Combined organic coagulants and photocatalytic processes for winery wastewater treatment

Due to the consumers demand for quality wines


Introduction
The production of wine is one of the largest industries in the world.
To maintain the quality of the product required by the consumers, various washing operations are necessary to be performed during the crushing and pressing of grapes as well as rinsing of fermentation tanks, barrels and other equipment or surfaces, and therefore, wineries, distilleries and other grape processing industries annually generate a large volume of wastewater (Malandra et al., 2003) with a pH ranging between 3 and 4, a chemical oxygen demand (COD) between 800 and 25, 000 mg O 2 /L (Petruccioli et al., 2002), high organic load of soluble sugars, organic acids, alcohols, polyphenols, tannins and structural polymers (Welz et al., 2016). Considering these values, the winery wastewater (WW) should not be directly conveyed to municipal wastewater treatment plants (WWTP) or released into the environment without proper treatment.
In this work, two processes are proposed to treat the WW, a physicalchemical process, namely coagulation-flocculation-decantation (CFD) process and a chemical process based in advanced oxidation processes (AOPs).
The CFD process is an easy and cheap process that allows the separation of the solids suspended in the wastewater. Many authors share the use of hydrolysable metal salts as coagulants, however there are several risks associated to its use, such as dialysis encephalopathy and prevalence of Alzheimer's disease (Ghernaout et al., 2018). Therefore, more ecologic products should be applied to guarantee the safety of disposed water. Plants such as Dactylis glomerata L., Daucus carota L., Festuca ampla Hack. and Tanacetum vulgare L. are considered invasive plants and large quantities of herbicide, such as glyphosate, are spent every year to eradicate these plants. These herbicides can get over-accumulated in soils as residues and consequently become a biological and environmental concern (Cerdeira and Duke, 2006;Yin et al., 2008). In addition, after the vintage, large amounts of rachis without value for winemaking are disposed, thus it is a cheap subproduct that could be recycled, along with the invasive plants and be transformed as coagulants. In this direction, the application of invasive plants and agriculture subproducts as coagulants, and the recycling of wastewater for irrigation, two problems become one solution.
To degrade the organic matter dissolved in the wastewater, advanced oxidation processes (AOPs) can be applied. The AOPs refers to all treatment processes that work at near ambient temperature and pressure, involving the generation of hydroxyl radicals (HO • ) with an oxidizing power of 2.80 V, which react immediately and non-selectively with most organic and organometallic pollutants until achieving total combustion or at least partial mineralization (Ganiyu et al., 2022). Depending on the type of oxidant used, AOPs can be designated HR-AOPs with the application of hydrogen peroxide (H 2 O 2 ) or SR-AOPs, with the application of persulfate (PS) and peroxymonosulfate (PMS). Among the HR-AOPs, the Fenton process is the most common. In this process ferrous iron (Fe 2+ ) reacts with H 2 O 2 to generate HO • radicals (Equation (1)). The ferrous iron is oxidized in this reaction to ferric iron and a second process occurs (Fenton-like), in which the Fe 3+ reacts with the H 2 O 2 producing hydroperoxyl radicals (HO • 2 ) (Equation (2)). The Fe 3+ reacts with the H 2 O in solution, been reduced to Fe 2+ (Equation (3)), however, this is a slow process (Lin et al., 2022). Fe 3+ + H 2 O 2 → Fe 2+ + HO • 2 + H + Equation 2 Fe 3+ + H 2 O → Fe 2+ + HO • + H + Equation 3 The application of ultraviolet radiation (UV) to Fenton process is able to increase the efficiency of the reaction, due to the faster regeneration of Fe 3+ to Fe 2+ (Equation (4)) and at the same time, the production of HO • radicals by photolysis of H 2 O 2 (Equation (5)) (Jorge et al., 2022b;Khan et al., 2021).
In the last decade, the employment of sulfate radicals (SO •-4 ) attracted a great attention on wastewater treatment applications (Shu et al., 2016), due to the high oxidizing potential (SO 2-4 ; E SO 2− 4 = 2.6 V).
The activation of persulfate is accomplished by various methods, such as heat or UV radiation (Equation (6)), transition metal (M n+ ) (Equation (7)) or alkaline pH (Equation (8)) (Chen et al., 2016;Dbira et al., 2019;Huie et al., 1991;Rodríguez-Chueca et al., 2017), as follows: The performance of oxidation process combining persulfate and hydrogen peroxide has been applied before on the treatment of trichloroethylene (Yan et al., 2013), chlorinated ethane (Ko et al., 2012), landfill leachate (Hilles et al., 2016), erythromycin (Li et al., 2017), and levofloxacin (Epold et al., 2015) due to the production of SO •-4 and HO • radicals simultaneously. However, most of these studies focus on the degradation of contaminants in distilled water, and these types of matrixes can't be compared to complex matrixes such as WW. In addition, to our knowledge, there is no information regarding the addition of both chemicals on the treatment of WW, and how both of them interact during oxidation process.
The main aim and major novelty of this work is the production and application of natural organic coagulants (NOC) to reduce the solid parts in the WW. It is also aimed in this work the application of an oxidation process combining hydrogen peroxide (H 2 O 2 ) and persulfate (PS) to degrade the dissolved carbon and to evaluate the effect of CFD process in the efficiency of oxidation process. Finally, it is aimed in this work the study of the separation of wastewaters (red and white) for the oxidation of organic matter. 30% w/w) and titanium(IV) oxysulfate solution 1.9-2.1%, were acquired by Sigma-Aldrich (Missouri, USA), sodium sulfite anhydrous (Na 2 SO 3 ) was acquired by Labkem, Barcelona, Spain. For pH adjustment, it was used sodium hydroxide (NaOH) from Labkem (Barcelona, Spain) and sulphuric acid (H 2 SO 4 , 95%) from Scharlau, (Barcelona, Spain).

