Virus removal of new and aged UF membranes at full-scale in a wastewater reclamation plant

Abstract

Ultrafiltration (UF) membrane technology is widely used in water recycling schemes as a physical barrier for the removal of human pathogens. Assessing the consequence of long-term operation (membrane ageing) as well operational extremes on pathogen removal performance is crucial for assessing health risks of human exposure to recycled water. In this study, challenge testing with MS2 bacteriophage (MS2) was undertaken to validate the integrity of full-scale UF membranes used for virus removal in a wastewater recycling plant (WRP). Validation was performed on new (6 months old) and aged (6 years old) membrane units. The impact of a hazardous event—in the form of a short-lived, high turbidity spike—was also assessed since this event affected the hydraulic throughput of exposed membrane units. Validation of the new membranes demonstrated a mean virus log removal value (LRV) of 3.0 log₁₀ (SD = 0.6), with 5th and 95th percentile values ranging between 2.3 and 3.9 log₁₀ units. Re-validation of the membranes exposed to the hazardous event as well as those that were aged demonstrated that the virus LRV was comparable and had reduced by 1.0 log₁₀ units, measuring 2.0 log₁₀ (SD = 0.2), however the LRV performance was more stable, where the 5th and 95th percentile values ranged between only 1.7 and 2.2 log₁₀ units. Very little research has previously examined the consequence of membrane ageing or hazardous events on the LRV performance of full-scale UF systems, and thus the findings presented here will facilitate improvements in risk management, in the design of installations and in the operation of UF membrane systems.

---

Water impact

This study examined the consequence of long-term membrane ageing and a hazardous event on the virus rejection performance of a full-scale ultrafiltration membrane systems used in a wastewater reclamation plant. Understanding virus removal performance under these conditions provided a better understanding of health risks associated with membranes employed in water reuse schemes and will facilitate improvements in risk management.

1. Introduction

Ultrafiltration (UF) membrane technology is widely used in water recycling schemes as a physical barrier for the removal of human pathogens including bacteria, protozoa and viruses. In Australia, implementation of treatment barriers such as UF requires validation to demonstrate that the process can provide the effective LRV of pathogenic microorganisms.¹ Information on the different validation methods available for low pressure membranes including direct and indirect techniques has been widely documented.^2,3 Among these methods, the pressure decay test (PDT) is one of the most frequently used for validating bacterial and protozoa removals, however the sensitivity of this approach does not guarantee virus log removal performance.² Microbial challenge testing using a human enteric virus surrogate is a method commonly used, and is generally performed simultaneously with the PDT under normal operating conditions.^4,5 Surrogates typically include laboratory-cultured bacteriophages such as MS2, which can be cultivated and dosed into the UF feed at very high numbers to demonstrate high magnitude log removal.

Validation is often performed on pilot plants by manufacturers or during or soon after the commissioning of newly installed membranes at full-scale in order to verify the log removal credits claimed by manufacturers. However, the LRV performance of a UF membrane can alter as membranes age due to pore expansion, detached fibres, pinhole puncture, abraded membrane surface, and compromised seal or o-ring.⁶ Understanding the impact of membrane ageing on pathogen LRV is important to manage pathogen risks and to inform asset management and develop accurate replacement strategies. Membrane autopsies have been used to characterise the membrane surface properties of aged membranes,^7,8 which cover the degree of foulant deposition and characteristics, changes in pore size and hydrophobic/hydrophilic properties and tensile strength;⁷ however data characterising the impact of these ageing effects on virus LRV is limited. To date, most of the published data has been collected from pilot and laboratory scale investigations and very little research has examined the consequence of membrane ageing on virus LRV within a full-scale system after long-term operation.

In addition to membrane ageing, the UF process, like many other critical water treatment barriers are subjected to deviations in feed water quality and operational conditions or failure of upstream processes or auxiliary equipment or a maintenance activity, that can adversely affect treatment performance and these are commonly termed ‘hazardous events’.^9,10 An example of a hazardous event experienced by UF membranes includes a sudden change in feed water quality, such as a high turbidity spike, which can result in the deposition of materials on the membrane surface. Although feed water turbidity can improve virus removal due to pore blockage and cake formation,¹¹ extremely high turbidity events can cause membrane damage from the deposition and trapping of materials on the membrane surface which causes scouring that has potential to impact membrane integrity via abrasion. Although these events are usually short-lived, characterisation of these events in terms of their consequence on pathogen LRV is limited yet crucial for understanding the associated health risks in the event of human exposure.¹⁰ Consequently, the assessment of these events has become an integral component of managing human health risks associated with the risk management of wastewater recycling schemes. However, to date, reports characterising the impact hazardous events on pathogen removal by UF treatment are scarce. In particular, data collected from full-scale units is limited.

