Main

Wastewater treatment is essential for protecting public health and the environment. However, treatment processes generate greenhouse gases (GHGs) while relying on energy and chemicals whose production also contributes to total emissions1. Globally, the Intergovernmental Panel on Climate Change (IPCC) estimated wastewater treatment emissions to be 0.38 Gt of CO2 equivalent (CO2e) in 2019, comparable to key industries targeted for decarbonization, including the chemical (0.37 Gt CO2e), cement (0.82 Gt CO2e) and iron and steel (1.35 Gt CO2e) industries2. Yet, recent studies found that the IPCC and US Environmental Protection Agency (US EPA) may underestimate methane (CH4) emissions from on-site processes alone by twofold3,4, and nitrous oxide (N2O) emissions factors are often oversimplified and/or inaccurate5. This study aims to improve understanding of wastewater GHG emissions, which are critical for meeting global climate targets as this sector grows with increasing population and expansion of essential sanitary services.

On-site emissions at wastewater treatment plants (WWTPs, also called water resource recovery facilities) are dependent on wastewater characteristics (for example organics and nitrogen), treatment processes, level of treatment and plant size3,5. N2O emissions span multiple orders of magnitude, but appear correlated with treatment objective5,6. Biogas used for on-site power generation can reduce imported energy, but leaking equipment and treated solids handling emits CH4 (ref. 3). In addition, upstream electricity emissions depend on power requirements of treatment processes and regional grid carbon intensity.

Given this complexity, emissions reduction strategies must consider key plant and geospatial characteristics, energy and material inputs, factors not currently accounted for in national emissions inventories7. Internationally, several studies report national wastewater treatment inventories (Supplementary Table 1), including on-site biogenic CH4 and N2O emissions7,8,9,10. However, they accounted for different emissions components inconsistently. For instance, the US EPA did not include upstream energy emissions7, and only one inventory considered on-site, non-combustion CO2 emissions9. Seiple et al. catalogued US facilities as part of their analyses to quantify the energy potential from wastewater sludge but did not estimate GHG emissions or energy consumption11.

Therefore, we developed an approach to inventory wastewater emissions based on facility location and treatment processes, accounting for on-site, upstream and downstream GHG sources. We estimated on-site emissions from treatment processes, on-site and upstream emissions from producing and using energy for facility operation and chemical production, and downstream emissions from off-site disposal of treated solid waste. We used this approach to generate a national emissions inventory of over 15,000 WWTPs across the contiguous USA. To capture uncertainty associated with various emission sources, we estimated potential emissions ranges using Monte Carlo simulations. In addition, we conducted a global sensitivity analysis to identify key parameters influencing emissions, providing guidance for future data collection and decarbonization efforts. Our approach allows policymakers and engineers to analyse trade-offs inherent to different treatment technologies, better informing interventions for reducing wastewater treatment emissions. This method can be adapted as new measurement studies improve accuracy of emissions estimates and associated emissions factors.

For clarity, the following terminology is used in all text and figures: CH4, N2O and fossil origin CO2 emissions are collectively described as ‘process emissions’ and refer to the gases produced on-site during biological wastewater treatment processes, unless otherwise specified. CH4 and N2O produced from biosolids disposal are distinguished as ‘landfill CH4’ and ‘land application N2O’. Emissions associated with natural gas include those from natural gas combusted on-site for heat (via boiler, dryer and incinerator) and chemical production, as well as upstream emissions from natural gas extraction and distribution. Electricity-related emissions are full fuel cycle, accounting for upstream and power plant emissions.

Characterizing US wastewater treatment facilities

Our inventory includes 15,863 WWTPs in the contiguous USA. For each facility, we assigned a treatment train based on publicly available data reported by the US EPA12,13,14,15. We modelled 49 treatment configurations, representing the major combinations of processes used in the USA, each with liquids treatment, solids treatment and additional optional processes (Fig. 1). We refer to each configuration using the alphanumeric codes included in Fig. 1, where letters indicate liquids treatment, numbers indicate solids treatment, and suffixes e and -p represent electricity generation and chemical phosphorous treatment, respectively (see full details in Supplementary Methods section 2.2.1). We used energy and chemical requirements adapted from previous process models16 and determined process emissions based on treatment target and existing unit processes.

