Multiple benefits of MBR and absolute-rated cartridge filters for RO pretreatment: Results from a full-scale facility

Increasing demands on water supplies combined with drought conditions and urbanization have led many water agencies to consider direct or indirect potable reuse (DPR or IPR) (NRC 2012; US EPA 2019). In California, IPR regulations require reverse osmosis (RO) and advanced oxidation, e.g., an ultraviolet light/advanced oxidation process (UV/AOP), to be part of the treatment train (Gerrity et al. 2013; Pecson et al. 2018a). These two treatment processes, plus the travel time for groundwater recharge or dilution requirements of reservoir water augmentation, meet the strict regulatory requirements for the removal of pathogens, salts, and trace organic contaminants (Pecson et al. 2018a, b).

RO for potable reuse applications requires pretreatment to reduce the impact of fouling and scaling (Jiang et al. 2017; Hoek et al. 2022; Liu et al. 2024). Fouling simultaneously reduces the productivity and performance of RO and increases the costs due to energy and maintenance (Jiang et al. 2017; Hoek et al. 2022; Nthunya et al. 2022). The causes of fouling and scaling can differ and may include inorganic scaling, biofilm growth, colloidal fouling, organic fouling, or a combination of inorganic and biological fouling (Li et al. 2007; Kimura et al. 2016; Hoek et al. 2022). The typical approach to RO pretreatment for potable reuse applications has been micro- or ultrafiltration (MF/UF) of secondary or tertiary wastewater (Gerrity et al. 2013; Tang et al. 2018). However, this approach of simple MF/UF-RO is not suitable for some applications, e.g., non-nitrifying wastewater facilities where additional nitrogen management is required; or decentralized potable reuse via ‘scalping plants’ where biological wastewater treatment must be part of the advanced treatment facility. In addition to providing a barrier against fouling, pretreatment can also provide additional pathogen removal credit (Adelman et al. 2024). MF/UF-based pretreatment trains already have well-established pathogen credits, but the use of an alternative train needs to achieve comparable levels of pathogen removal.

Interest is therefore growing in alternative pre-RO treatment processes for certain potable reuse applications. For example, ozone combined with biological activated carbon (ozone–BAC) has been the subject of recent research showing the potential for realizing multiple benefits from alternative pre-RO treatment processes (Trussell & Pisarenko 2024), and further such research on other novel process trains would be beneficial. Membrane bioreactors (MBRs) have recently been explored as direct pretreatment options for RO. The MBR process includes a suspended growth bioreactor followed by solids separation using membranes, with opening size generally in the MF/UF range but operated under high-solid conditions with a surface cake layer. MBRs are advantageous over conventional wastewater treatment due to their reduced footprint, increased solids inventory and cell residence time, greater ability to nitrify/denitrify, and improved solids separation (Comerton et al. 2005; Le-Clech 2010; Zanetti et al. 2010). MBRs are of particular interest in instances where an existing wastewater facility must be retrofitted to manage nitrogen, or where a new biological process must be implemented in a decentralized facility. The membrane separation step in the MBR process removes suspended solids and pathogens by a combination of size exclusion and capture by the cake layer (Chaudhry et al. 2015), so MBR is theoretically at least as effective of a pathogen barrier as MF/UF. If MBR is to be used in a potable reuse system, it would be desirable to use direct MBR-RO treatment and avoid the cost of including additional MF/UF downstream. Currently, there is conflicting evidence on the impact of pretreatment by MBR on RO fouling. Qin et al. (2006) found that MBR-RO was more effective than MF/UF-RO in a pilot-scale system. However, more recent experience has shown that MF/UF pretreatment tends to be more effective in practice, while MBR-RO plants appear to be prone to higher rates of fouling and shorter clean-in-place (CIP) intervals than conventional MF/UF trains (WRF 2023). Additional data are required to better understand the impact of MBR on RO fouling.

While research has primarily focused on the efficacy of various other pretreatment processes, RO systems for potable reuse also include cartridge filters (CFs). CFs are most commonly added to satisfy warranty requirements for the RO membranes, but few studies have explicitly characterized their impacts on RO performance and pathogen removal. CFs used for potable reuse RO systems typically have a nominal pore size rating of around 5 μm, and they are typically monitored only by periodic grab samples for silt density index which likely produces data of insufficient frequency to optimize fouling control and pathogen removal. Hoek et al. (2022) evaluated the performance of typical 5 μm nominally-rated CFs at a brackish groundwater RO plant and found that they increased RO fouling due to the growth and sloughing of biofilms. There is little literature overall on CFs as a unit process in a pre-RO train, and the literature does not elucidate their capability for removal efficiency and fouling control.

