Combination of Coagulation–Flocculation–Decantation and Ozonation Processes for Winery Wastewater Treatment

This research assessed a novel treatment process of winery wastewater, through the application of a chemical-based process aiming to decrease the high organic carbon content, which represents a difficulty for wastewater treatment plants and a public health problem. Firstly, a coagulation–flocculation–decantation process (CFD process) was optimized by a simplex lattice design. Afterwards, the efficiency of a UV-C/ferrous iron/ozone system was assessed for organic carbon removal in winery wastewater. This system was applied alone and in combination with the CFD process (as a pre- and post-treatment). The coagulation–flocculation–decantation process, with a mixture of 0.48 g/L potassium caseinate and 0.52 g/L bentonite at pH 4.0, achieved 98.3, 97.6, and 87.8% removals of turbidity, total suspended solids, and total polyphenols, respectively. For the ozonation process, the required pH and ferrous iron concentration (Fe2+) were crucial variables in treatment optimization. With the application of the best operational conditions (pH = 4.0, [Fe2+] = 1.0 mM), the UV-C/ferrous iron/ozone system achieved 63.2% total organic carbon (TOC) removal and an energy consumption of 1843 kWh∙m−3∙order−1. The combination of CFD and ozonation processes increased the TOC removal to 66.1 and 65.5%, respectively, for the ozone/ferrous iron/UV-C/CFD and CFD/ozone/ferrous iron/UV-C systems. In addition, the germination index of several seeds was assessed and excellent values (>80%) were observed, which revealed the reduction in phytotoxicity. In conclusion, the combination of CFD and UV-C/ferrous iron/ozone processes is efficient for WW treatment.


Introduction
Portugal is a Mediterranean wine producer, with an approximated vineyard area of 191,000 ha and a wine production value of 6.4 MhL, in 2020 [1]. This extensive wine production requires large amounts of water, to perform several activities that are necessary to ensure the quality of the wines, such as the floor and equipment washing, rinsing of the transfer lines, barrel cleaning, bottling facilities, filtration units, etc. [2][3][4]. The high consumption of water inevitably leads to the generation of large amounts of winery wastewaters (WW). The environmental impact of wastewater from the wine industry is highly detrimental if it is released without proper treatment, causing water pollution, soil degradation, and damage to the vegetation by odors and gaseous emissions [2,5]. Currently, the most widespread decontamination approaches are activated sludge reactors [6]; however, the seasonal character of these wastewaters makes it difficult for the microorganisms to adapt. In addition, during the vintage stage, the pollutant loads and wastewater volumes that are produced are higher, requiring longer retention times in the activated sludge reactors, which makes these biologic reactors oversized for most of the year [7].  Phytotoxicity tests were performed by germination of onion, cucumber, lettuce and corn seeds (standard species recommended by the US Environmental Protection Agency, the US Food and Drug Administration, and the Organization for Economic Cooperation and Development [32]) and determined by Equation (1) in accordance to Varnero et al. [33] and Tiquia et Tam [34], as follows: where GI is the germination index, N SG,T is the arithmetic mean of the number of germinated seeds in each extract (wastewater), N SG,B is the arithmetic mean of the number of germinated seeds on standard solution (distilled water), L R,T is the mean root length in each extract (wastewater) and L R,B is the mean root length in control (distilled water). Let Xi be defined as the removal of a given indicator (%) of water contamination (turbidity, TSS, TOC, COD and total polyphenols) that is achieved by a treatment (Equation (2)) [12], as follows: X i (%) = C 0 −C f C 0 * 100 (2)  where C 0 and C f are the initial and final concentrations, respectively, of parameter i.

Phenolic and Chromatic (CIELab) Characterization
Color intensity (CI) and hue were determined by the OIV method [35], total polyphenol index (TPI) was determined by Curvelo-Garcia method [36], total phenols, nonflavonoids and flavonoids were determined according to Kramling and Singleton [37]. Total anthocyanins were analyzed by SO 2 bleaching method, described by Ribéreau-Gayon et al., [38], colored anthocyanins (CA), total pigments (TP) and polymeric pigments (PP) were analyzed by the method described by Somers and Evans [39], and total tannins were determined by the LA method [40]. All samples were analyzed by a spectrophotometer (GENESYS TM 10 series spectrophotometers). The absorption spectra of WW samples were recorded with a Shimadzu UV-2101 spectrophotometer (Shimadzu, Kyoto, Japan) scanned from a range between 380 and 770 nm, with 5 nm distance, using 1 cm path length quartz cells. Data were collected to determine L (lightness), a (redness), and b (yellowness) coordinates using the CIELab 1976 method. This allows reliable quantification of the overall color difference of a sample when compared to a reference sample (blank). Color differences can be distinguished by the human eye when the differences between ∆E * ab values are greater than two units, in accordance to Spagna et al. [41]. All analyses were performed in duplicate. Table 2 resumes the formulas used in this work. Table 2. Formulas for phenolic composition and chromatic (CIELab) determination.

