mainly by water column density gradient (Ramírez & Bicudo, 2005). The epilimnion nutrient pool comes from the balance between losses due to sedimentation and increased nutrient flows from the hypolimnion. Thus, the nictemeral variation can be greater than changes that occur on an annual cycle (Diemer et al., 2010; Godoy et al., 2018; Neu et al., 2014).

The fish farming environment is highly complex in relation to chemical composition, physical parameters, nutrients and biota and there is an interdependence in relation to these parameters. Therefore,the process of qualitative and quantitative assessment of this complex environment is generally challenging when all environmental variables are considered simultaneously (Olsen et al., 2012).

Response surface methodology (RSM) is widely used in the optimization of chemical experiments (Chiavelli et al., 2019; Sharifi et al.,2018), food improvement (Malekjani & Jafari, 2020; Yolmeh & Jafari,2017), and water experiments (Beuckels et al., 2015; Keshtegar &Heddam, 2017; Okoli & Ofomaja, 2019). RSM is a mathematical model that evaluates the influence of several individual factors and their interactions at the same time. This technique is used to design experiments and can help reduce the number of required experiments (Bezerra et al.,2008; Myers et al., 2011). RSM can be used to evaluate interactions between several intrinsic factors of the aquatic environment in limnologic studies such is in the present work.

In view of the many characteristics that farming activity involves sustainability that must be met, management practices must be adopted in order to ensure a comfortable environment for the animals to grow. In this sense, the objective of the present work was to evaluate the nictemeral and vertical dynamics of limnological characteristics in a fish production pond using: (a) ANOVA and Pearson?s correlation; (b)analysis of the principal components (PCA); and (c) response surface methodology (RSM) to better understand the behavior of the limnological variables in fish ponds.

2.Materials and methods

2.1.Water collection

The study was conducted in the municipality of Dourados-MS, Brazil,at coordinates 2213’

.

57

.

3

S

5459’

.

17

.

8

W

(Fig.1), in a region with a tropical climate that is dry in the winter and humid in the summer. The experiment took place in a tank lined with geomembrane with dimensions of 20 m long x 9.75 m wide x 0.80 m deep, totaling 156 m,with no renewal or water circulation, and only with replacement of evaporation losses made once a week. The stocking density was 1.53 fish per m(totaling 239 tilapia of ≈108 g per fish) that were fed according to the proposal by Boranga et al. (2018).

The experiment setup followed Diemer et al. (2010); Neu et al.(2014) with adaptions for Geo-membrane pond conditions. Water samples were collected in days without rain in November 2019, totaling 18 collections. In order to verify the nictemeral dynamics, eight samples were taken during a 24-h period in each day of collection, with a 3-h interval between samples, totaling eight different time points per day:09:00 a.m., 12:00 p.m., 3:00 p.m., 6:00 p.m., 9:00 p.m., 00:00 a.m., 3:00 a.m. and 06:00 a.m. hours. Response surface methodology (RSM) was followed by water collection trough procedure setup in section 2.4.1, the same spots were used to verify the nictemeral dynamics.

To determine the vertical dynamics of the water column, the water samples were taken at 9:00 a.m. on the day of collection - at three depths: 5 cm (Epilimnion), 35 cm (Metalimnion), and 70 cm (Hypolimnion), in the central region of the pond (please see Fig.1), with the help of hoses and dark bottles to preserve the characteristics of the water.

2.2.Water analysis

In all water collections, the following parameters were analyzed:water temperature (C), dissolved oxygen (mg L), pH, alkalinity (mg L), electrical conductivity (

μ

S cm), and total dissolved solids (ppm)were evaluated

in situ

using a Hanna digital potentiometer. For the analysis of ammonia (mg L), nitrite (mg L), orthophosphate (mg L), and total phosphorus (mg L) the samples were preserved in dark bottles and refrigerated (5C) till further analysis. The analyzes were performed using the techniques suggested by APHA (2017), and the results were read on a UV–Vis Alfakit AT100 spectrophotometer.