In this work, the impact of both MBR and absolute-rated 1 μm CFs was evaluated on the fouling of RO and the achievement of pathogen log reduction value (LRV) credit. This site, located in California, had a novel treatment train configuration consisting of MBR, CF, RO, UV/AOP, and chlorine disinfection followed by post-stabilization. The objective of this study was to evaluate pathogen barrier monitoring and RO fouling control by an MBR process along with an absolute-rated 1 μm CF system at a full-scale advanced treatment plant. If this novel pre-RO alternative treatment train can demonstrate multiple benefits in these areas, there will be many other potential applications for it in advanced treatment facilities.

This study is focused on the Sustainable Water Infrastructure Project (SWIP) implemented by the City of Santa Monica. The SWIP advanced water treatment facility (AWTF) is a scalping facility that takes approximately 10% of the City's average municipal wastewater supply to produce a maximum of 1.2 million gallons per day (MGD) or 4,540 m³/d of advanced-treated recycled water. The AWTF can accept local stormwater runoff as up to 30% of its influent volume, but during this study period, no stormwater flow was recorded. The advanced-treated recycled water is used for non-potable applications (e.g., irrigation, toilet flushing, cooling towers) by private and public customers, and it will be used for IPR through direct injection of advanced-treated recycled water for groundwater augmentation beginning in spring 2025.

There are two MBR trains, each consisting of an anoxic basin, an aerobic basin, and membrane separation. The MBR is intended to achieve carbonaceous biochemical oxygen demand (BOD) removal, full nitrification with partial denitrification, and solids separation. The biological basins operate with 4,000–9,000 mg/L mixed liquor suspended solids (MLSS); a 2 mg/L dissolved oxygen (DO) target in the aerobic zone; and a cell residence time of 1–2 weeks to provide adequate conditions for nitrifiers. The MBR membranes are Suez ZeeWeed 500d-422 membranes with two 52-module cassettes per train. These membranes are ultrafiltration-rated with 0.04 μm nominal pore size, and they operate with 12 min production cycles at gross flux between 7.8 and 17.6 gfd (13–30 LMH) followed by 0.5 min relaxation cycles to control cake formation. Return activated sludge (RAS) flows back to the anoxic basin, with a recycle ratio of 4–5 times the plant flow. Waste activated sludge (WAS) is wasted to control MLSS in the MBR system.

There are two CF-RO trains as part of the process. CFs are provided upstream of RO to protect the membranes from damage by passively filtering MBR effluent. Although conventional RO systems generally employ CFs with nominal ratings around 5 μm to satisfy warranty requirements, the SWIP AWTF selected finer absolute-rated 1 μm cartridges that sought to achieve pathogen removal credit. These CFs are Harmsco HC/170-LT2 filters made of pleated microfiber media. Supplier validation results based on the US EPA Long-Term 2 Enhanced Surface Water Treatment Rule have shown >3.6 log removal of 2 μm Cryptosporidium surrogate beads up to 30 psi (2.1 bar) pressure drop and up to 100 gpm (379 L/min) flow per 125 ft² (11.6 m²) cartridge, equivalent to 1,152 gfd or 0.8 gal/ft²-min (1,956 LMH) gross flux. Based on this validation testing, these cartridges achieved the US EPA and the National Sanitation Foundation certification for the removal of cyst-sized particles, and they are accepted by DDW for potential pathogen reduction credits. The CF system included two redundant vessels per train operated in a duty/standby configuration, with five cartridges per vessel to remain below the maximum validated flow rate.

The two RO trains, provided by H₂O Innovation, are designed as a three-stage process with a 14:6:3 array of 6 L vessels and an intermediate booster at Stage 2. The membranes are Toray TMG20D-400 polyamide brackish water elements with 400 ft² (37 m²) area each. The RO trains were operated from 9 to 11 gfd (15–19 LMH) overall permeate flux and 78–85% recovery. The RO system removes dissolved inorganic and organic components in the CF filtrate, as well as pathogens. To control scaling and to protect the RO membranes, feed conditioning included antiscalant dosing, pH reduction with sulfuric acid, and chloramine addition.

