Application of membrane processes for expressway rainwater runoff treatment and reuse

Due to population growth, climate change, and anthropogenic pressure, water resources are gradually decreasing all over the world^1,2. The forecasting made by International Water Association indicates that by 2030 the world is projected to face a 40% global water deficit³. As water resources become more limited, the concept of water reuse is becoming important⁴. Recently attention has been paid to the use of rainwater which can be easy-to-collected water resource^1,2,3. The reuse of rainwater enables saving of surface and groundwater resources, reducing energy consumption and minimalizing (diminishing) stormwater runoff^5,6,7,8,9. The rational management of rainwater flow also helps to close the hydrological cycle¹.

A literature review indicates that studies on harvesting and reuse of rainwater for potable or non-potable water production focus mainly on rainwater collected from various roof coverings^2,7,10,11,12. However, another important dimension attracting little attention is the possibility of rainwater gathering from hardened/sealed surfaces, i.e., parking lots, traveler’s service places, and expressways. Surface runoff from these areas is collected in reservoirs, which are part of the road drainage system. Such a system allows for the collection of larger volumes of water. So far, these reservoirs have not been widely analyzed in the context of the possibility of recovering rainwater collected in them. The quality of collected runoff in these reservoirs is affected by the way in which the road infrastructure surfaces are developed (type of pavement, number of car parks, etc.), the way in which the roads are operated and maintained (winter road maintenance, de-icing), and the characteristics of the precipitation (timing and intensity of rainfall, rainfall pollution, etc.)^13,14,15. The main pollutants are mainly suspended solids, petroleum hydrocarbons, heavy metals, chlorides, Na, Mg, Ca, biogenic compounds (nitrogen, phosphorus, and potassium), micropollutants (e.g. aromatic hydrocarbons, grated parts of tires, brake linings, used vehicle parts, and road surfaces), inorganic salt ions, and pathogenic microorganisms^3,13,15. The technologies for treating this type of water for potential reuse, as described in the literature, include bioretention, rain gardens, constructed wetlands¹⁶, sustainable biochar¹⁷, sand filters, and vegetated channels¹⁸, as well as swales¹⁹. Numerous studies recommend a “treatment train” approach tailored to the intended end use of the reclaimed water²⁰. Published research on the application of membrane processes remains limited and primarily focuses on the separation of one or several specific contaminants from stormwater, rather than on the final quality of treated stormwater and its potential for reuse^21,22. In other studies, membrane-based methods are restricted to selected processes, such as ultrafiltration (UF)²³. Conducted research indicates that runoffs from sealed areas are characterized by high content of inorganic pollutants and low biodegradability, hence the use of membrane filtration methods for their treatment can bring a good results^2,3,13. Membrane processes are techniques with high treatment efficiency and a large amount of permeated water capacity³. Use of them can minimize the number and complexity of rainwater treatments to reach the water quality reuse standards, which is important in the context of water reuse. However, the application of membrane technology in the field of expressway rainwater treatment and reuse has not been comprehensively discussed. Therefore, the aim of this paper is to assess the possibility of using selected membrane processes: ultrafiltration (UF), nanofiltration (NF), and reverse osmosis (RO) for the purification and recovery of rainwater from expressways and passenger service areas. Therefore, the article evaluates the efficiency and performance of selected membrane processes in rainwater treatment as well as the validity of their use in relation to potential reuse of rainwater.

In this context, this article discusses an innovative issue that has not been widely described in the literature to date: the possibility of reusing rainwater from motorways by applying various membrane techniques for its purification, especially since no comprehensive research results are available on this topic. This issue is particularly important in view of the growing water shortage and the need to implement sustainable solutions in water management. Analysis of the proposed technologies allows not only to assess their effectiveness in removing pollutants characteristic of road runoff, but also to determine their potential for application in engineering practice. As a result, the approach discussed may constitute an important step towards closing the hydrological cycle, increasing the resilience of water systems to the effects of urbanization and climate change, and contributing to effective rainwater management.

Site characteristics and sample collection

The rainwater used in the study was taken from the reservoir of the stormwater drainage system for the National Road No. 8, with the catchment area beginning at kilometer 648 + 841. The total catchment area is 12.6 hectares, of which 48.5% consists of impervious surfaces, while the remainder is comprised of green areas. The drainage system collects rainwater from the main road, shoulders, the Sielachowskie road junction, and access roads. The rainwater runoff flows gravitationally through an 800 mm collector into a retention reservoir, from which samples were taken for analysis. The reservoir and collectors are watertight.