Coagulants Characterization
The FTIR spectra were obtained by mixing 2 mg of coagulants with 200 mg KBr. The powder mixtures were then inserted into molds and pressed at 10 ton/cm 2 to obtain the transparent pellets. The samples were analyzed with a Bruker Tensor 27 spectrometer and the infrared spectra in transmission mode was recorded in the 4000-400 cm −1 frequency region. The microstructural characterization was carried out with a scanning electron microscop (FEI QUANTA 400 SEM/ESEM, Fei Quanta, Hillsboro, WA, USA) and the chemical composition of the different catalysts was estimated (Table 3) using the energydispersive X-ray spectroscopy (EDS/EDAX, PAN'alytical X'Pert PRO, Davis, CA, USA). The textural parameters of samples were obtained from N 2 adsorption-desorption isotherms at 77 K using a Micromeritics ASAP 2020 apparatus (TriStar II Plus, Micromeritics Instrument Corporation, Norcross, GA, USA). The samples were degassed at 150 • C up to 10 −4 Torr before analysis. The specific surface area (SBET) was determined by applying the Gurevitsch's rule at a relative pressure p/p 0 = 0.30 and according to the Brunauer, Emmet, Teller (BET) method from the linear part of the nitrogen adsorption isotherms. Different pore volumes were determined by the Barrett, Joyner, Halenda model (BJH model).
The results obtained by BET analysis (Table 4) showed that bentonite exhibited a mesoporous structure with a specific surface area of 8.8 m 2 /g, a total pore volume of 0.045 cm 3 /g and a particle size of 4.0 nm. The respective isotherms can be classified as type II, where unrestricted monolayer-multilayer adsorption occurs, and the behavior of the hysteresis loops can be associated with type H3, which usually corresponds to aggregates of plate-like particles forming slit-like pores [56], which is in agreement with these material structures. Potassium caseinate had a specific surface area of 1.0 m 2 /g, PVPP's surface area was not quantifiable. Total pore volume and particle size were not quantifiable for potassium caseinate and PVPP, which indicates that these coagulants have a microporous structure. The shape of its N 2 adsorption-desorption isotherm was a type I isotherm, typical of microporous solids with relatively small external surfaces, as defined by the International Union of Pure and Applied Chemistry (IUPAC) [57]. The results obtained by BET analysis (Table 4) showed that bentonite exhi mesoporous structure with a specific surface area of 8.8 m 2 /g, a total pore volume o cm 3 /g and a particle size of 4.0 nm. The respective isotherms can be classified as t where unrestricted monolayer-multilayer adsorption occurs, and the behavior of th teresis loops can be associated with type H3, which usually corresponds to aggreg plate-like particles forming slit-like pores [56], which is in agreement with these m structures. Potassium caseinate had a specific surface area of 1.0 m 2 /g, PVPP's surfa was not quantifiable. Total pore volume and particle size were not quantifiable for sium caseinate and PVPP, which indicates that these coagulants have a microporou ture. The shape of its N2 adsorption-desorption isotherm was a type I isotherm, typ microporous solids with relatively small external surfaces, as defined by the Intern Union of Pure and Applied Chemistry (IUPAC) [57].

Coagulation-Flocculatuion-Decantation Experiments
CFD experiments were performed in a jar-test apparatus (ISCO JF-4). Severa were performed using 500 mL of effluent in 1000 mL beakers. Fixed conditions w

Coagulation-Flocculatuion-Decantation Experiments
CFD experiments were performed in a jar-test apparatus (ISCO JF-4). Several trials were performed using 500 mL of effluent in 1000 mL beakers. Fixed conditions were set as the following: pH 4.0, rapid mixing 150 rpm/3 min, slow mixing 20 rpm/20 min, temperature 298 K, sedimentation period 12 h. The CFD experiments were developed with the statistical software Minitab 18.0 (State College, Pennsylvania, USA), and applied the simplex lattice design (SLD) as follows: 1.
Application of bentonite, potassium caseinate and PVPP with a lower and upper dosage of 0 and 1.0 g/L, respectively; 2.
Application of maximum mixture dosage of 1.0 g/L.
All the experiments were performed in triplicate and the observed standard deviation was always less than 5% of the reported values. Statistical analysis was performed by OriginLab 2019 software (Northampton, MA, USA).

Ozonation Experiments
Batch experiments for the ozonation process were performed using an air pump IDEA-R AP2 (1.8 W/1000 ccO 2 /min) from SICCE. Oxygen was converted into ozone by an ozone generator 1KNT-24 (25 W) of 1000BT-12 with low working noise (<30 DB). This ozone generator uses high-output corona discharge ozone tube, together with a high Óow ball bearing cooling fan (48 CFM) to ensure steady ozone output (up to 300 mg/hr). Figure 2 shows the setup of the ozone reactor used in this work.
The photoreactor was fitted with a Heraeus TNN 15/32 lamp (14.5 cm in length and 2.5 cm in diameter), mounted in the axial position inside the reactor. The spectral output of the low-pressure mercury vapor lamp emitted mainly (85-90%) at 253.7 nm and about 7-10% at 184.9 nm. The experiments were carried out as follows: All the experiments were performed in triplicate and the observed standard deviation was always less than 5% of the reported values. Statistical analysis was performed with OriginLab 2019 software (Northampton, MA, USA).  All the experiments were performed in triplicate and the observed standard deviation was always less than 5% of the reported values. Statistical analysis was performed with OriginLab 2019 software (Northampton, MA, USA).

Simplex Lattice Design-Model Establishment
In Table 1, it is observed that the WW has high levels of turbidity (1040 NTU), total suspended solids (TSS) (2430 mg/L), and organic content (1962 mg C/L). The CFD process is essential to reduce these parameters, to improve the ozonation treatment efficiency, since the presence of refractory compounds is considered to be hydroxyl radical scavengers. In addition, without suspended solids, light can better penetrate through the solution, to trigger ozone decomposition [11]. The CFD experiments were performed by a simplex lattice design (SLD), in which seven different tests were performed, to optimize the combination of coagulants. The experimental and predicted values obtained after the assessment of turbidity, TSS, chemical oxygen demand (COD), and total organic carbon (TOC), are listed in Table 5. The regression models of the four responses were established by linear regression fitting (Equations (3)-(6)), as follows: Validation of the statistical design model is a very important parameter, in order to assess the relevance of the obtained results. Univariate analysis of variance (ANOVA) provides an extremely powerful and useful tool for the statistical tests of different factors and their interactions in experiments [58]. The regression-adjusted average squares and linear regression-adjusted average squares, allowed the calculation of the Fisher ratios (F-value), which allowed the determination of the statistical significance (p-value). Table S1 (see Supplementary Materials) shows that the P-value for regression, linear regression and quadratic regression of turbidity, TSS, COD, and TOC, is non-significant for a p < 0.05 [59].
To establish the reliability of the results that were obtained by the SLD statistic design, it was necessary to evaluate other sets of parameters. By the performance of linear fit, the regression equations were obtained. These equations established a relation between the observed results and the predicted results, which were evaluated by the calculation of R 2 and R 2 adjusted. By observation of the R 2 values, turbidity, TSS, COD, and TOC had values of 99.5, 97.9, 95.5, and 99.7%, respectively, which were greater than 80.0%, ensuring a very good model fit [60,61]. The R 2 adjusted that was obtained from the turbidity, TSS, COD, and TOC regression equations (97.3, 87.7, 73.1, and 98.1%, respectively), was also high, indicating the strong significance of the model [61,62].
The adequacy of the model was also evaluated by use of diagnostic plots, such as a normal probability plot of the standardized residuals, and the plot between individual residual values and the fitted values ( Figure 3). The normal probability plot shows the distribution of the residual value, which is defined as the difference between the predicted (model) and observed (experimental) values. From the normal probability plot, the population is claimed not to be normally distributed if, and only if, the n points do not fall close to a straight line [63].
After analyzing the plots between individual residual values and the fitted values, it was observed that the turbidity, TSS, COD, and TOC residual values were scattered randomly around zero, which was similar to Harbi et al., [64], who observed a uniform scattering of the residual values after the performance of an SLD statistic design on a PCR detection method for nitrite reductase genes.

