The Brazilian Ministry of Health Decree n. 2914/2011 [28] constitutes the most recent legal arbitration instrument on the potability of drinking water in Brazil. However, simply complying with the quality standards established by this regulation does not always guarantee that the treated water is, in fact, in consumption condition due to the successive appearance of new chemical and microbiological pollutants. This indicates that conventional unit processes may no longer be sufficient to remove a wide variety of pollutants from various sources. More recent, complete, and selective technologies have emerged as possible solutions to solve the problem. Therefore, it was decided to compare the environmental performance of the drinking water route from effluents to those of the conventional route (chemistry) and other variants comprising the biological route and treatment by ultrafiltration. Although still underexplored, the latter options are promising.

Table 1 describes the quality standards for different raw material sources (surface water and domestic effluent) used by the different treatments and characterizes their product (drinking water). The indices referring to surface water were obtained from Resolution n. 357/2005 stipulated by the National Council on the Environment [29], which establishes a classification for Brazilian water bodies and sets environmental guidelines for their frameworks.

For this verification, it was recognized that surface water would be collected from Class 3 water bodies, which becomes fit for consumption only after being subjected to conventional or advanced treatment. The requirements from CONAMA Resolution n. 430/2011 [30] served to specify the domestic effluent, while Decree n. 2914/2011 was adopted for drinking water characterization. The data for Biochemical Oxygen Demand (BOD), Total Suspended Solids (TSS), and phosphates (PO₄^3-) for domestic effluents are not arbitrated by Resolution n. 430/2011, as they are specific to each region. Consequently, these parameters were obtained from a sewage treatment plant located in the state of São Paulo. All other values displayed in Table 1 refer to the maximum tolerable limits for each parameter.

For the current analysis, potable reuse technology (REU) is feasible from the association between MBR and RO. Meng et al. [31] warn, however, that the MBR will reach its full efficiency (η ≈ 97%) of removing the total Chemical Oxygen Demand present in the feed stream of that system only if the effluent is subjected to preliminary treatments. This intervention takes place through grate operations, applied for coarse solid removal and filtration, for sand extraction. The gratings consist of a bundle of iron (or steel) bars placed in the transverse plane, perpendicular or inclined to the bottom of the channel from which the effluent flows. The spacing between the grids must be designed to avoid significant load losses during the flow. Sand extraction seeks to protect pumps and abrasion equipment and prevent silting. The sandbox size must be selected so as to prevent the retention of sedimentable organic solids [31].

For the REU system assessed herein, the preliminary treatments consist of a grid comprising a mechanical cleaning system and a sandbox coupled to a rotary sieve and helical conveyer thread. The average electricity consumption for the railing was estimated at 1.71 Wh/m³ drinking water, while the separation process involving the litter box and its coupled devices totalled 13.8 Wh/m³.

MBR is a combination of conventional biological wastewater treatment and submerged membrane filtration. The conceptual approach is technically similar to that applied in a traditional wastewater treatment plant, except for the separation of activated sludge and treated wastewater. Replacing sedimentation that would take place in a clarification tank, this operation occurs in a membrane of the MBR unit [32]. Given its characteristics, the MBR unit holds the largest electric consumption of the whole arrangement, of 534 Wh/m³ [33].

After leaving that stage, the effluent follows for disinfection via ultraviolet radiation. Low-pressure mercury vapor lamps produce UV radiation at λ = 254 nm, the optimal wavelength for disinfection and ozone destruction. The UV lamp never comes into contact with the water, as it is housed in a quartz glass sleeve inside the water chamber, and its specific electricity consumption is about 6.42 Wh/m³. Some ultrapure water systems also use λ = 185 nm UV units to reduce Total Organic Carbon [34].