Knowledge of aged membrane performance and impacts of hazardous events will facilitate improvements for the risk management of UF membranes employed in water recycling schemes and provide options for the design of installations and the need to be vigilant in operations. Therefore, the aim of this study was to examine the consequence of membrane ageing and a hazardous event on the LRV performance of a full-scale UF system. Challenge testing was performed using MS2 to assess and compare virus LRV performance on membranes that were new, aged, and those that were exposed to a hazardous event.

2. Material and methods

2.1 Reuse plant description

A full-scale validation exercise was conducted on a wastewater recycling plant (WRP) in Southern Australia, Australia. The WRP receives chlorinated secondary effluent which is screened (200–500 μm), and stored in two covered feedwater storage basins (each has a volume of 4 ML). Any treated effluent that is not reused is discharged to sea. The automatic control system pumps the effluent from the feedwater basins to the UF membrane system pre-chlorinating (residual of 0.5 mg L⁻¹) in transit to prevent biological fouling. The WRP operates a pressurised modular system comprised of eight UF units with each containing 120 modules. The membranes are hollow fibre polyvinylidene difluoride (PVDF) membranes with a nominal pore size of 0.04 μm, with an asymmetric structure. All membranes function with outside to inside operation and have a design flux of 52.1 L m⁻² h⁻¹. Post UF, the filtrate is disinfected by UV and chlorination for further protozoa and virus removal, where it is then stored in two recycled water storage basins. The WRP has a maximum design flow of about 36 ML d⁻¹. The primary use of the recycled water is for unrestricted municipal irrigation, industrial applications, and dual reticulation schemes. The WWTP and WRP plants were validated to achieve a combined LRV credit of 6.5, 5.0 and 5.0 log₁₀ for viruses, protozoa and bacteria respectively. The contribution provided by the new UF membranes for viruses was 2.5 log₁₀, subject to compliance with critical control parameters which included pressure decay rate (PDR) (<4.8 kPa min⁻¹) and effluent turbidity (daily average <0.15 NTU or not >0.3 NTU for 30 continuous minutes).

Although the membranes were six years old, like most reuse schemes, the units were operated intermittently, which was based on seasonal demand. Therefore to allow comparison between units, data on the cumulative UF filtration run time is provided in Table 1. Membrane units were rotated in duty to minimise storage time and were stored per the manufacturer's recommendations. For short-term storage of <7 days, membranes where soaked following a chlorine clean-in-place (CIP) using a fresh chlorine solution of 200 ppm. For long term storage of >7 days membranes were soaked in a chlorine solution of 5.0 ppm, with a chlorine contact time limit of 100 000 ppm h.

Table 1 Summary of operational conditions and water quality for each validation exercise

Parameter	New membrane	Aged membrane	Aged membrane + hazardous event
UF units	2, 4, 5, 6	2, 6	4, 5, 8
Cumulative UF filtration run time in days	16	196 (2), 188 (6)	124 (4), 227 (5), 144 (8)
System flux (L m^−2 h^−1)	52	41 (2), 43 (6)	43 (4), 42 (5), 47 (8)
Sampling days	3	1
Filtration run time setpoint (min)	30	20
System test flow rate (L s^−1)	120	100
Number of duty unit	1	1
MS2 stock dosing rate (L min^−1)	0.7	0.5
MS2 sampling times (min)	1, 12, 20, 28	1, 7, 14, 20
MS2 batching tank volume (L)	800	140
MS2 feed dose (log10 PFU per mL)	5.0 ± 0.4	5.0 ± 0.1
Feed turbidity (NTU)	1.5	0.8
Feed pH	6.8–7.2	6.8
Feed TOC (mg L^−1)	12.2	8.3
Feed temperature (°C)	15–16	16
Feed chlorine (free and total) (mg L^−1)	<0.1	<0.1
PDR (kPa min^−1) – pre validation	1.2	2.7
PDR (kPa min^−1) – post validation	1.3	2.7
Filtrate turbidity (NTU)	0.05–0.07	0.05–0.06

