Treatment of river water using modular gravity-driven ultrafiltration (GDU) for individual contingency water supply

Water source pollution is a serious problem for almost all countries around the world and poses a threat to the community that depends on the water source for daily activities. The contamination of river water is dominated by the human activities, which is manifested through the microorganism concentration in the water bodies (Fadzilah Ghazali et al. 2018). The increasing demand for clean water has prompted the reuse of even contaminated river water as alternative water resources. River water is usually reclaimed for several reasons including for drinking, agricultural purposes and many more. River water reclamation involves the removal of impurities such as turbidity, organic and inorganic chemicals, suspended solids, pesticides, nitrate, microbial contaminants and other pollutants (Sahu et al. 2020). The water quality from the source is crucial to human health (Govindan et al. 2015; Guimarães et al. 2018; Zhang et al. 2019). For surface water treatment that produce potable water, presence of Escherichia coli in the produced water is one of the indicators that determine the efficacy of the treatment method.

E. coli is always used as indicator of faecal contamination in water bodies, as it is usually released through the deposition of faecal material (Ishii & Sadowsky 2008). E. coli can cause various water borne diseases such as gastrointestinal contagion like non-bloody or bloody diarrhoea and haemorrhagic colitis (Baschera et al. 2019). E. coli contamination were reported either in most probable number (MPN) or colony-forming unit (CFU) per 100 mL. In natural municipal wastewater, the typical ranges of E. coli concentration is within 10⁵–10⁸ MPN/100 mL while the median infectious dose is in the range of 10⁶–10¹⁰ MPN/100 mL (Pillai et al. 2011). World Health Organization (WHO) has classified E. coli concentrations based on low-, intermediate-, high- and very high-risk levels for the range of 1–10, 11–100, 101–1,000, >1,000 (E. coli/100 mL) respectively, according to the Guidelines for Drinking-water Quality (Kayser et al. 2013). Due to the absence of centralized system at remote or isolated area, accessibility to the microbiological risk-free potable water is challenging (Tang et al. 2021). The presence of pathogenic microorganisms in drinking water such as E. coli, which is a member of the faecal coliforms group of bacteria (Pérez-Vidal et al. 2019) has become an increasing concern around the world. Zhang et al., summarized that stormwater contained around 500–16,000 MPN/100 mL of E. coli (Zhang et al. 2014). It has been declared that the pathogenic microorganisms are the culprit for various waterborne diseases, such as diarrhoea and gastrointestinal, which in turn cause about 2,000,000 deaths per year (Hashim et al. 2020).

Membrane filtration technology has become an interesting alternative for water treatment process due to its smaller footprint and its ability to polish the quality of the water (Ang et al. 2011; Guo et al. 2012; Uribe et al. 2015). Membrane filtration is an effective isothermal and physical method used to remove the pathogen from the water. For example, a minimum of 2 log reduction of E. coli using ultrafiltration and 1.5 log reduction using nanofiltration has been reported (Krzeminski et al. 2017). In one of the studies using treated wastewater for irrigation, it was found that the secondary treatment method could remove about 2 log of the initial bacteria concentration whereas the ultrafiltration membrane could achieve 3 log reduction on average (Vergine et al. 2017). Recent investigation using pot ceramic filter technology to improve the domestic water quality was also reported with 2 log reduction value (LRV) for E. coli and 1 LRV for Salmonella spp. from initial concentration (10⁴, 10³ and 10² CFU/mL) of E. coli and Salmonella spp., respectively (Pérez-Vidal et al. 2019).

Among the membrane filtration technologies, ultrafiltration (UF) is considered one of the favoured technologies for reclaiming wastewater or plays the role as pre-treatment unit for other water treatment technologies (Michael-Kordatou et al. 2015). Gravitational driven ultrafiltration (GDU) is known as a membrane filtration process that requires minimal energy during the filtration. In GDU membrane system, water will be driven through the membrane based on hydrostatic pressure while the microorganism and some contaminants with bigger size can be removed. In that case, membranes with optimum pore size that balance between the flux (driving force) and the removal efficiency must be achieved. Previous studies indicated that during algal bloom period, the gravity-driven membrane (GDM) system was able to remove toxic cyanobacterial metabolites (microcystins). However, a much lower flux was obtained due to the algae attachment on the surface that reduced the membrane flux (Kohler et al. 2014; Silva et al. 2018).

