Abstract
The negative effects of nutrient pollution in streams, rivers, and downstream waterbodies remain widespread global problems. Understanding the cost-effectiveness of different strategies for mitigating nutrient pollution is critical to making informed decisions and defining expectations that best utilize limited resources, which is a research priority for the US Environmental Protection Agency. To this end, we modeled nutrient management practices including residue management, cover crops, filter strips, grassed waterways, constructed wetlands, and reducing fertilizer in the upper East Fork of the Little Miami River, an 892 km2 watershed in southwestern Ohio, United States. The watershed is 64% agriculture with 422 km2 of row crops contributing an estimated 71% of the system’s nutrient load. The six practices were modeled to treat row crop area, and among them, constructed wetlands ranked highest for their low costs per kilogram of nutrient removed. To meet a 42% phosphorus (P) reduction target for row crops, the model results suggested that the runoff from 85.5% of the row crop area would need to be treated by the equivalent of 3.61 km2 of constructed wetlands at an estimated cost of US$2.4 million annually (or US$48.5 million over a 20-year life cycle). This prompted a series of projects designed to understand the feasibility (defined in terms of build, treatment, and cost potential) of retrofitting the system with the necessary extent of constructed wetlands. The practicalities of building this wetland coverage into the system, while leading to innovation in unit-level design, has highlighted the difficulty of achieving the nutrient reduction target with wetlands alone. Approximately US$1.2 million have been spent on constructing 0.032 km2 of wetlands thus far and a feasibility analysis suggests a cost of US$38 million for an additional 0.409 km2. However, the combined expenditures would only achieve an estimated 13% of the required treatment. The results highlight the potential effectiveness of innovative design strategies for nutrient reduction and the importance of considering realistic field-scale build opportunities, which include accounting for acceptance among landowners, in watershed-scale nutrient reduction simulations using constructed wetlands.
- constructed wetlands
- cost effectiveness
- implementation
- innovative design
- nutrient pollution
- watershed modeling
Introduction
Implementation of constructed wetlands for nutrient reduction from row crops (RC) in a large watershed in which they are the dominant source has been the pursuit of a watershed partnership in southwestern Ohio for over a decade. Constructed wetlands have long been considered as an approach to mitigate excess nitrogen (N) and phosphorus (P) (hereafter referred to collectively as “nutrients”) in municipal, industrial, and agricultural discharges (Kadlec et al. 2000; Kadlec and Wallace 2008; Crumpton 2001; Crumpton et al. 2008). They are known to be effective nutrient sinks when they are designed and maintained to accommodate sufficiently long hydraulic residence times (e.g., >1 d) relative to their hydraulic loading rates (Crumpton et al. 2008, 2020; Cheng and Basu 2017; Fisher and Acreman 2004; Land et al. 2016). Much of the research on the design and performance of constructed wetlands has been conducted at the unit-level, where one wetland or a series of cells comprising a wetland treatment system is evaluated by in-out type measurement and modeling (USEPA 2000; Kadlec and Wallace 2008). In the context of treating excess nutrients from nonpoint agricultural sources specifically, unit-level constructed wetlands treating small farming catchments are often studied as an edge-of-field nutrient reduction practice for farm runoff (Williams et al. 2016; Mendes 2020, 2021; Gordon et al. 2021a; Soldo et al. 2022; Weeber et al. 2022; Mitchell et al. 2022). Farmers can apply for financial and technical assistance from the USDA National Resources Conservation Service (NRCS) for adding a constructed wetland for nutrient runoff reduction as an approved practice under different programs like the Environmental Quality Incentives Program (EQIP) (USDA NRCS 2023). Determining how multiple wetland installations across a landscape combine to effect watershed-level nutrient loads has largely been approached as a modeling exercise and is less well understood (Gordon et al. 2021a). Real-world examples that measure the effectiveness of multiple constructed wetlands in the aggregate are needed because there may be barriers to implementation at the small scale or emergent properties at the larger scale that are otherwise difficult to account for with modeling exercises.
To understand the role that constructed wetlands play in meeting nutrient reduction goals in watersheds that drain to impaired waters, like when part of implementing total maximum daily load (TMDL) requirements (USEPA 2018), practitioners may need to look beyond the field or parcel level to account for the hydrologic and nutrient loading properties of the larger watershed. Studying the potential for wetlands at this scale requires the integration of watershed nutrient budgeting and tracking tools and simulations of process-level performance or expected removal rates of multiple wetland installations of varying size and complexity. Such research on wetland effectiveness at watershed scales also has a dense and ongoing research history (Day et al. 2003; Mitsch et al. 2005; Tomer et al. 2013a; Hansen et al. 2018, 2021; Kalcic et al. 2018; Evenson et al. 2021; Lemke et al. 2022; Crumpton 2001; Zammali et al. 2021). These effectiveness studies have been informed by others that focus on the relevance of the patchwork of isolated wetland systems in headwatersheds and along the corridors of larger rivers (Evenson et al. 2016; Cohen et al. 2016; Thorslund et al. 2017; Czuba et al. 2018). This track of research is important to understanding the natural role wetlands play in regulating flow permanence and water quality in downstream waterbodies and the decisions about their jurisdictional protection under the Clean Water Act in the United States (Leibowitz et al. 2018; USEPA 2015; Leibowitz 2003). Effort has focused on modeling the interactions between spatial location, landscape pattern, relative extent and structural variability on nutrient fate and transport processes (Evenson et al. 2021, 2023; Golden et al. 2021, 2023).
The Soil Water Assessment Tool (SWAT) is a popular model used for characterizing the potential effectiveness of constructing wetlands in watersheds (Krysanova and White 2015; Ikenberry et al. 2017; CARD 2019). SWAT is sometimes paired with other tools and/or an optimization algorithm to assess the cost-effectiveness of wetlands in achieving nutrient reduction targets set by TMDLs or similar regulations (Gordon et al. 2021a; Weeber et al. 2022; Yang and Best 2015; Amin et al. 2020; Heiskary and Markus 2001; Preston et al. 2011; Iavorivska et al. 2021; Singh et al. 2019). Cost-effectiveness, in this context, is a relative term indicative of the cost per unit of treatment. Cost-effective installations at the unit or watershed scale are those that have lower cost per unit treatment (e.g., US$ kg−1 P removed) compared to other practices, among design alternatives, or different spatial configurations. Berkowitz et al. (2020) noted that optimization and identification of wetlands for P retention requires analyses at both the watershed and field scales. They coupled SWAT watershed simulations with field-scale soils analysis to identify wetland placement opportunities. Hansen et al. (2021) used multiobjective optimization to evaluate cost-effectiveness and integrated three biophysical models, including SWAT and two stream corridor models, finding that fluvial wetlands were the single-most cost-effective management action for reducing both nitrate (NO3−) and sediment based on the relative location of sources and sinks over a large watershed in the upper Mississippi River Basin. The diversity among studies highlights that turning watershed-scale prospects for the effectiveness of constructed wetlands into implementation realities requires contemplating multifaceted engineering designs and different opportunities for placement. Furthermore, there may be barriers to implementing the optimal model-based scenarios because of cost limitations and/or landowner perceptions and preferences (Brooker et al. 2021; Martin et al. 2022; Berkowitz et al. 2020; Kalcic et al. 2014). While these examples demonstrate how watershed-scale modeling simulations can be used to optimize wetland effectiveness compared to other practices or farmer adoption preferences, for the most part, studies focusing on the on-the-ground realization of these exercises appear lacking. Never, to our knowledge, have enough funds been available at once to cover a projected watershed-scale constructed wetland need (optimized or not). Because of this, there is an inevitable shift to designing and constructing wetlands in a piecemeal fashion. This is likely to require such resources and time that rarely allows a capacity to compare the real expenditures to the original cost-effectiveness projections.