---

2.2 UF membrane preparation and operation

The validation of the new UF membrane installation was performed on units 2, 4, 5, and 6 and repeated over three separate days in 2010 (Fig. 1). Re-validation of the aged membrane was performed six years later over a single day in 2015, which included units 2, 4, 5, 6 and 8. The aged membranes from units 4, 5 and 8 were also exposed to the hazardous event, which is described below. The units chosen were representative of the range of filtration hours and volume of the system i.e. the maximum (unit 5) and minimum (unit 4) were selected (Table 1). Challenge testing was performed after the membranes were chemically cleaned to ensure all reversible fouling was removed to represent conservative conditions in relation to MS2 rejection. Membrane cleaning involved a chlorine CIP wash involving sodium hypochlorite (200 mg L⁻¹ w/v) or sulphuric acid (0.1% w/v) and citric acid (0.5% w/v). All hydraulic parameters confirmed that the CIP of all units was successful as indicated by an increase in unit flow and decreases in transmembrane pressure (TMP) and resistance (R) post cleaning.

Fig. 1 Schematic of the UF membrane challenge testing configuration showing the validated units coloured in grey.

PDT was carried out for all units at the beginning and the end of the test day. All testing had the minimum required test pressure of 100 kPa for 3 μm particles.¹² PDTs involved applying greater than 100 kPa of air pressure on the filtrate side, and then measuring the pressure decay rate (PDR, kPa min⁻¹) on the filtrate side over a 2 minute period. All PDRs were below the target limit of 3.5 kPa min⁻¹ for the plant.

During validation of the new membranes, the feedwater was pumped at a steady flow rate of 120 L s⁻¹ (60 L s⁻¹ per unit) over a filtration run time of 30 minutes providing a flux of 52 L m⁻² h⁻¹. When validation was repeated on the aged membranes, the feed flow rate was reduced to 100 L s⁻¹ (50 L s⁻¹ per unit) with a flux of 41–47 L m⁻² h⁻¹, which provided a filtration run time of at least 20 minutes. This was to ensure adequate time for sampling given the run times of the aged units had decreased as a result of increased automatic backwash frequency (which was controlled by TMP).

2.3 MS2 stock cultivation and quantification

MS2 was selected as the virus surrogate. Highly concentrated stock culture of MS2 (10⁹ PFU per mL) was grown in suspension for the validation. Briefly, E. coli (ATCC700891) with an approximate optical density at 600 nm of 0.6–0.7 was cultured overnight in tryptone soy broth (TSB, Oxoid) with MS2 (10⁵ PFU per mL). The culture was centrifuged (20 min, 3750 rpm) and the supernatant filtered (0.2 μm) and frozen at −80 °C until use. The frozen MS2 stocks were added to the batching tank which contained chlorine free feedwater. Samples were taken from the batching tank during validation to monitor survival of the stock. Trip controls were also stored at the site for the duration of the testing day and returned for analysis.

MS2 was quantified by the plaque assay method (soft agar double layer).¹³ TSB (Oxoid) was supplemented with ampicillin and streptomycin (150 μg mL⁻¹) was inoculated with an antibiotic resistant (streptomycin/ampicillin) strain of F⁺E. coli (ATCC700891) and incubated at 37 °C on an orbital shaker at 125 rpm until log-phase was reached. Once the bacterial cells reached log-phase, soft agar over layers (3 mL, TSB with 0.7% bacteriological agar (Oxoid)) were augmented with log-phase host E. coli (100 μL) and sample (0.1 mL new and 1 mL aged ) prior to pouring over a tryptone soy agar (TSA) plate containing ampicillin (150 μg mL⁻¹). MS2 (ATCC 15597-BI) were serially diluted in TSB to 100 PFU per mL and used as controls. The inoculated plates were incubated at 37 °C overnight and counts were expressed as plaque forming units per 100 mL (PFU/100 mL). All analysis was performed in duplicate.