Reagents and winery wastewater sampling
The winery wastewaters were collected from a cellar located in the north of Portugal. The samples were collected in plastic containers, transported to the laboratory and stored at −40 • C.

Analytical techniques
Different physical-chemical parameters were determined in order to characterize the WW, such as total suspended solids (TSS), chemical oxygen demand (COD), biological oxygen demand (BOD 5 ), total organic carbon (TOC) and total polyphenols (TPh). The main wastewater characteristics are shown in Table 1.
Total polyphenols (TPh) were measured by Folin-Ciocalteu method (Singleton and Rossi, 1965). Chemical oxygen demand (COD) analysis was carried out using a COD reactor from HACH Co. (Loveland, CO, USA) and a HACH DR 2400 spectrophotometer (Loveland, CO, USA) for colorimetric measurement. Biochemical oxygen demand (BOD 5 ) was determined using a respirometric OxiTop® IS 12 system (WTW, Yellow Springs, OH, USA). Dissolved organic carbon (DOC) samples were analyzed by direct injection of filtered samples into a Shimadzu TOC-L CSH analyzer (Shimadzu, Kyoto, Japan), equipped with an ASI-L autosampler, and with an NDIR detector, the H 2 O 2 concentration was determined using titanium (IV) oxysulfate (DIN 38 402H15 method) at 410 nm using a portable spectrophotometer from Hach (Loveland, Colorado, USA). The iron was analyzed by atomic absorption spectroscopy (AAS) using a Thermo Scientific™ iCE™ 3000 Series (Thermo Fisher Scientific, Massachusetts, USA). The turbidity was determined by a 2100 N IS Turbidimeter (Hach, Loveland, CO, USA), the TSS were determined by a portable spectrophotometer (Hach, Loveland, CO, USA), the pH was determined by a 3510 pH meter (Jenway, Cole-Parmer, UK) and conductivity was determined by a portable conductivity meter, VWR C030 (VWR, V. Nova de Gaia, Portugal), in accordance to the methodology of the Standard Methods (APHA et al., 1999). The FTIR spectra of plant coagulants was obtained by mixing 2 mg of the plant powder 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 an IRAffinity-1S Fourier Transform Infrared spectrometer (Shimadzu, Kyoto, Japan) and the infrared spectra in transmission mode were recorded in the 4000-400 cm −1 frequency region. The microstructural characterization was carried out with a scanning electron microscopy (FEI QUANTA 400 SEM/ ESEM, Fei Quanta, Hillsboro, WA, USA). Mineral characterization (iron, copper, sodium, potassium, calcium and magnesium) was obtained by digestion of 500 mg of powder samples with nitric acid and hydrogen peroxide for 24 h. Then the samples were transferred to a Dry Block Heater reactor (Techne, Cole-Parmer, UK), at 60 • C. The temperature was gradually raised to 80 • C, 100 • C, 120 • C until reach 150 • C. The samples were cooled, and then a matrix solution (1.5 ml de HNO 3 p.a. conc. in 1000 mL distilled water) was added. Finally, the cations concentrations were analyzed by atomic absorption spectroscopy (AAS). 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 (S BET ) was determined by applying the Gurevitsch's rule at a relative pressure p/p 0 = 0.30 and according to the Brunauer, Emmett, 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).

NOC preparation
All the plants used on this work were collected on the district of Vila Real, north of Portugal. The species Daucus carota L., Dactylis glomerata L., and Festuca ampla Hack. were collected at the GPS location 41 • 17 ′ 14.4 ′′ N 7 • 44 ′ 15.2 ′′ W. The species Tanacetum vulgare L. and Vítis vinifera L. were collected at the GPS location 41 • 18 ′ 25.2 ′′ N 7 • 42 ′ 06.2 ′′ W. On Table 2, it is indicated the plants sub-species, part collected for this study and the herbarium number attributed by UTAD for plant's identification.
All the vegetable parts collected were washed and dried in an oven at 70 • C for 24 h. Them they were grounded into powder using a groundnut miller. The powder was then left to cool and stored in a tightly closed plastic jar.

