Water scarcity is a growing sustainability challenge facing many regions. The recent drafting of a National Water Reuse Action Plan in the U.S. [1
] is evidence of a pressing need to address this challenge, as well as a reminder of work to be done to identify suitable reuse sources, end uses and treatment approaches.
Onsite non-potable water reuse (NPR) is one option to alleviate water scarcity challenges, particularly in large cities [2
]. Onsite NPR systems capture and treat water generated within or surrounding a building, such as mixed wastewater or source-separated graywater, for reuse in toilet flushing, clothes washing and irrigation [3
]. Besides alleviating water scarcity, onsite NPR can reduce the burden on existing drinking water and wastewater treatment systems, reduce building sewer fees, inspire community innovation and foster water system resilience through redundancy and source diversification [4
Demonstration projects across the country are showing that onsite treatment of rainwater, stormwater, graywater and blackwater is technologically achievable and publicly acceptable [4
]. Some cities are even requiring onsite reuse for certain new construction. San Francisco, for example, has an ordinance requiring new commercial, mixed-use or multi-family buildings over 250,000 square feet (23,226 square meters) to include onsite NPR [8
]. Accordingly, a growing body of guidance literature has led to a risk-based framework for public health protection for onsite NPR [9
], including pathogen log reduction targets (LRTs) to inform the selection of treatment configurations that achieve health risk benchmarks [12
]. Still, onsite NPR systems are not widespread and guidance on proper technology selection and best design practices is lacking.
As onsite NPR becomes more common, there is an opportunity to incorporate additional measures of economic and environmental sustainability to inform the adoption of integrated urban water management (IUWM) principles [14
]. A central tenant of IUWM is that potential options must be comprehensively evaluated in terms of economic, social and environmental aspects, requirements echoed in broader discussions of urban water system sustainability [16
]. This facilitates greater transparency in the decision-making process [19
] and helps identify problematic tradeoffs that can lead to negative consequences [21
There are few examples of integrated assessments of the financial, social and environmental aspects of onsite NPR. Schoen et al. [22
] used environmental, cost and quantitative microbial risk assessment (QMRA)-derived risk metrics to compare conventional and alternative water and sanitation systems, including onsite NPR. The alternative systems incorporating NPR had reduced environmental impacts, local human health impacts and cost compared to the conventional, centralized option, but their cost was highly variable compared to onsite sanitation options without NPR. However, only one NPR system was assessed and the technologies evaluated were designed for residential households, not large buildings.
Several studies have evaluated community-scale NPR systems using individual economic, environmental or human health metrics. The influence of treatment system capacity, degree of decentralization and treatment system technology has been evaluated using life cycle assessment (LCA) [23
], LCA and life cycle cost assessment (LCCA) [26
] and QMRA [10
]. Both Cashman et al. [26
] and Kavvada et al. [24
] found that design flow or capacity economies of scale strongly influenced cost and environmental performance of decentralized membrane bioreactors (MBRs), with clear advantages for larger systems. However, they only evaluated larger, community-scale NPR systems, which have different distribution and collection requirements and pathogen risk profiles than single building systems. Hendrickson et al. [23
] used LCA to compare a novel building-scale wetland treatment system for onsite NPR to a centralized conventional wastewater treatment plant and centralized NPR system. Although they found the wetland to be significantly less efficient than the conventional wastewater treatment plant, results showed the onsite wetland had energy consumption advantages when compared to centralized NPR, in line with suggestions that constructed wetlands can be a low-energy reuse option [27
]. Schoen et al. [10
] found that onsite MBR treatment of source-separated graywater or mixed wastewater at the large building scale could meet current human health benchmarks but that additional disinfection barriers would more robustly protect against protozoan pathogen risk, a conclusion echoed by a recent review of membrane treatment performance [30
]. Similarly, the pathogen reduction performance of constructed wetlands is low relative to other biological processes [31
], requiring still greater protection than MBRs.
To develop sound design and implementation guidance that can be widely adopted, it is critical that system economic, environmental and human health aspects be evaluated simultaneously to avoid burden shifting. For example, do the previously identified environmental benefits of decentralized MBRs [25
] translate to single building applications, especially when coupled with more robust disinfection processes necessary for adequate human health protection (e.g., [10
])? Likewise for low-energy treatment wetlands [23
]. How does reduced potable water demand affect net system cost and environmental impacts? How are cost and environmental impacts affected by the unique requirements of additional, in-building collection and distribution piping for onsite NPR systems? When designed to the same standards, which system costs less? To our knowledge, such thorough analyses of onsite NPR systems have yet to be conducted.
