On-site treatment of used wash-water using biologically activated membrane bioreactors operated at different solids retention times

Sanitation in developing countries requires development of low-cost and robust technologies with low maintenance requirements. However the current absence of adapted technical solutions hampers sanitation development. Eawag and the Austrian design studio EOOS developed the ‘blue diversion toilet’ in response to the Reinvent the Toilet Challenge of the Bill & Melinda Gates Foundation (Larsen et al. 2015). The blue diversion toilet operates ‘off the grid’; that is, without connections to piped water, sewer system, etc. It separates urine, feces and used wash-water. Used wash-water is treated on-site for further reuse. Safe reuse of wash-water requires removing the organic substrate and pathogens in addition to removing color and odor to make the water appealing.

Gravity-driven membrane (GDM) ultrafiltration is successfully applied to the decentralized production of high-quality drinking water (Clasen et al. 2009; Peter-Varbanets et al. 2010; Derlon et al. 2012). During GDM ultrafiltration, biofilm formation on the membrane surface is tolerated and allows for long-term (several months) operation of the membrane system without cleaning or maintenance (Peter-Varbanets et al. 2010). Previous studies on GDM ultrafiltration reported stable permeate fluxes of 4–11 L m⁻² h⁻¹ (Peter-Varbanets et al. 2010, 2011). However, an increasing organic substrate concentration in feed water decreases the level of flux stabilization. Also, flux stabilization was observed only during GDM ultrafiltration of surface water, and in some experiments with diluted wastewater (total organic carbon (TOC) < 15.3 mg L⁻¹) (Peter-Varbanets et al. 2010, 2011). In other experiments with diluted wastewater (6.6–23.8 mg TOC L⁻¹) a steadily decreasing flux was observed (Peter-Varbanets et al. 2011). On the other hand, used wash-water contains significantly higher concentrations of organic substrate (300–1,600 mg COD L⁻¹) (Künzle et al. 2015). Thus, development of the blue diversion toilet requires an evaluation of the suitability of GDM ultrafiltration for on-site treatment of used wash-water. The system based on GDM ultrafiltration used for treatment of used wash-water is referred to as the biologically activated membrane bioreactor (BAMBi) (Künzle et al. 2015).

Used wash-water results from anal cleansing, hand washing and toilet cleaning (Künzle et al. 2015). Thus, it contains feces, urine, soap and traces of blood. Particles represent a large fraction of the organic substrate. Degradation of the organic substrate in the particulate form first requires hydrolysis followed by microbial utilization. Thus, if the hydrolysis rate is slower than the inflow of organic particles, particles will accumulate in the reactor. Another factor influencing degradation of organic particles is degradability of the particles. A high solids retention time (SRT) can lead to degradation of non-biodegradable particles and no net accumulation of solids after long-term operation (Laera et al. 2005; Sperandio et al. 2013). It is not clear to what extent organic particles accumulate and if this phenomenon will be detrimental for system operation. Such investigation is required to evaluate how much maintenance and control of the system in terms of solids removal will be required.

The objectives of this study were: (i) to evaluate if BAMBi can be successfully applied to on-site treatment of used wash-water and ultimately represents a suitable technology for development of the blue diversion toilet; (ii) to study the effect of operating conditions in terms of suspended SRT varying from 7 days to full retention, aeration conditions, and organic loads on the system performances in terms of permeate flux and substrate/nutrient conversion; and (iii) to investigate the mechanism of particle degradation in BAMBi. Three BAMBis were operated at different suspended SRTs of 7 and 102 days, and at full retention for several months in two phases. In phase #1 the organic loading rate (OLR) was variable while in phase #2 the load was stable. Solids, organic substrate and nutrient concentrations were monitored as well as permeate flux. To evaluate the composition of sludge formed in BAMBi, microscopy and particle size distribution measurements were performed.

Used wash-water was prepared by mixing tap water with human feces, human urine and Bul Star Washing Soap (Uganda) at concentrations of 5 g L⁻¹ feces, 10 mL L⁻¹ urine and 0.75 g L⁻¹ soap. Fresh human feces and urine were collected at Eawag (Dübendorf, Switzerland). Total and volatile suspended solids (TSS and VSS), chemical oxygen demand (COD), nitrogen and phosphorus concentrations of used wash-water are provided in Table 2. Particulate organic substrate represented 85% of total organic substrate.