A recent study has compared the application of microfiltration (MF) vs. UF in treating algae-polluted lake water in terms of the membrane performance and water quality through a GDM system (Truttmann et al. 2020). The results showed that UF-GDM showed more permanent-fouling resistance compared to the MF-GDM system. A stable flux of 2–4 L/m² hr was achieved using GDM filtration for secondary wastewater treatment and more importantly the fouling layer could be removed via physical cleaning (Wang et al. 2017). Another study was carried out to recycle rainwater with a steady flux of 6 L/m² hr and a similar finding concluded that physical flushing is adequate to recover 90% of the flux (Ding et al. 2017).

The GDM process has been reported to demonstrate a limited removal capacity of natural organic matter (NOM) (Tang et al. 2018). Before the conventional UF process, coagulation was employed to remove colloidal and soluble organic matters to prevent membrane fouling. In one of the studies, the GDM system including a pre-coagulation system, in-line coagulation system, and system without coagulation were operated for treating Yangtze River water. Results showed that all the systems revealed an excellent (96%) ammonia removal efficiency. However, due to the nitrification process, the coagulation was extended from 3 days to 5 days (Huang et al. 2021). Nonetheless, it was notable that GDM can diminish the concentrations of inorganic pollutant such as Mn²⁺ in the manganese contaminated surface water (Tang et al. 2020).

The advantage of the GDM process is its flexibility and can be used at any place even without electricity supply especially at the remote area. Nonetheless, the GDM setup suffers from lower flux (hydrostatic dependent) and is susceptible to fouling due to its mode of operation. In that case, the balance between the membrane flux and rejection performance must be evaluated according to the needs (Pineda et al. 2021). Due to their simplicity, GDM filtration systems are notably suitable to supply water for remote or disaster-struck area (Tobias & Bérubé 2020). Furthermore, according to World Health Organization (WHO), the GDM system could meet the performance requirements of ‘highly protective’ household drinking-water treatment options in term of bacteria removal. It is therefore considered a suitable approach for a decentralized drinking-water treatment system (Schumann et al. 2020).

In most cases, the construction of a point-of-use system does not come along with the pre-treatment system due to limited space and site accessibility. In light of the above considerations, the purpose of this work is to carry out water filtration in a flexible single or double GDU membrane unit to filter the contaminated water. The extent of membrane fouling under mild hydrostatic pressure and its performance in terms of flux and permeate water quality were evaluated. In this study, we are using a membrane with bigger pore size to enhance the flux due to its low hydrostatic pressure. Its water quality could then be polished by allowing the system to be operated under single or double unit in series.

In this study, E. coli was used as the feed solution. E. coli broths were cultivated continuously in six 250 mL conical flasks containing 13 g of nutrient broth dissolved in 1 L distilled water associated with 0.195 g/L peptone, 0.039 g/L yeast extract, 0.078 g/L NaCl and 0.013 g/L D (+) – glucose for 3 days to achieve sufficient cell density. The E. coli cells were cultivated at temperature of 252 °C for 24-hours using an orbital shaker (IKA, KS 501 Digital) at 100 rpm. All experiment materials were autoclaved for 15 min at 121 °C before cultivation. The E. coli cells' images were captured using a light microscope (OLYMPUS, BX53) at a magnification of 20x and 40x as shown in Figure 1. In general, E. coli cells exist in a regular rod-like shape with an average size of about 1.0 μm width and 2.0 μm in length (Ibrahim et al. 2015). The E. coli suspension was diluted to 100,000 times to enable cell count. Two E. coli cell suspension namely 5,000 and 10,000 MPN/100 mL were prepared by diluting the stock solution with the initial cell density of 320,000 MPN/100 mL. The calibration curve of E. coli cells concentration (MPN/100 mL) against absorbance reading (OD600) is developed in order to determine the concentration of cells in the permeate.

