Operating a pilot-scale nitrification/distillation plant for complete nutrient recovery from urine

Separating urine at the source is an effective approach to recover nutrients from wastewater, given that urine contains most nutrients, which humans excrete (Larsen & Gujer 1996). Researchers have targeted specific nutrients (e.g. nitrogen, phosphorus) and developed technologies to reclaim them from the liquid (Larsen et al. 2013). Alternatively, our research group opted for another procedure: water is separated from urine, leaving behind a concentrate, which contains all nutrients. After extensive laboratory work (Udert et al. 2003; Udert & Wächter 2012), a pilot plant was started up at Eawag's main building, Forum Chriesbach. The plant operates in two stages: first, half of the total ammonia (⁠ and NH₃) in urine is biologically converted into nitrate ⁠, which is then concentrated with a distiller.

The aim of the nitrification step is to prevent ammonia (NH₃) losses and malodor. During nitrification bacteria convert half of the total ammonia into non-volatile nitrate and, as the pH drops from pH 9 to values around 6, the other half is stabilized as non-volatile ammonium ⁠. About 90% of the organic substances are mineralized including compounds which are responsible for the malodor (Troccaz et al. 2013). The resulting stabilized solution can be concentrated without any substantial nitrogen losses in order to simplify transportation, storage or direct application as fertilizer.

Four main factors determine the suitability of the pilot-scale reactor for nutrient recovery: first, nitrification rates determining the reactor size; second, the energy demand; third, stability of the biological processes; and fourth, the quality of the concentrate produced.

High urine nitrification rates of 380 mg N L⁻¹ d⁻¹ were observed in a laboratory-scale moving bed biofilm reactor (MBBR) at temperatures of 25 ± 0.3 °C and over a period of 80 days (Udert et al. 2003). However, stronger changes in temperature and urine composition have to be expected for the pilot-scale reactor, which may influence long-term performance.

The primary energy demand for the overall process (nitrification and distillation) has been calculated as 30 W cap⁻¹ based on laboratory results and literature data (Udert & Wächter 2012). This is two and a half times higher than for the treatment of municipal wastewater in a conventional wastewater treatment (Maurer et al. 2003). The energy consumption was estimated for the urine composition given in Udert et al. (2006). The concentrations from urine collected in urine-diverting toilets may, however, not reach equally high values (Goosse et al. 2009), which will cause a higher energy demand for water removal by distillation.

The stability of the biological process is particularly challenging, as it requires the well-tuned interplay of ammonia-oxidizing bacteria (AOB) and nitrite-oxidizing bacteria (NOB). A simple pH control was proposed to achieve high process stability in a urine nitrification reactor (Udert & Wächter 2012). The pH value is kept in a narrow range by intermittent dosage of stored urine based on the following principle: when no urine is pumped, the pH decreases due to nitrification; during urine dosage, pH increases due to the high pH value and alkalinity of stored urine. Udert & Wächter (2012) tested this strategy for the steady-state operation of a laboratory reactor, but not for reactor start-up, which is the most critical phase in reactor operation.

The nitrification process converts 50% of the total ammonia to nitrate, and heterotrophic bacteria degrade 90% of the organic substances (measured as chemical oxygen demand) (Udert & Wächter 2012). Most other compounds though are not degraded or removed during nitrification and become concentrated during distillation. If the concentrate is supposed to be used as a fertilizer, its composition needs to satisfy legal standards with respect to various compounds (e.g. heavy metals; Council Regulation (EEC) 1991).

This paper summarizes our previous experience with the nitrification/distillation pilot plant at Eawag. The results provide important information for further research and pilot studies. We address the nitrification rates, energy demand and the quality of the concentrate as fertilizer based on the results from the pilot-scale installations. The factors influencing process stability are discussed based on laboratory nitrification experiments.

Urine was collected from urine-diverting flush toilets (NoMix toilets, Roediger Vacuum, Hanau, Germany) and waterless urinals through a separate piping system at Eawag's main building (Dübendorf, Switzerland). In the building, an average of 110 L of urine is collected daily (only working days). In two 1,000 L tanks, women's urine and men's urine are stored separately before they are treated in the nitrification/distillation process. Urine from the women's storage tank is less concentrated by a factor of approximately two with respect to all compounds compared with men's urine (Table 1). The main reason for the lower concentration of women's urine, is the collection with NoMix toilets only: leaky valves allow some flushing water to enter the collection tank.