To address these questions, we design onsite NPR treatment systems for a large, mixed-use building that meet defined human health risks guidelines [12
] and comprehensively evaluate their health, cost and environmental aspects using QMRA, LCCA and LCA, respectively. Scenarios evaluated include different core biological treatment technologies used to treat either combined wastewater or source-separated graywater to meet the partial, full, or excess supply of building non-potable water demand. We focus on MBRs and treatment wetlands as core biological processes due to their prevalence in existing case studies and their demonstrated robust treatment performance at small scales. Specifically, we evaluate aerobic membrane bioreactors (AeMBRs), anaerobic membrane bioreactors (AnMBRs) and recirculating vertical flow wetlands (RVFWs). AeMBRs are a common, commercially viable treatment option (Hai et al., 2019). AnMBRs were investigated to explore the energy recovery potential of onsite wastewater treatment (Cashman et al., 2018). RVFWs were selected as a lower-energy, natural treatment option that relies on active recirculation to achieve a smaller land requirement than traditional constructed wetlands (Arden and Ma, 2018; Gross et al., 2007).
When reviewed in all metrics, results across human health, cost and environmental impact metrics show that some onsite NPR options perform consistently better than others, while exceptions provide insight into optimal configurations under specific contexts. The AeMBR system tends to perform best compared to the RVFW and AnMBR; however, its performance may be skewed based on its more advanced state of development. Unlike the AeMBR, AnMBR technology at this system size is not widely commercially available and the limited data available for LCI development come primarily from lab- and pilot-scale studies [2
]. Results show that, aside from energy, even the mixed-wastewater AnMBR system with intermittent sparging (i.e., the design variation expected to perform best of all AnMBR variations) is outperformed by the AeMBR system (Figure 5
, Figure 6
). While the AnMBR system is able to produce enough biogas to decrease its energy impact relative to the AeMBR and RVFW systems, this benefit is offset by the costs associated with the additional unit processes required to address its limited ability to remove nutrients—mainly nitrogen—as well as the inclusion of metrics related to, but distinct from, cumulative energy demand such as global warming and particulates. In addition, due to requirements that the system maintain a specific internal temperature, thermal recovery is not suitable as a pre-treatment option, reducing the potential benefit that can be realized from the AnMBR system. The future optimization of AnMBR technology may provide different impact and cost outcomes.
The RVFW systems also do not perform as well as the AeMBR systems in terms of human health protection, cost and most LCA metrics. Although the RVFW has been shown to perform more consistently than other constructed wetland types in terms of organics and pathogen removal [31
], its material and energy costs generally exceed those of the AeMBR systems when additional unit processes needed to meet effluent microbial risk guidelines are incorporated. For example, the RVFW biological process uses only 0.26 kWh/m3
of electricity (Table S13
, Scenario 2) compared to 0.43 kWh/m3
for the AeMBR biological process (Table S9
, Scenario 2). However, the RVFW system uses 0.76 kWh/m3
compared to 0.74 kWh/m3
for the graywater AeMBR system. This finding highlights the challenge for wetlands to maintain their ‘low-energy’ competitive advantage when subject to more rigorous effluent guidelines, as is the case for San Francisco’s Living Machine wetland system which uses more than 2 kWh/m3
(Hendrickson et al., 2015). Conversely, passive wetlands not subject to human health-based effluent guidelines can require less than 0.1 kWh/m3
]. In addition, the variable pathogen reduction performance of the RVFW results in the system’s modeled risk exceeding the health benchmark despite being designed with a sufficiently high LRV.
In terms of source water type, systems treating source-separated graywater outperform those treating mixed wastewater for all metrics except cost (Figure 4
and Figure 6
b). Cost differences are not large, however; the difference between the NPV of graywater and wastewater AeMBRs for the Full Treatment Scenario is less than 20%, while differences in environmental benefits can be far greater. Specifically, graywater versions require less energy for aeration, generate fewer screenings and have lower emissions of methane and nitrous oxide. This is consistent with suggestions that, owing to its lower concentration of organics, pathogens and nutrients [48
], treatment of source-separated graywater may be more efficient that mixed wastewater [17
Incorporation of a thermal recovery unit to offset natural gas use associated with hot water heating demonstrates further tradeoffs. Although the thermal recovery unit reduces global warming and fossil fuel impacts of the AeMBR system with minimal additional cost, it results in much higher acidification and particulates impacts as well as overall higher energy use. These larger impacts result from the additional electricity required to run the heat exchanger (4.1 kWh/m3
, Table S9
) and depend on the emission factors associated with the fuel mix of the San Francisco power grid. These results will vary across the country depending on the local electrical grid mix as well as the type of hot water heater used (i.e., natural gas or electric) and should be explored further.
