Modelling the benefits of urine source separation scenarios on wastewater treatment plants within an urban water basin

Nowadays, cities stakeholders and utilities have to face multiple challenges related to wastewater treatment. Limiting nutrients discharge is still a priority for protecting large river basins from eutrophication, while energy saving and nutrients recovery are increasingly needed. According to the Water Framework Directive of the European Union, Paris utilities invest tremendous effort to obtain low nitrogen concentrations in the effluents rejected from treated wastewater emerging from the city, to insure the good surface water status of the Seine River. Yet, optimization of nitrogen removal in intensive nitrifying and denitrifying biological active filters is still a complex challenge, while the high loading rate and variation of carbon source affect both nitrite accumulation (Rocher et al. 2015) and N₂O emissions (Bollon et al. 2016). A regular retrofitting of existing wastewater treatment plants (WWTP) could help tackle the problem, but alternative approaches such as urine source separation (USS) is also considered (Guest et al. 2009; Larsen et al. 2009). Despite the theoretical advantages of source separation, stakeholders and utilities need planning tools and scenarios assessment for successful implementation of USS technology, especially in a large city such as Paris with various urban typologies and densities.

Urine separation gained wide merit as it could simultaneously reduce nutrient loads and help in recovering valuable nutrients from the separated waste streams (Larsen et al. 2013; Jimenez et al. 2015; Flanagan & Randall 2018; Volpin et al. 2018; Spuhler et al. 2020). In the last decade, innovative technologies for nutrients recovery from urine were investigated and are still under development (Larsen et al. 2021). Nitrogen and phosphorus were recovered using electrodialysis (De Paepe et al. 2018; Volpin et al. 2020), forward osmosis (Volpin et al. 2019; Courtney & Randall 2022) or bio-electrochemical systems (Zamora et al. 2017; De Paepe et al. 2020; Liu et al. 2022). Nitrification-denitrification was also studied (Udert & Wächter 2012), as well as magnesium dosage to achieve struvite precipitation from source separated urine (Hug & Udert 2013; Udert et al. 2016). The most mature solution is based on combined partial nitrification and distillation which allows to convert urine into a nutrient rich solution that can be used as a fertilizer (Fumasoli et al. 2016). Otherwise, solutions that are still under development consist of two-step processes including struvite precipitation and ammonia recovery as ammonium sulfate using microbial electrolysis cell (MEC) (Igos et al. 2017) or transmembrane chemisorption (TMCS) (Pradhan et al. 2019; Gonzalez-Salgado et al. 2020; Besson et al. 2021). These technologies offered promising results, however their implementation was, in the majority, limited to lab-scale and pilot-scale.

Urine source separation should be further considered for its potential consequences at the wastewater treatment plant. Considering all the constraints and objectives to be considered for WWTP management, reducing overall operating costs (chemicals and energy) is a complex challenge for operators (Azimi & Rocher 2017). The most advanced approaches focused on multiple control strategies to reduce specifically the aeration costs (Varhelyi et al. 2018), the methanol addition as an external carbon source for denitrification, or iron dosage (Kazadi Mbamba et al. 2019). Urine contributes almost 80% of the nitrogen and 50% of the phosphorus to the total nutrients load in the influent wastewater (Larsen & Gujer 1996). Consequently, adopting a USS based approach could diminish and modify the pattern of nutrients load that arrives at the WWTP, helping operators to save energy for aeration and chemical consumption for nitrogen and phosphorus removal (Ishii & Boyer 2015; Jimenez et al. 2015). The potential benefits of urine source separation (USS) was globally assessed by life cycle analysis and costs analysis (Bisinella de Faria et al. 2015; Igos et al. 2017), using average data in the context of European countries. However recent work has also demonstrated the importance of urbanism typology in the assessment (Besson 2020; Besson et al. 2021), making it crucial to consider local information for each case-study.

