Journal of Environmental Management 90 (2009) 1909–1917

Contents lists available at ScienceDirect

Journal of Environmental Management

journal homepage: www.elsevier.com/locate/jenvman

Real-life efficiency of source separation

L. Rossi a,*, J. Lienert b, T.A. Larsen b a Swiss Federal Institute of Technology (EPFL), ENAC ISTE ECOL, Station 2, 1015 Lausanne, Switzerland b Eawag, Swiss Federal Institute of Aquatic Science and Technology, U¨berlandstrasse 133, 8600 Du¨bendorf, Switzerland article info abstract

Article history: Urine source separation (NoMix technology) is a promising innovation in wastewater management. To Received 3 July 2008 improve and further develop NoMix technology, it has been implemented in four Swiss households and Received in revised form at our research institute (Eawag). We conducted measurements during one year on frequency of 9 December 2008 usage (in households 5.2/person/day for weekdays, and 6.3/person/day for weekends), flushing behavior Accepted 8 January 2009 (30–85% small flushes), and recovered urine. We calculate the amount of urine effectively recovered per Available online 5 February 2009 voiding in NoMix (138 ml/flush in households; 309 ml/flush in women’s toilets at Eawag), and waterfree (225 ml/usage). We estimate urine recovery in the households to be maximally 70–75% Keywords: Flushing pattern of the expected quantity, leaving room for technical and behavioral improvements. Based on sampling of Nitrogen loading N and P concentrations, we suspect nitrogen losses in the extended urine piping system. For households NoMix technology and workplaces, the daily and weekly flushing pattern is recorded. Our results are in accordance with Toilet use literature data from a shorter period but with more people. These results represent a good dimensioning Urine recovery basis for future urine source separation applications. An example of extrapolation to an entire watershed is presented. The flushing pattern corresponds well with the typical nitrogen loading of a treatment plant. Ó 2009 Elsevier Ltd. All rights reserved.

1. Introduction per capita for the investment in NoMix installations, assuming a life expectancy of 15 years for these installations. CSO control is also Urine source separation has the potential to become an integral relatively expensive due to the high price of CSO tanks (in part of mainstream wastewater treatment where high nutrient Switzerland about 1000 US$/m3). In one example, a CSO tank of elimination rates are required (Larsen et al., 2007). Although the 2000 m3 was necessary to avoid ammonia risks in the environ- potential is of course larger in areas with no existing wastewater ment, which corresponds to a total investment of about 2 M US$. management infrastructure, there is still room for improvements of Rossi et al. (2004) calculated that with the same money invested in the existing European wastewater treatment plants with NoMix at-source control of urine, about 2100 US$/NoMix installation could technology (Lienert and Larsen, 2007a) – especially if one wants to be invested, however assuming a life expectancy of these installa- obtain high nutrient removal and recycling rates and at the same tions equal to that of CSO tanks. Of course, NoMix technology does time optimize the total energy demand (Maurer et al., 2003; Wil- not allow for the control of suspended solids that play an important senach and van Loosdrecht, 2006). Urine source separation is role during rain events in urban areas. However, if urine in CSOs promising with respect to nutrient removal optimization, because was avoided, one could concentrate on particulate matter without ca. 80% of N and 50% of P from domestic wastewater at treatment the trade off experienced today between suspended matter and plants is contained in urine (Larsen and Gujer, 1996). Furthermore, soluble pollutants (urine also contains about 30% of dissolved NoMix technology may rapidly prove economically interesting, not organic carbon (DOC) and a large part of the soluble phosphorus in only for savings on the treatment plant, but also where combined raw wastewater). sewer overflow (CSO) tanks are primarily necessary due to high One may also want to extend the life time of an existing, over- ammonia loadings. Based on the savings at treatment plants, loaded treatment plant, either by peak shaving (Rauch et al., 2003) Maurer et al. (2005) set approximate benchmarks of 260–440 US$ or by separating advanced nutrient elimination from removal of organic matter (Wilsenach and van Loosdrecht, 2004). If NoMix technology shall rely on local storage and subsequent transport of urine in combined sewers as first suggested by Larsen and Gujer * Corresponding author. Tel.: þ41 21 693 57 80; fax: þ41 21 693 56 70. E-mail addresses: luca.rossi@epfl.ch (L. Rossi), [email protected] (J. Lie- (1996) (for explanations see below), avoiding urine in CSOs is of nert), [email protected] (T.A. Larsen). special importance, and storage capacity must be optimized in

