Pathogen reduction co-benefits of nutrient best management practices

Step 1: Define BMPs in Chesapeake Bay TMDL implementation

To define the types and extent of BMP implementation proposed to meet the TMDL, we used information generated by Chesapeake Bay Watershed Model scenarios developed for and run by the US EPA Chesapeake Bay Program (CBP). The scenarios are based on the Phase 2 WIPs provided by all of the Chesapeake Bay jurisdictions to EPA and indicate acreage or number of BMPs that are anticipated to be implemented within given areas of the watershed. Using the information provided by US EPA CBP, we calculated the difference in estimates between the “2009 baseline” and the “with TMDL” model run scenarios. Thus, our estimates of BMP implementation represent the change in pathogen loads resulting only from actions taken to meet the TMDL.

Step 2: Identify pathogen reduction efficiencies for BMPs

We then conducted a literature review of the BMPs identified in the Chesapeake Bay WIPs to investigate the potential efficiency of agricultural, urban, and septic BMPs in reducing pathogens at edge of field or edge of (small) streams. The WIP efforts identified to have the potential to affect the delivery and concentration of these waterborne pathogens include (categorized by source):

1. Agricultural: pasture and grazing management, nutrient management on crop fields, livestock waste management, restricted stream access, plantings and other structural practices to reduce nutrient and sediment runoff.

2. Urban: detention and retention ponds, impervious surface reduction, street sweeping, forested riparian buffers, bioswales, afforestation.

3. Septic: connecting septic systems to sewers, septic pumping, and on-site septic upgrades.

4. Wastewater treatment plants: new and enhanced treatment of municipal waste.

The literature review was conducted by employing each of the BMPs and pathogen-related terms as keywords (e.g., “restricted stream access” or “riparian buffer” and “fecal coliform” or “E. coli” or “bacteria”) using Google Scholar, EBSCO, and Google as search engines. Pathogen reduction estimates from peer-reviewed journal articles, documentation prepared by state agencies for compliance with TMDLs, and best practice guidance reports from state agencies and universities were all considered for the purposes of this paper. Data were included in the analysis only if they could be matched to the BMP form and function and were relevant to the pathogens being evaluated.

The range of pathogen removal efficiencies varied widely from −6% to 99%, where negative efficiencies increased pathogen concentrations as a result of the BMP implementation (Table 1). It was observed that fecal indicator bacteria (FIB), which includes fecal coliform and E. coli, were most often evaluated as a surrogate for a variety of pathogens (Marion et al., 2010; EPA, 2012a). Therefore, to make the best use of the literature, we averaged the efficiencies of fecal coliform and E. coli into a single FIB value to associate with each BMP in the analysis.

Agricultural practices showed a range of efficiencies at removing fecal coliform and E. coli (28–100%), but the average performance per practice was above 50% for all practices except wetland and stream restoration. Studies also showed high efficiency of grassed buffers at removing cryptosporidium (93–99%). Stormwater practices showed a wider range of removal efficiencies (−6–99%) than agricultural practices when looking across the range of practices. However, a few practices were responsible for the cases of low performance (bioswales, street sweeping, and septic pumping). The majority of practices had average efficiencies of 48% or greater.

Step 3: Estimate baseline pathogen loads

To estimate the baseline pathogen load, we required an understanding of pathogen sources and deliveries to water bodies, for the given level of management practices implemented in the baseline scenario. A study of the Upper Potomac River Basin, the portion that lies above the fall line¹, provided the best available information about how pathogens were being produced, transformed, intercepted and, finally, delivered downstream (Vann et al., 2002). That study estimated average annual edge-of-stream (EOS) pathogen loadings for a period that roughly corresponded to 2000–2010. The 2010 scenario was a projection of land use and population changes expected to occur by 2010 combined with 2000 estimates of non-point source BMPs and wastewater loads, and 2010 estimates for septic conditions. We use the 2010 model results as if they occurred in conditions equivalent to the baseline scenario developed by the CBP.

Vann et al. (2002) estimated loadings by land use type, using models similar to those of the CBP but modified to include pathogen movement and transformation and a wide variety of data sources on fecal sources. Data on livestock, geese, deer and human populations; National Pollution Discharge Elimination System (NPDES) and wastewater emissions; and other sources were used to inform modeling of pathogen loads by land use type. The CBP watershed model was adapted to include bacterial fate and transport, and the loads by land use were calibrated using pathogen concentrations measured at monitoring stations, primarily within the main channel of this major Chesapeake tributary. The model was also combined with data from surface water intakes to estimate the downstream delivery factors for the Potomac River basin.

