Advancing the Economic and Environmental Sustainability of the NEWgenerator Nonsewered Sanitation System

Achieving safely managed sanitation and resource recovery in areas that are rural, geographically challenged, or experiencing rapidly increasing population density may not be feasible with centralized facilities due to space requirements, site-specific concerns, and high costs of sewer installation. Nonsewered sanitation (NSS) systems have the potential to provide safely managed sanitation and achieve strict wastewater treatment standards. One such NSS treatment technology is the NEWgenerator, which includes an anaerobic membrane bioreactor (AnMBR), nutrient recovery via ion exchange, and electrochlorination. The system has been shown to achieve robust treatment of real waste for over 100 users, but the technology’s relative life cycle sustainability remains unclear. This study characterizes the financial viability and life cycle environmental impacts of the NEWgenerator and prioritizes opportunities to advance system sustainability through targeted improvements and deployment. The costs and greenhouse gas (GHG) emissions of the NEWgenerator (general case) leveraging grid electricity were 0.139 [0.113–0.168] USD cap–1 day–1 and 79.7 [55.0–112.3] kg CO2-equiv cap–1 year–1, respectively. A transition to photovoltaic-generated electricity would increase costs to 0.145 [0.118–0.181] USD cap–1 day–1 but decrease GHG emissions to 56.1 [33.8–86.2] kg CO2-equiv cap–1 year–1. The deployment location analysis demonstrated reduced median costs for deployment in China (−38%), India (−53%), Senegal (−31%), South Africa (−31%), and Uganda (−35%), but at comparable or increased GHG emissions (−2 to +16%). Targeted improvements revealed the relative change in median cost and GHG emissions to be −21 and −3% if loading is doubled (i.e., doubled users per unit), −30 and −12% with additional sludge drying, and +9 and −25% with the addition of a membrane contactor, respectively, with limited benefits (0–5% reductions) from an alternative photovoltaic battery, low-cost housing, or improved frontend operation. This research demonstrates that the NEWgenerator is a low-cost, low-emission NSS treatment technology with the potential for resource recovery to increase access to safe sanitation.

Section S1. QSDsan input assumptions for NEWgenerator simulation Table S1. Input waste characteristics for process simulation of the general case.

Frontend
The frontend unit design consisted on an existing frontend toilet unit in QSDsan, assuming 1 seated toilet and 1 urinal per 25 people (4 seated toilets and 4 urinals for 100 users). In the increase system capacity scenario, the frontend was scaled up accordingly (Section S5). The frontend consisted of housing, seated toilet, urinal, fan, and miscellaneous parts (e.g., pipe, lighting, floor, etc.). The annual O&M cost was calculated as 7.5% of the total capital costs which would consist of replacements, labor, and maintenance. Direct methane and N2O were from the collection of waste were calculated accordingly (Section S6).

Foundation
The foundation unit was based off of the concrete foundation pad design used to support the NEWgenerator in the India and South Africa field trials. A foundation area of 4.8m 2 and thickness of 0.1143 m (4.5 in) was implemented. It was assumed to have the same duration as the 25-year system lifetime and would not require any O&M.

Pretreatment
The pretreatment unit was based off of the pretreatment design in the South Africa field trial. It was consisted of the bar screen, pretreatment tank, piping, and feed pump. All components except the feed pump were assumed to have the same duration as the 25-year system lifetime, where feed pump had a 6-year lifetime according to the BOM. The O&M requirements consisted of pump replacements and labor associated with pump replacement and bar screen cleaning.

Sludge Pasteurization
The sludge pasteurization unit was designed to utilize the biogas and sludge produced from the AnMBR unit as a fuel source to treat the sludge on-site according to ISO 30500 solids output requirements. A scaled-down hydronic heat exchanger system and pump from an Omni Processor 28 was used to pasteurize the sludge and assumed to service 10 NEWgenerators, therefore factor of 0.1 was applied for economic and environmental analysis of one NEWgenerator. The pasteurization method heats the sludge by biogas combustion at 70 degrees Celcius for 30 minutes to achieve the LRV and maximum concentration requirements. 29 The process was assumed to have 90% biogas combustion efficiency or 10% biogas loss during the where the methane in the lost portion would be directly emitted to the atmosphere. 30 A 10% heat loss from the combustion heat transferred to the sludge was also assumed. 31

S3. Learning curve for scaled production
A learning curve was used to conservatively estimate costs at scaled production. For the NEWgenerator, a production scale of 100,000 units was assumed. Each individual capital cost component or item, including Bill of Materials (BOM) specific items and additional items for the complete system, were assessed on its relevance to scaled serial production. The percentage of capital costs to be scaled from the total captail costs, was determined to be 65%. Indicating that 65% of the all capital cost items can be optimized with a learning curve as they could be massmanufactured at a scaled production. The following generalized leaning curve function (Eq.1) was used to conservatively estimate the capital cost of the 100,000 th unit produced and calculate a conservative estimate of user cost.
CN is the cost of the N th unit using the C1, the first unit cost (which we assume to be the design team provided BOM or estimated cost at a single quantity), L the minimum cost limit, N the number of units, and b the learning curve exponent. assuming a certain rate of learning and efficiency improvement in the production process over time. Therefore, as the number of units increase the learning curve will approach the minimum cost limit asymptotically. A minimum cost limit of 1.5% of the first unit cost was assumed. 17,32 The learning curve exponent represents the rate of decrease in unit cost as additional units are manufactured (Eq. 2).
The learning curve percentage of 92.5% was assumed, which means that as the output doubles, the relative production cost is 92.5%. This means that the second unit costs 92.5% of the first unit, and then the fourth unit costs 92.5% of the second unit, which continues unit 100,00 units. This learning curve percentage of 92.5% for scaled material costs falls within the conservative range (typically 90-95%), 16 with lower values (e.g., 70%) is found to be aggressive. Learning curve percentages vary depending on the industry and type of production.