0301-4797/$ – see front matter Ó 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.jenvman.2009.01.006 1910 L. Rossi et al. / Journal of Environmental Management 90 (2009) 1909–1917 order to avoid transport during rain. Transport of locally stored Faeces etc. urine in combined sewers makes sense in two fundamentally different situations: ‘wave transport’ or ‘peak shaving’. In both versions, urine is stored in the household during daytime and NoMix toilets transported through combined sewers at night. For ‘wave trans- Urine port’, the idea is to obtain as concentrated a urine solution as possible at the treatment plant for further processing, and the Urine tank Sewer release of urine from the storage has to be synchronized accord- with valve Waterfree ingly, taking transport times into account (Huisman et al., 2000). operated by urinals For ‘peak shaving’, one aims at the opposite effect: to level out the remote control nitrogen loading of the treatment plant over 24 h in order to optimize nitrification (Rauch et al., 2003). There are other (combinations of) transport possibilities, but the most difficult optimization of storage capacity follows from the transport in Fig. 1. Principle of urine source separation technology (NoMix toilets) with controlled combined sewers. A stakeholder perspective on different urine release of urine to the sewer system. storage and transport options is presented in Borsuk et al. (2008). For ‘wave transport’, high safety levels would have to be set because overflows would be severe with concentrated urine in the sewer. To study in detail the storage of urine at source in detail, data For this reason we have focused on the peak shaving option, which were collected over one year in four apartments equipped with would have immediate advantages at the treatment plant (opti- NoMix toilets in a Swiss city. The residents were only little informed mizing nitrification) and at the same time would allow for learning about the reason for the NoMix toilets before they moved into the whether real-time control of urine storage tanks is reliable enough apartments. During the study, information and coaching were to assure safe transport also of concentrated urine in combined intensified in order to secure continuation of the study that had sewers. For peak shaving, however, the optimization problem has some negative consequences for the residents (frequent visits of two parameters. Leveling out the nitrogen load of the treatment researchers, practical problems, etc.). A total of ten persons lived in plant and holding back urine in household storage tanks when it the four apartments, including one child of around six years, which rains are opposed strategies, which call for integrated manage- we counted as half a person as approximation in order to account ment. Rauch et al. (2003) demonstrated that in a typical catchment for the smaller amount of urine produced by children of that age around Zu¨ rich in Switzerland, a storage capacity of 10 l per toilet and less toilet use per day (Laak, 1974; Friedler et al., 1996b; and the simplest possible real-time control strategy could produce Almeida et al., 1999; Nijman et al., 2002). The characteristics of a 30% reduction in the peak loading of nitrogen (corresponding to inhabitants (age, occupational activities, socio-economics, etc) are a 30% increase of capacity of the treatment plant) and at the same described elsewhere (Thiemann, 2006). The German company time could reduce the amount of raw urine discharged during rain Roediger produced the NoMix toilets (www.roevac.com), and the events by 50%. A real-time control strategy may be implemented, Swiss company Geberit the water cistern, equipped with two using the electrical network or wireless technology to give a signal flushing modes (www.geberit.com). A small flush corresponds to to empty the recipient. ca. 3 l of water, a big one to ca. 6 l. A closing mechanism opens when In all cases, a necessary pre-condition for a successful integra- people sit on the toilet, allowing urine to be collected separately in tion of NoMix technology into the existing sewer system is a proper the front of the toilet. When people rise to flush, the urine outlet is functioning and understanding of the household sanitary devices: closed, and the entire toilet bowl is cleaned with a normal flush, the NoMix toilet and storage tank. To this end, we must look at the without diluting the urine. To estimate the frequency of toilet uses, technology where it is actually implemented, i.e. in the . an electrical signal was generated when people pushed the full or For a systematic discussion of the difficulties associated with pilot the small toilet flush button, and it was stored on a logger (Eltek projects in , see Lienert and Larsen (2007b). 1000 Squirrel) that recorded the date and time of each flush. In Hence, the aim of this study was to better understand urine order to avoid registering multiple flushes (if people press the same collection and storage in order to optimize the NoMix technology. button several times), the registration of the next flush event was Rather than relying on sociological data (questionnaire surveys, blocked during an interval of about 17 s. This time corresponds to diaries) as others, we used technical measurements to collect data. the filling of the toilet reservoir. The urine from the NoMix toilets We also discuss some of the advantages and shortcomings of this was stored in a 200-litre recipient in the cellar of the house. An approach. The comparison of our results with other literature ultrasonic probe (Milltronics PL-372 probe) measured the urine information will help to define a dimensioning basis for future level every 10 min (acquisition of data in a Microdaq VC-301 implementation of urine separation at source. In this paper, we will process recorder) and converted it to liters according to a linear extrapolate our results to an entire watershed in order to validate rating curve. our approach. The same type of data were collected at a working place, at Eawag, the Swiss Federal Institute of Aquatic Science and Tech- 2. Material and methods nology, during one year. We registered the time of toilet use on a logger in the men’s and in the women’s bathrooms, each equip- 2.1. In house experimental set-up ped with one NoMix toilet (www.dubbletten.nu). In this NoMix toilet, urine is flushed with a small amount of water, and feces with The basic idea of NoMix technology is the separation of urine a normal flush. The amount of flush water for the urine outlet is and feces by means of NoMix toilets and in some cases waterfree supposed to be about 0.1–0.2 l, but in this case it was actually only urinals at the household or institutional level. Urine is stored in 8 ml (6 direct measurements in the toilet itself). The men’s bath- a tank located in or near the households/institutions and if inte- room was further equipped with one conventional toilet and three grated in conventional sewer-based wastewater treatment, feces different waterfree urinals: Ernst (www.system-ernst.de), Uridan are evacuated through the existing sewer system to the wastewater (www.uridan.ch), and Urimat (www.urimat.com). The frequency of treatment plant (WWTP; Fig. 1). use was estimated using a motion detector. The urine from L. Rossi et al. / Journal of Environmental Management 90 (2009) 1909–1917 1911 the waterfree urinals and NoMix toilets was stored in two 500 l 12 tanks situated one floor below, separately for men and women. An ultrasonic probe (Milltronics PL-372 probe) was installed on the 10 urine tanks, during one year for the men’s tank and three months for the women’s tank. All information about toilet use or urine 8 levels in tanks was stored in a database (CODEAU, 1996), allowing a validation of the time dependent variables over long periods. Statistical analysis was performed with the Data Desk software 6 (www.datadesk.com). Social enquiry has been conducted in households through personal interviews and through question- 4 naires at the working place (Lienert and Larsen, 2006). 2 2.2. Validation of the results at the level of a watershed and per day (all apartments) Amount of toilet use per habitant 0 In order to extrapolate the results from at-source control to Mon Tue Wed Thu Fri Sat Sun a watershed level, a measuring campaign was performed during 6 weeks in 2002/2003 on the wastewater treatment plant (WWTP) Extreme value (>95%) Adjacent value Ergolz1 situated in Canton Baselland, Switzerland. The watershed 75% comprises 23,500 habitants (no age distinction). For flow o Extreme value (95%) Median 25% measurements, an ultrasonic probe (Milltronics PL-372 probe) was installed at the entry of the WWTP on a platform, upstream of a combined sewer overflow (CSO). Information was collected every o minute on a Milltronic Logger. The relation between level measurements and flow was established through flow measure- Fig. 2. Pattern of average daily use of toilet per habitant in households over an average ment from a Venturi channel installed downstream of the CSO at week. the entry of the WWTP. This Venturi channel was inspected and calibrated three months before our measuring campaign. During 3. Results rain events, the amount of water discharged to the CSO tank was recorded in order to be able to calibrate also at high water flows. 3.1. Pattern of toilet use in households Ammonia was measured in-line with a WTW selective membrane device (AmmoLyt 700 IQ ammonium sensor, in Fig. 2 shows the pattern of toilet use during an average week conjunction with the AmmoLyt ammonium combination elec- (Monday–Sunday) for all the 9.5 residents in the four private trode). Data were collected each minute on the IQ Sensor-Net households investigated. This figure summarizes more than one system from WTW. The system was maintained by automatic, in- year of measurements. During weekends, the average number of situ air cleaning using compressed air. The probe was installed on toilet uses was significantly larger than during weekdays (week- a swimming mounting device at the entry of the WWTP, allowing days: 5.2 flushes/person/day; weekends: 6.3 flushes/person/day; T- for measurement in a turbulent flow regime, without perturbations test, p 0.0001 < a ¼ 0.05; Table 1). The variations are relatively due to or sanitary wastes. The probe was calibrated high, as we consider one year of measurements, including holiday before use and also on a weekly basis during the measuring periods. campaign. Ammonia analyses in the laboratory were performed on In Fig. 3, the hourly flush frequency per apartment is shown. We manual samples of 1 l, 5–10 min after sampling (Dr. Lange chose percentage of usage in order to be able to compare the four #LCK303). The same sample was analyzed 24 h later, after complete apartments directly without taking into account the differences in decomposition of urea present in the effluent. On the basis of these number of flushes per day. The patterns represent the use of toilets manual measurements, the on-line ammonia measurements were over 24 h for an average weekday or weekend. In order to facilitate converted to total urea and ammonia N (a linear correlation was the presentation, the standard deviation is not illustrated. found with R2 ¼ 0.92). Since flow and concentration were both There are of course differences between the four apartments, measured every minute, the resulting load of ammonia/urea-N was but in general the pattern is easily recognizable, especially on obtained with the same high resolution. weekends where the results are less influenced by the resident’s