RO is a technology that uses a permeable membrane to remove particles, molecules, and even ions from drinking water. According to this approach, pressure is applied to overcome the osmotic pressure driven by chemical potential solvent differences. RO can remove different types of dissolved and suspended chemical species and biological ones (i.e., bacteria) from industrial and potable water. The effect is that the solute is retained on the pressurized side of the membrane, and the purified solvent (water) passes to the other side [35]. The RO unit designed for REU demands 686 Wh/m³. The treatment is completed with a new application of UV radiation, which is performed alongside hydrogen peroxide (H₂O₂) to complete water oxidation. The UV/H₂O₂ system must act as an Advanced Oxidative Process and, therefore, this intervention requires an energy expenditure of 151 Wh/m³ [33], [36].

The REU design was carried out considering the quality specifications of the domestic effluent and drinking water (Table 1) for a total product flow of 3600 m³/h. In addition, it was admitted that the removal of organic matter would occur through nitrification. For modelling the phenomena that occur during the process, equations, stoichiometric constants, and kinetic coefficients obtained from Warsinger et al. [37] and Henze et al. [38] were used. The grid was specified as a set with bars with thickness t = 15.0 mm and spacing s = 2.00 mm, to be installed at an θ = 80° inclination relative to the bottom of the effluent flow channel. In addition, concerning preliminary treatments, a 2.00 mm mesh aperture was defined for the rotating sieve [36].

The MBR tanks were designed considering Purch PSH-1800 type modules from Koch Membrane Systems Inc, with S = 1800 m² of the individual area [36]. It was estimated that 100 filtration modules are required for the permeate flow treatment. For this context, sludge recirculation flow of Q_S = 310 m³/h was also calculated. The air consumption of an MBR unit is expressive. According to Tchobanoglous et al. [39] and Cassano and Basile [40] the submerged membrane system makes use of Ω_m = 0.25 m³/h·m², at the same time as a further Ф_a = 5.30 m³/h·m³ is consumed for aeration of the biological reactor. Under the study conditions, the association of these performances resulted in a total air flow Q_a,t = 143 m³/h (at 20°C), which will be maintained by rotary blowers from VentBras Ltda of the R500 type, at ω = 2,250 rpm, pressure drop ΔP = 40.0 kPa and installed power E_ae = 3.40 kW. The premise that the slurry recirculation pump must exceed a manometric head of h_mt = 5.00 m resulted in an operating power for this equipment of E_rp = 1.00 kW. The tailings from this treatment step will be pressed before proceeding to final disposal in an industrial landfill [36].

The RO operation was modelled using the Water Application Value Engine – WAVE^®, a freeware design software conceived and distributed by DuPont de Nemours Inc. [41]. To this end, it was assumed that the liquid flow would cross the membrane only once. The effluent enters the unit at T = 25°C and P = 370 kPa and acidity index pH = 7.0. The membranes selected to act in RO are thin-film composite [42], with dimensions of ϕ = 20.3 cm and L = 1.00 m and pore diameter of ϕ_p < 0.001 µm. These elements are made of polyamide and are, therefore, resistant to extreme pH (2.0–12) oscillations; conversely, the material is poorly tolerant to chlorine (Cl < 0.10 ppm). The system operates at a pressure P = 1500 kPa, and the total active filtration area is S = 343,370 m². Under these conditions, the RO reached expressive water recovery (η = 80%) and salt and silica rejection (ψ_salt ≈ ψ_SiO2 > 99%) rates [36].

The Basic Sanitation Utility Company of the State of São Paulo (SABESP) holds the public services concession for water production and sewage treatment in the region. Most of the 119 m³/s of drinking water it generates comes from chemical treatment. However, during the last decade, SABESP has made significant investments in biological and ultrafiltration techniques to expand water supply with adequate quality standards for RMSP [43]. These arguments have led to selecting such technologies as the conventional processes for water production assessed in this study.

Chemical Treatment. The first alternative to be compared to potable reuse comprises the chemical treatment process (CHE) of raw surface water. The scheme consists of the serial association of the coagulation, decanting, filtration, disinfection and oxidation, and final treatment stages [43].

