Nitrate behaviors and source apportionment in an aquatic system from a watershed with intensive agricultural activities

Abstract

Nitrate pollution in aquatic systems caused by intensive agricultural activities is a serious problem in the Sanjiang Plain. In this study, a dual isotope approach (δ¹⁵N–NO₃⁻ and δ¹⁸O–NO₃⁻) was employed to identify potential nitrate sources (atmospheric deposition, AD; NO₃⁻ derived from soil organic matter nitrification, NS; NO₃⁻ derived from chemical fertilizer nitrification, NF; and manure and sewage, M&S) and transformation processes occurring in the Abujiao River watershed located in the Sanjiang Plain. The Bayesian model (stable isotope analysis in R, SIAR) was utilized to apportion the contribution of the potential sources. In this watershed, the nitrate concentrations in the surface water were low (mean ± SD = 1.15 ± 0.84 mg L⁻¹), and were greatly influenced by precipitation and land use conditions during the two sampling periods (the high flow period, September; the low flow period, November). On the contrary, in the ground water, high NO₃⁻ concentrations were observed (7.84 ± 5.83 mg L⁻¹) and no significant temporal variation in NO₃⁻ was found during the sampling periods. The sampled water δ¹⁸O–NO₃⁻ values suggest that the nitrification process was not the main N cycling process, because most of the measured δ¹⁸O–NO₃⁻ values were above the expected δ¹⁸O–NO₃⁻ from nitrification throughout the sampling periods. Both the chemical and isotopic characteristics indicated that the signs of de-nitrification were absent in the surface water. However, significant de-nitrification processes were observed in the ground water for all sample periods. Results from the SIAR model showed that source contributions differed significantly during the two sampling periods. During the high flow period, chemical fertilizers and soil N fertilizer equally contributed to the major sources of nitrate in the surface water. In contrast, manure and sewage sources dominated the source contribution during the low flow period (November). This study suggested that with the assessment of the behaviors and sources of NO₃⁻, effective nitrate reduction strategies and better management practices can be implemented to protect water quality.

---

Environmental impact

Nitrate pollution in aquatic systems caused by intensive agricultural activities is becoming an increasingly serious problem in the Sanjiang Plain, north eastern China. However, to control the main contamination remains challenging, partly because information on the nitrate behaviors and sources in aquatic systems is still limited. In this study, the transformation and origin features of nitrate in the aquatic systems of the Abujiao River watershed, Sanjiang Plain, are discussed. Water mixing and de-nitrification are recognized as the two major processes which cause a decrease in the nitrate in surface water and groundwater in this watershed. Chemical fertilizers and manure and sewage dominate the nitrate source contribution in the surface water during the high and low flow periods, respectively.

1 Introduction

The Sanjiang Plain in northeastern China, which is one of the most important food bases in China, was once covered by wetlands and forests with precious black soil and abundant water resources. In order to address the food crisis, the Sanjiang Plain has experienced four large-scale agricultural development activities since the 1950s, and lots of wetland and forest areas have been exploited as crop land (which increased from 7870 km² to 47 330 km²).¹ Intensive agricultural activities have caused eco-hydrological changes, soil structure and function degradation, and a serious loss of black soil resources (decreasing from 60–100 cm to 16–72 cm), coupled with serious non-point nitrogen pollution in aquatic systems.²

The soil structure and function degradation and serious non-point nitrogen pollution in the Sanjiang Plain have had a significant influence on food security and the biogeochemical cycles in the ecosystem. Researchers have reported that the use of fertilizers on farmland is the major non-point source contributing to the nitrate pollution of aquatic ecosystems.³ The land-use and soil property variation were suggested as the most sensitive factors to control the non-point nitrogen pollution output.⁴ However, those studies were mainly based on the combined interpretation of chemical data with statistical techniques, GIS and mathematical models, and the limited information on the nitrate sources and nitrogen (N) cycling process often caused a significant deficit in calculating the balance between the nitrogen loading and controlling the nitrate contamination in this high latitude watershed.⁵ Moreover, although some research on the source and fate of nitrate had been conducted in the other agricultural areas of China,^6–8 the special black soil properties as well as seasonal freezing and thawing processes in the Sanjiang Plain inevitably resulted in a significantly different N cycling process.⁹ Meanwhile, due to agricultural activities, the original landscape and ecological structure have been changed, and single natural ecosystems were replaced by a hybrid ecosystem which was more complex for the identification of the sources and behaviors of nitrate.¹⁰ Therefore, the identification of how nitrate from various sources moves through the landscapes to water and which N cycling processes alter these sources during transport are critical in prioritizing effective nitrate reduction strategies.

The use of the isotope approach to identify potential pollutant sources is the current research trend and has been reported to be most diagnostic.¹¹ Numerous researchers have documented that the isotopic compositions of nitrate (N and O) act as a useful tool to identify nitrate sources and to track the processes of the N cycle in aquatic ecosystems.^12–14 δ¹⁵N–NO₃⁻ values have proven more useful for distinguishing NO₃⁻ derived from ammonium fertilizer, soil organic matter, and animal manure in aquatic ecosystems,¹⁵ whereas δ¹⁸O–NO₃⁻ values have proven more useful for identifying NO₃⁻ derived from nitrate-bearing fertilizer and atmospheric deposition of NO₃⁻.^15,16 The N cycling processes often involve coupled fractionation; for example, de-nitrification often results in a net change rate of 1 : 2 in δ¹⁸O–NO₃⁻ : δ¹⁵N–NO₃⁻ of the residual pool in water, suggesting that N fractionation is twice the amount of oxygen fractionation in many environments.^12,17 The model of stable isotope analysis in R (SIAR) has been used successfully to predict the quantities of the relative contributions of different nitrate (NO₃⁻) sources with uncertainties in the isotope fractionation.^8,18,19 Additionally, hydro-chemical characteristics can provide some conventional information to help distinguish nitrate sources and N cycling processes.²⁰ Thus, a combination of nitrate isotopes, SIAR, and hydro-chemical information may provide more detailed information regarding the sources and potential N cycling processes in intensive agricultural activities that impact on watersheds.

In this study, we researched the nitrate sources and N cycles in aquatic systems from a watershed with intensive agricultural activities and mixed land use types in the Sanjiang Plain. By analyzing hydro-chemical characteristics (Cl⁻, NO₃⁻, NH₄⁺, and DO) and dual nitrate isotopic signatures, the potential nitrate sources and N cycling processes were examined systematically during the different sampling periods. The SIAR model was used to quantitatively apportion the contribution of potential sources. The objectives of this study were to (1) determine the factors which best explain the variability and isotope composition of nitrate in aquatic ecosystems, (2) identify the main N cycling processes in different water bodies and (3) apply the source apportionment model SIAR to estimate the proportion of each source of NO₃⁻ enrichment.