The recipient of the stormwater is the Supraśl River. The estimated quantities of rainwater discharged to the recipient are as follows: maximum hourly water volume Qh_max = 52.2 m³/h, average daily water volume Qd = 89.0 m³/d, maximum annual water volume Qy_max = 32,491.4 m³/a. Before being introduced into the retention reservoir, the rainwater is pre-treated in grassy ditches, sedimentation manholes of individual storm inlets, a sedimentation tank, and a petroleum-derived substances separator in the main drainage channel. The sedimentation tank is constructed as a reinforced concrete well with a diameter of 2,000 mm and a capacity of 5.0 m³. The separator used is a lamella separator with a nominal capacity of 60 L/s, which retains 97% of oil-derived pollutants, and a hydraulic capacity of 600 L/s. Additionally, before being discharged into the recipient, the water is treated in a sedimentation tank and petroleum-derived substances separator on the outflow channel.

The retention reservoir has a capacity of 900 m³, and the maximum outflow from the tank is 58 L/s. The tank’s slopes are inclined at a 1:2 ratio, the bottom has a cross slope of 2%, and the effective depth of the tank is 1.6 m. The tank is designed to be watertight.

Rainwater was sampled from the reservoir using a peristaltic pump, with the suction nozzle placed 10 cm below the water surface, and collected in 1.5-liter bottles for chemical analysis and 20-liter polyethylene containers for membrane processes. For microbiological analysis, samples were taken directly into sterile bottles and, along with the 1.5-liter bottles, immediately transported to the laboratory in a portable cooler.

This article contains the results of pilot-scale studies. The extension of the sampling campaign will be included in future studies. Broader generalization requires multi-site and multi-season campaigns. The representativeness of the published data is limited, but due to the averaging of the quality of rainwater retained in the reservoir, it can serve as a basis for assessing the possibility of using membrane processes for its treatment and use for specific purposes.

Membrane processes setup and operation

The rainwater purification process was carried out using the RO20NS.1 laboratory membrane system, which was designed and constructed in such a way as to reflect the actual operating parameters of membrane systems on a technical scale. The membrane unit allows UF, NF and RO processes to be carried out with different types of tubular membranes and with the possibility of configuring the system operation. The filtration process was carried out in a cross-flow system with the concentrate being returned to the feed tank supplying contaminated water to the UF/NF/RO system. Before each experiment, the clean water flux (J₀) was determined using ultrapure water. A commercial rural membrane of PCI Membranes was used in the experiment with characteristics given in Table 1.

Before direction to the UF, NF and RO systems the collected rainwater was first screened with a 1 μm mechanical filter to remove suspended particles. Additionally, before the UF and NF processes a filter with activated carbon (AC) filling was used. Activated carbon filtration was applied to eliminate contaminants that might not be effectively retained by membranes. Due to the characteristics of RO membranes, the use of an activated carbon filter was omitted, assuming that the high degree of contaminant separation achieved through reverse osmosis would render additional carbon filtration unnecessary.

Each membrane process was carried out at a constant temperature of 25 °C and constant velocity of 18 L/min. The process was operated with a water recovery of 70%. The volume of permeate was monitored to determine the permeate flux (J) using the equation:

where: V_p – permeate volume (L), t – time (h), S– membrane area (m²), k – temperature correction factor. The removal rate of salts expressed as a electroconductivity value (R_EC) was calculated according the following formula:

where: EC_p – electroconductivity of permeate (mS/cm), EC_n – electroconductivity of inlet (leachate) (mS/cm). For all membrane process configuration a concentration factor of EC in concentrate (CF_EC) was determined as:

where: EC_c – electroconductivity of concentrate (mS/cm).

The degree of rainwater recovery was calculated from equation:

where: V_p – volume of filtered rainwater (permeate) and V_in – volume of rainwater supplied/inlet. Treatment efficiency was evaluated by monitoring quality parameters in permeate obtained after filtration and calculation of removal effect:

where: C_p –contaminant concentration of permeate (mg/L), C_n – contaminant concentration of inlet (leachate) (mg/L).

After ultrafiltration, nanofiltration and reverse osmosis process, membrane was physically washed using distilled water. Next, chemical cleaning procedure has been carried out. According to the manufacturer’s instructions, a 0.25% (v/v) solution of Ultrasil 11 was used, with which the membranes were rinsed for a period of 30–60 min. The membrane cleaning was carried out at a temperature of 50 °C and a pressure of 0.3 MPa. After both physical and chemical cleaning, its effectiveness was assessed by measuring the value of permeate flux (m³/m²·s).

A flux reduction ratio associated with fouling (inorganic and organic) FRR was calculated on the basis of the ratio of water flux through clean membrane before filtration process and water flux after physically cleaning:

where: J₀ -is the water flux through clean membrane before filtration process and J_phs – is the water flux after physically cleaning.

