Introduction

Nitrous oxide (N2O) is a primary stratospheric ozone-depleting substance and the third most important long-lived greenhouse gas (GHG)1, with a global warming potential 273 times that of CO2 over a 100-year horizon2. As an important source of N2O emissions to the atmosphere3, estuaries account for approximately 33% of the global oceanic N2O emissions, despite covering only 0.4% of the world’s ocean area4. However, N2O emissions from estuaries remain a poorly constrained component in the global N2O budget3. Many densely populated estuaries receive a substantial amount of wastewater discharged from surrounding urban areas5. The wastewater input, together with dynamic anthropogenic–estuarine interactions, leads to strong spatial variations and great uncertainties in estuarine N2O emissions5,6. Accurately characterizing their spatial heterogeneity requires high-resolution monitoring, which is critical to improving our understanding of regional and global N2O budgets.

Estuarine N2O is primarily produced through biochemical processes, including nitrification, denitrification, nitrifier denitrification, and dissimilatory nitrate reduction to ammonium (DNRA)7. The production of estuarine N2O is governed by nutrients, organic carbon, dissolved oxygen, microbial activity, coastal vegetation, and hydrologic conditions8,9,10,11. In particular, anthropogenic sources, especially wastewater discharge, are critical contributors to estuarine nutrient input12, accounting for an estimated ~50% of the total nitrogen load in estuarine and riverine watersheds worldwide13,14. Discharges from wastewater treatment plants are typically characterized by low carbon-to-nitrogen ratios15 and low chemical oxygen demand to total nitrogen ratios16. These conditions stimulate denitrification activity due to the abundant NO3-N load, resulting in a high N2O:N2 ratio with limited availability of dissolved organic carbon (DOC) (as electron donor) relative to NO3-N (as electron acceptor)17,18,19.

To estimate estuarine N2O emissions, the Intergovernmental Panel on Climate Change (IPCC) employs dissolved nitrate and the emission factor (EF5e), assuming a linear relationship between N2O-N and NO3-N20. However, previous studies have suggested substantial uncertainties in the estimated global estuarine emissions when using the current IPCC EF5e13,21,22. Considering that the factors affecting N2O emissions in natural estuaries differ from those influenced by anthropogenic activities23,24, the use of a universal EF5e for all estuaries worldwide is questionable13,25. Furthermore, the response pattern of N2O to N load could differ when there is high nutrient input with altered C:N stoichiometry in wastewater-influenced estuaries13,19. Especially, it remains unclear whether a further increase in N input would linearly translate to elevated N2O emission. In urban rivers and lakes, biological saturation17, characterized by nonlinear relationships such as efficiency loss models26 and exponential models27, has been observed due to excessive nutrient input. Therefore, it is imperative to re-evaluate the relationship between N2O emissions and nutrient concentration, as well as the applicability of the current EF5e in human-influenced estuaries.

To address these knowledge gaps, we carried out a high-resolution measurement campaign for N2O concentrations and fluxes in the Pearl River Estuary (PRE), China—a typical wastewater-influenced estuary. We quantify the impact of wastewater discharge on estuarine N2O emissions and identify the likely biochemical processes that shape N2O dynamics and their spatial heterogeneity in the PRE. Through a meta-analysis, we further examine the distinct patterns of N2O emission in response to nitrogen input from wastewater in global estuaries. Despite their relatively small surface area, wastewater-influenced estuaries emit disproportionately high levels of N2O. Meanwhile, due to the progressive biological saturation of N2O production, where the supply of available nitrate exceeds biological demand, the EF5e in wastewater-influenced estuaries is significantly lower than in undisturbed estuaries and the current IPCC’s global level. The findings of this study demonstrate the important role that wastewater nitrogen discharges play in shaping estuarine N2O emissions, highlighting the need to revise the emission factor and consequently the regional and global N2O inventories.

Results and discussion

Heterogeneous N2O concentrations and high emission fluxes in the PRE region

Our high-resolution monitoring campaign, consisting of ~30,000 real-time N2O concentration measurements (Fig. 1a and Supplementary Data in the figshare repository: https://doi.org/10.6084/m9.figshare.24129774), reveals substantial spatial variations in dissolved N2O concentrations across the PRE region. Concentrations ranged from 9.1 to 132.2 nmol L−1 with an average of 31.7 ± 26.5 nmol L−1 (Fig. 2a). The vast majority of PRE exhibited N2O supersaturation (140.91–1723.07%) (Fig. 2b), suggesting that the estuary is a net source of atmospheric N2O28. N2O concentrations, in general, follow an inland-to-sea gradient, with higher mean concentrations observed upstream of Humen (zone i) at 69.9 ± 27.7 nmol L−1, decreasing to 13.1 ± 2.3 nmol L−1 downstream of the estuary (zone iii) (Fig. 2a). Concentration ranges in zones i, ii, and iii were 30.5–132.2 nmol L−1, 13.0–86.4 nmol L−1, and 9.1–20.7 nmol L−1, respectively, while the corresponding saturation ranges were 415–1723%, 177–1119%, and 141–284%. Along the west-to-east transect, slightly higher N2O concentrations were observed on the west side (12.8 nmol L−1) compared with the east side (9.1 nmol L−1) (transect III). Our high-resolution monitoring also provides detailed insights into the spatial heterogeneity of N2O concentrations. Notably, we observed a hotspot of N2O concentration to the northwest of Inner Lingding island (Fig. 2a), where N2O concentrations reached 25.5 ± 3.9 nmol L−1, significantly higher than the surrounding areas (p-value < 0.001).