2 Materials and methods

2.1. Description of the study area

Our study focused on an intensive agriculture-impacted watershed (the Abujiao River catchment, latitude N 47°22′11′′–47°28′14′′, longitude E 134°0′45′′–134°20′47′′) that has mixed land use types and is a typical representative of the Sanjiang Plain of Northeast China (Fig. 1). The Abujiao River flows eastward and finally flows into the Wusuli River, with an elevation decline from 129 m to 38 m. This region is typically characterized as having temperate and continental climates, with a cold and long winter and a warm summer. In early spring and late winter, the temperature is low, with a lowest monthly average temperature of approximately −19.27 °C. Summers are short and muggy and the annual average temperature is 2.94 °C. After October, temperatures fall down to below 0 °C, with no rainfall occurring, and the ground is gradually covered by snow. The air temperature often rises above 0 °C in April, and the frozen topsoil begins to thaw from the surface downwards in late April.⁴ These freeze–thaw cycles have a strong influence on soil erosion in April. The average annual rainfall is 583.18 mm, most of which falls in summer (from July to September). According to the Genetic Soil Classification of China (GSCC), the primary soil types in this region include dark brown soil, bleached Baijiang soil, chernozem soil, meadow soil, and black soil.

Fig. 1 Study area and sampling sites (G, groundwater; S, surface water; WW, wastewater).

The Abujiao River Basin is a typical intensive agricultural activity-impacted watershed and its catchments cover 142.5 km², with 56.12% for agriculture, 11.78% for wetland, 3.97% for construction land, 25.41% for forest, and 2.72% for other land use types (Fig. 1). Rice paddy (Oryza satiua L.) is the dominant crop (accounting for 57% of the total cropping area) in the river basin, which is planted in early May and harvested in October. The extensive application of fertilizer takes place from the end of April to July. Fertilizer N is almost all in the reduced form, i.e., urea (36.67%) and ammonium (63.33%), NH₄HCO₃ and NH₄Cl. Nitrogen from manure (human and animal wastes) and sewage is directly poured into the river system, which makes an important contribution to the N input into the river basin.

2.2. Samples and analysis

According to the information on temperature and rainfall in the study area (Fig. 2), we found a significant difference between September and November. Considering the convenience and safety of sampling campaigns, we selected those months as the study periods. In order to address the seasonal variability of the distribution and isotope composition of nitrate and the underlying causes, we investigated nitrate from surface water (two sampling campaigns: 18–20/9/2012; 7–9/11/2012) as well as ground water (three sampling campaigns: 19/9/2012; 28/9/2012; 10/11/2012) during the high flow period in September 2012 and the low flow period in November 2012 (Fig. 2). Water samples were collected from upstream to downstream of the Abujiao River along the main stream, comprising 20 surface water samples and 13 groundwater samples (all samples from wells for drinking or irrigation) (Fig. 1). Detailed information on the ground water samples can be found in Table 1.

Fig. 2 Daily average temperature and rainfall distribution at the Abujiao watershed in 2012.

Table 1 Detailed information on the ground water samples

Sample	Well depth (m)	Well purpose	Land use	Sampling times
G01	25	Drinking	Urban	Three times (19/9/2012; 28/9/2012; 10/11/2012)
G02	78	Drinking	Urban	Three times (19/9/2012; 28/9/2012; 10/11/2012)
G03	32	Irrigation	Corn	Two times (19/9/2012; 10/11/2012)
G04	106	Irrigation	Corn	One time (19/9/2012)
G05	30	Irrigation	Soybean	Two times (19/9/2012; 10/11/2012)
G06	27	Irrigation	Rice	Two times (19/9/2012; 10/11/2012)

---

Wastewater, rainwater, soil and synthetic fertilizers were also sampled because they represented the primary sources of NO₃⁻ in the area. Untreated waste water samples (n = 6) were collected at the two urban sewage outfalls (Fig. 1) and three sampling campaigns were conducted (the sampling times were the same as for the groundwater sampling campaigns). All rain water samples (n = 4) were collected at an agricultural experiment station located in the studied watershed in September 2012 (from 9/9/2012 to 30/9/2012). Top soil samples (0–10 cm, n = 12) of different land use types (including 3 forest soils, 3 wetland soils, 2 unfertilized soybean soils, 2 unfertilized corn soils and 2 unfertilized rice soils) were obtained in the studied watershed in September and November 2012. Samples of urea and ammonium (n = 6), which were used widely in this watershed, were collected at the local market. Soil and fertilizer samples were preserved in polyvinyl chloride bags and frozen (−20 °C) until required for further analyses.

At each water sample site, pH, temperature (T) and dissolved oxygen (DO) were measured using portable equipment (YSI, USA). Then, three replicates of surface water samples at 0–20 cm depth at the center of the stream were collected using Teflon containers. All sampling equipment was pre-cleaned with non-phosphate detergent and rinsed thoroughly with deionized water. The collected water samples were filtered in situ through 0.45 μm cellulose-acetate filter paper for chemical and isotope analysis.

In the laboratory, anion concentrations in the water samples, including NH₄⁺, NO₃⁻, and Cl⁻, were determined using ion chromatography (Dionex, DX-600 and ICS-2100, USA), with a precision of 5%. Dissolved nitrate was converted to solid AgNO₃ according to a reported approach.²¹ AgNO₃ was then analyzed for δ¹⁵N using a DELTA plus XL mass spectrometer connected to a CE EA1112 C/N/Sanalyzer, and for δ¹⁸O using a MAT 253 mass spectrometer connected to a high-temperature conversion elemental analyzer. International standard IAEA-N3 was used for δ¹⁵N and δ¹⁸O calibration. The reproducibility of duplicate analyses was <0.9‰ for δ¹⁸O (averaging ±0.3‰), and <0.3‰ for δ¹⁵N (averaging ±0.2‰).

In the laboratory, plant debris and large stones were picked out from each soil sample, which was then ground, sieved and screened using 100 mesh before being freeze-dried again for use. A total of 0.5 g of soil samples was taken. Inorganic carbonates were removed using 0.5 mol L⁻¹ HCl for acidification, and samples were then rinsed with deionized water until the filtrate was neutral. Silver nitrate solution was used to detect the presence or absence of Cl⁻ residues, and then samples were freeze-dried and inorganic carbon was removed. All soil sample and fertilizer sample analysis was conducted in the Test Center of the Chinese Academy of Forestry, Key Laboratory of Stable Isotope, using a mass spectrometer (MAT253). The analytical precision of δ¹⁵N is ±0.2‰.

