The samples were collected in the open air in a 20 L high-density polypropylene container and kept under controlled conditions. Rainwater samples were collected during different rainy days in a rural area, located at coordinates 19–40′04.93′N and 99'.71‴56.69′E in Toluca, State of Mexico, Mexico. The rainwater collection was carried out during the rainy season of 2019, from May to September. Point A in Figure 1 corresponds to a rural area. This region is surrounded by agriculture, livestock, and some industrial zones. So, these multiple anthropogenic activities emit different pollutants into the atmosphere, which subsequently precipitate with the rain.

The batch EC experiments were conducted with a monopolar cell in a 600 ml glass reactor. The volume treated was 500 ml of rainwater (Figure 2), the sample pH was 6.25 ± 0.15, sea salt and sodium sulfate were added as supporting electrolytes at 5 × 10⁻³ M (0.034 g L⁻¹), and experiments were carried out at 400 rpm and 293 K. The electrochemical cell consists of an aluminum anode and cathode. The dimensions of the Al electrodes were 20 × 2.5 × 0.2 cm, with an anodic area of 50 cm², and the distance between the electrodes was 1.0 cm. The current intensity: 0.05—0.15 A (1 and 3 mA/cm², of current density, respectively), was supplied with a Steren power source and monitored with a Steren multimeter model UL-28 at different times between 0 and 17 min.

The reagents used as supporting electrolyte were food grade sea salt and sodium sulfate Anhydrous Powder (J.T. Baker, Analytical Grade). The samples were characterized before and after treatment, according to the standard methods (American Water Works Association, and Water Environment Federation et al., 2017), pH (pH meter HANNA HI98190), turbidity (HF Scientific^® Micro100 Laboratory Turbidimeter), chemical oxygen demand (COD), biochemical oxygen demand (BOD₅), nitrates, and ammoniacal nitrogen (HACH DR5000) and metals, including Mn, Fe, Cu, Zn, Pb, Na, and Al, were determined using Atomic Adsorption (SpectrAA 240 FS spectrophotometer).

The equipment used for the physical chemical analyses were spectrophotometer UV-VIS 1, Model: DR 6000, Brand: HACH, spectrophotometer UV-VIS, Model: Cary 1E, Brand: Varian, conductometer CG/EM, Model: Orion Star A212, turbidimeter Model: micro 100 turbidimeter, Brand: Itfscientific.inc. and potentiometer Model: ION 7000 Brand: OAKTON. Analytical balance: BEL engineering.

Optimization of the electrochemical process using the RSM has been used due to the possibility of optimizing several parameters simultaneously in order to obtain the experimental design and mathematical modeling (Suárez-Escobar et al., 2016). A CCD was used to evaluate the factors that significantly affect and optimize the operating conditions for the removal of pollutants from rainwater by the EC system. The variables studied are described in Table 1.

The proposed experimental design consists of two factors (time and current) and a categorical factor (type of electrolyte A and B), with two blocks of 13 treatments. The experimental design and the statistical analyses were performed using Minitab 19 (Minitab Inc. United States) and the Design Expert^® program (Design Expert, Stat-Ease, Inc. United States). As response variables, COD and turbidity were monitored for their availability and viability in the determinations, both to find the best conditions and for comparative purposes with other rainwater treatment methods.

The factors used for the design of experiments were selected according to the literature (Mousazadeh et al., 2021), focusing mainly on current (since the reactions to electrogenerate the coagulant depend on it), the reaction time (because of the amount of coagulant), and the supporting electrolyte (which depends on it too), to determine the effectiveness of the applied current.

The characterization of the rainwater collected indicated the presence of contaminants such as heavy metals, COD, turbidity, nitrates and ammoniacal nitrogen.

According to the obtained results, the rainwater presented a slightly acidic pH of 6.25, although it is possible that this increase is due to the expected difference between samples. The existence of acid rain is defined by the pH of the rain; thus, the rain that falls on the ground is considered acid rain when its pH is lower than 5.6. The causes of acidity in rain can be from anthropogenic or natural activities. Anthropogenic activities, such as the burning of fossil fuels from transportation or industrial activities, release sulfuric acid and nitric acid into the atmosphere, which are known as inorganic acids. Meanwhile, sources of rain acidity in rural areas come from volatile biogenic organic carbon emissions, such as formic acid, oxalic acid, isoprene, propene, acetic acid, formic acid, or oxalic acid (Payus et al., 2020).

