This study is focused on the tidal Potomac River estuary, which includes the section of the river from Washington D.C. to its confluence with the Chesapeake Bay (Fig. 1). The Potomac River estuary begins as tidal freshwater, becoming oligohaline ∼ 30–50 km below Washington D.C. and mesohaline at its mouth approximately 160 km below Washington D.C. (Jaworski et al., 1992). The Potomac River estuary can be seasonally stratified (Hamdan and Jonas, 2006), especially in the southern portion of the system where intruding, saline bottom water from the main stem of the Chesapeake Bay leads to density-driven estuarine circulation patterns (Elliott, 1976, 1978; Pritchard, 1956). Mixing is most evident at the estuarine turbidity maximum (Hamdan and Jonas, 2006), ∼ 60–80 km down-estuary of Washington D.C., and the water column is generally well mixed above the estuarine turbidity maximum zone in the tidal fresh and oligohaline regions of the estuary (Crump and Baross, 1996; Sanford et al., 2001).

The watershed draining to the Potomac River estuary is classified as 58 % forested, 23 % agricultural, and 17 % urban, based on Maryland Department of Planning data for 2002 (Karrh et al., 2007a). Based on the Chesapeake Bay Program (CBP) model it was estimated that, during 2005, total inputs of nitrogen were 33 % from agriculture, 20 % from urban (e.g., stormwater runoff and leaky sewers), 19 % from point sources (wastewater treatment plants and industrial releases), 11 % from forest, 10 % from septic, 6 % from mixed open land, and 1 % from atmospheric deposition to water (Karrh et al., 2007b). The CBP model is developed using long-term monitoring data and the non-point loads are estimated from a variety of sources including land cover and agriculture records (Karrh et al., 2007b).

The Potomac River estuary also receives N inputs from the Blue Plains wastewater treatment plant, located in Washington, D.C. In 2009 Blue Plains discharged 2.3 mg L-1 of NO3- and 3.7 mg L-1 of TN, on average, and exported loads of approximately 2300 kg day-1 of NO3- and 3900 kg of TN. Overall, Blue Plains treats and discharges approximately 1.06 × 109 L day-1, almost 5 % of Potomac River's annual discharge. In the past several decades, Blue Plains has undergone several technological improvements with phosphorus removal in the 1980s and enhanced N removal beginning in the year 2000. Since the implementation of advanced wastewater treatment technologies at Blue Plains, there has been a significant decrease (p < 0.01) in the concentration of nitrate in effluent discharge, from an average of 7.2 ± 0.3 mg L-1 before the year 2000 (years 1998 and 1999) to an average of 4.1 ± 0.4 mg L-1 directly after 2000 (years 2001 through 2008).

Surface and bottom water N and carbon data collected by the Maryland Department of Natural Resources (DNR) and accessed through the Chesapeake Bay Program's Data Hub website (Chesapeake Bay Program, 2013) was used to look at historical (1984 to 2012) monthly nutrient concentrations from stations located longitudinally along the Potomac River estuary (Fig. 1). In particular, these data were used for analyzing the spatial and temporal trends for dissolved and particulate forms of N and dissolved organic carbon (DOC) in the Potomac River estuary prior to and during this study.

Water chemistry samples along the Potomac River estuary were collected monthly for 1 year from April 2010 to May 2011; from 12 to 160 km below the Blue Plains wastewater treatment plant (See Fig. 1). Water was collected from the surface (top 0.5 m) and bottom water depths. Surface water samplings from 6 above to 12 km below the Blue Plains wastewater treatment plant effluent outfall were collected seasonally during this time (Fig. 1). Water temperature and salinity were also measured during each water chemistry sampling.

Surface samples for δ15N-NO3- and δ18O-NO3- isotopes of dissolved nitrate were filtered (0.45 µm), frozen, and shipped to the UC Davis Stable Isotope Facility (SIF) for analysis. The isotope composition of nitrate was measured following the denitrifier method (Casciotti et al., 2002; Sigman et al., 2001). In brief, denitrifying bacteria are used to convert nitrate in samples to N2O gas, which is collected and sent through a mass spectrometer for determination of the stable isotopic ratios for N and O of nitrate (15N / 14N and 18O / 16O). Values for δ15N-NO3- and δ18O-NO3- are reported as per mill (‰) relative to atmospheric N2 (δ15N) or Vienna Standard Mean Ocean Water (VSMOW; δ18O), according to δ15N or δ18O (‰) = [(R)sample/(R)standard – 1] × 1000, where R denotes the ratio of the heavy to light isotope (15N / 14N or 18O / 16O). For data correction and calibration, UC Davis SIF uses calibration nitrate standards (USGS 32, USGS 34, and USGS 35) supplied by NIST (National Institute of Standards and Technology, Gaithersburg, MD). The long-term standard deviation for nitrate isotope samples at UC Davis SIF is 0.4 ‰ for δ15N-NO3- and 0.5 ‰ for δ18O-NO3-. Previous studies (Kaushal et al., 2011; Kendall et al., 2007) indicate that the relative amounts of δ15N-NO3- and δ18O-NO3- can be used to determine specific sources of nitrate (i.e., fertilizer, nitrification, atmospheric, or sewage derived nitrate).