In order to reduce water loss during aeration of the biological reactor, the air from Eawag's pressurized air supply is humidified in a column containing distilled water prior to injection into the reactor. In the nitrification reactor, the instrumentation comprises pH, temperature, and dissolved oxygen probes (Tophit CPS491D, Oxymax COS61D, Liquiline CM448, Endress & Hauser AG, Reinach, Switzerland). Between the nitrification reactor and the distiller, an intermediate storage tank is installed, which holds 600 L nitrified urine corresponding to the volume of one distillation batch. The average concentration factor during distillation is between 20 and 25.

The distiller is a commercially available industrial vapor compression vacuum distiller (KMU-Loft Cleanwater, Hausen, Germany). It is set to operate at approximately 500 mbar working pressure. The vapor compression occurring in the vacuum pump is the sole source of heating, as the heated vapor is recycled into a heat exchanger located in the distiller's sump. The distiller is operated in semi-batch mode, i.e. the liquid volume in the sump is kept constant at 20 L by compensating for evaporated liquid with new influent. Thus, over a distillation batch, the concentration in the sump continuously augments up to a given level. Once a set volume of liquid has been distilled, the process is halted and the remaining concentrated liquid is drained from the distiller. In the case of our set-up, the distilled water is recycled into the toilet flush system. In the distiller, temperature, pressure, electric conductivity (as a proxy for concentration), and electricity consumption are measured and recorded. In both, the urine supply tanks and the intermediate storage tank holding the nitrified urine, differential pressure sensors (Vegaflex 61, VEGA Grieshaber KG, Schiltach, Germany) record the liquid level.

Laboratory experiments were conducted to test the start-up procedure. Each of the two reactors had a volume of 7 L. The pH and oxygen concentrations were measured continuously. Data were stored in a data logger (Memograph M, RSG40, Endress & Hauser, Reinach, Switzerland), which was also used to control the influent pumps (SCi-Q 400, Watson Marlow, Falmouth, UK). The airflow was maintained at 1 L min⁻¹ using a flow controller (EL-FLOW, Bronkhorst, Reinach, Switzerland), resulting in dissolved oxygen concentrations above 7 mg L⁻¹. The reactors were stirred magnetically at 500 rpm. The temperature was maintained at 25 °C with a thermostat (F32, Julabo Labortechnik GmbH, Seelbach, Germany).

For start-up, the reactors were filled with 0.7 L of activated sludge from the nitrification tank of Eawag's municipal wastewater treatment plant (WWTP), 140 mL of men's urine, and tap water. The reactors were operated with suspended biomass but no biofilm carriers. Men's urine was added automatically according to the pH control mechanism described above (Udert & Wächter 2012). The two reactors were operated in two different pH intervals: reactor 1 (R1) between 5.80 and 5.85 and reactor 2 (R2) between 6.2 and 6.25. As the low pH setpoints were initially not reached, four and two times 50 mL of urine were added manually within the first 15 days to R1 and R2, respectively.

Chloride, sulfate, phosphate, nitrate, potassium and sodium were analyzed with ion chromatography (Metrohm, Herisau, Switzerland). Magnesium, calcium, iron, copper, zinc, manganese, cobalt, copper, chromium, cadmium, nickel, lead and boron were determined with inductively coupled plasma optical emission spectrometry (Ciros, Spectro Analytical Instruments, Kleve, Germany). The total ammonia, total nitrite (nitrite and nitrous acid) concentration as well as the chemical oxygen demand (COD) were measured photometrically with cuvette tests (LCK 303, LCK 342, LCK 614 Hach-Lange, Berlin, Germany). Total inorganic and organic carbon (TIC and TOC) were measured with a TIC/TOC analyzer (IL550 OmniTOC, Hach-Lange, Berlin, Germany).

To characterize the distribution between suspended and attached nitrifying bacteria, biomass samples of the suspended biomass and of the carrier material were taken from the pilot-scale reactor. Samples were taken on 1 day in month 15 and on a second day, in month 29. The nitrification rates were substantially different in both cases (see Table 2). Furthermore, biomass samples were removed from each of the two laboratory reactors after 142 days of reactor operation. The biomass samples were stored at −20 °C prior to analysis. DNA was extracted using the FastDNA SPIN Kit for Soil (MP Biomedicals, Santa Ana, CA, USA) and analyzed by Research and Testing Laboratory (Lubbock, TX, USA) with bacterial tag-encoded FLX amplicon pyrosequencing using the primer pair 341F (5′-CCTACGGGNGGCWGCAG-3′)/785R (5′-GACTACHVGGGTATCTAATCC-3′) targeting the bacterial 16S rRNA gene pool (Herlemann et al. 2011).

