The potential of carbon markets to accelerate green infrastructure based water quality trading

This study evaluates the economics and emissions of water treatment technologies in the CONUS. The CONUS was divided into smaller sections as designated by the United States Geological Survey’s Hydrological Unit Code (HUC) regions. To maximize geographic resolution of this analysis, HUC 12 sub-watersheds were used wherever possible. However, it was assumed that nutrient trading could take place at the HUC 6 waterbasin level and, therefore, all results were aggregated to the waterbasin level³⁷. Geodatabase files for various HUC regions were downloaded from the United States Geological Survey’s Watershed Boundary Dataset³⁸. Data associated with HUC 12 sub-watersheds was aggregated from United States EPA’s EnviroAtlas database³⁹. EnviroAtlas provides national data layers at the HUC 12 sub-watershed level with many of these data layers being derived from data with a resolution of 30 m. Full details of which data was required is discussed in each subsection. It was assumed that nutrient trading could take place within each waterbasin (HUC 6), therefore stricter requirements placed on existing facilities could only be satisfied by gray or green treatment methods within the same waterbasin.

Wastewater nutrient data

Geographically resolved nutrient loading data compared to water quality targets for point source dischargers in the CONUS motivates the water treatment trade study. Therefore, 2022 data from the Nutrient Model (Hypoxia Task Force Search) created by the EPA was used⁴⁰. This data is provided through the EPA’s Water Pollutant Loading Tool⁴¹. The Nutrient Model was created by EPA to provide access to aggregated nitrogen and phosphorus loads (including modeled loads) for facilities across the United States. As such, data is provided for wastewater treatment facilities with current EPA NPDES permits with facility information, total annual wastewater flow, total nutrient loads, and maximum allowable nutrient loads (if applicable). In total, 53,055 data entries were provided for 29,335 unique facilities. Data consists of both discharge monitoring report (28,318) and modeled (24,737) nutrient loads for both nitrogen (27,238) and phosphorus (25,817). Additionally, each data point was associated with a HUC 12 sub-watershed code so analysis could be evaluated on a geospatially resolved level. An overview of the input data including number of facilities, mean daily wastewater flow, mean nitrogen concentration, and mean phosphorus all aggregated to the waterbasin level can be viewed in Fig. 4. For analysis of all technologies in this study, a 40 year time horizon was assumed.

a Number of wastewater facilities in each waterbasin. b Total mean flow in millions of gallons per day in each waterbasin. c Mean Nitrogen concentration of the wastewater (mg L⁻¹) in each waterbasin. d Mean Phosphorus concentration of the wastewater (mg L⁻¹) in each waterbasin.

Gray treatment methods

Gray nutrient treatment technologies outlined in the EPA’s report title Life Cycle and Cost Assessments of Nutrient Removal Technologies in Wastewater Treatment Plants were used in this analysis²⁸. The EPA report estimated the costs and GWP of 8 alternative wastewater treatment technologies to treat excess nitrogen and phosphorus in wastewater streams. Costs and GWP were also provided for a 9th ‘baseline’ technology, but it was excluded from this analysis because its primary design was not focused on nutrient removal and had low nutrient remediation potential. Details of the gray treatment methods can be seen in Table 2. Each gray nutrient treatment technology was assigned a treatment level in the EPA report. These levels range from Level 2 to Level 5 based on their ability to achieve target effluent nutrient concentrations. These concentration levels are 8 mgN L⁻¹ and 1 mgP L⁻¹ for Level 2, 6 mgN L⁻¹ and 0.2 mgP L⁻¹ for Level 3, 3 mgN L⁻¹ and 0.1 mgP L⁻¹ for Level 4, and 2 mgN L⁻¹ and 0.02 mgP L⁻¹ for Level 5. Level 1 designates that no effluent concentration is specified and has been excluded from this analysis accordingly.

