N₂O Emissions from Aquatic Ecosystems: A Review

Abstract

Emissions of nitrous oxide (N₂O) from aquatic ecosystems are on the rise due to the dramatic increase in global reactive nitrogen input by anthropogenic activities (e.g., agricultural nitrogen fertilizer use). However, uncertainties exist in the estimation of aquatic N₂O budgets due to limited knowledge of mechanisms involved in aquatic N₂O emissions, as well as the N₂O flux measurements and modelling. To give a full picture of aquatic N₂O emissions, this review discusses the biotic and abiotic mechanisms involved in aquatic N₂O emissions, common methods used in aquatic N₂O flux measurements (including field measurement methods and formula simulation methods), and alternatives for aquatic N₂O budget estimation. In addition, this review also suggests that stable isotope technology is promising in the application of aquatic N₂O source partitioning.

1. Introduction

Nitrous oxide (N₂O), whose single-molecular global warming potential is 273 times (100-year timescale) greater than carbon dioxide, is the third most important long-term greenhouse gas (LLGHG) and accounts for about 7% of the total radiative forcing of LLGHG in 2020 [1,2]. N₂O has a long residence time of 110–180 years in the atmosphere [3]. N₂O is stable in the troposphere, but when it escapes into the stratosphere, it is photolyzed into NO radical by ultraviolet light, which further destroys the stratospheric ozone layer [4,5]. After the prohibition of the usage of chlorofluorocarbon in The Montreal Protocol on Substances that Deplete the Ozone Layer, N₂O has become the most important ozone-depleting substance currently being emitted [4,5].

N₂O emissions increased rapidly during the Industrial Revolution [6], mainly due to the increased reactive nitrogen (Nr) input through the Haber–Bosch process [7,8]. According to the latest field network observation data, the global mean atmospheric N₂O concentration reached a new high (333.2 ± 0.1 ppb) in 2020, which is 123% of the pre-industrial level, and is still increasing at an average annual growth rate of 0.25% [2].

Aquatic ecosystems, including inland waters, estuaries, and oceans, have long been considered as net N₂O sources [9,10]. To meet the food demand of the dramatically growing world population, the application of chemical fertilizers increased from 11.46 Mt in 1961 to 113.29 Mt in 2020 to promote crop production (Figure 1). Only 15–70% of fertilizer N is taken up by plants, and some is leached into inland waters and further transported into estuaries and oceans [11,12,13]. Correspondingly, riverine N₂O emissions went through a growth of 91.5% from 0.15 Tg N₂O-N yr⁻¹ in 1961 to 0.29 Tg N₂O-N yr⁻¹ in 2016 [14]. It is estimated that the total N₂O emissions from global rivers and streams account for 10–15% of the total anthropogenic N₂O emissions [15]. More than 80% of anthropogenic nitrogen emissions from aquatic ecosystems come from the northern mid-latitudes, consistent with the geographic distribution of nitrogen use and population [16]. The annual N₂O emissions of the source streams and lakes are 0.36 and 0.52 Tg N₂O yr⁻¹, respectively, which are affected by the discharge of wastewater from agricultural activities, human life, and industrial production [10].

The global biogeochemical cycles of nitrogen in marginal oceans are being strongly interfered with, particularly in estuaries and coastal areas that serve as major transport channels between terrestrial and ocean ecosystems [17,18,19]. Although coastal areas account for only about 18% of the world’s ocean area, they contribute to about 41% of global marine N₂O emissions [20,21,22]. The N₂O emission of the global coastal continental shelf areas affected by human activities was 0.6 Tg N₂O-N yr⁻¹ [23]. The N₂O emissions from estuaries and coastal areas are 0.23 and 0.28 Tg N yr⁻¹, respectively, from 2007 to 2016 (Figure 2) [1,20], suggesting that coastal areas are an important source of global N₂O emissions.

The ocean is the second natural source of N₂O, but with the increase in atmospheric deposition of nitrogen compounds, the anthropogenic N input to the ocean is 1 Tg N yr⁻¹ [24] and the Intergovernmental Panel on Climate Change (IPCC) assumed that the marine N₂O emissions range from 2.5 to 4.3 Tg N yr⁻¹ with an average of 3.4 Tg N yr⁻¹ from 2007 to 2016 (Figure 2) [1,25,26,27]. The global ocean contribution to N₂O emissions is estimated to be between 10% and 53%, with an average of 21%, but ongoing environmental changes are affecting ocean N₂O cycling and emissions to the atmosphere [28,29]. Although it is clear that the ocean is a major natural contributor of N₂O to the atmosphere, quantitative estimates remain uncertain [28,30,31].

This review aims to summarize the current state of knowledge of the role of aquatic ecosystems (inland waters–coastal regions–oceans) in global N₂O emissions and discuss methods of flux measurement and budget quantification, interpret mechanisms involved in N₂O production, and point out the power of stable isotope techniques in N₂O source partitioning.

Figure 2. The N₂O budgets of different aquatic ecosystems and their isotopic values. The N₂O emission data of each aquatic ecosystem in units of Tg N₂O-N yr⁻¹ [1,10,20]. The δ¹⁵N isotope value in the figure represents the isotope ratio of ¹⁵N and ¹⁴N, and the site preference (SP) represents the enrichment difference between the proximal N and distal N of the O atom in the N₂O molecule; data derived from [32,33,34,35,36].

