Isotopic constraints confirm the significant role of microbial nitrogen oxides emissions from the land and ocean environment.

Nitrogen oxides (NOx, the sum of nitric oxide (NO) and N dioxide (NO2)) emissions and deposition have increased markedly over the past several decades, resulting in many adverse outcomes in both terrestrial and oceanic environments. However, because the microbial NOx emissions have been substantially underestimated on the land and unconstrained in the ocean, the global microbial NOx emissions and their importance relative to the known fossil-fuel NOx emissions remain unclear. Here we complied data on stable N isotopes of nitrate in atmospheric particulates over the land and ocean to ground-truth estimates of NOx emissions worldwide. By considering the N isotope effect of NOx transformations to particulate nitrate combined with dominant NOx emissions in the land (coal combustion, oil combustion, biomass burning and microbial N cycle) and ocean (oil combustion, microbial N cycle), we demonstrated that microbial NOx emissions account for 24 ± 4%, 58 ± 3% and 31 ± 12% in the land, ocean and global environment, respectively. Corresponding amounts of microbial NOx emissions in the land (13.6 ± 4.7 Tg N yr-1), ocean (8.8 ± 1.5 Tg N yr-1) and globe (22.5 ± 4.7 Tg N yr-1) are about 0.5, 1.4 and 0.6 times on average those of fossil-fuel NOx emissions in these sectors. Our findings provide empirical constraints on model predictions, revealing significant contributions of the microbial N cycle to regional NOx emissions into the atmospheric system, which is critical information for mitigating strategies, budgeting N deposition and evaluating the effects of atmospheric NOx loading on the world.

INTRODUCTION

Atmospheric nitrogen oxides (NOx) loading influence human health (e.g. respiratory and cardiovascular diseases, acute bronchitis) [1], tropospheric chemistry (e.g. precipitation acidity, aerosol and ozone formation) [2-4], climate [4] and economic development [5]. In past decades, anthropogenic NOx emissions have significantly increased the fluxes of atmospheric NO3− deposition [6-8], altered N cycles in both terrestrial and marine ecosystems [9-12] and thus affected microbial NOx emissions to the atmosphere [13]. Hence, it is pivotal to accurately constrain land and ocean NOx emissions to the atmosphere to mitigate human-induced NOx emissions, budget NO3− deposition fluxes and evaluate the eco-environmental and climatic effects of atmospheric NOx loading. However, it has long been challenging to accurately constrain land- and ocean-to-atmosphere NOx emissions due to uncertainties over microbial N cycles in both land and ocean.

In marine environments, the oil combustion of marine traffic transportation is a known source of NOx emissions [14-20]. According to the European Monitoring and Evaluation Programme Meteorological Synthesizing Centre West model, NOx emissions from oil combustion in the ocean averaged 6.4 ± 0.8 Tg N yr−1 (5.0–7.8 Tg N yr−1) [14-20]. However, the microbial N cycle occurring in the ocean is the other significant source of NOx emissions [21-24]. First, earlier studies based on molecular analysis and lab culture experiments have confirmed that multiple kinds of bacteria associated with several processes of microbial N cycles can produce NO, e.g. ammonium-oxidizing bacteria, nitrite-oxidizing bacteria, methanotrophic bacteria and denitrifying bacteria [25-29]. Second, nitrification in the oxic layer of the ocean is a significant source of NO [22] and NO can be produced in biofilms and marine sediments [30]. Third, Ulva prolifera (forming a belt on a vertical concrete wall in the upper intertidal zone at low tide) was the primary contributor to the high NO concentrations during the late-bloom period [31]. Meanwhile, the photolysis of NO2− and NO3− (in the surface water and on particles) or alkyl nitrates or dissolved organic matter may also be the sources of atmospheric NO in the ocean [32-35]. However, due to its high reactivity [36], NO would be involved quickly into the NOx cycle in the atmosphere [34]. Accordingly, it has long been difficult to accurately observe microbial NO emissions in the ocean [24]. Until now, microbial NOx emissions from the ocean and their fractional contribution to total NOx emissions from the ocean have not been quantified [21-24]. Hitherto, owing to the lack of microbial NOx emissions, the NOx from oil combustion has long been assumed as the total ocean NOx emissions in reports of the Intergovernmental Panel on Climate Change (IPCC) [20].