The setup of bench top dead-end gravitational driven ultrafiltration (GDU) membrane system is shown in Figure 2. The GDU membrane module comprises of 30 pieces of commercial polyvinylidene fluoride (PVDF) hollow fibre membranes (Hangzhou Waterland Environmental Technologies Co. Ltd, Zhejiang, China) with a length of 60 cm assembled vertically in a 30-cm-height tubular column in two-pass (folded membrane) to give a total filtration area of 0.0792 m². The hollow fibre used in this study is associated with inner diameter, outer diameter and nominal pore size of 0.7 mm, 1.4 mm and 0.03 μm, respectively. The column for the tubular hollow membrane module was made of polyvinyl chloride (PVC) pipe of 30.0 cm length and 0.3 cm wall thickness and internal diameter of 2.0 cm.

The setup was placed on the bench with the retort stand to maintain the hollow fibre membrane module in the upright position to prevent any differences, which gave the maximum hydrostatic pressure of 30 cm, which is equal to the maximum of 30 mbar. In this study, there were two configurations of GDU membrane, namely a single membrane module and a double membrane module, in series. The feed solution was poured into the membrane module from the top. In this study, the filtration was carried out in batch filtration mode where the feed solution was poured from the top to the filtration module and then the solution level allowed to deplete over time throughout the UF process. The permeate passed through the membrane wall into the lumen and collected at the bottom of the filtration module. For every feed solution with different E. coli concentration, a fresh and clean hollow fibre membrane was used for the UF process without undergoing recycling or cleaning. The permeate was collected at the bottom of the column using 250 mL of glass beaker and measured as weight using an electric balance (FX-3000i, A&D Company, Japan) connected to data-acquisition system.

The concentration of E. coli solution in the permeate was obtained based on absorbance at wavelength of 600 nm (Beal et al. 2020) using Cary 60 UV-Vis Spectrophotometer (Agilent Technologies, Santa Clara, USA). The concentration of E. coli solution was determined based on a calibration curve plotted for UV absorbance against the predetermined cell density.

The mean pore size of the hollow fibre PVDF membrane was determined using a porometer (Porolux 1000, Benelux Scientific, Belgium). The composition and morphology of the deposit layer on membrane surface, before and after the ultrafiltration process, were examined under scanning electron microscopy (FEI Quanta 450, United Stated), coupled with energy dispersion spectrometry (EDS, EDAX company, USA). The surface hydrophilicity of the clean and fouled hollow fibre membranes were analysed using a dynamic contact angle analyser instrument (Lauda LSA200, Germany) via sessile droplet method. The contact angle using deionized water as probe liquid was analysed using Surface Meter™ version 1.2.1.9 (LAUDA Scientific) software. The contact angle was measured immediately after dropping 5 μl of deionised water on the surface. Three measurements were repeated to minimize its error and the results were reported as an average value (Liu et al. 2012; Kumar et al. 2013; Hebbar et al. 2017).

A real water sample was collected from Kerian River located at (5°09′50.2″N 100°28′35.0″E), Nibong Tebal, Penang. 8 litres of river water sample were collected to determine the total suspended solids (TSS), mixed liquor suspended solids (MLSS), mixed liquor volatile suspended solids (MLVSS), chloride, ammonia, nitrate, nitrite, sulphate, chemical oxygen demand (COD), hardness, turbidity, pH, conductivity and total dissolved solids (TDS). Then, the river water sample was left for 24 hours settling and the supernatant was collected for the UF process using the GDU membrane system.

The bacteriological examination was conducted by Malayan Testing Laboratory Sdn. Bhd. Penang. All bacteriological analyses were carried out according to the Standard Method for the Examination of Water and Wastewater. The method used by the laboratory for the bacteria testing depended on the type of bacteria. For the microbiological parameter the permeate sample was examined for four types of widely used bacterial indicators including total coliform, E. coli, Staphylococcus aureus and Pseudomonas aeruginosa. E. coli and total coliform count were quantified by using MacConkey broth, and E. coli was incubated at 42–44 °C for 24 hours while the coliform count was incubated at 37 °C for 24 hours. The bacteria were then enumerated based on APHA 9221 E method. The presence of S. aureus and P. aeruginosa was assessed following the British Pharmacopoeia as described in 2014 XVI B (Ishak et al. 2021).