Figure 2. The N₂O budgets of different aquatic ecosystems and their isotopic values. The N₂O emission data of each aquatic ecosystem in units of Tg N₂O-N yr⁻¹ [1,10,20]. The δ¹⁵N isotope value in the figure represents the isotope ratio of ¹⁵N and ¹⁴N, and the site preference (SP) represents the enrichment difference between the proximal N and distal N of the O atom in the N₂O molecule; data derived from [32,33,34,35,36].

2. Mechanisms Involved in N₂O Production

2.1. Biological Pathways

N₂O production refers to the conversion of any other N species into N₂O [37]. In aquatic ecosystems, there are five different microbe-mediated pathways that are important for N₂O production (Figure 3): (1) Nitrification is catalyzed by ammonia-oxidizing archaea (AOA) or ammonia-oxidizing bacteria (AOB). NH₄⁺ is stepwise oxidized into hydroxylamine (NH₂OH) and nitrite (NO₂⁻), which is further oxidized into NO₃⁻ by nitrite-oxidizing bacteria (NOB), and N₂O is produced as a by-product of NH₂OH oxidation [38]. (2) Denitrification includes bacterial denitrification and fungal denitrification. The former is usually an anaerobic dissimilatory reduction process of NO₃⁻ to NO₂⁻, NO, N₂O, and N₂ with reductase enzymes of nitrate reductase (Nar), nitrite reductase (Nir), nitric oxide reductase (Nor), and nitrous oxide reductase (Nos), while the latter lacks the gene encoding N₂O reductase (NosZ) and fail to reduce N₂O to N₂. So far, the reduction of N₂O to N₂ by Nos during denitrification is the only known biological pathway of N₂O consumption [4]. (3) Nitrate reduced to ammonium (DNRA) is a process that is primarily driven by nitrate-reducing bacteria and nitrite-reducing bacteria; N₂O is produced as a byproduct of NO₂⁻ reduction [39]. (4) Nitrifier denitrification refers to the process in which denitrification is conducted by nitrifiers instead of denitrifiers [40]. (5) Coupled nitrification and denitrification is the oxidation of NH₄⁺ to NO₃⁻ under an aerobic or micro-oxygen environment and the reduction of NO₃⁻ to N₂O under anaerobic or hypoxic conditions [41,42,43,44,45,46].

Microbial nitrification and denitrification are biological pathways that contribute largely to aquatic N₂O production [47,48,49]. In heavily polluted estuaries, the contribution of nitrification and denitrification to N₂O emission ranges from 4.52% to 38.1% and from 61.9% to 80.3%, respectively [32,50]. The anoxic zone of the estuary accounts for 90% of the total estuarine N₂O production, which is very likely from denitrification [35,51,52]. The contribution of denitrification to total N₂O emissions is about 44%, 14%, 1%, and 11% in global continental shelves, marine oxygen minimum zones (OMZs), estuarine, and freshwater ecosystems (except groundwater), respectively [16]. When the estuary is affected by ocean acidification, N₂O release from the sediment is stimulated mainly by bacterial denitrification, while in a neutral environment, N₂O production is dominated by fungi [53].

2.2. Abiotic Pathways

Since the intermediates (NH₂OH and NO₂⁻) of microbial nitrification and denitrification are highly chemically reactive, they are quickly chemically converted into N₂O when leached into the environment [54]. Under acidic conditions, NO₂⁻ is protonated to HNO₂, which is more reactive than NO₂⁻ and easily reduced to N₂O by soil organic matter or reduced transitional metals in the environment [55]. The chemical pathway in which NO₂⁻ is chemically reduced to N₂O is called chemodenitrification (Equations (1) and (2)). Ferrous iron (Fe (II)) and lignin-derived soil organic compounds are common reductants in chemodenitrification [44,56,57,58,59,60]. In general, structural iron plays a more important role in chemodenitrification compared with dissolved free Fe (II) ions [61]. In contrast to chemodenitrification, the chemical oxidation of NH₂OH becomes quicker when the pH increases because NH₂OH is more stable in acidic conditions [54]. When NH₂OH is leached into the environment, it can be easily oxidized to N₂O by oxidative transition metals, such as Fe (III) and Mn⁴⁺ (Equations (3) and (4)).

$$4{Fe}^{2+}+2{NO}_2^{-}+5H_{2}O\to 2{Fe}^{3+}+N_{2}O+6H^{+}$$

(1)

$$2{Mn}^{2+}+2{NO}_2^{-}+6H^{+}\to 2{Mn}^{4+}+N_{2}O+3H_{2}O$$

(2)

$$4{Fe}^{3+}+2{NH}_{2}OH\to 4{Fe}^{2+}+N_{2}O+H_{2}O+4H^{+}$$

(3)

$$2{Mn}^{4+}+2{NH}_{2}OH\to 2{Mn}^{2+}+N_{2}O+H_{2}O+4H^{+}$$

(4)

More and more studies demonstrate that chemodenitrification and chemical hydroxylamine oxidation play a significant role in global N₂O production [45,62,63]. At least 15–25% of N₂O formation in coastal marine sediments is caused by chemodenitrification [44]. In the reaction of Fe (Ⅱ) with NO₂⁻, the percentage of NO₂⁻ converted to N₂O ranges from 11% to 52%, reflecting a considerable difference in the degree of chemodenitrification reactions [56]. In intertidal sediments, N₂O production with NO₂⁻ as a precursor averaged from 70% to 80%, indicating that the N₂O was largely catalyzed by fungal denitrification and abiotic reactions such as chemodenitrification [59]. Abiotic NH₂OH oxidation is also an important source of N₂O in coastal ecosystems; it has been reported that NH₂OH produced by nitrifying bacteria is rapidly oxidized to N₂O by active Mn (III/IV)-oxidizing minerals [64]. It has been proved that abiotic N₂O production plays an important role in coastal N₂O emissions, while their roles in streams and seawater are still unclear [37,44,59,64,65].