Fig. 1: Overview of treatment configurations.
figure 1

Top: modelled unit processes for each treatment train, with all major energy-consuming processes. Bottom: Variations in treatment train are determined on the basis of the biological reactor and its configuration (middle, in green) and solids management (right, in blue). The following processes may also be included or excluded (indicated with the dashed lines): primary treatment, chemical phosphorous removal and power generation. Naming convention for each treatment train is based on the combination of biological reactor and solids management. The asterisk prefix is added to indicate the presence of primary treatment, and the suffixes -p and e indicate the presence of chemical phosphorous removal and power generation, respectively. All nomenclature codes are depicted in boxes next to the relevant process in this figure. Detailed descriptions of each treatment train can be found in Supplementary Tables 2 and 3.

Full size image

Figure 2a includes the total number of facilities and corresponding treated flow for each treatment configuration. Basic activated sludge (code A) accounts for 4,753 facilities (30% of total facilities) treating 23 billion m3 per year (34% of annual flow). However, there are over 8,000 lagoons (aerobic, anaerobic, facultative and unclassified) across the country, making them the most abundant technology despite treating only an estimated 7.8% of national flow. Thirty-two per cent of facilities use aerobic digestion (code 3) for solids treatment, and 19% use anaerobic digestion (codes 1 and 2). Treatment configurations with anaerobic digestion treat 64% of total flow, while those with aerobic digestion treat only 17%, indicative of the large size of facilities with anaerobic digestion. In addition, we identified 321 facilities recovering energy from biogas production, accounting for 10% of plants with anaerobic digestion. Facilities with energy recovery are large, with an average flow rate (139,000 m3 per day, equivalent to 37 million gallons per day) greater than that of 98.6% of WWTPs included in this study.

Fig. 2: Treatment train distribution and associated emissions.
figure 2

a, Count and flow. b, Electricity and natural gas consumption intensity. c, GHG emissions intensity. d, Annual GHG emissions. e, Key unit processes and treatment targets of treatment trains. Identical configurations with and without primary treatment are combined using brackets ([*]). Treatment train codes are defined in Fig. 1, and trains beginning with L- indicate lagoons (L-a, aerobic; L-n, anaerobic; L-f, facultative; L-u, uncategorized, calculated as the flow-weighted average of aerobic and anaerobic/facultative lagoons). In b, hatches indicate energy used for chemical production (acetic acid, hypochlorite and lime). In c and d, the electricity emissions reflect national average carbon intensity of electricity production. In b–d, error bars reflect results of 10,000 Monte Carlo simulations, also summarized in box plots (whiskers indicate 5th and 95th percentiles, boxes indicate the 25th to 75th percentiles, midlines indicate the 50th percentiles and diamonds represent the mean). Colours indicate different emission types: red indicates on-site GHG emissions, with shade indicating gas type: CH4 (dark), N2O (medium) and CO2 (light); yellow indicates upstream emissions associated with electricity usage; blue indicates emissions associated with natural gas extraction, transportation and on-site combustion. Emissions associated with biosolids handling have limited contribution to the total emission and are not included in breakdown bars for figure clarity, but values remain included in box plot results. The sum of the median value of each emission type (bar plots) in c and d tends to be lower than the median of total emissions (box plots) due to right-skewed underlying distributions, where extremely high values can disproportionately influence aggregate median. In e, circles indicate the presence of specified unit processes or treatment targets. AD, anaerobic digestion; P removal, phosphorous removal; N removal, nitrogen removal.

Source data

Full size image

Aside from lagoons, the most common treatment train is basic activated sludge coupled with aerobic digestion (code [*]A3), accounting for 22% (3,557) of all facilities and treating 10% of national flow (6.7 billion m3 per year). Accounting for a lower proportion of all facilities (7.3%), basic activated sludge with anaerobic digestion ([*]A1 and [*]A1e) is the configuration that treats the most wastewater (17%). All nutrient transformation configurations combined (codes D, E, F and G) account for a small portion of total facilities (13%) but treat 39% of total national flow (27 billion m3 per year).