Coagulation occurs from a dosage of aluminium sulphate (Al₂(SO₄)_3(s)), which acts as a fluidity reducer when creating available Al³⁺ ions. If dispersed in an aqueous medium, the electrolyte associates with hydroxylates (OH^-), producing a colloidal solution. The solid organics present in the raw water adhere to the solution by electronic affinity, being further removed by decantation [44]. Considering the raw water characteristics adopted for modelling the treatment routes (Table 1), an addition of 40.2 g Al₂(SO₄)₃/m³ was admitted to the coagulation step. Residual amounts of suspended material are removed by filtration.

The filter medium consists of materials with different granulometry, from gravel to fine sand, and the latter should be replaced periodically [45]. The replacement rate of this bed was estimated as 1.50 g/m³. An adjustment of the acid content of the water is performed immediately after filtration to avoid pipe corrosion by adding lime (6.20 g/m³) and sodium silicate (1.44 g/m³). Water disinfection occurs by the addition of chlorine gas (Cl_2(g)). Although it presents dangers (explosion, corrosion, potential to produce HCl) and occupational problems (dermatitis, burn, visual disturbances, bronchial lesions, pulmonary irritation) during preparation and handling, chlorine is still used for this purpose due to its high oxidation potential [46].

According to the Brazilian Ministry of Health, drinking water delivered to the consumer must contain a minimum concentration of 0.20 ppm residual Cl [43]. In the present case, an administration of 2.02 g Cl₂/m³ was considered at this stage of the process. Organic matter and microorganism removal are completed by the addition of 41.3 g/m³ of Oxygen (O_2(g)) which, like chlorine, acts as an oxidant for these contaminants when it turns into ozone (O₃).

At the end of the process, the water receives complementary treatment to remove pollutants that give it colour, turbidity, odour, and taste, and fluoride is also added to prevent tooth decay. The withdrawal of organoleptics was represented in the model by an activated carbon bed, which must be 661 mg/m³ due to saturation [39]. The fluoridation intensity varies according to local conditions (e.g., climate and temperature). However, the Ministry of Health recommends a residual content of 0.7 ppm F in drinking water [43]. In addition to chemicals, the process makes use of electrical energy to drive and move equipment. The overall consumption of this utility was estimated at 323 Wh/m³ [39].

Biological Treatment. The biological filters (BIO) is another consecrated route for water treatment. Due to this characteristic, the system does not use chlorinated agents, and the disinfection process is carried out only by O_2(g). The biological filters develop microorganisms capable of removing dissolved solids and gases (CO₂ and N₂ in excess) and adding O₂, as well as degrading organic chemicals (organic matter) and inorganic compounds (ammonia and nitrites) present in the water. Three classes of aerobic microorganisms are noteworthy regarding biological filters: (i) heterotrophic bacteria that metabolize dissolved carbonaceous compounds; (ii) Nitrosomonas that use ammonia (NH₃) as substrate producing nitrites (NO₂^-); and (iii) Nitrobacter that convert nitrites to nitrates (NO₃^-) [47].

The other BIO treatment stages and their sequencing are similar to those that make up the chemical process. The amounts of reagents and assets adopted to build this model are the following: Al₂(SO₄)₃ 63.6 g/m³; sand 1.15 g/m³; lime 5.42 g/m³; O_2(g) 1.74 g/m³; and activated charcoal 227 mg/m³. The need for oxygen pumping to the filters and the circulation of volatile suspended solids cause the electricity consumption for the system to be higher than its counterpart, reaching 693 Wh/m³ [48].

Ultrafiltration. The Ultrafiltration technology (UFT) is more recent than the processes described above. This alternative uses a pressure gradient (or concentration) to displace contaminated water through a semipermeable membrane. Thus, both suspended solids and high molecular weight solutes (colloids and viruses) are retained in the retentate stream. In contrast, water and low molecular weight solutes pass through the membrane, forming the permeate [49], [50]. UFT has been used to replace secondary filtration stages (coagulation, flocculation, sedimentation) and tertiary stages (sand filtration and chlorination disinfection) in water treatment plants or to operate in separate systems in isolated regions. In such functions, its use has grown exponentially [51].