Fig. 1: Real-time high-resolution monitoring trajectory in the PRE and surrounding wastewater treatment plants.
figure 1

a Monitoring trajectory and sampling sites. Along the monitoring trajectory (red curve), ~30,000 data points of N2O concentrations and environmental parameters (Chlorophyll a, dissolved oxygen (DO), oxidation reduction potential (ORP), salinity, conductivity, total dissolved solids (TDS), turbidity, pH, and temperature) were collected. Yellow circles indicate sampling sites where biogeochemical parameters (NH4-N, NO2-N, NO3-N, dissolved inorganic nitrogen (DIN), dissolved organic carbon (DOC), and total phosphorus (TP)) as well as ammonia-oxidizing and denitrifying microbial genes with qPCR (AOA amoA, AOB amoA, nirS, nirK, nosZ, and narG) were analyzed. The entire PRE is divided into three zones: (1) Upstream of Humen (upper estuary; zone i); (2) Inner Lingdingyang (middle estuary; zone ii between I and II); (3) Outer Lingdingyang (lower estuary; zone iii between II and III). b Distribution of wastewater treatment plants and built-up areas around the PRE. Wastewater treatment plants shown in (b) are national key monitoring wastewater treatment plants in Guangzhou, Shenzhen, Dongguan, Foshan, Zhongshan, Zhuhai, and major wastewater treatment plants in Hong Kong and Macao. The inset shows the study area in red on the map of China. The base map is from RESDC data (http://www.resdc.cn/).

Full size image
Fig. 2: Spatial distribution of N2O concentrations, saturations, and air–water fluxes in the PRE.
figure 2

a Dissolved N2O concentrations. b N2O saturations. c N2O fluxes. Spatial distributions were interpolated from high-resolution, real-time data (Supplementary Fig. 3) using Kriging interpolation. The yellow box in (a) indicates the N2O concentration hotspot to the northwest of Inner Lingding Island. The base map is from RESDC data (http://www.resdc.cn/).

Full size image

Similarly, water–air N2O flux exhibited substantial spatial variability in the PRE, varying by up to two orders of magnitude (5.5–308.3 μmol m−2 d−1). The average N2O flux was 58.5 ± 65.7 μmol m−2 d−1 (Fig. 2c), which is three times the global estuarine average (18.2 μmol m−2 d−1; range: 2.4–199.2 μmol m−2 d−1)3. The average flux density in the PRE was comparable to that reported in other human-impacted estuaries, such as Schelde (66.6 μmol m−2 d−1)29, Thames (69.1 μmol m−2 d−1)30, and Humber (76.6 μmol m−2 d−1)31 estuaries in Europe, Werribee estuary (78 μmol m−2 d−1)32 in Oceania, and Tokay estuary (77.3 μmol m−2 d−1)33 and Adyar estuary (44.3 μmol m−2 d−1)34 in Asia.

Our high-resolution monitoring enables reliable estimation of N2O emissions. As suggested by previous observations in the PRE, there are no strong seasonal variations in N2O concentrations/emissions within the middle and lower estuaries17. In comparison, the upstream of Humen exhibits clear seasonal patterns: N2O concentrations and fluxes in winter and spring are 0.75–2.5 times (on average 1.5 times) higher than in summer and fall17,24. Considering this seasonal variability in the upstream of Humen, we estimate an annual emission of 1.05 × 109 (9.22 × 108–1.23 × 109) g N2O from the entire PRE to the atmosphere. This emission level is comparable to the total emissions from 19 European inner estuaries31, which cover an area of approximately 1840 km2 and emit 1.35 × 109 g N2O yr−1. Given that the median annual global N2O flux3 is 0.23 (range: 0.13–0.44) Tg, the PRE contributes 4.6‰ of global estuarine N2O emissions while covering only 1.4‰ of global estuarine areas. Especially, despite comprising only ~8% of the PRE’s surface area, the upstream section (zone i) accounts for up to 42.16% of N2O emissions. When expressed in terms of 100-year global warming potential (GWP), N2O emissions (6.3 × 109 mol yr−1 CO2-equivalent) are equivalent to ~21% of CO2 emissions (3 × 1010 mol yr−1 CO2)35 or 161% of CH4 emission (3.92 × 109 mol yr−1 CO2-equivalent)36 from the PRE. This amount is also close to ~9.3% of the CO2 sink in the coastal systems of the entire China37.

Impacts of wastewater discharges on N2O emissions in the PRE region

Increasing urban wastewater load can greatly enhance estuarine N2O production5,21. Surrounded by numerous wastewater treatment plants (Fig. 1b), the PRE received 5.55 × 108 tons of treated wastewater in 202138. When further accounting for indirect discharges from surrounding cities, this figure could reach up to 6.80 × 109 tons39. Assuming a dissolved inorganic nitrogen (DIN) concentration in wastewater effluent of 15 mg L−1, which follows the highest discharge standard (Class 1A) for total nitrogen (TN) in municipal wastewater treatment plants in China (GB18918-2002), DIN input is estimated to be (5.95–72.90) × 108 mol. Considering the relatively consistent wastewater discharges over time, we estimate a DIN load of (0.16–2.00) × 107 mol d−1 for both dry and wet seasons.