To perform a geographically resolved analysis, costs and GWP of each gray treatment technology were adjusted based on the location of the facility being evaluated. For gray treatment technologies, only the electricity grid mix was assumed to vary geographically. Treatment costs and GWP presented in the EPA report assumed the 2010 United States average electrical grid mix was used for water treatment and all cost information was presented in 2014 dollars. We assumed a linear increase in energy demand between Level 2 and Level 5, which is likely conservative as some estimates suggest an exponential increase in energy use approaching Level 5⁴². Electricity prices were updated using the mean state electricity prices as reported by the United States Energy Information Administration’s 2021 Annual Energy Outlook, which is the most recent annual outlook available⁴³. To approximate the emissions associated with electricity use in various geographic regions in the United States, the United States EPA’s Emissions & Generation Resource Integrated Database (eGRID) was used⁴⁴. Since Energy Information Administration electricity prices and eGRID mixes are not aggregated to the sub-watershed level, the GeoPandas library in Python was used to compare the shapefiles for United States states and eGRID regions to HUC 12 sub-watersheds⁴⁵. If two states or eGRID regions overlapped a sub-watershed region, the state or eGRID region which overlapped a larger area of the sub-watershed was assigned to the sub-watershed. All technology costs and electricity prices were converted to 2022 dollars using historical Consumer Price Index data provided by the United States Bureau of Labor Statistics using the mean annual Consumer Price Index values for all items and the United States city average was used^46,47. The electricity prices and GWP was updated for each gray treatment method in each HUC region using the total electricity demand (kWh m⁻³) presented in the EPA report and 2021 electricity values using Eq. (1):

where X represents the technology’s cost or GWP, i represents the gray technology method, w represents the waterbasin value, US represents the United States mean value, ElectricDemand represents electricity demand of nutrient treatment for each gray technology, and Y represents the geographic specific cost or GWP of electricity.

Green treatment methods

Green non-point source nutrient treatment methods range from minimally invasive nutrient fertilizer reduction to land altering constructed wetlands^48,49. For this analysis, 7 green treatment methods were considered, all of which are implemented on agricultural farmland (Table 3). These include 3 barrier treatment methods which are applied at the edge of the field (saturated buffers, woodchip bioreactors, and constructed wetlands) and 4 land treatment methods (nutrient rate reduction, split nutrient application, cover crops, and no-till farming). Some of these treatment methods treat both nitrogen and phosphorus, while others only treat one of the two nutrients considered. Mean nutrient removal percentages and treatment costs came from the 2016 Illinois Nutrient Loss Reduction Strategy report⁴⁹. All values used within this analysis fall within the range of values reported in the literature^{48,50,51,52,53,54,55,56,57,58,59,60,61}.

One limitation to published values on green treatment methods is that they are presented in terms of the cost for the farmer to implement the technology, not the costs that would be incurred by a utility encouraging the adoption of these technologies to avoid new gray infrastructure upgrades. Therefore, some of the technology costs (i.e. applied nutrient reduction and no-till farming) are negative because they are cheaper than conventional farming practices. Since this analysis was performed from the utilities perspective, it was assumed that the utility would incur the costs of technology adoption, but would not claim the benefits of cost saving practices. Therefore, it was assumed that the technology costs of the negative cost technologies would be zero.

Additionally, it was assumed that farmers would need to be financially incentivized by the utility to implement green nutrient treatment methods. Therefore, it was assumed that the utility would pay farmers $31 acre⁻¹ year⁻¹ for land treated with green treatment methods, which is the mean value reportedly paid to farmers in 2021 by the Soil and Water Outcomes Fund⁶². This incentive payment is in addition to the green technology costs paid for by the utility. Barrier treatment methods which only need to be installed once, were only assumed to pay incentive fees during the first year of operation. Land treatment methods are applied yearly and, as such, the incentive fees were paid out annually. Lastly, constructed wetlands require up to 6% of the treated farmland acres to be converted to a wetland⁴⁸. As a conservative estimate, it was assumed that this land was productive farmland and the utility would need to rent the land from the farmer at the mean land rental prices as reported by the USDA’s 2022 land cash rental prices in order to compensate farmers for reducing their farm size⁶³. Farmland rental prices were reported at a state level and were applied to each waterbasin based on the states which the waterbasin resided in. If the waterbasin covered land in multiple states, the land rental prices were calculated using a weighted mean based on the number of agricultural acres in each state. Land rental prices were assumed to stay constant over the life of the project.