Electricity and natural gas requirements

We determined the electricity (kWh m−3) and natural gas (MJ m−3) needed for each treatment train (Fig. 2b) by expanding existing models16 to account for the full range of treatment configurations in the USA. In this section and below, we report median results from our Monte Carlo simulation, with 5th and 95th percentiles reported subsequently in brackets. In Fig. 2b, error bars (5th and 95th percentiles) reflect the national profile for electricity carbon intensity, highlighting the variability possible based on regional electricity mix (Supplementary Methods section 2.6.1). The least energy-intensive configurations include anaerobic/facultative lagoons (L‑n/L‑f, median [5th and 95th percentiles] 0.17 [0.14–0.21] kWh m−3) and a trickling filter with anaerobic digestion (*C1e, 0.19 [0.083–0.29] kWh m−3). These three configurations have no natural gas requirement. Conversely, the most intensive configurations use membrane bioreactors for biological nutrient removal (*D1e and *D3) which consume: 1.6 [0.83–2.3] kWh m−3 electricity and 5.8 MJ m−3 natural gas (*D1e, natural gas not subject to uncertainty for this configuration; Supplementary Table 24); and 1.7 [1.0–2.5] kWh m−3 and 4.5 [3.9–5.2] MJ m−3 (*D3). The electricity usage in these two configurations is primarily for powering bioreactors, and the natural gas is for producing acetic acid used in nutrient removal.

With increased pumping and aeration for nitrification, the top electricity-consuming configurations are those with biological nutrient removal for secondary treatment (codes D, E, F and G). Within the top ten energy-consuming treatment configurations, process electricity accounts for over 80% of total electricity consumption. Electricity for producing chemicals accounts for a higher portion of emissions in biological phosphorous removal processes (codes D and G), with acetic acid representing 9–22% of total electricity use in the baseline model inputs (mean 15%). Consequently, the regional grid largely determines electricity emissions rather than chemical manufacturing. Biological phosphorous removal requires the most natural gas, with an average of 90% used for acetic acid production, although this proportion decreases when solids are stabilized with incineration. Reducing natural gas dependency will require more sustainable chemical manufacturing and selection, as well as reduced reliance on incineration.

Carbon dioxide, methane and nitrous oxide emissions

Across different treatment trains, CH4 and N2O combined account for 73–96% of median process emissions (fossil origin CO2, CH4 and N2O from biological wastewater treatment), with CO2 exceeding 11% only when organics removal (codes A, B and C) is coupled with aerobic digestion or incineration (codes 3, 4, 5 and 6). The highest process emissions originate from facilities with nitrification (code E), anaerobic or facultative lagoons (L-f or L-n) and anaerobic digesters (codes 1 and 2) (Fig. 2c). Nitrification with anaerobic digestion (*E1[e]) produces 0.69 [0.33–1.3] kg CO2e m−3, with 53% [24–84%] from CH4 and 40% [8.8–72%] from N2O. Anaerobic and facultative lagoons produce 0.94 [0.13–2.3] kg CO2e m−3, 96% [70–99%] of which is CH4. However, each configuration’s contribution to annual emissions depends on the abundance and flow rate of facilities (Fig. 2d). Nationwide, nitrifying facilities (code E) are the largest contributors, generating an estimated 11 [6.9–17] million metric tonnes (MMT) CO2e per year, followed by activated sludge facilities (code A), which contribute 7.5 [5.3–11] MMT CO2e per year despite a much lower flow-normalized emission rate. By contrast, all types of lagoons contribute 3.0 [1.5–4.9] MMT CO2e per year.

Nationwide emissions by treatment configuration

Total emissions include process emissions, energy emissions and downstream emissions from treated biosolids land application or landfill disposal. We did not include emissions from sewer systems, effluent discharge and facility construction. To facilitate comparison between generic configurations, we calculated emissions intensity (kg CO2e m−3) for each treatment train (Fig. 2c,d) using the national profile for electricity carbon intensity, while calculations elsewhere associated with inventoried facilities use the carbon intensity of the plant’s regional electricity supply (Supplementary Methods section 2.6.1). Highest total emissions are from configurations removing nutrients with membrane bioreactors: *D1e (1.8 [1.2–2.6] kg CO2e m−3) and *D3 (1.5 [0.85–2.4] kg CO2e m−3). Fifty-nine per cent [39–77%] (*D1e) and 71% [45–89%] (*D3) of emissions are from energy consumption, attributable to the electricity requirement of membrane systems and the natural gas needed for acetic acid production. The train with the next-highest carbon intensity is *G1 (1.1 [0.73–1.7] kg CO2e m−3). Here, 60% [40–76%] of emissions are produced through biological treatment (CH4, N2O and CO2) and 37% [22–56%] are from electricity and natural gas use. Across treatment trains, biosolids disposal at most accounts for 12% [4.6–26%] of total emissions (*C4). Finally, by comparing median emission values from identical configurations with and without energy recovery through combined heat and power (CHP), we found that energy recovery reduces total emissions by only 5.2–13%. Understanding whether emissions are driven by treatment processes or electricity requirements will inform decarbonization strategies, particularly as the US electricity generation mix continues to evolve17.