In addition to wastewater discharges, nutrient inputs from submarine groundwater and upstream river sources are also important. By applying the 224Ra mass balance model, submarine groundwater discharges in the PRE are estimated to contribute water flows of (4.5–10) × 108 m3 d−1 and (1.2–2.7) × 108 m3 d−1 in the dry season and wet season, respectively40. These flows result in DIN loads of (1.0–43) × 107 mol d−1 and (0.26–11) × 107 mol d−1 during the respective seasons40. For riverine nutrient input, the estimated values of DIN loads are ~8 × 107 mol d−1 during the dry season and ~20 × 107 mol d−1 during the wet season40. The contribution of nutrient input from wastewater discharges is comparable to that from submarine groundwater and upstream rivers, suggesting its vital role in regulating the nutrient budget and stoichiometry in the PRE (Fig. 3a). Additionally, the relative high TN to total phosphorus (TP) ratios observed in the aquatic system are associated with submarine groundwater discharge40 and rapid improvements in wastewater treatment41, which can potentially contribute to N2O emissions in the estuary42.

Fig. 3: Influence of wastewater discharges and drivers of N2O emissions in the PRE.
figure 3

a N input from wastewater discharge, submarine groundwater, and upstream rivers, and spatial variations of key physicochemical parameters DIN, salinity, and DO. Upward/downward arrow denotes increasing/decreasing trend from upstream to downstream. Detailed spatial distributions of all physicochemical parameters are in Supplementary Fig. 5. b Relationship between N2O concentration and DOC:NO3-N concentration. c Relationship between dissolved N2O concentration and NO3-N concentration. d Relationship between dissolved N2O concentration and NH4-N concentration. e Abundances and significance of spatial differences in ammonia-oxidizing and denitrification genes in the PRE. Symbols “***”, “**”, and “*” denote different significance levels with p-values <0.001, 0.01 and 0.05, respectively. f Linear regression between N2O concentrations and abundances of denitrifying genes (nirS and nirK). g Linear regression between N2O concentrations and abundances of ammonia-oxidizing genes (AOA and AOB amoA). Shaded areas in (b, c, d, f, and g) show the 95% confidence bands.

Full size image

The upper section of the PRE exhibited higher concentrations of DOC and DIN (Supplementary Fig. 4), primarily influenced by the influx of carbon and nitrogen from nearby urban wastewater treatment plants. With 97 centralized wastewater treatment plants in Guangzhou and Dongguan surrounding the area, the upstream of Humen received 175,081 thousand tons of treated wastewater in 2021, resulting in a discharge intensity per unit area more than four times higher than that in the middle and lower sections38. The increasing DIN loads can lead to the formation of hypoxic and anoxia zones in estuaries43. In the upstream section of the PRE, lower dissolved oxygen concentrations were observed (Supplementary Fig. 5), which can further stimulate denitrification and result in a high N2O:N2 ratio under high NO3-N concentration conditions19 (Supplementary Fig. 4).

Compared with natural estuaries, the discharge from wastewater treatment plants in human-influenced estuaries exhibits lower carbon-to-nitrogen ratios, potentially leading to different N2O production rates and yields. N2O concentrations upstream of Humen (zone i) showed a negative correlation with the ratio of DOC to NO3-N (R2 = 0.681, p-value < 0.001) (Fig. 3b). This correlation suggests a possible link between N2O emission and the nitrogen reduction process17. With limited DOC relative to NO3-N, denitrification slows down or halts at N2O or NO2− instead of fully converting to N2. It is noteworthy that the average ratio of DOC to NO3-N upstream of PRE was significantly lower compared to five other subtropical estuaries in China21, where denitrification is favored during N2O production. In addition, dissolved N2O showed a relatively stronger correlation with NO3-N concentration (R2 = 0.901, p-value < 0.001) than with NH4-N concentration (R2 = 0.220, p-value < 0.05) in the upstream of Human (Fig. 3c, d), indicating a higher contribution of denitrification and DNRA to N2O production compared with nitrification and nitrifier denitrification14.

The presence of abundant ammonia-oxidizing and denitrifying genes in the PRE (Fig. 3e) indicates that N2O production is collectively mediated by nitrifying and denitrifying microorganisms. In general, the total abundance of denitrifying genes (nirK, nirS, narG, and nosZ) was much higher than that of AOA amoA and AOB amoA, with the upstream section of Humen showing higher gene abundances than the middle and lower sections (Fig. 3e). N2O concentrations showed a positive correlation with the abundance of denitrifying bacterial nirS and nirK genes in the upstream section (R2 = 0.74, p-value < 0.05) (Fig. 3f), suggesting the important role of denitrification in upstream N2O production21. In comparison, N2O concentrations exhibited a stronger correlation with the abundance of ammonia-oxidizing genes than with that of denitrifying genes in Lingdingyang (Fig. 3g). The contribution of nitrification in the upper and middle sections is also supported by the positive correlations between N2O concentration and NH4-N substrate (Fig. 3d), aligning with observations in other anthropogenically influenced estuaries21.

Overall, the combined effects of abundant nutrient input, low C:N ratios, and relatively low dissolved oxygen levels induced by wastewater discharges collectively stimulate N2O emissions. Stronger denitrification occurs in the upstream section of the PRE, which intensifies the spatial heterogeneity of N2O distribution and complicates the biogeochemical processes associated with N2O emissions.