It should be noted that while the denitrifier method converts sample NO3- and NO2- to N2O gas, in marine systems, NO2- has been shown to complicate interpretations of the N and O isotopes of NO3- if it remains unaccounted for (e.g., Fawcett et al., 2015; Marconi et al., 2015; Rafter et al., 2013; Smart et al., 2015). This is partially because during the reduction of NO3- and NO2- to N2O by the denitrifiers, the O isotope effects are different (and thus need to be corrected for). In addition, the δ15N of NO2- can be extremely different from that of NO3-, potentially further complicating interpretation of the data. We found that in the Potomac Estuary stations TF2.1 through LE2.3 (stations from the top of the estuary to the bottom of the estuary) the mean nitrite concentration from 2010 to 2012 is 0.013 mg L-1, with a minimum of 0.0055 mg L-1 and maximum of 0.0183 mg L-1. The mean nitrite is about 2.4 % of the mean nitrate+nitrite concentration. Based on the literature (Fawcett et al., 2015), this level of nitrite is still high enough to have some impacts on the nitrate isotope values, with differences up to 5 ‰ for both N and O isotopes of nitrate when using the denitrified method with and without nitrite mixed with nitrate in the samples (Casciotti and McIlvin, 2007).

To distinguish between the different potential nitrate sources we used a Bayesian isotope mixing model (Parnell et al., 2010, 2013; Xue et al., 2012; Yang and Toor, 2016). For the Bayesian isotope mixing model, the Stable Isotope Analysis in R (SIAR) package was used to determine the fraction of nitrate in each sample from four different sources: wastewater, atmospheric deposition, nitrification, and nitrate fertilizer (Parnell et al., 2010, 2013; Xue et al., 2012; Yang and Toor, 2016). The SIAR mixing model is able to incorporate uncertainty in nitrate source estimates based on the uncertainty in the nitrate source endmembers (Parnell et al., 2010, 2013; Xue et al., 2012; Yang and Toor, 2016).

Nitrate source endmember values for δ15N-NO3- and δ18O-NO3- were obtained from the literature, except wastewater nitrate, which was obtained from this study. The endmember values for δ15N-NO3- and δ18O-NO3- were -10.3 ± 1.7 and 10.1 ± 1.5, respectively, for nitrate from nitrification (Mayer et al., 2001); 0 ± 3 and 22 ± 3, respectively, for NO3- fertilizer (Mayer et al., 2002); and 3 ± 3 and 69 ± 5, respectively, for atmospheric nitrate (Burns and Kendall, 2002; Divers et al., 2014). The wastewater δ15N-NO3- and δ18O-NO3- endmember values (31.5 ± 7.8 and 11 ± 4.5, respectively) were based on averaging the effluent nitrate isotope values measured monthly from Blue Plains during the study period. The nitrification source represents NO3- from nitrification in the water as well as nitrification of ammonia fertilizer in the watershed. The fertilizer source represents synthetically produced NO3- fertilizer, not the more common ammonia fertilizer. Animal manure was not used as one of the endmembers because this source is more significant in the upper Potomac River, above Washington, D.C. compared to the lower Potomac River watershed. For example, there are 171 concentrated animal feeding operation (CAFOs) in upper Potomac compared to 25 CAFOs in the lower Potomac below Washington, D.C. (US EPA, 2016).

Due to the variability in nitrate source endmembers, the mixing model was used primarily for illustrative purposes and should be viewed with caution.  For example, there can be high variability in the nitrification source endmembers because nitrate from nitrification can come from ammonia fertilizer, manure fertilizer, particulate organic matter within the water column, etc. The nitrate from nitrification will therefor carry a range of nitrate isotope values reflecting its original source (Kendall et al., 2007). Also, because denitrification is known to cause the increase in δ15N-NO3- and δ18O-NO3- values through isotopic fractionation in approximately a 2:1 relationship (Divers et al., 2014; Kendall et al., 2007), this isotopic enrichment can complicate the identification of wastewater nitrate. For example, water samples with increased wastewater nitrate, based on the mixing model, may also indicate denitrification has played a role in the isotopic levels of the sample nitrate. As a result, there is a potential to overestimate the contribution of nitrate from wastewater if denitrification is occurring in the estuary.

An additional method using plots of salinity vs. NO3- concentration or NO3- isotopes was used to assess whether there is conservative mixing (dilution), or mixing with additional NO3- sources down-estuary, or losses of NO3- through biotic uptake or denitrification (Middelburg and Nieuwenhuize, 2001; Wankel et al., 2006). Mixing line equations for NO3- concentrations were based on equations 1–3 from Middelburg and Nieuwenhuize (2001) and isotopes mixing lines were based on equation 4 from Middelburg and Nieuwenhuize (2001). The mixing line equations and endmember values used for salinity and nitrate isotopes are provided in the Supplement (Table S2). Based on those equations, the salinity vs. NO3- concentration mixing lines are linear, while the mixing lines for NO3- isotopes are nonlinear (Middelburg and Nieuwenhuize, 2001). Wankel et al. (2006) suggest that when nutrient concentrations fall above the mixing line, this indicates an additional source to raise the concentrations, while concentrations that fall below the mixing line indicate that there is a nutrient sink (e.g., denitrification, assimilation). For nitrate isotopes, when the δ15N-NO3- and δ18O-NO3- values fall above this mixing line, this could indicate an additional source or the fractionation of nitrate from assimilation or denitrification that would increase the heavy isotope levels, while isotope values below the mixing line could indicate an additional source of nitrate with lighter isotope values, such as from nitrification or fertilizer sources (Wankel et al., 2006).

