A stable isotope approach and its application for identifying nitrate source and transformation process in water

Abstract

Nitrate contamination of water is a worldwide environmental problem. Recent studies have demonstrated that the nitrogen (N) and oxygen (O) isotopes of nitrate (NO₃ ⁻) can be used to trace nitrogen dynamics including identifying nitrate sources and nitrogen transformation processes. This paper analyzes the current state of identifying nitrate sources and nitrogen transformation processes using N and O isotopes of nitrate. With regard to nitrate sources, δ¹⁵N-NO₃ ⁻ and δ¹⁸O-NO₃ ⁻ values typically vary between sources, allowing the sources to be isotopically fingerprinted. δ¹⁵N-NO₃ ⁻ is often effective at tracing NO⁻ ₃ sources from areas with different land use. δ¹⁸O-NO₃ ⁻ is more useful to identify NO₃ ⁻ from atmospheric sources. Isotopic data can be combined with statistical mixing models to quantify the relative contributions of NO₃ ⁻ from multiple delineated sources. With regard to N transformation processes, N and O isotopes of nitrate can be used to decipher the degree of nitrogen transformation by such processes as nitrification, assimilation, and denitrification. In some cases, however, isotopic fractionation may alter the isotopic fingerprint associated with the delineated NO₃ ⁻ source(s). This problem may be addressed by combining the N and O isotopic data with other types of, including the concentration of selected conservative elements, e.g., chloride (Cl⁻), boron isotope (δ¹¹B), and sulfur isotope (δ³⁵S) data. Future studies should focus on improving stable isotope mixing models and furthering our understanding of isotopic fractionation by conducting laboratory and field experiments in different environments.

Introduction

Nitrate contamination of water is a worldwide environment problem that, in most cases, can be attributed to human activities. Ecologically, excess nitrate (NO₃ ⁻) in surface waters can cause eutrophication, a condition that leads to excessive algal and/or plant growth and the severe degradation of the aquatic ecosystem. The consumption of NO₃ ⁻ polluted water is also of concern as it can (1) induce methemoglobinemia in humans, particularly in infants, and (2) increase the risk of cancer and infectious disease (Galloway et al. 2008). As a result, the World Health Organization has set a maximum limit of 10 mg/L of NO₃ ⁻-N for drinking water, whereas the EU (EC 1998) established a threshold of 50 mg/L of NO₃ ⁻-N· for human consumption. An EU Council Directive (EEC 1991) has also been formulated to protect freshwater against NO₃ ⁻ contamination. Nitrate contamination of water is difficult to control. Data from an ETC Water Technical Report (EEA 2010) indicated that from 1992 to 2008, NO₃ ⁻ concentrations in European rivers declined slightly, although some rivers and aquifers showed significantly increasing NO₃ ⁻ concentrations. In the USA, NO₃ ⁻ is also perceived as a problem. For example, NO₃ ⁻ concentrations have increased since 2000 within waters of the upper Mississippi and Missouri Rivers (Murphy et al. 2013), and a study released by UC Davis (Thomas and Lund 2012) warned that the NO₃ ⁻ contamination of groundwater would likely worsen in the coming years, potentially resulting in an 80 % increase in human health risks by 2050.

In light of the above, the protection of surface and drinking water supplies from NO₃ ⁻ contamination is a vital component of any water resource management plan. The protection of surface and groundwater from NO₃ ⁻ pollution requires the identification of the primary NO₃ ⁻ sources within the catchment and the relative contributions of NO₃ ⁻ to the water body. Simply put, it is impossible to control the influx of NO₃ ⁻ to the water if the source(s) is unknown. However, given the wide variety of potential NO₃ ⁻ sources that may exist within a catchment, and the non-conservative behavior of nitrogen (N) as it is dispersed from a source, it is often difficult to determine the predominant NO₃ ⁻ sources using conventional water quality monitoring techniques.

In order to overcome these shortcomings of source identification using monitoring methods, investigators have commonly used three alternative approaches to identifying N sources: (1) export coefficient modeling (Johnes 1996), (2) the analysis of an agricultural pollution potential index (APPI) (Petersen 1991), and (3) non-point source pollution models. Export coefficient modeling is used to identify N sources by calculating N loads from all potential sources. The key step is to determine export coefficients for each source, a process that requires significant observational data. The APPI is a compressive function based on natural and anthropogenic factors, including a runoff index (RI), sediment production index (SPI), chemical use index (CUI), and a people and animal load index (PALI). A shortcoming of this method is that weight determinations of the four indices are subjective and there is no standard approach to assigning the subjective values. Both the export coefficient modeling and APPI methods calculate total N loads rather than the inorganic form of N and neither consider N transformation processes. Non-point source pollution models require a large amount of observational data although they involve N transformation processes. Thus, they are difficult to apply to the areas with limited data.

An approach that is growing in popularity is the use of the N and oxygen (O) isotope systems (¹⁵N-NO₃ ⁻ and ¹⁸O-NO₃ ⁻) to determine both the predominant sources and their relative contributions to a water body. The approach may also serve as a valuable tool to understand N transformation processes (Sun et al. 2011). This approach has been successfully applied during the past several decades to streams and rivers (Mayer et al. 2002; Kaushal et al. 2011; Ma et al. 2015), groundwater (Li et al. 2007; Barnes and Raymond 2010; Jin et al. 2015), and coastal and estuarine systems (Wankel et al. 2009; Korth et al. 2013, 2014; Wong et al. 2015). The approach is based on the realization that NO₃ ⁻ originating from different sources will exhibit differing isotopic compositions. These distinct isotopic “fingerprints” can then be used to quantitatively determine the amount of NO₃ ⁻ that is coming from each of the potential N sources. For example, fertilizers and sewages have been shown to possess a unique range of ¹⁵N-NO₃ ⁻ values (Kendall et al. 2007; Xue et al. 2009), while soil microbial N and atmospherically deposited N often exhibit a unique range of ¹⁸O-NO₃ ⁻ values (Kendall 1998). N transformation processes such as nitrification, assimilation, and denitrification may also lead to distinct isotopic compositions.

This paper summarizes the current state of the N and O isotopes of nitrate as applied to NO₃ ⁻-contaminated water, including (1) identification of NO₃ ⁻ sources in water; (2) assessment of N transformation processes; and (3) the combined use of chloride (Cl⁻), boron (¹¹B), and sulfur (³⁵S) isotopes with the isotopes of N and O.

Utilization of stable isotope approach to determine nitrate sources

Isotope values of nitrate sources

Natural N has two stable isotopes, ¹⁴N and ¹⁵N. Natural O has three stable isotopes, ¹⁶O, ¹⁷O, and ¹⁸O. The abundances of these isotopes in the atmosphere are as follows: ¹⁴N (99.6337 ‰), ¹⁵N (0.3663 ‰) (Junk and Svec 1958), ¹⁶O (99.759 ‰), ¹⁷O (0.037 ‰), and ¹⁸O (0.204 ‰) (Cook and Lauer 1968). The isotopic ratios of both N and O are expressed using the per mille (‰) notation which presents the isotopic ratio in a sample relative to the ratio in a standard, such that

$$ \delta \left({\mbox{\fontencoding{U}\fontfamily{wasy}\selectfont\char104}} \right)=\kern0.5em \frac{{\left(\mathit{\mathsf{R}}\right)}_{\mathrm{sample}}-{\left(\mathit{\mathsf{R}}\right)}_{\mathrm{standard}}}{{\left(\mathit{\mathsf{R}}\right)}_{\mathrm{standard}}}\times \mathsf{1000} $$

(1)

where R is the ¹⁵N/¹⁴N or ¹⁸O/¹⁶O ratio of the sample and standard and δ¹⁵N values are reported relative to atmospheric air (AIR) and δ¹⁸O values are reported relative to the Vienna Standard Mean Ocean Water (VSMOW)(Gonfiantini 1978).

