Quantifying the Fate of Wastewater Nitrogen Discharged to a Canadian River

Addition of nutrients, such as nitrogen, can degrade water quality in lakes, rivers, and estuaries. To predict the fate of nutrient inputs, an understanding of the biogeochemical cycling of nutrients is needed. We develop and employ a novel, parsimonious, process-based model of nitrogen concentrations and stable isotopes that quantifies the competing processes of volatilization, uptake, nitrification, and denitrification in nutrient-impacted rivers. Calibration of the model to nitrogen discharges from two wastewater treatment plants in the Grand River, Ontario, Canada show that ammonia volatilization was negligible relative to uptake, nitrification, and denitrification within 5 km of the discharge points. 2018-06-20, EarthArXiv preprint 1 1


Introduction:
Nitrogen (N) is essential for life but can be present in the environment in excess of growth requirements due to human activities. N is a common point-source pollutant to aquatic systems from waste-water treatment plants (WWTPs). Nitrate (NO 3 -) and total ammonia nitrogen (TAN; where TAN includes both ammonia (NH 3 ) and ammonium (NH 4 + )) are the two inorganic N forms that determine the critical loads beyond which aquatic ecosystems experience eutrophication or acidification (Posch et al 2001, Schindler et al 2006. The fate of these inorganic N species is a key determinant in the health of ecosystems and the services they provide to humans. TAN can be both a fertilizer of and detriment to aquatic life. At elevated concentrations, NH 3  Many processes remove N from aquatic ecosystems. By understanding the relative contributions of each process and the factors that affect their rates, the environmental fate of N loading to aquatic ecosystems can be predicted (Iwanyshyn et al 2008). Successful nutrient mitigation strategies in larger aquatic ecosystems rely on using smaller, tractable ecosystems as realistic and replicatable systems (Schindler 1998, Sharpley et al 2009, Webster et al 2003, Dodds and Welch 2000, Withers and Lord 2002. The concept of nutrient spiralling in streams was developed to describe the cycling and transport of nutrients in small lotic ecosystem (Newbold et al 1981(Newbold et al , 1982(Newbold et al , 1983 and is based on downstream changes in nutrient concentrations. Isotope tracer experiments, where 15 N-enriched compounds are added and the tracer followed through different pools, have improved spiralling techniques.(e.g. Mulholland et al 2000, Tank et al 2000, Earl et al 2006, Hall et al 2009, Mulholland et al 2004. In a similar fashion, low nutrient streams can be spiked with nutrients and changes in the nutrient pulse can be used to understand ecosystem metabolism of nutrients (e.g. Davis andMinshall 1999, Hall andTank 2003). These studies are often restricted to short lengths of streams where the hydrology can be well characterized and to smaller systems in general. The understanding of nutrient spiralling in large impacted rivers is often confounded by a heterogeneous river morphology, frequent run-of-the-river 49 50 dams, groundwater and multiple nutrient inputs, and consequently relies on the intensive work conducted in these smaller systems supplemented by sampling campaigns of both concentration and stable isotopes of N species. Further, observed values are a cumulative result of a plethora of contemporaneous N cycling processes with rates that change in relative importance with distance from inputs and time of day. Disentangling the relative rates of these processes in large rivers is greatly aided by the additional information supplied by stable isotopes and the development of numerical model (Denk et al 2017).
Stable isotope studies in rivers have shown that (i) NH 4 + is preferentially incorporated into the food web compared to NO 3and (ii) some TAN is lost to volatilization to the atmosphere while some is nitrified to NO 3 - (Loomer 2008, Murray 2008, Hood et al 2014. Denitrification results in N attenuation in rivers, but to a lesser extent in well oxygenated rivers (Rosamond et al 2011, Laursen and Seitzinger 2002, 2004. The rates of these processes change from day to night in response to the release  (Murray 2008, Fourqurean et al 1997, Fry et al 2000, Savage and Elmgren 2004. The isotopic labelling of benthic biofilm by differing NH 4 + and NO 3sources has recently been describe (Hood et al 2014, Loomer et al 2014, Peipoch et al 2014. Here, we build on these studies by developing and testing a model that uses changes in concentrations and natural abundance stable isotopic ratios to quantify the contributions of the various nitrogen-removal pathways in nutrient-impacted rivers. We applied this model to quantify WWTP nutrients. The objectives of this research are to (1) quantify changes in concentrations and δ 15 N values of TAN and NO 3with distance downstream from WWTPs; (2) develop a parsimonious process-based model for N cycling and the fate of WWTP N in rivers, and assess model performance with field measurements; and (3) provide model-based estimates of the rates of nitrification, denitrification, NH 3 volatilization, and N assimilation in WWTP plumes in a river impacted by both WWTP and agricultural nutrient inputs.