3. Assessment of Aquatic N₂O Budget

3.1. N₂O Flux Measurements

Floating chambers, including static and dynamic floating chambers, are the most commonly used methods to determine the spatiotemporal variation in aquatic N₂O emissions [66]. The floating static chamber refers to using a top-sealed box placed on the surface of the water to collect the N₂O emitted from the surface water through diffusion (Figure 4a). The N₂O concentration in the chamber will be analyzed at regular time intervals in the laboratory [67,68], so that the N₂O fluxes can be calculated according to the increasing rate of N₂O concentration in the chamber over time, as follows [67,68,69,70]:

$$F_{N2O}=\frac{n_t-n_0}{A\times t}$$

(5)

where F_N2O (mol m⁻² d⁻¹) is the N₂O flux through the water–air interface; n_t (mol) and n₀ (mol) are the mole numbers of N₂O in the chamber at time t and 0, respectively; A (m²) is the surface area of the chamber in contact with water; and t (day) is the time for which the concentration of the gas increases linearly.

In contrast, the dynamic floating chambers are based on continuous measurements of gases in the chamber (Figure 4b) [66,71]. At a given airflow rate, the dynamic chamber can reach a steady state in a short time after placement and it can reduce headspace concentration build-up in the gas chamber [66,72]. The gas flow rate extracted by the gas analyzer is about 0.52 L min⁻¹, which is lower than the flow rate of N₂-loaded gas controlled by the mass flow controller (7.72L min⁻¹) [66]. In addition, an injection port is added to the intake carrier pipe (near the syringe in Figure 4b), which can be used to inject a known volume of greenhouse gases for calibration. In terms of ebullitive fluxes, the method can detect bubbles containing more than 1.6 × 10⁻³ mL of gas and a larger chamber surface area in contact with water will improve the ability to trap more bubbles in a given time [66]. It has been found that continuous monitoring data can accurately reflect the 24 h N₂O emission concentration [71]. The flux can be determined by following equation [66]:

$$F=(\frac{d(\theta \frac{d{C}_D}{dt}+C_D)}{dt}+\frac{{\theta }_D\frac{{dC}_D}{dt}+C_D-C_0}{{\theta }_{DC}})\times \frac{V_{DC}}{A_{DC}}$$

(6)

where F indicates the instantaneous flux (g m⁻² s⁻¹); C_D is the N₂O concentration (g m⁻³) measured by the gas analyzer; C₀ is the potential concentration of the influent gas (g m⁻³); θ_DC and θ_D are the residence time (s) in the dynamic chamber and the cavity of the gas analyzer, respectively; V_DC is the volume (m³) of the headspace of the dynamic chamber; and A_DC is the area (m²) of the dynamic chamber in contact with the aquatic ecosystem.

The uncertainty of static floating chamber methods mainly comes from the natural turbulence at the water–air interface when deploying floating chambers, so this method is recommended for low fraction and low wave conditions [10,20]. The impact of natural turbulence can be reduced by minimizing the chamber size, smoothing the surface, employing calm water conditions, and taking measurements during stable meteorological conditions. The gas detection accuracy of a dynamic floating chamber is affected by the carrier gas flow and the surface water area covered by the floating chamber. Therefore, it is necessary to set appropriate experimental parameters and select a detector with high sensitivity. For example, the internal pump of the ultra-portable greenhouse gas analyzer (UGGA, Model 915-0011, Los Gatos Research, Inc., Mountain View, CA, USA) can be used to extract the gas in the chamber and determine the concentration of greenhouse gases and water vapor at 1hz [66].

The water–air gas exchange model is another method to indirectly estimate the aquatic N₂O flux (Figure 4c) [73]. It is based on the principle that the gas transfer at the water–air interface depends on the combination of the gas diffusion rate and concentration gradient [74]. The gas exchange model aims to calculate the N₂O flux (F_N2O) according to the N₂O water–air gas transfer velocity (k_w) using the following formulas [15,42,67,75,76,77]:

$$F_{N2O}=k_w(C_{obc}-C_{eq})$$

(7)

$$k_w=k_{600}(\frac{S_c}{600}{)}^{-0.5}$$

(8)

$$S_c=2055.6-137.11t+4.3173\times t^2-0.054350\times t^3$$

(9)

where F_N2O and k_w are in the units of μmol·m⁻²·d⁻¹ and m·d⁻¹, respectively; C_obs is the measured concentration (μmol L⁻¹) of dissolved N₂O in the water, C_eq (μmol L⁻¹) represents the air-equilibrated concentration of dissolved N₂O, which is calculated according to the temperature, air pressure, and salinity of the sampling sites; S_c is the Schmidt coefficient in fresh water; and t refers to the in situ temperature. k₆₀₀ is the normalized value to a Schmidt coefficient of 600 at 20 °C, which is generally calculated by empirical hydrology parameter models.

The uncertainties of the water–air gas exchange method are associated with how the wind or water turbulence flow affects gas exchange across the water–air interface [10]. Several different equations for k₆₀₀ can be used to calculate a series of fluxes to reduce the deviation [15,20]. Equations that include the slope and velocity of a river are the best at predicting the speed of gas transfer, while equations that include depth terms have the strongest correlations (

Table 1. Equations for predicting the k₆₀₀ (m d⁻¹) based on stream velocity (V, in m s⁻¹), slope (S, unitless), depth (D, in meters), discharge (Q, in m³ s⁻¹), the Froude number (Fr = V/(gD)^0.5), and wind speed at a 10 m height (W, in m s⁻¹). All p-values for the regressions are 0.001. “-” indicates no measurement.