2.3. Calculations and statistical analysis

(1) Expected δ¹⁸O_nitrate from nitrification. When nitrification occurs the expected δ¹⁸O nitrate can be calculated using the following equation:²²

Expected δ¹⁸O_nitrate = 2/3δ¹⁸O_water + 1/3δ¹⁸O_O2 (1)

where expected δ¹⁸O_nitrate is the expected δ¹⁸O value of nitrate when nitrification occurs, δ¹⁸O_water is the δ¹⁸O value in the water and δ¹⁸O_O2 is the δ¹⁸O value in the air. It should be noted that this expected local nitrate δ¹⁸O range was rather simplified. There would be several uncertainties in the results of the theoretical calculation of nitrification, and those uncertainties might derive from isotope fractionation and the ratio of oxygen incorporation from H₂O and O₂ not being 2 : 1 in the different environments.^16,18,23 However, previous studies have proved that under the limited data conditions, this theoretical estimation might offer a representative indicator for the local δ¹⁸O of nitrate oxidized from reduced N.^7,13,24

(2) Rayleigh equation for the de-nitrification process. The changes in δ¹⁵N–NO₃⁻ and δ¹⁸O–NO₃⁻ values during the de-nitrification process can be expressed as a simplified Rayleigh equation:¹⁵

δ_R = δ_R0 + ε ln(f) (2)

where δ_R is the δ¹⁵N or δ¹⁸O value of the reactant nitrate at time t, δ_R0 is the initial δ¹⁵N or δ¹⁸O value of the nitrate, f is the remaining fraction of the nitrate, which equal to the remaining substrate concentration divided by the initial substrate concentration, and ε is the enrichment factor.

(3) Source apportioning mixing model (SIAR). The contributions of nitrate sources to a mixture in the aquatic ecosystems were quantified using the SIAR model. The system model used in SIAR can be expressed as follows:

S_jk ∼ N(μ_jk,ω²_jk) (4)

C_jk ∼ N(λ_jk,τ²_jk) (5)

ε_ij ∼ N(0,σ²_j) (6)

where X_ij is the isotope value j of the mixture i, in which i = 1, 2, 3, …, i and j = 1, 2,3, …, j; s_jk is the source value k of isotope j (k = 1, 2, 3, …, k) and is normally distributed with mean μ_jk and standard deviation ω_jk; p_k is the proportion of source k, which needs to be estimated using the SIAR model; c_jk is the fractionation factor for isotope j of source k and is normally distributed with mean λ_jk and standard deviation τ_jk; and ε_ij is the residual error representing the additional unquantified variation between individual mixtures and is normally distributed with mean 0 and standard deviation σ_j. A detailed description of this model can be found in literature.^25–27

(4) Statistical analysis. All statistical analyses were carried out using software SPSS version 20. After testing for normality (Kolmogorov–Smirnoff test) and homogeneity of variances (Levene test), two-way ANOVA tests were performed to assess the effect of time (season) and space (location) on the isotopic ratios and chemical concentrations. When assumptions of homogeneity of variances or normality were not reached, non-parametric Kruskall–Wallis ANOVAs were performed followed by post hoc non-parametric paired comparison tests.

3 Results and discussion

3.1. The variation of dissolved nitrogen (DIN) and isotopic composition in surface water

The physicochemical properties and isotopic composition of nitrate of all surface water samples are depicted in Fig. 4. There were considerable differences in the N species and isotopic compositions in the surface water among various land use types. At urban sampling sites, the highest NH₄⁺, Cl⁻, δ¹⁵N–NO₃⁻ and δ¹⁸O–NO₃⁻ levels in the surface water samples were observed (Fig. 4). These relatively high values appeared similar to the values observed for pollution of sewage origin, which indicates that untreated wastewater inputs were the principal factor impacting these sites. However, the isotopic composition of nitrate did not show a significant difference among the other sites (i.e., forest and wetland) and the lower δ¹⁵N–NO₃⁻ values of nitrate obtained at those sites were intermediate between the values measured for soil and NF (Fig. 6). For agricultural sampling sites, the highest value of NO₃⁻ (Fig. 4C) was observed during the high flow period. Meanwhile, the Cl⁻ decreased dramatically from S04 (urban sampling sites) to S05 (agricultural sampling sites) (Fig. 4A). These results suggested that the loading of nitrate might be associated with agricultural activities. The spatial variation of nitrate contents demonstrated that there was a significant nutrient contribution from human activities and land use was a major influencing factor for nitrate concentration and isotopic composition in this watershed.

Fig. 3 Records of daily precipitation and temperature at the Abujiao River watershed for the different sampling periods.

Fig. 4 Varying trends of (A) Cl⁻, (B) DO, (C) NO₃⁻ (D) NH₄⁺, (E) δ¹⁸O–NO₃⁻ and (F) δ¹⁵N–NO₃⁻ along the Abujiao river during the two sampling periods.

A significant temporal variation in the isotopic composition of the surface water was also observed (Fig. 4). Compared with the high flow period, the samples of surface water were characterized by relatively high δ¹⁵N–NO₃⁻ values (increasing from 5.59 ± 3.91‰ to 8.65 ± 2.99‰) and low δ¹⁸O–NO₃⁻ values (decreasing from 9.83 ± 1.69‰ to 7.00 ± 1.56‰) during the low flow period. This finding could be attributed to the significant variation in precipitation between the samples collected during the two periods. During the two sampling periods, precipitation and temperature conditions were significantly different (Fig. 3). The high volumes of precipitation with low δ¹⁵N–NO₃⁻ and high δ¹⁸O–NO₃⁻ values in September might decrease the δ¹⁵N–NO₃⁻ and increase the δ¹⁸O–NO₃⁻ values in the river.²⁸ Meanwhile, at the agricultural sites, the NO₃⁻ level changed dramatically between the two sampling periods. During the high flow period, the highest level of NO₃⁻ with a low isotopic composition was observed at sampling site S05 which was located in the agricultural area. These unusually high nitrate concentrations reflected that the NO₃⁻ might be mainly derived from chemical fertilizers and soil N, which was controlled by precipitation.²⁹ In contrast, during the low flow period, the effects of precipitation were reduced, because there was little supplemental water (or material) from crop land. Simultaneously, sewage waste from up stream (the urban sites) had a serious impact on the sites in the agricultural area, which resulted in the NH₄⁺ dominating the DIN with relatively high Cl⁻ and δ¹⁵N–NO₃⁻. These results indicate that precipitation may be another important factor that affects DIN and nitrate isotopic composition in surface water, especially in the rainy season.

3.2. The variation of DIN and isotopic composition in ground water

Compared with the surface water samples, in the ground water samples there were significant differences in the physicochemical properties and nitrate stable isotope ratios (Table 2). The NO₃⁻ in the ground water (7.84 ± 5.83 mg L⁻¹, mean ± SD) was the dominant species (>95%) of the DIN, and the concentration was much higher than that in surface water (1.15 ± 0.84 mg L⁻¹). Meanwhile, the δ¹⁵N–NO₃⁻ and δ¹⁸O–NO₃⁻ values in the ground water (18.95 ± 6.84‰ and 15.90 ± 5.62‰, respectively) were higher than those in the surface water (7.12 ± 3.74‰ and 8.42 ± 2.15‰, respectively). In contrast, NH₄⁺ (0.13 ± 0.18 mg L⁻¹) and DO (0.7 ± 0.2 mg L⁻¹) were observed in ground water at much lower concentrations than those in surface water (NH₄⁺ was 1.67 ± 1.53 mg L⁻¹ and DO was 5.3 ± 1.3 mg L⁻¹).