Our objective was to describe the effort that the East Fork Watershed Cooperative, a federal, state, and local partnership, has undertaken to implement constructed wetlands to help mitigate excess nutrient runoff in a large watershed in southwestern Ohio and discuss the lessons learned within the context of meeting TMDL nutrient reduction targets. We compare modeled cost-effectiveness to realized cost-effectiveness. We present the results of SWAT modeling and applying the Agricultural Conservation Planning Framework (ACPF) (Porter et al. 2018; Ranjan et al. 2019; Tomer and Nelson 2020) independently of SWAT to identify suitable locations for constructed wetlands. The TMDL is being written for the upper East Fork of the Little Miami River watershed (UEFW) (OhioEPA 2021). The nutrient reduction is intended to address recurring harmful algal blooms in a downstream reservoir used for flood control, recreation, and as a source of drinking water.
The primary source of the excess nutrients in the UEFW is row crops (RC) (soybean [Glycine max {L.} Merr.] and corn [Zea mays L.]). Our assessment fits within the bigger context of understanding the cost-effectiveness of mitigating nutrient pollution using a variety of agricultural conservation practices (ACPs). Constructed wetlands, one of the ACPs, were grouped into the following three categories for our analysis: (1) units treating small RC catchments, often referred to as edge-of-field wetlands and akin to those contemplated by Mitchell et al. (2022), Gordon et al. (2021b), Gordon et al. (2021a), and Lemke et al. (2022); (2) those situated in the floodplain terraces adjacent to larger stream and river channels, like those deemed most cost-effective by Hansen et al. (2021); and (3) larger wetland treatment complexes, situated to receive high hydraulic loading rates, typically by active pumping, yet large enough in aerial extent to meet residence time requirements to be considered effective, such as the systems studied by Kadlec and Hey (1994), Mitsch et al. (2014), and Jacquemin et al. (2022). Each of these alternatives was considered within a framework of meeting the estimated construction extent needed for achieving the TMDL nutrient targets.
The constructed wetland implementation research is diagramed in figure 1 and evolved over a decade of partnership effort. It consisted of interacting timelines of watershed modeling and wetland construction that included the following components: (1) a SWAT application for modeling cost-effectiveness of alternative ACP treatment scenarios, (2) demonstration sites of innovative wetland designs meant to optimize nutrient removal, (3) studying suitable locations for wetland construction using the ACPF tool, and (4) project assessment, which consisted of combining modeled loading with demonstration site specifics and/or project-specific hydrologic analysis to assess treatment potential. We combine this information with real implementation costs to arrive at an operational cost effectiveness for comparing projects. Finally, we return to the watershed scale to assess the feasibility of meeting nutrient reduction targets with wetlands. Feasibility is gauged based on realized construction costs relative to those projected from modeling. This paper discusses what we know to date about the realized cost-effectiveness of a decade of effort trying to implement constructed wetlands in the UEFW.
General timeline and conceptual linkages among the parallel tracks of watershed modeling, wetland construction, and the methods used to assess projects that culminate in estimates of operational cost effectiveness developed for this study. CIG = Conservation Innovation Grant; EOF = edge-of-field; ACPF = Agricultural Conservation Planning Framework; NRW = nutrient removal wetland; EFWA = East Fork Wildlife Area; EFRR = East Fork Riparian Reserve; and TMDL = total maximum daily load.
Materials and Methods
Research Approach. A SWAT application in a case study watershed is used to model the cost-effectiveness of different ACPs. These model results are presented as unit costs of nutrient removal at watershed scale. Along with model estimates of RC specific nutrient loads and proposed TMDL reduction targets, the unit costs are used to project total treatment costs using different ACP types. Constructed wetlands proved an attractive option. So, we set out to explore and demonstrate the performance of different wetland designs. During this time, we became aware of the ACPF tool and used it to identify potential locations for wetland construction in the watershed. After ground-truthing the ACPF results, 12 locations were specifically pursued for future edge-of-field builds and one other was built on using a hybrid design. Also, a large area of public land identified independently was evaluated for its feasibility for wetland conversion. Finally, for all the actual and potential build sites we returned to the SWAT model’s loading simulations to help characterize the cost effectiveness of the different wetland projects studied. The sequencing through time and interrelated tracks of research are diagramed in figure 1. This methods section is staged according to this sequencing of events. Note that the ACPF results were not directly integrated with the SWAT application in this study, largely because the SWAT application was developed before the ACPF was made available for broad use.
Case Study Watershed. The 892 km2 UEFW is the upper half of the East Fork of Little Miami River (EFLMR) watershed (figure 2a), occupying 1,300 km2 of the Pre-Wisconsonian Drift Plan Level IV ecoregion (Woods et al. 1998) in the southwestern corner of Ohio, United States. The UEFW drains to Harsha Lake, an 8 km2 reservoir that was built by the US Army Corps of Engineers (USACE) in 1978 for flood control, but also supplies source water and recreation.
Upper East Fork of the Little Miami River watershed (UEFW) maps: (a) SWAT subbasin watershed discretization with location of constructed wetland study sites (A: CIG, edge-of-field; B: Williamsburg, floodplain; C: EFWA; pump-and-treat; and D: EFRR pocket cells/channel restoration), and (b) watershed land cover patterns and distribution.
Harsha Lake is hypereutrophic with both N and P concentrations consistently exceeding state water quality targets for inland lakes (OhioEPA 2010). It has had an annual harmful algae bloom since at least 2008 (Smucker et al. 2021). Ohio Environmental Protection Agency (EPA) has designated the beneficial use of the lake as impaired under its Clean Water Act section 303(d) reporting requirements (OhioEPA 2014, 2021). Additionally, 50% of 51 stream and river sites surveyed in the UEFW are impaired for aquatic life (OhioEPA 2014). These impairments require TMDLs, which are in the final phases of development (OhioEPA 2020).
The watershed’s nutrient load is dominated by nonpoint sources. Row crop agriculture comprises 44% of the watershed land cover (ca. 422 km2) and accounts for an estimated 71% of the nutrient load (figure 2b), with soybeans occupying approximately two-thirds and corn one-third of the annual crop cultivation. This crop mix is a product of poorly drained soils, which can be a problem for early season planting. This favors the selection of soybeans, which can be planted later in the season. The soils are also prone to gully erosion and fragipans. Tile drains are not prevalent, nor are there many livestock operations. Cultivated pastureland occupies 16% of the land, and deciduous forest accounts for 30%. Low density residential land is 8% of the land cover. Soils in the UEFW are classified among 13 different types, but the poorly draining Clermont, Avonburg, and Rossmoyne classes are dominant (32%, 24%, and 21% of watershed area, respectively) based on the Soil Survey Geographic Database (SSURGO) (USDA NRCS 2018).