QMRA results showed that when accounting for variable performance, annual risk of the RVFW wetland treatment systems exceeded the health benchmark. So, although the systems comply with the recommended LRVs, the actual risks may exceed the benchmark some of the time due to variation in treatment performance. To avoid this, the LRVs assigned to processes should be conservatively based on the worst or 5th percentile performance rather than average performance.
There remains outstanding uncertainty that was not included in the results but could change the predicted rankings with additional information. No LRVs were found for the RVFW, only for more rudimentary wetland systems. For the RVFW system to perform better than the MBR system, the RVFW LRVs for viruses and protozoa would need to exceed 3.0 for graywater treatment without ozone or increase to 1 and 2.5, respectively, for wastewater treatment with ozone. In addition, due to a lack of monitoring data in distributed systems, performance data for the MBR and ozone units and the UV failure frequency were primarily derived from centralized municipal treatment systems. If distributed operation and maintenance is less rigorous than for centralized treatment, then the annual health risk from exposure to pathogens in treated non-potable water could increase.
Operational cost data for all systems were based on centralized municipal treatment systems as no single design or operational standard currently exists (Tables S19 and S20
, also see [2
] for additional discussion). Further research is needed to demonstrate what level of monitoring and operational control is necessary to provide consistent, protective treatment performance.
A formal uncertainty assessment was not carried out for either the cost or environmental analysis, and appropriate caution should be used when interpreting the presented results. For example, both the RVFW and AnMBR treatment systems are innovative technologies in their infancy and less widely implemented than AeMBRs, which are commercially available. The novelty of the former contributes both to wide uncertainty around the underlying inventory values and the potential for future improvements. Moreover, economies of scale can have a significant effect on the cost and environmental performance of these systems [25
]. Additional research is needed to explore the effects of building size and occupancy on the environmental and economic cost of these systems.
The underlying inventory data are specific to the San Francisco region, most notably utility fees, electrical grid mix, and centralized wastewater and drinking water treatment infrastructure. While most aspects of building and treatment systems are applicable to other regions, environmental impacts and costs associated with regionally specific factors do have a substantive effect on the absolute magnitude of results. For example, tradeoffs shown for thermal recovery incorporation will vary with grid mix, and differences may be more pronounced if the unit is used to offset an electric hot water heater rather than natural gas. Still, the influence of local conditions is expected to be less pronounced when focusing on comparative performance rather than the overall magnitude of individual metrics.
4.2. Decision Analysis
The objective of this study is to show how integrated metrics can comprehensively characterize onsite NPR options for large buildings, not to make a single recommendation. To be incorporated into a decision-making process, local stakeholder values must be applied to the provided, incommensurable metrics. For example, these results could be used as input to a multiple-criteria decision analysis (MCDA), where each metric is assigned a weight and a final ranking of options is made based on the context-specific combination of weighted metrics [50
]. Cole et al. [20
] provide a useful framework for MCDA in an IUWM context, as illustrated through a stakeholder-driven process to implement a dual water supply system.
This study presents the results of an integrated assessment that examines onsite NPR options for a large mixed-use building. Human health risk, cost and several environmental impact indicators were used to evaluate the effects of treatment system type, source water selection and treatment system capacity system performance. No one option performed best, though several general conclusions can be drawn from those options and approaches that performed consistently well. Although the context of the study was based on an actual building, the relative conclusions are intended to be broadly applicable.
According to health, cost and environmental indicators, AeMBRs tended to perform better than AnMBRs and RVFWs, as the latter are both challenged by the need for additional pre-treatment (RVFW), post-treatment (AnMBR) and disinfection processes (RVFW). Cost was the only indicator by which mixed-wastewater versions of each treatment technology had a comparable advantage over their graywater counterparts due to the cost of additional piping required for source-separated graywater collection. In terms of environmental indicators, graywater versions outperformed mixed-wastewater versions, largely due to lower energy inputs and reduced emissions associated with treating lower-strength wastewater. Thermal recovery from graywater to offset natural gas use associated with onsite hot water heating improves system GWP and FDP, but at the expense of large increases in AP, PMFP and CED. Last, displacement of potable water consumption is a key determinant of total system cost and environmental performance. Systems designed to meet, but not exceed, onsite non-potable demand performed best.