Simplex Lattice Design-Model Optimization
To improve the CFD optimization, an SLD statistic design was performed, in which three different coagulants, potassium caseinate (X1), activated sodium bentonite (X2), and PVPP (X3), were tested under seven different combinations, with the evaluation of turbidity, TSS, COD, and TOC (Table 5), under the following operational conditions: pH 4.0, rapid mixing 150 rpm/3 min, slow mixing 20 rpm/20 min, temperature 298 K, and sedimentation period 12 h. The maximum dosage (1.0 g/L) and pH 4.0 were selected, based on the works of Cosme et al. [65,66], who performed coagulation with these coagulants on the natural pH (3.3-3.6) of wine. Additionally, potassium caseinate has an isoelectric point of 4.6 [67] and, at a pH higher than 4.6, potassium caseinate is dissolved. The activated sodium bentonite that was used in this work presents a pH of 7.4, which corresponds to its isoelectric point (pH = 7); however, bentonite becomes electropositive at pH In Figure 3, the formation of a straight line was observed and the residual values were normally distributed on both sides of the line, which indicated that the experimental points of turbidity, TSS, COD, and TOC were reasonably aligned with the predicted values.
After analyzing the plots between individual residual values and the fitted values, it was observed that the turbidity, TSS, COD, and TOC residual values were scattered randomly around zero, which was similar to Harbi et al. [64], who observed a uniform scattering of the residual values after the performance of an SLD statistic design on a PCR detection method for nitrite reductase genes.

Simplex Lattice Design-Model Optimization
To improve the CFD optimization, an SLD statistic design was performed, in which three different coagulants, potassium caseinate (X 1 ), activated sodium bentonite (X 2 ), and PVPP (X 3 ), were tested under seven different combinations, with the evaluation of turbidity, TSS, COD, and TOC (Table 5), under the following operational conditions: pH 4.0, rapid mixing 150 rpm/3 min, slow mixing 20 rpm/20 min, temperature 298 K, and sedimentation period 12 h. The maximum dosage (1.0 g/L) and pH 4.0 were selected, based on the works of Cosme et al. [65,66], who performed coagulation with these coagulants on the natural pH (3.3-3.6) of wine. Additionally, potassium caseinate has an isoelectric point of 4.6 [67] and, at a pH higher than 4.6, potassium caseinate is dissolved. The activated sodium bentonite that was used in this work presents a pH of 7.4, which corresponds to its isoelectric point (pH = 7); however, bentonite becomes electropositive at pH 3.6-4.0 [68] and, therefore, after bentonite addition to the wastewater, its pH was corrected to 4.0. The variable charge of clays is affected by the pH, due to the ionization of its external hydroxyl groups. Therefore, the external sites may acquire positive (OH + 2 ) or negative charge ( − O − ), whether the pH is lower or higher than its isoelectric point (pH at the point of zero charge) [69,70]. In the work of Guimarães et al. [69], when the clay was mixed with WW at pH 4.0, a mechanism of adsorption occurred, in which a high concentration of amphoteric flavylium species were adsorbed by the clay. Therefore, pH 4.0 is the best pH for the performance of the CFD process, with activated sodium bentonite. The PVPP was tested in this work, based on the research of Labord et al. [71], in which it was reported that PVPP had a high affinity for polyphenols, according to tests conducted in wine treatment. PVPP is a high-molecularweight polymer, which is insoluble in water [38]. It interacts with polyphenols via H bonds between their CO-N linkages and phenol groups [72]. Figure 4 shows the interactions of activated sodium bentonite, potassium caseinate, and PVPP, with the colloids present in the wastewater. When bentonite was added, a mechanism of adsorption and charge neutralization occurred, attracting the colloids present in the WW to the interlayer region of the clay [73]. A similar mechanism was observed with the employment of potassium caseinate, which due to the different charges, attracted the negatively chargeed particles to the positively charged proteins, producing heavy aggregated particles, which precipitate by gravity [74]. PVPP acted by a mechanism of adsorption and interparticle bridging. The polymer chains of PVPP adsorbed the colloids from the WW, as a result of (1) coulombic (charge-charge) interactions, (2) dipole interaction, (3) hydrogen bonding, and (4) van der Waals forces of attraction [75]. In addition, PVPP creates a tridimensional net that adsorbs on the available surface sites of other particles (such as bentonite and potassium caseinate), thus creating a "bridge" between the particles surfaces, resulting in larger particles that settle more efficiently by gravity.
tested in this work, based on the research of Labord et al. [71], in which it was reported that PVPP had a high affinity for polyphenols, according to tests conducted in wine treatment. PVPP is a high-molecular-weight polymer, which is insoluble in water [38]. It interacts with polyphenols via H bonds between their CO-N linkages and phenol groups [72]. Figure 4 shows the interactions of activated sodium bentonite, potassium caseinate, and PVPP, with the colloids present in the wastewater. When bentonite was added, a mechanism of adsorption and charge neutralization occurred, attracting the colloids present in the WW to the interlayer region of the clay [73]. A similar mechanism was observed with the employment of potassium caseinate, which due to the different charges, attracted the negatively chargeed particles to the positively charged proteins, producing heavy aggregated particles, which precipitate by gravity [74]. PVPP acted by a mechanism of adsorption and interparticle bridging. The polymer chains of PVPP adsorbed the colloids from the WW, as a result of (1) coulombic (charge-charge) interactions, (2) dipole interaction, (3) hydrogen bonding, and (4) van der Waals forces of attraction [75]. In addition, PVPP creates a tridimensional net that adsorbs on the available surface sites of other particles (such as bentonite and potassium caseinate), thus creating a ''bridge'' between the particles surfaces, resulting in larger particles that settle more efficiently by gravity.  The statistical analysis that was obtained from the SLD design was converted into an optimization chart, which related the best results that were obtained from the CFD process. In accordance with the optimization chart ( Figure S1), the mixture of 0.48 g/L potassium caseinate (X 1 ) and 0.52 g/L bentonite (X 2 ) could reach a maximum removal of COD, TOC, turbidity, and TSS of 56.5, 44.3, 99.6, and 98.1%, respectively. Therefore, the optimal conditions that were selected for the CFD process were as follows: 0.48 g/L potassium caseinate, 0.52 g/L bentonite, pH 4.0, temperature 298 K, rapid mix 150 rpm/3 min, slow mix 20 rpm/20 min, and sedimentation time 12 h. After 12 h of sedimentation (Table S2, Figure 5), a 98.3, 97.6, 48.0, and 44.6% removals of turbidity, TSS, COD, and TOC were observed, which are similar to the values predicted by the optimization chart in Figure  S1. Considering the Portuguese Decree Law nº 236/98 for residual water discharge, it was observed that TSS achieved the legal value (60 mg/L). mix 20 rpm/20 min, and sedimentation time 12 h. After 12 h of sedimentation (Table S2, Figure 5), a 98.3, 97.6, 48.0, and 44.6% removals of turbidity, TSS, COD, and TOC were observed, which are similar to the values predicted by the optimization chart in Figure S1. Considering the Portuguese Decree Law nº 236/98 for residual water discharge, it was observed that TSS achieved the legal value (60 mg/L).