Cost-Effectiveness Modeling. SWAT was used to simulate the effectiveness of different nutrient management practices (Arnold et al. 2012; Santhi et al. 2014). The UEFW was discretized into 116 smaller subbasins to begin the SWAT set-up (figure 2a). A “base case” (BC) model was developed using a total of 20 land use classifications, 9 of which are agricultural crop rotations (2 two-year, 2 three-year, 1 four-year, and 4 five-year rotations) (Karcher et al. 2013). The multiyear rotations repeat over the course of the simulation. BC uses 13 soil classifications, three slope intervals (0% to 2%, 2% to 5%, and >5%), and includes a septic system classification in the land use layer. The land uses, soils, and slope characteristics are used by SWAT to derive unique hydrologic response units (HRUs) for each subbasin. With the HRU approach, hydrologic and nutrient fate and transport processes are spatially disconnected (Pignotti et al. 2017), but the relatively small sizes of subwatersheds and the number of landscape types combine in the application to the UEFW to produce output at the spatial scale similar to individual land parcels (Githui and Thayalakumaran 2011; Karcher et al. 2013). Wastewater treatment plant discharges are included in the model. A simulation begins January 1, 2000, after a three-year warm-up period and ends on December 31, 2018. The model was calibrated with data obtained from over a decade-long routine nutrient monitoring program (weekly water sampling) at a site on the river’s mainstem just upstream of the lake where the US Geological Survey (USGS) also maintains a stream gage (03246500) for continuous flow monitoring. More specifics on model set-up and performance can be obtained from OhioEPA (2021).
Model adjustments to simulate ACPs were made based on the SWAT literature, including the Wabash Water Quality Trading Report (CTIC 2011), among others (Arabi et al. 2007a; Waidler et al. 2011; Francesconi et al. 2015). The scenarios modeled included residue management, cover crops, filter strips, grassed waterways, constructed wetlands, and decreasing fertilizer applications on RC by 20%. All ACP scenarios were modeled separately by adjusting a version of the BC model with no tillage, as partners suggested this was a trending practice for farmers in the system. A separate modeling scenario for each practice was simulated watershed wide, which means that all the RC areas in each subbasin of the SWAT model that had them were effectively “treated” by each ACP type. The model output at the UEFW outlet for the BC no tillage and the BC no tillage with ACP scenario was differenced. We assessed the relative differences in terms of unit area responses of treating RC, i.e., kg harc−1 of total N and P removed. Combined with practice specific cost data (described in further detail below), we studied the relative cost-effectiveness of the six different alternatives in terms of US$ kg−1 removed. The modeling workflow was based on that developed in CTIC (2011), which used SWAT to conduct a nutrient trading market feasibility analysis for the Wabash River watershed, Indiana. We used SWAT 2012 ver. 664 (https://swat.tamu.edu/software/) (TAMU 2019) with the ArcSWAT interface (Olivera et al. 2006). SWAT-CUP facilitated the calibration, validation, and generated uncertainty distributions for the ACP simulations (Abbaspour 2015). We used the Sequential Uncertainty Fitting (SUFI2) algorithm for the analyses of uncertainty.
Uncertainty was represented by the distribution of effects on total N and P produced from running the model 100 times for each ACP scenario, with each run occurring under a different set of parameter values (Arabi et al. 2007b) (table S1). The approach assumes the modeled stream reach-specific effects (one reach per subbasin) are the same with and without ACPs. In all cases except for the constructed wetlands simulation, the effect on nutrient loading occurs for RC only. For constructed wetlands, the scenario was implemented on a subbasin scale, in-effect intercepting the HRU loads from all HRUs in the subbasin. The fraction of subbasin flow that is routed through the modeled constructed wetland was set equal to the fraction of RC area relative to the total, and their spatial extent was used to size the wetland using a ratio of 1:100, wetland area to RC area. This poses some inconsistency when normalizing across ACP types based on the subbasin contributing RC area (i.e., kg ha−1 y−1), which is assumed negligible given that RC dominated in most subbasins that contained them. A constructed wetland was not modeled in subbasins with no RC. Algorithm for modeling wetland plants directly was not available when the SWAT application was set-up and calibrated.
Cost data (2018$) for the ACPs were collected from the USDA NRCS Field Office Technical Guide website for Clermont County, Ohio (https://efotg.sc.egov.usda.gov). Agricultural conservation practice costs were tallied for the official EQIP practices reflecting the simulation scenarios including Residue Management, No-Till/Strip Till (Practice Code 329); Cover Crops (Practice Code 340); Filter Strips (Practice Code 393); Grassed Waterways (Practice Code 412); and Constructed Wetlands (Practice Code 656). As an example, table S2 provides a work-up of the cost estimate used for constructed wetlands, including adjustments for the foregone income from taking land out of production (Gordon et al. 2015). We also projected the future operations and management costs using a 3% inflation rate. Finally, we estimated the present value of the future costs using a 5% discount rate. The cost-effectiveness of each ACP type was calculated as the installation cost per hectare of RC treated divided by the modeled watershed-wide nutrient reduction per RC hectare when the ACP was added to the model to treat all RC areas. A median US$ kg−1 nutrient removal metric was used for comparing among ACPs, which was calculated from the estimated installation cost with and without the different opportunity cost estimates. One ACP is considered more cost-effective than another if the median cost of reducing a kilogram of N and/or P is lower. With the uncertainty analysis providing a distribution of nutrient reductions for each ACP type, we can study the cost-effectiveness at different reduction percentiles to be more or less conservative in the estimation of treatment efficiencies.
Conservation Innovation Grant Project: Innovative Edge-of-Field Wetland Project. The watershed partnership began focusing on constructed wetlands over a decade ago, when the Clermont County Soil and Water Conservation District received a Conservation Innovation Grant (CIG) (USDA NRCS 2019) to fund a modified urban stormwater basin design coupled with a subsurface flow wetland cell as a retrofit for an eroding grassed waterway originally funded by NRCS’s Conservation Reserve Program (CRP). The proposed design was innovative given that the low permeability and high clay content of the soils in the UEFW produce erosive stream flows in the headwaters, making poorly placed grassed waterways prone to failure.
The modified design consisted of an in-channel detention component linked to a subsurface-flow vegetated bed (i.e., a wetland cell) (figure 3a). The goal was to detain water using the existing active channel such that a fraction of it could be routed to the wetland cell for enhancing nutrient removal. The detention component and wetland cell were sized to hold runoff from a less than 2.5 cm, 1.44 month storm with a residence time of 21.7 hours. The 213 m long × 9 m wide detention area was produced with a check dam positioned just downstream from an in-channel riser (i.e., Hickenbottom weir) designed to intercept flow and route it through a buried pipe to the 122 m long × 3 m wide wetland cell for further treatment (figures 3b, 3c, and 3d). Sediment deposition and particle-bound nutrient removal was expected in the detention area, while the wetland cell would remove inorganic nutrients. Flows exceeding the detention capacity overflow the check dam, bypassing the wetland cell through the natural drainage channel for the small catchment (78 ha).
Conservation Innovation Grant (CIG) construction project layout and images. (a) Aerial view of catchment with treatment areas and monitoring site locations (stars). Yellow boundary is drainage for detention area, red boundary is additional drainage to wetland cell/bypass channel. (b) General layout of treatment area. (c) Image of detention area looking upstream from driveway/culvert. (d) Image of wetland cell looking downstream.
Construction of the CIG wetland was finished in 2014. Prior to construction, and beginning in 2011, stream flows were periodically measured, and nutrient species were sampled weekly. After construction, the channel flows could be continuously monitored upstream (sysIN) and downstream of the treatment system (sysOUT), and the wetland cell was isolated with influent (wcINF) and effluent (wcEFF) monitoring stations (i.e., four total). Monitoring stations were outfitted with area-velocity meters, or a bubble-type flow meter associated with a control device, for the in-channel and wetland cell sites, respectively, and had automated samplers for obtaining water samples for total suspended solids (TSS) and nutrient species during stormflow conditions. Nutrient species included total nitrogen (TN), nitrite-nitrate nitrogen (NO2-3), ammonium (NH4), total phosphorus (TP), and dissolved reactive phosphorus (DRP). The weekly grab sampling effort at the most downstream point (sysOUT) continued throughout. Approximately three years of continuous monitoring were used to determine the CIG project’s effectiveness. Seasonal baseflow averages for TSS and nutrient concentrations calculated from four years of weekly grab sampling data were assigned to the days when automated or grab sampling had not occurred (figure S1). Storm events impacting the system were captured by the automated sampling equipment.