Nitrate in water comes from a variety of sources, including atmospheric deposition, fertilizers and soil N, industrial wastewater, sewage, and manures. All these sources tend to have unique δ¹⁵N-NO₃ ⁻ values or δ¹⁸O-NO₃ ⁻ values (Table 1).

Table 1 The reported δ¹⁵N-NO⁻ ₃ and δ¹⁸O-NO⁻ ₃ values

Atmospheric N is transferred to the Earth’s surface in aqueous solution through rain, snow, fog, and gaseous and particulate species. δ¹⁵N values measured within atmospheric NO₃ ⁻ are usually in the range of −15 to 15 ‰ (Kendall et al. 2007). The δ¹⁸O values of atmospheric NO₃ ⁻ are higher than the δ¹⁵N values ranging from +14 to +75 ‰ (Kendall 1998).

Fertilizers applied to agriculture and urban green spaces are important NO⁻ ₃ sources in water. δ¹⁵N values of inorganic fertilizers have been found to be lower than those of organic fertilizers (Kendall et al. 2007), the former generally ranging from −5.9 to +6.6 ‰ (Bateman and Kelly 2007), whereas the latter from +2 to +30 ‰ (Kendall et al. 2007), respectively. δ¹⁵N-NO₃ ⁻ values of most soil are low ranging from 0 to +8 ‰, while soils polluted by fertilizer and animal waste exhibited average δ¹⁵N-NO₃ ⁻ values of +4.7 ± 5.4 and +14.0 ± 8.8 ‰ (Kendall 1998). δ¹⁵N-NO₃ ⁻ values of industrial wastewater are reported less than 10 ‰ (Li et al. 2005). Compared to fertilizers and soils, sewage and animal waste have higher δ¹⁵N-NO₃ ⁻ values as shown in Table 1.

The δ¹⁸O-NO₃ values of soil and animal waste are low. The δ¹⁸O-NO₃ ⁻ values of nitrate fertilizers are distinct. Amberger and Schmidt (1987) reported that δ¹⁸O-NO₃ ⁻ values of NO₃ ⁻ generated from nitrification of ammonium fertilizers are in the range of −5 to +15 ‰, while δ¹⁸O-NO₃ ⁻ values of synthetic fertilizers are in the range of +17 to +25 ‰.

Stable isotope approach for identifying nitrate sources

As discussed above, nitrate in water comes from a variety of sources. During runoff and infiltration events, NO₃ ⁻ from these sources is transported into surface water and groundwater. Once in the water, the NO₃ ⁻ represents a mixture from all of the differing sources, and determining the amount from each source has proven to be a difficult problem.

The stable isotope approach has been applied to solve this problem. Due to different ways in which N is formed within these potential sources, they are likely to have distinct δ¹⁵N-NO₃ ⁻ values. For example, the δ¹⁵N-NO₃ ⁻ values in fertilizers are low because they are produced by the fixation of atmospheric N₂ (Kendall et al. 2007); manure and sewage have higher δ¹⁵N-NO₃ ⁻ values attributed to ammonia (NH₄ ⁺) volatilization (Kendall et al. 2007), a process that results in a large enrichment of ¹⁵N in the residual NH₄ ⁺ (which is subsequently converted into ¹⁵N-enriched NO₃ ⁻). Thus, natural waters with limited N pollution from manure and sewage will exhibit lower ¹⁵N values (Mayer et al. 2002; Deutsch et al. 2006; Hales et al. 2007). In contrast, if NO₃ ⁻ from manure and sewage enters into the water, the ¹⁵N values will increase, as the isotopic composition of water will move toward that of the NO₃ ⁻ source if no N transformation occurs. As for δ¹⁸O-NO₃ ⁻, atmospheric NO₃ ⁻ has higher δ¹⁸O-NO₃ ⁻ values compared with other nitrate sources (e.g., fertilizers, soil nitrogen). Thus, if unprocessed atmospheric NO₃ ⁻ enters into water, the ¹⁸O-NO₃ ⁻ values will increase. As a result, the source of the NO₃ ⁻ can be identified by analyzing the isotopic composition of nitrate in the water. Similarly, if NO₃ ⁻ from multiple sources enters into water, the isotopic composition of water is a product of the physical mixing of NO₃ ⁻ from these sources, and all these sources can, at least theoretically, be identified and quantified using an isotope mixing model.

The use of N isotope data to determine NO₃ ⁻ sources in water can be traced back to the 1970s. Kohl et al. (1971) estimated the source of NO₃ ⁻ in waters of the Sangamon River in the USA on the basis of natural δ¹⁵N values. The results showed that half of the NO⁻ ₃ was derived from soil and half from unfractionated NO₃ ⁻ found in fertilizer. However, the study by Kohl et al. (1971) was questioned because they discarded potential effects of isotope fractionation on the identified fingerprint as well as the variations in isotopic values within the soil (Hauck et al. 1972). Shearer and Kohl (1988) later suggested that ¹⁵N was better suited to study NO₃ ⁻ transformation processes than to trace NO₃ ⁻ sources. Consequently, the use of ¹⁵N to identify NO₃ ⁻ sources in river waters draining agriculture areas in the USA was limited until the late 1990s. Similarly, δ¹⁸O was rarely used to identify NO₃ ⁻ sources, primarily because of analytical limitations in the technique at the time. With the advancement of isotope technology, of particular importance was the development of the microbial denitrified method (Sigman et al. 2001; Casciotti et al. 2002) which appeared in the early 2000s, a method which can simultaneously analyze for both N and O isotopes, and N and O isotopes were frequently used in identifying NO₃ ⁻ source and have been shown to serve as a powerful tool to identify NO₃ ⁻ sources in water.