A box model was used to estimate net fluxes of TN, NO3-, and nitrate isotope loads along the Potomac River estuary using methods modified from Officer (1980), Boynton et al. (1995), Hagy et al. (2000), and Testa et al. (2008), which are widely used methods for tracking nutrient fluxes in estuaries between different salinity zones. First, the Potomac Estuary was divided into six boxes in order to accommodate adequate sampling stations per box, and to evaluate net fluxes at key locations along the estuarine gradient (Fig. 2). Next, due to the Potomac Estuary having a semi-diurnal tidal cycle, where there is movement back and forth across boundaries of the box model, mean monthly freshwater discharge inputs to the first box (USGS, 2014) and interpolated salinity values (measured monthly from surface and bottom waters throughout the system) were used to calculate advective and diffusive exchanges of water and salt between adjacent boxes. Salt balances were then used to compute net exchanges at the boundaries of the six model boxes, similar to previous estuarine box model studies (e.g., Boynton et al., 1995; Hagy et al., 2000). Average monthly TN, NO3-, and NO3- isotope concentrations (collected from the surface and bottom water at each station, except for NO3- isotopes, which were collected from the surface only) were multiplied by net estimated exchange values at the box boundaries and summed to calculate the N load leaving or entering each box. In order to calculate the loads for NO3- isotopes, the δ15N-NO3- and δ18O-NO3- values in per mill (‰) were converted to concentrations (µg L-1) by multiplying the NO3- concentration of the sample by R, the ratio of the heavy to light isotope (15N / 14N or 18O / 16O). Fluxes were estimated for each month during the sampling period and then averaged to find seasonal estimates of N fluxes for the Potomac. The box model results were used to compute (1) the total inputs of N, (2) the  % inputs of loads from Blue Plains, (3) the net export of N to the Chesapeake Bay, (4) the  % of Blue Plains inputs that are exported, (5) the net loss in loads along the estuary, and (6) the contribution of N loads from the Chesapeake Bay through tidal inflow.

To account for uncertainty in monthly load estimates, error propagation (using standard errors) was used for each of the hydrologic and nutrient inputs to the model. For example, the error in discharge data came from averaging the mean daily discharge for each month, the error in concentrations came from averaging the surface and bottom water concentrations, and the error in N from atmospheric deposition came from averaging the weakly deposition data for each month. These uncertainties in the inputs to the box model were then propagated for each of the box model calculations, similar to Filoso and Palmer (2011).

Inputs to the box model include total monthly precipitation data based on averaging data from three stations along the Potomac Estuary (precipitation data are from the NOAA National Centers for Environmental Information, Climate Data Online); monthly estimates of atmospheric deposition for NH4+, NO3-, and DIN (obtained from the National Atmospheric Deposition Program/National Trends Network); NO3- concentrations and isotope levels in atmospheric deposition (from Buda and DeWalle, 2009, for the nearby central Pennsylvania region for the year 2005, which was a similar year hydrologically); freshwater and N inputs from the land (from Chesapeake Bay model output from 2005); surface and bottom water nutrient and salinity concentrations (from MD DNR); and inputs from the Blue Plains wastewater treatment plant. Also, while there are no USGS gages located along the Potomac Estuary, there is one USGS gage (USGS 01646580) located about 16 km up-river from Blue Plains, directly above the estuary (where the hydrodynamics of the river cease being tidally influenced) and this gage was used to account for freshwater inputs into the first box. The model also takes into account water temperature and evaporation.

In the box model we made two assumptions regarding the 14 other WWTPs that are dispersed along the estuary below Blue Plains. All but one of these WWTPs have tertiary treatment (the other has secondary treatment; www.epa.gov/npdes). These other WWTPs have a combined TN load that is 32 % of the TN load from Blue Plains. While the loads from these WWTPs are indirectly accounted for in the box model due to their impact on the concentrations in the estuarine water, it was not feasible to directly incorporate the loads from each WWTP into the box model estimates and thus there may be some added uncertainties. However, we can first assume that the estimated decline in nitrogen loads from the Blue Plains wastewater treatment plant to the mouth of the Potomac River estuary results in conservative estimates. The additional load from the other WWTPs only adds to the loads estimated further down-estuary, and consequently the measured loss in N load from the Blue Plains wastewater load down-estuary (the difference between the loads at the mouth and at the head of the estuary) is a conservative estimate because it is less then would be expected, underestimating biological assimilation and removal. Second, for modeling purposes, we also assume here that the loads from the 14 other WWTPs have little effect on the nitrate isotope signal. While δ15N-NO3- and δ18O-NO3- isotope values were not measured directly for the 14 other down-estuary wastewater treatment plants, based on the literature, the values for average WWTP nitrate isotopes are typically lower (∼ 10 ‰ for δ15N-NO3- and ∼ 0 for δ18O-NO3-) compared to 31.5 for δ15N-NO3- and 11 ‰ δ18O-NO3- for Blue Plains (Kendall et al., 2007; Wang et al., 2013; Wankel et al., 2006). As a result, we expected the other WWTPs to have a similar or an even less pronounced wastewater isotope signal compared to Blue Plains, which has biological nitrogen removal (i.e., denitrification is promoted within the Blue Plains WWTP), elevating the δ15N-NO3- and δ18O-NO3- isotope values at Blue Plains more (Kendall et al., 2007). Consequently, the estimated nitrate loads down-estuary incorporate inputs from Blue Plains and inputs from the other WWTPs. They are considered conservative estimates because the additional WWTPs only add to the TN loads and wastewater NO3- isotope signal, so any decline in an isotope signal that we attribute to Blue Plains would likely be greater if data availability permitted us to specifically parameterize the isotope values for additional WWTP inputs.