Model Equation	R2	Slope	y-Intercept	Ref.
k600 = (VS)0.89 × D0.54 × 5037	0.72	0.92 ± 0.024	0.98 ± 0.17	[8]
k600 = 5937 × (1–2.54 × Fr2) × (VS)0.89 × D0.58	0.76	0.94 ± 0.022	0.76 ± 0.16
k600 = 1162 × V0.85S0.77	0.54	0.91 ± 0.036	0.91 ± 0.24
k600 = (VS)0.76 × 951.5	0.53	0.82 ± 0.037	0.92 ± 0.24
k600 = VS × 2841 + 2.02	0.55	1.0 ± 0.038	−4.8 × 10−3 ± 0.26
k600 = 929 × (VS)0.75 × Q0.011	0.53	0.92 ± 0.036	0.81 ± 0.24
k600 = 4725 × (VS)0.86 × D0.66 × Q−0.14	0.76	0.95 ± 0.023	0.57 ± 0.17
k600 = 1.91 × e0.35W	-	-	-	[80]
k600 = 0.314 × W2 − 0.436 × W + 3.99	-	-	-	[81]
k600 = 1.0 + 1.719 × (V/D)0.5 + 2.58 × W	-	-	-	[82]
k600 = 17.19 × V0.5 × D−0.5 + 2.58 × W + 1.0	-	-	-	[83]

) [78], and other wind-speed-based or wind speed–hydrological models also used to calculate k₆₀₀ (Table 1). The gas flux calculated using the multi-model average k₆₀₀ is conservative but comparable to other models [15]. Field measurements of N₂O obtained from the above bottom-up methods are often used to estimate N₂O emissions in regional and global aquatic ecosystems but basically show high spatiotemporal heterogeneity [10,79].

3.2. Estimation Based on Emission Factor (EF)

IPCC Tier 1 assumes that the N₂O emissions from aquatic systems can be estimated based on the emission factor (EF) of total N leached into aquatic systems [47,84,85,86,87]:

$${\mathit{N}}_{\mathbf{2}}\mathit{O}\mathbf{-}{\mathit{N}}_{\mathit{e}\mathit{m}\mathit{i}\mathit{s}\mathit{s}\mathit{i}\mathit{o}\mathit{n}}=\mathit{E}\mathit{F}\mathbf{\times }\mathit{T}\mathit{o}\mathit{t}\mathit{a}{\mathit{l}}_{\mathit{N}}$$

(10)

where N₂O-N_emission, Total N, and EF are in units of kg N₂O-N yr⁻¹, kg N yr⁻¹, and kg N₂O-N kg⁻¹ N, respectively [1,85]. The default EF value was set at 0.0075 kg N₂O-N/kg N by IPCC in 1997 and revised down to 0.0025 kg N₂O-N/kg N for the convenient estimation of N₂O emission from rivers [86,88,89]. But this recommendation was later revised to 0.0026 kg N₂O-N/kg NO₃-N for rivers, reservoirs, and down-streams, while the EFs for outer estuaries, coastal seawater, and open ocean are still missing [85,89].

In practical application, the data acquisition of N leaching and runoff required to determine the EF of all aquatic ecosystems based on the IPCC definition is incomplete and difficult [10,90,91,92]. Therefore, many studies choose to estimate the EF using the actual NO₃⁻ concentrations in focused aquatic systems [10,89,93,94]:

$$\mathit{E}\mathit{F}\mathbf{=}\frac{{\mathit{C}}_{\mathit{N}\mathbf{2}\mathit{O}}\mathbf{-}\mathit{N}}{{\mathit{C}}_{\mathit{N}{\mathit{O}}_{{\mathbf{3}}^{\mathbf{-}}}}\mathbf{-}\mathit{N}}$$

(11)

where C_N2O−N (mg·L⁻¹) and C_NO3−N (mg·L⁻¹) are concentrations measured in the water.

Uncertainty may be introduced when using EF-based estimation [42] because it ignores potential differences in spatial and temporal N delivery efficiency [90], and multiple sources of input may result in the supersaturation of N₂O [7,78,95,96]. Moreover, N₂O in aquatic ecosystems is mainly produced and consumed through nitrification and denitrification pathways, but the EF-based method does not take into account that these processes may differ significantly under different conditions in diverse waters [68,97]. It is suggested that N₂O₋N/NH₄⁺ should be used to calculate EF in river sections with the ammonium nitrogen pollution type, and N₂O₋N/DIN (DIN ≈ NO₃⁻ + NH₄⁺) should be used to calculate the EF when both NO₃⁻ and NH₄⁺ are considerable N pollutants [98].

Data analysis based on global observations shows that the N₂O fluxes in rivers are positively correlated with the concentrations of NH₄⁺, NO₃⁻, and DIN [90]. N₂O concentrations and EF values show spatial differences associated with different land use types along water bodies [89,99]. The values of EFs could range from 0.000028 to 0.022, corresponding to the variability in the regional environmental condition (

Table 2. N₂O fluxes and EFs of different rivers.