Table 2 Mean values (±standard deviation) of δ¹⁸O and δ¹⁵N in nitrate and physicochemical properties of surface water (n = 20) and groundwater (n = 13), and results of one-way ANOVA

Samples		Variable
		DO mg L^−1	Cl^− mg L^−1	NH4^+ mg L^−1	NO3^− mg L^−1	δ^15N–NO3^− ‰	δ^18O–NO3^− ‰
SW	(Mean ± SD)	5.3 ± 1.3	15.48 ± 10.29	1.67 ± 1.53	1.15 ± 0.84	7.12 ± 3.74	8.42 ± 2.15
GW	(Mean ± SD)	0.7 ± 0.2	23.21 ± 20.62	0.13 ± 0.18	7.84 ± 5.83	18.95 ± 6.84	15.90 ± 5.62
Statistics	F	132.45	2.00	12.26	21.14	38.41	30.08
	Sig.	<0.001	P = 0.168 > 0.05	0.05 < P = 0.02 < 0.01	<0.001	<0.001	<0.001

---

Similar to the surface water samples, large spatial variations were in evidence for nitrate and nitrate stable isotope ratios in the ground water (Table 3). Compared to the agricultural sampling sites, higher Cl⁻, NO₃⁻, δ¹⁵N–NO₃⁻ and δ¹⁸O–NO₃⁻ were observed at the urban sampling sites. The relatively high Cl⁻, δ¹⁵N–NO₃⁻ and δ¹⁸O–NO₃⁻ in those ground water samples suggests a higher proportion of isotopically enriched nitrate sources (i.e., manure waste sources). We also noticed that the highest δ¹⁵N–NO₃⁻ and δ¹⁸O–NO₃⁻ values were observed at sampling site G01 (26.27‰ and 22.28‰ for δ¹⁵N–NO₃⁻ and δ¹⁸O–NO₃⁻, respectively) which are much higher than those in manure and sewage (15.32 ± 0.83‰ for δ¹⁵N–NO₃⁻ and 9.11 ± 0.93‰ for δ¹⁸O–NO₃⁻). These results indicate that a nitrogen cycle process had happened and this process could lead to the δ¹⁵N–NO₃⁻ and δ¹⁸O–NO₃⁻ values increasing. The de-nitrification process, which was confirmed by the low DO concentrations and the linear relationship between δ¹⁵N–NO₃⁻ and δ¹⁸O–NO₃⁻ in the ground water (see Section 3.3), might explain these extraordinarily high nitrate stable isotope ratios.

Table 3 Sampling data and results of the ground water samples^a

Sample site	Date	Cl^−	NH4^+	NO3^−	DO	δ^15N–NO3^−	δ^18O–NO3^−
		mg L^−1	mg L^−1	mg L^−1	mg L^−1	‰	‰
a n.a. means below the detection limit.
G01	19/9/2012	52.48	n.a.	11.05	0.6	22.36	18.48
	28/9/2012	47.24	n.a.	9.94	1.0	24.12	20.86
	10/11/2012	50.86	n.a.	7.43	0.8	26.27	22.28
G02	19/9/2012	37.08	n.a.	17.37	1.1	24.16	18.21
	28/9/2012	37.11	n.a.	16.85	0.4	25.77	20.27
	10/11/2012	39.36	n.a.	15.98	0.8	26.47	21.88
G03	19/9/2012	9.75	n.a.	3.23	0.6	17.00	14.02
	10/11/2012	6.37	n.a.	2.98	0.4	18.12	14.4
G04	19/9/2012	5.37	0.37	1.49	1.0	18.83	15.04
G05	19/9/2012	3.62	n.a.	4.13	0.6	7.05	9.07
	10/11/2012	3.37	n.a.	6.08	1.1	6.21	2.02
G06	19/9/2012	4.60	0.46	1.02	0.7	14.65	14.95
	10/11/2012	4.50	0.68	0.97	0.7	15.34	15.22

---

In contrast with the surface water samples, the effects of precipitation were not observed in the ground water, as can be seen from the following. Firstly, there were no observable variations (P > 0.05, two-way ANOVA tests) in the nitrate and nitrate stable isotope ratios in the ground water during the two sampling periods. If the ground water samples were affected by precipitation, relatively high δ¹⁸O–NO₃⁻ and low δ¹⁵N–NO₃⁻ would be observed. Furthermore, only a slight fluctuation of Cl⁻ was observed between the different sampling periods. Due to the dilution effect of the precipitation, Cl⁻ in the ground water would be changed dramatically under its influence. It should be noted that the de-nitrification process was also observed in those samples, and the isotopic fractionation effects might conceal the effects of precipitation.

3.3. Transport and transformation of nitrate

(1) The variation in chloride concentration. Chloride can be an ideal indicator of sewage and manure input and dilution because the chemical behavior of chloride in natural water is conservative and its concentration can change only by mixing within the river system. The ratio of NO₃⁻/Cl⁻ can provide more information to help distinguish the effect of dilution from de-nitrification on N removal processes.^13,30 In this study, the concentrations of Cl⁻ in the surface water ranged from 5.54 to 38.74 mg L⁻¹ (15.48 ± 10.29 mg L⁻¹) (Fig. 4) and varied greatly between sites. A sharp decline in NO₃⁻ was observed in the river between the forested sampling site (S02) and urban sampling site (S03). A reasonable explanation is that the surface water from the forested sampling site (S02) mixed with wastewater that came from urban sewage outfall presenting low NO₃⁻ and high Cl⁻. During the high flow period, the sudden decline in Cl⁻ and increase in NO₃⁻ concentration were observed at site S05. Meanwhile, positive correlations (R² = 0.765, Fig. 5A) between Cl⁻ and NO₃⁻ levels were also observed downstream (from site S05 to S10). On the contrary, a weak negative correlation (R² = 0.331) was found for all sampling sites during the low flow period. These results suggest that the dilution process (the surface water mixing with runoff caused by precipitation and drainage from farmland) were dominant and other processes such as de-nitrification occurring in rivers contributed little to the variation of NO₃⁻ in the surface water during the high flow period. However, this dilution process decreased during the low flow period due to the limited water supplement.

Fig. 5 (A) Relationship between Cl⁻ and NO₃⁻ concentrations, and (B) relationship between Cl⁻ μmolar concentration and NO₃⁻/Cl⁻ molar ratio in the Abujiao River.