Table 1 Annual emissions from wastewater treatment in the USA, reported as median values followed by 5th to 95th percentile values in brackets
Full size table

We conducted a sensitivity analysis using Spearman’s rank correlation to identify the following key drivers of emissions (P < 0.05 and an absolute ρ value >0.2) across treatment configurations: wastewater influent chemical oxygen demand (COD) and total nitrogen (TN), emissions factors for CH4 and N2O, electricity requirement and grid carbon intensity (Fig. 3). Across nearly all configurations, emissions are most sensitive to the CH4 emissions factor, even when anaerobic digestion is not present. By contrast, N2O emissions factor and influent TN are primarily, although not exclusively, important for nutrient removal facilities. COD is a statistically significant driver only in lagoon systems, for which fugitive CH4 emissions are directly proportional (as reflected in CH4 emissions factors). These findings underscore the importance of further refining emissions factors and influent wastewater variability through representative measurement campaigns and public datasets.

Fig. 3: Sensitivity of emissions estimates to key input parameters.
figure 3

Values represent Spearman’s ρ value for significant parameters (P < 0.05 and absolute ρ value >0.2), indicating their influence on per volume emission for each treatment configuration. An absolute ρ value closer to 1 (darker red) indicates the corresponding parameter has stronger influence on the emission result. EF, emissions factor.

Source data

Full size image

Nationally, wastewater treatment produces 47 [41–55] MMT CO2e per year (Table 1), based on a national-scale Monte Carlo simulation. CH4 dominates annual emissions (41% [31–54%]), followed by N2O (24% [16–38%]) and electricity (23% [19–27%]). Wider percentile ranges for CH4 and N2O reflect higher uncertainty and greater variability in available emissions data. Electricity emissions show lower uncertainty, reflecting less variation in estimates. While emissions from biosolids remain poorly characterized, our estimate of these emissions is only 1.1 [0.53–2.0] MMT CO2e per year, or 2.2% [1.1–4.3%] of national emissions. Emissions associated with natural gas production, distribution and combustion are similarly low (1.4 [1.2–1.5] MMT CO2e per year).

National distribution of GHG emissions from WWTPs

Figure 4 depicts the geographic distribution of emissions from selected treatment configurations: anaerobic digestion, *E1[e], F1[e] and lagoons (see Supplementary Fig. 1 for similar plots for each treatment train). Note that we report only baseline median values here for visual clarity. Emissions are highly distributed but cluster with major population centres, although nutrient removal configurations, particularly nitrification (*E1[e]), are more abundant in the eastern half of the country. The high density of anaerobic digestion facilities indicates that novel aerial methane measurement techniques may be promising for leak detection when used to survey large geographic areas18,19. While facilities with lagoons account for 9.6% of emissions in our inventory, these facilities are small, highly distributed and largely of unknown operation, thus posing a potential challenge to mitigation efforts. In aggregate, we found that the largest 10% of emitters account for 82% of emissions (Fig. 5a), and there is a strong linear relationship between total emissions and total flow (Fig. 5b). The largest ten emitters in the country (0.06% of facilities, treating 9.9% of national flow) generate 4.8 MMT CO2e per year and disproportionately account for 11% of total emissions from all WWTPs (Fig. 5c).

Fig. 4: Geographic distribution of emissions from selected treatment processes and trains.
figure 4

a, All facilities with anaerobic digestion (codes 1 and 2). b, *E1 and *E1e. c, F1 and F1e. d, All lagoons (aerobic, anaerobic, facultative and unclassified). Legends from inner to outer rings represent the annual emission of 0.01, 0.25, 1 and 2 MMT CO2e, respectively. A non-shaded legend indicates that no WWTP has emissions in the corresponding range.

