Linking transformations of organic carbon to post-treatment performance in a biological water recycling system

Ozone, electrolysis and granular activated carbon (GAC) were examined as potential post-treatments to follow a household-scale biologically activated membrane bioreactor (BAMBi), treating a wash water containing trace urine and feces contamination. Each post-treatment was evaluated for abilities and reaction preferences to remove or transform dissolved organic carbon (DOC), chemical structures that contribute color, and assimilable organic carbon (AOC), which can support bacterial regrowth. Batch treatment with each technology demonstrated an ability to remove ≥95% DOC. Ozone demonstrated a reaction selectivity through increased reaction rates with larger compounds and color-contributing compounds. Electrolysis and GAC demonstrated generally less-selective reactivity. Adding post-treatments to full-scale systems reduced DOC (55–91%), AOC (34–62%), and color (75–98%), without significant reaction selectivity. These reductions in DOC and AOC were not linked to reduction of bacterial concentrations in treated water. Reductions in bacterial concentrations were observed with ozone and electrolysis, but this is credited to oxidation chemicals produced in these systems and not the removal or transformations of organic materials.


Section S.1. Blue Diversion Autarky Toilet wash water context and composition
The BAMBi systems utilized in this study have been originally designed to serve the Blue Diversion Toilet (BDT) as part of the Bill and Melinda Gates Foundation Reinvent the Toilet Challenge. The BDT is a source separating toilet designed to provide safe sanitation to urban parts of the world that may have limited access to safe water sources and have no access to sewer facilities (Larsen et al. 2015). The toilet was designed to create one waste stream of urine and one waste stream of feces, so resources can be recovered from each waste. The original version of the BDT featured collection and centralized treatment of the urine and feces, while the updated version, called the Blue Diversion Autarky Toilet (BDAT), includes integrated urine and feces treatment. Both versions provide a water recycling component to supply safe water for hand washing, anal cleansing and flushing out the bowl where urine and washing water is collected. Though the BDT is a source separating toilet, field testing has demonstrated that 1-2% of the urine and feces produced by users are able to enter the water recycling system.
The wash water utilized in this study is formulated from actual human waste and soap to recreate what was observed in field testing. Human feces and urine were collected at Eawag (Dübendorf, Switzerland).
Bags of feces were frozen, diluted 100% in tap water and homogenized in a blender before feeding. Urine was added to the feeding tank after up to 24 h storage at room temperature. Soap was prepared from components: 140 g/L sodium dodecyl sulfate (VWR, Radnor, Pennsylvania, USA), 50 g/L glycerol (Merck, Darmstadt, Germany), 10 g/L NaCl (Merck, Darmstadt, Germany), and 0.72 g/L lactic acid (Alfa Aesar, Ward Hill, Massachusetts, USA) in tap water. The chemical characteristics of the concentrated feed water were approximately 1200 mg/L TOC, 300 mg/L total N, 100 mg/L-N combined ammonium and ammonia, and 200 mg/L chloride. The daily loading of carbon and nutrients divided by the daily volume of water processed in the BDT would create a wash water strength similar to that of greywater. Figure S1 depicts the 3 BAMBi units operated in this study, their feeding and waste plumbing and how post-treatments were integrated into each system. Coarse bubble (~3 mm diameter) aeration was introduced directly below the membrane module at a rate of 3 L/min. The daily feeding of 3 L/day was divided in a series of 40 events. Each event consisted of 75 mL of the concentrated feed and 1425 mL of water from the CWT being pumped back to the BAMBi. Treated water was removed from the CWT for each event to maintain a constant system volume. The combined 1.5 L of wash water added to the BAMBi for each event represents the combined inputs from hand washing, anal/personal cleansing and washing out the toilet bowl, as produced in the Blue Diversion Autarky Toilet. The 40 events correspond to 40 usages of the toilet, and the distribution of events throughout the day was based on expected usage, with high frequency at breakfast, lunch and dinner. The activation of the pumps controlling the concentrated feed, permeate, waste and the return of water from the CWT to the BAMBi was managed using process control hardware (Endress + Hauser AG Reinach BL, Switzerland), and automation softwares Codesys (3S-Smart Software Solutions GmbH, Kempten, Germany) and CitectSCADA (Schneider Electric, Rueil-Malmaison, France). Each BAMBi was initially started with 50 L of tap water, 0.5 L of activated sludge (from a conventional wastewater treatment plant in Switzerland) and 0.5 L of sludge from a urine nitrification reactor (Fumasoli et al., 2016). BAMBis were then operated for 2 months to allow the bacteria to acclimate and performance to stabilize before the more intensive data collection for this study began. No fouling control was performed on the membrane during testing. Approximately 2.75 L/d of CWT water was wasted to maintain system volume (~0.25 L/d lost to evaporation).

Section S.3. Details of LC-OCD and DOC measurements
The separation range for the LC-OCD column was with a range of 100-20,000 Da, and the lower quantification limit for DOC testing on the LC-OCD was 50 µg/L. All glassware for LC-OCD sampling and analysis was heated in a clean furnace to a temperature of 450 ˚C and held at this temperature for 12 h to incinerate any trace carbon residue. All samples for LC-OCD and DOC testing were each spiked with of 46 g/L sodium thiosulfate solution (0.25 mL per 50 mL of sample) to quench any oxidants.

Section S.4. Details of AOC, TCC and ICC processing
Samples for AOC processing were collected using 50 mL syringes and first filtered, with prewashed (50 mL of deionized water) 0.2 µm polyethersulfone filters (Pall Port Washington, New York, USA), into 60 mL glass vials and 1 mL of inoculum was added. Mixed samples were divided into three 45 mL glass vials, capped with PTFE-lined lids, incubated for 3 days at 30 ˚C without light and measured for TCC. AOC was calculated using the equation 1 µg AOC = 10 7 cells (Hammes et al. 2005). Inoculum was prepared from equal parts of (i) Evian water (Évian-les-Bains, France), (ii) permeate water from a BAMBi system fed the same as in this study, (iii) permeate water from a BAMBi system fed a synthetic greywater and (iv) water collected from the Chriesbach stream (Dübendorf, Switzerland). The BAMBi waters and the stream water were twice centrifuge washed (10 min at 3500 g) and re-suspended in mineral buffer (LeChevallier et al. 1993). The TCC of the inoculum was approximately 100 cells/mL. All glassware was muffled at 450 ˚C for 12 hours. All TCC, ICC and AOC samples were spiked with 0.25 mL of 46 g/L sodium thiosulfate solution per 50 mL of sample. The lower limit of quantification for TCC and ICC was 970 cells/mL and the lower limit for AOC quantification was 2.1 µg/L. Figure S2. GAC batch testing. NH4-total indicates that these data represent the combined values for ammonia and ammonium species. 8 Figure S3. Size-distribution profiles from GAC batch testing. Each sample is identified by the corresponding percentage of DOC removed by the treatment at the time of sample collection. This figure includes a peak that occurs at the same elution time as the biopolymer calibration peak, however this peak has not been included in the main text because we believe this peak may be contamination.