To use the Vann et al. (2002) results to estimate baseline loads for the land uses in the entire Potomac River basin and the Chesapeake Bay watershed, we converted the EOS loads to per acre loadings per land use type (Table 2). We evaluated only the three land uses being modified by the BMPs used in the analysis for the purposes of converting EOS loads, as described below. We then multiplied the per acre loads for acreages of pasture, cropland, and urban for the baseline scenario to estimate baseline loads.

This method relies on transferring results of sophisticated models for the upper Potomac to two different scales of analysis (Potomac River and entire Chesapeake Bay watersheds), to provide a rough estimate of TMDL implementation effects at these scales. Clearly, using data from a portion of the Potomac River basin to represent either the whole Potomac River basin or the entire Chesapeake Bay watershed requires making considerable assumptions about the similarity of patterns and processes at these two scales. We have greater confidence in the Potomac River basin results because the Potomac River basin would be expected to be more similar to the originally modeled area than the Chesapeake Bay as a whole. The Potomac River basin may be a reasonable model for the entire Chesapeake Bay watershed because it makes up over one-fifth of the Chesapeake Bay watershed and has proportions and distribution of land use types that are similar to the entire Chesapeake Bay watershed. However, the Potomac River basin differs from the Chesapeake Bay watershed in that it has slightly more urban land and pasture and less forest (Table 3), and BMPs were applied in different proportions to the whole Bay, as shown in the Results and Discussion section.

Step 4: Estimate change in pathogens due to the TMDL

As described earlier, the acreage of BMPs implemented due to the TMDL was derived by subtracting the baseline BMP implementation from the “with TMDL” scenario. Each BMP was associated with a particular land use and quantified in terms of the acres of that land use that were affected by implementation of the BMP. For example, prescribed grazing was associated with pasture, and the percentage of total pasture under prescribed grazing was used to estimate changes in pathogen loads.

Because of data limitations, only a subset of BMPs that are capable of reducing pathogen loads were used in our analysis. BMPs were omitted from analysis if they were not measured in terms of acreage in the state WIPs, or if efficiencies were specific to baseline conditions that could not be accurately measured. For example, omitted BMPs include those measured as pounds of manure transported outside of the watershed and miles of stream restored. Also, some cropland practices in widespread use, such as continuous no-till, can be effective at reducing pathogens, but only when applied to cropland receiving manure; lack of sufficient data on manure handling prevented their inclusion. Omitting these practices, as well as point source practices, such as septic and wastewater treatment plant upgrades, tends to make our study more conservative in terms of the TMDL effectiveness for reducing pathogens because practices that are expected to be implemented as part of the WIPs were not counted, and some of these practices have been demonstrated to be highly effective at reducing pathogen loads (Table 1).

To estimate the change in pathogen loads (measured as FIB) delivered to the main channel as a result of applying a subset of BMPs from the WIPs, we applied Eq. (1): (1) $\Delta FI{B}_{DS}=\underset{l}{{{\displaystyle \sum }}^{\text{}}}\left(\underset{b}{{{\displaystyle \sum }}^{\text{}}}\frac{{\left(BMPAcres\right)}_{b,l}}{{\left(Totallandarea\right)}_l}{\left(\text{\%}FIBreduction\right)}_b\right){\left(EOSload\right)}_l{\left(\text{\%}DSDelivery\right)}_l$ where

• b is the BMP applied and

• l is the land use type.

Equation (1) shows that delivery of pathogens to the main channel depends on edge-of-stream (EOS) loads and downstream (DS) attenuation of pathogen loads.

BMP Acres represents the acres of a given land use treated with a given BMP. The Total land area per land use (l) was derived from the baseline scenario. The %FIB reduction was the average removal efficiency for fecal coliform and E. coli for a given BMP. The proportion of treated acres to total acres in a given land use was multiplied by the percentage reduction for a given practice, and then these values were summed for all BMPs affecting a land use to generate a weighted sum representing the percentage reduction in pathogen loads expected for a given land use. The expected percent reduction for a given land use was multiplied by the baseline load for that land use to generate the EOS load (cfu/yr). Finally, the DS load was estimated by multiplying the EOS load by 21%, which was the average delivery ratio for all Potomac River segments modeled in the study by Vann et al. (2002).

Potential magnitude of benefits

Whether reductions in pathogens reduce human illness from water contact or shellfish ingestion is a function of the ability of the pathogens to produce disease in people, probability of exposure to the pathogens, pathogen concentration, the number of people exposed, and the characteristics of the people that influence their susceptibility to disease (e.g., Soller et al., 2003). A quantification of all these effects was beyond the scope of this project. However, we gathered some existing information and data to suggest the potential order of magnitude of benefits.