To achieve satisfactory efficiency indexes even when operating water containing high suspended solid contents (typical raw surface water available in many Brazilian regions), ultrafiltration should be preceded by a chemical pre-treatment promoting fluid disinfection. The pre-treatment modelling considered O_2(l) (34.3 g/m³) and sodium hypochlorite (NaClO: 4.02 g/m³). It was further admitted that the filter medium would consist of a hollow fibre membrane made from polyvinyl n-fluoride [–(CH₂CHF)_n–]. The electrical demand for ultrafiltration was calculated as 372 Wh/m³ [41], [52].

The operations that integrate each of the analyzed procedural schemes are depicted in Table 2. The method used for classification was based on the conventional ordering of water treatment plant stages (preliminary, primary, secondary, and tertiary treatments). The exception was the ultrafiltration operation, whose pre-treatment acts simultaneously as primary and secondary treatments. Table 2 also specifies the scenarios established for performing the study.

The elaboration of the diagnoses followed the methodological guidelines established by the ISO 14044 standard [53] for attributional-type LCA applying a 'cradle-to-gate' application scope. The analyses were carried out considering the production of 1.0 m³ of drinking water according to quality standards set by the legislation. The Product Systems that define each treatment were described schematically in Figure 1. The REU modelling was based on primary data collected for the 2016–2017 period in a unit located in the state of São Paulo. The models that represent the environmental performances of the other treatment alternatives were constructed using literature data (secondary data), taking into account the average production technology, parameters, and operating conditions practiced in the state of São Paulo. No multifunctional situations were identified in any of the investigated arrangements.

The Life Cycle Impact Assessment (LCIA) was developed from two perspectives. The first quantified resource consumption by each process in terms of Primary Energy Demand (PED) and Water Depletion (WD). The PED results were described by the Cumulative Energy Demand (CED), v. 1.13 method [54], which makes it possible to classify such depletions as to their origin in terms of renewable (biomass; solar, wind and geothermal; and water) and non-renewable (fossil; nuclear; and biomass) energies. The PED indicator was selected due to the known energy-intensive nature of the analyzed processes.

Conversely, impacts related to WD were estimated using the ReCiPe Midpoint (H), v. 1.12 method [55], which totals decreases in raw water from various sources (river, lake, ocean, and underground). The inclusion of this impact category allows for the verification of the amount of raw water spent in each system to produce 1.0 m³ drinking water. The consumption sets up the yield achieved by each technology from a life cycle perspective. The second perspective took into account environmental impacts associated with emissions. To perform this analysis, the Global Warming Potential (GWP) categories were applied, accounting for their effect magnitudes via the method proposed by the Intergovernmental Panel on Climate Change in 2013 [56]. GWP is one of the most important environmental problems to be faced by modern society. Therefore, proposals concerning improvement actions that aim to reduce impacts in this domain must consider GWP when they are planned.

The LCIs that composed the conventional treatments were originated from the following datasets present in the Ecoinvent database [57], [58]: CHE: ‘Tap water {RoW}| tap water production, conventional treatment | Alloc Def, U'; BIO: ‘Tap water {RoW}| tap water production, conventional with biological treatment | Alloc Def', U'; and UFT: ‘Tap water {RoW}| tap water production, ultrafiltration treatment | Alloc Def, U'.

These data sets were selected considering technological similarities between the processes they portray and their counterparts currently in operation in the state of São Paulo. The (few) cases presenting heterogeneities were the object of LCI adaptation procedures to Brazilian systems conditions. Maintenance of the criteria established for the modelling was a constant concern during this process. In the specific cases of the treatments that operate from surface water (CHE, BIO, and UFT), the energy consumed to pump the raw water fed to the systems was estimated considering an average distance of L = 500 m between the river and the treatment. For UFT, the shelf life of each ultrafiltration unit module was set at 10 years.