Effect of pH
In Section 3.1, the CFD process was optimized by employing an SLD statistical design. However, the CFD process was insufficient to substantially decrease the high organic carbon content that was present in the WW, thus, the application of a further chemical oxidation process, based on the reaction between ozone and a catalyst (Fe 2+ ), under UV-C radiation, was necessary. In the work of Huang et al. [76], it was observed that a combination of iron with O3 reached higher dissolved organic carbon (DOC) removal (53%), regarding O3 alone (32%), in the treatment of pharmaceutical wastewater. However, to our knowledge, the O3/Fe 2+ /UV-C process was never applied to the treatment of WW. Therefore, the O3/Fe 2+ /UV-C process was optimized, to act as a CFD complementary process. In order to maximize TOC removal, different pH's (4.0, 7.0, 9.0, and 11.0) were tested, under the following operational conditions: [Fe 2+ ] = 1.0 mM, ozone flow rate 5 mg/min, air flow 1.0 L/min, agitation 350 rpm, time 600 min, and a UV-C mercury lamp (254 nm). In Figure 6, the O3/Fe 2+ /UV-C system is represented in closed symbols, while the blank experiments, at pH 4.0 (O3, UV-C, and O3/UV-C), are represented in open symbols. The effect

Effect of pH
In Section 3.1, the CFD process was optimized by employing an SLD statistical design. However, the CFD process was insufficient to substantially decrease the high organic carbon content that was present in the WW, thus, the application of a further chemical oxidation process, based on the reaction between ozone and a catalyst (Fe 2+ ), under UV-C radiation, was necessary. In the work of Huang et al. [76], it was observed that a combination of iron with O 3 reached higher dissolved organic carbon (DOC) removal (53%), regarding O 3 alone (32%), in the treatment of pharmaceutical wastewater. However, to our knowledge, the O 3 /Fe 2+ /UV-C process was never applied to the treatment of WW. Therefore, the O 3 /Fe 2+ /UV-C process was optimized, to act as a CFD complementary process. In order to maximize TOC removal, different pH's (4.0, 7.0, 9.0, and 11.0) were tested, under the following operational conditions: [Fe 2+ ] = 1.0 mM, ozone flow rate 5 mg/min, air flow 1.0 L/min, agitation 350 rpm, time 600 min, and a UV-C mercury lamp (254 nm). In Figure 6 Table S3. It was observed that with different initial pH (4.0, 7.0, 9.0, and 11.0), there was a TOC removal of 31.9, 28.6, 11.2, and 35.8%, respectively, for O 3 ; 33.1, 0.9, 0.0, and 4.1%, respectively, for UV-C; and 57.5, 37.6, 36.5, and 38.0%, respectively, for O 3 /UV-C. These results were in agreement with the work of Hassanshahi and Karimi-Jashni [77], which observed that the pH had little effect in the COD removal of gray water, by the ozonation process.
increase in pH > 4.0, could be explained by the precipitation of iron at alkaline pH, in the form of iron hydroxide ([Fe 2+ ] leached = 12.6, 35.4, 42.6, and 188.8 mg Fe/L, respectively, for pH 4.0, 7.0, 9.0, and 11), which decreases the conversion of H2O2 to HO • and reduces the transmission of the radiation [88]. Therefore, based in these results, pH 4.0 was selected as the best pH for the O3/Fe 2+ /UV-C process studied.