The Cl⁻ sources in water from various areas generally include natural sources (dissolution of minerals), agricultural chemicals (i.e., KCl), animal waste, septic effluent, and road salt. Chemical fertilizers generally have high nitrogen contents with low chloride contents. Sewage effluents had high Cl⁻ contents and low ratios of NO₃⁻/Cl⁻. Elevated nitrate contents were observed in water samples with agricultural inputs. The values of NO₃⁻/Cl⁻ will decrease if de-nitrification removes nitrate or if plants take up the nutrients.³¹ Thus, the ratio of NO₃⁻/Cl⁻ and the Cl⁻ concentration can provide additional information to help distinguish the different input sources.

In this study, we did not collect the runoff and drainage from farmland directly. However, we noticed that during the high flow period, extremely high NO₃⁻ levels were found at agricultural sampling sites. Meanwhile, the Cl⁻ decreased from sampling sites S05 to S07 (in the agricultural areas) and gradually stabilized at 5–7 mg L⁻¹. Therefore, the min and max of NO₃⁻ at those sites and the stabilized Cl⁻ were used as the end values of agricultural input. Some untreated wastewater samples were collected at the two sewage outfalls. These data were invoked as the end values of manure and sewage inputs. Meanwhile, some precipitation samples were also collected at this watershed, which were regarded as giving the end values of rainfall. However, it was difficult to define the accurate range of the end value of the de-nitrification process. Because the de-nitrification process could remove nitrate, which causes the values of NO₃⁻/Cl⁻ to decrease, the end value of the de-nitrification effect might exist under the end values of agricultural input and manure and sewage input.

The plot of NO₃⁻/Cl⁻ molar ratios versus Cl⁻ concentration is illustrated in Fig. 5B. The results in Fig. 5B show the extreme variation in nitrate sources during the two sampling periods. During the high flow period, agricultural inputs heavily influenced the distribution of nitrate in the river. In contrast, during the low flow period, manure and sewage inputs and the de-nitrification process may dominate the nitrate distribution. These results also suggest that anthropogenic activities were a leading contributor to nitrate pollution in the whole watershed.

(2) Nitrification process. Nitrification is an important biologically mediated reaction that causes the N and O isotopes to fractionate and leave heavier isotopes in the residual species. Fractionation factors for δ¹⁵N and δ¹⁸O during nitrification vary depending on the local conditions and rates of reaction. Laboratory experiments showed that the NO₃⁻ produced from the nitrification process generally gains one-third of its O from atmospheric O and two-thirds from the surrounding water.²² Based on this rule, the δ¹⁸O–NO₃⁻ derived from the oxidation of reduced N forms could be calculated using eqn (1). The δ¹⁸O–O₂ value was +23.5‰ and the δ¹⁸O–H₂O values in the water ranged from −17.81‰ to +2.52‰, which were calculated using the δ¹⁸O value in precipitation water from a nearby International Atomic Energy Association (IAEA) station in Haerbin.³² The calculation results show that the δ¹⁸O–NO₃⁻ derived from nitrification of reduced N forms ranged from −4.04‰ to + 6.15‰. This local theoretical range for nitrate δ¹⁸O is shown in Fig. 4 and 6 (the horizontal dashed lines). In this study, we did not use this model to discuss the nitrification process in ground water. The reasons were as follows. Firstly, NH₄⁺ levels (most samples were under the detection limit) in all samples were much lower than those of NO₃⁻ in the ground water during all sampling periods. Furthermore, this simple model may exhibit defects as mentioned in Section 2.3. Additionally, during the nitrification process, there was a different apportionment pattern of O between O₂ and H₂O between the surface water and the ground water. Therefore, using this equation directly would lead to great uncertainty in our results. Finally, the fractionation caused by the de-nitrification process in ground water would conceal the signs of the nitrification process.

At forest and wetland sites, the values of δ¹⁸O–NO₃⁻ in the surface water were within the range of calculated values during all sampling periods. This result indicates that microbial nitrification might play a significant role at those sites. At urban sites, untreated wastewater containing high levels of NH₄⁺ was poured into the river directly. NH₄⁺ decreased down the river (from sampling site S05 to S07) with NO₃⁻ increasing and δ¹⁵N–NO₃⁻ decreasing. To our knowledge, two processes might explain this phenomenon in the surface water: one is the nitrification process, and the other is mixing with other sources (often containing higher NO₃⁻, and lower NH₄⁺ and δ¹⁵N–NO₃⁻). The chemical behavior of chloride in natural water is conservative and its concentration can change only by mixing within the river system. Therefore, we combined the information of water chemistry (such as Cl⁻) and tried to explore the reduction mechanism of NH₄⁺ at those sampling sites.

Supposing that there was only a nitrification process in the river, the Cl⁻ concentration should not vary when the NH₄⁺ changes. However, our data at those sites did not support this hypothesis. As can be seen in (Fig. 4A and D), the trend and extent of Cl⁻ were consistent with those of NH₄⁺. This indicates that some nitrate sources might flow into this river. It should be noted that those sites were located in agricultural areas, and the runoff and drainage from farmland (compared with untreated wastewater, this water often contained lower NH₄⁺, Cl⁻, δ¹⁸O–NO₃⁻ and δ¹⁵N–NO₃⁻) might flow into the surface water through some small ditches. Moreover, a similar decrease in δ¹⁸O–NO₃⁻ was found at those sites, which also suggests that the surface water might mix with runoff and drainage from farmland. Finally, nor did the values of δ¹⁸O–NO₃⁻ support the above hypothesis, because the δ¹⁸O–NO₃⁻ values at the other sites were all above the typical upper limit of δ¹⁸O–NO₃⁻ from nitrification of reduced N (Fig. 4E). Three processes (mixing with precipitation, addition of synthetic NO₃⁻ fertilizer, and de-nitrification) might produce the heavier δ¹⁸O–NO₃⁻ in water. The contribution of synthetic NO₃⁻ could be readily eliminated, because few of the synthetic N fertilizers were applied in this watershed. Precipitation might be one of the most important factors contributing to the relatively high values of δ¹⁸O–NO₃⁻ in the surface water, especially for the high flow period. Thus, these results suggest that nitrification might not play an important role in reducing NH₄⁺ levels at those sites compared with other processes (such as dilution effects and precipitation).