The preliminary treatments and the MBR system were disregarded for the purpose of preparing the LCI for the REU. This decision was made due to a methodological circumstance linked to the LCA technique. Although technologically necessary to compose the arrangement, the interventions carried out by these steps, especially MBR, on domestic effluents seek to reduce their contamination levels up to (at most) the limits provided for in Brazilian legislation for disposal in water bodies. As wastewater treatment is a compulsory activity in the country, the environmental loads generated in these stages must, therefore, be attributed to the fulfilment of the function 'disposing of effluents in the environment with quality standards consistent with the relevant legislation', which is different from that assessed by the study. Potable reuse is effectively configured only after disinfection provided by UV radiation, followed by the transformations carried out by RO and reduction of Total Organic Carbon, with a second dosage of UV rays. Although disregarded by the REU performance evaluation, the preliminary treatments and the MBR system were designed in detail so that the contamination levels of the wastewater upstream of the REU were known, without which inaccurate stage specification and dimensioning would occur. Another adjustment made to the inventory that describes the REU was to take into account only electricity consumption.

The environmental loads resulting from the manufacture, use, and disposal of chemicals consumed in the process (i.e., O₃ and H₂O₂) were also disregarded (Figure 1). These contributions were below the standard of 1.0% of the cumulative mass contribution adopted as a general cut-off criterion for modelling product systems [53]. Conversely, the stages of transportation and final disposal of solid residues in industrial landfills were considered for the purpose of inventory construction.

Although the RO and MBR units were modelled by different computational resources, care was taken to harmonize the results of these stages' design through the preparation of material and energy balances. This procedure also extended to the preliminary treatments and disinfection and oxidation stages that also comprise the arrangement. The environmental charges originated from electricity generation were computed from secondary data taking into account the national grid (BR grid) composition for 2016 [59]. According to this scenario, the hydroelectric source provides 68% of all energy generated in the country, followed by natural-gas fuelled thermoelectric plants (9.1%), biomass (8.2%), coal (4.2%), and oil derivatives (2.4%), as well as wind generation (5.4%) and nuclear (2.6%).

Figure 1.

Product System Schemes for the scenarios under analysis in the present study

According to the results described in Figure 2, electricity consumption represents the main source of PED impacts for all treatments. Contributions ranged from 48% of the total for the category, as in the UFT case, to 100% observed for REU. The relationship between this contribution (PED_El) for the total Primary Energy Demand for each scenario and the electricity consumption of the respective treatment unit is imminent. Treatments with low electricity requirements, such as CHE (323 Wh/m³) or UFT (372 Wh/m³), registered the lowest PED_El values among the analyzed options. In contrast, this index was significant enough in treatments with high requirements – i.e., BIO (694 Wh/m³) and REU (844 Wh/m³) – to result in energy demand as the primary individual impact cause in the category.

Figure 2.

Contributions of each treatment for PED [MJ/m³] discretized by type (or class) of precursors

As in all situations, electricity is provided by the BR grid of 2017, and it was possible to establish a direct correlation between the global energy demand of each system and its impacts share for the category. The result of 5.63 MJ/kWh is compatible with that obtained for the Norwegian matrix (3.86 MJ/kWh), for which, as in the case of the BR grid, the primary energy generation source (~ 95%) in 2017 was hydropower [60].

On the other hand, the BR grid proved to be less impactful in terms of PED than the energy mixes of China (10.9 MJ/kWh), Germany (11.2 MJ/kWh), Japan (11.4 MJ/kWh), France (12.0 MJ/kWh) and the United States (12.4 MJ/kWh). In those energy mixes, fossil sources represented 88–93% of the total electricity during the same period [60]. However, even though non-renewable sources contributed only 19% of the Brazilian matrix, this share represented about 40% of the PED impacts. This is because hard coal, raw natural gas, and crude oil – the natural resources from which energy sources are obtained – exhibit high specific energy values of 30.7 MJ/kg, 38.3 MJ/m³, and 45.8 MJ/kg, respectively [54].