Table 1 Average number of toilet uses and average urine collected in households and a working place (standard deviations in parentheses).

Number of toilet uses (¼flushes) per person Amount urine collected per toilet use Amount urine collected per day (ml/ and daya (ml/use) person/day)a Households average (Roediger NoMix 5.6 (1.4) 138 (45) 804 (265) toilet) Households weekdays (Roediger NoMix 5.2 (1.2) 144 (43) 671 (232) toilet) Households weekend (Roediger NoMix 6.3 (1.3) 135 (43) 970 (340) toilet) Eawag women (Dubbletten NoMix 309 (292) toilet) Eawag men (Dubbletten toilet plus 225(87) urinals)b

a Because it was not possible to identify the number of individuals using the Eawag NoMix toilets, we cannot identify the number of flushes per person and the urine collected per person. Values in this column cannot be estimated multiplying column 2 and 3, as the distributions are not similar. b Note that 88% were uses of the waterfree urinals, only 12% were NoMix toilets uses. 1912 L. Rossi et al. / Journal of Environmental Management 90 (2009) 1909–1917

12% Apart. 1 Weekdays Apart. 2 10% Apart. 3 8% Apart. 4

6%

4%

2% % of toilet use per hour 0% 4-5h 5-6h 6-7h 7-8h 8-9h 00-1h 1h-2h 2h-3h 3h-4h 9-10h 10-11h 11-12h 12-13h 13-14h 14-15h 15-16h 16-17h 17-18h 18-19h 19-20h 20-21h 21-22h 22-23h 23-24h