Another assumption was made for the box model related to estuarine mixing. Although portions of the lower estuary can be seasonally stratified, we assumed each box to be well mixed vertically as no bottom water isotope values were available to constrain a two-layer box model. This assumption is supported by other bottom water data that are available and by samples taken along the width of the estuary. For example, we have conducted the box model and other analyses with and without bottom water isotope data and found minimal change in results (Fig. S1 in the Supplement). Our measurements of various biogeochemical signatures at the station close to the estuarine turbidity maximum suggests that there is intense mixing at this site, and prior studies have documented extensive mixing in the freshwater tidal portion of the system (Elliott, 1976, 1978; Pritchard, 1956). Also, it can be assumed that, because wastewater effluent inputs are freshwater, much of the effluent plume would likely not sink in the more dense estuarine waters moving up from the bay. Additionally, our box model estimates of net fluxes was compared to a complex, three-dimensional hydrodynamic model (described below) that incorporates stratification, and this comparison provided support for the low impact of assuming mixing in our approach.

Only surface water samples were analyzed for δ15N-NO3- and δ18O-NO3- isotopes, and as a result our box model was not able to directly incorporate the potential impacts of stratification on the estimated flux of NO3- isotopes. However, while seasonal stratification has been found close to the mouth of the Potomac Estuary (Hamdan and Jonas, 2006), using documented nitrate bottom water isotope values from near the mouth of the estuary (Horrigan et al., 1990) we calculate that incorporating bottom water isotope values would have a minimal impact on the flux estimates of our box model, particularly when not including spring 2011 (Fig. S1). But when including spring 2011, and using the reported values of 10 ‰ for bottom water δ15N-NO3-, based on Horrigan et al. (1990), in boxes 5 and 6, where stratification is most likely, our estimates for the flux of δ15N-NO3- from these boxes increases by 20 % on average, and the net loss in load from box 1 to box 6 increases by 12 % on average. This indicates that our estimates are conservative because, by not using bottom water, we estimate a smaller net loss in δ15N-NO3- (Fig. S1).

For the box model we also assumed the estuary to be well mixed laterally. In terms of potential variability for samples taken at different locations along the width of the estuary, for surface water samples, on average, a 6 ± 3 % difference was found in δ15N-NO3-, a 7 ± 3 % difference in δ18O-NO3-, a 24 ± 8 % difference in NO3-, and a 15 ± 3 % difference in TN (based on samplings that were done at two or more locations along the same longitudinal transect at approximately the same distance down-estuary, but at different locations horizontally at that location). Consequently, the nitrate isotope values and NO3- and TN concentrations appear to show that the estuary is fairly well mixed laterally.

To assess the accuracy of the box model assumptions and results, estimated net fluxes of total N were compared to simulation output from the Chesapeake Bay Water Quality Model. This model was developed by the US EPA to aid in efforts to set total maximum daily loads (TMDLs), the maximum amount for each pollutant that can occur, for the Chesapeake Bay (Cerco et al., 2010), and it combines a 3-D hydrodynamic model (CH3D) with a water quality model (CE-QUAL-ICM). Simulation output data were available for 1996, 2002, and 2005. We selected a simulation year (2005) because it had similar river discharge conditions to 2010, and compared modeled net fluxes of TN at three boundary locations to estimates at the same (or nearby) box model boundaries.

Statistical analyses were performed using the statistical package R (R Development Core Team, 2013). Linear regression was used to test for significant changes in stream chemistry and nitrate isotope data with distance down-estuary. Repeated-measures analysis of variance (ANOVA) was used to test for seasonal differences in nitrate isotopes trends with distance.

Longitudinal patterns of dissolved inorganic nitrogen (DIN) in the lower Potomac River showed an increase in concentrations near and directly below the Blue Plains wastewater treatment plant and then a steady decline in concentrations down to the Chesapeake Bay (Fig. 3a). The implementation of tertiary treatment in 2000 coincided with a significant drop in annual average DIN concentration directly down-estuary of the Blue Plains WWTP (from 1.7 ± 0.02 to 1.3 ± 0.01 mg L-1, p < 0.05; Fig. 3a) when comparing years directly prior (1997–1999) and the years directly after 2000 (2001–2005). However, the impact of the wastewater treatment plant improvements on reducing longitudinal patterns of DIN was only apparent for the first 30 km down-estuary. After this, both the pre- and post-2000 DIN concentrations overlapped (Fig. 3a). As DIN decreased longitudinally down-estuary of the wastewater treatment plant, there was also a small but significant increase in total organic nitrogen (TON) after the year 2000 (p < 0.01, Fig. 3a), not including the last sample near the mouth of the estuary, which is likely influenced by tidal inflow.

There were seasonal variations in DIN concentrations along the Potomac River estuary with the greatest concentrations in the winter and spring (Fig. 3b). There is also a steeper decline in DIN with distance during fall, winter, and summer compared to the spring (p < 0.05, Fig. 3b). The average molar ratio of DIN to PO4-3 (N : P ratio) showed an initial increase, then a decrease as estuarine salinity started to increase (Fig. 3c). During the summer and fall, the N : P ratio fell below the Redfield ratio (16:1, the atomic ratio of nitrogen and phosphorus found in oceans and phytoplankton), around 40 km down-estuary and stayed below 16, which indicated a shift from P to N limitation. During the winter and spring, the N : P ratio never fell below 16 and increased steadily after 50 km down-estuary (Fig. 3c). There was also a significant negative relationship between NO3- and DOC concentration during the study period (p < 0.01, Fig. 4).