Analytical methods

Collected rainwater from the reservoir and after filtration processes was analyzed to determine its quality and the efficiency of membrane separation. pH and conductivity (EC) were determined by a pH/conductivity meter (HQ40D, HACH). Anions and cations were determined using the chromatographic method with conductometric detection (Ion Chromatograph ICS 5000+, Dionex). Total nitrogen was measured with use of the spectrophotometric method (Spectrophotometer DR6000, HACH). Suspended solids (SS) and Total Dissolved Solids (TDS) were analyzed with use of the weight method. Total organic carbon (TOC) was examined using the infrared spectrophotometry method (TOC Analyzer multi N/C 3100, Analytik Jena). The gas chromatography–mass spectrometry method (Gas Chromatograph GC/MS 7890B TripleQuad, Agilent Technologies) was used for Polycyclic Aromatic Hydrocarbons (PAHs) analysis. Metals were determined using the inductively coupled plasma – mass spectrometry method (ICP/MS Spectrometer 8800 TripleQuad, Agilent Technologies). The obtained results were the mean value of three determinations carried out simultaneously. All analyses were made in the Laboratories of Faculty of Civil Engineering and Environmental Science.

For microbiological parameters rainwater samples and permeate obtained from membrane processes were examined for total Coliforms and Escherichia coliform, number of fecal Enterococci, total number of microorganisms at 22±2 °C after 72 h and total number of microorganisms at 36±2 °C after 48 h. All microbiological analyses were made in the voivodeship sanitary and epidemiological station in Bialystok using the following procedures: PN-EN ISO 9308-2:2014-06 (Coliforms and Escherichia coliform), PN-EN ISO 7899-2:2004 (Enterococci), PN-EN ISO 6222:2004 (total number of microorganisms).

Collected rainwater characteristics

Rainwater runoff from the expressway was characterized prior to the membrane filtration to determine its quality (Table 4). The tested samples are characterized by a pH value of 8 and elevated content of TDS (1.19 g/L) and EC (2.04 mS/cm). The obtained values were higher than those obtained by Kayhanian et al.²³: TDS_mean – 87.3 mg/L, EC_mean – 96.1 µS/cm. However, the values of TSS – 0,013 g/L, TOC – 15.08 mg/L and turbidity – 26.6 NTU do not exceed or are similar to these in Kayhanian et al.²³ study. Analyzed rainwater contained a high value of Cl⁻ (567.28 mg/L) and Na⁺ (663.82 mg/L), which can be the consequence of contamination by road and street maintenance products. Other anion and cation concentrations were lower than reported by Köse-Mutlu¹ and Marszałek and Dudziak² in roof rainwaters. All analyzed metals were in concentration not higher than 56 µg/L with the highest concentrations of Al, Cu and Fe. Similar results were noted by Marszałek et al.²⁴ and Kayhanian et al.²³. As for the microbiology, there are not many analyses of run-off from road surfaces to which the results could be compared. The obtained data indicate that the E. coli content is lower and the Coliform content higher than in the studies of rainwater from roof surfaces analyzed by Shiguang et al.⁷. The characteristics of the quality of runoff from the expressway indicate that membrane methods may be appropriate processes for its treatment.

Membrane performance

The collected sample of rainwater was filtrated in three configuration of membrane process: AC-UF, AC-NF and RO. Due to the nature of the membrane and pore size, the permeate flux is always higher in ultrafiltration than in NF or RO filtration. Therefore, in assessing the permeate flux characteristics, the ratio J/J₀ was used, where J₀ is the flux through the clean membrane before the filtration process and J is the flux during rainwater filtration. This approach was also used in studies of Al Ashhab et al.²⁵, Kusworo et al.²⁶, Kusumocahyo et al.²⁷ and Shiguang et al.⁷.

The results of filtration process are presented in the Figs. 1 , 2, 3 and Table 2. In order to better reflect the treatment capacity of the different membrane processes, the x-axis in the Figures were expressed as the filtered volume of permeate.

The average ratio of initial observed flux was 83%, 48% and 19% for AC-NF, RO and AC-UF filtration respectively. Ultrafiltration was characterized by the greatest drop in flow despite the highest permeate flux values in the UF process. A drastic drop of permeate flux (which was below 20% of initial flux) was observed during the filtration of the first 2 dm³ of permeate, i.e. in the first 30 min of the filtration process. It was caused by the formation of a cake layer at the membrane area. According to a study conducted by Ouyang et al.²⁸ the cake layer is the dominant fouling mechanism in UF filtration compared to MF and NF processes²⁸. The AC-NF and RO processes were characterized by a lower decline of flux during the filtration the first 2 dm³ of permeate. The observed decrease was above 80% for AC-NF and 50% for RO. It can be concluded that pore blockage may have occurred at the beginning of the operation, resulting in a decrease in permeate flux, which is consistent with studies by other authors^1,7. A further decrease in permeate flux was expected because filtration was conducted without any backwashing procedure. However, after about 60 min of filtration, a stabilization of the permeate flux was observed in all experimental configurations, and it reached an almost constant value in AC-NF process. Similar permeate flux stabilization have also been documented by some authors and explained by the development of heterogenous structures in the fouling layer on the membrane surface^7,29,30. The strongest stabilization flux and the lower permeate decline were observed for nanofiltration processes, which can be attributed to the highest hydrophilicity of the NF membrane (Table 1). The high level of hydrophilicity contributes to less accumulation of polysaccharides and proteins on the membrane surface and to lightening membrane fouling and clogging, resulting in a higher permeate flux⁷.