Apart. 1 12% Weekend Apart. 2 10% Apart. 3

8% Apart. 4

6%

4%

2% % of toilet use per hour 0% 4-5h 5-6h 6-7h 7-8h 8-9h 00-1h 1h-2h 2h-3h 3h-4h 9-10h 10-11h 11-12h 12-13h 13-14h 14-15h 15-16h 16-17h 17-18h 18-19h 19-20h 20-21h 21-22h 22-23h 23-24h

Fig. 3. Toilet use profile for the four different apartments during weekdays (above) and weekends (below). The vertical axis represents the percentage of toilet use during discrete one-hour intervals (sum over 24 h ¼ 100%).

different working situations. Not surprisingly, people tend to go to Fig. 4 (left). In the same figure, the distribution of the amount of the toilet a few hours later in the morning during weekends. urine per toilet flush (Roediger NoMix toilet) is presented on the right. An average value of about 138 ml/per flush (45 ml) is esti- 3.2. Urine production in households mated (Table 1).

The total daily amount of urine collected in the tank was 3.3. Pattern of toilet use at working place evaluated considering 174 days of measurements (results not illustrated). The average amount of urine collection per day The NoMix toilets at Eawag are situated in the public area differed significantly (T-test, p 0.0001 < a ¼ 0.05) between bathrooms in order to ensure access also to visitors who took part weekdays (6.37 (2.20) l/day) and weekends (9.22 (3.23) l/day; in different acceptance studies (Pahl-Wostl et al., 2003; Lienert and Table 1). Larsen, 2006). The total number of toilet uses per day is presented The correlation between the total number of flushes per day in Fig. 7, for a period of 13 months. Extreme values are mainly with the amount of urine collected in the urine tank is presented in linked with social events. The toilets are located in a public area

25 25

20 20 y = 0.137x R2= 0.3227 15 15

10 10 the tank (l) 5 5

0 Number of data in each category

0 0 Daily amount of urine collected in 30 60 90

02040 60 80 100 120 150 180 210 240 270 300 Daily number of toilet use (-) Volume of urine per toilet use (ml/flush)

Fig. 4. Correlation between total number of toilet uses and amount of urine collected in households (left) and distribution of urine volume per toilet use (ml/flush) in households (right). L. Rossi et al. / Journal of Environmental Management 90 (2009) 1909–1917 1913

200 Even if the average number of flushes per day stayed relatively constant during the measurement period in the households * (5.8 1.4 average total toilet uses per day), we observed a devel- 160 opment in the use of the small flush button (Fig. 7). In apartment 4, * the initial high proper usage stabilized at about the same level (about 50–65% use of the small button) as in the apartments 1 and 120 3, where initial proper usage was low. In apartment 2, there was * a general low proper usage, stabilizing at the very low level of 10% * use of the small button towards the end of the period with a short 80 intermittent phase where some improvement had been observed. For the working place, the frequency of small flushes varied 40 considerably over time for the women’s NoMix toilet (Fig. 7

Total amount of toilet use at bottom; mean values: 49.5% 18.0 small flushes for women, and working place per day (n/day) 29.7% 10.7 for men). This very low frequency of small flushes of 0 * * the women’s NoMix toilet can be linked to the presence of students Mon Tue Wed Thu Fri Sat Sun during a few weeks at the beginning of 2005, as illustrated by the increase of toilet uses during the same period. The very high Fig. 5. Number of toilet uses per day at a working place (Eawag; only men’s toilets i.e. three waterfree urinals and one NoMix toilet). frequencies of small flushes during the period April–June 2004 (80–88%) are difficult to explain, but occur during a period of less near restaurant, auditorium and meeting room facilities. It is thus frequent use of the toilet. A general positive trend over the entire not possible to evaluate the average use of toilet per person. measurement period was observed between the proportion of We correlate the amount of urine stored in the two tanks with small flush uses and the total number of toilet uses, but this the number of usages recorded in the men’s (Figs. 5 and 6 left) and correlation is weak (R2 ¼ 0.54, results not illustrated). For the men’s women’s bathroom (Fig. 6 right). The correlation is high for the NoMix toilets, the percentage of small flush was low, as expected: men’s (R2 ¼ 0.82), but low for the women’s (R2 ¼ 0.29) bathroom. Due to the presence of urinals nearby, this toilet is mainly used for The totality of urine collected in waterfree urinals reaches the . Nevertheless, a constant decrease of small flush urine tank (i.e. no losses). As more than 88% of the toilet visits are frequencies of the men’s toilet was observed, independently from related with urinal use, the correlation between daily number of the number of total toilet uses (R2 ¼ 0.4). toilet use and volume of urine collected daily is thus higher for the men’s toilets (Fig. 6, left) than for women (Fig. 6, right). The weak 3.6. Extrapolation to an entire watershed correlation for the women’s NoMix toilet use could be explained with physiological differences. Moreover, the duration of the The urine production estimated for the four households (average investigated period is longer for the men’s toilets (392 days) than values, from Fig. 3) can be compared with on-line measurements of for women (52 days). The volume of urine per usage had an average the WWTP Ergolz1 conducted during weekdays and weekends. In value of 225 ml/usage for men (i.e., including urinals) and 309 ml/ this specific case, the average load for weekdays estimated by our flush for women (only Dubbletten NoMix toilets; Table 1). own on-line measurement system (176 kg N/day) was comparable (T-test, p ¼ 0.0407 < a ¼ 0.05) with the average daily load measured 3.4. Concentrations of nutrients in urine tanks by the WWTP itself over one year (183 kg/day). Results of daily profile of NH4–N are presented in Fig. 8 for weekdays (top) and for Concentrations of nutrients (NH4,Ptot,PO4, and Ptot) in the weekends (bottom). The results have been normalized by dividing different urine tanks are presented in Table 2. the average hourly amount of N by the average daily load. The two curves agree very well. In view of the small number of 3.5. Frequency of small and large flushes in the households and at apartments, this is rather surprising. Although it cannot be the working place excluded that this is partially due to chance, the results indicate that already on the level of an apartment house one may observe In Fig. 7, the mean daily use of the small button is compared to more or less the same time pattern as in a small catchment, at least the percentage of total flushes for households (top) and for the when averaged over time. In this catchment, two thirds of the working place (bottom). population is situated near the WWTP so that there is only a small