Chemical consumption also has an impact on PED. For CHE, such effects are mainly associated with Al₂(SO₄)₃ (309 J/m³) and O_2(l)(330 J/m³). The Al₂(SO₄)₃ is produced from aluminium hydroxide [Al(OH)₃], whose synthesis requires natural gas (5.06 MJ/kg) to meet thermal needs. O₂ processing occurs with cryogenics, a technology that presents an appreciable electrical demand (1.42 kWh/kg) to achieve high separation efficiencies. For the BIO treatment, PED inputs are due to the manufacture of NaOH (57.4 J/m³), which requires 1.33 kWh/kg when being carried out by electrolysis and, once again, via Al₂(SO₄)₃, whose production, in this case, affects the category at 490 J/m³.

Finally, for the UFT, contributions derived from chemicals are manifested through O_2(l) (274 J/m³) and sodium hypochlorite NaClO (45.9 J/m³). NaClO synthesis from chlorination of soda demands 1.62 MJ/kg of heat for being endothermic. Obtaining potable water by ultrafiltration deserves closer attention to PED impacts due to waterworks, which involve structural system aspects. In this segment, the highlights are glass fibre and epoxy resin, regularly used in the manufacture (or for maintenance) of UFT units. Their PED throughout the production is 1.31 MJ/m³ and 549 J/m³, respectively. Such performances are due to the consumption of heat (glass fibre: 14.4 MJ/kg; epoxy resin: 2.91 MJ/kg) resulting from natural gas burning and electricity (glass fibre: 1.61 kWh/kg; epoxy resin: 333 Wh/kg).

GWP contributions described a behaviour similar to that observed for PED for all the analyzed arrangements, also due to the influence of electricity consumption. REUs performance clearly defines the degree of intensity of this interference since electricity consumption is the only source of impact for the category. CO₂ of fossil origin (CO_{2, f}) appears as the primary impact precursor for the scheme, representing about 63% of its contributions to the GWP. Its emissions are mainly due to the combustion of natural gas (44.6 g/m³), coal (35.0 g/m³), and oil products (17.6 g/m³) in thermoelectric plants. Methane (CH₄) emissions are due to losses that occurred during the extraction, processing, and transportation of natural gas (1.52 g/m³) and coal mining (135 mg/m³), representing 33% of GWP_REU impacts.

About 60% of GWP impacts from the CHE life cycle are due to the power consumption of the treatment itself. A further 10.8 gCO_{2 eq}/m³ are assigned to this treatment due to CO_2,f emissions during O₂ liquefaction, so that it will be transported and stored in conventional treatment before being used as an oxidizing agent. Other 12.6 g/m³ of CO_2,f are also emitted due to the heat generation to be used in the Al₂(SO₄)₃ processing. In addition, another 5.56 kg and 2.17 CO_{2 eq}/m³ are attributed to CHE due to the use of lime (CaO) and chlorine (Cl_{2 (g)}). GWP impacts associated with lime originate from CO_2,f emissions from limestone (CaCO₃), decomposition, a process regularly applied to obtain this compound. The endothermic character of the reaction (Δh_298°C = 182.1 kJ/mol CaO) predisposes the internal temperature of the calcination furnaces to be maintained above 900 °C [61]. This generally occurs through natural gas burning, which generates 292 g CO_2,f /kg CaO as a combustion by-product. Another 786 g CO_2,f/kg CaO are also released into the atmosphere due to the calcination of the CaCO₃ itself. Chlorine production occurs through electrolysis, which consumes between 414–500 kJ/kg Cl_2,(g) [62]. If generated from the BR grid of 2017, this amount of electricity provides 1.74–2.10 g CO_2,f/m³.

Also, the electricity consumption of the BIO unit was responsible for the main GWP contributions (~ 75%). As with CHE, the processing of Al₂(SO₄)₃ and lime are significant sources of contribution to the category among the chemicals used in the process due to CO₂, _f emissions, in this case totalling 16.5 and 4.82 g/m³, respectively. Fossil carbon dioxide releases were also identified in the manufacturing cycles of sodium phosphate (Na₃PO₄: 2.87 g/m³), sodium hypochlorite (NaClO: 2.20 g/m³), sodium hydroxide (NaOH: 1.78 g/m³) and activated silica (SiO₂: 1.66 g/m³).