For benchmarking source separation scenarios, a dynamic influent generator (Gernaey et al. 2011) was recently adapted for modelling the modification of wastewater composition depending on urine separation strategies (Bisinella de Faria et al. 2020). Previously, simulation studies confirmed the importance of modelling tools to assess multilevel urine separation (Rauch et al. 2003; Rossi et al. 2009). A stochastic model was used to predict urine production and to provide detailed information about the dynamics at a local level while implementing a source separation scenario, and concluded an almost 30% reduction in the ammonia peak loads during dry weather periods (Rauch et al. 2003). In addition, researchers performed the life cycle assessment of urine source separation, and their findings helped in the design of future applications of urine source separation through the extrapolation of their analysis to an entire watershed (Rossi et al. 2009; Ishii & Boyer 2015; Badeti et al. 2021; Hilton et al. 2021). With the recent adaptation of the influent generator, researchers proposed a more deterministic tool for investigating and implementing various source separation scenarios at the city scale (Bisinella de Faria et al. 2020; Besson et al. 2021). The implementation of urine source separation needs large case studies that could identify all of the pros and cons that may emerge from urine source separation. If the advantages of such a strategy seem numerous, they have to be compared to the constraints of collection, transport and treatment of the urine collected separately (Boyer & Saetta 2019). Therefore, it is essential to use simulation approaches for assessing real case studies, which would help in defining and building the appropriate scenarios for urine source separation.

Previous studies in the literature focused on urine source separation at large scale without considering the location and the potential use of buildings, despite the fact that it is less costly to implement urine separation in offices using a reduced number of toilets for a large volume of urine collected. Moreover, no study has focused on the effect of area of collection and the length of the sewer on the influent pattern at the wastewater treatment plant.

The aim of this study was to assess the urine source separation scenario in a large urban watershed (Marne Aval, MAV, eastern Paris) considering different implementation strategies. A modelling tool was calibrated considering sewer characteristics, number of inhabitants and typology. Different implementations are compared either in households or in working buildings. Modification of dynamic wastewater patterns are calculated and benefits are concluded in terms of aeration costs and chemical consumption at the wastewater treatment plant. Finally, global estimations of greenhouse gas emissions and energy consumption are determined and discussed.

The Marne Aval (MAV) urban catchment, situated in the eastern part of Paris, generates the wastewater of 210,123 population equivalents (PE). This area is divided into 16 subareas and 62% of them have a separate rainwater collection. The working population corresponds to around a quarter of the PE (offices, working buildings) whereas three-quarters of the total PE are assumed to be households (house and residential buildings). The wastewater is transported and treated by the SIAAP (Greater Paris Sanitation Authority) at the Marne Aval Wastewater Treatment plant (MAV Plant). In the current study, the active working population was assumed to be equally divided between men and women. For simulating urine source separation, two typologies of implementation were compared: a homogeneous urine separation deployed in the entire city at various levels (both households and office buildings), and a more selective option with urine separation installed in working areas and office buildings (Table 1). For the first option, different levels of deployment were calculated (20, 40, 60%) for estimating the proportionality of the benefits. The intermediate level (40%) was used in this paper and named ‘scenario 1’ for comparison with source separation at the work place.

Two specific scenarios (named scenario 2 and 3) were considered for urine separation at work places and during work hours only (between 8:00 a.m. and 5:00 p.m.). Scenario 2 simulated the collection of 90% of men urine at work places using urinals (men workers estimated at 26,265 PE), assuming that during work hours toilets are mainly used for urination. This scenario focuses on male urine because of the greater availability of dry male urinals on the market, although female urinals also exist. The scenario is still valid for a mix of male and female urinals. Finally, scenario 3 corresponds to 90% urine source separation of all employees (men and women estimated at 52,530 PE), using diverting toilets, which are assumed to be very efficient in source separation (Table 1). Dry urinals were considered without any flush volume, whereas reduced volumes of water were considered for diverting toilets (Bisinella de Faria et al. 2020).

This study was conducted using the phenomenological influent generator (Gernaey et al. 2011) that was specifically modified to include urine source separation (Bisinella de Faria et al. 2020), and modified to consider the production from the work place. The tool is programmed on Matlab and is a flexible simulation tool that can be easily modified. The code is available on request from the author. In this study, the results were selected to represent typical weeks during dry weather periods.