During each season, except spring, δ15N-NO3- values increased sharply at the Blue Plains outfall, from 9.3 ± 1.4 ‰ up-estuary to 25.7 ± 2.9 ‰ at the outfall (p < 0.05), and then rapidly decreased within 2 km down-estuary of the Blue Plains WWTP to 15.7 ± 2.2 ‰ (p < 0.05, Fig. 5a). During the summer and fall, the δ15N-NO3- values showed the largest increase near the effluent outfall (except for one very high winter value) and then a significant decrease (p < 0.05) with distance down-estuary. There was also a slight increase in δ15N-NO3- and δ18O-NO3- values from 1 to 6 km down-estuary (Fig. 5a, b). During the winter and spring, the δ15N-NO3- and δ18O-NO3- values remained relatively constant throughout the estuary, even near Blue Plains (Fig. 5a, b), while during the summer and fall the δ15N-NO3- and δ18O-NO3- values steadily declined after 6–10 km down-estuary (Fig. 5a, b). At the mouth of the estuary, the δ15N-NO3- values for all seasons were roughly equivalent (Fig. 5a). During the summer and fall, the δ18O-NO3- values showed a steady decrease after 12 km down-estuary, while they increased during spring and winter (Fig. 5b).

Based on the nitrate isotope mixing model, nitrate contributions from wastewater ranged from 80 ± 13 % at the wastewater outfall to 57 ± 11 % within the first 1 km down-estuary. Wastewater nitrate contributions then decreased to 44 ± 14 % at the confluence of the Potomac River estuary with Chesapeake Bay (Fig. 5c). When we multiply the percent wastewater nitrate by the nitrate loads estimated from the box model for the top and bottom of the estuary (results below), we calculate that there was a 70 ± 31 % loss in wastewater NO3- along the estuary annually. Nitrate from nitrification (of N from upriver manure or ammonia fertilizer and also Blue Plains wastewater N) increased from 13 ± 12 % at the wastewater outfall to 29 ± 22 % at the confluence of the Potomac River estuary with Chesapeake Bay (Fig. 5c). Nitrate from fertilizer increased from 6 ± 6 % at the wastewater outfall to 22 ± 22 % at the confluence of the Potomac River estuary with Chesapeake Bay (Fig. 5c). Nitrate from atmospheric deposition changed little along the Potomac Estuary from 1 ± 1 at the wastewater outfall to 5 ± 5 at the confluence with the Chesapeake Bay (Fig. 5c). At the last two sampling stations near the mouth of the Potomac River estuary, NO3- from fertilizer showed an increase, while NO3- from nitrification showed a corresponding decline (Fig. 5c).

The Blue Plains effluent and Potomac River samples within 20 km down-estuary of the wastewater treatment plant showed a significant positive relationship between δ15N-NO3- and δ18O-NO3- (p < 0.05; Fig. 6a). When denitrification and biotic uptake occurs, plotting δ15N-NO3- vs. δ18O-NO3- shows a 2:1 relationship (Kendall et al., 2007). The Blue Plains effluent samples showed approximately a 2.4 to 1 relationship. The samples within 20 km down-estuary showed a 3:1 ratio (Fig. 6a). The nitrate samples within the first 6 km showed a 2.4 to 1 relationship (Fig. 6a). There were also seasonal differences in the relationship between δ15N-NO3- and δ18O-NO3- (Fig. 6b); spring, summer, and fall were characterized by close to a 2:1 relationship between δ15N-NO3- vs. δ18O-NO3-, while winter showed a ∼ 8:1 relationship.

Because salinity is a conservative tracer, plots of salinity vs. NO3-, δ15N-NO3-, and δ18O-NO3- can indicate effects of mixing between water at the tidal freshwater section with water from the mesohaline section of the Potomac River estuary. Deviations from the mixing lines can indicate additional sources or biological transformations (Middelburg and Nieuwenhuize, 2000; Wankel et al., 2006). Surface water NO3- concentrations and nitrate isotopes fell on (for δ18O-NO3-) or slightly below mixing line (for δ15N-NO3-) during the spring (Fig. 7a,b,c), which indicated mostly conservative mixing (dilution or inputs from sources with low δ15N-NO3-, like nitrification). But during the summer and fall, the NO3- concentration and isotope values fell well below the mixing lines. During the winter, the values fell both above and below the mixing line (Fig. 7a, b, c), which indicated non-conservative mixing.

Our comparisons of box model net exchange estimates with simulation output provided by the Chesapeake Bay Program eutrophication model (“Bay Model”) revealed similar TN loads between our results and the Bay Model in the winter, spring, and fall, with the largest differences in the models evident in the summer months at the boundary location where tidal fresh transitions to oligohaline conditions and at the mouth of the estuary (Table S3 and Figs. 8 and 9). Even so, these differences are smaller than a factor of 2 for winter and spring and for most of the summer and fall. Despite the assumption of complete mixing in our box model, this is a good agreement considering the simplification of hydrodynamics inherent to a box modeling approach when compared to the highly constrained CH3D hydrodynamic modeling platform (Cerco et al., 2010). The Potomac Estuary is well mixed along two-thirds of its length, and this likely contributes to our success in applying a single-layer box model to this system. The box model also permitted estimates of TN loads at smaller spatial scales than the three boundaries available from the Chesapeake Bay Program, which could enable a better interpretation of where Blue Plains effluent was subject to transformations in the oligohaline portion of the estuary (Fig. 8). The caveat here is that box-modeled summer loads should be interpreted with caution because they show the greatest differences from the CH3D model.

Results of the box model indicate that an annual average of 8.4 × 106 ± 4.8 × 106 kg yr-1 of TN are exported to the Bay and the net loss in load for TN along the estuary (from Blue Plains to the mouth of the estuary), attributed to assimilation, burial, and denitrification, was 9.1 × 106 ± 5.1 × 106 kg yr-1 of TN. Using an N burial rate of 2.49 × 106 ± 3.1 × 105 kg yr-1 (Harris, unpublished data), a denitrification rate of 6.17 × 106 ± 8.3 × 104 kg yr-1 (Cornwell et al., 2016) and a fisheries yield rate of 0.82 × 106 kg yr-1 (Boynton et al., 1995), we see that our box model estimate is nearly balanced by independently estimated values for these loss terms. On a mean annual basis, denitrification accounts for about 68 ± 1 % of the loss in TN, burial is estimated to account for 27 ± 3 % of the loss in TN, and assimilation into fisheries accounts for approximately 9 % of loss in TN load along the Potomac Estuary.