Regions	Aquatic Ecosystems	N2O Flux (μmol·m−2·d−1)	NH4+ (mg/L)	NO3− (mg/L)	EFs (%)	Ref.
Southeast China	Min River Basin	−0.84–3.12	0.01–2.3	1.9–11.9	0.043–0.93 a	[103]
	Estuary of Min River	−2.9–4.3	0.01–0.2	0.4–2.7	0.029–0.25 a	[104]
Northeast China	Liao River Basin	9.36–8539.2	0. 2–15.5	0.02–9.6	0.01–2.2 b	[98]
	Daliao River and Estuary	4.6–145.1	-	-	-	[105]
North China	Lower Haihe River Basin	−7.2–160.8	0.1–0.2	0.3–0.5	0.20–0.39 a	[106]
	Beitang Drainage River and Dagu Drainage River	44.6 ± 39.4	3.3–10.6	-	0.007–0.1 b	[107]
	Duliujian River, Yongdingxin River, and Nanyuhe River	26.9 ± 31.2	3.3–10.6	-	0.003–0.09 b	[107]
Eastern China	Middle and lower reaches of the Yangtze River	0.1–35.0	0.2–0.3	2.6–6.9	0.033–0.053 a	[22]
	Chaohu River Basin	0.4–2102.6	0.3–12.5	0.5–0.7	0.027–0.80 a	[88]
	Xin’an Tang River in Taihu region	35.3–86.7	0.5–1.2	3.0–7.2	0.040–0.044 a	[108]
Southwest China	Chongqing metropolitan river network	4.5–1566.8	-	3.7 ± 2.5	0.47 a	[99]
	Qingshui Stream, Taohua Stream, Panxi Stream, Xiaojia River, Fenghuang Stream, and Jiuqu River in Chongqing	10–4940	0.1–1.6	1.1–9.6	0.14–0.29 a	[15]
Western China	Lancang River	7.70–26.00	0.18–0.45	0.26–0.53	0.41–0.61 a	[109]
	The upper reaches of Yangtze River, Yellow River, Lancang River, and Nu River	9.4 ± 6.2	-	-	0.1 a	[68]
South China	Shenzhen River, Dashahe River, Xixiang River	-	0–9.6	0.3–15.8	0.0028–0.44 a	[110]
	Estuary of Pearl River	31.9 ± 7.5	-	-	-	[76]
		37 ± 15	0–8.46	0–14.26	0.13 a	[111]
USA	San Joaquin River	22.6–177.4	-	0.2–15.5	0.003–0.21 a	[67]
	Kalamazoo River	−7.7–228.7	0–0.4	0.003–27.4	0.05–0.014 a	[42]
	Connecticut River	28.9	0.03–0.07	0.5–2.3	0.09–0.44 a	[84]
Canada	Grand River	−35–4200	-	-	-	[68]
	Ontario Streams	0.48–199.2	0.002–0.4	0.5–12.8	-	[101]
Sweden	Uppsala watershed	92.5–150	<0.1	4.4–22.1	0.11 a	[100]
Japan	Tama River	-	0.4–0.7	8.0–30.9	0.019 a	[94]
France	Haut-Loir watershed	-	0.03–0.05	27.9–50.5	0.014–0.095 a	[112]
	Seine River	2.3–193.2	-	22.1–31	0.01–0.028 a	[95]
Sub-Saharan Africa	Congo River	−52–319	0.04	0.4	0.0024 a	[113]
Kenya	Mara River	13.7	0.02–0.19	0.65–1.77	0.031–0.04 a	[114]
Malaysia	Lupar and Saribas Rivers	21.3	0.012–0.12	0.3–2.3	0.1–1.19 a	[48]
South Asia	Cochin Estuary	0.5–29.7	0.02–0.5	0.4–2.6	0.21–0.37 a	[115]
	Adyar River and Estuary	0.24–122.4	7.6–77	0.3–2.8	0.03–1.76 a	[116]
UK	Upper Thurne River	129.8	0.4	7.1	0.11 a	[91]
	Wensum, Eden, and Avon Rivers	50.0	-	19.8–37.7	0.006–0.24 a	[102]
New Zealand	LII River	20.6–273.4	-	11.1–23.5	0.023–0.048 a	[75]
	Ashburton River	8.6–16.3	-	0.09–6.02	0.17–1.69 a	[117]

^a The N₂O concentration in water divided by the NO₃⁻ concentration. ^b The N₂O concentration in water divided by the concentration of NO₃⁻ and NH₄⁺. “-” indicates no measurement.

). The concentrations of dissolved N₂O and NO₃⁻ were usually higher in agricultural and urban rivers [89]. N₂O flux may be determined by NH₄⁺ concentration in some urban river networks, while NO₃⁻ plays a more important role in agricultural watersheds [42,67,100,101]. For example, in the Liaohe River basin in northeast China, sewage and aquaculture wastewater discharge leads to NH₄⁺ (0.2–15.5mg/L) pollution in the river, while those polluted by NO₃⁻ (0.02–9.6mg/L) are mainly distributed in agricultural areas, and the EFs are 0.4456 and 0.0005, respectively [98]. Therefore, the EFs for inland waters and estuaries should fully consider the influence of specific factors such as river basin land use type and water pollution type, and adopting region-specific EF values is more appropriate for the estimation of regional and global aquatic N₂O emissions [10,89,102].