A continuous GDU module was set up to allow extensive membrane fouling occurred under continuous mode (Figure 3), followed by backwash cleaning under constant hydrostatic pressure (Figure 4). The GDU system comprised two parts. The bottom pipe on the right consists of 30 pieces of hollow fibre membranes (PVDF, 0.03 micron) with a length of 30 cm assembled vertically to give a total filtration area of 0.0792 m². The top pipe on the right consists of a 30 cm-height empty tubular column, connected to the bottom pipe that was filled with membranes, to provide hydrostatic pressure of 60 mbar. Both the vertical pipes on the left were empty pipes for feed water recycling purposes. The overflow of the feed solution was channelled back to the river water feed tank through the top horizontal pipe and the empty pipes on the left. The purpose of continuous filtration at higher constant hydrostatic pressure is to increase the extent of fouling (resembling multiple use of depleting hydrostatic pressure).

Backwashing was performed in a flow direction opposite to the UF process by forcing deionized water (NANO Pure water purification system, Barnstead) with resistivity greater than 18 MΩ-cm through the PVDF hollow fibre membrane (inside out). It was performed by reversing the membrane module and allowing the pure water to be forced into the lumen and then penetrate through the pores (Figure 4). The process of cleaning was repeated for 5 cycles before subsequent filtration to determine the flux recovery percentage. The flux recovery was calculated by dividing the flux after cleaning by its initial flux whereas the rejection recovery was determined by comparing the rejection of the membrane after fouled and cleaned with the initial rejection of the membrane for each feed solution (pure water and water samples). The adopted cleaning strategy is to detach foulants from the membrane surface especially within the membrane pores via hydrostatic force.

In this study, two initial concentrations of E. coli cells (5,000 MPN/100 mL and 10,000 MPN/100 mL) were fed to the GDU membrane system under depleting mode and filtered via two different membrane configurations namely single and double filtrations modes. The feed solutions were allowed to deplete until the end. Figure 5 shows that during single membrane filtration configuration, the highest flux obtained for 5,000 MPN/100 mL of E. coli concentration solution was 16.17 L/m² hr. Then, the flux was declined to 11.52 L/m² hr after 12 min of GDU filtration and finally dropped to 0.21 L/m² hr after 75 min of total filtration time (column emptying). For 10,000 MPN/100 mL of E. coli concentration, similar maximum flux was obtained after 10 minutes of operation and dwindled to 0.20 L/m² hr after 75 min. Throughout the 75 min of the filtration period, suspension with higher E. coli density (10,000 MPN/100 mL) depleted faster than the suspension with lesser cell density due to its thicker cake layer.

In contrast, as shown in Figure 6, when the suspension was filtered using double modules (in series configuration), the maximum permeation flux was much lower (reduced 90%) compared to the single filtration module. However, the flux was smoothly decreased compared to the filtration using single module. For the low cell density system, the maximum flux around 1.58 L/m² hr was obtained after 12 minutes of filtration. However, the fluxes gradually decreased to 0.08 L/m² hr after 10 minutes of filtration regardless of the initial feed cell density. This showed that the overall flux was controlled by the second module and the similar depletion curve showed that membrane fouling of the second module was not a critical issue after foulants (cells) were removed by the first module. As expected, the alleviation of total flux was due to the reducing hydrostatic pressure in the second module, but the smooth flux declination of both suspensions showed that fouling became insignificant for the second module.

Figure 7 shows the removal efficiency of GDU membrane system toward E. coli cells. The single filtration module GDU membrane system could moderately remove the E. coli with a rejection rate of 90.22 ± 1.19% and 86.35 ± 1.45% for initial E. coli cells concentrations of 5,000 and 10,000 MPN/100 mL, respectively. As the rejection is concentration dependent, the E. coli rejection rate could be enhanced by allowing the rejection in series using a two modules system. As shown in Figure 6, the removal of E. coli was enhanced with the use of double filtration module of GDU membrane system and the superiority of double filtration module over single filtration module of GDU membrane system was verified. The rejection performance was improved drastically by applying double filtration module of GDU membrane system operated in series configuration. The rejection percentages of 99.03 ± 0.39% and 97.70 ± 0.37% for E. coli cells were achieved for initial feed concentrations of 5,000 and 10,000 MPN/100 mL, respectively.