Coagulation-flocculation-decantation experimental set-up
The CFD experiments were performed in a conventional model Jar-Test apparatus (ISCO JF-4, Louisville, USA), using 500 mL of WW in 1000 mL beakers. In the CFD process, it was tested 5 NOCs (DG seeds, DC seeds, FA seeds, TV seeds and rachis and one metallic based coagulant (ferric chloride, 10% w/w). The CFD process was optimized as follows: 1) 0.1, 0.5, 1.0 and 2.0 g/L of coagulant were added to 500 mL of red WW samples at pH 3.0, 5.0, 7.0, 9.0 and 11.0, under the experimental conditions: fast mix 150 rpm/3 min, slow mix 20 rpm/20 min, temperature 298 K, sedimentation time 12 h. After the sedimentation, the supernatant was withdrawn from a point located 2 cm below the top of the liquid level of the beaker to determine the DOC, turbidity and TSS removal; 2) Pre-determined optimum values (pH and dosage) of NOC and ferric chloride obtained in (1) were added to the red WW samples. Then the stirring process (rpm/min) was varied under different fast mixslow mix conditions (Table 3). After the sedimentation time, the supernatant was withdrawn to determine the DOC, turbidity and TSS removal.

Photocatalytic experimental set-up
The AOPs were performed in a cylindrical photoreactor with a volume of 500 mL. The reagents were added to the WW with a DOC 0 = 400 mg C/L, under constant agitation (350 rpm) at ambient temperature (298 K) with a reaction time of 240 min. For the reactions involving radiation, it was coupled to the reactor, a UV-C low pressure mercury vapor lamp (TNN 15/32) -working power = 15 W (795.8 W/m 2 ) and λ max = 254 nm (Heraeus, Hanau, Germany). The oxidation process was performed, as follows:   At the end of the experiments, sodium sulfite anhydrous was applied to quench the reactions. To determine the removal percentage of the parameters in study, it was applied Equation (9). Let X i be defined as the removal of a given indicator (%) of water contamination (turbidity, TSS, DOC, COD and total polyphenols) that is achieved by a treatment (Peres et al., 2004): where C 0 and C are the initial and final concentrations, respectively, of parameter i. 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).

Characterization of NOCs
All the five NOCs were analyzed by FTIR to assess the nature of the active groups in the powder. By observation of Fig. 1, three bands were detected at 3444.9, 3390.9 and 3344.6 cm −1 , which could be attributed to the presence of hydroxyl groups (OH) present in the proteins, fatty acids, carbohydrates and lignin (Baptista et al., 2017;Kushwaha et al., 2014;Vunain et al., 2019). The band 2922.2 cm −1 could be attributed to C-H stretching of aliphatic structures assigned to fatty acids and lipids (Chung et al., 2018). The band 2848.9 cm −1 could be attributed to symmetric and asymmetric C-H vibration of methyl groups (Kushwaha et al., 2014) present on lipids, proteins (Chung et al., 2018), tannins and tannic acids (Prozil et al., 2014). The band 1741.7 cm −1 could be associated to the C--O stretching in ketones (Vítis vinifera L.) (Mageshkumar and Karthikeyan, 2016), and lipids (Zimmermann et al., 2017). The bands 1662.6 and 1518.0 cm −1 were associated to the presence of N-H bending and C-N stretching of amide I and amide II respectively, and proteins. The band 1425.4 cm −1 could be associated to O-H bending from cellulose. The band 1165.0 cm −1 could indicate the C-O stretching from polysaccharides (Rana et al., 2018). The band 1022.3 cm −1 indicated the C-O stretch which was an indication of the lignin structure (Kushwaha et al., 2014), or the presence of aliphatic amines with C--N type bond (Mageshkumar and Karthikeyan, 2016). The band 775.4 cm −1 could indicate the presence of N-H bending form proteins (Rana et al., 2018). It was also observed many peaks from 590.0 to 440.0 cm −1 which confirmed the presence of halogenated alkanes (Mageshkumar and Karthikeyan, 2016). Fig. 2 shows the physical structure of the NOCs. It is observed that NOCs have a heterogeneous and disformed surface, with a porous morphology, which provides a lot of free spaces available, that can be important because they can facilitate the adsorption of particles with smaller size. It is also observed by the SEM images that NOCs present a chainlike a spherical structures, which are similar with the SEM images of Moringa oleifera seeds, observed in the work of Boulaadjoul et al. (2018).
The mineral composition of the NOCs was also studied. Iron and copper were observed to be the lowest minerals in NOCs constitution (Table 4), with a maximum of iron detected on DC seeds (6.3 mg/L) and a maximum of copper detected in DC seeds and rachis (1.2 mg/L). Regarding sodium, potassium, calcium and magnesium, it was observed on Table 4 high concentration levels on all NOCs. DC seeds were observed to have higher concentration of these minerals. Iwahashi et al. (1982) observed in studies performed on leaves, petioles and stems of cucumber plants, that during the plant's growth, the leaves lose most of these minerals. However, in petioles and stems the concentration remains almost the same during the stationary phase, which could explain the concentrations detected on the seeds and rachis.
The specific surface area and pore characteristics of NOCs are shown in Table 5. These results showed that all NOCs had a small specific surface area, ranging from 0.00 to 0.45 m 2 /g, which indicates that NOCs are organic materials without mesoporosity. The shape of its N 2 adsorption-adsorption isotherm was of type I isotherm typical of microporous solids having relatively small external surfaces as defined by the International Union of Pure and Applied Chemistry (IUPAC) classification (Thommes et al., 2015).

