Impacts of onsite greywater reuse on wastewater systems

Reducing domestic water demands by onsite greywater reuse (GWR) for toilet flushing and garden irrigation has the potential to contribute to alleviation of the stress on natural water resources. GWR affects the quality and quantity of domestic wastewater. Studies on GWR to date hardly gave any attention to the effects of these changes on domestic wastewater quantity and quality, and on urban sewer systems and wastewater treatment plants (WWTPs) performance. Therefore, the research presented in this paper aimed at substantiating conclusions on the extent of the effects that onsite GWR systems, if widely implemented, may have on the urban sewer systems and WWTPs. This will better allow the promotion of GWR schemes, which are expected to result in significant water savings and increase the sustainability of urban water use.

Greywater is generally defined as domestic sewage excluding the streams generated by toilets (blackwater) and in many cases also excluding the streams from kitchens (dark greywater) (Friedler 2004). Onsite GWR for toilet flushing and garden irrigation is thought to be an efficient tool for reducing potable water demand (Campisano & Modica 2010; Penn et al. 2012), decreasing the need for exploiting new (and usually costly) water sources, and for enhancing sustainable urban water use. Friedler (2004) has shown that, as the demand for greywater within the urban environment (i.e. for toilet flushing and garden irrigation) is lower than its production, only the less polluted streams from the showers and washbasins (light greywater) come in sufficient quantity to be treated and reused.

With an increase of GWR, wastewater discharges from the houses decrease, with the main reduction occurring during the morning and evening peaks (Penn et al. 2012, and others). In upstream links of the sewer system, where flows are low, the decrease in flow reduces gross solid (GS) movement and deposition may occur. Further downstream, where flows are higher, their movement improves (e.g. Penn et al. 2014; Mattsson et al. 2015; Penn 2015; Spence et al. 2015). However, sediment accumulation and possible resulting blockages in upstream links of the sewer systems are not necessarily an outcome of these effects. Flows in the sewer, especially in the upper parts, are un-steady and non-uniform. A momentary high flow might be sufficient to move stationary solids out of the sewer stretch, and hence may prevent blockages. The effect of these higher pulses of flows on GS movement requires further examination, and is presented in this paper. GWR also affects the wastewater composition. With high reductions in flow and a smaller reduction in pollutant loads (as a result of the needed treatment before its reuse), pollutant concentrations in the wastewater do increase (Penn et al. 2012). This may well affect the performance of the WWTP, an effect which also is further examined and presented in this work.

As a means to examine the effects of GWR on sewer systems and WWTPs, an integrated stochastic simulation system for GWR in urban areas was developed and implemented for a real network (separate sewer system) and WWTP in Israel. An up-to-date user-friendly simulation system, SIMBA# (Alex et al. 2013; Ogurek et al. 2015), served as a platform. The simulation model starts at the single house scale, goes through the sewer network and ends at the WWTP outflow. It includes stochastic generators of domestic wastewater streams and GS, a municipal sewer network model which includes full hydrodynamic flow simulations, a GS transport module, and simulation of biochemical processes occurring within the WWTP. Setup and implementation of the stochastic generator and the GS transport simulator were previously developed and described by Schütze et al. (2014) and Penn (2015). The next sections of the present paper provide a short introduction to each of these modules. A short description of the previously developed modules, and a broader description of their integration with the WWTP and of the implementation of the integrated model for the examination of the effects of GWR on a represented sewer system and WWTP, will be presented in more detail subsequently.