The net load (kg day-1) of TN, NO3-, and δ15N-NO3- decreased down-estuary during each season (Fig. 10a–c, p < 0.05 for winter and spring and p < 0.1 for summer and fall). N loads were highest along the estuary during spring and winter (Fig. 10), and there was a greater decline in TN loads on average from box 1 to box 6 during winter and spring (a loss of ∼ 27 000 ± 15 000 and 50 000 ± 52 000 kg day-1, respectively; Table 1) compared to summer and fall (a loss of ∼ 7000 ± 8000 and 15 000 ± 13 000 kg day-1, respectively). However, the summer and fall months showed a greater percent decline in TN (75 ± 75 and 112 ± 95 %, respectively) compared to winter and spring (54 ± 40 and 36 ± 43 %, respectively). The relatively high errors are primarily from the larger uncertainty found in the last box, at the mouth of the estuary, due to the larger size of this box and greater uncertainty in fluxes at the mouth of the estuary; the uncertainties are much smaller further up-estuary (See Fig. 10a). NO3- and δ15N-NO3- follow the same seasonal patterns as TN. Also, winter, along with summer and fall, showed a greater percent decline in NO3- and NO3- isotope loads compared to spring (Table 1).

The percent contribution of TN inputs from the Blue Plains wastewater treatment plant to the main stem of the Chesapeake Bay ranged from 8 to 47 % (Table 1). The contribution was significantly lower during the winter and spring (10 ± 13 and 8 ± 1 %, respectively) compared to summer and fall (38 ± 3 and 47 ± 13 %, respectively, Table 1), when TN fluxes from all sources are relatively low. The percent of Blue Plains wastewater TN inputs that are exported to the Chesapeake Bay ranged from < 4 to 71 %, and they were highest in the spring (71 ± 20 %, Table 1), with an annual average of 28 ± 6 % of blue plains TN exported from the estuary. There were also N inputs to the Potomac river–estuarine continuum from the Chesapeake Bay during each season, except spring, due to higher flows (Tables 1 and 2) because flow in spring was too high to allow the inputs from the Bay that occurred in the other seasons. NO3- and δ15N-NO3- follow the same seasonal patterns as TN, showing the greatest percentage of inputs from Blue Plains exported during the spring.

The decrease in DIN concentrations with distance down-estuary is largely from denitrification, assimilation, and burial, as indicated by the inverse relationship between NO3- concentrations and DOC and TON concentrations, the NO3- isotope data, and N mass balance data. Dilution from tidal marine waters plays a minor role in the decrease in DIN and the incoming tidal waters may even contribute to DIN as suggested by the decrease in DIN slope after 130 km down-estuary (Boynton et al., 1995), depending on the season. The installation of tertiary wastewater treatment technology at Blue Plains in the year 2000 showed a significant drop in DIN concentrations within 20–30 km of Blue Plains. However, the DIN concentrations below 30 km down-estuary were approximately the same based on an annual average, before and after the year 2000. One explanation is that the dissolved wastewater N is completely assimilated into particulate organic matter, supported by the inverse NO3- vs. TON or DOC relationships (Figs. 3a and 4), or removed by denitrification (as suggested by the isotope data) within the first 10 km down-estuary, and thus the majority of DIN below 30 km is from other inputs than the Blue Plains wastewater treatment plant. For example, there are 14 other smaller wastewater treatment plants along the Potomac River estuary, which contribute a total of about 1.02 × 109 L day-1 (almost as much as the amount Blue Plains contributes) and they could offset further decreases in NO3- concentrations down-estuary. Also, our isotope mixing model data show that nitrification (likely of upriver manure or ammonia fertilizer inputs) and fertilizer are important sources further down-estuary, and 42 % of the land use along the Potomac Estuary is agriculture (Karrh et al., 2007b). A second explanation could be related to a change in N : P ratio with distance down-estuary. Specifically, there was a rise in estuarine salinity around 30 to 50 km down-estuary and a coinciding increase in dissolved PO4-3 concentration (typical of the estuarine salinity gradient; Jordan et al., 2008). When the N : P ratio fell below the Redfield ratio of 16:1, the estuary could shift from P limitation to N limitation (Fisher et al., 1999). The potential shift from P to N limitation occurred 40–50 km down-estuary, around the estuarine turbidity maximum, which is associated with higher estuarine bacterial productivity (Crump and Baross, 1996), and may be driving DIN removal further down-estuary.