Effect of pH vs coagulant dosage
The WW was observed to have a high content of turbidity and TSS (Table 1), therefore, a CFD process was applied to reduce these parameters. Considering the lack of information regarding the efficiency of the NOCs developed for the treatment of WW, in this step the pH and coagulant dosage were varied and DOC, turbidity and TSS removal was accessed ( Figure S.1). The results showed that NOCs and ferric chloride were influenced by both pH and coagulant dosage. Although, at pH 3.0 it is observed higher DOC removals with the application of NOCs, an evaluation of the turbidity and TSS removal showed that all coagulants achieved their highest efficiency at pH 3.0 as well. However, the dosage required for each coagulant was observed to be different (2.0, 2.0, 0.1, 0.1, 0.1 and 0.5 g/L, respectively, for DG seeds, FA seeds, DC seeds, TV seeds, rachis and ferric chloride).
The actions of NOCs in treatment of the WW lies in the presence of water-soluble dimeric proteins present in the seeds and rachis, as previously detected by the FTIR analysis. These proteins work as natural polyelectrolytes, with a mechanism that combines adsorption (due to the porous structure observed in the SEM images), charge neutralization, particle bridging of destabilized particles, enmeshment and precipitation (Beltrán-Heredia et al., 2012;Camacho et al., 2017;Vunain et al., 2019). Due to these mechanisms, when NOCs are applied to the WW, the protein in the seeds binds to the negatively charged particles that cause turbidity, such as clays, bacteria, dust, among others, to form heavier flocs which precipitate.
When ferric chloride is added to the WW at pH 3.0, a number of parallel and sequential reactions occur. Initially, the ferric chloride is dissociated to yield the trivalent Fe 3+ ions (Equation (10)). The trivalent ion Fe 3+ then hydrate to form the aquometal complex Fe(H 2 O) 3+ 6 . This aquometal then lose protons to form a series of soluble mononuclear species, such as Fe(HO) 2+ and Fe(HO) + 2 (Howe et al., 2012).
These species promote positive or negative charges, depending on

Optimization of mixing parameters on coagulation process
To achieve a good flocculation, the mixing time must be considered  Table 3. The results showed that coagulants DG seeds and FA seeds required faster agitation speed (mix 5) to achieve higher removals. On the other hand, DC seeds, TV seeds, rachis and ferric chloride achieved the best removals wit application of lower agitation speed (mix3, mix 2, mix 2 and mix 1, respectively respectively, for DG seeds, FA seeds, DC seeds, TV seeds, rachis and ferric chloride (Fig. 3).
The application of higher agitation speed affected the performance of the coagulants, because, as the particle size increases, the faster mixing speed agitation broke up the existing flocs as a result of disruptive forces and the collision efficiency of the particles in a shear field decreased (Domínguez et al., 2007). It is also necessary to consider the dynamic   balance between floc growth and breakage which often leads to a steady-state floc-size distribution, where the limiting size depends on the applied shear rate. If the shear rate applied is increased to a certain point, the pre-formed flocs can be broken. In the case of sweep-floc coagulation, this breakage in not fully reversible, and flocs do not completely re-form when the original shear conditions are restored, which explains the decrease in efficiency, when agitation speed is increased.