Field Site:
The Grand River is the largest river discharging into the Canadian side of Lake Erie ( Figure 1).
Almost 1 million people live in its watershed and more than half of those rely on the river for drinking water. There are 30 wastewater treatment plants of varying sizes in the watershed where agriculture is the dominant land use (80%). We have previously studied the N and O 2 cycling in the Grand River (Rosamond et al 2011, Jamieson et al 2013, Venkiteswaran et al 2014, 2015. Here, we focus on two large WWTPs in the central part of the watershed that serve a combined population of about 230,000. Ecosystems the size of the Grand River are not amenable to experimental isotope tracer additions but nevertheless afford us the opportunity to assess many of the processes resultant from the discharge of nitrogen-rich WWTP effluent. These processes include assimilation of NH 4 + by primary producers, nitrification of NH 4 + to NO 3 -, loss of NH 3 to the atmosphere via volatilization, denitrification of NO 3 -, and dilution of both NH 4 + and NO 3 -. Rather than simply a point-source addition of nutrients to a pristine ecosystem, WWTP effluent in the Grand River increases nutrients in an already nutrient-rich system (Venkiteswaran et al 2015).
The upstream Waterloo WWTP serves an urban population of approximately 120,000 and discharges a mix of NH 4 + and NO 3via a pipe on the west side of the river. The plume hugs that bank of the river for several km downstream. At baseflow, WWTP discharge accounts for 10-25% of river flow along this reach. The downstream Kitchener WWTP serves about 205,000 and discharges mostly NH 4 + via a diffuser in the middle of the river. The plume hugs the east bank of the river for several km downstream before several large river bends result in lateral mixing. The river is about 50m wide through the entire sampling area. Together, the WWTPs discharge about 900 tonnesN/yr (Table S1).
In the study reach, the Grand River flows over the stony and sandy Catfish Creek till (Karrow 1974).
This forms a substrate for the patchy growth dominated by the macroalga Cladophora spp. and Below each of the two WWTPs, eight sampling points were established based on availability of access to the river ( Figure 1). The first site was immediately downstream of the effluent discharge point, one was a few hundred kilometres downstream, and the others about every 800 m to 1000 m for about 5 km (Table S2). At each site, samples for NH 4 + , NO 3 -, Cl -, DOC, δ 15 N-NH 4 + , and δ 15 N-NO 3were collected from the centre of the plume as identified by in situ measurement of conductivity (YSI 556 MPS). Samples were collected in HDPE bottles and immediately chilled in a cooler for transport to the laboratory, filtered to 0.45 µm, and kept cold (4°C) until analyses. Samples for NH 4 + and δ 15 N-NH 4 + were immediately acidified to pH 4 with HCl and frozen until analyses. In situ measurements of temperature and pH were made (YSI 556 MPS) with reported accuracy on pH and temperature of ± 0.2 units and ± 0.15°C, respectively. To account for dilution of the effluent plume by river water, Clat these elevated concentrations was assumed to be a conservative tracer and NH 4 + and NO 3 -2018-06-20, EarthArXiv preprint 5 concentrations were adjusted accordingly.

Analyses:
Anion is converted to NH 3 by increasing the sample pH; NH 3 is trapped in a filter pack containing a 1 cm GF/D filter, acidified with H 2 SO 4 , trapped in a PTFE packet. The filter is dried and analysed for δ 15 N on a Carlo Erba 1108 elemental analyzer (EA) coupled to a Micromass Isochrom isotope-ratio mass spectrometer (IRMS). Precision of δ 15 N-NH 4 + analysis was ± 0.3‰. δ 15 N-NO 3was measured via the AgNO 3 method. Briefly, sample volumes were reduced by evaporation, SO 4 2was removed by barium precipitation, and NO 3was collected on anion exchange resin in a column. After being eluted from the column, AgO was added to precipitate AgNO 3 , which was analyzed on the same EA-IRMS as above.