Disproportionately high N2O emissions from global wastewater-influenced estuaries

To better understand the impact of wastewater discharges on estuarine N2O emissions, we investigate the nitrogen input from wastewater44 and associated N2O emissions in 83 estuaries worldwide through a meta-analysis (Fig. 4 and Supplementary Data in the figshare repository: https://doi.org/10.6084/m9.figshare.24129774). On a global scale, dissolved N2O concentrations were positively correlated with wastewater nitrogen input (R2 = 0.15, p-value < 0.05) (Fig. 5a), suggesting a higher contribution of exogenous nitrogen to N2O emissions in wastewater-influenced estuaries. Based on the levels of nitrogen from wastewater discharges, we classify global estuaries into two groups. Compared with estuaries experiencing lighter wastewater influence (nitrogen < 2 Mg N km−2), highly wastewater-influenced estuaries (nitrogen ≥ 2 Mg N km−2) exhibited significantly higher N2O flux (p-value < 0.001, t-test) (Fig. 5c), with a median emission level of 32.20 μmol m−2 d−1 (95% CI: 13.30–48.10 μmol m−2 d−1).

Fig. 4: Wastewater nitrogen in global estuaries.
figure 4

a Distribution of wastewater nitrogen input in global estuaries. b Zoomed-in view of estuarine wastewater nitrogen in China. c Zoomed-in view of estuarine wastewater nitrogen in the PRE. Wastewater nitrogen input is shown in log10(g N). Green circles in (a) represent the 83 estuaries covered in the meta-analysis. Zoomed-in view of different continents are shown in Supplementary Fig. 6.

Full size image
Fig. 5: N2O emissions influenced by wastewater nitrogen discharge in global estuaries.
figure 5

a N2O concentration in relation to wastewater nitrogen discharge, b EF5e in relation to wastewater nitrogen discharge. c Levels of N2O concentration, saturation, emission flux, and emission factors (EF5e) in two groups of estuaries worldwide with different levels of wastewater nitrogen discharge (estuaries with high human impacts: wastewater nitrogen ≥2 Mg N km−2; estuaries with low human impacts: wastewater nitrogen <2 Mg N km−2). Details of estuaries are summarized in Supplementary Data in the figshare repository: https://doi.org/10.6084/m9.figshare.24129774. Symbols “***”, “**”, and “*” in (c) denote p-values < 0.001, 0.01, and 0.05, respectively. Shaded areas in (a) and (b) show the 95% confidence bands.

Full size image

Estuaries facing increasing anthropogenic stress are playing an increasingly important role in predicting future global N2O emissions and assessing climate feedback. Globally, approximately 52% of the total nitrogen load in estuarine and riverine watersheds originates from anthropogenic sources13, aligning closely with the estimate of 56% (based on the ratio of anthropogenic to total nitrogen additions from land) used in calculating anthropogenic estuarine emissions in global N2O budget14. Enhanced nitrogen load resulted in an additional 2.6–9.9 Gmol yr−1 of N2O-N emission from estuaries and rivers13. Meanwhile, estimates of estuarine N2O emissions are associated with large uncertainties due to the interaction of human influences with biochemical processes, leading to limited constraints on both marine and global N2O budgets. Notably, global N2O emissions were estimated to be ~17.0 Tg N yr−1 14, with ~4.2 Tg N yr−1 originating from the ocean45. Therefore, enhancing the accuracy of estuarine N2O emission estimates is crucial to better constrain its contribution to marine and global budgets, especially under anthropogenic perturbations.

N2O emission factors in wastewater-influenced estuaries are significantly lower than the current IPCC EF5e

We further calculate N2O emission factors (EF5e) for PRE and 82 other estuaries globally. Across all estuaries, N2O emission factor is negatively correlated with nitrogen input from wastewater discharge (R2 = 0.34, p-value < 0.05) (Fig. 5b), suggesting a reduced contribution of nitrogen input to water-to-air N2O diffusive emissions in response to increases nutrient substrate. The N2O emission factor for PRE is 0.00052 ± 0.00022 kg N2O-N/kg NO3-N, while the median EF5e for highly wastewater-influenced estuaries globally is 0.00078 (95% CI: 0.00034–0.00116) kg N2O-N/kg NO3-N. Both values are significantly lower than the IPCC’s default value of 0.0026 kg N2O-N/kg NO3-N20. Even the median EF5e for estuaries with low wastewater influences globally—0.00138 (95% CI: 0.00086 to 0.00181) kg N2O-N/kg NO3-N—is lower than the IPCC level. The discrepancy suggests that the current global estuarine N2O emission factor used by IPCC is overestimated by up to an order of magnitude, particularly in estuaries with high wastewater influences.

Previous studies have also revealed large uncertainties associated with the IPCC’s EF5e estimate13,21. Mechanistic modeling approaches that kinetically limit the extent of biochemical reactions have been used to estimate the ranges of emission factors for estuaries13. The results consistently indicate that IPCC’s emission factors are too high to be universally applied across all estuaries. For example, Kroeze et al. applied the IPCC default emission factor to estimate global estuarine N2O emissions at 100 Gg N yr−1 with a TN input of 44 Tg yr−1 46. However, if we consider the latest TN load estimate of 97.6 Tg yr−1 13, estuarine N2O emissions would be approximately 220 Gg N yr−1 when using the IPCC’s default emission factor, significantly higher than the estimate of 60–155 Gg N yr−1 from that based on kinetically mechanistic modeling13,14.