Section S4. Country-specific analysis
Due to the variety of different contextual economic and environmental factors, the user costs and GHG emissions for the NEWgenerator is dependent on the country of deployment. To capture these location-specific differences in the five countries of interest (China, India, South Africa, Senegal, and Uganda) in our analysis, input parameters were changed to reflect the expected conditions in each country. These country-specific parameters include household electricity price, energy mix GHG, price level ratio, tax rate, maintenance labor wage, vegetal protein, animal protein, caloric intake, food waste ratio, Liquified Petroleum Gas (LPG) price, and Sodium Chloride (NaCl) price. For specific parameters where country-specific data could not be obtained   The Sludge Pasteurization unit is an exception to the above user scaling category method where a linear scaling is utilized in addition to the sludge service assumption. This assumed that this unit will service 10 NEWgenerators, therefore factor of 0.1 was applied for economic and environmental analysis of one NEWgenerator.  Figure S1. The daily user cost and annual user GHG emissions were simulated based on the impact of increasing users (increasing hydraulic throughput and loading rate) at NEWgenerator general case. Two different energy configurations were simulated: photovoltaic (green), and gridtied (blue). The user capacity was simulated from 50 users to 600 users for user cost and emissions. The median, 25 th /75 th , and 5 th /95 th are depicted by the solid line, shaded region, and dashed line, respectively, to represent the range of results from uncertainty analysis.

Section S6. Environmental impact assessment
Life cycle assessment (LCA) was performed to characterize the life cycle GHG emissions of the NSS system across the construction and operational stages. All sources of GHG emissions were normalized to global warming potential (GWP) with a functional unit of kg CO2-eq·cap -1 ·y -1 . The ecoinvent v3.2 database 27 was used to acquire inventory data for all materials and processes and translated to GWP using the U.S. EPA's Tool for the Reduction and Assessment of Chemicals and Other Environmental Impacts (TRACI 2.1 v1.03). 34 The emission sources from direct impacts from excreta, construction, electricity, and O&M impacts were calculated. Construction impacts were calculated for materials and processes of each component from the bill of materials (BOM), vendor websites, or manufacturers. If masses were not available from the BOM, vendors, or manufacturers the mass and materials for components were estimated and calculated using data or specifications available on dimensions, density, etc. Global or rest of world GHG emissions were used from the ecoinvent v3.2 database instead of country-specific for all analyses. Electricity impacts were calculated using the energy demand from the NEWgenerator BOM for the grid-tied configuration and the unit grid-electricity environmental impacts. The country-specific local grid electricity intensity was used for the contextual analysis. The energy produced from the photovoltaic configuration exceeded the NEWgenerator energy demand, therefore no environmental impacts from grid electricity were considered as it was not necessary. O&M impacts from consumables (production of) from the system processes and replacement components were calculated using the NEWgenerator maintenance schedule. Direct GHG emissions from excreta using during collection and treatment (e.g., user interface/frontend, containment, conveyance, pretreatment) were estimated. Direct GHG emissions of biogenic methane and N2O released from bodily waste degradation and dissolved methane were considered in the frontend, effluent, and sludge pasteurization unit. Methane accounts for 28 times greater at 28 and N2O approximately 265 times greater global warming impact than carbon dioxide 35,36 and was considered that way in emissions calculations (Table S9). For the process model, the direct methane and N2O were calculated as a proportion of COD and nitrogen in the input waste stream. Dissolved methane remaining in the effluent after COD degradation in the anaerobic process was assumed to have direct GHG impact.

Section S7. Uncertainty and sensitivity analysis
The NEWgenerator QSDsan simulation involved 150 input parameters which values cannot be specified exactly (e.g., contextual differences, future change, data availability, numerous manufacturers). For each uncertain parameter a distribution (e.g., uniform, triangular) over a range was generated, where the range was defined The coefficient values range from -1 to 1, but for the purpose of this report absolute values of coefficients were used (0 to 1). The stronger the correlation between input parameter and output occurs with a larger the absolute value. The sensitivity analysis for input parameters by category (capital, O&M, direct) is detailed for daily user cost ( Figure S2) and annual GHG emissions ( Figure S3). Figure S2. The Spearman's rank correlation coefficient for total, capital, and O&M (including labor, electricity, consumables, and component replacements) daily user cost for photovoltaic and gridtied energy configurations. The size of the bubble indicates value of the correlation, with the larger bubble indicating higher correlation between the parameter and daily user cost which indicates relative sensitivity that the daily user cost has to the uncertainty of specific input parameter. Figure S3. The Spearman's rank correlation coefficient for total, capital, O&M (including electricity, consumables, and component replacements), and direct annual GHG emissions for photovoltaic and grid-tied energy configurations. The size of the bubble indicates value of the correlation, with the larger bubble indicating higher correlation between the parameter and annual GHG emissions which indicates relative sensitivity that the annual GHG emissions has to the uncertainty of the specific input parameter.

S8. Targeted improvement scenarios
The targeted improvements (described in Scenario 2 through 4 of Methods) were simulated for the photovoltaic configuration NEWgenerator at general case for detailed breakdown ( Figure S4) and differences ( Figure S5). An additional scenario of LPG for sludge pasteurization only (no biogas) was explored, however resulted in negligible differences for user cost and GHG emissions from the baseline.