ACPF Modeling for Edge-of-Field Wetland Construction Opportunities. Using grant funds from the US Fish and Wildlife Service (USFWS) National Fish Passage Program, the ACPF tool was applied in 11 of the most heavily RC dominated HUC12s in the UEFW. For the grant, we had proposed to use ACPF to identify small catchments suitable for constructing “edge-of-field” wetlands, which the tool refers to as nutrient removal wetlands (NRWs) (Tomer et al. 2013b, 2015) (https://acpf4watersheds.org/about-acpf/). ACPF uses digital elevation models and other geodata layers to place structural ACP types on the landscape (Porter et al. 2018). For the UEFW HUC12s, the following parameter values and criteria were input for determining potential locations for NRWs: site outside the existing stream network, a maximum drainage area threshold of 34.4 ha, a minimum spacing distance of 150 m between potential site locations, a minimum impoundment height of 0.9 m and buffer height of 1.5 m, a pooled area to drainage area ratio between 0.5% (1:200) to 2.0% (1:50), and buffered area to pooled area ratio <4.0. Note, the ACPF tool did not become more broadly applicable until several years after our original cost-effectiveness modeling was completed.
Williamsburg Off-Channel, Floodplain Wetland Project. Concurrent to exploring wetland sites in small catchments, floodplain terraces along the EFLMR mainstem were also explored. Floodplain wetlands have the potential to intercept flow from upstream and provide treatment of excess nutrients from agricultural runoff and other sources. However, the unit-level treatment efficiency of these systems is likely lower than the CIG wetland and the ACPF NRW type systems due to the limited residence time of a large portion of the hydrograph volume. The partnership identified a floodplain parcel owned by the town of Williamsburg for further focus. Approximately 61,200 ha of the UEFW drains through the river corridor at this location. In the 1950s, the Williamsburg property had a 1.21 ha impoundment constructed by adding an earthen berm within the floodplain to hold water pumped from the river prior to treatment for drinking water. The impoundment had been abandoned decades ago. The watershed partnership had been working with the USFWS and the Ohio River Basin Fish Habitat Partnership to design and construct off-channel (or floodplain) wetlands for riverine fish habitat improvement and offloading flows that would otherwise contribute to riverbank erosion, when some potential synergies were recognized. A proposal to the H2Ohio program included constructing an off-channel wetland optimized for nutrient removal at the Williamsburg site.
Wetlands situated in floodplains that are connected to the adjacent channel have long been considered critical habitats for riverine productivity and biodiversity (Junk et al. 1989; King et al. 2003). The USACE lists off-channel (floodplain) wetlands among options for restoring wetlands under its §404 permitting process (USACE 2010; USEPA 2015). These systems are typically designed to have a bi-directional connection with the river, filling when river levels increase under stormflows and draining when levels recede. Because they are excavated into the floodplain terrace, flood waters are retained within permanent pools providing critical habitat for aquatic species. While potentially beneficial for wildlife and reducing riverbank erosion (Hawley et al. 2023; Dierauer et al. 2012; Guida et al. 2015; Asselman et al. 2022), their structure and hydraulic properties had not been previously geared toward nutrient removal. The Williamsburg project was justified as an experimental effort to innovate the off-channel wetland concept for enhancing nutrient removal.
East Fork State Park and Wildlife Area Wetland Feasibility Study. In 2022, the Ohio Department of Natural Resources (ODNR), Division of Parks and Watercraft, secured H2Ohio funding to consider a potential water quality project in a large area of public lands situated in a high bank where it empties into Harsha Lake. The East Fork Wildlife Area (EFWA) is owned by USACE but is leased under a managerial agreement to the ODNR. It consists of 1,095 ha, including 11.27 km of the EFLMR mainstem. The ODNR’s Division of Engineering’s bid request called for a formal feasibility study of the EFLMR at its confluence with the lake. It was awarded to Coldwater Consulting LLC, Galena, Ohio, who, in turn, included watershed partners and their data to inform the effort (ODNR 2022). The objectives of the feasibility study included determining design alternatives, estimating project costs, and nutrient removal benefits for (1) the construction of floodplain wetlands in the meander bends of the EFLMR just downstream of the Williamsburg project but within the USACE property boundary and (2) the conversion of fields managed for dove habitat to flow-through wetland treatment cells.
The effort included analysis of geodata for soils, natural wetlands, flood maps, and a digital elevation model (Clermont County, Ohio Geographically Referenced Information Program) to map the existing topography of the region, identify alternative catchment treatment areas, grading requirements, and determining the general hydrologic patterns of the area under consideration. Water data, including river flows from the USGS gage noted above from June of 2017 to June of 2022, reservoir elevations from the USGS gage on Harsha Lake (https://waterdata.usgs.gov/monitoring-location/03247040/), and nutrient concentrations from grab sampling effort of the watershed partners were used to assess the feasibility of constructed wetland permeance, inform the placement of pump works, and to estimate nutrient reductions.
Cost-Effectiveness Comparisons Among Projects. Given that only the CIG demonstration study had real measures of nutrient reduction, we use an alternate definition of cost-effectiveness to compare sites. We calculate an operational cost-effectiveness for this study as the unit area costs of treating runoff from RC in the UEFW based on the construction costs (real or estimated) divided by an estimate of the equivalent treated RC area (i.e., US$ ha−1). The equivalent treated RC area was calculated from the fraction of the total RC hydrologic load (HL) for the watershed (i.e., 2.14 × 108 m3 y−1) estimated to be intercepted by each site studied and the total RC area in the UEFW (equation 1):
1
The estimate for intercepted RCHL was obtained from the UEFW SWAT model’s HRU-level output for the edge-of-field NRWs in the small agricultural catchments or was based on the RCHL fraction of total HL for the EFLMR mainstem that is offloaded or that would be pumped directly to the wetland(s) at the Williamsburg and EFWA sites, respectively. This normalization approach assumes that all RCHL intercepted is treated equally among the site’s wetland designs, which could not be validated without direct monitoring. The normalization scheme provides a means of making a direct comparison among studied build sites and can be related to the original simulations of the watershed model at the 50th percentile of the uncertainty distribution (i.e., the median) under the same assumptions.
We used the same assumption of equivalency in treatment efficiencies among the sites that were studied for each site’s estimated intercepted RCHL. We also estimate the percentage of watershed-scale P treatment potentially accounted for by the project. This calculation uses the equivalent RC area treated by the build and the RC area that would need to be treated to meet the watershed P reduction requirement established by the UEFW TMDL. This estimation also assumes that all sites receive the same annual mass of P per unit volume of RCHL.