VSS, COD (<0.45 μm), NH₄⁺ and NO₃⁻ were measured weekly. Reactor walls were cleaned before sampling, while the membrane surface remained untouched to prevent biofilm detachment. Bulk liquid in the reactor was gently stirred while a subvolume of 3–4 L was withdrawn from the sludge removal valve. The subvolume was mixed using a magnetic stirrer during sampling. Samples for soluble compounds analyses were filtered using 0.45 μm Marcherey-Nagel Nanocolor 50 chromafil GF/PET membrane filters prewashed with deionized water; 50–150 mL sludge samples were filtered with 7–12 μm Marcherey-Nagel 640w 90 mm ashless filters for VSS measurement. Two samples were analyzed for each reactor. COD analysis was performed in duplicates using Hach Lange LCK614 and LCK314 tests. Ammonia was measured with a flow injection analyzer (Foss FIA star 5000 Analyzer, Gerber Instruments). Nitrate was analyzed using ion chromatography (Ionenchromatograph 881 Compact IC, Metrohm).

The total DNA content in 1 mL sludge samples were stained with Syto9 (S-34854, Invitrogen, Switzerland) in a 1:100 dilution. To prevent background fluorescence, brief low-speed centrifugation and phosphate buffered saline solution washing steps were performed. Biomass in the sludge was observed using confocal laser scanning microscopy (CLSM) (Leica SP5, Wetzlar, Germany). Image acquisition was performed by Leica LAS AF software 2.6v. using a 488 nm laser line for the green fluorescence signal (Syto9), and for transmitted light signals.

Particle size distribution was measured using laser diffraction (Mastersizer 2000, Malvern, UK). Laser power was set to run at 78–82 watts. The sample was injected into the measurement chamber at an obscuration rate of 10–20%. Measurements were done in triplicates and average size distributions were determined by scattered light using the Frauenhofer model.

Dissolved COD concentrations were similar despite an increasing solids concentration with increasing suspended SRT (Figure 2). COD concentrations of 61 ± 2 mg COD L⁻¹, 81 ± 3 mg COD L⁻¹ and 61 ± 2 mg COD L⁻¹ were measured in the permeate of high-, medium- and low-SRT reactors, respectively (i.e. COD removal efficiencies of 92.8 ± 0.3%, 88.9 ± 0.7% and 89 ± 1%, respectively). During phase #1-1, low aeration resulted in formation of anoxic zones at high and medium suspended SRT, with measured oxygen concentrations of 0.05–6.1 mg O₂ L⁻¹. At low suspended SRT, oxygen concentrations of 6.2–6.6 mg O₂ L⁻¹ were measured during phase #1-1. To prevent formation of anoxic zones, aeration was increased in phase #1-2. Oxygen concentrations thus increased to above 6.9 mg O₂ L⁻¹. At high suspended SRT, ammonia concentrations were similar to influent concentrations until day 50. After day 50, ammonia concentrations dropped below 1.6 mg N L⁻¹ (Figure 2). Nitrate concentrations remained below 1 mg N L⁻¹ until phase #1-2. In phase #1-2 and phase #2 ammonia concentrations remained stable at below 1 mg N L⁻¹, while nitrate concentrations increased to 6–88 mg N L⁻¹. In medium- and low-SRT reactors, ammonia concentrations dropped below 1 mg N L⁻¹ when phase #1-2 started, while nitrate concentrations did not increase immediately.

Our results indicate that on-site treatment of used wash-water can be successfully achieved using BAMBi. A key question was whether flux stabilization could be observed despite the high strength of used wash-water. We demonstrated that flux stabilization was achieved over several months without cleaning the system (Figure 3). An earlier study on BAMBi pointed to the need for more research on optimum sludge removal protocol (Künzle et al. 2015). We found that flux stabilization was independent of solids retention. Thus there is no need to control sludge concentrations in the system to maintain a stable production of recycled water. During operation at medium and high suspended SRT, flux stabilization was observed between 0.5 and 1 L h⁻¹ m⁻² prior to biofilm sloughing. Biofilm sloughing increased the flux to about 1.5 L h⁻¹ m⁻² in both reactors. These levels of flux stabilization observed with used wash-water are lower than those reported for filtration of surface water and diluted wastewater (Peter-Varbanets et al. 2010, 2011). However flux stabilization at 1–2 L h⁻¹ m⁻² was recently reported during treatment of greywater using a novel fixed fiber biofilm membrane reactor operated without fouling control (Jabornig & Podmirseg 2015). This flux is comparable to those measured in the current study. Earlier studies show that higher influent TOC concentrations reduce the level of flux stabilization (Peter-Varbanets et al. 2010; Derlon et al. 2013). However predation can lead to higher permeate flux for water with high TOC content, and worms can enhance particulate organic carbon (POC) removal in GDM filtration (Derlon et al. 2013). Our results indicate that changing the influent TOC concentration through variation in the OLR after flux stabilized did not affect the flux level. Microscopy of suspended solids in the system also showed high worm activity at one point after flux had stabilized; however later no worms were detected (data not shown).