These results showed that separation of E. coli using this UF membrane is concentration dependent and permeation of E. coli through the membrane is viable. However, the E. coli cells exhibits cylindrical shape with a size of about 2 μm in length and 0.5 μm width. Based on the membrane nominal pore size of 0.03 μm, it is unlikely that E. coli cells could penetrate the UF membranes. According to the literature on the separation of E. coli suspension solution with 10⁷ CFU/ml, the permeability of the bacteria through the UF membrane was in the range of about 10⁻³%, which is near to perfect permeation restriction (Kobayashi et al. 1998). The most probable reason that E. coli permeation occurred in this work was due to a defect on the membrane surface. This situation is aggravated by the E. coli cells that deposited on the membrane surface, which negatively impact the membrane rejection and permeation flux. However, by employing the two modules configuration, the microbial removal efficiencies could reach up to 99% due to the additional resistance to microbial penetration.

The SEM images in Figure 8 reveal that E. coli was deposited on the membrane surface and exhibited both normal rod-shape and budding structure. The membrane surface of single module and double module suffered from different extents of permeate flux declination, which was ascribed to a difference in formation of foulants. Past studies demonstrated that the E. coli cells have the capability to promote the formation of biofilm (Levantesi et al. 2012; Yang et al. 2017). During the filtration, microorganisms attached on the surface of the membrane layer secreted extracellular polymeric substances (EPS), which consist of proteins and carbohydrates (Kochkodan & Hilal 2015). The secreted EPS developed microbial colonization on the membrane surface and led to biofouling. As can be clearly seen in Figure 8(a) and 8(b), the foulant layer in the first module was severely fouled with aggregated E. coli compared to the second modules of GDU membrane module as shown in Figure 8(c) and 8(d).

Unlike the other foulants, the surviving microbes can still multiply by using the biodegradable substance and proliferate in the feed solution over the time, although 99.9% of the biofoulants have already been eliminated (Nguyen et al. 2012). The cells' attachment pattern on the membrane surface increased rapidly and continuously until the end of the experiment regardless of the permeation flux (Jadoun et al. 2018). Figure 8 shows that only aggregated cells of E. coli can be observed on the membrane surface of the first module of the GDU membrane system. It was established that the relevant water-related pathogenic bacterial species will speed up the formation of biofilm, which leads to a reduction in the efficiency of the GDU membrane system (Mackowiak et al. 2018).

Since a visual inspection of SEM images was not able to allow quantitative analysis of inorganic constituents in the GDU membrane system, energy-dispersive X-ray (EDX) analysis was performed and the results are shown in Figure 9. EDX results show that water contaminated with E. coli cells for both 5,000 MPN/100 mL and 10,000 MPN/100 mL E. coli suspensions contain similar constituents, which included carbon (C) and oxygen (O), magnesium (Mg), aluminium (Al) and silicon (Si) elements. The increase of the elemental peaks of Mg and from UF of water contaminated with 5,000 MPN/100 mL to 10,000 MPN/100 mL of E. coli cells, demonstrated that the deposited elements were also fouled predominantly by inorganic substances. As can be seen in Figure 9, the elemental peaks of O and C became sharper over the increasing concentration of E. coli cells, indicating that more EPS was secreted by the feed with higher cell density and gradually built up on the membrane surface. The EPS secreted by the E. coli cells, which are made of polysaccharides and protein, could build up the fouling layer and eventually increase the fouling rates of the membrane surface of the hollow fibre membrane used in the GDU membrane system. Besides, the silica content also increased for suspension with higher cell density. This could contribute to the faster development of fouling layer at the initial stage.