Effect of Fe 2+ Concentration
In the previous section, as the pH had a great influence on the removal of TOC from the WW, by the ozonation process, was described. It was also observed that the treatment using the combination of O3/UV-C generated H2O2, which interacted with Fe 2+ to produce HO • radicals. Therefore, in this section, the Fe 2+ concentration was varied (0.5-2.0 mM) under the following operational conditions: pH 4.0, ozone flow rate 5 mg/min, air flow 1.0 L/min, agitation 350 rpm, and radiation UV-C mercury lamp (254 nm).
In Figure 7, a TOC removal of 59.0, 63.2, and 66.7% (Table S4), was observed, for 0.5, 1.0, and 2.0 mM Fe 2+ , respectively. These results agreed with Quiroz et al., [89], who stated that the application of iron with ozone improved the COD removal of industrial wastewater. In addition, in the work of Piera et al. [82], it was reported that Fe 2+ can interact with O3 and produce HO • radicals (Equations (23) and (24)), as follows:  The results in Figure 6 showed a lower TOC removal efficiency for a single O 3 application, regarding O 3 /UV-C. These results can be explained by two mechanisms through which ozone can degrade organic pollutants, such as direct electrophilic attack (Equation (7)) and indirect attack, through the formation of hydroxyl radicals (Equations (8) and (9)) [78,79], which are shown as follows: where R is the organic solutes and Rox is the oxidized organic products.
where [TOC] 0 and [TOC] t are the TOC concentrations at time 0 and t in mg C/L. As the O 3 /Fe 2+ /UV-C reaction proceeds, the concentration of TOC decreases. Another measure of the rate of a reaction, relating concentration to time, is the half-life, t 1/2 , which is the time required for the concentration of TOC to decrease to half of its initial concentration. We can obtain an expression for t 1/2 , for a first-order reaction, as described by Equation (22) [87], as follows: It was observed that k pH 4.0 = 1.67 × 10 −3 min −1 > k pH 9.0 = 1.18 × 10 −3 min −1 > k pH 11 = 1.07 × 10 −3 min −1 > k pH 7.0 = 9.41 × 10 −4 min −1 . The decrease in the kinetic rate, with the increase in pH > 4.0, could be explained by the precipitation of iron at alkaline pH, in the form of iron hydroxide ([Fe 2+ ] leached = 12.6, 35.4, 42.6, and 188.8 mg Fe/L, respectively, for pH 4.0, 7.0, 9.0, and 11), which decreases the conversion of H 2 O 2 to HO • and reduces the transmission of the radiation [88]. Therefore, based in these results, pH 4.0 was selected as the best pH for the O 3 /Fe 2+ /UV-C process studied.

Effect of Fe 2+ Concentration
In the previous section, as the pH had a great influence on the removal of TOC from the WW, by the ozonation process, was described. It was also observed that the treatment using the combination of O 3 /UV-C generated H 2 O 2, which interacted with Fe 2+ to produce HO • radicals. Therefore, in this section, the Fe 2+ concentration was varied (0.5-2.0 mM) under the following operational conditions: pH 4.0, ozone flow rate 5 mg/min, air flow 1.0 L/min, agitation 350 rpm, and radiation UV-C mercury lamp (254 nm).
In Figure 7, a TOC removal of 59.0, 63.2, and 66.7% (Table S4), was observed, for 0.5, 1.0, and 2.0 mM Fe 2+ , respectively. These results agreed with Quiroz et al., [89], who stated that the application of iron with ozone improved the COD removal of industrial wastewater. In addition, in the work of Piera et al. [82], it was reported that Fe 2+ can interact with O 3 and produce HO • radicals (Equations (23) and (24)), as follows: Fe 2+ + HO 2 → Fe 3+ + HO 2 (26) The results were fitted into a pseudo-first-order kinetic rate, and the following order was observed: k 2.0 mM = 1.72 × 10 −3 min −1 > k 1.0 mM = 1.67 × 10 −3 min −1 > k 0.5 mM = 1.34 × 10 −3 min −1 . Due to the low differences in the kinetic rate that was observed between 1.0 and 2.0 mM Fe 2+ , and due to the high costs that are associated with the application of 2.0 mM Fe 2+ , a ferrous iron concentration of 1.0 mM was selected as the best concentration for the O3/Fe 2+ /UV-C process. Bolton et al. [90] proposed figures of merit for electric-driven photocatalysis. For firstorder kinetics, they proposed the electric energy per order (E EO ) as the electric energy, in kWh that are required to reduce the concentration of a pollutant concentration C by one order of magnitude, according to Equation (27), where P is the rated power [kW] of the AOP system, V is the volume [L] of water or air treated in the time t [h], and TOCi and TOCf are the concentrations of total organic carbon at initial and t times.
Batch operation (27) The results are presented in Table 6 and, as expected by increasing the Fe 2+ concentration, the E EO values decreased. Lower E EO values (in kWh•m −3 •order −1 ) correspond to higher removal efficiencies, in terms of electrical power consumption [90]. The application of UV-C radiation alone achieved the lowest E EO value (1720 kWh•m −3 •order −1 ) in comparison to the O3/UV-C and O3 alone (2153 and 2996 kWh m −3 •order −1 , respectively), due to the low power of the UV mercury lamp. However, the reaction kinetics of the UV-C alone, were much lower (k UV-C = 7.17 × 10 −4 min −1 ) in comparison to the other treatments, thus more energy will be required to achieve similar TOC degradation. With the application of the O3/Fe 2+ /UV-C system, the results in Table 6 clearly indicated that with 0.5 mM Fe 2+ , the reaction took too much time, increasing the electric power consumption. However, little differences in the E EO values were observed when 1.0 and 2.0 mM Fe 2+ were After 600 min of reaction, an Fe 2+ concentration of 1.84, 12.60, and 41.92 mg Fe/L, respectively, was observed, for 0.5, 1.0, and 2.0 mM. With the application of 0.5 mM Fe 2+ , there was a low concentration of Fe 2+ present in the solution, to react with H 2 O 2 . The use of higher concentrations of ferrous iron resulted in a higher TOC removal, although the increase from 1.0 to 2.0 mM Fe 2+ was very mild, possibly because scavenging reactions, between Fe 2+ and HO • radicals, may have occurred, as observed in Equations (25)-(26) [19], as follows: The results were fitted into a pseudo-first-order kinetic rate, and the following order was observed: k 2.0 mM = 1.72 × 10 −3 min −1 > k 1.0 mM = 1.67 × 10 −3 min −1 > k 0.5 mM = 1.34 × 10 −3 min −1 . Due to the low differences in the kinetic rate that was observed between 1.0 and 2.0 mM Fe 2+ , and due to the high costs that are associated with the application of 2.0 mM Fe 2+ , a ferrous iron concentration of 1.0 mM was selected as the best concentration for the O 3 /Fe 2+ /UV-C process.
Bolton et al. [90] proposed figures of merit for electric-driven photocatalysis. For first-order kinetics, they proposed the electric energy per order (E EO ) as the electric energy, in kWh that are required to reduce the concentration of a pollutant concentration C by one order of magnitude, according to Equation (27), where P is the rated power [kW] of the AOP system, V is the volume [L] of water or air treated in the time t [h], and TOC i and TOC f are the concentrations of total organic carbon at initial and t times.
Batch operation (27) The results are presented in Table 6 and, as expected by increasing the Fe 2+ concentration, the E EO values decreased. Lower E EO values (in kWh·m −3 ·order −1 ) correspond to higher removal efficiencies, in terms of electrical power consumption [90]. The application of UV-C radiation alone achieved the lowest E EO value (1720 kWh·m −3 ·order −1 ) in comparison to the O 3 /UV-C and O 3 alone (2153 and 2996 kWh m −3 ·order −1 , respectively), due to the low power of the UV mercury lamp. However, the reaction kinetics of the UV-C alone, were much lower (k UV−C = 7.17 × 10 −4 min −1 ) in comparison to the other treatments, thus more energy will be required to achieve similar TOC degradation. With the application of the O 3 /Fe 2+ /UV-C system, the results in Table 6 clearly indicated that with 0.5 mM Fe 2+ , the reaction took too much time, increasing the electric power consumption. However, little differences in the E EO values were observed when 1.0 and 2.0 mM Fe 2+ were applied, and, considering the high iron leaching that was observed with 2.0 mM Fe 2+ , the application of 1.0 mM Fe 2+ becomes the best choice. Table 6. Evaluation of first-order kinetic rate (k), half-life (t 1/2 ) and electric energy per order (E EO ) through the O 3 /Fe 2+ /UV-C process at different Fe 2+ concentrations (0.5-2.0 mM). P Ozonator = 0.025 kW, P UV-C = 0.015 kW, t = 10 h, V = 0.5 L. The comparative values of E EO , for the degradation of active pharmaceutical ingredients, real textile wastewater, and organic pollutants, by the ozonation process, are displayed in Table 7. In addition, a comparative study is also shown, in which the winery wastewater was treated by a UV-C/PMS/Co(II) system. The results showed that the winery wastewater is very difficult to treat (1843 kWh m −3 ·order −1 ), with higher energy requirements, regarding the treatment of active pharmaceutical ingredients, real textile wastewater, and organic pollutants. These results were in agreement with Rodriguez-Chueca et al. [19], who observed a high E EO for the treatment of WW (173 kWh·m −3 ·order −1 ). The application of ozonation processes in the treatment of WW is insufficient, with scarce examples of organic matter removal and energy consumption. Most of the authors studied the degradation of emerging contaminants and textile dye removal [91][92][93]; however, the degradation of the organic matter load of winery wastewater requires more demanding treatments, with higher energy consumption [20].