50 20

y = 0.247x y = 0.2862x 2 40 R = 0.8199 16 R2 = 0.2919

30 12

20 8

10 4 Daily amount of urine Daily amount of urine collected in the tank (l) collected in the tank (l)

0 0 04080120160 010203040 Daily number of toilet use (-) Daily number of toilet use (-)

Fig. 6. Urine collected in the tank and number of toilet uses per day at Eawag (Left: men’s bathroom, n ¼ 392; Right: women’s bathroom, n ¼ 52). 1914 L. Rossi et al. / Journal of Environmental Management 90 (2009) 1909–1917

Table 2 Measured nutrient concentrations in urine and literature values (interval of measurements in brackets).

3 3 3 3 n NH4 gN m Ntot gN m PO4 gp m Ptot gp m NH4/PO4 [] Average physiological values (Ciba-Geigy, 1977) 7700 (as urea) 9200 510–1340 8.3 (5.7–15) Households 5 1905 (1872–2012) 2259 (2150–2300) 387 (378–413) 430 (402–436) 4.9 Working place, men’s bathroom 21 5027 (4510–6510) – 232 (213–284) – 22 Working place, women’s bathroom 1 3795 4590 200 19 Jo¨nsson et al., 1999 2125–3576 2350–3631 308–313 9.2 Jo¨nsson et al., 1997 3516–3900 3626–4000 306–350 11 time lag between wastewater production in the houses and its source separated urine. Assuming that all flushes would be related arrival at the WWTP. Obviously, the exact time of arrival of the to voiding of urine, our results on the number of toilet uses are ammonia morning peak at the treatment plant does not only relatively high, since the inhabitants also visit toilets outside their depend on the ‘lag time’ between production of urine in the homes (Fig. 2; Table 1). Medical literature reports a mean of 5.1 household and arrival at the treatment plant, but also on the urine voids per day (standard deviation 1.8; Burgio et al., 1991). In ‘morningness’ of the population in question (Dı`az-Morales and that study, the men had 4.8 urine voids per day (n ¼ 403), and the Sa`nchez-Lo` pez, 2004). It is clear from the data that the (more rural) women 5.6 (n ¼ 263); 94% of all participants voided fewer than 9 population in the catchment rises a little earlier than the (more times per day. Toilet uses related to defecation alone (without urban) people whose urine production was observed. The nearly ) explains only about 2.2% of total toilet uses (Friedler exact match in timing is of course purely coincidental. et al., 1996a), and cleaning actions have been recorded at a rate of only 1.1 applications per household and week (Friedler et al., 4. Discussion 1996b). Therefore, one may speculate that the toilets in our study were used for flushing away other wastes. A similar 24-hour toilet 4.1. Pattern of toilet uses in households use profile as in our study (Fig. 3) was found in Southern England, where weekend patterns differed from weekday usage, particularly The pattern of toilet uses is an important aspect if one wants to in terms of the timing and magnitude of the morning peak (Friedler construct rational and economic storage and drainage systems for et al., 1996b). We conclude that the results are plausible although probably a little high, and that even on weekdays, the household Apart.1 Apart.3 toilets are used quite often. The use of loggers instead of diary Apart.2 Apart.4 sheets, as used in other studies, has advantages and disadvantages. 80% The technical solution is independent of the enthusiasm, accuracy, 70% and reliability of people participating and can thus be extended over very long periods, as in our case (more than one year). Thus, 60% we can cover more realistic situations occurring during the 50% measuring period, including exceptions such as social events, 40% holidays, guests at home, etc. There is also no effect from diaries 30% that may cause people to change their toilet habits. A major 20% disadvantage is that it does not distinguish between different uses 10% of the toilet, and no specific data about users are available. flush uses per month 0% Average percentage of small 4.2. Urine production in households and at the working place 12.01 01.02 02.02 03.02 04.02 05.02 06.02 07.02 08.02 09.02 10.02 11.02 12.02 01.03 02.03 The total physiological amount of urine produced by the human Tot Women No Mix men metabolism is estimated at 1.349 l/day/person (0.412) in a study Tot men No Mix Women of about 80 people (Krieg et al., 1986). The amount of urine collected with the Dubbletten NoMix toilet was estimated between 90% 900 0.61 and 1.09 l/person/day (Jo¨nsson et al., 1999). In our study, the