Effect of AOP variation
In the previous sections, it was observed that CFD process could only remove organic carbon to some extent. Therefore, a more suitable treatment process is required to decrease the DOC present in the WW. In this section, different AOPs ((1) H 2 O 2 , (2) UV-C, (3) UV-C + H 2 O 2 , (4) UV-C + Fe 2+ , (5) H 2 O 2 + Fe 2+ (Fenton process), (6) H 2 O 2 + UV-C + Fe 2+ (photo-Fenton process)) were tested, to evaluate which is the most efficient in the degradation of DOC. In Fig. 4(a) it is shown the DOC removal in function of time and in Fig. 4(b) is shown the H 2 O 2 consumptions.
The results showed that application of H 2 O 2 , UV-C + Fe2+ and UV-C had little oxidation power, revealing low organic carbon degradation (12.3, 13.7 and 15.6%, respectively). The combined action of H 2 O 2 + UV-C achieved higher DOC removal with 32.9% in 240 min. The Fenton process, which involved the application of H 2 O 2 + Fe 2+ achieved a DOC removal OF 22.9% after 240 min. The application of photo-Fenton process achieved the highest DOC removal, with 84.9%.
Considering these results, the relative efficiencies of the processes tested above are in the following order: The highest efficiency observed with the photo-Fenton process is due to the production of more hydroxyl radicals (HO • ) than Fenton process and other processes, due to the higher consumption of H 2 O 2 , as observed in Fig. 4(b). These results indicate that the UV-C lamp has a synergistic effect when combined with the Fenton process enhancement in the degradation of the organic matter.
With the combination of UV and H 2 O 2 the chemical bonds of the organic compounds present on the WW can be directly photolyzed by UV radiation and rapidly improved in the presence of H 2 O 2 (Ince et al., 1997;Teixeira et al., 2022). Adding Fe 2+ to UV-C and UV-C light can remove DOC, but to a smaller percentage. At last, the H 2 O 2 alone has the lowest influence on DOC removal, thus the results allow the conclusion that the oxidizing power of H 2 O 2 is not enough to bring about the degradation of organic carbon.
In all the experiments, the DOC degradation can be described as a pseudo first-order kinetic, in which ln[DOC] = -kt + ln[DOC] 0 and a plot is charted applying ln[DOC/DOC 0 ] as a function of time (t). regarding TOC concentration as it is observed from the data in Table 6. Also, to have a better knowledge on the removal of organic carbon, it is presented the time necessary to reduce to 50% the initial concentration of total organic carbon from the WWthe half-life time (t 1/2 = 0.693/k).
As it can be seen in Fig. 4 and comparing k values and t 1/2 in Table 6, photo-Fenton process is the most efficient process on DOC removal regarding other processes. This fact is due to the synergistic effect of UV-C radiation. The low-pressure mercury vapor lamp TNN 15/32 has little power, only 15 W, however the use of a simple UV lamp in Fenton reaction is very important to increase the efficiency of DOC degradation. This is because the UV light wavelength can significantly influence direct formation of HO • radicals by photolysis of the H 2 O 2 (as observed in Fig. 4(b)) as well as the photo-reduction of Fe 3+ to Fe 2+ , increasing the reaction's kinetic rate. In order to have a complete knowledge of the optimal conditions for the process of photo-Fenton process, the next steps of this work are to assess the capacity of photo-Fenton process to oxidize organic carbon, evaluating each parameter fixing the values of the others.

Optimization of pH on photo-Fenton oxidation
In this section, the pH was varied (3.0-11.0) under the following operational conditions: [DOC] 0 = 400 mg/L, [Fe 2+ ] = 2.5 mM, agitation 350 rpm, reaction time 240 min, radiation UV-C lamp (254 nm), temperature 298 K. Fig. 5(a) shows that the pH significantly influences the mineralization of organic carbon by photo-Fenton process, with the highest removal obtained at pH 3.0 (84.9%). With application of pH 5.0, 7.0, 9.0 and 11.0, the DOC removal results showed a significant decrease

Table 6
Pseudo first-order rate kinetics (k) and half-life (t 1/2 ) of DOC removal from WW. Means in the same column with different letters represent significant differences (p < 0.05) within each parameter (k and t 1/2 ) by comparing AOPs. In these reactions three factors influences the oxidation process of photo-Fenton: (1) the iron, (2) the hydrogen peroxide and (3) the pH. The decrease of activity for pH values below the optimum is understandable considering that Fe 3+ forms different species in solution, and the quantum yield of light absorption by Fe 3+ is directly depending on the specific species responsible for the absorption. The main specie observed at pH 3.0, Fe(OH) 2+ (H 2 O) 5 is one with the largest light absorption coefficient and quantum yield for HO • radical production, along with Fe 2+ regeneration in the range of 280-370 nm (Perez et al., 2002), which explains the high DOC removal at pH 3.0. For higher pH solutions(alkaline), the presence of Fe(OH) 2 + (H 2 O) 4 dominates, but the solution becomes unstable with Fe(OH) 3 precipitation (Benkelberg and Warneck, 1995;Perez et al., 2002), which blocks the decomposition of hydrogen peroxide catalyzed by the ferrous iron (Lucas and Peres, 2007). In Fig. 5(b), it is observed that by increasing the pH > 3.0, the H 2 O 2 consumption decreases significantly, thus decreasing the DOC degradation efficiency of the photo-Fenton process. These results are in agreement with the work of Çalık and Çifçi (2022), who observed that Fenton process achieved the highest efficiency in the treatment of a textile industry wastewater at pH 3.0. However, unlike this work, the consumption of H 2 O 2 was much lower (8.82 mM), than the concentration consumed in this work. The difference between the treatment of textile dyes and WW was previously reported by Rodríguez-Chueca et al. (2017), who observed that the degradation of organic matter from WW required more demanding treatments with higher dosages of reagent regarding the treatment of dyes.