2.3.Statistical analysis

The data obtained were grouped by depths and times and submitted to analysis of variance and according to the normality of Shapiro-Wilk test (Shapiro & Wilk, 1965), when significant differences were observed, Tukey’ test was applied at a significance level of 5% (Tukey,1949, pp. 99–114). Pearson’s correlation analysis was used to verify correlations between the selected abiotic variables (Pearson, 1895)which established that coefficients, positive or negative, ranging from 1 to 0.7 are considered strong, 0.7 to 0.3 are considered moderate, and 0.3 to 0.0 are considered weak.

Fig.1.Location of the geo-membrane pond site in the Mato Grosso do Sul state, Brazil.

2.4.Principal components analysis

The most applied methodology in environmental studies is the principal component analysis (PCA), which performs an exploration of the correlation structure between the constituent variables of the database. This process produces a smaller number of more important variables that reflect the original set providing a database useful to assessing the studied environment (Olsen et al., 2012).

The vertical variations were tabulated, and the data set was used for principal analysis components (PCA), which decomposes the original and allows for the reduction of the dimensionality of the data by performing a linear transformation and rejecting a part of the components with smallest variations.

The use of the PCA was to provide an overview of the relationship between the stratification of the cultivation tank and also to evaluate the influence of limnological assessments on the stratification of the reservoir in the discrimination. A 3 ×11 data matrix was used with the three stratifications investigated being: epilimnion, metalimnion and hypolimnion in the rows and the columns being: temperature, depth, dissolved oxygen, pH, alkalinity, conductivity, solids, ammonia, nitrite,orthophosphate, and phosphate.

The original parameters and the PCs are correlated by the loadings factor, explaining the weights of the PCs in the original parameters(Tabachnick & Fidell, 2013). For this PCA analysis, only PCs with eigenvalues greater than 1 (Kaiser, 1960) were taken into account. For statistical analysis, the statistical program R was used (R Core Team,2019).

2.4.1.Surface response methodology

Different depth levels and times from CCD were analyzed using the response surface analysis to visualize the interactive effects (Montgomery, 2017). Limnological variables were modeled following polynomial regression. The model coefficients describe the estimated effects of their respective parameters on the variable response in the model and a generic expression can be written as follows:

where Y is the response and the predicted dependent parameters of the limnological variables,

b

,b

,

b

,b

,b

,b

,b

,b

, and

b

represent the linear and constant effect of time (A), the linear effect of depth (B),active effect of depth and hours, neffect of depth and meffect of hours, respectively (Montgomery, 2017). The optimization analysis for the predicted responses was performed using R software (R Core Team,

Table 1 Independent variables evaluated in the limnological dynamics in Nile tilapia farming tank and the respective coded and real values for the central composite design.

＊ Levels of variables: factorial points (±1), central points (0), axial points (±

α

).2019) with the package described by Lenth (2009). The significant difference was established at p

<

0

.

05, and the model with a statistical difference was chosen.

3.Results and discussions

3.1.General overview of the response datasets

In order to have an overview of the parameters of the responses presented, a descriptive statistic must be presented (Hatvani et al.,2015), as shown in Table 2. In the data overview, alkalinity, ammonia,conductivity, and total solids did not vary to a high degree between the all-day, epilimnion, metalimnion, and hypolimnion groups. On the other hand, temperature, oxygen, pH, nitrite, phosphate, and orthophosphate had considerable variation between the groups of samples through-out the day and the epilimnion, metalimnion, and hypolimnion.

In general, the variability, based on the variation coefficients (CV;Table 2), of the water quality variables observed are low in responses that are not so strongly linked to the presence of animals, such as temperature, alkalinity, and total solids. Further details and justifications will be shown in the tests of averages of the samples of the all-day data and vertical dynamics (sections 3.2, 3.3, and 3.4).