80% 800 Total number of toilet uses amount of urine collected with the Roediger NoMix toilet was in 70% 700 the same range in households (Table 1). The amount of urine collected for each toilet use is relatively low compared with clinical 60% 600 values of bladder voiding volume (an average value of 260 ml/ 50% 500 voiding; Krieg et al., 1986; Burgio et al., 1991). This indicates that there are practical problems associated with the use of NoMix 40% 400 toilets in these households that are open to interpretation. First, we 30% 300 suspect that the number of toilet visits (Fig. 2; Table 1) may be the 20% 200 result of a measurement artifact, caused perhaps by more frequent flush uses per month cleaning and use of the toilet for other purposes than reported in 10% 100 Average percentage of small other studies (e.g., waste disposal). Other explanations are related 0% - to compliance and to the construction of the Roediger toilet: Urine is only captured if people actually sit on the , which in

08.03 10.03 12.03 02.04 04.04 06.04 08.04 10.04 12.04 02.05 04.05 06.05 08.05 10.05 a private household may not appeal to all men. In our institutional setting, the sitting position is not required. Crouching suffices due Fig. 7. Frequency of small flushes to the total number of flushes in households (top) to a different construction of the Dubbletten NoMix toilet, and men and in the working place (bottom). The curves represent the average percentage of small flush values per month. For the working place, the total number of NoMix toilet can use the urinals if they want. According to our related ques- uses (men and women) is also presented. tionnaire survey, 72% of the users sit on the NoMix toilets in the L. Rossi et al. / Journal of Environmental Management 90 (2009) 1909–1917 1915

12% Measurements WWTP 10% Flush profile (households)

8%

6%

4%

2%

0% 4-5h 5-6h 6-7h 7-8h 8-9h 3h-4h 9-10h 10-11h 11-12h 12-13h 13-14h 14-15h 15-16h 16-17h 17-18h 18-19h 19-20h 20-21h 21-22h 22-23h 23-24h

10%

Measurements WWTP 8% Flush profile (households)

6%

4%

2%

0% 4-5h 5-6h 6-7h 7-8h 8-9h 3h-4h 9-10h 10-11h 11-12h 12-13h 13-14h 14-15h 15-16h 16-17h 17-18h 18-19h 19-20h 20-21h 21-22h 22-23h 23-24h