Mass balance indicates that TN and NO3- loads decreased down-estuary each season (despite inputs from the 14 other wastewater treatment plants down-estuary). The 8.4 × 106 ± 4.8 × 106 kg yr-1 of TN exported to the Bay annually is close to the 14.1 × 106 kg yr-1 estimated by Boynton et al. (1995) for the lower Potomac Estuary. The net loss in load for TN along the estuary (9.1 × 106 ± 5.1 × 106 kg yr-1), attributed to burial and denitrification, was also similar to the sum of the burial and denitrification rates estimated by Boynton et al. (1995) for the lower Potomac (13.3 × 106 kg yr-1 of TN). Also, our comparison of net losses in TN along the estuary with independent estimates of burial (Harris, unpublished data), denitrification rate (Cornwell et al., 2016), and assimilation (Boynton et al., 1995) closely align with our estimate for the net loss in load for TN along the estuary. The large loss in TN load attributed to denitrification (68 ± 1 %) is supported by the NO3- isotope data indicating that there was likely denitrification (and assimilation) of NO3-, particularly within 6 km down-estuary from the Blue Plains wastewater treatment plant. Over seasonal timescales, there was a greater percent decline in TN loading during summer and fall, likely due to warmer temperatures and increased biological transformation (attributable to high rates of phytoplankton uptake and detrital deposition; Eyre and Ferguson, 2005; Gillooly et al., 2001; Harris and Brush, 2012; Nowicki, 1994), which suggested that the urban river–estuarine continuum may be more efficient at removing TN during the summer and fall. Compared to summer and fall, winter also had a relatively high percent decline in NO3- loads possibly driven by the higher concentrations typically found in winter months, which could result in quicker assimilation through first-order reaction rate kinetics (Betlach and Tiedje, 1981). Since there was no evidence for denitrification during the winter, burial could also be a mechanism for the relatively high decline in winter months, which is typical of higher flows (Boynton et al., 1995; Milliman et al., 1985; Sanford et al., 2001). However, more work is necessary to evaluate the fate of nitrate using ecosystem process-level measurements.

The higher exports of TN and NO3- to Chesapeake Bay during the winter and spring are due to greater N inputs from the upper and lower watershed and/or greater flow rates. The proportion of N exports attributed to Blue Plains wastewater treatment plant was the highest in the spring, likely due to shorter water residence times (Table 2), resulting in less time for biological uptake, removal, or burial of N. The greater decline in N loads during the spring, however, may be attributed to multiple factors, such as greater N loads being imported from the upper estuary and higher concentrations, compared to summer and fall (Table 1) and thus driving greater losses (from burial and denitrification) due to first-order reaction rate kinetics (Betlach and Tiedje, 1981) similar to winter (described above), stratification that is characteristic of higher flows (Boesch et al., 2001), and increased burial rates due to greater sediment loads during higher flows (Milliman et al., 1985; Sanford et al., 2001). As mentioned previously, more work is necessary regarding linking ecosystem processes and microbial dynamics with the fate of nitrate in the estuary. Nonetheless, the decline in TN and NO3- loads down-estuary each season provide strong evidence for the transformation and retention of N along estuaries.

The Potomac River estuary was a transformer of wastewater N inputs from the Washington D.C. metropolitan area to its confluence with Chesapeake Bay. The values for δ15N-NO3- above the wastewater treatment plant were relatively high, suggesting upriver sources may primarily be from animal waste (Burns et al., 2009; Kaushal et al., 2011; Kendall et al., 2007). This is consistent with a previous study which found that 43 % of N inputs to the upper Potomac River are from manure (Jaworski et al., 1992), while the lower Potomac River has more fertilizer and fewer combined animal feeding operations (CAFOs; US EPA, 2016). Effluent inputs from the Blue Plains wastewater treatment plant significantly increased the δ15N-NO3- values even further, yet this NO3- signal from wastewater disappeared after 20–30 km down-estuary. The increase in δ15N-NO3- and δ18O-NO3- values within the first 1 to 6 km down-estuary suggests denitrification or assimilation of nitrate, due to the lighter δ14N-NO3- and δ16O-NO3- isotopes being preferentially denitrified or assimilated and leaving behind the heavier nitrate isotopes (Granger et al., 2004, 2008; Kendall et al., 2007). But the gradual decline in both δ15N-NO3- and δ18O-NO3- values from 6 to 160 km down-estuary indicates nitrification dominates this portion of the estuary (supported by the nitrate isotope mixing model results) because the process of nitrification, which converts ammonia to nitrate results in lighter nitrate isotopes being generated through fractionation (Kendall et al., 2007; Vavilin, 2014). However, the decline in δ15N-NO3- and δ18O-NO3- loads corresponding to the decline in overall NO3- loads down-estuary also suggests that the heavy nitrate isotopes are being removed as well as the light isotopes. The disappearance of δ15N-NO3- and δ18O-NO3- down-estuary, where NO3- concentrations are very low (∼ 0.01 mg L-1), may indicate that assimilation or even denitrification is occurring on the remaining heavy δ15N-NO3- or δ18O-NO3- after the lighter δ14N-NO3- or δ16O-NO3- is all used up (Fogel and Cifuentes, 1993; Vavilin et al., 2014; Waser et al., 1998a, 1998b).

Seasonal differences in the longitudinal trends for δ15N-NO3- and δ18O-NO3- suggest differences in biological transformations of nitrate due to differences in water temperature, hydrology, and/or N inputs. The δ15N-NO3- values from effluent inputs were higher in warmer months due likely to higher denitrification rates in the wastewater treatment plant associated with warmer water temperatures (Dawson and Murphy, 1972; Pfenning and McMahon, 1997), resulting in elevated δ15N-NO3- values produced by isotopic fractionation (Kendall et al., 2007; Mariotti et al., 1981). An increase in δ15N-NO3- between 2 and 6 km down-estuary during summer and fall (Fig. 5b) further shows increased denitrification or biological uptake due to warmer water temperatures and fractionation (Eyre and Ferguson, 2005; Gillooly et al., 2001; Harris and Brush, 2012; Nowicki, 1994). The significant drop in δ15N-NO3- beyond 10 km down-estuary during summer and fall may have been due to mixing with other N sources and increased nitrification (Wankel et al., 2006), indicated by the salinity mixing line results. During the spring, there was also a significant decline in δ15N-NO3- between 10 and 160 km down-estuary, but this was likely attributed to dilution and nitrification, based on the conservative mixing results. The lack of a significant change during the winter, may be due to shorter residence times (Table 2) and cooler temperatures, contributing to lower biological transformation rates. Further down-estuary, near the mouth of the estuary, the increase in δ18O-NO3- in winter and spring might indicate denitrification in the estuary but in spring nitrate seems conservative based on the salinity mixing plots. The decline in δ18O-NO3- down-estuary in summer and fall suggests that processes other than denitrification in the estuary are controlling the δ18O-NO3-, such as nitrification.