We analyze the relationship between the observed N2O and NO3-N concentrations (Supplementary Table 1) using three models that incorporate biochemical reaction kinetics in N2O production (Fig. 6). The linear model assumes a 1st-order response, in which N2O concentrations are directly proportional to NO3-N concentrations47. The efficiency loss model describes a power relationship with an exponent (or order) less than one, indicating that N2O production by the biota increases with NO3-N availability, but the efficiency of N2O concentrations relative to NO3-N concentrations declines26. The third model employs Michaelis–Menten uptake kinetics, representing a saturation of N2O production as the supply of available NO3-N exceeds biological demand48.

Fig. 6: Relationships between N2O concentrations and NO3-N concentrations characterized by linear model, efficiency loss model, and Michaelis–Menten kinetics models in estuaries worldwide.
figure 6

a Global estuaries. b European estuaries. c Asian estuaries. d North American and Oceanian estuaries. Goodness-of-fit statistics including residual sum of squares (RSE), adjusted R2, root mean square error (RMSE), and Akaike information criterion (AIC) are summarized in Supplementary Table 1.

Full size image

Globally, the functional relationship between biological N2O production and NO3-N concentration is best described by the efficiency loss model (R2 = 0.49, p-value < 0.05) (Fig. 6a), in which the efficiency of N2O production decreases with increasing NO3-N availability. This efficiency loss model is particularly evident in microbial communities adapted to chronic loading, showing high flexibility in response to increasing nutrient concentrations and thus a delayed saturation49. In European and Asian estuaries (Fig. 6b, c), better performance is observed for the Michaelis–Menten model with R2 = 0.42 and 0.28, respectively. These results show that N2O production initially increases linearly with NO3-N concentration but then plateaus at around 1600 nmol L−1 when NO3-N concentration exceeds ~3 mg L−1, similar to observations in wastewater-influenced inland water17. This plateau is mainly attributed to progressive biological saturation in processing nitrogen with increasing DIN load50,51. While the number of favorable denitrification and nitrification sites at the water–sediment interface remains relatively constant52, the efficiency of denitrification and nitrification decreases with increasing nitrogen inputs, eventually leading to relatively stable N2O emissions at high NO3-N levels.

Estuaries with high levels of dissolved N2O saturation are widespread along the coasts of West Europe, Southern Asia, and Eastern Asia, including estuaries such as Clone and Trent estuaries in the UK (Fig. 6b) and the Yellow River and Tianjin River estuaries in China (Fig. 6c). It is noteworthy that data for European estuaries are primarily from the 1980s and 2000s. In comparison, for Asian estuaries, especially in rapidly developing countries such as China and India, the projected increase in nutrient load3 is expected to maintain high N2O emissions. In North America and Oceania (Fig. 6d), a linear model is a better fit for the response of N2O to NO3-N, indicating that the increase in nitrogen input can linearly translate to elevated N2O emissions under conditions of low substrate concentration. This pattern is consistent with observations in estuaries such as Werribee Estuary in Australia and Chesapeake Bay in the USA. Additional fitness statistics for the three models in different estuaries are summarized in Supplementary Table 1.

Refining the IPCC estimate to constrain global estuarine N2O emission contribution

The assumption of a linear response of N2O-N to NO3-N by IPCC is not always valid, as demonstrated by the results discussed earlier. Therefore, it is crucial to develop a global mechanistic model that explicitly captures the spatial and dynamic relationship between substrate availability and N2O emissions. Such a model would provide a more accurate representation of the complex processes occurring in estuaries.

A mechanistic modeling approach for predicting N2O emissions was developed previously13, which incorporates water residence time as a key factor. This refinement adds an important kinetic dimension to existing N2O estimates and better constrains the extent of denitrification and nitrification processes. Additionally, considering piecewise EF5e under different nitrogen load conditions could be an alternative approach to improving the estimation of estuarine N2O emissions. Previous studies have also shown the potential of using regression models based on bottom-up data to derive global riverine N2O emission factors across continents and climate zones52. Despite these advancements in models, the combination of monitoring efforts and modeling endeavors remains essential for improving the accuracy of estuarine N2O emission estimates.

Specifically, it is crucial to reconsider the role of denitrification and its contribution to N2O emissions in human-influenced estuaries. The assumption made by the IPCC that nitrification produces twice as much N2O as denitrification53,54 has been a subject of controversy. Recent studies have reported global nitrification and denitrification fluxes in estuaries to be approximately 0.69 Tmol yr−1 and 0.66 Tmol yr−1, respectively13, indicating a higher contribution from denitrification than suggested by the IPCC. Observations in various estuaries, such as Humber33, Clone55,56, Tianjin57, Mulan river21, and Pear River estuaries, have demonstrated intense N2O production associated with denitrification, particularly under high NO3-N concentrations as shown in Fig. 6.

In the presence of high DIN loads and low oxygen conditions, denitrification can be stimulated18, and N2O is not efficiently consumed as an electron acceptor, resulting in a high N2O:N2 ratio (higher than 5%) in estuaries influenced by urban wastewater19. This phenomenon is common in eutrophic waters and the upper reaches of estuaries3. In comparison, in organic-rich environments, denitrification potential is limited by nitrate availability19. In the case of the PRE, influenced by wastewater discharges, the ratios of DOC:DIN and DOC:NO3-N are 0.51 ± 0.09 and 1.25 ± 0.29, respectively, much lower than the levels found in global rivers (average DOC:NO3-N: 55)58 and agriculture/forestlands (median DOC:DIN: 10)52. Globally, there is a negative correlation between estuarine DOC:NO3-N and N2O concentration (R2 = 0.23, p-value < 0.05) (Fig. 7a) but a positive correlation with EF5e (R2 = 0.85, p-value < 0.01) (Fig. 7b). Therefore, in addition to nutrient, organics and C:N stoichiometry should considered as important variables for estimating N2O emissions, especially in denitrification-dominant estuaries.