FIB are correlated with a number of illnesses caused by bacteria and viruses, and the illness that has been most consistently and clearly linked to water contact is increased risk of gastroenteritis (Kay et al., 1994; Fleisher et al., 1998; Wade et al., 2010), although other diseases have also been observed, including respiratory illnesses, ear infections, and skin rashes (Fleisher et al., 1998; Fleisher et al., 2010). Skin diseases (infections and rashes) have been most closely linked to non-point sources of pathogens (Fleisher et al., 2010), whereas gastroenteritis is more clearly linked to sewage (Wade et al., 2010). The gastrointestinal illnesses caused by shellfish consumption have been linked to concentrations of Vibrio spp. (Hlady & Klontz, 1996), but Vibrio concentrations are widespread in the marine environment and are not highly correlated with fecal coliform (DePaola et al., 2000) and only weakly correlated with nitrogen concentrations (Pfeffer, Hite & Oliver, 2003; Eiler, Johansson & Bertilsson, 2006; Johnson et al., 2010). However, concentrations of Vibrio spp. have been linked to increased sediment suspension in some cases (Vanoy, Tamplin & Schwarz, 1992; Pfeffer, Hite & Oliver, 2003; Fries, Characklis & Noble, 2008).

The number of people harmed is indicated by the cases of reported illness that can be linked to waterborne pathogens. Table 5 presents reported illnesses from Maryland and Virginia. These states were included because these are the states with Chesapeake Bay shoreline, but other swimmable water bodies would also be affected by pathogens.

These data on illnesses suggest a potential order of magnitude of illnesses caused by pathogens, but are not an accurate accounting, for three reasons. First, these numbers may be underestimates of true disease incidence due to water contact because only a fraction of illnesses are likely to have been identified and reported. Many more cases of gastrointestinal illnesses are likely to occur than to be reported (Hlavsa et al., 2014). Further, anecdotal information suggests that skin rashes and infections due to water contact are not an uncommon ailment in the Chesapeake Bay (Kobell, 2011; Kobell, 2013), particularly in the warmest months. These cases are not usually reported but have been documented elsewhere (Wade et al., 2010). Second, this particular set of cases may not be representative of risk associated with swimming in the Chesapeake Bay or its tributaries. Data collected by the Centers for Disease Control and Prevention using different reporting criteria found that 70% of reported illnesses due to waterborne pathogens were from pools or other treated water and 30% were from open (untreated) water, such as lakes and oceans (Hlavsa et al., 2014). Third, the relationship between animal-derived pathogens and human illnesses is poorly understood. Many dose-response relationships are based on pathogens from sewage sources, not agricultural sources.

Because of these data limitations, an estimation of the reduction in cases of disease was beyond the scope of this effort. However, we know that at least some water bodies contain dangerous levels of pathogen concentrations (evidence provided by TMDLs and cases of disease) that will be reduced in different proportions depending on the extent of BMP implementation in the watershed. The ability to reduce areas of high pathogen concentrations in areas with a high probability of exposure will have the most potential to create benefits.

Potential value of reduction in pathogens

This analysis suggests that these reductions can provide the following benefits to people in the Chesapeake Bay watershed. First, those who are in contact with the water (commercial fishermen, recreational anglers, boaters, and swimmers) are likely to have improved welfare due to illnesses avoided and may increase the number of trips they take. Second, more risk-averse recreators, who might currently avoid the water, might be induced to recreate in the Chesapeake Bay, in response to improved water safety (Lipton, 2004). Third, increased safety of shellfish could benefit commercial watermen, the burgeoning aquaculture industry, and seafood consumers. We would expect welfare increases from additional recreation trips, increased safety per trip, lowered costs of production for producers, and safer shellfish for consumers.

Table 6 summarizes the potential values of avoiding illnesses and beach closures as identified in the economic literature.

Given the potential number of beach users in the Chesapeake Bay³, the total economic value of pathogen reductions due to the TMDL could be substantial when aggregated over the total number of beach users. Virginia had 29 days of beach actions (notifications and closure days) out of a total of 6,900 beach days in 2012 (open days multiplied by number of beaches), and an average of 56 beach actions per year from 2007–2012 (VDH, 2014; EPA, 2012b; EPA, 2013b). Maryland had 139 days of beach actions out of a total 6,501 beach days in 2012, and an average of 196 beach actions per year from 2007–2012 (EPA, 2013a).

Furthermore, these estimates do not include avoided costs due to reduced shellfish bed closures, lost wages and medical bills due to illness, or costs associated with stream miles impaired due to pathogens. Over 9,000 stream miles in Virginia (EPA, 2013c), over 4,000 miles in Maryland (MDE, 2012), and 190 miles in Pennsylvania (EPA, 2013c) are impaired for E. coli and fecal coliform⁴. A reduction in impaired stream miles would decrease both administrative costs associated with listing impaired waters and costs associated with developing and implementing TMDLs as required by Clean Water Act Section 303(d) (EPA, 2015).

Future efforts to estimate the total economic value of the TMDL with respect to pathogen reductions are required to fully understand the potential of the action.