During the past decade, many studies have used N and O isotopes to investigate the contributions of NO₃ ⁻ to streams, rivers, and groundwater from terrain characterized by differing land use (Mayer et al. 2002; Burns et al. 2009; Barnes and Raymond 2010; Ohte et al. 2010; Kaushal et al. 2011; Liu et al. 2014). These investigations found that rivers draining areas of agricultural and urban land had higher NO₃ ⁻ concentration and δ¹⁵N-NO₃ ⁻ values (>5 ‰), indicating that sewage and manure were the main NO₃ ⁻ sources. As expected, both NO₃ ⁻ concentration and δ¹⁵N-NO₃ ⁻ values are typically lower (<5 ‰) in forested watersheds (Barnes et al. 2008; Ohte et al. 2010; Schwarz et al. 2011). It has been suggested that these lower values for NO₃ ⁻ in stream waters in forested watersheds result from the atmospheric deposition of NO₃ ⁻, a process that also leads to steam water acidification (Durka et al. 1994), which refers to NH₄ ⁺ assimilated and nitrified by microorganisms and contributes to NO₃ ⁻ in runoff and soil nitrification (Burns et al. 2009). Recent studies (Wang et al. 2013; Ma et al. 2015) found that groundwater and rivers receiving industrial wastewater also have lower δ¹⁵N-NO₃ ⁻ (<5 ‰). High δ¹⁸O-NO₃ ⁻ values derived from atmospheric nitrate are always reduced through nitrogen biological processes and are similar to those values derived from fertilizers, sewage, and soils. Thus, oxygen isotope may be not effectively used to identify nitrate sources derived from human activities but may be used to distinguish nitrate derived from unprocessed atmospheric nitrate from nitrogen biochemical processes such as nitrification. Koszelnik and Gruca-Rokosz (2013), for example, argued that NO₃ ⁻ from wells and ditches located in hardwood forests was derived from the soil where nitrification generated δ¹⁸O-NO₃ ⁻ values in the range of those typically associated with nitrification. Similarly, Tobari et al. (2010) estimated the contribution of atmospheric NO₃ ⁻ in streams draining forest watersheds in Japan by using O isotopes. The results indicated that the contribution of atmospheric NO⁻ ₃ in stream water was higher in young forests than in old forests. The higher δ¹⁸O-NO₃ ⁻ values discharged from young forests were attributed to atmospheric NO₃ ⁻ that was not altered because of the lower level of plant NO₃ ⁻uptake by the younger trees. In shallow groundwater, NO₃ ⁻ distributions are also influenced by land use. In the West Lake watershed, eastern China (Jin et al. 2015), nitrogen and oxygen isotopic data indicated that nitrate sources in groundwater were soil N from forest, chemical fertilizers and manure in the tea garden, and domestic sewage from the old residential area in forest and urban area. In deep groundwater, Murgulet and Tick (2013) found that δ¹⁸O-NO₃ ⁻ and δ¹⁵N-NO₃ ⁻ values in the coastal aquifer of the Alabama, USA, indicated that the mixing ratios of sewage and fertilizer were consistent with the study area’s land use. Singleton et al. (2005) found that industrial sources contributed to the major nitrate plumes in groundwater below Hanford chemical processing facilities in south-central Washington, USA. Human activities have a direct effect on δ¹⁵N-NO₃ ⁻ and δ¹⁸O-NO₃ ⁻ values in water. Mayer et al. (2002) found that mean ¹⁵N-NO₃ ⁻ values in water increased directly with increases in the area of agricultural and urban lands in the watershed. Burns et al. (2009) also found that in the urban watershed of Lisha Kill in the USA, δ¹⁸O-NO₃ ⁻ values were significantly related to the amount of runoff from impervious areas.

Temporal changes in system hydrology have been shown to significantly influence δ¹⁵N-NO₃ ⁻ and δ¹⁸O-NO₃ ⁻ values, potentially complicating the use of N and O isotopes as tracers of NO₃ ⁻ provenance. Kaushal et al. (2011), for example, found that δ¹⁵N-NO₃ ⁻ values were higher during low-flow conditions than during high flow in urban watersheds, suggesting that δ¹⁵N-NO₃ ⁻ values decreased with increasing runoff from agricultural watersheds. These results were consistent with those of Silva et al. (2002). In contrast to δ¹⁵N-NO₃ ⁻, δ¹⁸O-NO₃ ⁻ values were higher during high-flow conditions than during low-flow conditions (Barnes and Raymond 2010; Kaushal et al. 2011; Ding et al. 2014). These differences can largely be attributed to the greater proportion of unprocessed atmospheric NO₃ ⁻ entering into streams or rivers directly because high recharge rates shortened residence times. In other words, NO₃ ⁻ enters the runoff without being dissolved in the soil solution and, in turn, without being denitrified as there is a decrease of soil NO₃ ⁻ dissolution. In groundwater, however, temporal changes seem to have few influence on δ¹⁵N-NO⁻ ₃ and δ¹⁸O-NO₃ ⁻ values (Baily et al. 2011; Wexler et al. 2012; Jin et al. 2015), which may be due to surface runoff entering shallow aquifers and mixing with groundwater there (Wexler et al. 2012). But, in the groundwater body of the Brussels sands (Belgium), δ¹⁵N-NO⁻ ₃ and δ¹⁸O-NO₃ ⁻ values displayed a clear temporal pattern due to groundwater recharge dynamics and cycling process of nitrogen in the soil nitrogen pool (Mattern S et al. 2011). Temperature also appears to effect δ¹⁵N-NO₃ ⁻ and δ¹⁸O-NO₃ ⁻ values. Burns et al. (2009) found that δ¹⁵N-NO₃ ⁻ values are positively related to air temperature in two agricultural watersheds with the correlation coefficient of 0.87 and 0.84, respectively. Barnes and Raymond (2010) found that δ¹⁵N-NO₃ ⁻ and δ¹⁸O-NO₃ ⁻ values were positively related to water temperature. These values are explained by higher rates of denitrification at warm temperatures resulting in increases in δ¹⁵N-NO₃ ⁻ values and δ¹⁸O-NO₃ ⁻ values by isotope fractionation (discussed in the “Denitrification” section). Catchment size influences δ¹⁵N-NO₃ ⁻ and δ¹⁸O-NO₃ ⁻ heterogeneities. For example, Danielescu and MacQuarrie (2013) found that δ¹⁵N-NO₃ ⁻ values in the smaller watershed of McIntyre Creek were less variable than those in the larger Trout River, whereas δ¹⁵N-NO₃ ⁻ values exhibited a logarithmic relationship with the catchment size in Canada. Ohte (2013) also found that δ¹⁸O-NO₃ ⁻ values in a river in central Japan were 10 ‰ lower than that those observed in the northeastern USA, where the former’s area is 100 times smaller than the latter’s.

The data highlighted above show that seasonal and spatial changes are the primary factors influencing δ¹⁵N-NO₃ ⁻ and δ¹⁸O-NO₃ ⁻ values. Seasonal factors include precipitation, runoff, and air or water temperature, and spatial factors include land use type such as impervious urban areas, agricultural farmland, and forests of varying age. These factors may not cause the alteration of isotopic compositions directly; rather, they may be the result of changes associated with microbial activities (Fig. 1) that control N transformation processes. In other words, the signatures are produced by isotopic fractionation (discussed in “Utilization of stable isotope approach to determine nitrogen transformation processes” section) that alters δ¹⁵N-NO₃ ⁻ and δ¹⁸O-NO₃ ⁻ values. Consequently, the isotopic composition (or fingerprint) of the NO₃ ⁻ within a particular source does not remain constant but may be altered when being transported into water.

Fig. 1

The main factors influencing isotopic fractionation of nitrogen and oxygen isotopes of nitrate

Quantification of contributions of nitrate sources

The traditional method for quantifying the relative contributions of NO₃ ⁻ from individual sources requires monitoring NO₃ ⁻ concentration and stream flow, which are used to estimate NO₃ ⁻ load in a particular period (e.g., day, month, season, or year). NO₃ ⁻ concentration is usually monitored less frequently than stream flow because of the higher costs of sampling and laboratory analyses (Quilbé et al. 2006). Moreover, monitor stations are usually sited at the outlet of the basin or watershed where waters contain NO⁻ ₃ from multiple sources. Thus, it may not be possible to use the traditional method to quantitatively estimate the contribution of each NO₃ ⁻ source. In these instances, contributions may be estimated using statistical mixing models.

Since the isotopic signature of the NO₃ ⁻ sources is distinct, it is possible to determine the source of the NO₃ ⁻ as the isotopic values in the water reflect the physical mixing of isotopes from two or more sources.

Basic mass balance isotope mixing models

N and O isotopes can be used to quantify the contributions of NO₃ ⁻ sources to water based on the assumption that any change in isotopic composition is the direct result of mixing of two or more sources of known composition.