The highest EC rejection rate – over 99% – was achieved during RO filtration which contributed to the high value of concentration factor of EC in concentrate, which reached the value of 2.25 (Figs. 2 and 3). In the case of nanofiltration and reverse osmosis, the EC rejection rate was stable throughout the process, resulting in very low standard error values of 0.02 and 0.37 for AC-NF and AC-UF, respectively (Fig. 2; Table 2). A stable rejection rate during the filtration process resulted in an almost constant EC concentration factor.

The average EC removal effect for AC-NF reaches the value of 42% and the value of the concentrate factor was 1.75 (Fig. 3; Table 2). Lower than expected EC removal efficiency was achieved for the UF process, which averaged 2%, giving a value of 1.05 for the EC concentration factor. Sherhan et al.³¹ obtained an 8% reduction of EC during ultrafiltration of the raw produced water. During filtration purification of pipe drinking water, a 13% of TDS was achieved by Yuan³². The low EC removal effect during ultrafiltration is due to the characteristics of the rainwater inlet, which is dominated by inorganic compounds and monovalent ions that are not retained by the ultrafiltration membrane.

The degree of concentration factor determines the possibility of its subsequent management, thus in this case its high values do not always have a positive effect. From this point of view, the optimal solution can be nanofiltration with sufficient rejection rates and EC concentrate factor below the value of 2. Table 2 presents basic statistical data on the membrane processes carried out.

The average permeate flux value for AC-NF filtration was 75 dm³/h m². This filtration required the most time (229 min) to achieve the assumed 60% recovery rate, i.e. 7 dm³ of permeate. The weak point of this process was the low EC rejection rate of 42%. This is in line with the Baker study according to which NF membranes provide enhanced water permeability due to lower hydraulic resistance, but rejection of small ions is inherently limited³³.

The use of reverse osmosis and the application of higher pressure (4.0 MPa) resulted in an increase in permeate quality (average EC rejection 99.6%). This gave in a shorter filtration time (212 min), which was due to a faster flow rate of 81.9 dm3/h m2. In turn, the UF process did not guarantee high-quality permeate. Due to the flow through the membrane of most monovalent ions, such as chlorides, sodium and potassium, the EC rejection rate was only 2.3%.

Membrane fouling

After finishing each filtration process, a physical cleaning of the membrane was performed. Based on measured values of flux before and after physical washing, a flux reduction rate associated with the cake layer was determined. The highest differentiation of flux before and after physical cleaning was for AC-UF process with the value of 96 L/h·m², which represented 40% of flux at the end of the filtration process. For AC-NF and RO filtration the differences were as follows: 15 L /h·m², 18 L /h·m², representing 10% and 20% of flux at the end of the filtration process, respectively. The results confirm the strong influence of the cake layer on the permeate flux, especially in the ultrafiltration process.

Based on measured values of filtration before and after cleaning procedures a flux reduction rates were calculated as described in the Materials and Methods section and presented in Table 3.

The flux reduction rate was highest during ultrafiltration process and reached the value of 87%. The FRR rate for AC-NF filtration achieved the value of 12.1%. The FRR rate for RO process was 50%. Because chemical cleaning procedure was conducted with alkaline product, it was effective in removing organic fouling, biofouling and colloidal fouling. This indicate that inorganic deposition was responsible for the remaining pore blockade. Inorganic fouling (scaling) was responsible for 47% permeability decrease for RO, 7.2% for AC-NF and 87.6% for UF filtration.

