Enhancing stormwater treatment through ultrafiltration: Impact of cleaning chemicals and backwash duration on membrane efficiency

Stormwater runoff is known to be a contributor of pollutants discharged into water bodies. Stormwater management is necessary for a variety of reasons, particularly because of the environmental damage stormwater can cause, such as disturbing ecosystems and endangering the lives of various species (Prudencio & Null 2018; Levin et al. 2020). Due to climate change and its related problems, such as water scarcity, treating stormwater for reuse as a water resource is becoming increasingly important (Barbosa et al. 2012). Controlling, storing, treating, and reusing stormwater runoff could result in mitigating water scarcity and also prevent the pollution of natural water sources (Rupak et al. 2010). For industries that have high water demands, natural water resources are currently the main water resource, but it might be practical to reuse treated stormwater for industry processes, reducing the pressure on natural waters. This issue is particularly of interest to regions where such resources are limited.

Various centralized or decentralized stormwater treatment systems have been studied at laboratory and pilot scales, as well as being implemented in a full scale (Saraswat et al. 2016). The most common systems are blue-green infrastructure, such as ponds, wetlands, and bioretention systems, and infiltration (Karlsson et al. 2010; Lange et al. 2020). Treatment ability, capacity, area requirement, costs as well as legal requirements are important factors for decision makers to consider when choosing a treatment method.

Membrane treatment, which is known to be an efficient treatment method in the water and wastewater industry (Arévalo et al. 2009; Yadav et al. 2021; Bilal et al. 2022), is a promising option for stormwater treatment. The membrane process has the potential to treat stormwater to a high degree and produces a high-quality permeate, which could be reused for industrial purposes or as a water source for non-potable and potable water applications.

Recently, a number of research studies have been carried out on applications of membranes for the treatment of stormwater. Forward osmosis (Li et al. 2014), a gravity-based membrane reactor (Du et al. 2019), as well as microfiltration (MF) and ultrafiltration (UF) (Ortega et al. 2019) have been tested. In a study by Ortega et al. (2019), both MF and UF processes removed total and settleable suspended solids by more than 90%, and fecal and total coliforms were reduced to below 300 colony forming units (CFUs)/100 mL. In comparison to MF, a more stable flux was achieved using UF, and flux decline occurred faster in the MF membrane process. It has been suggested that stormwater treated with UF could be used for irrigation, flushing toilet, or washing streets.

UF membranes can be used for stormwater treatment because of two reasons. First, it has a higher removal efficiency in comparison to MF for the separation of pollutants from stormwater (Ortega et al. 2019). Second, UF membranes can separate stormwater pollutants, i.e., total suspended solids (TSS), turbidity, biochemical oxygen demand, and oils, well enough to meet the regulations for reuse (Ortega et al. 2019; Kaykhaii et al. 2023). However, research about membrane applications for stormwater treatment is still limited. During the membrane filtration process, the particles attach to the membrane surface and very small particles can stick inside the pores, which result in fouling. There are some common methods to disturb and partially remove the fouling layer on membranes, e.g., by physical methods (e.g., backwashing with gas or liquid, and forward flushing) and in combination with chemical cleaning. Depending on the water quality of the feed, the chemicals needed for membrane cleaning may differ (Cardew & Le 1999). Stormwater often contains large and fine-grained mineral particles, metals (Lindfors et al. 2020), oils, poly-aromatic hydrocarbons, and other organic micro-pollutants (Müller et al. 2020). However, the proportion of particulate organic matter is low in comparison to municipal wastewater (Lindfors et al. 2020; Philip et al. 2021). Previous studies mainly focused on the application of membranes for drinking water, industrial water, or wastewater treatment. Since the quality of stormwater is significantly different from other types of (waste)water, our study is highly relevant, especially with respect to stormwater reuse for different purposes. The insights that the study at hand provides on methods for membrane cleaning will help to design membrane processes for stormwater treatment, which will open new opportunities and facilitate the application of stormwater as a water resource. Effective membrane cleaning has a positive effect on the flux recovery and membrane lifetime, and it is important to find an efficient method to remove fouling substances from the membrane surface and pores.

The aim of this research study was to evaluate the effects of different backwash durations (varying from 30 to 60 s) and different combinations of cleaning chemicals (NaOH with or without NaOCl followed by HCl) on the efficiency of the membrane process used for stormwater treatment. Filtration productivity, backwash efficiency, as well as pure water reversibility and permeability of the membrane were the parameters assessed. The fouling layer on the membrane surface after the experimental runs was characterized using scanning electron microscopy (SEM) and energy-dispersive X-ray spectroscopy (EDX).