2.4 Feedwater preparation

Prior to commencing the challenge testing, the feedwater basin (Fig. 1) was de-sludged to minimise suspension of suspended solids, filled and then isolated to allow the chlorine to decay from the chlorinated secondary effluent process. Prior to and during the challenge testing, the pre-UF chlorination system was also isolated. Low concentrations can inactivate MS2 and therefore chlorine concentrations (both free and total) were closely monitored during the validation exercises to ensure that concentrations were below the limit of detection. Feedwater characteristics for each challenge test are summarised in Table 1.

2.5 MS2 dosing unit set-up

A schematic representation of the MS2 dosing set-up is shown in Fig. 1. The MS2 dosing unit was comprised of a batching tank filled with chlorine-free feedwater, a variable speed mixer to ensure a homogenous stock solution, and a peristaltic dosing pump (Verdex Dura, Model D10). The MS2 stock was continuously dosed at flow rates of 0.5–0.7 L s⁻¹ into the suction side of a single duty UF feedwater pump in order to achieve a target membrane challenge dose of 10⁵ PFU per mL. The UF feedwater pump was operated at 100–120 L s⁻¹. Dosing commenced 10–30 minutes prior to the challenge tests, with a duty membrane unit operational, to ensure that the inlet pipe manifold was primed with an even concentration of MS2.

2.6 Sampling

MS2 grab samples were collected at UF unit feed and filtrate sampling taps. The taps were continuously running for the entirety of the challenge. Sampling was performed at four equidistant points in a single filtration cycle in order to account for the volumetric concentration factor that may occur in between start up and draw down of the membrane units. For the 30 minute filter run (new membranes) MS2 sampling was conducted at time intervals of 1, 12, 20 and 28 minutes, whereas for the 20 minute filter run time (aged membranes) the sampling time intervals were 1, 7, 14, and 20 minutes. Total organic carbon, dissolved organic carbon, and UV_254nm were analysed by National Association of Testing Authorities (NATA) certified laboratory (AWQC, South Australia). All samples were placed on ice and analysed on the same day of collection.

2.7 Online instrumentation description

Online instrumentation monitored feed and filtrate water quality parameters which included pH, temperature, free and total chlorine, conductivity and turbidity. Key online UF performance parameters included TMP, R, and flow rate.

2.8 Hazardous event description

Four and half years after UF installation, a sudden change in feedwater turbidity occurred during operation in the beginning of filtration, where the turbidity exceeded the limit of the online analyser detection of 100 NTU. This was a result of the inlet manifold retaining a large amount of sediment that was left over from a prior feedwater basin desludging exercise, which utilised the same feedwater pump and inlet manifold (Fig. 1).

Out of the eight UF units, only three were impacted by the turbidity spike including units 4, 5 and 8. The sudden turbidity spike resulted in rapid increases in TMP which reduced filtration run times to 3–8 minutes before the units were forced into backwash. A dramatic backwash flow frequency followed to dislodge particles accumulated during the filtration process. Despite chemical cleaning procedures carried out on the affected membranes, sonic testing and weighing of the UF membranes identified units 4 and 8 as having potential integrity breaches due to the accumulation of mud (data not shown). Subsequently, a module autopsy was conducted by selecting a ‘worse case’ module from each of the exposed units (4 and 8) to assess the impacts of this hazardous event on the UF membranes to help inform an asset replacement strategy. The autopsy included an assessment of quality (productivity, selectivity, anti-fouling propensity) and durability methods (mechanical robustness, chemical properties, morphological properties, fouling and clogging). This confirmed that the major foulant composition was mud from the high turbidity event and that the membrane surface displayed signs of abrasion which was likely to be attributed to the fibres moving within the mud (data not shown). Whilst these characteristics are highly likely to be a result of the hazardous event other characteristics such as low embrittlement and calcium and silicon fouling were seen which are characteristic of aged membranes.^7,8 As the viral LRV performance of the aged and impacted units was unclear, revalidation by challenge testing the same four units (2, 4, 5 and 6) including unit 8 was conducted to understand the impact of membrane ageing and the impacts of a hazardous event on membrane integrity and virus removal.

2.9 Data analysis

The average LRV was calculated for each sample time point from paired feed and filtrate samples which were analysed in triplicate for MS2. The Blom formula (eqn (1)) was used to calculate LRV percentiles and assign plotting positions needed to generate frequency distribution plots to show the data distribution. Here the plotting position or percentile (P_i), is a function of the rank i and sample size n. Data was arranged in this format because it describes parameter variability.