Four main streams were defined in the influent generator (different blocks): 1) the Total Wastewater stream (TWW) that includes wastewater from households, industries and other contributors without any urine separation; 2) the Total Urine stream (TU) that simulates profiles of all produced urine; 3) the Urine Stored and Collected stream (USC) that corresponds to the user-specific urine source separation and collection; and 4) the Wastewater stream (WW) that represents the influent entering the treatment plant and represents the combined contribution from households, industries, rainfall, and infiltration (Bisinella de Faria et al. 2020). In addition, the influent generator was modified to create the composition of urine that would be collected in different buildings, such as residential buildings, houses or even office buildings. Regarding the simulations of different scenarios of USS, the user can define the general parameters according to the supposed scenario and accordingly defines the percentage of urine retention and the size of the catchment area. Also, it is possible to modify the flush water volumes for urine source-separated toilet (New flush water), to simulate USS using men urinals or low flush toilets (Bisinella de Faria et al. 2020).

Dynamic flow patterns were generated for daily profiles, weekly profiles and yearly profiles. For the influent wastewater entering the treatment plant, two major peaks were distinguished, around 8:00 a.m. and 6:00 p.m. (Langergraber et al. 2008). In the current study and during several set of simulations, the average total wastewater flowrate (TWW) from households was considered to be 209 L.PE⁻¹.d⁻¹, while the average urine flowrate was equal to 1.36 L.PE⁻¹.d⁻¹ (Bisinella de Faria et al. 2020).

For the set-up of the influent generator, several hydraulic parameters needed to be calibrated specifically regarding the size and characteristics of the catchment area. Indeed, the wastewater residence time in the sewer increases with the surface of the catchment area, and leads to delaying and buffering of the diurnal peaks. In this work, the hydraulic parameters were adapted in order to obtain a sufficiently good prediction of influent flowrates profile measured at the MAV Plant. The influent generator describes the sewer network by a number of interconnected pipes and reservoir tanks that correspond to a number of ‘subareas’. A subarea is defined by a surface A (m²) corresponding to the surface of each aforementioned collection tank (Gernaey et al. 2011). Consequently, the number of subareas and their surface are two specific parameters that needed to be identified for each catchment area depending on its size and properties. A series of simulations were conducted successively, to identify the appropriate number of subarea and the tank surfaces A (m²) that best fit the measured wastewater profile entering to the treatment plant. Statistical analysis based on the normalized root-mean-square deviation (NRMDS) and R² methods were used to compare the simulated profiles and the experimental one (Fig S2). A number of subareas of (seven) and a tank surface of 2000 m² corresponded to the optimal set of parameters (Tables S1 and Table S2).

The modification of nitrogen and COD influent loads were used to estimate the reduction in aeration needs (nitrification) and methanol consumption (denitrification). Reference data were obtained from the plant operators of MAV: extensive measurements of influent ammonium, wastewater flow rates, methanol dosing and aeration flow rate were recorded with high frequency. The wastewater treatment plant is equipped with two stages of biofilters to achieve nitrification and denitrification of the influent wastewater. The first stage of biofilters (DN/N with internal recirculation of 100% of the influent flowrate) achieves 90% nitrification. The second stage of biofilters (DN with external carbon source) achieves the post-denitrification to reach 4.5 mgNO₃-N/L (effluent setpoint) with an average methanol dosage of 3.43 kg MeOH/kg N-NO₃ removed. These ratios were used to calculate the potential savings at the treatment plant for each scenario of urine separation. The oxygen demand for nitrification and COD removal were calculated based on the concentration in the rejected effluent at 45 mg COD/L, 3.5 mg NH₄-N/L and 4.5 mg NO₃-N/L. The details of calculations are available in section 3 of the Supplementary Information.

In order to investigate the sensitivity to the size of the watershed, a different watershed simulation was performed using three subareas and a tank surface of 2000 m². The number of PE treated was reduced by a factor 3/7. This factor is equal to the reduction ratio between the new number of subareas (three) compared to the default one (seven). The scenario without source separation and scenario 1 and 3 were simulated. The results in terms of ammonium mass flow entering the WWTP, the aeration need and the methanol use are discussed hereafter. It should be noted that the variation of subareas etc. only influences the peaks and not the averages.