Fig. 8. Profile of N–NH4 measured during weekdays (above) and weekends (below) at the entry of the WWTP Ergolz 1 and comparison with the mean number of flushes for all four apartments together (from Fig. 3). institution (Lienert and Larsen, 2006). Moreover, the low correla- because this small flush apparently does not make much difference tion between the total number of flushes per day in households and as compared to not flushing. the amount of urine collected in the tank (Fig. 4, left) confirms that Measurements in the urine recipient were not precise enough to there are problems caused by toilet design or compliance of the determine the exact amount recovered from each urine-related residents, e.g., because urine is lost when somebody does not sit on flush. For this reason, we consider only the daily amount of urine the toilet. production in our results. If future work is conducted on urine For working places, we found a low correlation between the tanks, the shape of the tank or the measuring device must be number of women’s toilet uses per day and the urine collected in adapted to allow more precise measurements. the tank (Fig. 6) and since the women used exclusively Dubbletten NoMix toilets (i.e., no urinals), there seem to be practical problems 4.3. Dilution of urine in urinals and NoMix toilets with the use of the NoMix toilets leading to large differences in urine recovery between the single visits to the toilet. Moreover, The interpretation of the data for phosphorus in urine is always because the data from the men’s toilets at Eawag are dominated by difficult (Table 2), since normally the probes are filtered and waterless urinals, we consider them a good proxy for the physio- therefore do not include precipitation of phosphorus. Therefore, logical reality amongst the users of this bathroom (Table 1) because the concentrations will depend on the amount of calcium and hardly any urine is lost when using a urinal. There are however magnesium ions added with the flushing water (Udert et al., some uncertainties, especially connected to the recording of urinal 2003b). In a comparison between our own pilot projects, it makes uses. Every cleaning action will result in a recording by the motion sense that the concentration in the Eawag tanks is lower than in the detector occurring 1–3 times a day. However, apart from this fact, household tanks, because the water hardness is higher in Du¨ ben- we assume that there is not much motion detection due to the dorf than in the city where the household pilot project was con- presence of people around the urinals and not using them. This ducted. At Eawag, urine from the women’s tank is more diluted assumption is also supported by the distribution between urinal than urine from the men’s tank. Since only little water (8 ml/flush) and toilet use being similar to that found in a user survey of NoMix is used for flushing the Dubbletten NoMix toilets, the prevalent use toilets in a vocational college, where 71% (4.3) of the men visited of urinals in the men’s room cannot explain the difference. a urinal (Lienert and Larsen, 2006) or in a U.K. survey with 76% In the apartments, we find very low nitrogen concentrations, urinal visits (Hills and Birks, 2004). Interestingly, the average leading to an extremely low N:P relation below the physiological volume recorded in the women’s toilet was higher than that of value (Table 2). We have no reason to doubt the results from the men, despite the risk of urine losses in the NoMix toilet. Since the laboratory or from sampling, but it may be caused by nitrogen flushing volume in the front part of the Dubbleten NoMix toilet is losses in the long urine conducting pipes connected to the NoMix extremely small, some women may omit to flush altogether toilets in these households. Since the system is ventilated to 1916 L. Rossi et al. / Journal of Environmental Management 90 (2009) 1909–1917 prevent smell and the pipes are long enough to lead to significant that NoMix sanitary technology still needs improvements for proper hydrolysis of urea (Udert et al., 2003a), there may actually be functioning in a real-life setting. This study thus contributes to a strong flux of nitrogen out of the system. This would equally lead a realistic assessment of the NoMix technology and will help to to a relatively low pH-value in the tank, and therefore to less further improve this promising innovation in urban water manage- precipitation of phosphorus, which would be consistent with the ment. Differences were observed between the physiological urine relatively high phosphorus values. Unfortunately, this inconsis- production and the amount of urine effectively collected by the tency was only detected after the toilets had already been removed NoMix system. This demonstrates that the use of simple counters from the households, so there was no opportunity to test the installed on toilets for estimating the urine collection potential will hypothesis. We suggest that this should be done in other pilot overestimate the effective urine collection by about 30%. projects, especially if they involve long, well ventilated piping. Generally, the data presented in this paper are in accordance with other studies conducted during shorter periods. They repre- 4.4. Frequency of small and large flushes in the households sent useful information for understanding voiding and flushing behavior in households and institutions. Extrapolated to an entire Using a small or a large flush in the households surveyed in this watershed, the flushing pattern and loads correspond well with the study is not much different from using a small or large flush of an typical nitrogen loading of a treatment plant. A global imple- ordinary toilet. Butler (1991) found that for seven households that mentation of urine separation at source in a watershed neverthe- had a toilet with dual-flush facility the average of all small flushes less requires definition of detailed urine storage strategies was 58%, corresponding well with the overall 61% small flushes (collection in private households, at institutions, or public places, found in this study over a longer period (Fig. 7). The results from etc) depending on the characteristics of the watershed. The data apartment 2 show that non-compliance is also possible and that presented in this study represent a good dimensioning basis for either awareness or technical problems may compromise the water future at-source urine control systems. saving effects of a double flush toilet. From personal feedback from the inhabitants of the apartments we know that the toilet paper was not always flushed away in the NoMix toilets with the small References flush. This resulted in double flushes and might have caused people Almeida, M.C., Butler, D., Friedler, E., 1999. At-source domestic wastewater quality. to systematically use the large flush. Urban Water 1, 49–55. The use of small and large flushes at the working place is more Borsuk, M.E., Maurer, M., Lienert, J., Larsen, T.A., 2008. Charting a path for innovative interesting from the point of view of NoMix technology. Due to the toilet technology using multicriteria decision analysis. Environmental Science & Technology 42 (6), 1855–1862. very small flush of the Dubbletten NoMix toilet, the water saving Burgio, K.L., Engel, B.T., Locher, J.L., 1991. Normative patterns of diurnal urination potential of these NoMix toilets is considerably larger than for across 6 age decades. The Journal of Urology 145, 728–731. a normal double flush toilet. Butler, D., 1991. A small scale study of wastewater discharges from domestic appliances. Journal of the Institution of Water and Environmental Management 5 (2), 178–185. 4.5. Extrapolation to a watershed scale Ciba-Geigy, 1977. Scientific Tables Geigy, Volume Body Fluids (Wissenschaftliche Tabellen Geigy, Teilband Ko¨rperflu¨ ssigkeiten), eighth ed., Basel, Switzerland (in German). The use of data logger allowed us to measure urine production CODEAU, 1996. Hydro-Meteorological Data Management Software. User Manual, over a very long period, including holidays and social events. Thus Modules I and II. Swiss Federal Institute of Technology (EPFL), Hydram, Lau- the reliability over time can be considered as very good. However, sanne, Switzerland. Dı`az-Morales, J.F., Sa`nchez-Lo` pez, M.P., 2004. Composite and preferences scales of the results obtained in this study are conducted on a small sample morningness: reliability and factor invariance in adult and university samples. of the entire population of a city. The reliability of the results, in The Spanish Journal of Psychology 7 (2), 93–100. term of population, and the extrapolation at larger scale represents Friedler, E., Brown, D.M., Butler, D., 1996a. A study of WC derived sewer solids. a real challenge. Nevertheless, the surveys by Friedler et al. (1996b) Water Science and Technology 33 (9), 17–24. Friedler, E., Butler, D., Brown, D.M., 1996b. Domestic WC usage patterns. Building that were conducted with a bigger sample of the population (135 and Environment 31 (4), 358–392. households) but during a much shorter period (seven days) ended Hills, S., Birks, R., 2004. Washroom behaviour and users’ perceptions of ‘‘novel’’ with similar results. The differences obtained with our study may water-efficient appliances. Water Science and Technology: Water Supply 4 (3), 13–23. be explained by the fact that a major part of the residents in our Huisman, J.L., Burckhardt, S., Larsen, T.A., Krebs, P., Gujer, W., 2000. Propagation of study work part-time. Moreover, as illustrated in our study, it is waves and dissolved compounds in sewer. Journal of Environmental Engi- necessary to cover a long period of investigation to account for neering–ASCE 126, 12–20. Jo¨nsson, H., Stenstro¨m, T.-A., Svensson, J., Sundin, A., 1997. Source separated urine variability in the measurements. The amount of urine collected -nutrient and heavy metal content, water saving and faecal contamination. with NoMix toilets in this study is also comparable with results Water Science and Technology 35 (9), 145–152. obtained in other socio-economic environments, but during Jo¨nsson, H., Vinnerås, B., Ho¨glund, C., Stenstro¨m, T.-A., 1999. Source separation of urine. Wasser & Boden 51 (11), 21–25. a shorter period of time (Jo¨nsson et al., 1999). Thus the results Krieg, M., Gunsser, K.-J., Steinhagen-Thiessen, E., Becker, H., 1986. Quantitative- presented in Table 1 represent a good set of information for analysis of clinical chemical-parameters in the 24-h urine and in the morning defining several scenarios of urine at-source control at a larger urine – a comparative-study. (Vergleichende quantitative Analytik klinisch- chemischer Kerngro¨ssen im 24-Stunden-Urin und Morgenurin). Journal of scale. This point has been confirmed by the extrapolation of our Clinical Chemistry and Clinical Biochemistry 24 (11), 863–869. in German. results to an entire watershed (Fig. 8). Additionally, differences in Laak, R., 1974. Relative pollution strengths of undiluted waste materials discharged urine concentrations collected in households or at working places on households and the dilution waters used for each. In: Winneberger, J.H.T. (e.g. a hypothesis of stronger dilution and lower urea concentration (Ed.), Manual of Treatment Practices. Arbor Science Publishers, Ann Arbor, MI, USA. than in morning urine) should also be considered in future urine at Larsen, T.A., Gujer, W., 1996. Separate management of anthropogenic nutrient source storage strategies. solutions (human urine). Water Science and Technology 34 (3–4), 87–94. Larsen, T.A., Maurer, M., Udert, K.M., Lienert, J., 2007. Nutrient cycles and resource management: implications for the choice of wastewater treatment technology. 5. Conclusions Water Science and Technology 56 (5), 229–237. Lienert, J., Larsen, T.A., 2006. Considering user attitude in early development of As with any innovation, the benefits and drawbacks should be environmentally friendly technology: a case study of NoMix toilets. Environ- mental Science & Technology 40 (16), 4838–4844. considered carefully before large-scale introduction of a technology Lienert, J., Larsen, T.A., 2007a. Soft paths in wastewater management – the pros and (Lienert and Larsen, 2007a). The results of this survey show clearly cons of urine source separation. GAIA 16 (4), 280–288. L. Rossi et al. / Journal of Environmental Management 90 (2009) 1909–1917 1917