Results and Discussion
Modeled Cost-Effectiveness of Agricultural Conservation Practices. The cost-effectiveness modeling results among ACP alternatives provided rationale for focusing on the implementation of constructed wetlands for nutrient reduction in the UEFW (figure 4). When interpreted within the context of meeting the TP and TN reduction requirements that will be set forth by the Harsha Lake TMDL (P. Gledhill, Ohio EPA, personal communication), the unit cost differences among ACPs modeled are relevant (table 1). Only constructed wetlands appear feasible based on the potential for meeting both N and P reduction requirements (227,107 and 50,553 kg y−1, respectively), the extent of RC in the UEFW (42,211 ha), and the 50th percentile treatment efficiency (24.5% or 1.40 kg ha−1) (see cell background colors in table 1 and the description in the legend). Conservatively, using the 5th percentile treatment efficiencies for implementation planning to meet the TP requirement, a combination of practices is required. For example, the lowest cost combination for TP would require maximizing constructed wetland area (422 ha), as well as filter strip area (2,110 ha), and would still require adding 870 ha of grassed waterways. Based on these modeling results, the watershed partnership focused its attention on opportunities for constructing wetlands. The partnership used the 361 ha required under the 50th percentile treatment efficiency to meet the TP reduction under the TMDL as a planning-level target, which is the equivalent of needing to treat the hydrologic load from 85.5% of the RC area, or 1.83 × 108 m3 y−1 and with a projected cost of ≈US$2.4 million per year over 20 years (table 1). The rationale for focusing the planning effort initially on wetlands was validated with realized outcomes from the CIG project.
Unit costs of total phosphorus (TP) and total nitrogen (TN) removal for the watershed wide model simulations of agricultural conservation practice (ACP) effects. The 5th, 50th, and 95th percentiles of the distribution of effects are plotted separately. 422 and 2,109 ha of wetlands and filter strips implemented, respectively. The 5th centile TN removal is absent for residue management because it was negative.
Agricultural conservation practice (ACP) implementation areas and costs to meet the draft total maximum daily load (TMDL) nutrient reduction goals in the upper East Fork of the Little Miami River watershed (UEFW). Data highlighted in red are infeasible; yellow highlight is feasible but only for one nutrient; and green highlight is feasible for both nutrients. Feasibility is based on existing area of row crops and modeling ratios used for ACP area to row crop area treated for structural practices (existing row crop area = 42,211 ha).
Conservation Innovation Grant Edge-of-Field Wetland Project Evaluation. The CIG project had an original construction cost of US$26,000.00. The area of the treatment system (detention area and wetland cell) as built is 0.20 ha and the contributing drainage area is 77 ha, resulting in a lower wetland to drainage area ratio than that used to size the modeled wetlands in the UEFW SWAT application. The check dam that creates the detention area has had to be repaired three times since construction. It was constructed using rip-rap sized stones (10 to 15 cm). We add an additional US$4,000.00 to the construction cost to account for the addition of a sounder check dam. The bulk of the cost for this project came from the medium aggregate (≈5 cm) used to line the wetland cell, adding more than was projected using the EQIP costs. Grading and excavating of the area were done by the farmer under a contract with Clermont SWCD. The actual costs were reasonably close to the projected costs used for pairing with the UEFW SWAT modeled treatment efficiencies, assuming that minimal operation and maintenance and that no replacements are necessary for 20 years (table S2).
The monitoring data were used to calculate cumulative system loadings over the nearly three-year observation period beginning December 1, 2014, and ending October 17, 2017 (figure S1). With the four locations monitored, removal estimates for the whole system (i.e., detention area plus wetland cell) and the wetland cell component were made separately (table 2). Of the total hydrologic load from the catchment, an estimated 26% of it passed through the wetland cell. For the monitoring period, the system removed 51%, 28%, and 37% of the TSS, TN, and TP load, respectively. A relatively low removal of P in the wetland cell was observed, but this was offset at the system scale by higher removal in the in-channel detention area The estimated annual flow through the system was 302,283 m3 y−1, which was 74% of the SWAT-modeled annual average for the subwatershed. The modeled hydrologic load from RC specifically was 81.5% of the subwatershed total, which equates to 0.18% of the annual hydrologic loading from RC that would need to be treated by constructed wetlands to meet the watershed TMDL TP target.
Constituent loads removed by the Conservation Innovation Grant (CIG) project, calculated as the difference between the cumulative load estimates over the 2.88-year monitoring period. Data in parentheses are load removal percentages relative to the total load on the system or wetland cell. System removals are unit area normalized to account for differences in drainage size between monitoring locations. Negative values indicate no removal.
ACPF for Edge-of-Field Wetland Construction Opportunities. The ACPF analysis identified potential locations for 61 NRWs. Ground-truthing these locations with land cover and aerial photogrammetry revealed that only 12 were sited in locations that would replace existing crop land or occurred at a CRP/forested edge. The rest were sited in low-lying forested areas. Among the 12 areas considered for build potential, ACPF estimated wetland-specific areas that ranged from 0.2 to 2.1 ha and drainage areas between 35 and 109 ha (table 3). The average treatment to catchment area ratio for the 12 systems is 1:75. The estimate for the total 20-year cost of constructing and maintaining these systems is US$891,016.90, using data from the EQIP-based cost analysis (table S2). A map of the ACPF NRW wetlands with an aerial of a specific example of 1 of the 12 systems considered appropriate is shown in figure 5. The total hydrologic load from RC that the ACPF wetlands would intercept is 2,383,449 m3 y−1. This is 1.31% of the annual RC hydroload treatment requirement. With the 12 sites identified, the watershed partnership sought approval by property owners to build because they were on private land. Following interviews with property owners for potential interest, none were interested.
Nutrient removal wetlands identified by the Agricultural Conservation Planning Framework (ACPF) analysis and screened for appropriateness as potential build sites.
Map of nutrient removal wetlands (NRWs) deemed appropriate for further consideration for mitigating row crop impacts with example of aerial photogrammetry used to ground truth the placement.
Williamsburg Site (Floodplain Wetland). The design objectives for optimizing the Williamsburg system included maximizing the number of events that activate inflow to the system, incorporating the existing impoundment to provide detention storage (figure 6a), and maximizing residence time of captured inflows and the opportunities for interaction with biogeochemically active areas for the flows entering and detained by the system. This last objective was accomplished by integrating a low gradient meandering wetland channel component.
Williamsburg wetland (a) preconstruction, (b) engineering drawing, (c) aerial image taken November 1, 2022; construction nearly completed.
Detention storage at the Williamsburg site was achieved by (1) adding a 60.9 cm diameter piped connection to the inlet forebay area through the old reservoir’s berm that was outfitted with a flap-gate valve that allows flows to enter but not exit via the same pipe; (2) excavating the bottom of the old reservoir further and creating an elevated peninsula meant to remain mostly dry with woody vegetation and to discourage short-circuiting of flows, and (3) adding a 15.2 cm diameter orifice-controlled pipe at the side adjacent the inlet. The orifice controls the rate of release from the reservoir to the meandering wetland channel, which the partnership has taken to referring as the “gut,” denoting that its role in the design of the system is similar anatomically and physiologically to an animal’s gastro-intestinal tract, fostering microbiologically mediated processes important to nutrient removal over a larger surface area (figure 6b). The inlet/forebay area was constructed to route flood flows to both the reservoir and the gut to expand opportunities for sedimentation and nutrient assimilation even after the reservoir water level has reached an equilibrium with the river and can no longer accept additional volume for storage and treatment after the event. The forebay and reservoir design included permanent pools for fish habitat, and it is anticipated that most of the areas, except the permanent pools, the tree peninsula of the reservoir, and the narrow thalweg and high banks of the gut, will colonize with emergent marsh vegetation. Construction of the Williamsburg wetland was completed in January of 2023 (figure 6c).