Model Setup
To interpret patterns in the data, a dynamic model ( (Figure 2). In the model, N 2 O produced by denitrification is allowed to accumulate rather than being further reduced to N 2 ; this choice was made because the N 2 O:N 2 ratio produced during denitrification varies widely and once nitrogen is removed from the TAN and NO 3pools, it is very unlikely to return to those pools especially in a system where N is in excess. Similarly, the biological assimilation of NO 3was not included given that NH 4 + is in excess. Metabolic costs suggest NH 4 + is the preferred source of nitrogen over NO 3for phytoplankton and aquatic plant (Mariotti et al 1982, Yoneyama et al 1991, Collier et al 2012 and that cycling of NH 4 + uptake is rapid (Mulholland et al 2000). Isotopic evidence suggests this is also true for macrophytes in the Grand For the Grand River, the gas exchange coefficient for O 2 has been estimated for its length with focus An initial best-fit solution for each set of field data was found by allowing the model to find a combination of rate constants (greater than or equal to 0), isotopic fractionation factors (between the lowest literature α values, i.e. the strongest values, and 1), and initial values that minimized the sum of squared errors between field data and model output.

Model development: Effect of N cycling processes on coupled N concentrations and isotopes
The coupling of concentrations and isotopes in a simple process-based model shows that the various N cycling processes result in different patterns at the river scale. These results suggest the model may reproduce the variety of expected patterns from each process in the model. Additionally, as we describe next, the dynamic features of each process are sufficiently distinct that we would expect the model to be identifiable. That is, we would expect to arrive at a tight estimate of the kinetic parameters given a sufficiently rich field data set. If this were not the case, then there would be less likelihood that a We fit the resulting model separately to the four field data sets and then, in each case, applied uncertainty analysis as described in Methods. The results varied, but from this preliminary analysis (results not shown) we discovered that in every case the available data was not sufficient to provide accurate estimates of the 8 free parameters. In particular, the k nit2 and α nit2 parameters could not be wellestimated from any of the datasets. Consequently, we reduced the model further, by removing NO 2and instead describe a single-step nitrification process (k nit1 and α nit1 ) where NH 4 + is oxidized to NO 3 -; justified given that NO 2concentrations are low compared to NO 3and TAN and not accumulating.

Discussion:
The process-based NANNO model was able to reproduce the observed dynamics in concentrations and the δ 15 N values of TAN and NO 3 - (Supporting Information Tables 3-6 (Tables S4-S7) but are better compared as the mass of N transformed by each process (Table 2). In three of four cases, NH 3 loss via volatilization was much lower than NH 4 + loss via update or nitrification (Table 2) In both Waterloo cases, denitrification played a modest role in reducing N concentrations ( Table 2). The mass and δ 15 N of river biomass are difficult to capture in the parsimonious NANNO model structure; model fitting may be improved if the release of TAN and NO 3by biomass contributes significantly to river N relative to WWTP effluent (Loomer et al 2014). Nitrogen uptake and release rates can be estimated with nutrient spiralling techniques but this analysis often conflates TAN and NO 3 -. It is therefore difficult to discern which N form is used, which is released, and how these results apply to a river with more than 100 km of upstream nutrient inputs. The degree of importance, if any, to dissolved organic N mineralization or N release from microbes and macrophytes in the nutrient-replete WWTP plumes is unknown.
Understanding the ecosystem effects of changes in nitrogen sources, such as altering WWTPs to produce only NO 3instead of NH 4 + in order to improve river O 2 concentrations, requires knowledge about which N enters the base of the foodweb via primary producers and consumers. In cases where δ 15 N-NO 3and δ 15 N-TAN values are far enough apart, or one is changing while the other is constant, the use of each by primary producers and consumers may be teased apart. NO 3uptake is associated with little to no isotopic fractionation (Mariotti et al 1981, Yoneyama et al 1998, 2001  We have presented a process based-isotopic model of key nitrogen species for use in nutrient plumes in rivers. The NANNO model successfully reproduced observed dynamics in TAN and NO 3 -concentrations and their δ 15 N values including seasonal differences in the way N species were processed. The ability to model these processes is a key step to making predictions about how improvements in WWTP effluent will affect receiving waters.

Acknowledgements:
Canada

Data Deposition:
Data and code are available as part of the NANNO package https://github.com/jjvenky/NANNO (reviewers can anonymously review the code at this URL; upon the paper's acceptance, a DOI will be obtained and used).