Water contact angle measurement were conducted to examine the changes of the membrane surface energy after fouling. The contact angle value of the membrane surfaces in the first module increased significantly after fouled by the cell suspension. The contact angle value of pristine PVDF hollow fibre membrane was 67.6°. After filtering 5,000 MPN/100 mL of E. coli suspension, the contact angle value was increased to 90.9°. Similarly, after ultrafiltration of 10,000 MPN/100 mL of E. coli suspension, the contact angle was increased to 95.1°. This means the hydrophobicity of the membrane was increased due to the adherence of E. coli cells on the membrane surface. The increased contact angle value was due to the increase of surface roughness after surface deposition of E. coli cells on the membrane surface. Besides, it might be due to the EPS, which interact with the membrane surface via hydrophobic interaction. Past investigation had established that two predominant factors which facilitate the formation of foulant layer on the membrane surface were the EPS and soluble microbial products (SMP) (Aslam et al. 2017). Therefore, it is most likely that higher contact angle value is due to the formation of biofilms from E. coli on the membrane surface.

On the other hand, the water contact angle value for the second filtration modules were only slightly increased to 70.9° and 71.5° for the ultrafiltration of 5,000 and 10,000 MPN/100 mL of E. coli suspension, respectively. These results implied that the fouling condition of the second filtration module were not as severe as the first filtration module. The fouling on the second module was dominated by the less hydrophobic inorganic materials such as the silica and magnesium. Thus, double filtration module is an effective method that could reduce the propensity of membrane fouling while enhancing the permeate quality.

In the operation of GDU membrane system, TSS and volatile suspended solid (VSS) are the main parameters that will impact UF performance, as stated in the recent study (Nguyen et al. 2014). The TSS and VSS obtained from the river water sample were 95.22 ± 1.86 mg/L and 14.17 mg/L, respectively. VSS is the organic or volatile suspended solids portion used as a measure or indicator of the microorganism present. In this case, the high TSS can cause a thicker layer of foulant on the membrane surface and a more severe concentration polarization, which eventually decreases the water permeation flux, while the VSS could change the surface hydrophilicity (Bhinder et al. 2018).

Kerian River water sample was also filtered using the single and double filtration modules to verify its performance. For single membrane filtration configuration, the highest flux of 13.07 L/m² hr was obtained. Then, the flux was falling off rapidly from 8.14 L/m² hr to 1.49 L/m² hr within 20 min and reached 0.16 L/m² hr at the end of the filtration process. On the other hand, when the river water was filtered using double modules, the initial flux increased to 2.45 L/m² hr within the first 12 minutes and the fluxes gradually decreased after reaching the peak due to reducing hydrostatic pressure. The flux was continuously declined to 1.39 L/m² hr within 30 min and reached 0.08 L/m² hr at the end of the filtration process. In this double GDU membrane system, the resistance of the permeation flux was transferred from the cake layer and biofilm resistance in the first module to the membrane resistance in the second module. The flux profiles shown in Figure 10 was mainly contributed by the membrane pore resistance as well as the reduction in the hydrostatic pressure. The implementation of two membrane filtration in series work effectively in removing the contaminant from river water thus improved the permeate water quality.

Table 1 shows the water qualities of the Kerian river water. Basically, the quality of the river water is quite poor with high turbidity due to suspended solids and contaminated with ionic compounds such as sulfates and total dissolved solid TDS.

The physicochemical and nutrient parameters of water quality of the Kerian River when filtered by the double filtration module of GDU membrane system are presented in Table 2. It was found that the quality of the permeate fulfilled the requirement of drinking water for almost all the quality indicators especially for the heavy metal contents, hardness, salt and colour. However, it was notable that the phenol and ammoniacal nitrogen concentrations are slightly higher than the maximum permitted concentrations due to its higher volatility. Based on the results, phenol in the permeate samples amounted to 0.006 mg/L, which was 3 times higher than the maximum permitted proportions of 0.002 mg/L. Since there are industrial and agricultural activities nearby Kerian River, the phenol compounds found in the discharged river water source was unavoidable. Ammoniacal nitrogen was noted to be 1.0 mg/L, which is higher than the desirable limit of 0.5 mg/L. The presence of ammoniacal nitrogen is most likely due to the waste generated from industrial activities, especially from palm oil mill effluent.