AOB sequences affiliated with the Nitrosomonas europaea lineage and NOB sequences with the genus of Nitrobacter (Figure 2). The predominance of these nitrifiers was not surprising, as AOB from the Nitrosomonas europaea lineage are often selected in environments with high ammonia concentrations (Koops et al. 2006), while NOB of the genus Nitrobacter are adapted to higher nitrite concentrations than Nitrospira (Nowka et al. 2015). The relative abundance of all other AOB or NOB was below 0.1%.

The nitrification rate increased to 630 mg N L⁻¹ d⁻¹ at higher temperatures of 26.3 °C (months 29 to 30). At temperatures of 27.0 °C, the maximal rate of 930 mg N L⁻¹ d⁻¹ (3.1 g N m⁻² d⁻¹) was reached, corresponding to a discharge of 120 L d⁻¹ of women's urine. The high rates were maintained for only 10 days, because not sufficient urine was available from the women's urine tank. In this phase, both the particulate COD and the relative abundance of nitrifying bacteria in suspension increased (Table 2, Figure 2). Hence, the nitrification rate increased at higher temperatures, because both suspended and attached nitrifiers contributed to the overall conversion rate.

Following the switch from women's to men's urine (months 36 to 40), the nitrification rate dropped to average values of 120 mg N L⁻¹ d⁻¹. The particulate COD concentrations decreased drastically to 220 mg L⁻¹. In the biofilm of the carrier material, nitrifying biomass competes with heterotrophic bacteria for oxygen and space (Hem et al. 1994). The increased attachment of bacteria to the carrier material (visual observation) after the switch to men's urine further boosted this competition, resulting in lower nitrification rates.

The energy demand for nitrification is mainly caused by aeration. Due to the fact that the airflow was used for mixing, the airflow was kept constant independently of the nitrification rate. The electric energy demand for nitrification is relatively low (11 Wh gN⁻¹) at high nitrification rates (640 mg N L⁻¹ d⁻¹), but it increases to values of 59 Wh gN⁻¹ for the low rates observed in men's urine (120 mg N L⁻¹ d⁻¹, Figure 3).

The electric energy demand for the overall process amounted to 71 Wh gN⁻¹ in the best case (640 mg N L⁻¹ d⁻¹, 1,790 mg NH₄-N L⁻¹ in influent). By assuming a conversion efficiency of 31% for electricity production (average European electricity mix; UCPTE 1994) and a nitrogen production of 8.8 g N cap⁻¹ d⁻¹ (Maurer et al. 2003), the primary energy demand amounted to 84 W cap⁻¹ (71 W cap⁻¹ for distillation and 13 W cap⁻¹ for nitrification), which corresponds to 1.9% of the overall primary energy demand in the European Union (4,430 W cap⁻¹ in 2012; The World Bank 2015).

The energy demand is clearly higher than the estimated primary energy demand of 10 W cap⁻¹ for the combination of conventional treatment of municipal wastewater and industrial fertilizer production (Maurer et al. 2003) (nitrification/pre-denitrification in WWTP, average N- and P-fertilizer production in Europe, P-precipitation in WWTP including energy for sludge transport and incineration). However, the overall energy demand can still be reduced significantly, and additional energy demands in urban water management, e.g. for flushing water provision, have not been considered so far. In fact, the lower flushing water requirement for urine-diverting toilets has been shown to be an important energetic advantage of urine separation (Ishii & Boyer 2015).

Collecting urine that is as concentrated as possible is an important requirement to save energy. If urine is collected in a more concentrated form, however, the aeration rate should be reduced due to the lower nitrification rate. Stirrers instead of the aeration could be used to provide sufficient mixing under such conditions. Alternatively, different reactor types, e.g. membrane bioreactors, could achieve lower energy demands.

The start-up of the urine nitrification reactor was the main challenge during the operation of the pilot-scale reactor, mainly due to the accumulation of nitrite and the growth of acid-tolerant AOB (see below). The urine addition to the pilot-plant was increased manually, and several trials were required for successful reactor start-up. To simplify the start-up procedure, a simple pH control was tested in two laboratory reactors.

The pH setpoints had an influence on nitrite production: nitrite increased to a maximal value of 109 mg N L⁻¹ in R2, while the nitrite concentrations remained below 1 mg N L⁻¹ at all times in R1 (Figure 4). NOB are inhibited by nitrous acid (HNO₂), but despite this inhibition NOB were still capable of removing the excess nitrite in R2 (Anthonisen et al. 1976), which was not always the case during the start-up of the pilot-scale application.