Source data

Full size image
Fig. 5: Distribution of total emissions in the USA.
figure 5

In each plot, individual points represent a single facility, and the y axis represents the fraction of total emissions. a–c, The x axes represent the fraction of total facilities, with facilities sorted from highest to lowest emissions (a); the fraction of total flow, with facilities sorted by flow rate from largest to smallest (b); and the magnitude of emissions from a single facility, with facilities sorted from lowest to highest emissions (c). The red line in b represents parity of x- and y-axis values (y = x).

Source data

Full size image

Discussion

We estimated emissions from wastewater treatment in the USA are 47 [41–55] MMT CO2e per year, with CH4 and N2O accounting for 66% [51–86%] of emissions. Compared with previous works (Supplementary Table 1), our approach integrates a greater number of emissions sources, including electricity, CO2 (fossil origin) and biosolids disposal, and allows a side-by-side comparison across key emission types. Our estimates of GHG intensity for specific treatment trains range from 0.33 [0.19–0.74] kg CO2e m−3 (*C6) to 1.8 [1.2–2.6] kg CO2e m−3 (*D1e), aligning with models by the US EPA (0.5–1.8 kg CO2e m−3), although the relative contributions of different components differ20. Largely due to a higher number of identified anaerobic digestion facilities, our national estimate of CH4 emission (19 [15–25] MMT CO2e per year) is approximately twice that of Song et al., who estimated 9.7 MMT CO2e per year by drawing on one of the four source datasets used in this study3 (see Methods for additional details). Our nationwide estimate for N2O (12 [7.9–17] MMT CO2e per year) is similar to values reported by Song et al. (11.6 MMT CO2e per year), calculated using the same underlying emissions factor data that we grouped by treatment objective, whereas Song et al. organized by bioreactor type5.

Methane emissions are the largest single contributor to annual emissions in our inventory. We found that 85% [74–92%] (16 [12–21] MMT per year) of CH4 is produced by treatment configurations with anaerobic digestion (code 1 or 2). Anaerobic digestion produces energy on-site in the form of biogas, while reducing the spatial footprint of WWTPs and providing additional opportunities for resource recovery21. However, our results indicate fugitive CH4 emissions currently outweigh climate benefits gained from renewable biogas. On a volumetric basis, energy recovery (code suffix e) reduces the median emissions by only 0.061 kg CO2e m−3 on average. By contrast, the average of median emissions from treatment configurations with anaerobic digesters is 0.34 kg CO2e m−3, or 0.30 kg CO2e m−3 higher than those with aerobic digestion. While our analysis uses data from the USA, these findings have broader global implications as concerns about GHG emissions from similar treatment trains are global. Recent studies in Europe have identified key sources of methane leaks in anaerobic digestion facilities, including non-gas-tight digestate storage22, biomass storage tanks and pressure relief valves23. In addition, minor technical fixes to existing infrastructure have been demonstrated to reduce methane emissions from biogas facilities by up to 46% (ref. 23), highlighting the potential impact of leak detection and repair programmes and the importance of immediately adopting such a programme globally.

We developed N2O emissions factors adapted from those compiled by Song et al. (2024), distinguishing on the basis of treatment objectives (organics removal, nitrification and denitrification)5. However, 79% of the underlying measurement data come from denitrifying systems24. Additional measurements are needed from conventional activated sludge and nitrifying systems. Given the high spatial and temporal variability in N2O emissions from wastewater treatment6, improving the characterization of N2O production is essential. In addition, current N2O mitigation efforts use aeration, feed and process optimization25, which are not captured in existing facility-level data. Understanding current operation strategies and their impact on emissions should be a focus of future research.

The relative importance of electricity generation (23% [19–27%] of total emissions) will decrease with grid decarbonization26. Biosolids disposal through landfilling or land application contribute only a small portion of total emissions (2.2% [1.1–4.3%]), but CH4 and N2O from biosolids are poorly studied. We used IPCC’s emissions factor for managed soils, which has a high uncertainty (0.002–0.018 kg N2O-N per kilogram of nitrogen)27 but aligns with the small number of recent studies where biosolids were used as an agricultural amendment28,29,30. Finally, our model did not account for CH4 produced through different practices for biosolids dewatering and on-site storage, which also contribute to fugitive CH4 emissions31.