The UF process using a GDU membrane system could efficiently remove the contaminants or foulant of the feed river water. However, UF is known to be a membrane that has poor retention of the dissolved ionic salt due to its bigger pore size. During the experiment, partial reduction of chloride, ammonia, nitrate, nitrite and sulfate concentration were observed. Based on the result, chloride was reduced from 7.83 mg/L to 4.16 mg/L and sulfate was successfully removed from 23.43 mg/L to 15.30 mg/L at the end of the filtration process. The poor chloride and sulfate removal was an expected phenomenon as UF membrane is known to be poor in retaining the ions. The average pore size of membrane is around 30 nm, which is much bigger than the size of the TDS like sulfate and chloride. However, its partial removal might be due to the blocking effect by the precipitate and not by the pore itself. Although at the beginning, the COD was 45.00 ± 2.65 mg/L, the settlement-assisted GDU membrane system was able to reduce the COD to an undetectable level, showing that the organic matter content can be effectively removed. It is important because COD is a crucial parameter in determining the water quality that represents organic loading in the water (Verma & Singh 2013).

Unexpectedly, merely 80% of the hardness in the river can be removed, which indicates that the hardness may appear in both dissolved and precipitated form. Most hardness is in particulate structure and easily retained on the membrane surface. On the other hand, the feed turbidity dropped significantly from an initial value of 67.07–0.31 NTU, which shows that the GDU membrane is very effective in suspended solids removal. No significant changes on pH were observed. Other parameter such as conductivity and TDS shows more than 50% of reduction after the filtration using single module of GDU with rejection of 57.14% and 59.08% respectively. It can be concluded that the proposed system is more effective in removing solid and bigger organic molecule which showed the typical separation characteristic of the UF membrane even though ultra-low pressure was applied (Wang et al. 2019).

The bacteriological examination of the permeate water are summarized in Table 3 and compared to the National Standard prescribed for drinking water quality. There were four parameters used for the bacteriological examination including coliform count, and the presence of common harmful bacteria such as E. coli, S. aureus, and P. aeruginosa. E. coli is the major species in the faecal coliform group. E. coli is considered to be the species of coliform bacteria that is the best indicator of faecal pollution and the possible presence of pathogens. The presence of E. coli is considered a more accurate indication of animal or human waste than the total coliforms. The bacteriological analysis showed the presence of coliform contamination in the river water feed solution with coliform count of 1,100 MPN/100 mL, whereas the coliform count in the river water permeate was found to be reduced by 96.45% to be 39 MPN/100 mL. The value of the coliform count was slightly higher than the desirable limit as well as permitted proportion based on drinking water standard (10 MPN/100 mL). For a GDU membrane system, such pathogenic microorganism may also be a source of contamination of the permeate quality due to the storage problem, which allowed the proliferation of the total coliforms. Similar findings were reported before, showing that indicator bacteria were detected after 24 hours of storage (Chaidez et al. 2016).

E. coli as much as 32 MPN/100 mL was detected in the river water together with S. aureus, which indicated that the river water was badly contaminated. Nonetheless, after filtration, all the pathogenic bacteria, namely E. coli, S. aureus, and P. aeruginosa contamination, were absent from the permeate. Based on the current acceptable upper limit for E. coli content in fresh surface water of 900 cfu/100 mL (Price & Wildeboer 2017), water with below 1 cfu/100 mL of E. coli concentration was good to be used as drinking water (Huang et al. 2020). Considering the low concentrations of the indicator components in the permeate and its efficacy in pathogenic bacterial reduction, it can be concluded that the developed double filtration module of GDU membrane system can meet the treatment requirement for Class I water production using this simple point-of-use GDU membrane system. Nonetheless, the permeate quality can be further polished for drinking water by using activated carbon to remove the phenolic compounds and UV deactivation for total coliform removal.

The SEM images of the fresh and fouled membrane underwent the filtration of river water are illustrated in Figure 11(a)–11(c). A significant difference of fouling layer development was observed on the surface of membranes. Based on Figure 11(b), a rough and dense foulant layer was formed on the membrane surface of first module as compared to the clean membrane surface (Figure 11(a)). The cake layer of the second filtration module exhibited a lesser and scattered foulant structure and formed more permeable fouling layers. The flake-form fouling layer on the first and second module is most likely the humic substance and crystalline salt from Mg and Si that present in the river water. Overall, the permeability loss in this case was due to the deposition of impurities within the membrane pore or on the membrane surface (Gao et al. 2011).