Most of the characterizations and generators for domestic wastewater streams and GS production found in the literature (e.g. Butler et al. 1995; Friedler & Butler 1996; Friedler et al. 1996; Langergraber et al. 2008; Gato-Trinidad et al. 2011; Rodríguez et al. 2013) are deterministic or deterministic models converted to stochastic ones by adding a stochastic component. These do not take into account all the many dependencies among wastewater generation events and quantities within a household. These dependencies are not rigid but elastic and can change with time, person, appliance, etc. As a result, domestic wastewater streams are subject to high variability and fluctuation (Friedler et al. 1996; Shteynberg 2015; Spence et al. 2015, and others). Characterization by averaging, as suggested in some of these models, attenuates the peaks of the momentary upstream flows, which may significantly affect simulation results of modelling GS movement. Further, most of these existing models do not provide a description of outflows from individual houses; thus they do not represent the behaviour in the upper parts of the sewer system, in which the variability is more pronounced (e.g. Blanksby 2006; Butler & Davies 2010) and where higher risk for blockages occurs. Moreover, most of the above-mentioned characterizations and generators do not differentiate between all of the individual domestic wastewater streams, which, however, is needed for a GWR analysis. Blokker et al. (2010) developed a detailed stochastic prediction of water demand patterns from different household appliances, taking into consideration many parameters and their dependencies. However, these require a large amount of statistical information, making it difficult to implement the model for an alternative case study. For this paper, a detailed stochastic generator for individual domestic wastewater streams and GS production, as developed by Penn et al. (2017) in an easy-to-use tool was used.

There has been some earlier work on GS characteristics and transport in sewer systems (Babaeyan-Koopaei et al. 1999; Schütze et al. 2000; Digman et al. 2002; Butler et al. 2003, 2005; Walski et al. 2011, and others). For example Schütze et al. (2000) have confirmed by a regression analysis the results of the detailed model by Babaeyan-Koopaei et al. (1999), establishing that a solid would move if and only if both water velocity and water level exceed a certain threshold. A linear relationship, with two calibration parameters, was found between the GS velocity and the wastewater velocity. Walski et al. (2011) also stated that a minimum flow Q should be exceeded in order for a GS to be moved. Gouda et al. (2001) modelled GS transport in a stochastic way, using Markov chains. However, none of these modelling approaches seem to have found their way in commonly applied simulation programs, nor have any easy-to-use simulators for GS transport been reported. The subsequent sections describe the implementation of a GS simulator in an easy-to-use simulation framework.

For planning and simulation of conventional WWTPs, diurnal patterns of flow and concentrations prove very useful. Successful procedures have been developed for their description by analytical functions, which are useful in particular when measurement data are not available or not sufficient. For example, Langergraber et al. (2008) defined diurnal variations of flow and pollutants in domestic wastewaters by means of a Fourier series with easily determinable coefficients. Several authors derived stochastic extensions of this concept (e.g. Rodríguez et al. 2013), which, however, fall short of the principle of simplicity. In the present work, the wastewater generation is modelled using the stochastic procedure outlined in subsequent paragraphs.

The simulation system that served as a platform for the research was SIMBA#, an integrated simulator for sewer systems, WWTPs and river water quality (Alex et al. 2013; Ogurek et al. 2015). The hydrodynamic modelling modules of SIMBA#, which are based on an extended version of the USEPA SWMM (Storm Water Management Model) (Rossman 2004, USEPA website) were applied for simulating wastewater flows. These yield the full dynamic solution of the Saint Venant differential equations. The stochastic wastewater and GS generator served input data to the hydrodynamic simulations of the flows and to the GS transport simulator which was developed as an add-on module in SIMBA# (Schütze et al. 2014). The quantity and quality of the wastewater flow and the GSs at the outlet of the sewer system served as input to the WWTP model.

For examination of the effects on the WWTP, a dynamic model of Netanyas WWTP has been set up, based on the Activated Sludge Model No. 3 (Figure 2). Based on information obtained from Netanya Water Ltd, the plant has an average dry weather flow rate of 39,000 m³/d. Although the real plant has an equalization tank at its inlet, which also has been included in the model, it is not being used for the simulation of the various scenarios, since most of the time it is not in use.

Overall, the WWTP consists of the following four process steps:

• Mechanical pre-treatment (primary clarifier, equalization tank).

• Biological wastewater treatment with typical plant layout (pre-denitrification).

• Anaerobic digestion of primary sludge.

• Aerobic stabilization of excess sludge (this represents a specific feature of Netanya WWTP. Thickened sludge is, after hydrolysis, further processed in a cascade of anaerobic tank, denitrification and nitrification. By adding air, the sludge gets further stabilized.