Variations in water quality parameters throughout the day, or between water layers are demonstrated in the literature (Karakassis et al.,2001; Neori et al., 1989). In shallow tanks, changes during the 24-h interval are quite evident, as demonstrated in the current study, a fact reported by de Assis Esteves (2011) for the concentration of dissolved oxygen in lakes. Although there is legislation in Brazil (Brasil, 2005)with limitations for the maximum limits of some variables, taking into account mainly environmental issues, there are countless studies that describe the tolerance and the ideal breeding range of fish species for the variables limnological studies (Bhatnagar & Devi, 2013; Mercante et al.,2018). Egna and Boyd (2017) compiled a series of information that is available in the work “Dynamics of pond aquaculture”, where they report all the changes that may be in the breeding environment, as well as the confidence intervals of some abiotic parameters.

3.2.Mean test of the response datasets

The means comparison test is used to assess the discrimination between established treatments. The mean values of monitoring the collected abiotic variables are shown in Table 3. Significant differences(p

<

0

.

05) were observed when comparing different hours of the day for the variables: temperature, dissolved oxygen, pH, alkalinity, ammonia,electric conductivity, nitrite, orthophosphate, and total dissolved solids.On the other hand, there were no significant differences observed (p

>

0

.

05) for the parameter phosphorus based on collection time.

The water temperature varied between 22.40 and 26.03C between the hours of 9:00 a.m. and 3:00 p.m., respectively. The recommended values for fish farming are between 20 and 29C, showing that the tolerable limits for the cultivation of tropical fish was maintained (Boyd& Tucker, 2015).

Dissolved oxygen ranged from 3.86 to 18.38 mg L, with the lowest record being at 6:00 a.m., just below the minimum recommended range for tilapia, 4 mg L(Boyd & Tucker, 2015; Sousa et al., 2018). Similar behavior was observed by Diemer et al. (2010) in an environment of farming native fish in net tanks expressing a drop in concentration at night. This observation may indicate that during this period, fish and oxygen producing organisms consume dissolved oxygen (Diemer et al.,2010; Green & McEntire, 2017; Sousa et al., 2018). The amount of oxygen dissolved in the water is influenced by the temperature (Mata et al.,2018). As the water temperature increases, the oxygen level decreases.The temperature is directly linked to the osmotic process of the fish, in addition, the physiological activities such as breathing, digestion,excretion, feeding, and movement contribute to higher oxygen consumption. However, oxygen levels can increase during the day due to

light-dependent photosynthetic processes, which is observed between 12:00 p.m. and 6:00 p.m., where the highest levels of O2 was observed.As a consequence, during the night, O2 levels decrease due to biological respiration and the biological decomposition processes of the sediment(Straˇskrabra & Tundisi, 2013). It is common to experience a continuous variation of oxygen during the (Neu et al., 2014), similarly to what occurred in the tank environment of the present study.

Table 2 Descriptive statistics of water quality variables for each of the homogeneous groups on the tank.

Table 3 Values of abiotic parameters during nictemeral analysis, collected in situ.

The pH of the water is a factor of intensity of acidity or basicity (Boyd et al., 2011). During the 24 h period pH fluctuated from 8.00 to 9.65.According to Bhatnagar and Devi (2013); Leira et al. (2017), the comfortable range for fish production is between 6.5 and 9.0. In the present work, the values were slightly above the tolerable ranges for the cultivation of fish. Due to the photosynthetic and respiratory activity of aquatic communities, this parameter may change during the day (Gao et al., 2016). In a fish farming environment, there are five factors that can cause changes in pH, namely respiration, photosynthesis, fertilization, liming, and pollution (Leira et al., 2017). The lowest pH values were found in the early hours of the morning, a fact that may be linked to a greater amount of carbon dioxide (CO). This observation may be attributed to the animals physiological processes, the presence of algae in the water, and even the low level of dissolved oxygen. The present experiment did not present acidic pH and apparently did not present toxicity to the fish since mortality was not observed, which may be related to the low concentration of ammonia in the pond.