Fig. 7: Dissolved N2O concentration and emission factor as functions of DOC:NO3-N in global estuaries.
figure 7

a Relationship between dissolved N2O concentration and DOC:NO3-N. b Relationship between EF5e and DOC:NO3-N. Red lines are linear regression fits, with shaded areas showing the 95% confidence bands.

Full size image

Taken together, the abundant nutrient loads and altered substrate stoichiometry ratio in highly wastewater-influenced estuaries, such as PRE, favor N2O production through nitrification and especially denitrification processes. Consequently, these estuaries often exhibit disproportionately high N2O emissions, which contribute positively to global warming. However, as the nitrogen load continues to increase, the dissolved N2O concentration gradually reaches a plateau due to progressive biological saturation. The systematic overestimation of estuarine N2O emission based on the IPCC’s emission factor mentioned above results from the assumption that N2O emission linearly responds to nitrogen input. Therefore, it is crucial to refine the relationship between N2O emission and DIN substrate and subsequently revise EF5e and global estuarine N2O emission estimates. This can potentially be achieved through appropriate modeling of the kinetic relationship between carbon and nitrogen substrates with estuarine N2O emissions. Additionally, the current study underscores the significance of comprehensive bottom-up monitoring efforts, as they are essential and indispensable for validating and improving the accuracy of models.

Methods

Study area

The Pearl River Estuary (PRE) is located between Guangzhou (23°10’ N) and the Wanshan Islands (22°00’ N), covering an area of ~1794.6 km2, where Pearl River, the second largest river in China in terms of annual water discharge (3.3 × 1011 m3 yr−1), flows into the South China Sea. Surrounded by metropolitan areas such as Hong Kong, Shenzhen, and Guangzhou, this subtropical estuary bears substantial influence from human activities. There are in total of 436 centralized wastewater treatment plants, including 199 national key monitoring plants (Fig. 1b) densely distributed throughout the PRE38,59,60. The designed capacity of these plants is 23,813 × 103 m3 d−1, of which 4942 × 103 m3 d−1 is directly discharged into the estuary38. In 2021, the PRE received 554,897 × 103 tons of treated wastewater discharge38, a figure that could potentially soar to 6,803,960 × 103 tons when further accounting for additional indirect discharges from surrounding cities39. Human activities, especially wastewater discharge, have led to hypoxia61 and eutrophication28 within the PRE, promoting elevated N2O emissions. Given the ongoing urbanization and industrialization trends, N2O emissions from the PRE are expected to further increase in the near future.

We carried out field measurements from October 28 to November 1, 2021, continuously monitoring dissolved N2O concentrations in the water column during a 5-day cruise. To better understand spatial variations in N2O emissions, we divided the entire study area into three zones based on both N2O concentrations and the geometry of the estuary. Zone i, upstream of Humen (upper estuary), spans from tidal channels in Guangzhou to the Humen Outlet, covering an area of ~143 km2. Zone ii, Inner Lingdingyang (middle estuary), extends from the Humen Outlet to Inner Lingding Island, encompassing an area of ~848 km2. Zone iii, Outer Lingdingyang (lower estuary), stretches from Inner Lingding Island to the Outer Estuary, covering an area of ~804 km2.

N2O concentration measurements

We conducted continuous N2O concentration measurements using a fast-response automated gas equilibrator62 attached to a gas analyzer (N2O M1-916; LGR, Canada) aboard a ship during the 5-day field campaign, covering the entire PRE (Fig. 1a). The measured N2O mole fraction was then converted to dissolved concentration in the water column using a device-specific calibration curve, accounting for both partial equilibrations within the device62 and temperature-dependent solubility63. We obtained in total 28,351 individual measurements of dissolved N2O concentrations, with the sampling locations recorded via GPS. We subsequently interpolated the spatial distribution across the entire PRE using Kriging.

We also measured multiple physicochemical parameters, including Chlorophyll a, dissolved oxygen (DO), oxidation reduction potential (ORP), salinity, conductivity, total dissolved solids (TDS), turbidity, pH, and temperature using a multi-parameter probe (EXO2, YSI, USA) (see Supplementary Data in the figshare repository: https://doi.org/10.6084/m9.figshare.24129774). In addition, we collected surface water samples at representative sites for biogeochemical and microbial analyses (yellow circles in Fig. 1a). Air temperature, air pressure, and wind speed were measured in situ with a portable anemometer (YGY-QXY). Annual air temperature and precipitation data were obtained from the National Meteorological Information Center (http://data.cma.cn/).