The chemical composition of the retention reservoir water indicated a high mineral load (TDS = 1.19 g/L; EC = 2,040 µS/cm), moderately alkaline pH (8.0), and elevated sodium chloride concentration (Na⁺ = 663.82 mg/L, Cl⁻ = 567.28 mg/L). Although sodium chloride itself is highly soluble and does not directly crystallize under typical membrane operation, its high concentration significantly increases the ionic strength of the feed water and changes solubility equilibria of less soluble mineral phases (e.g., Fe(OH)₃, Al(OH)₃, CaCO₃, CaSO₄). This can lead to intensification of concentration polarization near the membrane surface. Under conditions of diverse chemical composition and the simultaneous presence of organic and inorganic substances, several physicochemical mechanisms may occur simultaneously, enhancing fouling phenomenon³⁴. The measured concentrations of iron (53.74 µg/L), aluminum (55.84 µg/L), and copper (46.02 µg/L) are within ranges reported to induce colloidal and chemical fouling in pressure-driven membrane systems^35,36. At the system’s pH (≈ 8.0), these metals undergo extensive hydrolysis. The formation of insoluble hydroxides leads to heterogeneous deposition on the membrane surface and within pore structures, contributing to pore blocking and surface coverage^35,36,37. Literature reports indicate that even trace amounts (tens of µg/L) of these metals can contribute to colloidal fouling and surface scaling by forming amorphous hydroxide. These amorphous hydroxide particles have high surface reactivity and can adsorb additional ions, including Ca²⁺ and Mg²⁺, promoting secondary scaling and fouling³⁸. Additionally, Cu²⁺ and Fe³⁺ can bridge negatively charged organic molecules, forming metal–organic complexes that adhere more strongly to the membrane surface^39,40.The calcium concentration (30.63 mg/L) is moderate but sufficient to form calcium carbonate (CaCO₃) or calcium sulfate (CaSO₄) scales, especially under locally elevated pH and ionic strength conditions. With an alkalinity typically associated with pH 8, CaCO₃ scaling is plausible even at moderate Ca²⁺ levels, as has been demonstrated in reverse osmosis systems operating with similar feed compositions (Antony et al., 2011). Magnesium (2.71 mg/L) and sulfate (16.03 mg/L) levels are comparatively low, suggesting that MgSO₄ scaling is less likely. However, both ions can co-precipitate with Fe and Al hydroxides, modifying the compactness of the fouling layer^38,41. Fouling is promoted by the simultaneous presence of organic compounds and divalent and monovalent cations, such as Ca²⁺ and Mg⁺. Despite the relatively low organic content (TOC = 15.08 mg/l), organic macromolecules can form complexes with metal ions, promoting the formation of organometallic colloids that deposit on the membrane and reduce its permeability⁴². The conducted analysis indicate that the rainwater from retention reservoir poses a risk of inorganic and complex fouling. Inorganic fouling is connected with Fe/Al hydroxides, CaCO₃ deposits, and NaCl-associated ionic clusters. The presence of these components corroborates the coexistence of inorganic scaling and organic adsorption, forming a compact, low-permeability fouling layer. Scanning electron microscopy (SEM) and energy-dispersive X-ray spectroscopy (EDS) analyses often reveal Fe, Al, Na, Cl, and Ca peaks in the fouling layer^43,44,45.

The material from which the membrane is made and its hydrophilicity/hydrophobicity also affect the degree of flux reduction. UF membrane used in this study was made of PVDV with strong hydrophobic properties (Table 1). The hydrophobicity of the UF membrane promotes adsorption of both organic and inorganic contaminants on the membrane surface, resulting in a high fouling rate (FRR = 87.7%). Additionally, low efficiency in reducing electrical conductivity (EC) and total dissolved solids (TDS) may contribute to further deposition of ions and metals on the membrane surface. Research has shown that the hydrophobic nature of both the membrane and the contaminants affects the fouling process. This explains the low removal rate of hydrophobic PAHs on UF membranes, which are adsorbed onto the membrane, causing pore blockage or the formation of a dense layer of contaminants^41,46,47.The nanofiltration membrane had the lowest fouling rate, and was characterized by the strongest hydrophilicity (Table 1). Enhanced hydrophilicity of the NF membrane means a more strongly bounded water layer at the membrane surface, which may act as a barrier for foulants and hence reduce fouling^48,49. Inorganic fouling was greater in reverse osmosis. This may be due to two reasons: first, there is a lack of AC pretreatment, which did not remove fouling-promoting compounds from the solution. Second, the membrane’s hydrophilic properties are lower compared to NF³⁸. The results confirm that the use of hydrophilic membranes, apart from obtaining greater permeability, also ensures a higher degree of removal of hydrophobic contaminants and allows for minimizing the fouling phenomenon.

Membrane process efficiency

The efficiency of conducted membrane process is presented in Table 4. The results do not include nitrogen and its compounds and phosphates, due to their oxidation and/or reduction during the filtration process.

General pollutants

The highest level of general pollutant removal effect was obtained after the RO process. The turbidity and TDS were almost completely removed from the inlet. The RO efficiency in TSS and TOC removal was > 99% and 79% respectively. The highest removal effect of AC-NF filtration was achieved for turbidity − 98%, and the lowest for EC and TDS − 42% and 50%, respectively. In the case of AC-UF, the removal efficiency of general pollution did not exceed 39%. Obtained results showed that UF membranes with AC pretreatment has a low efficiency in removing organic carbon. By contrast, AC-NF membranes’ efficiency for organic carbon is 68%. This finding is in line with Upadhyayula et al.⁵⁰ studies that indicated that UF membranes are not effective at removing natural organic substances. The low efficiency (68% for UF and 79% for RO) may be attributed to the presence of mineral salts in the treated solutions, which significantly reduce the effectiveness of organic substance separation⁵¹. According to Kabsch-Korbutowicz⁵¹, the presence of sodium in amounts exceeding 100 mg/dm³ causes a decrease in the retention coefficient because inorganic cations neutralize the charge of the membranes and organic substances, resulting in a lack of electrostatic interactions between these elements. The concentration of Na + in the feed directed to filtration was 664 mg/dm3, resulting in a decrease in the retention of organic substances.