Optimization of Fe 2+ concentration on photo-Fenton oxidation
In this section, the effect of Fe 2+ concentration on the mineralization of organic carbon of red WW was studied. The Fe 2+ concentration was varied within the range of 0.5-5 mM ( Fig. 6(a)) and results showed a DOC removal of 81.6, 84.1, 84.9 and 82.5%, respectively, for 0.5, 1.0, 2.5 and 5.0 mM Fe 2+ .
The results indicated that the extent of DOC degradation increased with the increase in initial Fe 2+ concentration, from 0.5 to 2.5 mM. However, the increase from 2.5 to 5.0 mM Fe 2+ decreased the DOC removal. These results can be explained by the redox reactions since HO • and HO • 2 radicals may be scavenged by the excess of Fe 2+ as observed by Equation 11 -Equation (13) (Lucas and Peres, 2009).
Fe 2+ + HO • 2 → Fe 3+ + HO -2 Equation 12   The addition of 2.5 and 5.0 mM Fe 2+ also had the highest consumption of H 2 O 2 , regarding 0.5 and 1.0 mM Fe 2+ (Fig. 6(b)), meaning that DOC removal was directly linked to the consumption of hydrogen peroxide, and at the same time to the production of HO • radicals.

Effect of KPS and PMS addition to photo-Fenton process
In order to increase the rate of DOC degradation, different concentrations of KPS and PMS (0.5, 1.0, 2.5 and 10 mM) were applied in combination with the optimal experimental conditions obtained in Section 3.2.3. The results shown in Fig. 7(a) and Fig. 7(c), revealed a DOC degradation of 91.2 and 88.0%, respectively, for 1.0 mM of KPS and PMS. As previously observed, the sulfate radicals can be activated either by UV radiation or application of a catalyst, and when applied, the reactions described by Equation (14) to Equation (20) takes place in DOC degradation (Khataee et al., 2016), leading to the production of sulfate radicals.
Equation 20 The results showed that KPS was more reactive than PMS, due to the fact that the photolysis of PMS produces 1 mol of sulfate radical and 1 mol of hydroxyl radical per each mole of peroxymonosulphate (Rodríguez-Chueca et al., 2017), and the photolysis of KPS produces 2 mol of sulfate radical (Boczkaj and Fernandes, 2017). Also the efficiency of KPS is explained based on the reactivity of the persulfate anion (S 2 O 2− 8 ; E • = 2.01 V), comparatively to PMS, in which the peroxymonosulfate anion presents lower reactivity (HSO − 5 ; E • = 1.85 V) (Boczkaj and Fernandes, 2017).
The addition of 10 mM KPS and PMS had the lowest degradation of DOC (74.1 and 71.0% respectively). These results can be explained by the competition between the organic matter present in the WW and the excess of persulfate in solution for the SO •− 4 and HO • radicals (Equation (21) to Equation (24)). Similar results were obtained in the works of Khataee et al. (2016) and Xu et al. (2019), in the treatment of textile dye acid red 73 and contaminant 2,4-dinitrotoluene, by the combined application of persulfate and H 2 O 2 .
One of this work's objective is the study of the separation of wastewaters from the production of red and white WW, to understand how the chemical composition of the WW influences the efficiency of the AOPs. In this section, the photo-Fenton, KPS-photo-Fenton and PMS-photo-Fenton were applied in the treatment of a white WW with the same DOC concentration (400 mg C/L) as the red WW. The results presented in Fig. 8(a), shows that application of photo-Fenton reached a DOC removal of 86.5% after 240 min of reaction. When KPS and PMS were added to the photo-Fenton process, DOC removal increased to 96.8 and 91.5%, respectively. The effect of KPS and PMS was similar to the treatment of red WW, however, the results obtained with addition of KPS revealed higher. Therefore, these results suggest that the white WW is more sensitive to degradation by KPS-photo-Fenton. These results support the initial objective, of separating the red and white WW, to facilitate the DOC degradation.