When the ammonia parameter was evaluated (Table 3), it was found that it varied from 0.25 to 0.38 mg L, showing a significant difference between the hours of the day (p

<

0

.

05). This nitrogenous residue comes from the catabolism of proteins, resulting from the decomposition of organic matter, feed debris, and fish excreta (Mata et al., 2018). The excretion of this residue by fish is derived from the digestion of proteins and the catabolism of amino acids, reaching the gills through the circulatory system, and released into water. The values were within the appropriate range for fish farming, below 0.60 mg L(Soderberg,2017). Constant checking of ammonia levels is necessary, as concentrations above the tolerated level may express less growth of the animals(Leira et al., 2017) can lead to damage to the animals’ immune system,generating cases of disease or mortality. Improper handling and a high concentration of ammonia can lead to the accelerated growth of phytoplankton that assimilates these elements, resulting in algal blooms,which cause damage to water quality.In the present work, the nitrite values obtained varied from 0.03 to 0.06 mg L, thus, the nitrite present in the water was within the appropriate range of up to 0.5 mg Lfor fish (Leira et al., 2017),however significant differences between the sampled times were observed (p

<

0

.

05). Nitrite is the intermediate compound in the bacterial nitrification of ammonia to nitrate (Cyrino et al., 2010) and can be toxic to fish. Nitrite levels can also be related the levels of Oand ammonia concentrations in the water (Mercante et al., 2018).Phosphorus levels showed no significant difference during the hours(p

>

0

.

05) and ranged from 0.006 to 0.052 mg L. Phosphorus in high concentrations can be problematic, this accumulation of nutrients in the water together with other factors such as high nitrogen accumulation leads to the phenomenon of eutrophication (Garcia et al., 2009). Even though phosphorus has low toxicity, the increase in the concentrations in the water favors the excessive flowering of algae, consequently causing degeneration of the water body (Garcia et al., 2009; Leira et al.,2017). High concentrations of phosphorus combined with a high concentration of Oin times of more light, obtained in the present work,evidenced the intense supply of organic matter in the system. Besides, in the aquatic environment, phosphorus occurs in the forms of total phosphorus, particulate phosphate, dissolved phosphate, and orthophosphate which is the form available to vegetables. Orthophosphate is also important for the development of phytoplankton, a nutrient necessary for natural fish farming (Pilarski et al., 2004). In this work, the orthophosphate values ranged from 0.02 to 0.16 mg L, which may be related to the high concentration of organic matter in the pond.The total dissolved solids were below the values allowed by Brazilian environmental legislation (500 ppm), ranging from 72.00 to 84.33 ppm.Electrical conductivity is directly linked to the amount of dissolved solids, and in this experiment presented values from 144.66 to 178.00

μ

S cm. Electrical conductivity indicates the presence of dissolved ions,such as nitrites, bicarbonates, carbonates, phosphates, and ammonia,reflecting on the water’s capacity to conduct electricity (Leira et al.,2017). It was observed that at 12 p.m. and 3:00 p.m., when the highest concentrations of dissolved oxygen were observed, the lowest values for electrical conductivity were also observed. These results may demonstrate the phytoplankton’s capacity to absorb nutrients like nitrogen and phosphorus available in the tank (Pilarski et al., 2004). It was observed in this work that in some moments, the electrical conductivity was above the recommended range of 20–150

μ

S cm(Soderberg, 2017). When the values of electrical conductivity are high, they indicate a high degree of decomposition and it becomes a parameter that assists in the assessment of nutrient availability in the aquatic environment (Soderberg,2017).

Phosphorus concentrations increased in conjunction with increased O(Table 4), during daylight hours, may indicate that there was an increase in photosynthetic activity by the algae in the fish tanks of this study. It can be seen that oxygen had a moderate negative correlationwith temperature and depth.