Fast-Response Automated Gas Equilibrator (FaRAGE) system

Continuous measurement of N2O concentration was based on the Fast-Response Automated Gas Equilibrator (FaRAGE) system62 coupled with a greenhouse gas analyzer (Supplementary Fig. 1). The FaRAGE system is a low-cost equilibrator designed to achieve rapid gas–water equilibration, making it well-suited for field campaigns. The FaRAGE operates as a flow-through system, introducing gas flow into a constant water flow to create a minimal headspace for continuous concentration measurements. It comprises two main components: a gas–water mixing unit and a gas–water separation unit. In the gas–water mixing unit, water is continuously pumped into a 50-mL chamber using a peristaltic pump at a rate of 250 mL min−1. They are then thoroughly mixed with the carrier gas. Efficient degassing is achieved through the jet flow entering the chamber and the generation of micro-bubbles in the water tube and bubble diffusor and is further enhanced in a 2-m long coiled tube (Tygon; inner diameter: 4 mm). The gas–water separation unit then separates headspace gas from water through gravity, after which the headspace gas is directed to the greenhouse gas analyzer for measurement. Our tests suggested that the response time of the FaRAGE system (including the response time of the gas analyzer) was 41 ± 2 s (Supplementary Fig. 2). Note that the measured concentration in the headspace was adjusted by considering the background N2O concentration in the carrier gas (N2, 99.9%). The FaRAGE system has been calibrated62 and applied to high-frequency, real-time monitoring of N2O and CH4 concentrations in the Three Gorges Reservoir64,65 and CH4 emissions in the upper reaches of the Mekong River66, which has enabled accurate high-resolution mapping of gas concentration distributions.

Flux calculation

The N2O flux across the water–air interface can be estimated using the following equation

$$F=k\times ({C}_{0}-{C}_{e})$$
(1)

where \(F\) (μmol m−2 d−1) is the flux, \(k\) (cm h−1) is the normalized gas transfer velocity depending on wind and water temperatures, \({C}_{0}\) is the N2O concentration measured in the water, and \({C}_{e}\) is the N2O concentration in equilibrium with the atmospheric concentration.

The N2O equilibrium concentration is calculated based on Henry’s law as63

$${C}_{e}={x}^{{\prime} }\,{K}_{0}\left(P-p{{{{{{\rm{H}}}}}}}_{2}{{{{{\rm{O}}}}}}\right)\exp \left[P\left(\frac{B+2\delta }{{RT}}\right.+\bar{v}\left(\frac{1-P}{{RT}}\right)\right]$$
(2)
$${{{{{\mathrm{ln}}}}}}\;p{{{{{{\rm{H}}}}}}}_{2}{{{{{\rm{O}}}}}}= \,24.4543-67.4509\left(\frac{100}{T}\right)-4.8489\;{{{{{\mathrm{ln}}}}}}\left(\frac{T}{100}\right)\\ -0.000544S$$
(3)

where \({x}^{{\prime} }\) is the mole fraction of a constituent of dry air (the global abundance of N2O was 333.2 ± 0.1 ppb in 2020 according to the Global Atmosphere Watch Programme’s in situ observation network; http://www.wmo.int), \({K}_{0}\) is the equilibrium constant (mol L−1 atm−1), \(P\) is the total pressure, \(p{{{{{{\rm{H}}}}}}}_{2}{{{{{\rm{O}}}}}}\) is the vapor pressure of water (atm), as a polynomial function of temperature (\(T\)) and salinity (\(S\)) in Eq. (3), \(\bar{v}\) is the partial molal volume, \(R\) is the gas constant (0.08205601 L atm mol−1 K−1), and the term \(\frac{B\,+\,2\delta }{{RT}}\) is calculated as

$$\frac{B+2\delta }{{RT}}=-\frac{9.4563}{T}+0.04739-6.427\times {10}^{-5}T$$
(4)

The gas transfer velocity (\(k\)) of N2O is calculated as67

$$k=0.251\times {u}_{10}^{2}{\left({{Sc}/{Sc}}_{(20^\circ {{{{{\rm{C}}}}}})}\right)}^{-0.5}$$
(5)

where \({Sc}\) is the Schmidt number, and the value for 20 °C is 600 and 660 in freshwater and seawater, respectively; \({u}_{10}\) is the 10-m wind speed and derived from wind speed measurements conducted at a height of 4 m68. The temperature-dependent Schmidt number \({Sc}\) for N2O in freshwater (\({{Sc}}_{{{{{{\rm{f}}}}}}}\)) and seawater (\({{Sc}}_{{{{{{\rm{s}}}}}}}\)) is calculated respectively using67

$${{Sc}}_{{{{{{\rm{f}}}}}}}=2141.2-152.56T+5.8963{T}^{2}-0.12411{T}^{3}+0.0010655{T}^{4}$$
(6)
$${{Sc}}_{{{{{{\rm{s}}}}}}}=2356.2-166.38T+6.3952{T}^{2}-0.13422{T}^{3}+0.0011506{T}^{4}$$
(7)

Freshwater condition is used for the upstream of Humen (upper estuary; zone i), and seawater condition is used for Lingdingyang (middle and lower estuary; zones ii and iii).

Biogeochemical analyses

Collected water samples were filtered on-site through 0.45 μm polyethersulfone (PES) syringe filters to measure concentrations of NO3-N, NO2-N, NH4-N, and DOC. The filtrates were collected in 500 ml plastic sampling bottles, and the pH was adjusted to below 2 by adding H2SO4. Concentrations of NO3-N, NO2-N, and NH4-N were then measured by UV spectrophotometry, naphthylamine hydrochloride spectrophotometry, and hypobromite oxidation methods, respectively. DOC concentrations were determined using a TOC analyzer (Shimadzu TOC-5000, Japan) through high-temperature catalytic oxidation. TP concentrations were measured using an ultraviolet–visible spectrophotometer (Agilent 1200) with the standard molybdenum blue method after persulphate digestion.