Application of combined CFD-AOPs
In the previous sections it was observed that the CFD process had a good capacity to decrease the solid particles in suspension in the WW, however, low DOC removal was observed with the application of this process. With application of AOPs, DOC removal achieved higher results, however, several disadvantages were associated to this process, such as long reaction times and high H 2 O 2 consumption. In this section, the CFD process was applied to the red and white WW as a pretreatment. This pre-treatment allowed a high removal of turbidity and TSS which were responsible for the WW dark color. In addition to turbidity and TSS, total polyphenols (TPh) removal were also evaluated, with results showing a 77. 8, 84.5, 80.9, 89.7, 80.9 and <0.5%, respectively, for application of DG seeds, FA seeds, DC seeds, TV seeds, rachis and ferric chloride in red WW. With application of NOCs to white WW, it was observed a 55.4, 76.1, 88.0, 92.2, >99.5 and <0.5%, respectively (Fig. 9). With exception to ferric chloride, NOCs revealed to be very efficient in the removal of TPh from WW. Considering that these NOCs were never applied before in the treatment of WW, the effects in TPh removal were still unknown, however, these results were similar to other plant-based coagulants such as derived from Moringa oleifera (MO). In the work of Rifi et al. (2022), a coagulant derived from MO showed a 96 and 86% removal efficiency of turbidity and TPh from a olive mill wastewater. The removal of TPh content is important because they are responsible for the WW color and represent a toxicity to the environment, due to the difficulty of microorganisms to degrade cycling compounds (Jorge et al., 2021(Jorge et al., , 2022a. The removal of turbidity, TSS and TPh from the WW by the NOCs influenced the DOC removal. The results in Fig. 10(a), showed that the application of NOCs in combination with KPS-photo-Fenton for the treatment of red WW achieved a DOC removal of 96.0, 95.9, 94.6, 94.6 and 96.3%, respectively, for DG seeds, FA seeds, DC seeds, TV seeds and rachis. When this treatment was applied to white WW ( Fig. 10(b)), it was observed a DOC removal of 97.4, 97.9, 97.4, 97.7 and 98.2%, respectively. NOCs improved DOC removal, clearly driven by the removal of turbidity, TSS and TPh from the WW, which from one hand blocked the radiation, reducing the rate of regeneration of Fe 3+ into Fe 2+ and on another hand, the particles in suspension competed for the HO • radicals. These results are in agreement with the work of Amor et al. (2015), who observed that the excess of sediments in the landfill leachate reduced the efficiency of photo-Fenton process and that combining CFD process with photo-Fenton increased the DOC removal.
When ferric chloride was applied, the opposite effect was observed, as DOC removal achieved 89.8 and 93.7%, respectively, for red and white WW, which are lower than the values obtained by application of KPS-photo-Fenton without CFD pre-treatment. These results can be explained by the presence of the excess of ferric iron present in the WW after treatment with ferric chloride (106.0 and 103.3 mg Fe/L respectively, for red and white WW), which acted as scavenger of SO •-4 and HO • (Lucas and Peres, 2009;Wang and Chu, 2012) radicals as observed on Equation (11) and Equation (25), which in turn decreased the KPS-photo-Fenton process efficiency on carbon degradation.

Equation25
Although the results showed high DOC removal with the application of combined CFD-photo-Fenton process, this parameter indicates only the dissolved carbon in the WW. Therefore, a more conclusive analysis is required, such as COD and BOD 5 , which indicates the carbon that is dissolved and undissolved. The application of NOCs in the treatment of red WW showed a COD removal of 24.9,29.5,29.5,33.4,25.6 and 49.7%,respectively (Fig. 11(a)). When NOCs were applied for the treatment of white WW, results showed a COD removal of 26.6,23.1,33.7,24.8,52.4 and 35.5%,respectively (Fig. 11(b)). The BOD 5 was also investigated, with results showing a removal of 1. 8, 1.8, 0.5, 12.7, 5.5  and 75.0%, respectively, for red WW and 20.0,32.5,22.9,22.9,83.9 and 33.6%, respectively for white WW. These results can be related with the removal of turbidity, TSS and TPh, in which their removal led to the removal of the undissolved carbon. With the application of KPS-photo-Fenton, the COD and BOD 5 removal reached the legislated value for wastewater discharge (150 and 40 mg O 2 /L, respectively) in accordance with the Portuguese Decree Law n • 236/98. The biodegradability (BOD 5 /COD) of the WW was evaluated, considering that both red and white WW presented low biodegradability (0.26 and 0.37). The application of NOCs was observed to have increased the biodegradability of red and white WW. With application of KPS-photo-Fenton, the biodegradability increased further in red WW pre-treated with DC seeds and rachis (0.61 and 0.98, respectively) and in white WW pre-treated with FA and DC seeds (0.58 and 0.75, respectively). These results were in agreement with the work of Barros et al. (2020), who observed that the combination of CFD process with Fenton process increased the biodegradability of the coffee processing wastewater (CPW) from 0.34 to 0.47. The results obtained in this work also showed that the matrix of the wastewater is important, considering that ferric chloride decreased the biodegradability of red WW (0.26) and increased the biodegradability of white WW (0.38). Although the combined CFD-KPS-photo-Fenton processes could improve wastewater biodegradability, it was not advisable to introduce the WW into a biologic process, because the COD concentration remaining was very low. Ideally, the photo-Fenton process should be used until legal discharge limits are achieved.