Evaluation of Ozone Consumption
In the previous sections, we described how to optimize the ozonation process, and it was observed that under the best operational conditions, pH 4.0, [Fe 2+ ] = 1.0 mM, ozone flow rate 5 mg/min, air flow 1.0 L/min, agitation 350 rpm and UV-C radiation (254 nm), a TOC removal of 63.2% was achieved. However, to understand the efficiency of the O 3 /Fe 2+ /UV-C process, the concentration of ozone dissolved in the wastewater was assessed (Table S5). The rate of ozone consumption was monitored by AccuVac Ampul procedure, and the ozone that was lost in the process was measured by the difference between the ozone mass flow rate at the inlet of the reactor and the ozone dissolved in the WW. Table 7. Reported values of electric energy per order (E EO ) in ozonation processes. This concentration of dissolved O 3 was observed to be much lower than the concentration that was observed in the work of Lucas et al. [23], considering that the air was supplied by a small air pump, but a high TOC removal kinetic rate was observed with this process.

Combination of Coagulation-Flocculation-Decantation with Ozonation Processes
The WW is very toxic when discharged into the environment without proper treatment, due to the high levels of organic carbon that are present in its composition (Table 1). Previous treatments, with CFD and O 3 /Fe 2+ /UV-C, achieved a high TOC removal; however, both treatments showed limitations and, therefore, in this section, the processes of CFD and O 3 were combined CFD/O 3 and O 3 /CFD processes, with the application of the CFD process' best operational conditions-0.48 g/L potassium caseinate, 0.52 g/L bentonite, pH 4.0, temperature 298 K, rapid mix 150 rpm/3 min, slow mix 20 rpm/20 min, and sedimentation time 12 h-and the ozonation process' best operational conditions-pH 4.0, [Fe 2+ ] = 1.0 mM, ozone flow rate 5 mg/min, air flow 1.0 L/min, agitation 350 rpm, and UV-C radiation (254 nm). Figure 9 shows a TOC removal of 63.2 and 66.1%, respectively, after the performance of O 3 and the combined O 3 /CFD processes (Table S6). A TOC removal of 44.6 and 65.5%, respectively, was also observed, after the performance of CFD and the CFD/O 3 processes. Clearly, there were not significant differences between both the combined treatments; however, after the analysis of the biodegradability, an increase, from 0.28 (raw WW) to 0.29 for O 3 /CFD and 0.40 for CFD/O 3 , was observed. Therefore, the application of CFD before the O 3 process can be more advantageous for the WW treatment. These results were in agreement with Liu et al. [94], who observed that the performance of the CFD process, as a pre-treatment, enhanced the ozonation process in the treatment of a landfill leachate. The results were also in agreement with Bu et al. [95], who observed that the performance of the CFD process, as a pre-treatment, followed by ozonation, enhanced the UV 254 removal efficiency and decreased the accumulation of organic matter in the treatment of wastewater. Considering the Portuguese Decree Law nº 236/98 for residual water discharge, only TSS achieved the legal value (60 mg/L), while COD (150 mg O 2 /L), BOD 5 (40 mg O 2 /L), and total polyphenols (0.5 mg gallic acid/L), failed to reach the legal thresholds. Nonetheless, the observed improvement in biodegradability showed that the WW can be sent to a biological reactor for further degradation of organic matter.
O3/Fe 2+ /UV-C process, the concentration of ozone dissolved in the wastewater was assessed (Table S5). The rate of ozone consumption was monitored by AccuVac Ampul procedure, and the ozone that was lost in the process was measured by the difference between the ozone mass flow rate at the inlet of the reactor and the ozone dissolved in the WW.
In Figure 8, it was observed that under a constant injection of 5 mg O3/L, only 0.165, 0.330, 0.190, 0.270, and 0.360 mg O3/min reacted with the Fe 2+ , to produce hydroxyl radicals (HO • ), respectively, at time of 120, 240, 360, 480, and 600 min −1 . This concentration of dissolved O3 was observed to be much lower than the concentration that was observed in the work of Lucas et al. [23], considering that the air was supplied by a small air pump, but a high TOC removal kinetic rate was observed with this process.