Table 4 Pearson’s correlation values between the variables analyzed.

Table 5 Factors, levels, and limnological variables for each procedure of the experimental design.

The negative correlation between dissolved oxygen and temperature is due to the difficulty in dissolving oxygen at elevated temperatures.Variables can affect dissolved oxygen through oxidation, photosynthesis, respiration, etc. When the negative correlation with depth is considered, it may be due to the oxygen consumption by the heterotrophic organisms present in the aquatic environment (Li et al., 2020;Oberle et al., 2019).

The positive correlation between phosphorus and orthophosphate indicates that the higher the concentrations of phosphorus, the greater the content of orthophosphate in the medium. The levels of orthophosphate in ponds are mainly derived from large amounts of organic and inorganic compounds, such as manure, fertilizers, and feed (Hatvani et al., 2015). This variable, together with the levels of nitrogen and phosphorus, favors the growth of algae which can increase the pH of the pond. Increases in temperature as in the case of the present study, lead to an increase in ammonia levels in the tank (Mercante et al., 2018), thus decreasing the levels of orthophosphate in the aquatic environment.

Water quality analysis should always be evaluated in fish farms, as it can help with either the success or failure of production. Due to the ever changing environment of aquaculture tanks, it is important to be able to assess water quality of the tanks at multiple depth, as well as at multiple times through out the day to ensure optimal production.

3.3.Response surface analysis methodology

The effect of the independent variables: time of collection and depth on limnological were evaluated through two central composite designs(CCDs) in conjunction with response surface methodology (RSM). The larger results for oxygen, ammonia, and phosphorus were achieved with experiments 11, 5–3, and 9 respectively (Table 5).

The response surface methodology was used to evaluate the individual effects and the interactions between the variables; time of sample collection and tank depth on the studied responses. The plot of residuals forms approximately a straight line (Please see Figs. S1A, S1B, and S1C)indicating that it does not violate the normality assumption of ANOVA.The ANOVA parameters for the models are presented in the supplementary material (Tables S1, S2, and S3). According to the ANOVA data,the parameter models that had no significance (p

>

0

.

05) were: temperature, pH, alkalinity, electric conductivity, nitrite, orthophosphate,and total dissolved solids. On the other hand, the parameters models that showed significance (p

<

0

.

05) were: dissolved oxygen, ammonia,and phosphorus. The empirical relationship between the independent variables and the responses obtained by the application of the RSM is given by the quadratic functions of the respective models, according to equations (2)–(4).

The predicted models for dissolved oxygen, ammonia, and phosphorus can be viewed through response surfaces (Fig.2) these samples show how depth affects the concentrations of dissolved oxygen,ammonia, and phosphorus in the water. The coefficient of determination(R) of the models were 0.98, 0.80, and 0.90 for oxygen, ammonia, and phosphorus respectively. This implies that 98, 80, and 90% of variation of the variable evaluated are explained by the independent variables. In addition, the Radjusted and Rpredicted should be ≈0

.

20 of each other to be in reasonable agreement, such as verified (0.97, 0.66, and 0.83 for oxygen, ammonia, and phosphorus respectively).

Fig.2.Three-dimensional response surface plot for: dissolved oxygen (A), ammonia (B), and phosphorus (C) as response of interaction between independent factors depth and time.

Fig.3.Actual and predicted value plot of (A) dissolved oxygen (mg L-1), (B) ammonia (mg L-1), and (C) phosphorous (mg L-1) plots.

A response surface methodolgy (RSM) is considered accurate for each of the response variables when there is similarity between the experimental and predicted data (Mourabet et al., 2017). The predicted data were consistent with the experimental data (Fig.3), which indicated a good fit between predicted values and the experimental data points. In our case, actual ×predicted presented

R

=0

.

98 and

R

=0

.

89for dissolved oxygen and phosphorus. For ammonia present

R

=0

.