Lienert, J., Larsen, T.A., 2007b. Pilot projects in bathrooms: a new challenge for Rossi, L., Lienert, J., Rauch, W., 2004. At-source control of urine to prevent acute wastewater professionals. Water Practice & Technology 2 (3), doi:10.2166/ wet-weather impacts of ammonia. 5th International Conference on Sustainable wpt.2007.0057 Download: http://www.iwaponline.com/wpt/002/03/default. Techniques and Strategies in Urban Water Management (NOVATECH), May htm. 2004, Lyon, France, vol. 2, pp. 919–926. Maurer, M., Rothenberger, D., Larsen, T.A., 2005. Decentralised wastewater treat- Thiemann K., 2006. Neue Dinge fu¨ r eine nachhaltige Entwicklung. Ansa¨tze fu¨ r einer ment technologies from a national point of view: at what cost are they Kulturpsychologie nachhaltigen Produktdesigns. PhD, University of Bern, competitive? Water Science and Technology: Water Supply 5 (6), 145–154. Switzerland, 280 pp (in German). Maurer, M., Schwegler, P., Larsen, T.A., 2003. Nutrients in urine: energetic aspects of Udert, K.M., Larsen, T.A., Gujer, W., 2003a. Biologically induced precipitation in removal and recovery. Water Science and Technology 48 (1), 37–46. urine-collecting systems. Water Science and Technology: Water Supply 3 (3), Nijman, R.J.M., Butler, R., van Gool, J., et al., 2002. Conservative management of 71–78. urinary incontinence in childhood. In: Abrams, P., Cardozo, L., Hkoury, S., Udert, K.M., Larsen, T.A., Gujer, W., 2003b. Estimating the precipitation potential in Wein, A. (Eds.), Incontinence. Health Publication, Paris, pp. 513–551. urine-collecting systems. Water Research 37 (11), 2667–2677. Pahl-Wostl, C., Scho¨nborn, A., Willi, N., Muncke, J., Larsen, T.A., 2003. Investigating Wilsenach, J.A., van Loosdrecht, M.C.M., 2004. Effects of separate urine collection on consumer attitudes towards the new technology of urine separation. Water advanced nutrient removal processes. Environmental Science & Technology 38, Science and Technology 48 (1), 57–65. 1208–1215. Rauch, W., Brockmann, D., Peters, I., Larsen, T.A., Gujer, W., 2003. Combining urine Wilsenach, J.A., van Loosdrecht, M.C.M., 2006. Integration of processes to treat separation with waste design: an analysis using a stochastic model for urine wastewater and source-separated urine. Journal of Environmental Engineering– production. Water Research 37 (3), 681–689. ASCE 132 (3), 331–334.