Permeate COD concentrations were also independent of sludge concentrations maintained in BAMBi (Figure 2). Thus BAMBi can be operated at a stable flux with a stable content of organic substrate in the permeate without regular sludge removal and membrane cleaning. Used wash-water could be treated with BAMBi despite high concentrations of particles. There is no need for pre-treatment to remove solids in used wash-water in the blue diversion toilet. Independent of solids retention time, COD removal efficiencies were from 89 to 93%. This is comparable to COD removal efficiencies in membrane bioreactors (MBRs) treating wastewater (van der Roest et al. 2002). Used wash-water treated with BAMBi had higher COD concentrations than greywater treated in two different types of biofilm membrane reactors (Jabornig & Favero 2013; Jabornig & Podmirseg 2015). However lower permeate COD concentrations were measured in BAMBi despite higher load of organic materials applied for this system in comparison with the systems treating greywater.

Simultaneous nitrification and denitrification occurred when the system developed anoxic zones (Figure 2). Reduced aeration led to incomplete mixing and formation of anoxic zones at the reactor bottom. The combination of anoxic zones and aerobic zones resulted in simultaneous nitrification and denitrification, and maximum nitrogen removal efficiencies of 97%, 87% and 86% were measured in high-, medium- and low-SRT reactors, respectively. Increased aeration in phase #1-2 limited formation of anoxic zones, and nitrate concentrations increased rapidly. In the low-SRT reactor, anoxic zones were not detected. Solids concentrations measured in the low-SRT reactor were low compared with medium- and high-SRT reactors. Thus, the development of anoxic zones in BAMBi required solids accumulation in the system. The same trend with combined nitrification and denitrification was observed also at low SRT, indicating that anoxic zones could have existed in the biofilm.

Particles smaller than 10 μm were abundant in the influent, but nearly absent in BAMBis (Figure 4). Thus, these particles either degraded, accumulated in the biofilm or flocculated. Our observations indicate that solids from the bulk liquid consisted of influent particles colonized by bacteria (Figure 5). Microscopic observations indicated that microorganisms were growing on the surface and not in the core of particles (Figure 5). Thus flocculation of the smaller particles seems unlikely. Bacteria colonizing the surface area of particles should lead to faster degradation of smaller particles because of a larger relative surface area to volume ratio compared with large particles (Dimock & Morgenroth 2006). A batch experiment measuring the rate of oxygen consumption (SI) confirmed that with a feed of smaller particles the maximum oxygen utilization rate increased. Total oxygen consumption after seven days was not significantly different, thus the extent of degradation was not affected by size in this experiment. This is similar to anaerobic digestion, where the degradation rate increases with decreasing protein particle size, but the extent of degradation is constant (Aldin et al. 2011). Operation of the system at near to realistic conditions prevented monitoring of biofilm development on the membrane surface. Measurements of particle size distribution were performed after biofilm sloughing events. Thus it is unlikely that the small particles (<10 μm) only accumulated in the biofilm.

• Biologically activated membrane bioreactors (BAMBis) were applied for treatment of used wash-water and successfully operated for an extended period of 220 days at stable flux. Suspended SRT and incomplete mixing had no effect on permeate flux and COD removal. Hence, there is no need to control SRT, resulting in a robust system that is easy to operate for a non-expert.

• Tuning aeration to develop anoxic zones in the aerobic reactor led to simultaneous nitrification and denitrification and increased nitrogen removal efficiencies. Nitrogen removal efficiencies of up to 86–97% were reached when the BAMBis were operated with low aeration intensities, leading to incomplete mixing.

• An important mechanism for degradation of particulate fecal matter is bacterial colonization of particulates. As colonization is surface limited, small particles are degraded at a faster rate than large particles.

• COD removal efficiencies in the range of 89–93% were achieved in three systems operated at vastly different suspended SRTs.

The laboratory staff and colleagues at the process engineering department at Eawag are thanked for help with sample analysis.