Once the design was finalized, separate approaches were taken to estimate hydraulic loading and residence time. Both relied on the continuous flow data from the USGS gage. One approach included spreadsheet calculations relying on inlet elevation, a return flow rate to the river limited by the channel slope and assumed roughness of the meandering channel section using the Manning’s equation. The other was a continuous simulation approach using an application of the Storm Water Management Model (Niazi et al. 2017) to model the hydraulic performance in terms of flow-through rates, water elevations, storages, and residence times in the three interconnected components (personal communication, J. Dorsey and M. Stoltzfus, The Ohio State University, Columbus, Ohio, May 18, 2022). We assume that all flows entering storage eventually drain through the meandering channel and that all 24 events activating the inlet produce a 0.86 m3 s−1 flow (given Manning’s flow constraints) through the channel for one day and received at least some treatment, which equate to 1.78 × 106 m3 y−1. The total annual hydrologic load treated by the Williamsburg project, therefore, is estimated to be 1.86 × 106 m3 y−1. Based on the UEFW SWAT model results, 58% of the total hydrologic load in the EFLMR that passes the Williamsburg site is attributed to RC runoff. We apply this percentage to estimate the RC hydrologic load that is off loaded to the Williamsburg system (i.e., 1.09 × 106 m3 y−1). The Williamsburg system is estimated to intercept 0.59% of the total RC hydrologic load treatment goal. The final construction cost was US$737,910.00 and given the novelty of the Williamsburg treatment system design, estimates for operation and maintenance are not available. The system was outfitted with flow and water quality monitoring equipment in February of 2023.
Second Construction Site Under the H2Ohio Grant: Hybrid Design. One among the 61 wetlands sited with the ACPF tool was within the East Fork Riparian Reserve (EFRR): a parcel donated to the Clinton County Parks district in 2004 (figure 7a). This was excluded from the original analysis because the ACPF NRW would have replaced existing forest. However, after a site visit and a study of the surrounding area, build potential was recognized for the Reserve’s lands. Its southeastern section (figure 7b) could have been designed as a large off-channel wetland system like the Williamsburg site. However, budget constraints demanded a different approach.
East Fork Riparian Reserve site (a) Agricultural Conservation Planning Framework Nutrient Removal Wetlands (ACPF NRW) drainages (yellow) and the drainage for the pocket wetland (blue); (b) reserve aerial image yellow lines are hiking paths; and (c) engineering drawings for stream reach restoration and six off-channel, pocket wetlands in the floodplain. Construction underway June 15, 2023.
A 315 ha RC dominated catchment drains through a small stream that enters the EFRR on the eastern side. Excessive runoff from the drainage’s RC fields was creating chronic streambank erosion, resulting in an entrenched channel with disconnected floodplain. This reach became the focus of the construction project. The treatment system design represents an offtake of the off-channel wetland concept by reconnecting the channel to the floodplain and routing water into six pocket wetland cells, each with separate bidirectional inlets, and restores 274 m of the entrenched channel with riffles, pools, log vanes, gentle banks, and native vegetation (figure 7c).
Combined, the pocket wetlands will occupy 0.28 ha of the floodplain terrace (~1 ha), with the remaining areas comprised of a restored meandering stream and reconnected floodplain. The pocket wetland cells will be excavated into the terrace for a depth range 0.46 to 0.61 m, and each with wintering holes for fish habitat between 1.22 and 1.83 m deep. Each of five separate inlets will be connected to the restored stream at ~15 cm above the downstream riffle’s crest. Given the dimensions of the restored channel this translates to inlet activations at roughly 5% of the estimated two-year peak discharge. It is expected that multiple storm events per month would offload to the pocket wetlands. For flows between 0.25 and ≈1.7 m3 s−1, the hydrologic load to the wetlands will only enter and exit through their designed connections. Flows above approximately 1.7 m3 s−1 will start accessing the reconnected floodplain terrace, traversing the entire valley floor. With this design schema, estimating hydrologic load intercepted by the EFRR wetland/floodplain treatment system is not straightforward and proved beyond scope. However, from the UEFW SWAT model, we estimated that 1,025,989 m3 y−1 of RC hydrologic load passes through the restored stream section, representing 0.56% of the annual RC hydrologic load treatment requirement for the watershed. An undetermined fraction of this percentage would be intercepted by the pocket wetland/floodplain terrace component of the project. The EFRR project is under construction at the time of writing and had a combined design and construction budget of US$402,901.00.
East Fork Wildlife Area Feasibility Study. Steep floodplain topography along most of the riverine-lacustrine transition section was found to limit the construction of any wetlands immediately adjacent to the river. However, converting the wildlife management fields to wetlands was considered a technologically achievable option for the EFWA. Given their elevation relative to the river, electrical pumping for water delivery would be required. River flow and nutrient grab sample data were used to support the consideration of two pump station alternatives for delivering water: one at an upstream location in the river, which would intermittently route comparatively smaller volumes of river flows somewhat more concentrated in nutrients to the converted fields, and the other downstream where the river has transitioned to lake conditions, which would allow for continuous pumping of a larger volume from the lake’s permanent pool. Water delivery from the river intake location would be limited by availability when the river is under low flow conditions in the late summer months.
The study determined that 21 fields exist for potential conversion to wetlands. These, along with the studied intake locations, proposed routes of the force mains for water delivery, and directions for gravity flows are shown in figure 8. To allow for a phased construction approach, three different layout options—one with three suboptions—were configured and costed separately based on different field arrangements and the amount of wetland construction that would be needed. Each option was organized such that water would be delivered to the field with the highest elevation (F6 and F16 in figure 8) and then would gravity feed through a series of wetland terraces excavated in stair-step fashion down gradient to ultimately discharge back to the main river channel (figure S2). The study estimated costs for a range of pump station sizes and wetland cell configurations to achieve an approximate 48-hour water residence time. Estimates of cost-effectiveness were made assuming TN and TP removals from searching the literature of 40% to 60% and 50% to 70%, respectively. Here we focus on the hydrologic loading capacity of the construction alternatives and their respective costs for comparison with the other presented projects (table 4).
Existing wildlife management fields (F#) considered for conversion to wetlands in the East Fork Wildlife Area feasibility analysis. Pump station location A draws water from the active river channel. Location B would be constructed in the zone where the river has transitioned to lake conditions with a permanent pool that would allow for continuous pumping all year long. F14 to F18 and F20 and F21 are located on the opposite side of the river. F1, F18, F19, F11 to F13, F20, and F21 were not feasible for a gravity flow/terraced design.
Construction options determined from the East Fork Wildlife Area wetland feasibility analysis.
Results suggest that site access challenges of the lake intake (Loc B) are offset by the greater volume of water treatment and result in a comparable cost-effectiveness for both pump station locations. If all fields were converted to wetlands, the percentage of the annual river flow intercepted and treated would be 6.9% and 11.0% for pump station locations A (river) and B (lake), respectively. With the same assumptions about watershed RC hydrologic loads as above, EFWA wetlands could intercept 6.8% for pump location A or 10.9% for pump location B, which would come at an estimated cost of US$29,951,921 or US$37,560,366, respectively.
Realized Cost-Effectiveness. Now that we have more thoroughly identified, studied, and/or built out 18 constructed wetland projects in the UEFW, actual or more realistic costs can be reported within the context to the original planning-level estimates from the model simulations. We find that build sites in small RC catchments would be more cost-effective per RC area treated than the innovative off channel build site at Williamsburg and the much larger treatment system requiring active pumping near the lake (table 5). The EFRR project could be considered a hybrid between the edge-of-field NRWs placed by the ACPF tool and Williamsburg project. However, because we are not able to provide a reasonable estimate of the intercepted and detained load at this time, we can only say that such a system is likely to be less cost-effective than small wetlands that can be built directly in the path of runoff. The pump and treatment systems configured for the wildlife area near the lake appear on the order of 3 to 10 times less cost-effective than constructed wetlands in the small catchments.