Similar to what was observed for Primary Energy Demand, epoxy resin (28.0 g CO_2eq/m³) and glass fibre (73.7 g CO_2eq/m³) used for waterworks appear as the most significant foci of GWP for ultrafiltration. As in the previous cases, CO_2,f is also the primary impact precursor. From the technical documents that describe sizing and maintenance actions in pressure vessels of RO units operating with seawater, Althaus et al. [57] estimated the average consumption of fiberglass and epoxy resin at 52.1 and 4.07 mg/m³ in UFT plants, at an operational life span of 15 years. Glass fiber production emits on average 34.4 g/m³ of fossil carbon dioxide due to its specific energy consumption of 1.61 kWh electricity and 14.4 MJ heat per 1 kg product.

Figure 3.

Contributions of each treatment for GWP [g CO2eq/m3] distributed by type (or class) of precursors

The remainder of CO_2,f emissions (24.9 g/m³) recorded by this production cycle comes from the processing of raw materials, such as Nylon 6 (33%), boric acid (27%), aluminium oxide (5.2%), and silica (2.9%) [57]. The most common epoxy resins are obtained by reacting epichlorohydrin with bisphenol A. The product of this interaction is treated to recover the epichlorohydrin, and filtered to increase the purity of the epoxy, which in this stage is in the liquid phase. Althaus et al. [57] reported the total energy (electricity and heat) demand of the process as 141 MJ/kg resin. These utilities generate CO₂ emissions, which, together with those resulting from syntheses of epichlorohydrin and bisphenol A, total 24.1 g/m³. The unit's electrical demand completes the GWP impact load, accumulating 37% of the contributions to that category.

Impacts in the form of WD can be considered performance indicators for conventional systems, as they verify the degree of water resources required for each treatment. From this perspective, UFT is slightly better (~ 3.0%) than CHE and BIO, registering almost full yields in converting raw water into potable water. For UFT, this performance can be attributed to the ultrafiltration capacity to separate suspended and dissolved matter from the liquid fraction.

The REU's impact regarding WD is restricted to water consumption to replace regular drains in thermoelectric (purging and evaporation) and nuclear (evaporation) plants. As these sources represented 24% and 2.6% of the electricity generated in Brazil in 2017, respectively, the associated specific consumption of raw water was 4.71 L/kWh. When analyzed alongside the individual energy demand of each treatment, this performance explains the impact fractions related to this utility (Table 4).

CHE and UFT recorded the most significant WD impacts associated with chemicals. In both cases, this is due to using O_2(l) in each technology, which corresponds to 41.3 and 34.3 kg/m³, respectively. The process to obtain O₂ generally occurs through cryogenics which, according to Althaus et al. [57], accumulates a water consumption of 354 L/kg O₂ to replace evaporative losses generated in the cooling tower.

The estimate was based on real-life data collected in a small air separation unit (1.0–2.0 t/h) and extrapolated to larger systems. In this case, the cooling water demand (circulating) ranged between 120–160 L/kg of liquefied product [63], and the makeup accounted for 1.0–3.0%_w/w of the circulating flow [64]. According to Recknagel et al. [65], a maximal makeup of 3.0 kg/h of water is required to cool 1.0 kJ/s of produced waste heat. Althaus et al. [57] applied the allocation procedure based on the economic criterion to distribute the environmental loads associated with the cryogenic process for O₂, N₂, and Argon (Ar). Thus, O_2(l) was responsible for about 31% of the consumption and emissions generated up to that life cycle stage. This decision projected an overall demand of 914 L raw water/kg O₂.

Of this total, about 560 L/kg O₂ are released to the environment as a liquid effluent from the process. However, except for temperature (~ 40 °C), none of the characteristics or properties of this stream were altered by the cooling process. This condition causes the eviction to be treated as 'credit' to CHE and UFT in terms of WD, which must, therefore, be charged against the full consumption of raw water.