Combination of Coagulation-Flocculation-Decantation with Ozonation Processes
The WW is very toxic when discharged into the environment without proper treatment, due to the high levels of organic carbon that are present in its composition (Table  1). Previous treatments, with CFD and O3/Fe 2+ /UV-C, achieved a high TOC removal; however, both treatments showed limitations and, therefore, in this section, the processes of CFD and O3 were combined CFD/O3 and O3/CFD processes, with the application of the CFD process' best operational conditions-0.48 g/L potassium caseinate, 0.52 g/L bentonite, pH 4.0, temperature 298 K, rapid mix 150 rpm/3 min, slow mix 20 rpm/20 min, and A factor that possibly explains the high organic carbon removal that was observed after the combined O 3 /CFD and CFD/O 3 processes, is the reduction in turbidity, TSS, and total polyphenols, which would otherwise prevent light from penetrating into the water and triggering the photo-Fenton reaction [11]. In Figure 10, the application of O 3 , O 3 /CFD, CFD, and CFD/O 3 , achieved a turbidity removal of 65.5, 99.9, 98.3, and 99.3%, respectively, and a TSS removal of 66.8, 98.3, 97.6, and 98.3%, respectively. In addition to turbidity and TSS, a total polyphenol removal of 95.1, 95.9, 81.2, and 99.3%, respectively, was also observed. These results were in agreement with Lucas et al. [23], who observed that the ozonation process could be beneficial for the removal of polyphenols (toxicity) from the winery wastewater, thus contributing to the safety of public and environmental health.

Effect of the Treatments in Phytotoxicity of Different Plants
In previous sections, was described how the combination of the CFD and ozonation processes were beneficial for the reduction in the organic carbon present in the WW. However, considering that the application of oenological coagulants in combination with the ozonation process, has never been implemented in wastewater treatment, its effects in vegetables are still unknown. Therefore, in this section, we evaluated the effects of the different treatments in the phytotoxicity of different species of vegetables. A series of tests were performed on the germination of seeds, in order to evaluate the phytotoxicity of the treatments, similarly to several authors [96][97][98][99], in two Dicotyledonae species (lettuce and cucumber) and in two Monocotyledonae species (corn and onion). In Figure 11, it was observed that WW had a phytotoxic effect on cucumber and lettuce seeds (GI = 51 and 0%, respectively). The performance of the treatments O 3 , O 3 /CFD, CFD, and CFD/O 3 increased the germination index (GI) of the cucumber seeds, to 113, 109, 71, and 117% (Table S7), respectively. The germination index of the lettuce seeds increased to 298, 88, and 249%, respectively, for O 3 , O 3 /CFD, and CFD/O 3 . These are excellent values that reveal the goodness of the studied treatments and the reduction in phytotoxicity. 0.29 for O3/CFD and 0.40 for CFD/O3, was observed. Therefore, the application of CFD before the O3 process can be more advantageous for the WW treatment. These results were in agreement with Liu et al. [94], who observed that the performance of the CFD process, as a pre-treatment, enhanced the ozonation process in the treatment of a landfill leachate. The results were also in agreement with Bu et al. [95], who observed that the performance of the CFD process, as a pre-treatment, followed by ozonation, enhanced the UV254 removal efficiency and decreased the accumulation of organic matter in the treatment of wastewater. Considering the Portuguese Decree Law nº 236/98 for residual water discharge, only TSS achieved the legal value (60 mg/L), while COD (150 mg O2/L), BOD5 (40 mg O2/L), and total polyphenols (0.5 mg gallic acid/L), failed to reach the legal thresholds. Nonetheless, the observed improvement in biodegradability showed that the WW can be sent to a biological reactor for further degradation of organic matter. A factor that possibly explains the high organic carbon removal that was observed after the combined O3/CFD and CFD/O3 processes, is the reduction in turbidity, TSS, and total polyphenols, which would otherwise prevent light from penetrating into the water and triggering the photo-Fenton reaction [11]. In Figure 10, the application of O3, O3/CFD, CFD, and CFD/O3, achieved a turbidity removal of 65.5, 99.9, 98.3, and 99.3%, respectively, As previously observed, some of the problems that are generated by this type of wastewater are the intense color and high turbidity, which absorbs the radiation and decreases the treatment efficiency. The intense color that is present in winery wastewaters appears mainly due to the presence of total phenols, non-flavonoids, flavonoids, total anthocyanins, colored anthocyanins, total pigments, and total tannins, present in the wines [100]. Considering that there is insufficient information regarding the impact of the CFD and ozonation processes in the phenolic composition of the WW, in this section, the efficiency of these treatments in the removal of total phenols, non-flavonoids, flavonoids, total anthocyanins, colored anthocyanins, total pigments, and total tannins, was evaluated ( Figure 12). The performance of the treatments O 3 , O 3 /CFD, CFD, and CFD/O 3 achieved a total phenol removal of 0.0, 5.1, 1.9, and 3.2%, respectively, a non-flavonoid removal of 0.0, 0.0, 2.6, and 2.6%, respectively, and a flavonoid removal of 0.0, 14.6, 0.0, and 7.3% (Table S8), respectively, which are responsible for the yellow color in wines [38]. In Figure 12, a high decrease in the total anthocyanins (50.0, 50.0, 25.0, and 100%, respectively) and colored anthocyanins (28.6, 10.0, 42.9, and 100%, respectively) was observed, which are linked to the red color of wines [101]. The reduction in these phenolic compounds had an effect in the removal of color from the wastewater, which was evaluated by a CIELab analysis (Table S9). The combined treatments O 3 /CFD and CFD/O 3 had negative values for ∆a* and ∆b* (−1.90 and −3.29), which indicated a reduction in the red and yellow color. The luminosity (L * ) increased from 0.0% (raw WW) to 99.7 and 100%, respectively, which meant that the phenolic compounds were directly linked to the dark-yellow color of the wastewater. The color removal, given by the Euclidean distance, was 99.74 and 100, after the combined treatments of O 3 /CFD and CFD/O 3 , which meant that color removal was perceptible by the human eye, which is in accordance with Spagna et al. [41].  [23], who observed that the ozonation process could be beneficial for the removal of polyphenols (toxicity) from the winery wastewater, thus contributing to the safety of public and environmental health.