The stormwater used for the experiments was collected from a manhole of a stormwater sewer located in a commercial and light industrial area in Luleå, Sweden (65°37′09.8″N and 22°03′23.6″E). The catchment area included roads, pavements, parking lots, and industrial buildings. Stormwater runoff samples were collected during two precipitation events in two 1 m³ tanks, transported to the laboratory, and kept at room temperature until used in the experiments. The recorded stormwater quality values are shown in Table 1. Prior to feeding the stormwater into the UF membrane module, the stormwater in 1 m³ tank was stirred for 10 min and then a feed sample of approximately 15 L was collected and sieved through a stainless-steel sieve with a pore size of 315 μm. The stormwater was sieved to remove particles bigger than 300 μm as these particles could lead to faster clogging and harm the structure of the UF membrane.

To investigate the effect of backwash duration and different cleaning chemicals on membrane productivity, backwash efficiency, and pure water reversibility and permeability, a total of 10 experiments were carried out (Table 2), four experiments with a backwash duration of 30 s, four experiments with a backwash duration of 45 s, and two experiments with a backwash duration of 60 s. The run order of the 10 experimental runs was fully randomized. The backwash duration was chosen in a range that was reasonable in relation to the membrane treatment time of 1 h. In preliminary experiments, it was found that a backwash duration of 15 s was not sufficient to adequately rinse the membrane surface and remove particles that had weaker bonds with the adsorptive layer on the membrane surface. This was because the transmembrane pressure (TMP) rise was rapid and this backwash duration was not sufficient to remove the contaminants from the membrane surface and the pipes.

To remove the foulants on the membrane surface, membrane chemical cleaning was evaluated using two chemical combinations. NaOH with or without NaOCl was evaluated as a chemical for cleaning of membrane and removing the organic pollutants followed by membrane acidic washing with HCl to remove inorganic pollutants and for neutralization. The alkaline solution was a 1 M solution of NaOH and adjusted to the pH of 12. When chlorine alkaline solution was used, a 200-ppm chlorine solution was prepared with NaOCl in water and NaOH was added to adjust the pH to 12. An HCl solution with a pH of 2 was used for acidic cleaning of the membrane. Table 2 shows which chemical combination is used for each experiment. NaOH and HCl were supplied by Merck and NaOCl by FF-Chemicals AB.

Each experiment started by cleaning the membrane with one of the chemical combinations (Table 2). One of the alkaline solutions was added to the feed stream and the membrane was filled with cleaning chemical and left standing for 10 min before the membrane was acid washed using HCl with the same cleaning process. The temperature of the cleaning solution was kept constant (20 °C) for all experiments. After chemical cleaning, the membrane was backwashed for 1 min followed by 2 min of forward washing with Milli-Q water, which has been recommended after chemical cleaning (Chen et al. 2003). Pure water flux (PWF) was measured before and after chemical cleaning. Three different pump flow rates were determined, and flux was recorded manually using a beaker and scale 10 times during 10 min (once per minute). The pressure was recorded by pressure sensors during this time. This process was repeated two more times with higher flow rates.

Following this, a TMP of 1.8 bar was set and the experiment was started. After 1 h of stormwater treatment, the permeate volume was measured, and if the average flux had declined by 30% compared to the start of the experiment, the TMP was increased by 0.1 bar. Every hour, the membrane was backwashed using a peristaltic pump (150 ml/min) and Milli-Q water for either 30, 45, or 60 s. These 1-h cycles were repeated until the TMP had increased by 0.3 bar (i.e., from a TMP of 1.8 bar at the beginning to a TMP of 2.1 bar). The permeate flux was measured at the beginning, middle, and end of each 1-h cycle.

Feed samples were taken at the beginning of each of the 10 experimental runs and were analyzed with respect to TSS, turbidity, pH, conductivity, total and dissolved metal concentrations, total organic carbon (TOC), and Cl⁻. In addition, before starting the experiment, a pretest was carried out using Milli-Q water and a permeate blank sample, which was analyzed for the same parameters as previously described. Samples for analysis of dissolved metals were passed through a 0.45 μm filter. TSS, conductivity, pH, and turbidity were analyzed immediately after sampling. For TOC and Cl⁻, samples were stored at −18 °C, and for total and dissolved metals, the samples were stored at 8 °C until taken to the external laboratory for analysis.