Similar to the costs of gray treatment technologies, the costs of green treatment technologies were received in 2016 dollars and were converted to 2022 dollars using historical Consumer Price Index data provided by the United States Bureau of Labor Statistics^46,47. The GWP of each green treatment method were estimated using life-cycle inventory data from the EcoInvent 3.71 database, using cut-off analysis, accessed through the software openLCA 1.10.3 (https://openlca.org), and calculated using the Traci 2.1 impact assessment methodology^64,65. The GWP estimate for constructed wetlands includes direct land use change effects which were calculated using IPCC methodology⁶⁶. Details of LCA calculations for each treatment method are discussed in the Supplementary Methods 2 section in the Supplementary Information.

Since each green nutrient treatment method requires different topology, infrastructure, or climate in order to be implemented; not every green treatment method could be applied in every waterbasin. Therefore, land limitations were added to green infrastructure on a waterbasin basis. These land limitations included the availability of tile-drained soil (saturated buffers and woodchip bioreactors), the availability of riparian buffer between agricultural land and discharge waterways (saturated buffers), the soil and climate to support wetlands (constructed wetlands), and the requirement of supplemental fertilizer application (nutrient rate reduction and split nutrient application). It was assumed that if the requirements were met in one part of the waterbasin, the requirements could be implemented in the rest of the waterbasin and the nutrient reduction strategy could be applied. For example, if tile drains were used on agricultural land in one part of the waterbasin, it was assumed that they could be added to all agricultural land in the waterbasin. Data for tile drain locations was acquired from Nakagaki et al. based on analysis from Sugg and data for riparian buffers, wetlands, and fertilizer application were acquired from the EPA’s EnvironAtlas^39,67,68. Details of each green treatment method’s requirements is provided in Supplementary Table 1 and maps of tile drainage, riparian buffers, wetlands, and fertilizer application availability in each waterbasin is presented in Supplementary Fig. 13.

One of the benefits of green nutrient treatment methods is that they can be used in combination with each other⁴⁸. This analysis considered all combinations of the 7 treatment methods proposed. Since each of the barrier treatment methods are applied at the edge of the field before discharge to the waterway, it was assumed that only one barrier treatment method could be used at a time. Conversely, no limitations were placed on the land treatment methods. Therefore, 63 unique combinations of green treatment methods were evaluated to find the best performing treatment methods in each watershed. For combined green treatment methods, it was assumed that costs, GWP, and nutrient removal efficiency were compounded. For example, if saturated buffers were combined with cover crops, their nitrogen cost would be $1.95 kgN⁻¹ + $3.90 kgN⁻¹ = $5.85 kgN⁻¹, nitrogen GWP would be 0.10 kg − CO_2eqkgN⁻¹ + 0.55 kg − CO_2eqkgN⁻¹ = 0.65 kg − CO_2eqkgN⁻¹, and their nitrogen removal efficiency would be 90% + 30%∗ (100% − 90%) = 93%.