Cost-effectiveness of constructed wetland alternatives studied.
From the UEFW-SWAT constructed wetlands simulation, an estimate of US$2,425,252.00 was considered to achieve the TP load reduction required under the UEFW TMDL. The watershed partnership has used US$1,048,581.00 in grant funds to construct 3.785 ha of wetland-based treatment systems. An additional ~US$38 million could be spent, while not appearing as a cost-effective expenditure, and another ~US$1 million would be needed if the reluctant landowners could be convinced to support the 12 projects identified as good candidates for wetland treatment systems. With this total of nearly US$40 million, we estimate that only 15% of the original target area and 13% of required treatment need could be achieved. The average annual cost estimate is roughly US$2 million, assuming all systems built would meet design performance for 20 years with minimal extra expenditures. We caution that a direct comparison of the planning-level cost and the six projects’ cost is not appropriate due to the different estimation approaches. However, given that only a small fraction of the need appears achievable at this time, it seems that the planning-level assessment largely underestimated the cost-effectiveness of constructed wetlands when the realities of finding suitable build sites and designing well performing systems are more challenging than anticipated.
Discussion. This watershed-scale analysis is unique. It assesses the practicalities for mitigating excess nutrients from RC by including results of modeling and on-the-ground realities experienced in the field while trying to place, design, and build constructed wetlands. The decadal history of the watershed partnership’s efforts to implement constructed wetlands for nutrient reduction in the UEFW reflects a framework for assessing feasibility that includes an upfront planning-level modeling analysis of wetland cost-effectiveness at watershed scale that should be informed by a process for identifying sites with real construction potential at the field scale. The efforts also show that building demonstration sites with measures of individual installation-level performance and detailed feasibility studies focused on gauging cost-effectiveness to meet watershed nutrient reduction targets is critical for further acceptance and support from partners, funding agencies, and stakeholders. We use the reported results of our experience to discuss these components.
Modeling exercises are necessary to gauge potential for constructed wetlands. In this study, the modeling objective was to compare the cost-effectiveness of several EQIP-type ACPs and to use the results to guide prioritization, which was akin to more recent studies like Uniyal et al. (2020), Liu et al. (2019a), Zammali et al. (2021), and Hansen et al. (2021). For the results to be meaningful to TMDL-type decisions (i.e., HUC12 watersheds and larger in Ohio, for example), modeling packages geared toward simulations at larger scales are necessary. We turned to SWAT for this purpose (Kalcic et al. 2018; Liu et al. 2019b; Yuan and Koropeckyj-Cox 2022). We found SWAT simulations to be too coarse for constructed wetland scenarios because wetlands are modeled in the program at the scale dictated by the subbasin delineation. The simulation of effectiveness was overestimated because watershed delineation is done without knowledge of where constructed wetland build sites are feasible. Even with the adjustments to the SWAT algorithm that allow for HRU-level wetland simulation (Evenson et al. 2016) and based on our experiences in the UEFW, a practical modeling scenario for cost-effectively constructing wetlands is not easily developed. The costs we included in the model were meant to reflect edge-of-field building sites on private agricultural lands, but we learned that these are not linearly scalable to other types of wetland treatment settings. Furthermore, we learned that these designs had to accommodate much larger flows, absorb costs associated with permitting processes such as archaeological, endangered species, and stream restoration, or account for electrical pumping and water distribution piping. As others have noted, farmer engagement, interest, and acceptance are crucial (Zammali et al. 2021; Sharpley et al. 2009; Kalcic et al. 2014; Prokopy et al. 2019). For the UEFW to make significant progress to meeting the TP reduction target, building constructed wetlands in the agricultural uplands to capture RC runoff proximal to its source would appear to be necessary. This is especially true if constructed wetlands are going to be a major player among alternatives for reducing nutrients in the system. Even if we based our cost-effectiveness calculations on the 95th percentile TP removal efficiency (i.e., 41%), we would still only meet 25% of the reduction requirement with all the wetland builds that we considered.
The ACPF tool proved a useful means of identifying potential build sites. ACPF is freely available and being used throughout the Midwest (Ranjan et al. 2019; Tomer and Nelson 2020; Duncan et al. 2021; Respess and Duncan 2021). We suggest that exploring potential wetland locations provided by ACPF results can be advantageous prior to performing subbasin discretization in SWAT for modeling constructed wetlands. In our case the opposite was done because ACPF was not available when we conducted our cost-effectiveness modeling (figure 1). This made it difficult to assign hydrologic and nutrient loads to the applicable NRWs because many of the NRW catchments overlapped UEFW SWAT subbasin boundaries. Likewise, without re-discretizing the watershed, the integration of the NRWs with the existing UEFW SWAT application for direct simulation is not straightforward. Indeed, this re-discretization is an ongoing effort in our partnership. Given the popularity of both SWAT and the ACPF, guidelines for their integration would likely provide more realistic watershed-scale nutrient reduction simulations in the future—akin to Gordon et al. (2021a) but with the results of ACPF more directly integrated in an application of SWAT. An important caveat, however, comes from our lesson learned that only a small fraction of those sited in the uplands of the UEFW (i.e., ~20%) were considered suitable build sites upon further study, and none of these could be implemented because of reluctance among landowners.
Another lesson learned was the importance of constructed wetland demonstration sites to achieve further buy-in and to validate model performance projections. It was the cost-effectiveness of the CIG project on a private farm that spurred the use of the ACPF tool in the UEFW. The innovations in design that were proposed and implemented to counteract the limitations posed by the poorly draining and erodible soils common in the watershed made partners more interested in the notion of subsequent installations and made the prospects more palatable to funding agencies. Given the measured cost-effectiveness of these designs, especially relative to the other wetland alternatives that are in progress in the watershed (table 5), revisiting opportunities in small agricultural catchments should be a priority. The ACPF tool can be configured to site water and sediment control basins (WASCOBS), grassed waterways, and farm ponds as well (Porter et al. 2018). With slight modifications, these alternatives can be configured to encourage the same nutrient removal processes that make wetland systems so attractive. A case in point is the CIG project that replaced a failing grassed waterway design. Therefore, it is rational to return to the ACPF tool for exploring options for additional projects based on these other alternatives in the UEFW agricultural catchments. With the estimates provided in tables 3 and 5 from this study, the total costs could be on par with that from the original watershed-wide SWAT analysis (i.e., ~US$2.5 million annually), but it would require on the order of 300 individual installations given the averages obtained from the 12 NRWs studied and their cost-effectiveness.
Based on the fraction of watershed runoff from RC that are expected to be intercepted by the Williamsburg project, our second demonstration site, its cost-effectiveness is projected to be about half that realized by ACPF-type wetlands (i.e., twice the cost to treat a hectare of RC). Even if funds were available, numerous installations would be necessary and given their orientation along the river corridor, upstream sites would “steal” efficiency from downstream ones. However, sites like Williamsburg can provide critical habitat for fish species and other wildlife, act as sediment sinks, and work to dissipate peak flows and reduce flood stages as additional benefits not necessarily realized by those smaller systems built to receive runoff directly from RC (Hawley et al. 2023). Because it receives loads from the river’s mainstem, the Williamsburg system will treat nutrients from other sources in the watershed, including legacy sources. Indeed, Hansen et al. (2021) found wide, slow-flowing, vegetated wetlands within the riverine corridor to be the most cost-effective management for meeting targets related to NO3− and sediment in a modeling effort for a large basin in Minnesota, United States. In this regard, the design being constructed at EFRR is a hybrid. Estimating its cost-effectiveness hinges on understanding the hydraulic loading rates and actual detention times. This will be an objective of the partnership going forward.