3.3. Model Simulation

The modeling of N₂O emissions from aquatic ecosystems can be mainly categorized into two types: statistical models and process-based models. Statistical models of N₂O emissions, including empirical models and semi-empirical models, rely on multiple linear regression analysis which can reflect statistical relationships between emission data and their controlling factors [7,31,118,119,120]. For example, the global Nutrient Export from WaterShed (NEWS) model is the first global applicable semi-empirical model used to estimate aquatic N load rates and N₂O emissions [47,119]. The model estimates N load rates as a function of human activities, such as nitrogen fertilizer use, human sewage, and atmospheric nitrogen deposition [31]. The estimated global N₂O emission from rivers and estuaries based on this model is 1.32 Tg N yr⁻¹ [119]. The global ocean N₂O emissions are estimated to be 2.45 ± 0.80 Tg N yr⁻¹ by a semi-empirical model based on the relationship between ocean primary productivity and the N cycle [30]. Machine learning algorithms, such as Random Forest or Monte Carlo simulation, can be used to identify complex relationships between multiple variables [9,10,26].

Process-based models for N₂O emissions are based on the biogeochemical and physical processes that drive N₂O emissions. For example, the Dynamic Land Ecosystem Model (DLEM) riverine module simulates N₂O emissions according to nitrification and denitrification rates, N inputs, N retention and release rates, oxygen consumption, and organic matter decomposition and estimated riverine N₂O emission to be 0.3 ± 0.06Tg N yr⁻¹ [14]. The Bern3D Earth system model calculates N₂O emissions based on denitrification and organic matter consumption and estimates an oceanic N₂O budget of 4.5 ± 1.0 Tg N yr⁻¹ [25]. Process-based models show good advantages to estimate aquatic N₂O emissions in the absence of measured data and to predict the response of N₂O fluxes to multifactor climate and environmental changes [25,79,121,122]. In addition, the process-based N₂O emission model can be integrated with hydrological models to predict changes in N₂O emissions in response to land use changes beside the water bodies [37,118,123].

Even though multiple models have been developed to predict aquatic N₂O emissions, their results vary from model to model, regarding both natural and anthropogenic sources. The total N₂O emission in rivers and oceans ranges largely from 0.03 to 1.05 Tg N yr⁻¹ and from 2.45 to 4.50 Tg N yr⁻¹, respectively (

Table 3. Estimates of global aquatic emissions of N₂O by different models or methods in units of Tg N yr⁻¹. “-” indicates no measurement.

N2O Emission	Natural	Anthropogenic	Total	Models/Methods	Ref.
Global	9.7 (8.0–12.0)	7.3 (4.2–11.4)	17	Process-based model (DLEM)	[79]
Inland and coastal waters	0.3 (0.3–0.4)	0.5 (0.2–0.7)	0.8	Process-based model (DLEM)	[79]
Rivers and estuaries	0.07	1.25	1.32	Semi-empirical model (NEWS)	[23]
Continental shelves	0.5	0.1	0.6	Semi-empirical model (NEWS)	[23]
Estuaries	-	-	0.148–0.277	Process-based N balance model	[120]
Rivers, reservoirs, lakes, ponds, streams	-	-	0.94	Empirical Monte Carlo simulation	[10]
Estuaries	-	-	0.26	Empirical Monte Carlo simulation	[10]
Riverine	-	-	0.3 ± 0.06	Process-based model (DLEM)	[14]
Riverine	-	-	0.073	Semi-empirical model	[9]
Riverine	-	-	0.03–0.035	Empirical multiple regression model	[90]
Riverine	0.105	0.945	1.05	Semi-empirical model (NEWS)	[124]
Oceans	3.4 (2.5–4.3)	0.1 (0.1–0.2)	3.5	Process-based model (DLEM)	[79]
Oceans	4.2 ± 1.0	-	4.2 ± 1.0	Empirical Random Forest model	[26]
Oceans	-	-	4.5 ± 1.0	Bern3D Earth System model	[25]
Oceans	-	-	2.45± 0.8	Semi-empirical N cycle model	[30]
Oceans	3.5	0	3.5	Semi-empirical model (NEWS)	[23]

). The uncertainties of N₂O-predicting models mainly come from the fact that (1) measurement data are highly limited due to the difficulties of field sampling and (2) the mechanisms involved in aquatic N₂O emissions are not yet fully interpreted [47].

3.4. Uncertainties

Most N₂O emission inventories rely on simple EF methods or model studies [47]. Field measurements provide the most realistic estimates of N₂O fluxes and can provide spatiotemporal datasets for model calibration and verification [10,37,96]. While more field measurements can yield more accurate EFs, they also bring uncertainty due to aquatic conditions and the heterogeneity in land use types along water bodies, leading to a wide range of N₂O estimations [14,90,120,125,126,127]. Moreover, the large-scale deployment of EF measurements is difficult and labor-intensive, and the assessment of aquatic ecosystem EFs using the IPCC method needs detailed N-relevant data, which are difficult to obtain and often missing in many studies [37,70,127]. Despite the limitations, field measurements play an important role in providing accurate estimates of N₂O fluxes and improving model frameworks. In the future, it is necessary to take into account river hydrological characteristics and measure more data for the revision of IPCC EFs [67,125].

There is considerable uncertainty in the assessment of N₂O emission models at both regional and global scales [7,9,118,120]. Uncertainties come mainly from the complexity of the aquatic ecosystems and inadequate measurement data [123,128]. More data networks need to be established to improve model prediction and validation performance [25,28,79]. The N₂O model comparison project is proposed to better identify and ultimately reduce these uncertainties between different models [121], and additional long-term studies measuring results in aquatic ecosystems are needed [28,47].