The lowest efficiency of AC-UF was observed for EC and TDS, which is due to the passage of monovalent ions and cations, i.e. Na, K, and mainly Cl⁻, through the membrane pores.

The removal efficiency of PAHs was highest for RO and AC-NF processes and was 57% and 52%, respectively. The slightly highest efficiency was obtained by Smol et al.⁵² in RO filtration, during which a 59% to 72% removal effect of PAHs was achieved. The removal efficiency of PAHs in the UF process was 38%. According to Maeda and Kim et al.^37,53 the low efficiency of PAH removal is due to the fact, that long-chain aliphatic hydrocarbons have lower solubility, so there is a risk of free oil formation or precipitation during operation as the recovery rate increases, which reduces their removal.

Anions and cations

The removal efficiency of the RO process for cations and anions was very high, at no less than 98% for the analyzed ions. For the AC-NF process, a decrease in removal efficiency was observed and the lowest AC-NF efficiency was recorded for monovalent ions: Cl⁻, K⁺, Na⁺ which was 47%, 52%, 50% respectively. The highest AC-NF efficiencies were recorded for sulfate ions, which are divalent ions with a larger molecule than chloride, sodium and potassium ions. Ion rejection is closely correlated with ionic charge density and hydrated radius. Divalent ions possess stronger electrostatic fields and larger hydration shells, leading to higher rejection in NF and RO processes. Monovalent ions such as Na⁺, K⁺, and Cl⁻ have smaller hydrated radii and weaker electrostatic interactions with the membrane, allowing partial permeation^1,54,55. As for ultrafiltration – the process eliminated chlorides, sulfates, potassium and sodium ions from the solution to a low extent (< 10%). The highest efficiency of UF was observed for F⁻, Ca²⁺ and Mg²⁺, which can be attributed to adsorption onto activated carbon incorporated into the UF system rather than membrane sieving. Activated carbon exhibits surface functional groups (–OH, –COOH) capable of binding divalent cations and fluoride via surface complexation or electrostatic attraction. Obtained results are in accordance with Kurniasari et al. findings, which reported comparable ion adsorption efficiencies⁵⁶.

Metals

The highest metal removal efficiency was observed for the RO processes, at over 90%. Only for Cr and Zn was this efficiency slightly lower: 80 and 87%, respectively. Lower removal for Zn and Cd can be consistent with their smaller hydrated radii/weaker complexation and susceptibility to charge screening⁵⁷. The metal removal efficiency of the AC-UF process can be ranked in the order Fe > Mn > Al > Pb > Cd > Cr. The resulting order in the series is in agreement with studies by Gierak and Leboda⁵⁸, which show that the individual metals sorb onto the activated carbon in a similar order as they are placed in the metal ion exchange affinity series.

Microbiology

Based on microbiological analyses performed it was found that all membrane processes are capable of retaining bacteria and viruses. The AC-UF, NF, AC-NF and RO process ensured the complete retention of coliform bacteria, Eschericha coli and Enterococci. However, it was observed that the number of organisms was not reduced, especially in AC-UF filtration. Similarly low UF efficiencies were obtained by Marszałek and Dudziak². The literature on the subjects proves that the cells of microorganisms can penetrate the pores of the membrane with diameters much smaller than the dimension of the cells themselves due to pressure deformation with filtering out the intercellular fluid^2,59. Other studies have shown that some bacterial cells are capable of passing even through theoretically impenetrable RO membranes⁶⁰.

Although all membrane configurations effectively removed indicator bacteria (Coliforms, E. coli, and Enterococci), the persistence of total microorganisms in permeates, particularly after UF and NF treatment, indicates that membrane separation alone may not ensure complete disinfection. The remaining microbial population is likely composed of heterotrophic, non-pathogenic microorganisms capable of passing through membrane pores or surviving mechanical stress during filtration. While these microorganisms do not necessarily pose a direct sanitary risk, their presence may promote biofilm formation within storage or distribution systems, leading to secondary contamination and gradual water quality deterioration. Therefore, to guarantee the microbiological safety of treated rainwater—especially when it is intended for applications involving human contact—an additional disinfection step, such as ultraviolet irradiation, ozonation, or chlorination, is recommended following membrane filtration.

The conducted research indicated that the ultrafiltration (UF) membrane can be used mainly to eliminate high molecular weight substances, total suspended solids (TSS), monovalent ions like Ca²⁺ and Mg²⁺ and bacteria. However, this method is not effective in removing all microorganisms from the analyzed rainwater. The advantage of NF is the higher retention of dissolved substances and metals. The disadvantage of the method is the low removal rate of chlorides, whose concentration in rainwater is generally high. Purification of rainwater with RO makes it possible to achieve a higher treatment intensity.