The demonstration of the performance of the MBR-CF-RO systems over a full range of operating conditions was performed during a continuous 14-day acceptance test per DDW requirements between 28 March and 10 April 2023. AWTF influent wastewater quality during this evaluation period is summarized in Table 1. This acceptance test phase was followed by a continued commissioning operational phase of the plant for the remainder of 2023. The AWTF was fed with 100% wastewater at an average daily influent flow rate of 1.29 MGD (4,880 m³/day), and the wastewater remained similar to the quality shown in Table 1 although the total dissolved solids ranged as high as 1,200 mg/L during some periods. The effluent water was monitored for compliance with standards for advanced-treated recycled water requirement standards and for potable reuse by groundwater augmentation via subsurface application (State Board 2015).

Key critical control points (CCPs) for the MBR-CF-RO process are presented in Table 2, and achievement of LRV credits for pathogen removal was determined by comparison of online data to the CCPs. Each process was intended to achieve LRV credits as shown in Table 2 to meet the minimum 12-log virus, 10-log Cryptosporidium, and 10-log Giardia goals for the AWTF overall. For MBR, filtrate turbidity is the key monitoring parameter for LRV credits, based on the Tier 1 process for MBR validation (WRF 2021). Other performance parameters such as pH, DO, oxidation–reduction potential (ORP), temperature, solids retention time (SRT), hydraulic retention time (HRT), MLSS concentration, transmembrane pressure (TMP), and membrane flux are also utilized as performance indicators for the biological treatment and membrane separation processes. CF monitoring is focused on CCPs for flow rate, filter flux, dP, and filtrate turbidity to stay within the range of the supplier's validation testing.

During the 14-day acceptance test, multiple data sets were collected as shown in Table 3, including continuous monitoring of online process instrumentation, manual data collection through laboratory testing of grab samples, and data reported via the control system. All water quality sampling and performance parameters were sampled, logged, and measured on Days 1, 7, and 14 as a minimum during the 14-day evaluation period. Reported averages were used to determine performance compliance for non-continuous parameters.

This dP is normalized to the temperature-corrected flux squared, reflecting the expected minor loss relationship between headloss across a CF system and the interstitial velocity at the cartridge openings. During the normal operation of a CF system, this normalized dP would be expected to increase in a roughly linear fashion as solids accumulation blocks some of the open area and increases interstitial velocity.

This normalization process accounts for the variation in specific permeate flux over time as a result of changes in water temperature and feed salinity; as well as variation in specific flux between stages as a result of increasing feed-concentrate osmotic pressure along the elements in series. By accounting for these conditions, the normalized specific flux is essentially a measure of the underlying permeability of the membrane elements, and changes in its value over time reflect the progression of fouling and scaling.

For all treatment processes, a two-sample unequal variance t-test was used to determine whether there was a statistically significant difference between Train 1 and Train 2 at the AWTF. The resulting p-values less than 0.05 were considered significant. In general, it was hypothesized that similar operating conditions and similar performance should be observed between the two identical trains, and comparing these two duplicate trains increases the robustness of the findings.

Confirmation of stable online performance and compliance with the CCPs partially satisfied the requirements in Table 2 to achieve LRV credit. To address the final requirement, the cartridge filter online performance was also observed during two intentional breach experiments. The online dP is the directly measured parameter that is expected to respond most quickly to a breach. However, dP across a cartridge filter will also vary as a result of changing flow through the cartridges, changing water temperature and viscosity, and the general progression of solids accumulation as shown in Figure 6. Therefore, to illustrate a potentially more robust way to maintain pathogen monitoring, the calculated parameters of TCSF per Equation (1) and normalized dP per Equation (2) were also plotted.

As expected, the fouling observed in Figure 8 was characteristic of non-mineral fouling. Mineral scaling leads to the rapid decline of normalized flux over the time scale of days, while non-mineral fouling leads to a more gradual decline over the time scale of days to weeks (Crittenden et al. 2023). A gradual decline in normalized flux indicates some form of organic fouling – through some combination of dissolved particles, colloidal particles, and biofilm growth – that apparently affected the feed-concentrate side of both stages. Both the actual permeability and the rates of fouling were similar between the two RO trains as no statistically significant difference was found between the two trains for the normalized flux values at Stage 1 (p = 0.33) or Stage 2 (p = 0.28). Importantly, similar rates of fouling were also observed between Stages 1 and 2 of each train. This is notable because although fouling of multiple stages by dissolved organics is a familiar observation, colloidal fouling is conventionally thought to mainly impact Stage 1 (Hydranautics 2014). The observed fouling was also observed to be reversible, as the normalized specific flux was restored to between 96 and 100% of its initial value in the evaluation period following CIP.