4. N₂O Source Partitioning with Stable Isotope Technique

4.1. Stable Isotopes of N₂O

Isotopes are atoms of the same element that have different numbers of neutrons, resulting in different atomic masses, such as ¹⁴N and ¹⁵N and ¹⁶O and ¹⁸O. The stable isotope technique is widely used in the qualitative measurement of N-transforming processes, and it is also an effective method to study the pathway of N₂O production [36]. N₂O is an asymmetric linear molecule with the proximal and distal N of the O atom called N^α and N^β, respectively. The difference in ¹⁵N enrichment of N^α and N^β is defined as site preference (SP) of N₂O [4]:

$${\delta }^{15}{N}^{bulk}=\frac{({\delta }^{15}{N}^{\alpha }+{\delta }^{15}{N}^{\beta })}{2}$$

(12)

$$SP={\delta }^{15}{N}^{\alpha }-{\delta }^{15}{N}^{\beta }$$

(13)

where δ¹⁵N^α and δ¹⁵N^β denote the relative ¹⁵N abundance of N^α and N^β, respectively.

Isotopes substituted by ¹⁵N or ¹⁸O are usually studied to provide information about the formation and decomposition processes of N₂O [129,130].

Obvious differences in δ¹⁵N and SP values were detected among each aquatic ecosystem (Figure 2). The δ¹⁵N-N₂O values of freshwater and seawater are −4.64 ± 9.84‰ and 6.63 ± 3.50‰, respectively [34]. The isotopic characteristics of N₂O are controlled by biological, chemical, and physical fractionations, as well as isotopic characteristics of N substrates [101]. Due to the isotope fractionation effect of each N₂O production pathway, the stable isotopic values can be used to analyze the relative contributions of the N₂O production pathways [43,57].

4.2. Isotopic Characteristics of Biological Processes

During nitrification, the isotope effect of N₂O production (δ¹⁵N-N₂O) by the oxidation of NH₄⁺ is −56.6 ± 7.3‰, but when NH₂OH is the only substrate, the δ¹⁵N-N₂O increases to −5.1 ± 12.0‰ [131]. In aquatic environments, the isotope effect of NO₃⁻ reduced to NO₂⁻ is −14.3 ± 9.7‰, while the δ¹⁵N-N₂O for denitrification and nitrifier denitrification from NO₂⁻ to N₂O can be represented by −14.9 ± 6.7‰ and −34.5 ± 0.7‰, respectively [131,132]. The ranges of δ¹⁸O-N₂O produced by fungal denitrification and NH₄⁺ by bacterial nitrification are between 27‰ and 42‰ and 20‰ and 26‰, respectively. The δ¹⁸O-N₂O difference is partly due to the different isotopic compositions of O-containing sources. In the process by which bacteria produce N₂O by oxidizing NH₄⁺, the O comes from oxygen and water vapor in the atmosphere. On the contrary, oxygen atoms are derived from NO₃⁻ or NO₂⁻ substrates during denitrification [133].

Tracing the path of N₂O production by using these SP values has become a research hotspot [6,41,132,134]. The SP values of N₂O produced by different paths have certain differences, ranging from −11‰ to 36.9‰ (Table 4 and Table 5). Unlike δ¹⁸O and δ¹⁵N, SP is thought to reflect the N₂O production mechanism while remaining independent of the substrate’s isotopic signature [135]. For example, when nitrifying bacteria produce N₂O, there may be a positive correlation between SP and δ¹⁸O-N₂O, resulting from nitrification by mixing high SP and ¹⁸O-enriched N₂O from hydroxylamine decomposition with low SP and ¹⁸O-depleted N₂O from nitrifying denitrification [43]. Compared with aerobic denitrification and bacterial denitrification, the N₂O process produced by fungal denitrification and nitrification has a higher SP value [36,136]. The SP value of the coupled nitrification and denitrification process is similar to that of denitrification, with a value of 0.1 ± 1.7‰, indicating that the isotopic SP value can distinguish nitrification from coupled nitrification and denitrification [132]. In addition, differences in the experimental environment or the presence of an unknown N₂O production pathway in the sample will influence the measurement of the SP value. Therefore, it is necessary to reduce uncertainties and limitations through more experiments in the future.

4.3. Isotopic Characteristics of Abiotic Processes

Dual isotope ecosystems of N and O have also made further progress in their application to distinguish abiotic processes. The SP reflects the difference in the regulatory reaction mechanism, and the end value is related to the final concentration and yield of N₂O [56]. For the abiotic reduction of NO₂⁻ by Fe (Ⅱ), the δ¹⁵N and δ¹⁸O of N₂O ranges from −19.8‰ to −3.0‰ and from 29.3‰ to 46.4‰, respectively [56]. The SP values of N₂O produced by NH₂OH oxidation and NO₂⁻ reduction processes are 30.1‰ and 29.5‰, respectively [135], while in the process of NO₂⁻ reduction catalyzed by Fe (Ⅱ), the SP value ranges from 0.4‰ to 35‰ [56,57,137]. In addition, the reported SP of chemodenitrification ranges from −4‰ to 26.5‰, which overlaps with the SP value of biological processes [45,56,57,61]. At present, although the SP value is an important indicator to identify the pathways of N₂O production, most of the studies are pure culture experiments conducted in the laboratory [56,57,61], and the differentiation characteristics of N and O isotope effects on N₂O produced by the coupling of NH₂OH and NO₂⁻ with Fe/Mn in various natural aquatic environments remain unclear, which limits the understanding of the mechanism of N₂O production.

Table 4. Summary of N₂O isotope characteristics produced by each pathway.