The US EPA provides, to the best of our knowledge, the only other national-level inventory of wastewater treatment emissions in the USA. The two inventories differ in that we included electricity-associated emissions to enable a side-by-side comparison of key emissions sources, reflecting the wastewater industry’s historic prioritization of achieving net-zero energy use1. Also differing from the US EPA, we did not include CH4 and N2O effluent discharge emissions because we focused on emissions that can inform decarbonization efforts. Comparing on-site emissions, our N2O estimate (12 [7.9–17] MMT per year) is comparable to the EPA’s value (16.3 MMT per year), although our CH4 emissions estimate (19 [15–25] MMT per year) is threefold greater than their estimate (5.6 MMT per year−1)7. The relative contributions of the two gases also differ, with CH4 contributing 63% [51–73%] (19 [15–25] MMT per year) of combined on-site CH4 and N2O emissions in our inventory but only 26% (5.6 MMT) in the US EPA inventory. This difference is meaningful, as the relative importance of each gas informs potential mitigation measures. CH4 and N2O emissions are produced by different wastewater treatment processes via distinct microbial pathways, which are impacted by local climate (for example, temperature and seasonal changes to operation6,32). In addition, atmospheric lifespan and global warming potential will impact the priority of mitigation strategies for these two gases.

There are several limitations to this work. Our analysis relies on data reported through the US EPA’s Clean Watersheds Needs Surveys (CWNS), which did not require reporting unit process data in the most recent 2022 survey. As a result, 75% of our inventory relies on unit process data from 2012, although we used supplemental datasets and manual verification to update the data where possible (Fig. 6 and Supplementary Methods section 2.2.2). Furthermore, our results indicate a strong linear relationship between cumulative national flow rate and emissions (Fig. 5b). However, most recent national flow data from the CWNS do not distinguish between observed and design flow at facilities, meaning we probably overestimated the volume of wastewater treated. In addition, we used uniform influent concentrations for all facilities and did not account for variations in effluent requirements based on local regulations. Given their importance in estimating GHG emissions, future CWNS surveys should require data reporting on treatment processes; biogas production and utilization; observed flow; and influent and effluent water quality.

Fig. 6: Dataset integration workflow for facility inventory, energy consumption and GHG emissions.
figure 6

References listed in the flow chart are as follows: aCWNS (2004, 2008, 2012, 2022)12,13,14,15, bUS EPA’s inventory of lagoons serving as the primary method of wastewater treatment at a given facility36, cWater Environment Federation survey of WWTPs with biogas production37, dUS Department of Energy’s survey of facilities with CHP38, eEPA annual report on biosolids production and disposal methods45, fbiosolid calculations outlined in the work of Seiple et al.11, gSong et al. (2023) methane emissions factors from WWTPs3, hSong et al. (2024) nitrous oxide emissions factors for WWTPs5, iIPCC emissions factors for land applied biosolids27, jUS EPA’s LandGEM model for estimating methane emissions from landfills46, kTarallo et al. (2015) modelled energy requirements for treatment trains16, lUS EPA estimate of lagoon energy requirements36, mGREET model41, nIPCC emissions factors for lagoons43.

Full size image

There are also several opportunities for further refining in future inventories, particularly with expanded data collection. We made the simplifying assumption that, within a particular treatment plant configuration, energy consumption is directly proportional to flow rate, despite the potential for increased efficiency at larger sizes33. In addition, the energy requirements for lagoons may not reflect current designs and operation34. Future inventories and resulting mitigation efforts will also benefit from improved CH4 and N2O emissions factors, which will require direct measurement studies across a range of representative facilities, climates and time periods. Moreover, for lagoons, we used current IPCC assumptions that aerobic lagoons do not produce CH4 and that anaerobic and facultative lagoons do not produce N2O. However, given the lack of fully uniform mixing, and the many connections between CH4 and nitrogen microbial metabolism35, it is likely that both gases are produced across all lagoon categories.

Our analysis considered only climate impact, but facility-level decision-making requires a more comprehensive environmental assessment. For example, trickling filter configurations are among those with the lowest total emissions. However, they have large spatial footprints and need adequate hydraulic head differences across the plant to limit pumping requirements. In addition, because anaerobic digesters are the largest source of on-site emissions, facilities using incineration (codes 5 and 6) compare favourably because they have lower on-site emissions. However, incineration requires adequate pollution control measures to prevent release of particulates, heavy metals and volatile organic compounds, considerations that we did not capture.