Energy consumption and operational cost
The combination of CFD with KPS-photo-Fenton proved to be very efficient in terms of organic matter degradation, however, these data are nor enough to answer the question of whether these processes are economical in terms of energy consumption and operational cost. To evaluate the energy consumption of the system, the electric energy per g of DOC (E EM ) was determined (Equation (26)).
Equation 26 This parameter is defined as the electric energy in kilowatts hours (kWh) required to degrade 1 g of DOC in wastewater, where P is the power of the UV-C lamp (kW), t is the time (h), V is the volume (dm 3 ) of the treated wastewater and DOC i and DOC f are the initial and final concentrations of DOC, respectively (Dükkancı, 2018). To access the cost of the treatment, the E EM was multiplied by medium value of electricity cost in Portugal (0.080 €/kWh). The reagent cost was also access, with reagents costs assumed as: 0.25 € L −1 for H 2 O 2 solution 30% (w/v), 0.082 € kg −1 for FeSO 4 •7H 2 O and 0.018 € kg −1 for KPS . The application of the combined CFD + KPS-photo-Fenton process increased the E EM values regarding the application of KPS-photo-Fenton process in both red and white WW treatment (Table 7), which undoubtably increased the cost associated to the energy consumption. This fact is related to the degradation of lower concentrations of DOC, due to the previous removal by CFD, in a similar time. However, when the total cost is analyzed, the results show a significant decrease of the total costs with application of the combined process. The only exception was the application of ferric chloride as a coagulant, which increased significantly the cost of treatment. One explanation is the reduction of the H 2 O 2 used for the reaction within the application of the combined process, which decreased the costs. When ferric chloride was applied, the excess of iron acted as a scavenger for HO • , decreasing the efficiency of the reaction, and at the same time increasing the costs of the WW treatment. These results are in agreement with the work of Lak et al. (2018), who observed that the application of the CFD process as a pre-treatment decreased the energy consumption of the oxidation process, in the treatment of landfill leachate.
Based in the results obtained in this work, it could be possible throughout the control of a certain number of variables in CFD (pH, coagulant concentration, mixing conditions, sedimentation time) and AOPs (pH, oxidant and catalyst concentrations and radiation source), to apply these treatments in an upgraded scale, with a reduced operational cost. However, certain challenges need to be in concern when upscaling the treatment from batch to pilot, such as (1) treatment in plants operated in a continuous mode, (2) higher concentrations of coagulant are required in full-scale wastewater treatment and (3) controlling the ratio of oxidants used in AOP treatments, which could lead to higher treatment costs. However, the availability and the low cost related with the plant sub-products will reduce the treatment cost of the oxidation processes, overcoming the usually significant costs of agro-industrial wastewaters treatments.

Conclusions
This work addresses the treatment of a WW by application of a physical-chemical process (CFD) followed by a chemical process (AOP). To perform CFD process, 5 different NOCs were developed and by FTIR analysis it is concluded that these materials are organic in nature, with proteins, fatty acids, carbohydrates and lignin in their constitution. In addition, SEM analysis shows that NOCs are porous materials, with adsorption capacity. Based in the results, it is concluded that NOCs are dependent on the pH of the WW, the dosage applied and mixing conditions, reaching a turbidity removal between 86.2 and 98.9%, a TSS removal between 85.0 and 94.9% and a TPh removal between 77.8 and 89.7%. Different AOPs were applied and with the results obtained, it is concluded that photo-Fenton process is the most efficient process. The photo-Fenton process was optimized, with application of KPS, with results showing an increase of DOC removal from 84.9 to 91.2%, respectively. In this work it was also explored the treatment of a red and white WW separately, with results showing a DOC removal of 91.2 and 96.8%, respectively, with application of KPS-photo-Fenton, thus it is concluded Fig. 10. Influence of CFD and KPS-photo-Fenton process in DOC removal of (a) red WW, (c) white WW and H 2 O 2 consumption of (b) red WW, (d) white WW. Fig. 11. Effect of CFD and CFD + photo-Fenton (PF) processes in the CFD and BOD 5 removal and biodegradability of (a) red WW and (b) white WW. Means in bars with different letters represent significant differences (p < 0.05) within biodegradability (BOD 5 /COD) by comparing treatment processes. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.) Table 7 Electric energy per mass (E EM ), energy cost, reagent cost and total cost after treatment of red and white WW. Means in the same column with different letters represent significant differences (p < 0.05) within each parameter (E EM , energy cost, reagent cost and total cost) by comparing treatment processes.