Stochastic generation of GS and individual wastewater streams from household appliances served as input to the hydrodynamic simulation of the flows and to the GS transport simulator. The generator was developed while considering the many dependencies between water-use events in a household, and the varying elasticity of these dependencies, both making the characterizing of these discharges by ‘classic’ stochastic distributions almost impossible. Data for the stochastic generator were derived from an end-use water-usage measurement campaign (and the resultant wastewater generation) (Shteynberg 2015). Generation of these streams was carried out by sampling with replacement from the observed data: in our context, sampling daily diurnal wastewater discharge patterns from domestic appliances (kitchen sink, showers, washbasin, bathtub, dishwasher, washing machine and toilet) of different households, while keeping the observed domestic and neighbourhood structures. The generated patterns were validated with flow measurements of a real sewer system located near the case study under consideration. Both areas are characterized by similar socio-economic characteristic and household size distribution (Penn et al. 2017). Diurnal patterns for 7 days of simulation were generated for non-GWR households. Diurnal patterns for GWR households were derived from these patterns.

Further details on the measurement campaign can be found in Shteynberg (2015) and on the stochastic generator and its validation in Penn et al. (2017).

The GS transport module employed here takes a microscopic approach, i.e. it tracks the movement of each individual GS along its way through the sewer system.

For the validation of the model, field experiments in a real sewer segment were conducted, where GS movement was traced by using RFID (radio frequency identification) and light sticks (tracked by a web-cam placed in a downstream manhole) combined with flow characteristics in the sewer segment (discharge, velocity and depth of flow). High compatibility between the simulation results and the actual measurements was obtained, showing a linear correlation with a high R² value of 0.85 (further description can be found in Penn (2015)).

For simplification of the modelling process, households in the modelled sewer network were grouped into 34 clusters, each discharging its wastewater and GSs to a different junction in the sewer system. Within the GS transport module, sewer segments were further aggregated to one pipe between each of these discharge junctions. A change in diameter or extreme change in slope disassembled the aggregated pipes into smaller aggregated sections (each having the same diameter and similar slope).

For the influent to the WWTP, GS and appliance discharge loads are merged. For the pollutant loads of the discharges from each household appliance, values as cited by Penn et al. (2012), these being based on Friedler et al. (1996), Almeida et al. (1999) and Friedler (2004), are used. For converting each individual GS to its chemical content the following pollutant loads contained in faecal GS (g/(cap·d)) have been used: chemical oxygen demand (COD) 31, N 1.5, P 0.6 and mass of 123 (Lange & Otterpohl 2000; Friedler et al. 2013).

The sewer system under consideration (as described above) represents just a small part of the system feeding into Netanya WWTP, with a dry weather flow of 1,364 m³/d, as compared with 39,000 m³/d arriving at the WWTP. In order to avoid having to model all of Netanya's sewer system, the outflow derived from the sewer system simulations was multiplied by a corresponding factor and attenuated by means of a Nash cascade (to allow for the consideration of flow attenuation within the system over a distance of about 10 km) and up, scaled in order to emulate the superposition with the inflows of the other subcatchments of the Netanya drainage system.

Output from the sewer system, greywater and GS simulation modules was then fed as influent to the WWTP model and allowed analysis of impacts of GWR on wastewater treatment For the activated sludge system, default parameters for domestic wastewater were used (Alex et al. 2015). The model performance was observed to be comparable to the real plant.

Three types of wastewater streams of GWR houses were considered:

• Type 1 (No GWR) – no GWR is practised and the combined domestic wastewater streams (incorporating discharges from all domestic sources) are discharged into the sewage system.

• Type 2 (GWR WC) – greywater generated from the bath, shower and wash basin (i.e. light greywater) is treated and used for toilet flushing. Excess light GW that is not used for toilet flushing is discharged to the sewage system as overflow, prior to treatment (as raw greywater).

• Type 3 (GWR WC IRR) – the same as type 2, but the excess greywater flow (after treatment) is used for irrigation.

The treatment applied in houses of type 2 and 3 was biological treatment by an RBC (rotating biological contactor) based system. Quality of treated GW was taken from data obtained from an experimental pilot-scale RBC based unit, situated at a married couples’ dormitory in the Technion – Israel Institute of Technology, which was operated for more than 3 years (Friedler et al. 2005; Aizenchtadt et al. 2009). These were considered as the baseline quality characteristics for toilet flushing and for toilet flushing and garden irrigation in the second and third scenarios, respectively.