Calculation methods and assumptions

In order to estimate the nutrient trading potential of green versus gray nutrient reduction technologies, multiple scenarios were assumed. The first scenario assumed that each of the wastewater treatment facilities evaluated were required to meet Level 2 nitrogen and phosphorus concentration limits of 8 mgN L⁻¹ and 1 mgP L⁻¹, respectively. These values were selected because they are the conservative limit that all gray treatment technologies can achieve based on their treatment level in the EPA report. The second scenario assumed that each of the wastewater treatment facilities evaluated were required to meet Level 5 nitrogen and phosphorus concentration limits of 2 mgN L⁻¹ and 0.02 mgP L⁻¹, respectively. These values were selected because they are the limit that the advanced reverse osmosis gray treatment technologies can achieve based on their treatment level in the EPA report. Two additional scenarios (Level 3 and Level 4) were also run to evaluate the sensitivity of results between the conservative Level 2 and advanced Level 5 scenarios. Concentration limits for nitrogen and phosphorus were 6 mgN L⁻¹ and 0.2 mgP L⁻¹ for the Level 3 scenario and 3 mgN L⁻¹ and 0.1 mgP L⁻¹ for the Level 4 scenario. Each scenario was evaluated independently of the other scenarios. For all treatment methods, analysis was performed on the facility level and nutrient trading was assumed to occur within each waterbasin.

For each scenario, all facilities where both nitrogen and phosphorus concentrations were lower than the specified limits were excluded from analysis. Additionally, each gray nutrient treatment technology had maximum concentration limits which they could decrease the effluent during treatment. It was assumed that the wastewater could be treated multiple times when the concentration was above this limit, but costs and GWP would increase by the multiple of the number of treatments required. To avoid the highest concentration scenarios which would exaggerate the gray treatment costs, facilities which required a mean nutrient concentration reduction greater than 5X the Level 2 treatable concentration limit were excluded from analysis. Facilities located outside CONUS (i.e. Alaska, Hawaii, Puerto Rico, Guam, United States Virgin Islands, and American Samoa) were also excluded due to their lack of HUC 12 sub-watershed data provided by the EPA’s EnviroAtlas³⁹. After data filtration to remove facilities residing outside the CONUS or with nutrient concentrations lower than treatable limits, 18,534 unique facilities remained for the Level 2 analysis (16,686 facilities treated for nitrogen, 14,444 facilities treated for phosphorus, and 12,633 facilities treated for both); 20,989 unique facilities remained for the Level 3 analysis (17,634 facilities treated for nitrogen, 19,369 facilities treated for phosphorus, and 16,045 facilities treated for both); 21,828 unique facilities remained for the Level 4 analysis (19,207 facilities treated for nitrogen, 20,110 facilities treated for phosphorus, and 17,514 facilities treated for both); and 22,386 unique facilities remained for the Level 5 analysis (19,769 facilities treated for nitrogen, 20,785 facilities treated for phosphorus, and 18,192 facilities treated for both).

In addition to gray treatment facility limitations, green treatment methods were limited by agricultural land availability within each waterbasin. Total area within each sub-watershed was calculated using the GeoPanda’s area function in Python. The percentage of crop land and pasture land in each sub-watershed as reported by EnviroAtlas were used to approximate the total agricultural land in each sub-watershed³⁹. Since nutrient trading was performed at the waterbasin level, sub-watershed values were aggregated to the waterbasin level to determine the maximum nutrient treatment of the waterbasin as a whole.

Additionally, some of the green treatment methods considered are already in use on farms throughout the CONUS, but limited information exists on their prevalence. The USDA’s 2017 agricultural census provides state-level tillage and cover crop data, but geographically resolved data is unavailable for the other green treatment methods⁶⁹. The most recent non-census data coverage data is provided by the Iowa Department of Agriculture and Land Stewardship in their Iowa Nutrient Reduction Strategy 2018-19 Annual Progress Report⁵⁸. The report states that of the total 30,600,000 acres of farm land in Iowa, 8,200,000 acres (26.8%) were no-till farmed, 5,700,000 acres (18.6%) were treated with nutrient management strategies (nitrogen rate reduction and split nitrogen applications), 973,000 acres (3.2%) used cover cropping, 107,000 acres (0.35%) were treated with wetlands, and 2000 acres (0.35%) were treated with either saturated buffers or woodchip bioreactors^58,70. To fill the gaps between the USDA census data and treatment methods considered, these values were applied to their respective green treatment methods across all waterbasins in the CONUS to provide a conservative estimate of land availability for additional green treatment applications. While accounting for land limitations, the maximum nutrient treatment potential of each green technology in each waterbasin was calculated using Eq. (2):