Quality of purified rainwater

The quality of permeate after different filtration processes is presented in Table 5. The pH value of rainwater after membrane processes ranged from 7.39 to 8.18, which is within the range of DWD and WHO Standards^61,62 (Table 5). The turbidity was not acceptable only in the analyzed sample after the AC-UF process; its value was over 3 times higher than the one recommended by WHO⁶². After the RO process, the EC reached value of 11.67 µS/cm. The TDS content after the RO process was 0.020 g/L, and is lower than the recommended limit value by WHO⁶². The TDS value of water subjected to the AC-UF pretreatment process is almost 2 times higher than the recommended permissible value by WHO⁶².

In the analyzed rainwater samples, high chloride concentrations were found after the AC-UF, and AC-NF processes. According to literature data, chloride concentrations exceeding 250 mg/L give water a salty taste and can cause health problems, especially in people with heart or kidney diseases⁶³. Similar to chloride concentration, the sodium value was acceptable only after the RO process and was lower than that recommended by Polish standards and international regulations^61,62.

Rainwater after the AC-UF and AC-NF was characterized by the high Cu concentration, which exceeds the permissible limit of 2.0 µg/L according to both WHO and DWD norms. The RO process reduces the copper concentration to 1.910 µg/L, which remains below the permissible value. Rainwater after the AC-UF process has the highest zinc concentration, which exceeds the WHO reference value (3.0 µg/L) and the DWD (5.0 µg/L). The AC-NF processes also show zinc concentrations above the standard (12.000 µg/L and 11.181 µg/L, respectively). The RO process reduces the zinc concentration to 3.589 µg/L, which is below the DWD reference value but above the WHO standard. Water obtained from the AC-UF, and AC-NF processes does not meet drinking water quality standards regarding copper and zinc content.

The persistence of copper and zinc in UF and NF permeates at concentrations exceeding WHO guideline values indicates a potential health risk if the treated water is intended for human consumption. Elevated Cu and Zn levels are likely derived from the corrosion of metallic infrastructure, vehicle wear, and particulate deposition on impervious road surfaces. The limited removal efficiency of these metals in UF and NF processes can be attributed to their presence as dissolved ions or complexed species that are poorly retained by such membranes. From a practical perspective, this observation underscores the necessity of incorporating an additional treatment step prior to UF and NF, such as adsorption onto activated alumina, ion exchange, or chemical precipitation, to achieve compliance with health-based standards. Alternatively, post-treatment options—such as coupling with RO. The potential accumulation of these metals during long-term water reuse applications should also be taken into account in future research.

The absence of Coliform bacteria, Escherichia coli, and Enterococci is a positive result, suggesting that the water does not contain pathogens and meets sanitary and epidemiological standards. Permeate after AC-UF process does not meet the requirements for water intended for human consumption due to the total number of microorganisms incubated at 22 °C and 36 °C. Furthemore, permeate after AC-NF and RO contains excessive amounts of microorganism incubated at 22 °C what disqualifies water for human consumption. Removing pathogenic organisms from rainwater does not protect it from secondary microbiological contamination, and therefore final chemical disinfection of the water is necessary before it can be used for drinking purposes.

The results of the analysis of rainwater quality after using membrane processes indicate that permeate obtained from RO filtration could be used for economic purposes, such as flushing toilets, washing floors, and watering green areas. The factor excluding its use for drinking purposes is the total number of microorganisms. Permeate from AC-UF and AC-NF processes does not meet the standards regarding Cu and Zn content specified in Polish and international regulations. In order to potentially use it, it is necessary to introduce pre-treatment to remove or minimize the content of these metals. Research shows that RO gives higher permeate quality but requires potentially higher energy use than NF and UF process. Reverse osmosis (RO), particularly in water-reuse applications, exhibits higher specific energy consumption (SEC) values of 0.5–3.0 kWh·m⁻³, with energy strongly influenced by feed salinity, pressure, and recovery efficiency. Nanofiltration (NF) shows SEC in the range of 0.2–2.0 kWh·m⁻³, depending on feed quality and recovery, while UF typically operates at very low transmembrane pressures, resulting in SEC values of about 0.05–0.3 kWh·m⁻³ for surface water treatment^49,64,65,66. However, detailed information on energy consumption, as well as techno-economic and life cycle analysis, require broader and more detailed analyses. The comparison with WHO and EU standards was used as a reference framework for evaluating the purification efficiency.