The standard method SS-EN 872:2005 was used to determine the concentrations of TSS. Total and dissolved metal concentrations (except Hg) were analyzed using inductively coupled plasma sector filed mass spectrometry (ICP-SFMS) according to ISO 17294-2:2016. For Hg, the standard SS-EN ISO 17852:2008 was used. The reporting limits (RLs) for the total and dissolved metal concentrations were 10 and 0.2 μg/L for Al, 0.5 and 0.05 μg/L for As, 1 and 0.01 μg/L for Ba, 0.2 and 0.1 μg/L for Ca, 0.05 and 0.002 μg/L for Cd, 0.2 and 0.005 μg/L for Co, 1 and 0.1 μg/L for Cu, 0.9 and 0.01 μg/L for Cr, 0.01 and 0.0004 μg/L for Fe, 0.002 and 0.02 μg/L for Hg, 0.4 and 0.4 μg/L for K, 0.2 and 0.09 μg/L for Mg, 0.9 and 0.03 μg/L for Mn, 0.5 and 0.05 μg/L for Mo, 0.5 and 0.1 μg/L for Na, 0.6 and 0.05 μg/L for Ni, 0.5 and 0.01 μg/L for Pb, 1 μg/L for P, 0.2 and 0.005 μg/L for V, and 4 and 0.2 μg/L for Zn. The atomic fluorescence spectroscopy (AFS) method (ISO 17852:2008) was used for Hg analysis, and RLs for total and dissolved Hg concentrations were 0.02 and 0.002 μg/L, respectively. TOC was analyzed according to DIN EN 1484(H3) and the RL was 0.5 mg/L. Turbidity was measured using the turbidity meter 2100N (Hach, Loveland, Colorado), pH was measured using a WTW pH 330 electrode (WTW, Wielheim, Germany), and the conductivity was measured using a CDM210 conductivity meter. Cl⁻ concentration was measured by ion chromatography according to the method CSN EN ISO10304-1.

The scanning electron microscope (SEM instrument) was provided by PENTAIR, Sweden, to characterize the fouling layer on the membrane at the end of the experiment and was a combined system for SEM and EDX. The scanning electron microscope was JSM-IT500HR type from JEOL.

A module integrity test was carried out by PENTAIR, Sweden, on the module before autopsy. The permeate side was pressurized with air (1–3 bar), while the module was submerged in water. Any leak would have become visible because any compromised fiber would cause a bubble-train.

A partial least square (PLS) model was derived to find the statistical relationship between factors (backwash duration and type of chemical cleaning) and responses (productivity, backwash efficiency, pure water reversibility, and permeability).

Increasing the backwash duration from 30 to 60 s resulted in a 20% increase in productivity, on average. Optimum productivity (58 LMH), however, was achieved with a backwash duration of 45 s, which was 3% higher than the productivity with a backwash duration of 60 s. Although the experiment could run for a longer time using 60 s of backwash (Table 3), the total volume of backwash water that was needed for 60 s of backwash was large and the increase in permeate volume was not considerably higher than when 45 s of backwash took place, which resulted in lower productivity compared to the experiments with a backwash duration of 45 s. According to the PLS model, an increase in backwash duration had a positive effect on productivity (R²Y = 0.42, Q² = 0.2).

Backwashing and membrane productivity are closely interconnected, and a trade-off between productivity and backwashing is challenging. Increasing the frequency of backwash, backwash duration, or pressure can potentially have a positive or negative effect on membrane productivity. Therefore, it is important to optimize the backwash duration to remove the fouling layer without reducing productivity. The experiments using 60 s of backwash duration were continued for 50 h and stopped at the end of the 50th hour due to not observing TMP increase. If they had been continued until a TMP increase of 0.3 bar (as the other experiments), the productivity would have been even lower. However, a backwash duration of 60 s resulted in a higher backwash efficiency which is beneficial and results in an improved ability to control permeate flux and keep it constant for a longer time. Efficient backwashing helps to reduce the number of chemical cleaning runs required, which is important because chemical cleaning reduces membrane lifetime (Park et al. 2018). Therefore, a longer backwash duration can be beneficial despite the trade-off with productivity.

The mean permeability (Equation (4)) was positively affected by an increasing backwash duration (R²Y = 0.9, Q² = 0.65). For 30, 45, and 60 s, the mean permeability was 30 ± 3, 37 ± 3, and 47 ± 2 LMH bar⁻¹, respectively. This effect might be different for 60 s of backwash if the experiment had been allowed to continue.