3. Results and discussion

3.1 Membrane operational performance

Table 2 compares the typical performance of the new, aged, and aged + hazardous event exposed membranes during normal operation which is based on parameters such as filtrate turbidity, flux, run time, PDR, TMP and R. During the six year operational period, the UF membrane performance in terms of filtrate turbidity remained unchanged regardless of the state of the membranes. However there was a clear impact of ageing on operational parameters such as filtration run time, flux, TMP, R and PDR (Table 2). The TMP, R and PDR had increased over time in all UF units and flux and the run time between backwashes decreased, which was expected of the membranes as they age. While the hazardous event had immediate impacts on hydraulic throughput and UF performance, the long-term performance of these units in terms of filtration run time, flux, TMP, PDR were comparable to the aged only membranes which suggests that the large turbidity spike may not have had lasting adverse impacts, beyond the normal signs of ageing. However the ability of TMP to recover after automated cleaning was somewhat reduced in hazardous event exposed membrane units 4 and 5, showing a 12% recovery post cleaning compared with 20–22% seen with the aged only membrane.

Table 2 Indicative operational performance of the membranes

Parameter	New membrane	Aged membrane	Aged membrane + hazardous event
a .
UF units	2, 4, 5, 6	2, 6	4, 5, 8
Filtration run time (min)	30	15	13
Average TMP (kPa)	47	69	65
Average resistance^a (× 10^12 m^−1)	3.2	6.0	5.6
PDR (kPa min^−1)	1.2	3.0	2.4
TMP before/after clean (kPa)	45/31	70/55 (2), 62/47 (6)	60/51 (4), 66/58 (5), 67/45 (8)
Resistance before/after clean (× 10^12 m^−1)	3.0/2.2	6.5/4.9 (2) 5.0/3.9 (6)	5.2/4.0 (4), 5.8/4.7 (5), 4.8/3.5 (8)
Filtrate turbidity (NTU)	0.05–0.07	0.060–0.062	0.060–0.064

---

3.2 Validation of the new membranes

Full-scale challenge testing using the viral surrogate MS2 was conducted on four newly commissioned UF units over three separate days. The water quality parameters for the validation experiments are listed in Table 1. The mean LRV performance of the individual units is compared in Fig. 2A–D which also show the LRV performance measured at different sample time intervals (1, 12, 20 and 28 minutes). Fig. 3 presents the data using a frequency distribution plot to show the degree of inherent variability in the pooled LRV data from all UF units, and from all three repeated validation exercises. The 5th and 95th percentiles ranged between 2.3 and 3.9 log₁₀ units with a mean LRV of 3.0 log₁₀, which is consistent with published LRV data determined from bench-scale^4,6,14 and pilot-scale investigations,¹⁵ using viruses or surrogates. Throughout each unit validation, R was monitored to assess membrane fouling and indicated that unit R value increased over the 30 minute test period for all units (still considered to be within normal operating conditions). Assessment of the LRV against unit R indicated a positive linear relationship between resistance and MS2 removal (Fig. 4). High R values indicates increased fouling which has also been shown to increase virus removal due to the reduction in pore size caused by the fouling layer.⁹ High LRVs can also be an artefact arising from variations (increases) in the feed challenge concentration,¹⁶ however the feed MS2 dose was consistent, measuring 5.0 ± 0.4 log₁₀ PFU per mL, in accordance with the US EPA Membrane Filtration Guidance Manual.¹² The different feedwater turbidity and unit R values which varied between each of the three validation days may account for the variability in LRV performance of the new membranes. As an indirect measure of potential pathogen breakthrough, filtrate turbidity was monitored online as an indicator of stable plant operation. The low filtrate turbidity and PDRs indicated that the challenged membranes were intact during operation (Table 1).

Fig. 2 MS2 LRV performance comparison of the new membranes (A–D), aged membranes (A and D), and aged + hazardous event exposed membranes (B, C and E). LRV data from the new membranes represent the mean ± 1 standard deviation of the three separate validation trials.

Fig. 3 Frequency distribution plot showing pooled data from all test units, comparing new (●) and aged (○) membrane LRV performance.

Fig. 4 Relationship between MS2 LRV and resistance for new (close symbols) and aged (open symbols) membranes.