Wastewater treatment is a growing sector inextricably linked to public and environmental health, and understanding its current climate change impact is critical for decarbonization efforts. This analysis provides a comprehensive inventory of over 15,000 WWTPs across the contiguous USA, including treatment configuration, energy requirements and emissions estimates. Although confined to the USA, a similar approach can be adopted elsewhere. In less data-rich contexts, the archetypical treatment configurations reported here can be selected on the basis of best fit with available information, or serve as a guide for future data collection. In the USA, we identify on-site emissions of CH4 and N2O as priorities for climate change mitigation efforts, expanding the historical focus of the wastewater sector beyond energy-associated emissions. In addition, data generated from this work can be used to determine the impact of specific policy levers on overall emissions, and to analyse the effects of adopting novel resource recovery and emissions reduction technologies.

Methods

Facilities inventory development

We integrated multiple national datasets to compile an inventory of wastewater treatment facilities in the USA, their energy requirements and GHG emissions (Fig. 6). Our inventory consists of 15,863 WWTPs that reported non-zero flow to the US EPA in 202215. Throughout this Article, WWTPs refer to facilities owned and/or operated by municipalities or other public entities, including treatment plants and lagoons. We determined the treatment train for each facility in our inventory, defined as a common set of unit operations designed to reduce wastewater pollution between the influent and effluent of the plant. We used unit processes reported across the aggregated 2004, 2008, 2012 and 2022 CWNS12,13,14,15, supplemented with additional publicly available data that provided more granular or recent information: the US EPA’s Lagoon Inventory Dataset36, the Water Environment Federation’s Water Resource Recovery Facilities Biogas Database37 and the US Department of Energy’s Combined Heat and Power Installation Database38.

Using reported unit process data, we assigned each facility one or more treatment trains, abbreviated using alphanumeric codes, based on those previously defined by Tarallo et al.16 but with modifications to reflect additional possible combinations of liquids and solids treatment processes (full details are provided in Supplementary Methods section 2.2.2) Each treatment train is a unique combination of unit processes depicted in Fig. 1, and each configuration includes liquids and solids treatment processes, with optional primary treatment, chemical phosphorous removal and CHP for energy recovery. Liquids (or secondary) treatment technologies include activated sludge configurations, trickling filters and membrane bioreactors. Solids treatment options include aerobic and anaerobic digestion, lime stabilization and two incineration methods (see Supplementary Tables 2 and 3 for additional details). With available unit process data, we assigned treatment trains to 10,962 facilities, approximately 69% of the national fleet.

Where unit process data were incomplete or unavailable, we assigned treatment trains on the basis of facilities of a similar size and geographic location. For the 1,991 facilities with only partial unit process data available (that is, plants that provide information on secondary treatment or solids management but not both), we assigned treatment trains on the basis of the most common treatment train of similar plant size in the same EPA region, considering key unit processes present (activated sludge, biological nutrient removal, aerobic/anaerobic digesters, lime stabilization, incineration and trickling filters). For the remaining facilities with either insufficient partial data or fully absent data (2,910 facilities), we assigned a treatment train on the basis of the most common treatment train of similar plant size in the same EPA region. Supplementary Fig. 2 and Supplementary Table 7 provide a size breakdown and additional details on facilities with missing data.

Facility-level emissions associated with energy

For all treatment trains in the national inventory, we calculated electricity and natural gas consumption, as well as on-site electricity generation from biogas utilization. We used energy calculations from the results of process models in GPS-X reported by Tarallo et al.16. Because our study includes treatment trains beyond those reported by Tarallo et al., we used mass and heat balances for unit processes to determine energy requirements for treatment trains that were not modelled previously16 (details in Supplementary Methods section 2.3.1).

To estimate the GHG emissions associated with electricity consumption, we assigned each facility a regional emissions factor (kg CO2e kWh−1). Specifically, we used existing model results that simulate the energy mix across 134 simulated subregions of the USA, reflecting state-level and utility boundaries and capturing the existing variation to maintain load–interchange–generation balance39. We calculated electricity emissions factor (kg CO2e kWh−1) for each geographic region using the corresponding total emissions (kg CO2e) divided by the net power generated, as reported in National Renewable Energy Laboratory (NREL) Standard Scenarios Cambium data for 202017,40. We used data from the mid-case scenario based on central parameter values, including future electricity consumption, fuel costs and technology selection. We also calculated full fuel cycle GHG emissions for electricity and natural gas using the Greenhouse gases, Regulated Emissions, and Energy use in Technologies (GREET) model41 (Supplementary Table 18).