where NT_i,w represents the possible nutrient treatment for each green technology (subscript i) in each waterbasin (subscript w), A_w represents the waterbasin total area, Pct_crop,w represents the percent of waterbasin area which is crop land, Pct_pasture,w represents the percent of waterbasin area which is pasture land, Pct_tech,i represents the percent of agricultural land currently treated with each green treatment method, N_mean−loss represents the mean nutrient loss per land area of agricultural land in the waterbasin, and Pct_{N−removal,i} represents the percent of nutrient removal for each green technology. The state-level nutrient runoff values as predicted by the 2012 regional United States Geological Survey’s Spatially Referenced Regression On Watershed attributes models were used to quantify nutrient loading from agricultural land in each waterbasin^{71,72,73,74,75,76}.

Analysis was performed first for all green treatment methods and combinations. The required nutrient treatment and the possible nutrient treatment were calculated on a waterbasin level as described in the previous paragraphs. If the available agricultural land in a waterbasin could not support the removal of the required nutrient load to meet the desired concentration limits, it was assumed that the maximum possible treatment would be applied based on the land available. The percentage of maximum nutrient treatment compared to the desired nutrient treatment was calculated and was used for nutrient treatment of all facilities within the waterbasin. Total land area required for nutrient remediation was also recorded to calculate farmer incentive payments. After the nutrient treatment loads were calculated for each wastewater facility, the new mean nutrient concentrations were calculated based on annual wastewater discharge. After final nutrient treatment loads were determined, the treatment costs (including farmer incentive and wetland costs) and GWP were calculated for both nutrients. Lastly, if both nitrogen and phosphorus were being treated, the total treatment costs and GWP of the facility were set by the nutrient which required more infrastructure. For example, if nitrogen required 500 ha of treatment to meet concentration limits and phosphorus required 1000 ha, the phosphorus treatment costs and GWP were assumed for treatment of both nutrients at the facility since both nutrients can be treated simultaneously for certain treatment methods. Comparison of the nutrient treatment levels for each green treatment method are presented in Supplementary Fig. 4.

After treatment costs and GWP were calculated for every wastewater treatment facility and each green treatment method (including combinations), an optimization was run to determine the maximum amount of nutrients that could be treated using green treatment methods in each waterbasin. In many waterbasins, multiple green treatment methods could treat the required nutrient load to reach the desired concentration limits. Therefore, a secondary optimization was performed to determine the minimum cost scenario and minimum GWP scenario when the maximum amount of nutrients were treated. Results for the minimum cost scenario are used for comparison to the gray treatment methods in the results section. The minimum GWP scenario was excluded from the primary results section because it has a breakeven carbon cost of $939 per tonne-CO2e when compared to the minimum cost scenario which is more expensive than direct air carbon capture technologies⁷⁷. Detailed results for both optimization scenarios is discussed in the Supplementary Discussion 5 section in the Supplementary Information.

Once costs and GWP were determined for each green treatment method, costs and GWP were calculated for each of the gray treatment methods. In order to ensure green and gray treatments were compared evenly, the gray nutrient treatment levels were set equal to those of the green maximum treatment scenarios even though they are not limited by agricultural land constraints. If these limitations were not placed on gray treatment technologies, they would treat more nutrients than the green treatment methods which would increase their treatment costs and emissions and exaggerate the benefit of green treatment methods. Costs for all treatment methods were originally calculated at the wastewater facility level using sub-watershed characteristics. For analysis purposes, results were aggregated from the facility level to the waterbasin level. Supplementary Fig. 14 provides a diagram of the analysis process.