The overall impact of each urine diversion scenario on global warming potential was calculated. For this purpose, the effects on the treatment plant operation (reduction of energy consumption, chemicals consumption, direct emission of greenhouse gases) was translated into global warming potential thanks to the Ecoinvent database (Wernet et al. 2016) (see Table S4 in the Supplementary Information for detailed emissions factors). The following aspects were considered:

• – The avoided energy consumption for aeration of nitrifying biofilters was calculated with reference to the current air consumption in the plant (333 Nm³/kg NH₄-N removed) and the oxygen demand calculated for the reference. French electricity mix was used (0.109 kgCO₂-eq/kWh) from Ecoinvent database (see Supplementary Information section 3 for detail of the calculation).

• – The avoided emission of nitrous oxide from nitrifying biofilter into air equivalent to 0.0245 kgN-N₂O/kgN_nitrified (Bollon et al. 2016) (in average for winter and summer campaign in another Parisian WWTP). The emissions of nitrous oxide dissolved in the effluent are not considered in this study.

• – Urine transport with weekly truck collection and 10 km of distance (round trip).

• – The direct and indirect GHG emissions for three alternative technologies for urine treatment: (1) the VUNA process which consists of a nitrification of urine followed by distillation (Fumasoli et al. 2016), (2) the ValuefromUrine technology with struvite precipitation and ammonia recovery by microbial electrolysis cell (MEC) (Igos et al. 2017), and (3) the struvite precipitation followed by the nitrogen recovery by transmembrane chemisorption (Besson et al. 2021). Nitrous oxide emissions from partial nitrification in the VUNA process were also considered with the preliminary result of 0.007 kgN₂O/kg-Nurine (Faust et al. 2021)_, and were integrated in the urine treatment impact. The emissions during storage (estimated at around 0.008 kgN₂O/kg-Nurine) was assumed to be prevented by the storage strategy proposed by (Faust et al. subm.). In this strategy, nitrified urine is fed directly on the granular activated carbon filter and not stored beforehand. For the three processes, only the GHG emissions for the treatment are considered (energy consumption, chemicals…) and not the use of the products.

• – The avoided fertilizers production replaced by the use of products from urine was also considered with ammonium nitrate and triple superphosphate substitution.

Several other aspects were not considered: the transport of products before fields application, the emissions into air of urine-based fertilizers and conventional fertilizer; the change of pumping energy consumption in the WWTP. Finally, as the AluFer (aluminium and ferric chloride) addition is performed for removal of the suspended solid in the preliminary clarifier and not for phosphorus removal, no reduction of its consumption were considered.

In order to generalize the results to other case studies, a sensitivity analysis was performed on two parameters: (1) the N₂O emissions factors in the WWTP were modified for the one measured in activated sludge (0.0058 kgN₂O-N/kgN_in (Bollon et al. 2013)), and (2) the European electricity mix (0.38659 kgCO₂-eq/kWh) where less nuclear energy is used and the production of electricity generates more GHG emissions.

Moreover, urine separation has a significant effect on the influent COD:N ratio. As shown (Figure 2(b) and Table 2), this ratio changed dynamically and on average increased from 8.1 (without USS) to 10.3 (with 40% USS). This change in the COD:N ratio was considered in the calculation of denitrification and nitrification demands at the treatment plant (recirculation, air consumption, methanol) in the estimations of benefits gained from implementing a USS scenario.

Another question is the sensitivity to the size of the watershed and the level of centralization of the treatment plant. To evaluate this sensibility, a simulation was performed in a smaller watershed. Simulations (Fig S4) showed the MAV watershed presented two ammonium load peaks that had approximately the same intensity, however the smallest watershed had a more intense morning peak and a smaller evening peak. With Scenario 1, the urine source separation allowed reducing the first peak to the same level as the second peak from the reference scenario for the small watershed (Figure S4), hence the treatment plant can be designed with a treatment capacity for the second peak and not the morning one. For instance, the air supply in the small watershed without USS should be designed for the peak demand of 34,000 Nm³/h, whereas the average demand is 15,000 Nm³/h. With urine source separation implemented, the air compressors would need to produce 27,500 Nm³/h to ensure appropriate treatment.

For this calculation, it was assumed that the internal recirculation rate was not modified between aerobic and anoxic biofilters, leading to the same proportion of nitrate that was recirculated to the pre-DN. Due to higher COD:N ratio, more residual COD was degraded aerobically in the biofilter, but this was not detrimental for the overall balance. However, another option would be to increase the internal recirculation to consume more nitrate and COD in pre-DN. The methanol saving would be even more important in that case, but the energy demand for recirculation would increase and should be considered for global evaluation.