Carbon dioxide, methane and nitrous oxide

We estimated total CO2, CH4 and N2O from biological treatment processes, referred to here as ‘process emissions’ for brevity. For CO2, we assumed a baseline of 11.9% of influent carbon is of fossil origin, and 53.5% of influent COD (508 mg l−1) is released as CO2 during biological treatment42. To determine CH4 production, we used emissions factors based on the presence of key unit processes, and for N2O production we used emissions factors based on treatment targets. Namely, anaerobic processes (anaerobic digestors and lagoons) emit CH4, and nutrient removal processes emit N2O. For CH4, we used data reported by Song et al. (2023) to assign emissions on the basis of whether a facility contains an anaerobic digester3. For facilities with anaerobic or facultative lagoons, we used emissions factors reported by the IPCC43. For N2O, we developed emissions factors using data compiled by Song et al. (2024) to determine the median value of literature-reported values categorized by treatment objective: organics removal, nitrification or full denitrification5 (Supplementary Table 20). N2O emissions factors are a function of influent TN, and we assumed a range typical of domestic wastewater (23–69 mg l−1)44. As with CH4, we used N2O emissions factors reported by the IPCC for aerobic, anaerobic and facultative lagoons43. For uncategorized lagoons, we calculated both CH4 and N2O emissions factors using a flow-weighted average of the three other types of lagoon in the contiguous USA. For specific emissions factors and additional details, see Supplementary Methods section 2.4.

Biosolids-associated emissions

To estimate biosolids production, we combined reported production rates, where data were available, with mass flow calculations where data were unavailable. For 2,124 facilities, we used the US EPA’s Biosolids Biennial Report for 2020–202145, which documents the volume of biosolids produced, and their ultimate disposal through incineration, landfilling or land application. For the remaining facilities, we calculated sludge production based on facility flow rate, as developed elsewhere11 and described in Supplementary Methods section 2.5.1. For land applied biosolids, we used the IPCC N2O emissions factor for organic soil amendments27. For landfills, we used the US EPA’s Landfill Gas Emissions Model (LandGEM) to calculate the CH4 emissions factor for municipal solid waste landfills46. Full details are included in Supplementary Methods section 2.5.2.

Uncertainty and sensitivity analyses

For all treatment trains, we conducted Monte Carlo simulations (N = 10,000) for 19 input parameters subject to uncertainty. We selected all uncertainty ranges and distributions (Supplementary Table 24) using the tiered selection criteria previously developed47,48, with minor modifications described in Supplementary Methods section 2.6. Sensitivity analyses evaluated the relative impact of all input parameters on total per-volume GHG emissions for each treatment train, using Spearman’s rank order correlation coefficients. In this analysis, we defined key drivers of uncertainty as those with a P value <0.05 and an absolute ρ value >0.2.

For national-level results, we also evaluated uncertainty around two additional inputs: local electricity carbon intensity and facility biosolids production. We determined baseline facility-specific carbon intensity using 134 geographic regions reported in NREL Standard Scenarios Cambium data for 202017,40 and determined emissions factors for each electricity production pathway using the GREET model41. For local electricity carbon intensity, we assumed a uniform distribution with 80% and 120% of the baseline value as the lower and upper bound, respectively. For biosolids, we calculated the baseline facility-level mass flow rate and determined disposal pathways as described above and in Supplementary Methods section 2.5. Next, we used facility-level data to calculate the national total biosolids mass flow to landfills and land application. For our uncertainty analysis, we assumed a uniform distribution with 80% and 120% of the national total biosolids mass flow rate as the lower and upper bounds, respectively. Full details are provided in Supplementary Methods sections 2.5 and 2.6.

Due to the large number of facilities included in this study, we report only median values for all facility-level results (Figs. 4 and 5 and Supplementary Dataset C). Specifically, for each WWTP, we used median emissions factors to estimate emissions from individual processes (for example, biological CH4 emissions and upstream electricity emissions) and summed across all processes at a given facility to determine total emissions.