Three GWR scenarios were modelled, simulated and analysed. In each scenario, all houses implemented the same type of GWR: scenario (1), the current situation in Israel, where none of the houses reuse greywater; scenarios (2) and (3), extreme implementation of GWR where all houses implement type 2 or 3 GWR (as described above), respectively. These scenarios are not likely to fully occur, but as they result in the highest water saving they can potentially be the most problematic when the movement of GSs is concerned.

Simulation results suggest that, with the increase in GWR, the momentary flows are reduced; thus reducing velocities and water depths (e.g. Penn et al. 2013) affect the movement of GSs.

Analysis of the simulation results of the WWTP for the GWR scenarios (GWR WC, GWR WC IRR) indicates that great care has to be taken with regard to nitrogen when applying GWR. Without additional measures, WWTP effluent tended to have too high nitrate concentrations. This can be attributed to unsatisfactory COD:N ratio in the WWTP influent: stationary GSs which accumulate in the sewer system and GW treatment before reuse remove part of COD, but not of N. Therefore, the influent COD loads do not correspond anymore to the incoming N loads (see also Table 1).

Therefore, additional measures are to be taken. These could include:

• addition of carbon source;

• application of processes for improved N treatment (e.g. anaerobic ammonium oxidation (Anammox));

• urine separation (to take out part of the N from the greywater).

As a means for the examination of the effects of various GWR scenarios on sewer systems and WWTPs, an integrated stochastic simulation system was developed and validated. Simulation of the model on a real case study showed that in extreme situations, as in upstream links of the sewer system, where flows are low, with high implementation of GWR, GSs may accumulate in the sewer pipes, and the risk of blockages may increase. With up to 40% savings in water consumption, when greywater is being reused for toilet flushing and garden irrigation, we strongly promote its implementation, especially in households not located in the most upstream links of the sewer systems. For households located in the most upstream links, milder forms of reuse should be considered (e.g. reuse only for toilet flushing); alternatively these sewer pipes can be constructed with smaller diameters, and enlargement of existing systems can be postponed.

Analysis of the simulation results of the WWTP for the GWR scenarios (GWR WC, GWR WC IRR) indicates that a WWTP might face challenges regarding the nitrification–denitrification processes; hence, too high nitrate concentrations in the WWTP effluent may occur. Therefore, additional measures such as addition of carbon source, application of processes for improved N treatment (e.g. Anammox) and urine separation (to take out part of the N from the greywater) appear to be advisable.

Subsequent research steps may include the examination of the effects of reduced flows on the transport of additional types of solids and the effect of disintegrating/dissolving solids (large solids being divided to smaller solids and later possibly being disintegrated/dissolved completely), increasing the concentrations of the liquid phase of the wastewater stream and, finally, impacting the influent concentrations of the WWTP. The movement of accumulated and aggregated solids may be different from the movement of the individual, the original, solids (as suggested for example by Littlewood & Butler (2003)). Furthermore, these larger solids are expected to affect sewer flow. Moreover, integrating stochasticity to the models describing the solids movement may constitute subject for further research.

An outcome of this research is a stochastic, quantitative and qualitative complex modelling of the entire sewer system under evolving future alternatives for reduction in water consumption. Currently this information is limited in the scientific and engineering literature. The integrated model can potentially serve as a tool for analysing any new sewer system or existing sewer systems subject to changes in flows, as with the result of GWR just as presented in this work or other water saving measures. The model can be simulated for predicting ‘hot spots’ for GS accumulation and consequently can serve as means to support prevention of blockages in these systems. The intelligent, and relatively easy to implement, method to stochastically characterize diurnal patterns of domestic wastewater streams can be adapted to additional case studies and data sets. Further, if done by intelligent considerations, the model can also be adopted in the broader scope of many other research projects related to sewer systems such as optimization of sewer system performances, rehabilitation of sewers, fate of pollutants in the surrounding environments (e.g. water sources), etc.

The research results are further expected to increase water systems efficiency and sustainability by assisting in long term planning of introducing water saving measures to the urban sector and in designing and managing the urban wastewater systems that are influenced by these systems.

The authors gratefully acknowledge the financial support of the German Federal Ministry of Education and Research (BMBF) and the Ministry of Science and Technology (MOST) of the State of Israel (Grant reference 02WA1263).