The MBR system performance (Table 4) remained within the required envelope for turbidity and gross flux and achieved 1 LRV credit for viruses and 2.5 LRVs for Giardia and Cryptosporidium. Online CF turbidity, flow, and dP (Table 5) also remained within the validated envelope for the absolute-rated cartridges which allowed the CF system to achieve 2.0 LRV credit for Cryptosporidium and 2.5 LRVs for Giardia. The CF system challenge testing demonstrated that breaches could be detected based on the monitoring of online dP. This AWTF is the first in California to achieve regulatory approval for such credits, and the contribution of both MBR and CFs to the overall LRVs is an important benefit of this pre-RO treatment alternative. This benefit has economic value as well – if MBR and CFs can achieve pathogen LRV credit, then the capital and operating cost of adding MF/UF can be avoided in facilities that select MBR-based biological treatment.

The mitigation of RO fouling in MBR-based treatment systems is also a topic of interest for further optimization and research. The results at this facility suggest that absolute-rated cartridges could be tested at other advanced treatment plants struggling with colloidal fouling, as control of the colloidal particle load in the size range of 1 μm and above led the MBR filtrate in this study to have lower fouling propensity compared with other MBR-RO facilities. Varying cartridge pore size rating and MBR biological process conditions could provide further insight into the relative contributions of the dissolved and colloidal organic particle load that drives RO fouling and elucidate the conditions that best control it.

In this study, an MBR system followed by an absolute-rated 1 μm CF system was demonstrated at a full-scale advanced treatment plant as the pretreatment processes for RO. This advanced treatment plant included two identical treatment trains, which provide additional robustness to the study results. The MBR system achieved effective biological treatment and low filtrate turbidity, and the CFs stayed within their flow, dP, and turbidity CCP limits. Breach testing of the CFs demonstrated a consistent response pattern in flux and dP data. Novel TCSF and normalized dP parameters were shown to provide easier detection of breach conditions using the online dP data, and the breach response patterns were consistent across the CF change-out cycle. These observed breach response patterns and the novel derived parameters are a useful contribution to the online monitoring of CFs as a pathogen barrier.

The combined pretreatment processes of MBR and CFs also achieved remarkably low non-mineral fouling rates at the downstream RO system. These observed fouling rates were lower than reported for another comparable MBR-RO treatment train and on par with non-MBR reuse treatment trains, and the overall normalized flux decline was consistent with 5–7-month CIP intervals. Regularly achieving such CIP intervals with MBR filtrate would be of great benefit for RO operability. The observation of a significant increase in dP over a 2-week time scale showed that the CFs were capturing colloidal particles even from low turbidity influent, and this resulted in a lower-than-expected fouling rate at the downstream RO. This study, therefore, provides a fundamental insight into the nature of the particle load that drives RO fouling when treating MBR filtrate: namely, colloidal particles in the size range of 1 μm and above are an important contributor to fouling propensity when they are not removed by pre-RO treatment.

The findings of this study allowed the Santa Monica SWIP AWTF facility to achieve the first permitted pathogen log removal credit for both MBR and CFs in California by online confirmation of compliance with CCPs alongside the CF challenge testing results. Achieved pathogen LRV credit included 1-log for virus and 2.5-log each for Cryptosporidium and Giardia across MBR, as well as 2-log for Cryptosporidium and 2.5-log for Giardia across the CFs. The results discussed herein will be of interest for utilities in any location considering RO-based treatment trains and how to effectively acquire LRV credits for MBR and CF pretreatment processes. For many potable reuse applications, particularly those where MBR is desirable for nitrogen management or decentralized treatment, the multiple benefits of pathogen removal credit and fouling control make MBR-CF-RO a potentially attractive alternative.

The authors thank the many people at Kiewit, Arcadis, PACE, PERC Water, Vertech, Stantec, and the City of Santa Monica who have been part of the Sustainable Water Infrastructure Project. Special thanks to Tyler Hadacek, Andrew Devries, Yamrot Amha, Anna Philipp, Michael Scott, Jim Borchardt, Breanne Padilla, Eric Gonzales, Gilbert Perez, Mario Meza, and Terrell Thompson.

The data analysis and writing of this paper was funded by the City of Santa Monica and the Stantec Institute for Water Technology and Policy. The Sustainable Water Infrastructure Project was financed by the City of Santa Monica. Funding assistance for this project came from the California State Water Resources Control Board Clean Water State Revolving Fund, along with California Prop 1 Stormwater, Los Angeles County Measure W, and Metropolitan Water District of Southern California Local Resources Program grant funds.

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

The authors declare there is no conflict.