The foulant from the river water on the membrane surface were further characterized by EDX. Figure 12 shows the EDX spectra result of fresh hollow fibre membrane and fouled membrane surface (second module) after the filtration process of real water. As shown in Figure 12, the fresh PVDF membrane mainly consists of carbon (C), oxygen (O) and fluoride (F). On the other hand, with the deposition of impurities from the river water on the membrane surface after filtration, the elements of C and F peaks were slightly weakened from 30.72 to 29.79 wt% and 55.99 to 49.91 wt%, respectively. Meanwhile, the intensities of elements such as O, magnesium (Mg) and silicon (Si) were profoundly escalated after filtering the river water with the increment by 72.42%, 91.09% and 94.55% respectively. This indicated that the surface of the second module was mainly fouled by the inorganic substance rather than the organic materials.

The contact angle value of the membrane surface increased significantly upon river water filtration. Again, the increased contact angle value was due to the increase of surface roughness after deposition of the organic materials partly contributed by the microorganism. The contact angle value of fresh hollow fibre membrane was 67.6° but the first module of the GDU membrane system was increased to 91.8° after filtration. Similar to the synthetic E. coli test, the value of the water contact angle for the second filtration module was slightly increased to 77.2°, which is much lower than the first module. During the river water filtration, the membranes' surface was exposed to microbial fouling by the attachment and multiplication of bacteria on the surface to produce biofilms (Sawada et al. 2012). The increased contact angle value for the several foulants might be due to the EPS from the pathogenic microorganisms and colloidal and soluble organic matter from the river water samples. On the other hand, the fouling on the second filtration module was dominated by the inorganic materials.

Under constant hydrostatic pressure (60 mbar) and single unit filtration, a constant flux of 19.52 L/m² hr can be achieved. Upon cleaning via backwashing and re-filtering with the same river water, the flux can be maintained at 18.31 L/m² hr. This proves that backwashing is adequate to remove the fouling layer on the membrane surface and within the membrane pores with 93.8% recovery. Despite the flux drop around 6%, high water flux recovery (93.8%) is achieved. Recent study stated that backwashing procedures are critical for the longevity and proper functioning of the low-pressure membranes. Therefore, this operational procedure has become a critical component in ensuring consistent water quality, security and ability to provide safe and clean water supply, especially in drinking water production (Chang et al. 2017).

A GDU unit was developed to produce drinking water by removing the contaminants and pathogens from the river water. The double filtration unit demonstrated the ability of the GDU membrane module to remove bacteria and to produce permeate that is save to be used as potable water. The GDU membrane system is able to perform excellent rejection of E. coli of 99.03 ± 0.387% and 97.70 ± 0.368% for initial E. coli concentrations of 5,000 and 10,000 MPN/100 mL, respectively. The maximum permeate flux of E. coli cells for single and double filtration module of the unit could reach around 16.17 L/m² hr and 1.58 L/m² hr within 10–20 minutes of filtration period. The first filtration module was able to remove most the foulants especially the bacteria and led to the low fouling propensity of the second filtration unit. Application of the GDU to real river water filtration showed that the unit is able to eliminate pathogenic waterborne bacteria but its efficacy in removing volatile components such as phenol and ammoniacal nitrogen is less satisfactory. The first module is able to remove most of the organic and biofoulants while the second module is able to remove the inorganic foulants. A slight flux decrement was noticeable but it was able to be cleaned through 5 cycles of backwashing (up to 93.8% flux recovery) with slight impairment of rejection. The obtained results are valuable for the design of point-of-use GDU for reducing drinking water problems in rural and remote areas.

This work was supported by the Malaysia Research University Network (MRUN) Collaborative Research Program (203/PJKIMIA/67216003), Ministry of Higher Education Malaysia.

Data cannot be made publicly available; readers should contact the corresponding author for details.