N2O Sources	Reaction Type	δ15N-N2O (‰)	δ18O-N2O (‰)	SP (‰)	Ref.
Ocean	-	8.8	60.7	35.7	[138]
	-	−6~21	-	-	[139]
	Biotic	3.5~5.5	35.5~41.5	0–8	[33]
Natural or agricultural soil	Biotic	−38~6	-	-	[139]
Antarctic soil	Biotic + Abiotic	−82 ± 1	23 ± 1	22 ± 1	[62]
11 rivers in southern Ontario, Canada	-	−22.6~10.7	47.4~51.5	-	[101]
Estuarine salt marsh	Biotic	−1.41~2.99	39.66~50.38	4.09~13.69	[129]
Troposphere	-	7.0 ± 0.6	43.7 ± 0.9	18.7 ± 2.2	[140]
Stratosphere	-	11.4	48.3	21.3
Weighted average of marine and land sources	Biotic + Abiotic	−25.5~1.0	9.8~37.4	−0.5 ~15.1

“-” indicates no measurement.

Table 4. Summary of N₂O isotope characteristics produced by each pathway.

N2O Sources	Reaction Type	δ15N-N2O (‰)	δ18O-N2O (‰)	SP (‰)	Ref.
Ocean	-	8.8	60.7	35.7	[138]
	-	−6~21	-	-	[139]
	Biotic	3.5~5.5	35.5~41.5	0–8	[33]
Natural or agricultural soil	Biotic	−38~6	-	-	[139]
Antarctic soil	Biotic + Abiotic	−82 ± 1	23 ± 1	22 ± 1	[62]
11 rivers in southern Ontario, Canada	-	−22.6~10.7	47.4~51.5	-	[101]
Estuarine salt marsh	Biotic	−1.41~2.99	39.66~50.38	4.09~13.69	[129]
Troposphere	-	7.0 ± 0.6	43.7 ± 0.9	18.7 ± 2.2	[140]
Stratosphere	-	11.4	48.3	21.3
Weighted average of marine and land sources	Biotic + Abiotic	−25.5~1.0	9.8~37.4	−0.5 ~15.1

“-” indicates no measurement.

Table 5. Summary of N₂O isotope characterization of biotic and abiotic processes.

Reaction Process	N2O Sources	δ15N-N2O (‰)	δ18O-N2O (‰)	SP (‰)	Ref.
Biotic	Oxidation of NH2OH and NH4+	-	-	32~35	[132]
	Denitrification of NO2− and NO3−	-	-	0
	NO2− reduction	−90~2	13~35	−11~0	[6]
	NH2OH oxidation	-	−10	-	[41]
		−68~19	22~24.5	13~37	[6]
		-	-	32.5~35.6	[132]
	Nitrification	−5.1 ± 12.0	-	-	[131]
		−56.6 ± 7.3	-	-	[131]
	Bacterial denitrification	-	-	23.3 ± 4.2, −5.1 ± 1.8	[134]
	Fungal denitrification	-	37.3 ± 1.3, 31.5 ± 0.5	37.1 ± 2.5, 36.9 ± 2.8	[133]
		-	-	15.8~36.7	[52]
	Nitrifier denitrification	-	-	0.1 ± 1.7	[132]
Abiotic	NO2− reduction	-	-	30.1 ± 1.7	[134]
	NH2OH oxidation	-	-	29.5 ± 1.1	[134]
	Chemodenitrification	2~11	4~10	26.5 ± 0.8	[61]
		−19.8~−3.0	29.3~46.4	0.4~26.0	[56]
		-	-	10~22	[57]

“-” indicates no measurement.

Table 5. Summary of N₂O isotope characterization of biotic and abiotic processes.

Reaction Process	N2O Sources	δ15N-N2O (‰)	δ18O-N2O (‰)	SP (‰)	Ref.
Biotic	Oxidation of NH2OH and NH4+	-	-	32~35	[132]
	Denitrification of NO2− and NO3−	-	-	0
	NO2− reduction	−90~2	13~35	−11~0	[6]
	NH2OH oxidation	-	−10	-	[41]
		−68~19	22~24.5	13~37	[6]
		-	-	32.5~35.6	[132]
	Nitrification	−5.1 ± 12.0	-	-	[131]
		−56.6 ± 7.3	-	-	[131]
	Bacterial denitrification	-	-	23.3 ± 4.2, −5.1 ± 1.8	[134]
	Fungal denitrification	-	37.3 ± 1.3, 31.5 ± 0.5	37.1 ± 2.5, 36.9 ± 2.8	[133]
		-	-	15.8~36.7	[52]
	Nitrifier denitrification	-	-	0.1 ± 1.7	[132]
Abiotic	NO2− reduction	-	-	30.1 ± 1.7	[134]
	NH2OH oxidation	-	-	29.5 ± 1.1	[134]
	Chemodenitrification	2~11	4~10	26.5 ± 0.8	[61]
		−19.8~−3.0	29.3~46.4	0.4~26.0	[56]
		-	-	10~22	[57]

“-” indicates no measurement.

5. Research Prospect

Aquatic ecosystems are important global N₂O sources, but uncertainties still exist in the quantification of the aquatic N₂O budget. To understand more precisely the aquatic N₂O emissions, further studies should be conducted to (1) establish more N₂O flux measurements in aquatic ecosystems, especially data-inadequate areas such as oceans and arctic regions; (2) interpret more clearly the mechanisms involved in aquatic N₂O production and its controlling environmental factors; (3) decrease the uncertainties of N₂O-predicting models and make the results of different models more comparable; and (4) advance the application of stable isotopic techniques in aquatic N₂O emissions regarding both mechanism studies and model establishment.