3.3 Re-validation of aged and hazardous event impacted membranes

The LRVs measured from units 2 and 6 are shown in Fig. 2A and D which represent the performance of the ‘aged membranes’ that did not encounter the hazardous event. The LRV performance measured at different sample time intervals during the unit run time (1, 7, 14, and 20 minutes) are also displayed. When comparing the individual unit's performance between the new and aged UF membranes, both units showed a 1.0 and 0.7 log₁₀ decrease in LRV performance over the 6 year period. For aged membrane units 4, 5 and 8 that were exposed to the hazardous event, LRV performance had decreased on average by 1.1 log₁₀ units, which was comparable to the aged only membranes (Fig. 2). The largest decrease in LRV performance was seen with unit 4, which on average decreased by 1.5 log₁₀ units, although this may have been an artefact of this unit achieving some of the highest LRVs when new. Interestingly, unit 4 had operated the least over the 6 year period, recording a total operational duration of 124 days (Table 1). Furthermore, the total run time of each unit (which varied between 124–227 days) appeared to have no relationship with LRV performance.

Frequency distribution plots of the pooled LRV data from all membrane units showed that that the median LRV measured from the aged membranes was 2.0 log₁₀, which was 1.0 log₁₀ units lower than new membranes (Fig. 3). There was however less variability in LRV performance when compared to the LRV data from the new membranes, where the 5th and 95th percentile values ranged between only 1.7 and 2.2 log₁₀ units. The stable LRV performance may have been attributed to the development of irreversible fouling layers which develop as the membranes age. The decline in the ability of TMP and R to recover after membrane cleaning was indicative of irreversible fouling (Table 2).

The stable LRV performance may also be reflective of the lack of change in unit R conditions, which unlike the new membranes, did not substantially increase over the 20 minute filtration run time (Fig. 4). An increase in unit R (and TMP) has been shown to be correlated with an increase in the reversible fouling layers and hence improvement in LRV;⁹ however the relationship between LRV and R on the aged membranes was less pronounced than observed with the new membranes (Fig. 4). This may have been attributed to the reduction in filtration run times of the aged membranes, but also because during the validation exercise, the feed turbidity and TOC concentrations were low which reduced the potential for reversible membrane fouling to develop during the run time (Table 1).

The LRV performance of the revalidated membranes showed the aged units exposed to the hazardous event were comparable to the aged only membranes, which reflects the similarities in the membranes operational performance presented in Table 2. The LRV performance data therefore complements the operational performance data to suggest that that the hazardous event had no greater long-term impact on membrane performance, beyond the impacts seen by normal ageing. Currently, there is very limited published data which details the impacts of membrane ageing on virus removal in UF membranes, where most of the research has been focused on assessing impacts of membrane fouling, chemical cleaning and performance factors (resistance and breach frequencies).^17,18 In Victoria, Australia, the health authorities require annual virus testing of membrane installations for high risk exposure schemes. Furthermore, a novel aspect of this study was the validation assessment was conducted on a full-scale WRP, under real operational conditions. This study also confirms the need to have multiple barriers for each pathogen. Chlorine disinfection is invaluable for virus removal and chlorine systems must be designed with sufficient robustness that they can be strengthened in case other barriers such as membranes experience a hazardous event or they age, or there are cost/benefits to reducing the membrane LRV accreditation, and delaying membrane replacement, offset with an increase in disinfection.

4. Conclusion

This study has assessed the impacts of membrane ageing and a hazardous event on virus removal performance of full-scale UF membrane system. The results showed there was more variability in LRV performance of the new membranes, probably as a result of reversible fouling. In comparison, the LRV performance of the aged membranes was more stable, which was indicative of irreversible fouling and lower turbidity/reversible fouling. The mean LRV of aged membranes had reduced by 1.0 log₁₀ and there was no difference between aged and those that were exposed to the hazardous event. These findings will facilitate improvements for the risk management of UF membranes employed in wastewater recycling schemes.

Acknowledgements

The authors would like to thank Stella Fanok and Dusty Rietveld and the Allwater wastewater treatment plant operators for their assistance in 2015 and the United Water/Veolia Water operations team including Craig Heidenreich in 2010 and Alexandra Keegan from SA Water. The authors would also like to acknowledge Suez for the membrane autopsy.