Microbial analysis

A total of 27 surface water samples were collected and filtered through 0.22 μm PES filters, and these samples were prepared in triplicate for microbial analysis. DNA from particle-associated and free-living bacterial communities was extracted from approximately 1.5 L homogenized surface water (filtered with membrane) using an E.Z.N.A. soil DNA Kit (Omega Bio-tek, Norcross, GA, USA). The DNA extracts were checked using 1% agarose gel. Subsequently, DNA concentration and purity were determined using a NanoDrop 2000 ultraviolet–visible spectrophotometer (Thermo Scientific, Wilmington, USA). After DNA extraction, real-time quantitative polymerase chain reaction (qPCR) was employed to estimate the abundances of ammonia-oxidizing (AOA and AOB amoA) and denitrifying bacterial genes (nirK, nirS, narG, and nosZ). The abundances of ammonia monooxygenase (AOA and AOB amoA) genes were amplified using primer pair sets arch-amoA-23F/arch-amoA-616R and CTO189f/CTO654r. For denitrifying gene abundances, we used primer sets cd3aF/R3cd, FlaCu/R3cu, nosZF/nosZR, and narG-f/narG-r, which encompass the known diversity of dissimilatory nitrite reductase (nirS and nirK), nitrous oxide reductase (nosZ), and nitrate reductase (narG). qPCR assays were performed in triplicate on an ABI7300 Thermal Cycler (Applied Biosystems, USA). Further details on qPCR primer pairs and amplification conditions can be found in Supplementary Table 2. A 10-fold dilution series was obtained from extracted plasmid DNA with a known copy number and was used to generate an external standard curve. The amplification efficiencies of the archaeal amoA, bacterial amoA, nirS, nirK, narG, and nosZ genes ranged from 89 to 104%, with all amplifications achieving R2 values greater than 0.99.

Global estuarine meta-analysis

To examine the impact of anthropogenic stress on estuarine N2O emission flux and associated spatial patterns, we conducted a meta-analysis on nitrogen input from wastewater44 and N2O emissions in estuaries worldwide. Our dataset includes estuarine emission data from Asia, Europe, North America, South America and Oceania, spanning from the 1980s to the present. We synthesized information on variable parameters such as dissolved N2O concentration, N2O flux, saturation, EF5e, NO3-N, and DOC:NO3-N (Supplementary Data in the figshare repository: https://doi.org/10.6084/m9.figshare.24129774). To determine coastal nitrogen distribution resulting from wastewater input, we used a high-resolution (~1 km) global dataset44, which considers sewered, septic, and untreated wastewater inputs in 142,625 watersheds or coastal areas. The propagation of nitrogen effluent from each pourpoint into coastal waters was quantified using a plume model based on a logarithmic decay function69. We extracted wastewater nitrogen input data for individual estuaries using ArcGIS 10.4. The delineation of estuarine boundaries, represented as irregular polygons, was cross-verified against other studies. For each estuary, we calculated grid-level statistics, including maximum, mean, and standard deviation, for wastewater nitrogen input using Python 3.7. The distribution of wastewater nitrogen in 83 estuaries worldwide is shown in Fig. 4 and Supplementary Fig. 6, with detailed statistics summarized in Supplementary Data in the figshare repository: https://doi.org/10.6084/m9.figshare.24129774.

Generating global estimates from a limited pool of data inevitably introduces uncertainties. These uncertainties primarily stem from the highly uneven spatial distribution of the available in situ measurements, which have been predominantly concentrated in industrialized regions of Asia, Europe, and North America, with limited research efforts in Africa and South America. Nevertheless, our study includes a substantial number of wastewater-influenced estuaries, enhancing our understanding of N2O dynamics in these environments.

For global estuarine data analysis, we performed linear regression and t-test in R. Prior to linear regression analysis, we identified and removed outliers based on box plots. Additionally, we assessed the statistical significance of differences in N2O concentrations, N2O fluxes, N2O saturations, and EF5e values using t-test.

Fitting linear, efficiency loss, and Michaelis–Menten kinetics models

We applied the linear model, efficiency loss model, and Michaelis–Menten kinetics model to evaluate the relationship between dissolved N2O concentration and nitrate concentration. To create a comprehensive global overview, we compiled estuarine data from both previous studies and the present study (Supplementary Data in the figshare repository: https://doi.org/10.6084/m9.figshare.24129774).

Linear regression47 was used to fit the 1st-order response of N2O concentrations to NO3-N concentrations:

$$C=k\times \left[{{NO}}_{3}^{-}\right]$$
(8)

where \(C\) represents the N2O concentration.

In the efficiency loss model26, the N2O production by biota increases with rising NO3-N concentrations, while the efficiency of N2O production relative to NO3-N concentrations declines. This model employs a power relationship in which the exponent is less than one \((n < 1)\):

$$C=k\times {\left[{{NO}}_{3}^{-}\right]}^{n}$$
(9)

The Michaelis–Menten kinetics48 model depicts the N2O production as the supply of NO3-N exceeds biological demand:

$$C=\frac{{C}_{\max }\times \left[{{NO}}_{3}^{-}\right]}{{K}_{m}+\left[{{NO}}_{3}^{-}\right]}$$
(10)

where \({C}_{\max }\) represents the maximum N2O concentration achieved in the estuaries, and \({K}_{m}\) is the required NO3-N concentration to achieve half of \({C}_{\max }\).

Throughout these regression analyses, we applied a log transformation to the dependent variables to reduce the leverage effects of high N2O concentrations. The assessment of model fit, including the residual sum of squares (RSE), adjusted R2, root mean square error (RMSE), and Akaike information criterion (AIC) are summarized in Supplementary Table 1.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.