Effect of the Treatments in Phytotoxicity of Different Plants
In previous sections, was described how the combination of the CFD and ozonation processes were beneficial for the reduction in the organic carbon present in the WW. However, considering that the application of oenological coagulants in combination with the ozonation process, has never been implemented in wastewater treatment, its effects in vegetables are still unknown. Therefore, in this section, we evaluated the effects of the different treatments in the phytotoxicity of different species of vegetables. A series of tests were performed on the germination of seeds, in order to evaluate the phytotoxicity of the treatments, similarly to several authors [96][97][98][99], in two Dicotyledonae species (lettuce and cucumber) and in two Monocotyledonae species (corn and onion). In Figure 11, it was observed that WW had a phytotoxic effect on cucumber and lettuce seeds (GI = 51 and 0%, respectively). The performance of the treatments O3, O3/CFD, CFD, and CFD/O3 increased the germination index (GI) of the cucumber seeds, to 113, 109, 71, and 117% (Table S7), respectively. The germination index of the lettuce seeds increased to 298, 88, and 249%, respectively, for O3, O3/CFD, and CFD/O3. These are excellent values that reveal the goodness of the studied treatments and the reduction in phytotoxicity.   the combined treatments of O3/CFD and CFD/O3, which meant that color removal was perceptible by the human eye, which is in accordance with Spagna et al. [41].

Conclusions
In this work, a WW was treated by two different processes, a CFD process, which employed activated sodium bentonite mixed with potassium caseinate, and the advanced oxidation process O3/Fe 2+ /UV-C. The combination of both treatment processes in a CFD/O3/Fe 2+ /UV-C system, proved to be a feasible method for the treatment of WW, and the main conclusions are summarized as follows: 3. The performance of a CFD process, by application of an SLD statistical design, allows a high removal of turbidity, TSS, TOC, and COD (98.3, 97.6, 44.6, and 48.0%); 4. The application of an ozonation process, under the best operational conditions-pH = 4.0, [Fe 2+ ] = 1.0 mM, ozone flow rate 5 mg/min, air flow 1.0 L/min, agitation 350 rpm, time 600 min, and radiation UV-C mercury lamp (254 nm)-achieves a TOC removal of 63.2%; 5. The O3/1.0 mM Fe 2+ /UV-C system is concluded to be very efficient in terms of energy consumption, with an E EO = 1843 kWh m −3 order −1 ; 6. The combined processes O3/CFD and CFD/O3 achieved high TOC removal (66.1 and 65.5%, respectively). It is also concluded that the performance of the CFD/O3 process achieves higher biodegradability (0.40); 7. It is concluded that the combined processes O3/CFD and CFD/O3 have lower phytotoxicity effects in the germination of plant seeds;

Conclusions
In this work, a WW was treated by two different processes, a CFD process, which employed activated sodium bentonite mixed with potassium caseinate, and the advanced oxidation process O 3 /Fe 2+ /UV-C. The combination of both treatment processes in a CFD/O 3 /Fe 2+ /UV-C system, proved to be a feasible method for the treatment of WW, and the main conclusions are summarized as follows: 1.
The performance of a CFD process, by application of an SLD statistical design, allows a high removal of turbidity, TSS, TOC, and COD (98.3, 97.6, 44.6, and 48.0%); 2.
The O 3 /1.0 mM Fe 2+ /UV-C system is concluded to be very efficient in terms of energy consumption, with an E EO = 1843 kWh m −3 order −1 ; 4.
The combined processes O 3 /CFD and CFD/O 3 achieved high TOC removal (66.1 and 65.5%, respectively). It is also concluded that the performance of the CFD/O 3 process achieves higher biodegradability (0.40);

5.
It is concluded that the combined processes O 3 /CFD and CFD/O 3 have lower phytotoxicity effects in the germination of plant seeds; 6.
The combined process O 3 /CFD and CFD/O 3 have the capacity to completely decolor the WW (L* = 100%), through the high removal of phenolic compounds; 7.
The combined process O 3 /CFD and CFD/O 3 is concluded to decrease the risk of public and environmental health problems.
These results showed that the combination of both treatments was essential to achieve the high degradation of organic matter from the WW. In the future, based on these results, new approaches can be explored, such as optimization of the hydraulic retention time and microbial elimination.