Performance and applicability of nanofiltration and reverse osmosis in stormwater reuse systems

Compared to previous research on roof runoff reuse and mixed urban catchments, the present study highlights the distinct challenges associated with expressway runoff, which contains elevated concentrations of traffic-derived pollutants—particularly chlorides (567.3 mg/L), sodium (663.8 mg/L), and heavy metals such as copper (46.0 µg/L) and zinc (28.3 µg/L). These results confirm that runoff from transport infrastructure is considerably more mineralized and metal-enriched than roof-collected rainwater, thus requiring more advanced treatment solutions. This distinction underscores the need for hybrid membrane–adsorption systems specifically designed for road stormwater treatment, integrating metal-selective polishing and intermittent RO operation to ensure compliance with environmental regulations.

As reported by Liu et al.³, membrane-based technologies—including ultrafiltration (UF), nanofiltration (NF), and reverse osmosis (RO)—offer significant potential for rainwater treatment and reuse due to their high contaminant removal efficiency and operational flexibility. In agreement with these observations, the present study confirmed the performance hierarchy UF < NF < RO, with electrical conductivity removal rates of 2%, 42%, and 99%, respectively. NF and RO effectively reduced total dissolved solids (TDS removal of 50% and 98%) and ionic species, whereas UF exhibited limited retention of monovalent ions (Cl⁻ 3%, Na⁺ 8%).

In the context of integrated reuse schemes, Alhussaini et al.⁶⁷ noted that coupling desalination and membrane processes enables flexible control over effluent quality depending on reuse objectives. The findings of this study suggest that an adaptive treatment train combining activated carbon–nanofiltration (AC–NF) as a pretreatment step, followed by periodic RO operation, can optimize both energy consumption and permeate quality under variable rainfall and salinity conditions. The RO permeate exhibited EC = 11.7 µS/cm, TDS = 0.020 g/L, and Cl⁻ = 2.36 mg/L, meeting WHO and EU standards for non-potable reuse.

The moderate removal of PAHs observed in this study (38–57%) aligns with the findings of Zolghadr et al.⁶⁸, who reported synergistic effects between advanced oxidation and membrane processes in urban stormwater reuse. Incorporating a mild advanced oxidation process (AOP) or an adsorption step prior to NF could further enhance the degradation of hydrophobic organic compounds and improve permeate quality.

Similarly, Tsanov et al.⁶⁹ demonstrated that hybrid membrane systems incorporating adsorption or oxidation pretreatment improve flux recovery and reduce fouling potential. In the present study, the AC–NF configuration achieved a normalized flux of 79%, which was 31% and 54% higher than that observed for RO and UF, respectively. This confirms that upstream activated carbon adsorption effectively mitigates reversible fouling and enhances organic matter retention.

Finally, the present findings are consistent with the review by Feng et al.¹⁶, which emphasized that although membrane-based technologies provide high removal efficiencies for suspended solids and organic matter, their effectiveness in removing dissolved salts and monovalent ions remains limited. This limitation, also noted by Liu et al.³, reinforces the suitability of NF and RO for the treatment of mineralized road runoff, where salt rejection (> 98%) constitutes a critical performance criterion.

The analyzed express runoff collected in reservoir showed high levels of EC, TDS, Cl⁻ and Na⁺ which suggested the potential effects of pollution from road and street maintenance products. Studies have shown insufficient effectiveness of UF in removing contaminants. The obtained permeate was characterized by increased content of chlorides, sodium, Cu and Zn. The results indicate the need to increase the effectiveness of UF by using membranes with different properties to reduce fouling and improve efficiency or to introduce pre-treatment prior to UF to minimize membrane clogging and increase the fouling separation effect.

Obtained results demonstrated good effectiveness of AC-NF in rainwater purification. However, future research should focus on increasing efficiency of Cu, Zn removal and reducing fouling by introducing appropriate process parameters (e.g. variable pH).

The permeate obtained from the filtration processes does not meet the requirements for microorganism content. The quality of the permeate after RO allows it to be used for economic purposes, such as flushing toilets, washing floors, and watering green areas. However, due to the high pressures used in the RO process, the process may be too energy-consuming. Therefore, the next challenge for research may be a study on replacing the RO process with an NF process with appropriately selected pretreatment.

The permeate obtained after reverse osmosis (RO) treatment demonstrates sufficient quality for non-potable economic applications, such as toilet flushing, floor cleaning, and irrigation of green areas. However, due to the presence of residual microbial activity and trace amounts of heavy metals, the permeate does not meet the potable reuse standards set by the World Health Organization (WHO) and the European Union (EU) directives. Consequently, its use should remain limited to non-potable purposes unless additional disinfection and polishing processes are implemented.

Although membrane processes effectively removed indicator organisms, residual heterotrophic microorganisms were detected in UF/NF permeates. RO provided the highest microbiological performance, but downstream disinfection is still required for compliance with potable reuse standards. It is therefore recommended to integrate a disinfection barrier—such as UV or low-dose chlorination—and to monitor potential regrowth during water storage and distribution.