A basic mass balance mixing model (Moore and Semmens 2008) is defined as follows:

$$ {\delta}_M={f}_1\left({\delta}_1+{\gamma}_1\right)+{f}_2\left({\delta}_2+{\gamma}_2\right){f}_n\left({\delta}_n+{\gamma}_n\right) $$

(2)

where f _n is the proportional contribution of the source to the mixture, δ _n is the isotopic signature of the source, and γ _n is the isotope-specific fraction of the source. According to the model, these equations can be solved exactly for the contributions of the different sources when the number of sources is less than or equal to the number of isotopes + 1. When applied to the combined use of N and O isotopes to determine the relative contributions of NO₃ ⁻ sources in waters, the model takes the following form as described by Xue et al. (2009):

$$ \begin{array}{c}\hfill {\delta^1}^5{N}_M={f}_1{\delta}^{15}{N}_1+{f}_2{\delta}^{15}{N}_2+{f}_3{\delta}^{15}{N}_3\hfill \\ {}\hfill {\delta^1}^8{O}_M={f}_1{\delta}^{18}{O}_1+{f}_2{\delta}^{18}{O}_2+{f}_3{\delta}^{18}{O}_3\hfill \\ {}\hfill 1={f}_1+{f}_2+{f}_3\kern9em \hfill \end{array} $$

(3)

Deutsch et al. (2006) used the model successfully to estimate relative source contributions of NO₃ ⁻ to tile drainage water. Voss et al. (2006) also demonstrated the usefulness of the mixing model for identifying NO₃ ⁻ sources by estimating the contributions of NO₃ ⁻ from various sources (including agriculture, pristine soil, and atmospheric deposition) in catchments of the Baltic Sea. Kaushal et al. (2011) used the model to estimate the contributions of NO₃ ⁻ from sewage and atmospheric deposition in an urban watershed in Baltimore, Maryland (USA).

Bayesian mixing models

The three-equation model (Eq. 3) discussed above can only estimate the contributions of two or three NO₃ ⁻ sources. In addition, the particular mass balance mixing model does not include an analysis of the potentially substantial and multiple sources of uncertainty (Moore and Semmens 2008) (e.g., those associated with the temporal and spatial variability in δ¹⁵N-NO₃ ⁻ and δ¹⁸O-NO₃ ⁻ values described earlier). To overcome these shortcomings, Moore and Semmens (2008)) developed a statistical stable isotope mixing model based on a Bayesian framework referred to as MixSIR. MixSIR considers the temporal and spatial variability of isotopes in the source materials (including those associated with fractionation), and the number of sources solved by several statistical treatments was unlimited.

Nevertheless, Jackson et al. (2009) argued that the MixSIR model could not accurately quantify contributions of NO₃ ⁻ from the sources, since the model could not provide a reliable estimate for additional unquantified error and beta distributions misled the prior proportions. So, he developed another statistical model also based on Bayesian framework referred to as SIAR, which used the Dirichlet distribution instead of beta distribution as the prior distribution. The model by Jackson et al. (2009) takes the form of the following:

$$ \begin{array}{c}\hfill {X}_{ij}={\displaystyle {\sum}_{k={1}^pk}^K\left({S}_{jk}+{S}_{jk}+{c}_{jk}\right)}+{\varepsilon}_{ij}\hfill \\ {}\hfill {S}_{jk}\sim N\left({\mu}_{jk},{\omega^2}_{jk}\right)\kern6.5em \hfill \\ {}\hfill {c}_{jk}\sim N\left({\lambda}_{jk},{\tau^2}_{jk}\right)\kern6.5em \hfill \\ {}\hfill {\varepsilon}_{jk}\sim N\left(0,{\sigma}_j^{2\kern1em }\right)\kern7em \hfill \end{array} $$

(4)

where X _ij is the observed isotope value j of the mixture i, in which i = 1,…, N; j = 1,…, J; S _jk is the source value k on isotope j (k = 1,…, K) that is normally distributed with a mean μ and standard deviation ω; p _k is the proportion of source k which is to be estimated by the model; c _jk is a fractionation factor for isotope j on source k that is normally distributed with mean λ and standard deviation τ; and ε _jk is the residual error representing additional unquantified variations between individual normally distributed with mean = 0 and standard deviation σ.

The SIAR model was used by Xue et al. (2012) to estimate the contributions of NO₃ ⁻ from several sources in the Flanders watershed, Belgium. The results showed that the contribution of manure and sewage was highest from 32 to 49 % to the water, whereas the contribution of ammonium fertilizer and NO⁻ ₃ precipitation was minimal, ranging between 2 and 8 %.

Compared with the basic mixing model, the MixSIR and SIAR models have advantage of considering isotope variations in the NO₃ ⁻ sources and fractionation processes. However, some factors may influence the accuracy of the models. First, the basic assumption that the isotopic values of the sources are unique and can be used to fingerprint a source may be violated since the isotope values of different sources may overlap (e.g., fertilizer and soil N, manure, and sewage). Second, the accuracy of the two models decreases with the increasing number of sources. Third, the larger range of isotope values within the sources, as a result of the discussed temporal and spatial factors, will reduce the accuracy of the models. In contrast, a smaller range of isotope values will increase the accuracy of the results.

In order to reduce and quantify the uncertainty of the models, the study area should be divided into multiple, smaller sections to reduce NO₃ ⁻ sources and the variability of isotopic values resulting from spatial factors such as land use. In addition, the isotopic composition of each potential source within the study area should be analyzed instead of using reported isotopic NO₃ ⁻ data provided by other studies.

Utilization of stable isotope approach to determine nitrogen transformation processes

Nitrogen exists in various chemical forms. Transformations between these chemical forms are often induced by microbes, during which N and O isotopic fractionation may occur. In most cases, isotopic fractionation results in an increase in the heavy isotopes within the residue N pool. As discussed above, isotopic fractionation can complicate their use for purposes of NO₃ ⁻ source identification because the original isotopic fingerprint of the source materials is altered. However, on the other hand, the nature and degree of isotope fractionation can also provide a powerful tool (Wexler et al. 2011) for tracing N transformation processes.

Stable isotope fractionation

Stable isotopes are subjected to kinetic isotope fractionation, which is the variations of isotope composition in different phase and compounds due to geochemical processes. Kinetic fractionations often result in the light isotope accumulating in the product and heavy isotope accumulating in the substrate (Sulzman 2007). The fractionation factor is defined as follows:

$$ a={R}_{\mathrm{p}}/{R}_{\mathrm{s}} $$

(5)

where R is the ratio of heavy to light isotopes in the instantaneous product (R _p) and substrate (R _s). The enrichment in the heavy isotope in the remaining substrate is characterized by an enrichment factor (ε) in ‰ which is defined as (Xue et al. 2009) follows:

$$ \varepsilon ={10}^3\left(a-1\right) $$

(6)

In a closed system, isotopic enrichment can be expressed by the Rayleigh equation (Xue et al. 2009) as follows:

$$ {\delta}_{\mathrm{S}(t)}={\delta}_{\mathrm{S}0}+\varepsilon 1n\left({S}_t/{S}_0\right) $$

(7)

δ _S0 and δ _S(t) are the isotopic compositions of the substrate at time 0 and t, and S ₀ and S _t are the concentrations of the substrate at time 0 and t, respectively.