Seasonally, the ∼ 2:1 relationship between δ15N-NO3- and δ18O-NO3- during spring, summer, and fall may indicate denitrification or assimilation, but the salinity mixing plots suggest minimal denitrification in the spring. The fact that the δ15N : δ18O ratio is between 1 and 2 for summer and fall may mean assimilation plays a role, which is supported by previous studies that found a 1:1 relationship for assimilation in the marine environment (Granger et al., 2004; Karsh et al., 2012, 2014). However, other previous studies suggest that a δ15N : δ18O ratio between 1 and 2 can also be caused by denitrifying bacteria (Granger et al., 2008; Lehmann et al., 2003). The divergence from 2:1 ratio may also be attributed to hotspots of denitrification, such as in hyporheic zones where nitrate is completely consumed by denitrification, resulting in no fractionation (Fogel and Cifuentes, 1993; Vavilin et al., 2014; Waser et al., 1998a, b). Additionally, the divergence from the 2:1 ratio in samples further down-estuary may indicate mixing between two or more NO3- sources, such as between atmospheric, marine, or nitrification (Kaushal et al., 2011; Wankel et al., 2006). Due to water column dissolved oxygen levels averaging over 4 mg L-1 (data from Chesapeake Bay Program, not shown), assimilation likely dominates NO3- removal in the water column, while denitrification likely dominates nitrate removal from the sediment, which is supported by previous work (Cornwell et al., 2014; Kemp et al., 1990).

Based on the nitrate isotope mixing model, the longitudinal trends in nitrate sources along the Potomac Estuary correspond to the other results of this study. The decline in wastewater nitrate matched the decline in nitrate concentrations and loads, while the slight increases in nitrification and fertilizer both correspond to a decline in N and O isotope values down-estuary and the increase in agricultural land use in the lower Potomac watershed. Future research would benefit from doing the mixing model separately using different endmembers for the different seasons in order to better constrain the differences between seasons. However, due to lack of data on the seasonality of fertilizer and nitrification endmembers, this was not feasible for the scope of this paper. Seasonal endmembers could provide more confidence because we found that seasonality and temperature mattered in the N sources and loads. Many isotopic studies do not always take this into account and typically just use literature values; our work showed that there are important seasonal variations, and in order to improve the isotope mixing model to capture differences between seasons, the seasonal changes in the endmembers may need to be captured.

Denitrification is likely a sink for NO3- during the summer and fall based on the increases in δ15N-NO3- and δ18O-NO3- within 6 km down-estuary and due to warmer water temperatures, while there is no evidence for denitrification in the winter due to reduced biological activities typical in cooler winter temperatures (Eyre and Ferguson, 2005; Gillooly et al., 2001; Harris and Brush, 2012; Nowicki, 1994). Nevertheless, nitrate removal was significant in all seasons, including winter, suggesting other mechanisms are important, as indicated by the salinity-based mixing lines.

Plots of salinity vs. NO3-, δ15N-NO3-, and δ18O-NO3- were used to provide evidence for conservative mixing, uptake, production, or contributions from other NO3- sources. NO3- concentrations fell below the mixing lines during the summer, fall, and winter, suggesting non-conservative mixing behavior due to the presence of a NO3- sink, such as assimilation, denitrification, or burial (Wankel et al., 2006). However, during the spring, NO3- concentrations fell on the mixing line, indicating that there were no important sources or sinks. This may be due to higher flows and shorter residence times in the spring (Table 2), which can result in fewer biological transformations of NO3-. In the salinity vs. δ15N-NO3- and δ18O-NO3- plots, when the isotope values fell below the mixing lines, this suggested the contribution of NO3- from sources with lower δ15N-NO3- and δ18O-NO3-, such as fertilizer inputs or nitrification, which produces nitrate with lower δ15N-NO3- and δ18O-NO3- values through fractionation (Kaushal et al., 2011; Kendall et al., 2007). An increase in nitrification down-estuary is likely attributed to the conversion of remineralized N to nitrate or from down-estuary inputs of wastewater ammonia that is converted to nitrate (Middelburg and Nieuwenhuize, 2001). During the spring, δ18O-NO3-, isotope values again fell mostly on the mixing line, which may indicate the Potomac River estuary is acting more like a transporter instead of a transformer (e.g., Kaushal and Belt, 2012), transporting NO3- without there being any significant sinks of NO3- or mixing with additional sources, likely due to lower residence times (Table 2) in the spring. However, the fact that during the spring the δ15N-NO3- values were slightly below the mixing line indicates that there may have been an increased amount of nitrate inputs from the watershed through runoff carrying nitrate produced by nitrification. During the winter, δ15N-NO3- values also fell above the mixing line for some samples, which suggested the contribution of heavy δ15N-NO3- from an additional down-estuary source (potentially from one of the 14 other wastewater treatment plants in the lower Potomac watershed). This was likely not the case during the summer and fall, when other sources and sinks may dominate due to greater biological activities (Eyre and Ferguson, 2005; Gillooly et al., 2001; Harris and Brush, 2012; Nowicki, 1994), or during the spring, when there is more conservative behavior due to higher flows. Even though only surface water salinity, nutrient, and isotope values were used in these mixing line plots, when bottom water nutrient and isotope data were averaged with the surface water values, the mixing lines plots and results did not change (data not shown).

The following can be found in the Supplement:

Additional site information and details on methods