(3) De-nitrification process. De-nitrification is an important mechanism for reduction of the NO₃⁻ load in aquatic environments via the transformation from dissolved NO₃⁻ to N₂ and N₂O. In contrast with nitrification, de-nitrification usually causes isotopic enrichment of the residual NO₃⁻. The extent of fractionation during de-nitrification is dependent on the fraction of the substrate pool that is consumed. During the de-nitrification process, the increase in δ¹⁵N–NO₃⁻ values appear to be between 1.5 and 2.0 times that of the δ¹⁸O–NO₃ values, and this relationship could be used to indicate the occurrence of nitrification.¹²Fig. 6 shows the isotopic compositions of nitrate in the study area. In this study, no positive relationship was found between δ¹⁵N–NO₃⁻ and δ¹⁸O–NO₃⁻ in the surface water. In contrast, positive relationships (δ¹⁸O–NO₃⁻ = 0.78 × δ¹⁵N–NO₃⁻ +1.15, R² = 0.89 and δ¹⁸O–NO₃⁻ = 0.88 × δ¹⁵N–NO₃⁻ − 4.42, R² = 0.78) were observed in the ground water and wastewater, respectively. Additionally, the DO concentrations measured at the sampling points (0.8 ± 0.2 mg L⁻¹ in ground water, 1.1 ± 0.3 in waste water) were ideal for de-nitrification (0.1 mg L⁻¹ to 2 mg L⁻¹).³³ These results indicate that the de-nitrification process dominated the N cycling process in these aquatic ecosystems.

Fig. 6 Ranges of δ¹⁸O and δ¹⁵N values for potential nitrate sources and values measured in Abujiao River.

To better understand the de-nitrification process in ground water, cross plots of δ¹⁵N–NO₃⁻ and δ¹⁸O–NO₃⁻versus ln[NO₃⁻] were used to evaluate the de-nitrification process, which could be described by eqn (2). Plotting nitrogen isotope ratios versus ln[nitrate concentration] could offer some useful information on the de-nitrification process. Microbial de-nitrification typically results in progressively increasing δ¹⁵N nitrate values as nitrate concentrations decrease. A significant negative linear correlation was expected between nitrogen isotope ratios and ln[nitrate concentration] when the de-nitrification process occurred.¹² Therefore, we checked our data and hoped to find this pattern by plotting nitrogen isotope ratios versus ln[nitrate concentration].

The results were interesting. When we put our data together, a weak positive correlation was observed, which suggested no significant de-nitrification process in the ground water. This result was not consistent with the dual isotope results. However, we noticed two obvious discontinuity points in Fig. 7 and those discontinuity points happened to coincide with the boundary of the sampling sites. Moreover, when we checked the data from each sampling site, a significant negative correlation (R² > 0.6, as expected) was observed between those data. These results indicate that de-nitrification might occur in the ground water, but the different land uses determined the sources of nitrate and their different original δ¹⁵N–NO₃⁻ and δ¹⁸O–NO₃⁻ values, which conceal the de-nitrification effect when using the δ¹⁵N–NO₃⁻ and δ¹⁸O–NO₃⁻versus ln[NO₃⁻] method, when those samples mixed in this study area. It is worth noting that the ε values in eqn (2) for δ¹⁵N and δ¹⁸O at the G02 site were much higher than for other sites (Fig. 7). This suggested that a higher degree of de-nitrification had taken place at this site. However, due to inadequate sampling numbers and hydro-chemical information, the reasons for this result need further study.

Fig. 7 Results from δ¹⁵N–NO₃⁻ and δ¹⁸O–NO₃⁻versus ln[NO₃⁻] analysis of ground water samples.

3.4. Nitrate source apportionment using SIAR

To estimate the contributions of different NO₃⁻ sources, two types of isotope δ¹⁵N–NO₃⁻ and δ¹⁸O–NO₃⁻ (j = 2) from four major NO₃⁻ sources (atmospheric deposition, AD; NO₃⁻ derived from soil organic matter nitrification, NS; NO₃⁻ derived from chemical fertilizer nitrification, NF; and manure and sewage, M&S) were measured over two periods (the high flow period in September and the low flow period in November). The δ¹⁵N and δ¹⁸O values of different sources were used as site specific isotope values. All sources were normally distributed (Kolmogorov–Smirnoff test, Table 4). As expected, all data were consistent with those reported in the literature. There were observable variations in nitrate isotope compositions in precipitation (n = 4; −0.70 ± 2.41‰ for δ¹⁵N–NO₃⁻ and 43.85 ± 4.93‰ for δ¹⁸O–NO₃⁻). In contrast, the values of the nitrate isotope compositions in sewage (the untreated wastewater) showed little change during the sampling period (n = 6; 15.32 ± 0.83‰ for δ¹⁵N–NO₃⁻ and 9.11 ± 0.93‰ for δ¹⁸O–NO₃⁻). Additionally, the N isotopic composition of soil samples across land use types ranged from +2.91‰ to +7.24‰ (n = 12; 5.81 ± 1.51‰). Chemical fertilizers (n = 6) were measured, and found to have a value of −1.12 ± 1.41‰. However, the δ¹⁸O values of the soil N and synthetic fertilizer sources were not directly measured. The nitrate from “NS” and “NF” sources were mainly produced by microbial nitrification. Therefore, in this study, the δ¹⁸O values of the soil N and synthetic fertilizer sources were set as the same as the δ¹⁸O values from nitrification. This calculation yielded a range of values from −4.04‰ to +6.15‰ for δ¹⁸O–NO₃⁻ produced via nitrification (as mentioned in Section 3.3). The calculation data were time-series data, and after statistical analysis, the mean ± SD of those data (1.24 ± 3.13‰) could be obtained. Therefore, the value of 1.24 ± 3.13‰ represented the δ¹⁸O–NO₃⁻ value of nitrification sources in the study.

Table 4 Summary of the δ¹⁵N–NO₃⁻ and δ¹⁸O–NO₃⁻ sources in the studied watershed^a

Sources	N	δ^15N–NO3^− (‰)		δ^18O–NO3^− (‰)		Statistics^c
		Mean ± SD	Referenced range^b	Mean ± SD	Referenced range^b	δ^15N (sig.)	δ^18O (sig.)
a Atmospheric deposition, AD; NO3^− derived from soil organic matter nitrification, NS; NO3^− derived from chemical fertilizer nitrification, NF; manure and sewage, M&S; N, number of samples; SD, standard deviation. b Cited from ref. 37. c Kolmogorov–Smirnoff test; if sig. > 0.05, test distribution is normal.
AD	4	−0.70 ± 2.41	−13–13	43.85 ± 4.93	25–75	0.894	0.82
M&S	6	15.32 ± 0.83	4–25	9.11 ± 0.93	−5–10	0.952	0.992
NS	12	5.81 ± 1.51	−3–8	1.24 ± 3.13	−10–10	0.792	0.877
NF	6	−1.12 ± 1.41	−6–6	1.24 ± 3.13	−10–10	0.683	0.877

---

Because a significant de-nitrification process was observed in the ground water, which altered the original isotopic composition by fractionation, to determine the degree of de-nitrification and to reproduce the value of the initial isotopic composition in ground water which was employed in the SIAR model, further research was needed. Therefore, in this study, the SIAR model was utilized only for the surface water. The fractionation factors for all sources [i.e., c_jk in eqn (5)] were set as zero, because no significant signs of de-nitrification were observed in the surface water (as described in Section 3.3). Measured isotope data from the surface water samples were used as inputs in SIAR. In order to identify the sources of nitrate in different land use areas, the studied watershed was divided into four sections: forest areas (including sampling sites S01 and S02), urban areas (including sampling sites S03 and S04), agricultural areas (including sampling sites S05, S06 and S07) and wetland areas (including sampling sites S08, S09 and S10). Two different runs of the SIAR model were performed. In the first run, for the forest areas, NF (fertilizer nitrification) and M&S (manure and sewage) as potential NO₃⁻ sources were excluded since in this area there were no fertilizer and sewage inputs, because this area was the headwater of the Abujiao River, and the human activities were limited. In the second, for other areas, all four sources were included in the calculation.