It has to be noted that, even with a European electricity mix (0.38659 kgCO₂-eq/kWh) and lower emissions of N₂O measured in activated sludge (0.0058 kgN₂O-N/kgN_in) (Bollon et al. 2013), the urine separation scenarios still presented a beneficial balance for the MEC and TMCS treatment technologies (see Table S5). The avoided production of nitrogen fertiliser is in this case the main contributor to this improvement. In the case of the VUNA process, the avoided emissions do not compensate the emissions from the treatment. The emissions of nitrous oxide in the partial nitrification of VUNA process is slightly larger than the one in the activated sludge process, and the emissions from the avoided fertiliser production is not sufficient to compensate the treatment. Thus, this scenario is more relevant for intensive treatment plant where high nitrous oxide emissions occurr.

Reduction in nitrogen loads is by default proportional to the level of implementation of urine source separation. Homogeneous implementation in the large urban area of MAV leads to a stable and significant reduction of nitrogen load during the day (from 29 to 31% for 40% urine source separation). In contrast, the specific deployment of USS in office buildings and work areas lead to selective reductions of daily nitrogen peaks. This strategy could potentially help manage the nitrogen peak load at wastewater treatment plants, which is a common problem that most plant operators encounter (Rauch et al. 2003; Rossi et al. 2009). However, considering the contribution of working area in MAV, the effect remained relatively limited, as only a maximum of 9% reduction was achieved (scenario 3).

Finally, the dynamics of nitrogen loads depend on the catchment size with different intensity of the morning and evening peaks. In small catchment areas, the level of nitrogen load reduction due to USS is slightly more variable than for a larger watershed. For scenario 3, in MAV watershed the reduction load had a standard deviation of 2.3% against 4.3% in the smaller watershed. Finally, even if the benefits of USS implementation in work areas (scenario 2 and 3) would be more interesting in a small watershed, this scenario had a relatively low impact on the shaving morning peak loads.

Aeration and methanol requirement both decrease in proportion with nitrogen load reduction, i.e. with the increased level of USS implementation. In addition, for the MAV plant, reduction of methanol dosage was more prominent compared to reduction in aeration, firstly because part of the denitrification is performed in the pre-DN biofilter with the influent COD, and secondly because COD:N ratio increased with USS implementation. As a result, to reach the same effluent ammonium concentration, a lower proportion of denitrification would be performed in post-DN with methanol. This phenomenon would be accentuated if the internal recirculation was increased (not considered here), but this would also require more energy to operate the recirculation pumps.

In the case of pre-denitrification, the decrease of nitrate removal leads to a reduction of COD consumption, which would then need to be oxidized in the aerated biofilter. However, COD is also present in urine, thus source separation of urine allows a reduction of COD mass flow and more specifically soluble readily biodegradable COD. Consequently the increase of COD:N is also accompanied by an increase of particulate COD fraction. Such particulate COD is efficiently captured by primary sedimentation and does not reach the biofilters in the MAV plant, and finally less soluble COD would feed the biofilters. Thus, the aeration consumption for COD removal increased slightly with source separation (by 9, 1 and 2% respectively for scenarios 1, 2 and 3), but the calculation of the total aeration demand still showed a decrease by 17% for scenario 1, and by 2 and 4% respectively for scenario 2 and 3.

Despite important reductions that could be accomplished in energy and methanol in wastewater treatment, these savings should be compared to additional requirements for urine management, such as treatment and disposal.

In this study, it was assumed that the separated urine was stored at the building level, collected by trucks and treated at a centralized place (Bisinella de Faria et al. 2015, 2020) with three different technologies. Previous researchers used a stochastic model for urine production to investigate the feasibility of combining urine separation with waste sewer design (Rauch et al. 2003). In a parallel study, we also evaluated the synergies, which can be found by a combined treatment of urine with rejected liquor from sludge treatment (Besson et al. 2021). Another approach focused on collecting the urine in separate tanks, which were equipped with valves that could be controlled remotely. This approach allowed a specific urine release in the sewer network at night to compensate for nitrogen peak loads in the influent wastewater during the morning peak (Rossi et al. 2009).