Nitrogen does not always behave in a conservative manner (Li et al. 2013; Badruzzaman et al. 2012) when transporting through aquatic systems, during which N transformation processes may occur. Both nitrogen and oxygen isotopic fractionations exist during N transformation processes, and the fractionation factors in various N transformation processes are different. For example, the fractionation factor during nitrogen fixation is small about 1.004 (Delwiche and Steyn 1970), whereas the factors during nitrification and denitrification are high about 1.035 (Mariotti et al. 1981) and from 1.01 to 1.03 (Mariotti et al. 1982), respectively. In addition, the changes of relationship between nitrogen and oxygen isotopes originated from isotope fractionation are distinct during various N transformation processes. For example, δ¹⁵N and δ¹⁸O values of residual NO₃ ⁻ increase in both processes (denitrification: 2:1, Burns et al. 2009; assimilation: 1:1, Granger et al. 2004). Nitrogen and oxygen isotopes of nitrate can be used to assess N transformation processes since different N transformation processes lead to a distinct isotopic fractionation signature. Lehman et al. (2004) pointed out that the N and O isotopes of nitrate allow the effects of simultaneous nitrification and denitrification to be separated and used isotope analysis of nitrate (N and O) in benthic chambers deployed in Santa Monica Bay to show that benthic nitrification co-occurred with denitrification. A recent study (Dale et al. 2014a, b) used an isotope box model to calculate nitrate fractionation in bottom water of offshore NW Africa and found low benthic isotope fractionation during nitrification and high isotope fractionation during denitrification.

Stable isotope approach for identifying nitrogen transformation processes

Denitrification

Denitrification refers to microbial reduction process that converts NO₃ ⁻ to gaseous products. It occurs in soils, aquifers, riparian zones, and benthic and river sediment where O₂ concentrations are less than 20 μΜ (Kendall et al. 2007). Denitrification reduces NO₃ ⁻ in water from terrestrial ecosystems, thereby reducing the degree of NO₃ ⁻ pollution.

During denitrification, the fractionation of both N and O isotopes occurs. Laboratory experiments (Vidal-Gavilan et al. 2013; Vavilin and Rytov 2015) showed that the N and O isotope fractionations were −13.0 and −17.1 ‰ for ε _N and −8.9 and −15.1 ‰ for ε _O. In riparian zones, nitrogen isotopic fractionation caused by denitrification is more apparent being on the order of about −18 ‰ (Sebilo et al. 2003). In groundwater, N isotope enrichment factors are reported in the range of −13.9 to −30 ‰ (Li et al. 2014 and references therein). Nitrogen enrichment factors associated with benthic denitrification appears to be smaller, ranging from −1.5 to −6 ‰ (Alkhatib et al. 2012), which is influenced by NO₃ ⁻ diffusion across the water-sediment interface, a process that is the rate-determining step (Sebilo et al. 2003; Lehman et al. 2004). As denitrification is a redox reaction, during which process, two types of electron donor including organic carbon and sulfur compounds participate in the reaction, so the denitrification fate and isotopic fractionation factors depend on the enrichment of electron donors besides NO₃ ⁻ concentration and anoxic conditions. For example, in the bottom of Pétrola lake with organic matter-rich sediment, isotopic enrichment factors are high (ε _N = −14.7 ‰; ε _O = −14.5 ‰)(Carrey et al. 2014), while in shallow saturated zone, Clague et al. (2015) found that low enrichment factors (ε _N = −1.1 to −9.6 ‰, ε _O = −1.0 to −7.2 ‰) caused by denitrification are a result of denitrification primarily occurring in the small pores, rather than in the freely moving bulk of the groundwater system, because of longer contact time between microbes, electron donors, and NO₃ ⁻.

Denitrification causes both δ¹⁵N and δ¹⁸O values of the residual NO₃ ⁻ to increase. Sigman et al. (2005) and Granger et al. (2008)) found that δ¹⁵N and δ¹⁸O values of residual NO₃ ⁻ increased at a ratio of 2 during denitrification. However, some other surface and groundwater studies found that denitrification caused δ¹⁸O and δ¹⁵N to increase at a ratio of more than 1:1 (Minet et al. 2012; Dale et al. 2014a, b) and some closed to 2 (Burns et al. 2009; Baily et al. 2011; Critchley et al. 2014; Wexler et al. 2014) in agricultural watersheds. However, these lower ratios were not found in urban watersheds (Kaushal et al. 2011), presumable because denitrification was limited by less organic carbon (Barnes and Raymond 2010; Wong et al. 2015). The reported ratio of δ¹⁵N-NO₃ ⁻ and δ¹⁸O-NO₃ ⁻ values (1.13 to 2.1) (Vidal-Gavilan et al. 2013; Wexler et al. 2014) as being indicative of denitrification is widely used in groundwater system and groundwater surface-water interactions. The relationship between enriched δ¹⁵N and δ¹⁸O values of the residual NO₃ ⁻ is used by Critchley et al. (2014) to provide evidence of in situ denitrification after a cross-injection experiment in glacial sand gravel aquifer. Izbicki et al. (2015) pointed out that storage of nitrate in thick unsaturated zones and dilution with low-nitrate groundwater rather than denitrification are the primary attenuation mechanisms for nitrate from septic discharges in an arid area in California, based on the weak relationship between δ¹⁵N-NO₃ ⁻ and δ¹⁸O-NO₃ ⁻ values, even though denitrifying bacteria are abundant. Gómez-Alday et al. (2014) also found that weak relationship between δ¹⁵N-NO₃ ⁻ and δ¹⁸O-NO₃ ⁻ values in regional groundwater flow from recharged areas to the lake indicates that NO₃ ⁻ is only partially denitrified. It should also be remembered that the identification of denitrification may also be hidden by the mixing of N and O isotopes from multiple sources. In groundwater, long groundwater transit times lead to the water being mixed sufficiently with seasonality infiltration events, which covered the nitrogen and oxygen isotopic signature of nitrate during denitrification process (Clague et al. 2015). The same case also occurs in the northern hardwood forests at the Hubbard Brook Experimental Forest in the White Mountains of New Hampshire, USA. Wexler et al. (2014) found that both δ¹⁵N-NO₃ ⁻ and δ¹⁸O-NO₃ ⁻ values of a few samples fell below the denitrification line, which might be attributed to a mix of partially denitrified NO₃ ⁻ in groundwater and the NO₃ ⁻ in stream water when a hyporheic flow path brought water from the stream to the shallow groundwater in the riparian zone. One possible means of differentiating denitrification from the mixing of source materials is to analyze the relationship between δ¹⁵N-NO₃ ⁻ values (or δ¹⁸O-NO⁻ ₃) and NO₃ ⁻ concentrations (Kendall et al. 2007). According to the Rayleigh equation (Eq. (7)), denitrification leads to an exponential relation between δ¹⁵N-NO₃ ⁻ values and NO₃ ⁻ concentrations. The δ¹⁵N-NO₃ ⁻ values and NO₃ ⁻ concentrations of mixtures of two sources must plot on a line between those values of two sources in δ¹⁵N-NO₃ ⁻ values versus NO₃ ⁻ concentration plots, if the NO₃ ⁻ concentrations of the two sources are identical, the mixing line is straight, and, otherwise, the mixing line is hyperbolic. Hence, if two sources are mixed, δ¹⁵N-NO₃ ⁻ (or δ¹⁸O-NO₃ ⁻) is linearly positively related to the reciprocal of NO₃ ⁻ concentration. In contrast, if denitrification occurred, δ¹⁵N-NO₃ ⁻ (or δ¹⁸O-NO₃ ⁻) is linearly negatively related to the logarithm of NO₃ ⁻ concentration. Wong et al. (2015) pointed out that the distribution patterns of the δ¹⁵N-NO₃ ⁻ versus NO₃ ⁻ concentrations and the inverse (1/NO₃ ⁻) and logarithmic forms of NO₃ ⁻ in an agricultural-urban coastal aquifer in Werribee Plain, Australia, are indicative of conservative mixing nitrate sources from sewage and fertilizers. Lohse et al. (2013) confirmed that NO₃ ⁻ sources in a small catchment in California were primarily derived from mixing of microbial NO₃ ⁻ and denitrification. Peng et al. (2012) analyzed NO₃ ⁻ sources and denitrification within the Chi-Chia-Wan watershed of Taiwan and pointed out that spatial variation of NO₃ ⁻ was attributed to mixing of different sources rather than denitrification.