Table 4 also highlights the unpolluted wastewater that corresponds to process outlets due to leaks in the pumping of raw or treated water, tanks and filter drainage, equipment washing, and internal system losses. These streams can be used in different functions, provided they receive some level of treatment. Therefore, instead of effluents, LCA considers them as by-products of potable water production technologies.

This status causes unpolluted wastewaters to be affected by the allocation procedures applied to share the environmental burdens generated up to their production. Conversely, as with the liquid water discharged from the cooling tower during cryogenic O₂ processing, unpolluted wastewater receives a credit for saving raw water, which should also be discounted from the global demand of each treatment. In this context, CHE and BIO exhibited expressive results, reducing their water depletion by 29.3 and 43.7 L/m³, respectively. BIO performance was sufficient enough that raw water consumption was lower than the produced potable water volume.

The fact that direct result comparisons did not identify a prevailing treatment only reinforces the importance of elaborating a Global Performance Index (GI_i) to associate all the impacts verified by the analysis for each alternative. This conduct to define GI_i has already been adopted in other situations, as in Hansen et al. [66] who compared different polystyrene processing routes, Donke et al. [67] in the analysis of a multi-purpose ethanol processing plant, and Guerra et al. [68] during a thermodynamic and environmental analysis of scaling up a cogeneration unit. In the field of water reuse, Sakamoto et al. [62] adopted the same approach to evaluate the partial closure of an oil refinery's water circuit. Similarly, Ronquim et al. [69] applied a kind of global performance indicator to increase water conversion in RO membranes of a system designed under a zero-liquid discharge concept. It consists of normalizing those dimensions, defining a reference scenario among the schemes. Given that the common purpose of all treatments is to produce potable water, the baseline scenario was selected as the one accumulating the highest raw water consumption to exercise this function. This criterion led to the choice of CHE (Table 1). The Normalized Impact Index for each scenario was estimated by applying the following equation:

$N{I}_i^k=\left(\frac{E{I}_i^k}{E{I}_{CHE}^k}\right),$ (1)

where: NI_i^k describes the Normalized Impact Index for the impact category k of arrangement i, EI_i^k refers to the Environmental Impact k for scenario i, and EI_CHE^k denotes the Environmental Impact k for the baseline scenario (CHE). The Global Performance Index value of the treatments was estimated by multiplying their respective NI_i^k to measure the intensity of potential synergy (or conflicts) among the analyzed dimensions. Such estimates were made as follows:

$G{I}_i={\displaystyle \prod \left(N{I}_i^k\right)},$ (2)

where GI_i is the Global Performance Index for scenario (i).

Table 5 describes individual results of NI_i^k and the corresponding GI_i for each treatment. It follows from the comparison between GI_i values that REU is the less impactful option for the production of 1.0 m³ of potable water among the options considered by this study. Although its individual contributions for Primary Energy Demand (NI_REU^PED) and Global Warming Potential (NI_REU^GWP) were 76% and 57% higher than those of the baseline scenario, respectively, its performance in terms of Water Depletion (NI_REU^WD) significantly reduced those deviations. In addition, REU has already shown better total performance than BIO and UFT even before considering raw water consumption.

Nevertheless, the standard deviation calculated for the NI_REU^k indices (sd_REU = 0.79) demonstrated a high degree of dispersion among those results. This disparity results from the essence of potable reuse. On the one hand, more energy resources are consumed to purify domestic wastewater compared to approaches that use raw water. On the other hand, reuse technology avoids (or at least restricts) impacts in the form of Water Depletion, which are inherent to conventional routes. In this context, and as expected, the normalized impact indexes of conventional treatments exhibited a low degree of dispersion (sd_BIO = 0.35; and sd_UFT = 0.36). Finally, it should be noted that NI_BIO^WD 1.0. Although slightly lower than those obtained by CHE, such results can be seen as indicators of the high efficiency of those arrangements compared to the baseline scenario, especially if viewed from a perspective of the water needs of large metropolises.