References

1. NRMMC-EPHC-AHMC, Australian Guidelines for Water Recycling: Managing Health and Environmental Risks (Phase 1), Natural Resource Ministerial Management Council, Environmental Protection and Heritage Council, Australian Health Ministers' Conference, Canberra, 2006.
2. H. Guo, Y. Wyart, J. Perot, F. Nauleau and P. Moulin, Low-pressure membrane integrity tests for drinking water treatment: A review, Water Res., 2010, 44, 41–57.
3. K. Farahbakhsh, S. S. Adham and D. W. Smith, Monitoring the integrity of low pressure membranes, J. Am. Water Resour. Assoc., 2003, 95, 95–107.
4. H. Huang, T. A. Young, K. J. Schwab and J. G. Jacangelo, Mechanisms of virus removal from secondary wastewater effluent by low pressure membrane filtration, J. Membr. Sci., 2012, 409–410, 1–8.
5. A. M. ElHadidy, S. Peldszus and M. I. Van Dyke, Effect of hydraulically reversible and hydraulically irreversible fouling on the removal of MS2 and phiX174 bacteriophage by an ultrafiltration membrane, Water Res., 2014, 61, 297–307.
6. A. Antony, J. Blackbeard, M. Angles and G. Leslie, Non-microbial indicators for monitoring virus removal by ultrafiltration membranes, J. Membr. Sci., 2014, 454, 193–199.
7. S. Phuntsho, A. Listowski, H. K. Shon, P. Le-Clech and S. Vigneswaran, Membrane autopsy of a 10 year old hollow fibre membrane from Sydney Olympic Park water reclamation plant, Desalination, 2011, 271, 241–247.
8. L. D. Nghiem and A. I. Schafer, Fouling autopsy of hollow-fibre MF membranes in wastewater reclamination, Desalination, 2006, 188, 113–121.
9. T. Trinh, A. Branch, B. van den Akker, P. Le-Clech, J. E. Drewes and S. J. Khan, Impacts of hazardous events on performance of membrane bioreactors, in Membrane biological reactors: theory, modeling, design, management and applications to wastewater reuse, ed. F. Hai, K. Yamamoto and C.-H. Lee, IWA publishing, 2014.
10. B. van den Akker, T. Trinh, H. M. Coleman, R. M. Stuetz, P. Le-Clech and S. J. Khan, Validation of a full-scale membrane bioreactor and the impact of membrane cleaning on the removal of microbial indicators, Bioresour. Technol., 2014, 155, 432–437.
11. Z. Yin, V. Tarabara and I. Xagoraraki, Human adenovirus removal by hollow fiber membranes: Effect of membrane fouling by suspended and dissolved matter, J. Membr. Sci., 2015, 482, 120–127.
12. US Environmental Protection Agency, Membrane Filtration Guidance Manual, 2005.
13. US Environmental Protection Agency, Method 1601: Male-specific (F+) and somatic coliphage in water by two step enrichment procedure, EPA number 821-R-01-030, Washington, DC, 2001.
14. A. M. ElHadidy, S. Peldszus and M. I. Van Dyke, An evaluation of virus removal mechanisms by ultrafiltration membranes using MS2 and φX174 bacteriophage, Sep. Purif. Technol., 2013, 120, 215–223.
15. O. Ferrer, S. Casas, C. Galvañ, F. Lucena, A. Bosch, B. Galofré, J. Mesa, J. Jofre and X. Bernat, Direct ultrafiltration performance and membrane integrity monitoring by microbiological analysis, Water Res., 2015, 83, 121–131.
16. A. Antony, J. Blackbeard and G. Leslie, Removal Efficiency and Integrity Monitoring Techniques for Virus Removal by Membrane Processes, Crit. Rev. Environ. Sci. Technol., 2012, 42, 891–933.
17. C. Regula, E. Carretier, Y. Wyart, G. Gésan-Guiziou, A. Vincent, D. Boudot and P. Moulin, Chemical cleaning/disinfection and ageing of organic UF membranes: A review, Water Res., 2014, 56, 325–365.
18. S. Robinson, S. Z. Abdullah, P. Bérubé and P. Le-Clech, Ageing of membranes for water treatment: Linking changes to performance, J. Membr. Sci., 2016, 503, 177–187.

---