70. Considering the complexity and dynamics of the fish farming pond, these results are considered good predictions of the behavior of the variables with the independent variables.

The quadratic effects of time and depth establish the functional relationship between the variables dissolved oxygen, ammonia, and phosphorus and can be used to predict the concentrations of these factors in the environment. The increase in depth resulted mainly in a significant decrease in the dissolved oxygen content (Fig.2A). The lower depths of the tank, have less availability to light which may lead to reduces photosynthetic production of oxygen. In addition, oxygen is consumed by the aerobic microbiota present in the bottom of the tanks Oberle et al. (2019) which can lead to a decrease in oxygen concentration at night, a fact that was evidenced in this study.

The fish tank environment is dynamic and contains both aerobic and anaerobic microorganisms which can directly influence water quality parameters, such as pH and ammonia content (Moriarty, 1997). The increase in depth resulted in an increase in the concentration of ammonia and phosphorus, that is released when micro-organisms consumer fish feces and excess feed found at the bottom of the tanks. In addition, oxygen is consumed by the aerobic microbiota present at the bottom of the tank, which leads to decreases in oxygen concentration at night, as observed in this study (Chang & Ouyang, 1988; Ghosh &Tiwari, 2008).

3.4.Vertical dynamics

Due to the physical and chemical stratification of fish ponds, water circulation can release harmful minerals and gases that have settled on the bottom of the ponds; this distribution of gases and minerals can lead to productions losses or mortality (Gunkel, 2003). Therefore, knowledge of the vertical profile of the physical and chemical parameters in the rearing tank is essential (Diemer et al., 2010).

Aquatic environments are dynamic, suffering variations in their physical and chemical characteristics that influence water quality(Diemer et al., 2010). The water column in this environment is divided into three layers, usually well defined: epilimnion, which corresponds to the superficial layer; metalimnion, an intermediate layer; and hypolimnion, the deepest layer (Godoy et al., 2018).

Table 6 shows the vertical dynamics of the limnological characteristics of the collected variables

in situ

. Temperature, dissolved oxygen,and pH expressed close values in the different layers, this behavior was also reported by Diemer et al. (2010) and Godoy et al. (2018), in the Itaipu-Rio Paran´a and Salto Caxias-Rio Iguau Reservoirs. For the present study, this behavior may be related to the depth of the pond, which is 75 cm, and, therefore, such variations occur with greater intensity.

The layer with the highest temperatures and dissolved oxygen was epilimnion, 24.82C and 14.25 mg Lrespectively, most likely explained by its proximity to the surface. The water in the epilimnion is heated by solar radiation and facilitating oxygen production by algae and atmospheric dissolution occurs. Similar results were obtained by Oberle et al. (2019) in an experiment on daytime oxygen stratification in shallow tanks. The variables with the highest eutrophic capacities, such as ammonia, nitrite, and phosphorus had higher values in the hypolimnion, which may be due to the different biochemical reactionspresent in the medium, such as the degradation of excreta and animal waste (Li et al., 2020).

Table 6 Values determined in the laboratory at different depths in fish farming tank.

Fig.4.A) Biplots of the 1st and 2 nd PCs, where:dots represent epilimnion;dots represent hypolimnion; anddots represent metalimnion observations in pond;rings correspond to the 65% confidence ellipses estimated using the mean vertical dynamics of pond. The variance explained for each PC is showed between parentheses. B) Loadings of the assessed water quality variables in the first two principal components, only ±≥0.50 loading are shown.

The highest concentration of nitrite was found in the hypolimnion (p

<

0

.