The ranges of contributions of each NO₃⁻ source to the surface water in the two periods are presented in Fig. 8. The SIAR mixing model outputs revealed a high variability in the contribution of the four potential NO₃⁻ sources. The surface water showed distinctly different source apportionment patterns during the two sampling periods. During low flow period, the contribution of “M&S” is the highest (mean probability estimate (MPE) 36.17%), followed by “NS” (MPE 35.69%), “NF” (MPE 22.20%), and “AD” (MPE 5.94%). Compared to the low flow period, a decrease in the “M&S” source contribution occurred (MPE from 36.17% to 22.88%), which at the same time resulted in an increased contribution of “AD” and “NF” (MPE from 5.94% to 15.46% and from 22.20% to 30.10%, respectively). Between the two sampling periods, the contribution of the “NS” sources was similar and the contribution of the source slightly decreased from 35.69 to 31.55. The SIAR outputs were in accordance with the results obtained using Cl⁻ concentration (Section 3.3) and the dual plot approach (Section 3.3), which suggests that the agricultural activities and urban wastewater dominated the nitrate distribution for the high and low flow periods. In general, the results from the SIAR model indicate that the sources from human activities (“M&S” and “NF”) contributed the most nitrate to the aquatic system in the Abujiao watershed (MPE 58.37% and 52.98% for low and high flow periods, respectively).

Fig. 8 Proportional contributions of four potential nitrate sources estimated by SIAR for different sampling periods in the Abujiao River watershed.

The ranges of contributions of each NO₃⁻ source to the surface water in four areas are presented in Fig. 9. According to the SIAR mixing model outputs, the contribution of the four potential NO₃⁻ sources showed high variability. As expected, the surface water showed distinctly different source apportionment patterns between the different areas. In forest areas, the contribution of “NS” is the highest (MPE 86.9%) and the contribution of “AD” only accounted for 13.1% (MPE). This source allocation pattern was similar to previous findings in other regions (the nitrification of soil organic matter as the dominant source for the stream in forest areas).^10,34,35 The highest contribution of “M&S” (MPE 64.43%) was observed in the urban areas. This can be explained by the fact that a large number of untreated wastewaters were poured into the river directly, which made an important contribution to the N input into the river basin. In agricultural areas, with other sources of water input (such as runoff and drainage from farmland), the contribution of “M&S” decreased quickly (MPE from 64.43% in urban areas to 34.91% in agricultural areas) while the contribution of the “NF” and “NS” showed rapid growth (MPE from 10.13% in urban areas to 31.81% in agricultural areas and MPE from 16.55% in urban areas to 22.95% in agricultural areas, respectively).

Fig. 9 Proportional contributions of four potential nitrate sources estimated by SIAR for different land use areas in the Abujiao River watershed.

In wetland areas, the proportion of “NF” and “NS” continued to increase. The highest contribution of “NF” (MEP 35.43%) was observed, followed by “NS” (MEP 29.27%). This was an interesting result, because the agricultural activities were limited (no direct anthropogenic input), so why did the contribution of “NF” continue to increase? Two possible reasons might explain this result. One reason is the input of the drainage from farmland through ditches. As shown in Fig. 1, the sampling site S08 was located at the junction of an agricultural area and wetland. Meanwhile, a tributary (a ditch for farmland drainage) directly flowed into the river, and the interchange site was located between the sampling sites S08 and S09. Therefore, the input of the drainage from farmland might cause the contribution of “NF” to increase. The other reason is the uncertainty of the calculation. Previous studies suggested that small variations in the isotope values of samples and the nitrogen cycle might result in significant changes in the source apportionment as estimated using SIAR.^8,36 Therefore, this might be another reason for this unexpected result.

It should be noted that in this study, NO₃⁻ sources, particularly AD and M&S, had relatively wide ranges of δ¹⁵N–NO₃⁻, δ¹⁸O–NO₃⁻ values, which caused the contribution of the estimated sources from “NF” and “NS” to show a high degree of uncertainty. Therefore, the uncertainty of the potential NO₃⁻ sources for this watershed should be given special attention and additional sources and water samples may be required to reduce uncertainty.

4 Conclusions

Owing to the food crisis, nitrate pollution in aquatic systems caused by intensive human activities is becoming an increasingly serious problem in the Sanjiang Plain in north-eastern China. However, to control the main source of contamination remains challenging, partly because the information on the nitrate sources and behaviors in aquatic systems is still limited. In this study, an intense human activity-impacted watershed with mixed land use (Abujiao River watershed) was selected as the study area. There were considerable differences in the physicochemical properties between the surface water and ground water, except for Cl⁻ levels. The NO₃⁻ in the ground water (7.84 ± 5.83 mg L⁻¹, mean ± SD) was the dominant species (>95%) of DIN during both studied periods, and the concentration was much higher than in the river (1.15 ± 0.84 mg L⁻¹) for all sample areas. Nevertheless, no water sample in the study area exceeds the WHO drinking water limit of 50 mg L⁻¹ NO₃⁻. The nitrate concentrations in the surface water were greatly influenced by precipitation and land use conditions during the two sampling periods (high flow period, September; low flow period, November). In contrast, there was no significant temporal variation in NO₃⁻ in the ground water.

The transformation processes were identified using Cl⁻ concentration and dual isotope approach (δ¹⁵N–NO₃⁻ and δ¹⁸O–NO₃⁻). The variation of Cl⁻ suggested that during the high flow period, the dilution process (the surface water mixing with runoff caused by precipitation and drainage from farmland) was dominant and other processes such as de-nitrification occurring in the rivers contributed little to the variation of NO₃⁻ in the surface water. However, this dilution process decreased during the low flow period due to the limited water supplement. The sampled water δ¹⁸O–NO₃⁻ values suggest that the nitrification process was not the main N cycling process in the study area, because most measured δ¹⁸O–NO₃⁻ values were above the expected δ¹⁸O–NO₃⁻ from nitrification throughout the sampling period. Both chemical and isotopic characteristics indicated that the signs of de-nitrification were absent in the surface water. However, significant de-nitrification was observed in the ground water for all sample periods in the study area.