By comparing three different treatment technologies, struvite and TMCS technology presented the lowest greenhouse gas emissions but the use of heat recovered from sludge incineration is beneficial to the technology.

Finally, the carbon footprint balance can be highly positive when all the management chain was considered (Figure 6 and Table S4). It should be noted that only the operational impacts are considered in this case. For considering infrastructure, both the possibilities of limiting the extension of existing plant (benefit) and the construction of new technologies (impacts) could be considered in future work, but this is still highly delicate as no full scale feed-back is available for such realization (Bisinella de Faria et al. 2015; Igos et al. 2017).

Our previous work showed that 50% urine retention also reduced total phosphorus by 19% (Bisinella de Faria et al. 2020). Considering such a reduction, it could also be possible to determine other benefits, like for instance reduction of the coagulant demand. However, this was not considered in this case-study for MAV because mineral coagulant dosage is mainly controlled by physical-chemical capture of suspended solids in primary treatment in the MAV plant.

Implementation of urine source separation at a significant level in a watershed needs to be conducted with an extensive planning for such alternative wastewater management. This is essential for anticipating and financing urine storage, collection and disposal, such as installing toilets and tanks in households, residential buildings or public places. Despite important reductions in energy, chemicals, and carbon footprint that were calculated for global implementation of USS in the entire catchment of MAV (scenario 1), the feasibility of this approach remains challenging.

First from a technical aspect, installing diverting toilets and dry urinals in existing buildings is more difficult than installing these techniques in new buildings under construction (Larsen & Gujer 1996; Larsen et al. 2001). The option of installing urine separation preferentially in new buildings should be considered. A driver for urine source separation would be the possibility to avoid a potential expansion of the wastewater treatment plant, thus retrofitting it into a water resource recovery facility (WRRF). Urine source separation could achieve shavings of peak loads without impacting significantly the wastewater flowrate that reaches the WWTP.

From a logistical perspective, it may be more feasible to implement urine separation in office buildings compared to separated residential houses. On average, one toilet is installed in a single house in which 2.2 people live (INSEE et al. 2019) whereas the minimal number of toilets in offices is only one toilet for 10 people (Article R4228-10 n.d.). The option of modifying the toilets in offices and public service buildings seems more realistic, but as demonstrated in this study the effect would be less significant and principally positive for shaving the nitrogen peaks in a small watershed.

Finally, from an economical point of view, the cost of toilet modifications, storage systems, urine transport and treatment installation should also be considered to give a complete vision of the scenarios. Here, it was assumed that treatment of separated urine was realized at a centralized level, which could limit the cost of infrastructures compared to small decentralized installations. Finally, the present work allowed quantifying for a real case-study the environmental outcomes of urine separation scenarios. These results could be possibly translated into cost reductions and money saving in future studies.

Different scenarios of urine source separation (USS) in a specific urban area (Paris catchment of Marne-Aval treatment plant) were simulated. The results of the study are highlighted as follows:

• (1)

  Urine separation reduces nitrogen peak loads on average, and the level of reduction was proportional to the level of deployment in each scenario (household or workplace). Besides nitrogen load, the COD:N ratio changed dynamically which need to be considered for calculating the modification of aeration and methanol dosage.

• (2)

  The dynamic influent generator can be adapted according to the size of the city, and can predict dynamic changes while confirming substantial benefits from specific urine separation during work hours.

• (3)

  The simulated influent profiles allowed quantifying the reduction in energy demand at the biofilters for nitrification and methanol savings at the post denitrification.

• (4)

  A significant benefit in terms of greenhouse gas emissions is demonstrated with the urine diversion strategies, for different recovery technologies, and this was especially due to the high level of N₂O emissions avoided in nitrifying biofilters and nitrogen fertilizer substitution.

This study was part of the MOCOPEE research program (MOdeling, Control and Optimization of wastewater treatment ProcEssEs, www.mocopee.com) and was supported by the SIAAP (Greater Paris Sanitation Authority). The authors are very grateful to Dr Kai M. Udert from Eawag and ETH for his careful review and guidance during this project.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.