Assimilation

Assimilation refers to the process during which inorganic nitrogen is transformed into an organic form by living organisms during biosynthesis. Creatures prefer to assimilate light δ¹⁴N rather than heavy δ¹⁵N. Some experimental laboratory and field results (e.g., Granger et al. 2004; Montoya et al. 1991) have shown that nitrogen enrichment factors fall within a large range of −30 to 0 ‰ during phytoplankton assimilation of NO₃ ⁻ and NH₄ ⁺. In the estuary, the nitrogen enrichment factors fall within a narrow range of −13 to −6 ‰ (Kendall et al. 2007). Algae prefer to assimilate NH₄ ⁺ rather than NO₃ ⁻ even when NO₃ ⁻ concentrations are higher than NH₄ ⁺ concentrations in the system (Dortch 1990). Consequently, NO₃ ⁻ assimilation is prevented by NH₄ ⁺ (Sugimoto et al. 2011). The δ¹⁵N and δ¹⁸O of residual NO₃ ⁻ are enriched at a ratio of 1:1 during NO₃ ⁻ assimilation (Granger et al. 2004). Wankel et al. (2009) point out that combined use of δ¹⁸O and δ¹⁵N allows the separation between the enrichment of δ¹⁵N through assimilation processes, which would cause a positive covariance, the production of NO₃ ⁻ through regeneration, and the sewage-derived nutrients.

Deutsch et al. (2009) analyzed NO₃ ⁻ assimilation in Elbe River in German. The results suggested that δ¹⁵N-NO₃ ⁻ and δ¹⁸O-NO₃ ⁻ increased at a ratio of 0.89:1, indicating that NO₃ ⁻ assimilation was the main N transformation process. Korth et al. (2013) found that in Szczecin lagoon, δ¹⁵N-NO₃ ⁻ and δ¹⁸O-NO₃ ⁻ values near the surface of the outflowing water increased at a ratio of 1.2:1 indicating that NO₃ ⁻ assimilation was occurring. In a groundwater-fed estuary, Wong et al. (2014) used δ¹⁵N-NO₃ ⁻ and δ¹⁸O-NO₃ ⁻ in the Werribee River estuary to show that the shallow groundwater contributed 60 to 76 % of the NO₃ ⁻ to the estuary, and assimilation removed ∼70 % of this groundwater-derived NO₃ ⁻ during dry periods. Swart et al. (2013) provided evidence that wastewater had been injected into the aquifer and leaked back to the surface based on the monitored δ¹⁵N-NO₃ ⁻ values. δ¹⁵N-NO₃ ⁻ and δ¹⁸O-NO₃ ⁻ values in their study indicate that dissolved nitrate in the Biscayne Bay originated from anthropogenic sources and experience assimilation process.

Nitrification

Nitrification is an oxidation process in which NH₄ ⁺ is transformed to NO⁻ ₃, a process mediated by several different kinds of autotrophic bacteria or Archaea. Nitrification has two steps. The first is NH₄ ⁺ oxidation to NO₂ ⁻, and the second is NO₂ ⁻ oxidation to NO₃ ⁻. Nitrification often occurs in aerobic conditions. At the Massachusetts Military Reservation (MMR), chemical and isotopic evidence indicated that nitrification occurred near the contact between the plume and the oxic overlying groundwater, whereas it did not occur in the anoxic zone (Böhlke et al. 2006). Wexler et al. (2012) found that nitrification occurred in surface after mixing of shallow groundwater and flow from surface drainage based on isotope composition of nitrate. In groundwater underlying the Civic Center area of Malibu, CA, isotope of nitrate associated with nitrogen concentrations indicated that nitrification also occurred in the saturated zone after mixing of treated sewage discharges and native groundwater (Izbicki, 2014). Savard et al. (2007) pointed out that in groundwater, nitrification takes place throughout all seasons. Some studies have indicated that the rate of nitrification was equal to or even higher than the rate of denitrification in streams (Webster et al. 2003; Arango and Tank 2008), rivers (Richardson et al. 2004), and valleys (Lohse et al. 2013).

Nitrification causes isotope fractionation; in this case, there is a decrease in δ¹⁵N-NO₃ ⁻ and δ¹⁸O-NO₃ ⁻ values in the residual pool. The most nitrogen fractionation is associated with the first step during which NH₄ ⁺ is transformed to NO₂ ⁻ (Kendall et al. 2007). Nitrogen enrichment factors during the nitrification process are in the range of −38 to −14 ‰ (Casciotti et al. 2003). The extent of N fractionation is influenced by NH₄ ⁺ concentrations in the system (Casciotti et al. 2003). Large amounts of NH₄ ⁺ stimulate nitrification and lead to a large fractionation of nitrogen isotope. However, as the NH₄ ⁺ pool is consumed, the nitrification rate decreases resulting in a slight fractionation of N isotope and a minor δ¹⁵N enrichment in NO₃ ⁻ (Böhlke et al. 2006). Böhlke et al. (2006) found that nitrification reaction was highly localized within a narrow zone of mixing between contaminated and uncontaminated water with an enrichment factor in the range of −20 to −9 ‰ in shallow groundwater. In unsaturated zone, nitrification rate and nitrogen isotope fractionation factors are affected by soil moisture. Laboratory experiment (Yun and Ro 2014) showed that with soil water potentials from −1100 to −11 kPa, nitrification rate increased from 2.3 to 9.8 mg N kg⁻¹ day⁻¹, and nitrogen isotope enrichment factor increased from −14.7 to −33.8 ‰.

Oxygen isotope fractionation is complex during nitrification. Previous studies have shown that O atoms within NO₃ ⁻ originated from water and dissolved O (Andersson and Hooper 1983; Hollocher 1984) formed during nitrification, one atom from dissolved O₂ and two atoms from water. Thus, the formed NO₃ ⁻ exhibits an isotopic signature equal to 2/3 δ¹⁸O of water and 1/3 δ¹⁸O of dissolved O (Eq. (8)) if no kinetic fractionation occurs. Consequently, δ¹⁸O-NO₃ ⁻ values formed during nitrification are in the range of −10 to +10 ‰ (Kendall et al. 2007).

$$ {\delta}^{18}{O}_{\mathrm{NO}3}=2/3\left({\delta}^{18}{O}_{{\mathrm{h}}_2\mathrm{O}}\right)+1/3\left({\delta}^{18}{O}_{{\mathrm{O}}_2}\right) $$

(8)