Two chemical combinations were tested for membrane cleaning: NaOH with or without NaOCl, followed by HCl. According to the PLS model, the comparison between the efficiencies of chemical cleaning when the alkaline solution did or did not contain NaOCl showed that the addition of NaOCl had no significant effect (PLS) on chemical cleaning efficiency. Efficiency was evaluated by comparing the pure water permeability (Equation (5)) after chemical cleaning and average reversibility (Equation (3)). Reversibility after chemical cleaning (determined from the PWF after chemical cleaning) varied widely, ranging from 37 to 255% (average 96 ± 67%, R²Y = 0.8, Q² = 0.75) for the two types of cleaning tested (Table 2). The performance of chemical cleaning depends on the particles adsorbed on the fouling layer, as well as the structure of the fouling layer and how compact this layer was (Chen et al. 2003). During chemical cleaning, chemical reactions take place between the chemical and foulants, detaching the particles from the membrane surface.

Pure water permeability was calculated after using the two types of chemical cleaning (NaOH with or without NaOCl followed by HCl). The pure water permeability after cleaning with NaOH followed by HCl averaged 97 ± 13 LMH bar⁻¹ and after cleaning with NaOH + NaOCl followed by HCl averaged 117 ± 15 LMH bar⁻¹. Stormwater contains microorganisms, and the addition of chlorine solution to the chemical treatment could be suitable to prevent biological growth on the membrane surface, which is crucial for prolonging the membrane lifetime, especially in applications where biological growth is unavoidable (Decarolis et al. 2001). However, chlorine is harmful to the environment, and its use in stormwater treatment should be avoided whenever possible (Parveen et al. 2022).

Looking at the contour plot, it can be seen that the aging of the membrane during the experiments negatively affects the ability of the chemical cleaning on recovering pure water permeability. It should be mentioned that according to the PLS model, reversibility was positively correlated with productivity, backwash efficiency, and permeability.

The integrity test after the experiments showed that the membrane used for these experiments was not leaking and that the pretreatment method chosen was sufficient (in terms of stormwater quality) to prevent damage that could be caused by pollutants or large particles in the stormwater.

From the EDX results, it appears that the foulants on the membrane surface were metals belonging to the group of inorganic substances. In the case of inorganic foulants, the use of citric acid, sodium metabisulfite, sulfamic acid, or sodium metaphosphate may be another option for chemical cleaning of the membrane (Cardew & Le 1999). However, the application of these chemicals and the effects on this type of membrane need further study. It could also be interesting to characterize and compare the composition of the different foulants on the membrane surface before and after chemical cleaning and backwashing.

In this study, the effect of chemical cleaning and regular backwashing on the efficiency of a UF membrane used for stormwater treatment was investigated. The results of the experiments showed that increasing the backwash duration resulted in a more stable TMP over time, higher backwash efficiency (62 ± 2, 82 ± 6, and 92 ± 4% for 30, 45, and 60 s backwash duration, respectively), and permeability (the mean permeability was 30 ± 3, 37 ± 3, and 47 ± 2 LMH bar⁻¹ for 30, 45, and 60 s, respectively). Productivity was highest at a backwash duration of 45 s and was 3% higher than at 60 s. Since productivity is affected by backwash duration and flow rate, a balance between productivity, permeability, and energy consumption must be found to find the optimal operating conditions.

The membrane was chemically cleaned before each experiment and two different combinations of chemicals were used. The use of NaOH with or without NaOCl, followed by HCl, resulted in an average pure water membrane reversibility of 96 ± 67%, restoring flux and reducing TMP. The pure water permeability was 97 ± 13 LMH bar⁻¹ after cleaning with NaOH and HCl and 117 ± 15 LMH bar⁻¹ after cleaning with NaOH + NaOCl followed by HCl, although the difference was not statistically significant according to the PLS model.

The surface of the fouled membrane was examined with a scanning electron microscope. This showed that the fouling layer consisted of mineral particles, e.g., significant amounts of Si and Fe. In addition, small amounts of Si, Al, Na, K, Ca, and Mg were found on the inner surface of the membrane.

This research was jointly supported by DRIZZLE Center for Stormwater Management (grant no. 2016-05176), which provided financial support for salary costs for supervision and Formas (grant no. 2016-20075) for salary and experimental costs. In addition, the authors would like to acknowledge Peter Rosander, Ivan Mantilla, Emmanuel Okwori, Haoyu Wei, Nikita Razguliaev, and Ivan Milovanovic for their help and support in the field and laboratory.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.