Lastly, the lessons learned from the EFWA feasibility analysis may improve designs in this study and elsewhere. One challenge is the capital costs associated with pumping water to overcome a relatively large hydraulic head and distance from the main channel. Because of these factors, it is five times more costly to treat a hectare of RC compared to that projected for the ACPF NRWs and actualized by the CIG installation. The availability of the large area of relatively easily convertible land led to the interest in exploring opportunities in the EFWA. Like the Williamsburg wetland, EFWA wetlands would treat other sources of excess nutrients in the watershed. Such a conversion was used to help improve water quality at another distressed lake in Ohio: Grand Lake St. Mary’s (Jacquemin et al. 2022). For Grand Lake St. Mary’s, the relief of the landscape surrounding the lake is much lower, making the infrastructure required for electrical pumping less costly. The added cost-effectiveness afforded by being able to pump continuously from Harsha Lake in the EFWA analysis suggests that adding a pumping component to the Williamsburg project could improve treatment capacity and overall cost-effectiveness. Indeed, the cost-effectiveness of Williamsburg is primarily constrained by the number of events that offload to the system. With the old reservoir built into the design, realizing its full potential for nutrient reduction may not be possible without adding a pumping component. Given its proximity to the river channel with small hydraulic head, pumping costs should be a fraction of those projected for the EFWA wetlands. How pumping at Williamsburg could change the cost-effectiveness calculation would be dependent on the balance between pumping rate, the retention time afforded by the built-in reservoir, and not over withdrawing from the river to upset baseflow conditions.
Summary and Conclusions
The cost-effectiveness of ACPs for nutrient reduction was modeled for the UEFW using an approach that normalized simulated reductions of N and P to unit area of RC, the dominant source of excess nutrients at the watershed scale. This allowed for parity in comparisons among practices and offered estimates for budgeting nutrient management to meet TMDL requirements for a reservoir impacted by harmful algal blooms. Constructed wetlands proved the most cost-effective among the simulated practices. This led to a decision by the watershed partnership to place attention on building wetlands in the system, which was bolstered by the measured cost-effectiveness of an innovational edge-of-field wetland design built in a small agricultural catchment. It is not that other ACP types and opportunities were not being pursed through the EQIP program as part of normal NRCS business. They were, but the focus of the partnership was on constructed wetlands and the TMDL requirement provided an ambitious target to shoot for.
As the partnership began considering build opportunities, the group recognized success depended on a large-scale effort, hundreds of hectares would need to be converted to reduce nutrient reduction set by a TMDL exercise. Areas in the agricultural uplands were evaluated using the ACPF tool to identify potential build sites. At the same time, lands along the mainstem river were studied. Because the projected effectiveness for nutrient reduction of floodplain wetlands was suspect, owing to their lack of ability to capture low flows in the absence of pumping and low residence time during large flows, a second demonstration site was proposed where a different innovative design could be tested. Additionally, the large concentration of public lands currently under wildlife management and near the reservoir was also identified for a formal feasibility study.
Place-based ground truthing and the direct expenditures that have resulted showed that there were far fewer build opportunities than were implicitly assumed by the SWAT modeling approach and that were considered supportive given the criteria used by the ACPF to identify wetland sites. Implementing an integrated approach of multiple ACPs via ACPF within the cost-effectiveness modeling framework is needed. Furthermore, a multipronged wetland modeling approach, similar to Hansen et al. (2021), is also needed to help characterize the potential relevance of systems constructed within the river corridor. Wetlands are only one part of a more holistic watershed approach needed to achieve the UEFW nutrient reduction targets. Many edge-of-field ACPs (some more or less palatable to different farmers/stakeholders) are needed in addition to an improved in-field management.
Although the innovative off-channel wetland design in this study appears less cost-effective than the smaller catchment builds, the innovation that came from a focus on nutrient reduction brings added benefit to this wetland design primarily selected for habitat improvement and peak flow control. Finally, while a large area on public land with fields that would be relatively straightforward to convert to wetlands was deemed feasible, the water delivery infrastructure made this the least cost-effective among the alternatives that were studied. Although invaluable for supporting watershed management decisions early on, the watershed-scale cost estimate for treating excess nutrients in the UEFW with constructed wetlands appeared underestimated given the realities of reluctant property owners and physiographic constraints on effective designs.
Based on over a decade of lessons learned from trying to implement constructed wetlands at a scale relevant to meeting a TMDL target for a large agricultural watershed in Ohio, it would have been wiser to explore opportunities for placement prior to conducting the watershed-scale cost-effectiveness modeling. Such a strategy would have made the planning and budgeting process more realistic. With a tool like ACPF, this is now possible. However, had we not prioritized constructed wetlands early on, we may not have been as adamant about optimizing their designs for nutrient reduction. This history led to innovations that have the potential to increase the nutrient reduction effectiveness of more typical edge-of-field and off-channel wetland designs in this and other future watershed mitigation projects.
Supplementary Material
The supplementary material for this article is available in the online journal at https://doi.org/10.2489/jswc.2024.00077. Upon publication, all observed field data and model output used in results will be made available at the Environmental Protection Agency Science Hub repository at https://catalog.data.gov/dataset/UEFW-ConstructedWetlands_MSDataFile_17Jul2023.
Acknowledgements
The list of contributors to the constructed wetlands effort in the East Fork of the Little Miami River watershed (UEFW) is quite long and difficult to make complete. None of the on-the-ground efforts would be possible without Jacob Hahn and Susie Steffensen (Clermont County, Soil and Water Conservation District) and Laura Lair and Bell Mellman (Clermont County Office of Environmental Quality). Lori Lenhart and Steve Anderson (formerly USDA NRCS and Farm Services Association, respectively) were essential for stakeholder engagement early on; Steven Reese (Hazen and Swayer) for his support on the CIG project design; William Carlson (Tetra Tech) for conducting the original ACPF analysis; Jay Dorsey and Matt Stoltzfus (The Ohio State University) for the SWMM modeling analysis based on the Williamsburg construction design that provided the hydraulic loading estimates used in this analysis; Dana Macke (USEPA) for leading the routine field sampling in the UEFW; Donnie Knight (USFWS) for promoting the innovative off channel wetland design idea; and Nora Korth, Kurt Cooper, Abi Raetz, Peter Tower, Shelby Acosta, and Katie MacMannis (Sustainable Streams) for their design and permitting assistance for the Williamsburg and EFRR wetlands. USEPA reviewers Tammy Newcomer Johnson, Ken Forshay, Mike Elovitz, and Susan Cormier provided constructive edits and comments on early versions of the manuscript. The bulk of wetland site locating, design, and construction was funded by Conservation Innovation Grant NRCS 69-5E34-11-037, USFWS Fish Passage Grants F20AC00088 and F21AS00413, USFWS Reservoirs Grant F21AS00513, ODNR H2Ohio Grant 6H20 725681 and Ohio EPA 319 Grant 21(h) EPA-11. This article has been reviewed by the Center for Environmental Measurement and Modeling and approved for publication. Mention of trade names or commercial products does not constitute endorsement or recommendation for use by the US Government. The views expressed in this article are those of the authors and do not necessarily represent the views or policies of the US Environmental Protection Agency.
- Received September 2, 2023.
- Revision received November 20, 2023.
- Accepted January 28, 2024.
- © 2024 by the Soil and Water Conservation Society
References
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.