01), even though it was not considered toxic to fish, nitrite had a higher value detected in the deepest region of the tank. This fact may be related to the nitrification processes that consist of the oxidation of ammonium ions (NH) to nitrite through Nitrosomonas bacteria. This process is aerobic and is dependent on the amount of available oxygen may be more intense in the hypolimnion due to the amount of material present at the bottom of the tank. Denitrification consists of reducing nitrate to molecular nitrogen, this process occurs mostly in anaerobic conditions and where there is a large amount of organic substrate.During times of an anaerobic conditions, usually observed in the hypolimnion, ammonia concentrations can reach high levels (Pereira &Mercante, 2018). As the pond showed good oxygenation as a whole, an intense nitrification process began, which resulted in the consumption of a large part of the accumulated ammonia found in the deepest part of the tank. The values found for nitrite suggest that it does not interfere with the cultivation of fish, with no adverse effects on mortality or feeding.

3.4.1.Principal component analysis (PCA)

The PCA breaks the data down into separate sets of scores and loads for stratification and variables. The most significant principal component (PC1 and PC2 - Fig.4A) and their loadings (loading values ≥0.50 -Fig.4B) generated from selected values from experiment were considered for the PCA analysis. Two principal components extracted in the current study explained 44.60% of the total variability in the dataset.

In the tank strata, it was clear that temperature, dissolved O, and pH, are responsible for the greatest variations in the data set (Fig.4B).The values of eigenvalues are the factors by which eigenvectors (PCs)are scaled (Jain & Shandliya, 2013). In the present study, the eigenvalues were 2.64 and 1.82 for PC1 and PC2 respectively, which were greater than 1, according to the Kaiser (1960) rules. The PCA analysis showed small discrimination of the water samples between epilimnion,and metalimnion (Fig.4A). It was observed that the hypolimnion varied from the other two groups. The differences between stratification in the respective farming tank may be due to the chemical composition as verified by Table 6.

In the present study, we can observe the importance of using different water quality variables to assess limnological dynamics in fish farming tanks in conjunction with statistical tools to seek a better understanding of the observed responses and their interrelationships. The use of different statistical tools, especially the analysis of surface responses can enrich the evaluation of the data collected in the tank and thus favor aiding in decision making to mitigate any quality-related water problems in the rearing tank.

4.Conclusion

It was possible to verify that changes occur in water temperature,dissolved oxygen, pH, alkalinity, electrical conductivity, total solids,nitrite, and orthophosphate during a 24-h period. Differences in nitrite and phosphorus were more pronounced at the bottom of the rearing tank. In addition, when we observe the limnological dynamics in the tilapia tank, we observed that the interaction between depth and time directly influences the levels of dissolved oxygen, ammonia, and total phosphorus distribution in the water.

Authorship contributions

Category 1

Conception and design of study:

Antonio Cesar Godoy

Igor de Oliveira Ferreira

Dacley Ney

Writting manuscritp:

Antonio Cesar Godoy

Lucas Ulisses Rovigatti Chiavelli

Jarred Hugh Oxford

Rˆomulo Batista Rodrigues

Igor de Oliveira Ferreira

Arypes Scuteri Marcondes

Claucia Aparecida Honorato da Silva

Dacley Ney

Category 2

Drafting the manuscript:

Antonio Cesar Godoy

Revising the manuscript critically for important intellectual content:

Antonio Cesar Godoy

Lucas Ulisses Rovigatti Chiavelli

Jarred Hugh Oxford

Rˆomulo Batista Rodrigues

Igor de Oliveira Ferreira

Arypes Scuteri Marcondes

Claucia Aparecida Honorato da Silva

Dacley Ney

Category 3

Approval of the version of the manuscript to be published:

Antonio Cesar Godoy

Lucas Ulisses Rovigatti Chiavelli

Jarred Hugh Oxford

Rˆomulo Batista Rodrigues

Igor de Oliveira Ferreira

Arypes Scuteri Marcondes

Claucia Aparecida Honorato da Silva

Dacley Ney

Acknowledgment

The authors state that there is no any conflict of interest. We would like to thank CAPES - Coordenaç˜ao de Aperfeiçoamento de Pessoal de Nível Superior for scholarship.

Appendix A.Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.aaf.2020.08.005.