The Bayesian model (stable isotope analysis in R, SIAR) was used to apportion the contribution of potential sources. The results from SIAR software show that source contributions differ significantly between the high flow period (September) and low flow period (November). In the high flow period, “NF” and “NS” equally contributed as the major sources in the surface water. In contrast, during the low flow (November) period, “M&S” and “NS” dominated the source contribution along the downstream part of the river. Meanwhile, the contribution of the four potential NO₃⁻ sources also showed high variability between the different land use areas. The uncertainty of the potential NO₃⁻ sources for this watershed should be given special attention and additional sources and water samples may be required to reduce uncertainty.

This study proved that agricultural and residents' activities controlled the NO₃⁻ distribution during the high and low flow period, respectively, and with the assessment of temporal variation and sources of NO₃⁻, better agricultural management practices and sewage disposal programs can be implemented to protect water quality in this high latitude watershed.

Acknowledgements

This research work was funded by the National Natural Science Foundation of China (no. 41171384; 41301529) and National Science & Technology Pillar Program of China during the Twelfth Five-year Plan Period (no. 216027).

References

1. K. Song, D. Liu, Z. Wang, B. Zhang, C. Jin, F. Li and H. Liu, Acta Geograph. Sin., 2008, 63, 93.
2. F. Hao, X. Lai, W. Ouyang, Y. Xu, X. Wei and K. Song, Environ. Manage., 2012, 50, 888–899.
3. Y. Cao, C. Tang, X. Song, C. Liu and Y. Zhang, J. Environ. Monit., 2012, 14, 2624–2633.
4. W. Ouyang, H. Huang, F. Hao and B. Guo, J. Hydrol., 2013, 495, 126–134.
5. M. B. David, L. E. Gentry, D. A. Kovacic and K. M. Smith, J. Environ. Qual., 1997, 26, 1038–1048.
6. T. Liu, F. Wang, G. Michalski, X. Xia and S. Liu, Environ. Sci. Technol., 2013, 47, 13412–13421.
7. J. Ding, B. Xi, R. Gao, L. He, H. Liu, X. Dai and Y. Yu, Sci. Total Environ., 2014, 484, 10–18.
8. L. Yang, J. Han, J. Xue, L. Zeng, J. Shi, L. Wu and Y. Jiang, Sci. Total Environ., 2013, 463, 340–347.
9. X. Pu, H. Cheng, Y. Shan, S. Chen, Z. Ding and F. Hao, Acta Agric. Scand., Sect. B, 2012, 62, 749–764.
10. D. A. Burns, E. W. Boyer, E. M. Elliott and C. Kendall, J. Environ. Qual., 2009, 38, 1149–1159.
11. C. Kendall, M. B. Young and S. R. Silva, Applications of stable isotopes for regional to national-scale water quality and environmental monitoring programs, in Isoscapes, ed. J. B. West, G. J. Bowen, T. E. Dawson and K. P. Tu, Springer, 2010, pp. 89–111.
12. C. Kendall, E. M. Elliott and S. D. Wankel, Stable Isot. Ecol. Environ. Sci., 2007, 2, 375–449.
13. F. Chen, G. Jia and J. Chen, Biogeochemistry, 2009, 94, 163–174.
14. S. D. Wankel, C. Kendall and A. Paytan, J. Geophys. Res.: Biogeosci., 2005–2012, 2009, 114.
15. C. Kendall and J. J. McDonnell, Isotope tracers in catchment hydrology, Elsevier, 1998.
16. B. Mayer, S. M. Bollwerk, T. Mansfeldt, B. Hütter and J. Veizer, Geochim. Cosmochim. Acta, 2001, 65, 2743–2756.
17. R. Aravena and W. D. Robertson, Ground Water, 1998, 36, 975–982.
18. D. Xue, J. Botte, B. De Baets, F. Accoe, A. Nestler, P. Taylor, O. Van Cleemput, M. Berglund and P. Boeckx, Water Res., 2009, 43, 1159–1170.
19. D. Xue, B. De Baets, O. Van Cleemput, C. Hennessy, M. Berglund and P. Boeckx, Environ. Pollut., 2012, 161, 43–49.
20. S. Bernal, A. Butturini and F. Sabater, Biogeochemistry, 2006, 81, 269–289.
21. S. Silva, C. Kendall, D. Wilkison, A. Ziegler, C. Chang and R. Avanzino, J. Hydrol., 2000, 228, 22–36.
22. K. K. Andersson and A. B. Hooper, FEBS Lett., 1983, 164, 236–240.
23. D. A. Burns and C. Kendall, Water Resour. Res., 2002, 38, 9-1–9-11.
24. D. Kaown, D.-C. Koh, B. Mayer and K.-K. Lee, Agric., Ecosyst. Environ., 2009, 132, 223–231.
25. A. C. Parnell, R. Inger, S. Bearhop and A. L. Jackson, PLoS One, 2010, 5, e9672.
26. J. B. Hopkins III and J. M. Ferguson, PLoS One, 2012, 7, e28478.
27. A. L. Jackson, R. Inger, S. Bearhop and A. Parnell, Ecol. Lett., 2009, 12, E1–E5.
28. S. Silva, P. Ging, R. Lee, J. Ebbert, A. Tesoriero and E. Inkpen, Environ. Forensics, 2002, 3, 125–130.
29. F.-J. Yue, C.-Q. Liu, S.-L. Li, Z.-Q. Zhao, X.-L. Liu, H. Ding, B.-J. Liu and J. Zhong, J. Hydrol., 2014, 519, 329–339.
30. D. Widory, E. Petelet-Giraud, P. Négrel and B. Ladouche, Environ. Sci. Technol., 2005, 39, 539–548.
31. C.-Q. Liu, S.-L. Li, Y.-C. Lang and H.-Y. Xiao, Environ. Sci. Technol., 2006, 40, 6928–6933.
32. δ¹⁸O values in precipitation water for 1989 were taken from Water Isotope System for Data Analysis, Visualization and Electronic Retrieval (WISER), GNIP – Global Network of Isotopes in Precipitation, International Atomic Energy Agency, Vienna, Austria, http://www.univie.ac.at/cartography/project2/wiser/index.php.
33. M. O. Rivett, S. R. Buss, P. Morgan, J. W. Smith and C. D. Bemment, Water Res., 2008, 42, 4215–4232.
34. R. T. Barnes and P. A. Raymond, Ecol. Appl., 2010, 20, 1961–1978.
35. K. B. Piatek, M. J. Mitchell, S. R. Silva and C. Kendall, Water, Air, Soil Pollut., 2005, 165, 13–35.
36. D. Xue, B. De Baets, O. Van Cleemput, C. Hennessy, M. Berglund and P. Boeckx, Environ. Pollut., 2012, 161, 43–49.
37. D. Xue, J. Botte, B. De Baets, F. Accoe, A. Nestler, P. Taylor, O. Van Cleemput, M. Berglund and P. Boeckx, Water Res., 2009, 43, 1159–1170.

---