Nitrite accumulated in R2 but not in R1, as the influent rate increased faster in R2 compared to R1 (Figure 4). The faster increase in the urine dosage can be explained by the faster growth of AOB at higher pH values (Fumasoli et al. 2015), which boosted the automated urine addition. AOB and not NOB control the influent regime, because protons are released during ammonia oxidation to nitrite.

The inflow reached a maximal rate of 240 mL d⁻¹ in R1 (pH 5.8) and 370 mL d⁻¹ in R2 (pH 6.2) corresponding to a hydraulic residence time of 28.8 d in R1 and 18.9 d in R2, respectively (Figure 4). In the course of the experiment the inflow rates decreased, which may be explained with the increasing salt concentrations; high salt concentrations are known to reduce AOB and NOB activity (Moussa et al. 2006).

Acid-tolerant AOB are a potential risk in urine nitrification reactors, because some acid-tolerant strains can decrease the pH to values as low as 2.2 (Fumasoli et al. in preparation), in the case where no or only very low amounts of urine are added (e.g. during holidays), which could buffer the pH value. Low pH values can inhibit NOB and lead to chemical decomposition of nitrite (Udert et al. 2005), during which significant amounts of harmful gases, such as nitric oxide (NO), nitrogen dioxide (NO₂), nitrous oxide (N₂O) and HNO₂ are released (Fumasoli et al. in preparation).

An appropriate selection of the pH setpoint is therefore important for urine nitrification. A too high pH setpoint can lead to the accumulation of nitrite and subsequent inhibition of NOB, whereas a too low setpoint can foster the selection of acid-tolerant AOB and a consequent process destabilization. Online nitrite monitoring would be a possibility to explore the maximum pH setpoint. Unfortunately, online sensors for the high concentrations to be expected in urine treatment are currently not available, but recent tests with an ultraviolet spectral probe are promising (Mašić et al. 2015).

The concentrate has a high nitrogen content, but it also contains other important nutrients, such as phosphorus and potassium (Table 1). Most of the compounds from the initial influent are recovered in the final concentrate. Exceptions are the organic substances, ammonia and TIC: organic substances are oxidized during the biological treatment, about half of the ammonia is nitrified to nitrate and nearly all TIC is lost due to CO₂ volatilization, mainly during the biological treatment.

Studies on pharmaceuticals removal showed that some antiviral and antibiotic compounds were substantially degraded in the nitrification reactor. Remaining pharmaceuticals can be removed after the nitrification process by adsorption onto activated carbon (Oezel Duygan et al. in preparation).

The heavy metal content in the concentrated nutrient solution is far below the general limits for fertilizers in organic agriculture (Council Regulation (EEC) 1991; Table 1). The heavy metal concentrations in the concentrate are close to the expected values taking into account the heavy metal concentrations in urine (Ronteltap et al. 2007) and a concentration factor of 20 to 25. The heavy metal content in urine is generally low, because heavy metals are mostly excreted with feces (Jönsson & Vinnerås 2013).

The greenhouse trials with synthetic solutions representing concentrated nitrified urine showed that the fertilizer performed equally or better than commercial fertilizers (Bonvin et al. 2015). The trials were conducted on Italian ryegrass (Lolium multiflorum) for a growth period of 11 weeks in sandy, slightly acidic (pH = 6.5) soil with moderate phosphorus levels.

The nitrification/distillation process produces a concentrated nutrient solution, which could be well suited as an integral fertilizer. Current challenges of the system are the high energy demand and the possible process destabilization due to nitrite accumulation or low pH values. The inhibition by nitrite as well as the growth of acid-tolerant bacteria could be prevented by online monitoring and controlling the nitrite concentration. This should be the main focus of further research. The energy demand can be decreased by collecting urine as concentrated as possible and by energetically optimizing the nitrification reactor.

We thank the Bill and Melinda Gates Foundation for the funding received for the VUNA project (www.vuna.ch; Grant No. OPP1011603) and Eawag for infrastructure funds to construct the reactors. We also thank Karin Rottermann and Claudia Bänninger-Werffeli for their dependable laboratory analyses, and Mathias Mosberger and Corine Uhlmann for reactor operation and sampling. Furthermore, we greatly appreciated the frequent knowledge and experience exchange with our project partners in South Africa, Christopher Buckley, Maximilian Grau, Sara Rhoton, and Lungiswa Zuma, who have been operating a similar reactor combination in Durban.