The electrical conductivity is 111 ± 1.5 μS/cm, which shows a low concentration of ions due to the anions, which include 3.82 mg/L NO₃ ⁻ and 14 mg/L N-NH₃. Corresponding to the US EPA Report on the Environment from 2014, most airborne NOx originates from combustion-related anthropogenic emissions sources, primarily fossil fuel combustion in electric utilities, high-temperature operations at other industrial sources, and from motor vehicles. It is worth noting the use of home furnaces and gas stoves, which may also produce a considerable quantity of NOx (Keresztesi et al., 2020). Heavy metals were characterized, and 840 μg/L Zn, 130 μg/L Al, 70 μg/L Mn, 32 μg/L Fe, 24 μg/L Cu, and 30 μg/L Pb were found, presenting higher values in contrast with those reported by Rastegari Mehr et al. (2019) This was with the exception of Zn, for which they report higher values (1,320 μg/L Zn) due to several anthropic activities such as vehicular, agricultural and industrial activities, as well as the wind that transports pollutants.

The organic pollutants were also characterized as: 52 mg/L COD, 5.07 mg/L BOD, the poor biodegradability index was 0.097 and a turbidity of 8 NTU. The anthropogenic sources of organics, anions and cations can be from biomass burning, fossil fuel combustion, soil properties and rock weathering, including fertilizers, forest fires and wind-blown dust (Rastegari Mehr et al., 2019; Payus et al., 2020).

Once the electrocoagulation experiments were carried out, the values of the final measurements of COD and Turbidity were acquired after the process. For each sample, the removal of COD and Turbidity was calculated and fitted to a polynomial model to obtain the optimal value of removal percentage for sea salt, as shown in Figures 3, 4.

1) % Removal COD = 113.83—12.22 J—3.23 t + 3.14 J² + 0.104 t² + 0.426 J^∗t

The statistical analyses were presented by response variable (COD and turbidity), and the optimal conditions were obtained simultaneously with the Minitab 19 software. The results of the analysis of variance (ANOVA) are presented in Table 2 to evaluate the proposed models, which involves the coefficient of determination (R²), the F value and the p value; when the latter are less than 0.05, they indicate that model terms are significant. In this case, electrolyte support, current density, J*E, J² and t² are significant model terms at the 95% confidence level (Keshtkar et al., 2019).

According to the ANOVA, the correlation coefficient R² were 0.97 and 0.55 for the COD removal efficiency and turbidity models, for Eqs 1, 2 respectively, which implied that the COD removal efficiency model can establish an adequate correlation between the independent variables and the responses. The equation in terms of actual factors can be used to make predictions about the response for given levels of each factor. On the other hand, for the turbidity removal efficiency model it is not likely to establish an adequate correlation. In addition, the R²adj values for COD removal efficiency and turbidity were 0.96 and 0.39 respectively, showing that only less than 4% and 61% of the variation could not be described by the models.

Figure 5A represents the response values for COD removal when the variables t (min) and the current density interact, allowing the optimum operating conditions to be established as: 3.26 mA/cm² of density current, pH 6.25, treatment time 11.38 min with sea salt, reaching 100% COD removal for the rainwater treatment system. Figure 3B contains color contour bands representing ranges of the fitted response values. The highest values of COD removal for rainwater are found on the right side of the graph, which coincides with the high values of mean current density and the relationship with contact time. The lowest COD removal values for rainwater are found on the left side of the graph, which coincides with the low values for current density. Figure 3C shows the Pareto graph, which confirms what was obtained in the ANOVA, where it is observed that the current density, the supporting electrolyte, the quadratic term of the current density and the interaction between the current density and the supporting electrolyte have a significant effect on the removal efficiency of the COD.

Figure 4A represents the response values of the turbidity removal when the variables t (min) and current density interact, which allows establishing the optimal operating conditions for the rainwater treatment system. Figure 6B contains contour bands representing ranges of the adjusted response values. The highest turbidity removal values for rainwater are found on the left side of the graph, which coincides with the highest values of current density and the relationship with contact time. The lowest turbidity removal values for rainwater are found on the left side of the graph, which matches with the highest values of current density and the relationship with the lowest value of contact time. Figure 6C shows the Pareto graph, where it is observed that the current density and the supporting electrolyte have a significant effect on the Turbidity removal efficiency.

Electrolysis time is an underlying factor leading to a significant difference in the removal efficiency. Figure 3 illustrates the removal efficiency of COD₀ = 52 mg/L at pH 6.55–7.89, with treatment time from 0 to 16 min at 3 mA/cm². The removal efficiency reached approximately 99.9% at 10 min, with a 25 mg/L Al dose. According to the kinetics, at 11 min the complete removal of COD was achieved, so no additional time is required. The produced amount of metal ions was calculated by Faraday’s Law equation (Eq. 1): m = I t M F Z (1) where M and Z in Eq. 1 represent Al molar mass (g/mol), and oxidation state in solution, respectively, and I the current (A). F is the Faraday’s constant equal to 96,485 C/mol. The amount of aluminum produced is m = 12.5   m g Al, the total volume is 0.5 L, then 25 mg/L Al was the optimal dose (72 mg/L Al(OH)₃). Additionally, a study carried out for the treatment of wastewater with the generation of 0.17 g of Al to treat 500 ml of wastewater for 15 min and with 0.04 A is reported (Molano-Mendoza et al., 2018; Padmaja et al., 2020).