Equation (8) has been used by some researchers to identify nitrification in water. For example, Thibodeau et al. (2013) found that nitrification was not the main N transformation process in the St. Lawrence River since δ¹⁸O-NO₃ ⁻ was negatively related to δ¹⁸O-H₂O. Lohse et al. (2013) found that the measured δ¹⁸O-NO₃ ⁻ values in a stream in Marin, CA, were consistent with calculated values based on Eq. (8); thus, so they concluded that nitrification occurred. However, some studies (Casciotti et al. 2010; Buchwald and Casciotti 2010) found that δ¹⁸O-NO₃ ⁻ values produced by nitrification were significantly higher than theoretical values in surface water. This might be attributed to evaporation (Böhlke et al. 1997), respiration (Kendall 1998), and higher δ¹⁸O values in soil (Wassenaar 1995) that are caused by seasonal changes or different nitrobacteria (Mayer et al. 2001). In groundwater, however, Mongelli et al. (2013) found that δ¹⁸O-NO₃ ⁻ values of groundwater in the central-southern Italy during nitrification process are lower than the expected δ¹⁸O-NO₃ ⁻ values, which might be explained by that the relative contribution of ambient O from surrounding water and O₂ did not always keep the rule of 2:1 (Mayer et al. 2001). In addition, a recent study (Kool et al. 2011) confirmed the exchange of O atoms between nitrite and water during nitrification in soil which also resulted in a lighter measured δ¹⁸O-NO₃ ⁻ values than predicted δ¹⁸O-NO₃ ⁻ values. Temperature and runoff are fundamental controls on δ¹⁸O-NO₃ ⁻ values from nitrification. Higher temperature stimulates microbial respiration with the enrichment of δ¹⁸O-O₂ (Kendall 1998), and produced NO⁻ ₃ is transported into waters. Savard et al. (2007) also noted that in temperate regions, δ¹⁸O-NO₃ ⁻ values produced by nitrification were largely determined by hydrological and climatic conditions (hydroclimatic index) and were significantly positively related to δ¹⁸O-NO₃ ⁻ values. Gammons et al. (2011) pointed out that diurnal variability of oxygen isotopic compositions of dissolved oxygen (δ¹⁸O-DO) should be considered in the analysis of δ¹⁸O-NO₃ ⁻ derived from nitrification, because δ¹⁸O-DO caused by plant or algae respiration would directly influence δ¹⁸O-NO₃ ⁻ values.

Combining nitrogen and oxygen isotopes of nitrate with chloride, boron, and sulfur isotope

From the above analysis, we have shown that determining the nature and degree of isotopic fractionation provides a means of assessing N transformation processes as well as NO⁻ ₃ source(s). However, other types of data may be combined with O and N isotopic information to more fully identify NO⁻ ₃ source(s) and N transformation processes. The following sections describe the use of chloride, boron isotope, and sulfur data to further our analysis.

Combining nitrogen and oxygen isotopes of nitrate with chloride

Chloride (Cl⁻) is a conservative substance, meaning that it moves with the water without any loss in elemental mass. It exists in a wide range of fertilizers, animal waste, and sewage. Thus, Cl⁻ can be combined with δ¹⁵N-NO₃ ⁻ and δ¹⁸O-NO₃ ⁻ to identify NO⁻ ₃ sources and N transformation processes. Silva et al. (2002) found that δ¹⁵N-NO₃ ⁻ values were positively related to Cl⁻ concentrations in an urban watershed, indicating that the main NO⁻ ₃ source was mixing between baseflow and stormflow. Chen et al. (2014) found that mixing processes played an important role in Taihu Lake in China by combining NO₃ ⁻/Cl⁻ data with δ¹⁵N-NO₃ ⁻. In a study on groundwater NO₃ ⁻ contamination in Osona, Spain (Vitòria et al. 2008), δ¹⁸O-NO₃ ⁻ values increased with a decrease in NO₃ ⁻/Cl⁻ ratios and exhibited a relative decrease in NO₃ ⁻ concentrations compared to Cl⁻. These trends indicated that denitrification was occurring. However, without the linear relationship between δ¹⁵N-NO₃ ⁻ and δ¹⁸O-NO₃ ⁻, NO₃ ⁻/Cl⁻ ratios instead of the NO₃ ⁻ concentrations dispense with the effect of other NO₃ ⁻ sources. Some studies on NO₃ ⁻ sources in rivers in Korea (Min et al. 2002; Shin et al. 2013) have found that water chemistry changed from a Ca-HCO₃- to Na-Cl-NO₃-dominated system when NO₃ ⁻ sources changed from soil organic N to sewage and manure. Dun et al. (2014) calculated the contribution of Xiao River to groundwater combining δ¹⁵N with Cl⁻.

Combining nitrogen and oxygen isotopes of nitrate with boron isotope

Boron (B) has two natural stable isotopes, ¹¹B and ¹⁰B, whose relative abundances are 80 and 20 ‰, respectively. B is widely used in fertilizer, manure, and household products (Kendall et al. 2007) and is difficult to remove by sewage treatment (Barth 2000). B isotopes never change with physical, chemical, and microbiological processes in water.

δ¹¹B values vary between NO₃ ⁻ sources. Xue et al. (2009), for example, found by summarizing existing data that δ¹¹B values in manures, fertilizers, and sewage were in the range of +6.9 to 42.1 ‰, +8 to 17 ‰, and −7.7 to 12.9 ‰, respectively.

Combining δ¹¹B with δ¹⁵N-NO₃ ⁻ can, in some cases, provide a clear understanding of NO₃ ⁻ sources. Widory et al. (2005) used a combination of δ¹¹B and δ¹⁵N-NO₃ ⁻ data to conclude that NO₃ ⁻ in groundwater in the Pyrenees region of France was derived from wastewater. Xue et al. (2013) identified the NO⁻ ₃ sources in five polluted areas of the Flanders watershed in Belgium. The results showed that relatively high ¹¹B values, which were consistent with those found in agricultural manure, were associated with groundwater beneath agricultural areas and in groundwater below areas of agriculture and horticulture. In contrast, lower δ¹¹B values that were within the range typically associated with fertilizers were found in groundwater associated with households.

As a conservative substance, B can also provide useful information for distinguishing the mixing of multiple sources from denitrification. Widory et al. (2013) found by analyzing δ¹⁵N-NO₃ ⁻, δ¹⁸O-NO₃ ⁻, and δ¹¹B values that NO₃ ⁻ in the Alsace region of the Rhine basin was the result of mixing of multiple sources accompanied with denitrification. More specifically, they found that ¹⁵N-NO₃ ⁻ and ¹⁸O-NO₃ ⁻ values increased by following a 2:1 ratio, indicating that denitrification was occurring whereas ¹¹B values varied between values typically associated with different source materials (e.g., fertilizer in the range of 2.4 ‰ and pig manure in the range 44 ‰).

Combining nitrogen and oxygen isotopes of nitrate with sulfur isotope

Sulfur (S) has four natural stable isotopes, ³²S, ³³S, ³⁴S, and ³⁶S, of which ³²S and ³⁴S are widely studied, whose natural abundances are 95.02 and 4.21 %, respectively. The general terrestrial range of stable S isotope values (δ³⁴S) is −50 to +50 ‰ (Kendall et al. 2007). It is useful to identify natural and anthropogenic sources by using ³⁴S-SO₄ ²⁻ (Pauwels et al. 2010; Urresti-Estala et al. 2015), since δ³⁴S-SO₄ ²⁻ values in natural and anthropogenic sources are distinct. δ³⁴S-SO₄ ²⁻ values in anthropogenic sources are in the range of −2 to +11 ‰, while those values in natural sources are in the range of +13 to +16 ‰ (Urresti-Estala et al. 2015). As to anthropogenic sources, δ³⁴S-SO₄ ²⁻ values in various sources are different. Sewage is characterized by a δ³⁴S-SO₄ ²⁻ value of 9.6 ‰ (Otero et al. 2008). Chemical fertilizers have lower δ³⁴S-SO₄ ²⁻ values in the range of −1 to 8.7 ‰, whereas those values of chemical detergents are higher with average value of 12.1 ‰ (Vitòria et al. 2004). For natural sources, mineralized soil organic sulfate has low δ³⁴S-SO₄ ²⁻ values in the range of −1.5 to +8 ‰ (Vitòria et al. 2004). The δ³⁴S-SO₄ ²⁻ values of atmospheric deposition are in the intermediate field between sulfate from anthropogenic sources (fertilizers and detergents) and that of geological origin (Krouse and Mayer 2000; δ³⁴S-SO₄ ²⁻ = −3–9 ‰).