In addition, Figure 3 shows the kinetic COD, kinetic constants and the correlation coefficients fitted to the second order model k = 0.4828 L/mg min⁻¹ and R² = 0.9925 C = d C d t = k 2 t 2 (2)

According to the species distribution diagram performed via medusa program (Supplementary Figure S1), the predominant species in the pH range monitored during treatment (6.25–7.89) is Al(OH)₃ which promotes coagulation-flocculation of colloidal particles present in rainwater, promoting the removal of COD, turbidity and some metals.

Eqs 3–5 show the electrochemical reactions and solution reactions carried out in the batch EC process that involve the oxidation of Al to Al³⁺. The H₂O reduction to H₂ and OH⁻ions is carried out on the cathode. The solution reactions show the formation of metallic hydroxides. Because aluminum hydroxide is an amphoteric hydroxide, the pH is a vulnerable factor for the formation of Al(OH)₃ (Eq. 5) flocs. The solubility of aluminum is in balance with the solid phase Al(OH)₃ depending upon the surrounding pH, since solid Al(OH)₃ is more common among pH 4.5 and 12, and higher pH 10, the soluble species Al(OH)₄ ^– is the prevalent species, as shown in the species distribution diagram (Eq. 6).

Anodic reactions: A l → A l 3 + + 3 e − (3)

Cathodic reaction: 2 H 2 O + 2 e − → H 2 ↑ +   2 O H − (4)

Solution reactions: A l 3 + + 6 H 2 O → A l ( O H ) 3 ( s ) ↓ + 3 H 3 O + (5) A l ( O H ) 3 ( s ) + O H − → A l ( O H ) 4 − (6)

During the electrogeneration of the coagulant (Al(OH)₃ (Eq. 5), small flocs are formed where both the colloidal organic matter and the metallic ions of the rainwater are adsorbed. The generation of H₂ bubbles in the cathode due to the reduction of water (Eq. 4) benefits the separation of the light flocs by the electroflotation process. Additionally, the bubbles contribute to the hydrodynamics of the reactor through convection, favoring quick contact among the species so that with the treatment time, the flocs increase in size, generating the coagulation process, and finally precipitation occurs (see Figure 6) (Mousazadeh et al., 2021; Shakeri et al., 2021).

The experiments were carried out with a combination of Al/Al electrodes in an electrolysis time range of 2.93–17 min. The effect of current (0.58–3.40 mA/cm²) was evaluated on COD and turbidity according to the experimental design (Table 3). From the results obtained, Figures 5, 6 show that the removal efficiency increases with the increase in current density with a time of 15 min, reaching 100% removal for COD and turbidity with 3.26 mA/cm2 of current and 11.26 min (Padmaja et al., 2020). The increase in removal efficiency with the higher current (3 mA/cm²) is due to the greater amount of coagulant resulting in more active sites for COD removal. Likewise, more oxidation of the anode is generated and, therefore, more Al ions according to Faraday’s law. In addition, as more Al is produced, it reacts with more OH⁻ ions to create aluminum hydroxide, which acts as an adsorbent that can adsorb/remove contaminants (Negarestani et al., 2020). In this study, the COD content decreased from 52 to 0 mg/L, showing to be more efficient compared to what was reported by Xu et al. (2021), who reported a decrease of 1.7 to 1.0 mg/L of COD through electrocoagulation coupled with a gravity-driven membrane for rainwater treatment (Du et al., 2019; Xu et al., 2021).

One of the most important parameters for the proper functioning of electrochemical cells is the supporting electrolyte, which ensures that the ionic conductivity is sufficient, and the applied current is used to the maximum (Maitlo et al., 2018).

According to the data in Table 3, the composite sample of rainwater represents 100% removal for turbidity and COD parameters using sea salt as the electrolyte, and 70.83% for COD and 99.27% for turbidity with sodium sulfate under the same conditions of treatment time (15 min) and density of applied current (3 mA/cm²) in both cases.