Due to the difference of sulfur isotopic compositions in natural and anthropogenic sources, a combination of sulfur isotope and nitrogen and oxygen isotopes of nitrate is successfully used to identify contamination sources in surface water and groundwater. Kaown et al. (2011) found that in rural areas of Yupori, Korea, a mixture of chemical fertilizers and manure is the main contamination sources in groundwater by analyzing stable isotopes of nitrate and sulfate in soils and shallow groundwater. In Taipei City of Taiwan, Hosono et al. (2011) used N and O isotopes of nitrate and sulfate to identify these ions sources and found that municipal sewage leaking from sewer pipes contribute in urban area, while chemical fertilizers contribute in the suburb area. In coastal regions, the combined analysis of stable isotope of nitrate and sulfate can not only be used to determine nitrate and sulfate sources, but also be used to distinguish these source in seawater from freshwater (Hosono et al. 2011; Saccon et al. 2013), which is not be understood by only analyzing stable isotopes of nitrate.

Besides identification of contamination sources, the nitrogen and oxygen isotopes of nitrate combined with ³⁴S-SO₄ ²⁻ can also provide a clear understanding of nitrogen transformation processes (Moncaster et al. 2000). Pyrite is an abundant sulfide mineral in natural aquifers which represents one of the potential electron sources (Hosono et al. 2014; Schwientek et al. 2008) for denitrification process. Pyrites existed in aquifer are sensitive to redox conditions and promote autotrophic denitrification processes in groundwater. During autotrophic denitrification process, the SO₄ ²⁻ produced through sulfide oxidation has a δ³⁴S_SO4 close to the δ³⁴S_FeS2 (Pauwels et al. 2010), and SO₄ ²⁻ concentration, δ¹⁵N-NO₃ ⁻, and δ¹⁸O-NO₃ ⁻ values increase while NO₃ ⁻ concentration and δ³⁴S_SO4 values decrease (Vitòria et al. 2008; Pauwels et al. 2010). Pauwels et al. (2010) pointed out that it was only from the association of both δ³⁴S-SO₄ ²⁻ and δ³⁴N-NO₃ ⁻ values that the occurrence of denitrification had become clear. δ¹⁸O-NO₃ ⁻ associated with δ³⁴S-SO₄ ²⁻ is also used by Vitòria et al. (2008) to prove the relationship between sulfide oxidation and denitrification in the Osona region of NE Spain. A number of field studies (Vitòria et al. 2008; Otero et al. 2009; Pauwels et al. 2010; Hosono et al. 2014) have demonstrated the occurrence of autotrophic denitrification coupled to oxidation of pyrite in aquifers based on isotope method by a combined analysis of sulfur and nitrogen and oxygen isotopes of nitrate. Hosono et al. (2014) studied a type of denitrification and occurrence of sequential anaerobic processes along the flow systems with N and S isotopes in Midori River watershed in the west of Japan and developed a model of N and S isotopic evolutional patterns, which is useful in understanding denitrification processes.

Overall, Cl⁻, ¹¹B, and ³⁴S can be used to address the shortcomings inherent in the application of nitrogen and oxygen isotopes of nitrate alone. Combining Cl⁻, ¹¹B, or ³⁴S with δ¹⁵N and δ¹⁸O will provide more information on potential NO⁻ ₃ sources and represents a fresh approach to distinguish between the changes in δ¹⁵N values caused by the mixing of N from different sources and denitrification. While these iron and isotopes are useful tools for studies on nitrogen dynamics, there is still some weakness of this method. For example, Cl⁻ may be more useful in those areas in which groundwater anion is mainly Cl⁻. As for boron isotope, various nitrogen sources have large ranges of δ¹¹B values (Pennisi et al. 2013) and the overlaps in isotopic values among these nitrate sources may increase the uncertainty for nitrate source identification; thus, a well-characterized δ¹¹B values for each nitrate source are required in a special area. As for sulfur isotope, some sulfates in aquifers do not derived from denitrification but from pyrite oxidation by newly formed ferrous ion (Fe³⁺) (Tarits et al. 2006) which result in a little unclear about understanding the occurrence of denitrification process.

Conclusions

The stable isotope approach can be used to identify both NO₃ ⁻ sources and N transformation processes without requiring the collection of a large amount of monitoring data. With regard to determining NO₃ ⁻ sources in water, as the NO₃ ⁻ moves into the receiving waters, the various δ¹⁵N-NO₃ ⁻ and δ¹⁸O-NO₃ ⁻ values of the natural and anthropogenic sources of nitrate lead to receiving waters having different isotopic composition of nitrate from original waters, and the isotope changes are attributed to the mixing of water from different NO₃ ⁻ sources, each characterized by a specific isotopic signature. By doing so, ¹⁵N-NO₃ ⁻ is effective at tracing NO₃ ⁻ sources from areas with different land use (Fig. 2). For example, δ¹⁵N-NO₃ ⁻ values are typically higher in agricultural and urban watersheds (where the NO₃ ⁻ is primarily derived from fertilizers, sewage, and manure) than those in forest watersheds, where the NO₃ ⁻ is from soils and atmospheric deposition. δ¹⁸O-NO₃ ⁻ values are more useful to identify NO₃ ⁻ sources from atmospheric NO₃ ⁻. Thus, the fingerprint based on N and O isotopes can be used to identify the predominant sources, and their relative contributions, in surface water and groundwater. Nitrogen and oxygen isotopes of nitrate can also be used to identify nitrogen transformation processes, such as nitrification, assimilation, and denitrification, since, during these nitrogen transformation processes, isotope fractionation leads to various changes in the isotopic composition of the residual N pool. However, isotopic fractionation associated with N transformation processes (described as in Fig. 2) is often temporally and spatially variable because they are largely driven by microbial activity, which reduces the accuracy of identifying NO₃ ⁻ sources and assessing N transformation processes. This problem may be reduced by combining the N and O isotopic data with other types of information including Cl⁻ concentrations and ¹¹B and ³⁵S isotopes, which widely exist in NO⁻ ₃ sources and are not influenced by physical, chemical, and microbiological process. Nevertheless, there is much room for improvement in the methods used to quantify NO₃ ⁻ sources, including the following.

Fig. 2

The potential nitrate sources and the main nitrogen transformation processes in soil (Heaton 1986), surface water, and groundwater

• 1. Stable isotope mixing models need to be further improved. The three models presented here can quantify the relative contributions of NO₃ ⁻ from delineated sources but have not been widely applied or tested. The reason may be that the accuracy of the models is low because variations in the isotope values of NO₃ ⁻ sources and fractionation factors have significant effect on results. The two Bayesian models are statistical in nature and are not based on analytical solutions but consider temporal and spatial variability in the isotopes and fractionation factors. The models might be improved by reasonably dividing target areas and by using specific isotopic compositions of measured for potential NO₃ ⁻ sources within these target areas instead of those values that have been reported by other studies.

• 2. The fractionation of N and O isotopes in various environmental surroundings requires further study. Stable N and O isotope methods can be used for qualitative analysis but are not suitable to quantify N transformation processes accurately. Isotope fractionation factors associated with N transformation vary in different environments as a result of many factors. Consequently, further laboratory and field experiments on isotope fractionation processes in different environments are clearly needed.