In this particular case, the two electrolytes used were 0.05 M Na₂SO₄ with a conductivity of 237 μS/cm, and for sea salt, the amount equivalent to that of sodium sulfate was added, registering a conductivity of 305 μS/cm; the ionic conductivity increases because in the composition of the sea salt there is 86% NaCl, so conductivity is improved and therefore the applied current is used properly during the treatment. This gives rise to reactions 3 and 4 for the electrogeneration of the coagulant with greater efficiency, thus achieving in a short time total elimination of COD and turbidity. Sea salt provides some additional benefits, like stabilizing the electrochemical process, whilst also providing other essential minerals such as Na, Cl, Mg, S, Ca, Br, and K for human use and consumption.

The Cl⁻contents are another advantage of electrocoagulation using sea salt, as the electrogeneration of Cl₂ (Eq. 7) is carried out at the anode, and subsequently ClO⁻is produced, which act as disinfectant (Yildiz et al., 2008; Jaiyeola, 2017). 2   C l −   → C l 2 + 2 e − (7) C l 2 + H 2 O → C l O − + H + + C l − (8)

Therefore, the generation of Cl₂ can act in two ways: 1) it can directly oxidize organic matter (including the cell wall of bacteria) and 2) Cl₂ with water generates ClO⁻which can continue to oxidize organic matter, that at the same time is adsorbed in the generated coagulant and is finally separated by flotation or sedimentation. Thus, it can be suggested that a disinfection was carried out (Aguilar-Ascon, 2019), since both parameters can be related to the elimination of possible microorganisms found in the suspended particles that generate turbidity and at the same time provide organic matter. Furthermore, it has been reported that rainwater contains a wide variety of microorganisms (Yildiz et al., 2008; Jaiyeola, 2017).

Some studies show different treatments and removal efficiencies for rainwater; Aljerf (2018) reported a modern PVC-filter filled with reused glass and crushed foam glass for the removal of major contaminants in rainwater to improve the physical properties and enhance the matrix of this potential source of water with important minerals. The turbidity parameter was 8.3 NTU, and total removal of COD after treatment was achieved. Heavy metals were also decreased: Al (100%), Mn (84.3%), Zn (98%), Fe (21.9%), and Pb (46.7%). Leong et al., 2018, reported the following removal data for the treatment of rainwater: COD (56% of 71 mg/L), turbidity (38% of 3.8 NTU), Cu (63%) and Zn (62% of 0.013 mg/L).

Table 4 shows the final characterization of rainwater before and after EC treatment, under optimal pH conditions set to 6.25, treatment time 11.38 min at 3.26 mA/cm² with 25 mg/L Al dose. The electrical conductivity was increased due to supporting electrolyte addition, from 115.16 to 256 μS/cm (+130.32%), mainly Na⁺ (from 0.06 to 17.52 mg/L). This value improves the physical-chemistry quality of rainwater according to the European Community. Final pH was increased at 8.26 ± 0.5. Turbidity was reduced from 8.0 to 0.08 (99.99%). COD, BOD, N-NH₃ were 100% removed. Metals were reduced significantly: Al (100%), Mn (84.29%), Fe (21.87%). Cu was slightly increased from 24 to 35 μg/L (48.5%), although this value is under the permissible limit. Zn was reduced from 840 to 17 μg/L (97.97%), Pb from 30 to 16 μg/L (46.6%).

For EC process the electrical energy costs were estimated as shown in Eq. 9: C e n e r g y   ( k W h m 3 ) = U   x   I   x   t E C 1000 ∗   V (9) where U is the cell voltage (2.4 V), I is the current (0.15 A), t_EC is the time of electrolysis (0.25 h) and V is the volume (0.0005 m³). With these values, energy consumption was 0.18 kWh/m³ (Taeyeon Kim et al., 2020) for EC treatment and the monetary energy production rate in Mexico is US $ 0.08 for industrial use per kWh, the energy cost is $ 0.014/m³.

Therefore, the treatment costs are accessible for the implementation of rainwater treatment, compared with other conventional methods such as aerated lagoons, where it is required to apply energy to air generators. Subsequently, it requires a disinfection process, with the produced sludge increasing costs (Mcnaughton et al., 2011). Rainwater filtration is an efficient treatment; however, it also requires disinfection.

It is important to mention that in the literature consulted, no information was found on EC treatments with rainwater, therefore there is no point of comparison regarding costs. However, there are publications that indicate the costs of treating wastewater from various source with EC are varied, as shown in the Table 5.

The energy consumption and costs calculated in this study were lower than reported in other studies, although they used to treat different types of wastewater, where costs are between 1.04 and 5.25 USD/m³ and energy consumption ranges from 4.5 to 95 kWh/m³ (Asselin et al., 2008; Geraldino et al., 2015; Sahu & Chaudhari, 2015; García-Orozco et al., 2021).