In the land environment, NOx emissions are mainly derived from coal combustion, oil combustion, biomass burning and microbial N cycles in substrates such as waters, soils and wastes [3,37-40]. Currently, emission amounts of NOx from coal combustion [10,41], oil combustion [42] and biomass burning [43,44] have been reported explicitly in national statistic yearbooks and emission inventories [45-47]. However, land NOx emissions from microbial N cycles have been observed chiefly for soils under natural vegetation and agriculture [40,43,48]. Therefore, estimates of NOx emissions from the land are based on limited empirical observations combined with process and statistical models and satellites used to scale up emissions [40,49,50]. Based on IPCC reports, microbial NOx emissions were budgeted at 5.6 Tg N yr−1 before 2001, increasing to 11.0 Tg N yr−1 when incorporating more observational data in the report of 2013 [40,49,50]. This doubling of emissions highlights a substantial underestimation of microbial NOx emissions in the land, which has shifted with additional measurements and better models. New methods are strongly needed to comprehensively constrain microbial NOx emissions from soils and many other unconsidered substrates (such as the surface water of rivers, lakes, swamps, etc.) and emission sources (such as wastewater, water treatment systems, solid wastes).

Here we provided a unique evaluation of the relative importance of the microbial NOx emissions in the land and ocean to the known fossil-fuel NOx emissions and then made a new budget for global microbial NOx emissions. First, we compiled stable N isotopes (δ15N values) of NO3− in atmospheric particulates (denoted as δ15Np-NO3- hereafter) in the land and ocean, respectively (detailed in ‘Materials and methods’ section) (Fig. 1 and Supplementary Table S1). Second, based on concentrations and δ15N of NOx, HNO3 and p-NO3− over the land, we estimated the δ15N of the initial NOx mixture from different emission sources in the atmosphere (denoted as δ15Ni-NOx, Supplementary Fig. S1) and the difference between δ15Np-NO3- and δ15Ni-NOx values (denoted as 15Δi-NOx→p-NO3-) (detailed in ‘Materials and methods’ section). By using 15Δi-NOx→p-NO3- (Supplementary Fig. S2), δ15Np-NO3- (Fig. 2) and δ15N of dominant sources of NOx emissions (coal combustion, oil combustion, biomass burning and microbial N cycles, Supplementary Table S2), we estimated the relative contributions of dominant NOx sources from the land and ocean, respectively, by developing a model of Stable Isotope Analysis in R code (detailed in ‘Materials and methods’ section). Finally, combining fractional contributions with corresponding amounts of fossil-fuel NOx emissions from the land and the ocean, we calculated the amount of microbial NOx emissions in the land and ocean, respectively (detailed in ‘Materials and methods’ section).

Figure 1.

The distribution of study sites with δ15Np-NO3- observations. Red and blue circles represent land sites (n = 91) and ocean sites (n = 134), respectively.

Figure 2.

δ15Np-NO3- values observed at land sites, observed at ocean sites and derived from ocean NOx emissions. Circles around each box show mean values of replicate measurements at each site (n) (replicate measurements at each site are 1−318 and 1−72 for land and ocean sites, respectively). The box encompasses the 25th to 75th percentiles; whiskers and lines in boxes are the SD and mean values, respectively. Different letters above the boxes show significant differences at P < 0.05.

RESULTS AND DISCUSSION

Different δ15N signatures of atmospheric p-NO3− between the land and ocean

Mean δ15Np-NO3- observed over terrestrial sites (4.7 ± 3.6‰; n = 91) was significantly higher (p < 0.05) than that observed for ocean sites (–3.5 ± 3.9‰; n = 134) (Fig. 2). This finding implied that human activities contributed relatively more 15N-enriched NOx to atmospheric NOx loading on the land than in the ocean.

First, the δ15Np-NO3- signal observed at land sites can represent land NOx emissions without a significant overprinting of marine sources. The net water vapor flux transported from the ocean to the land accounted for only 10% of the total water evaporation over the ocean [51,52]. According to the existing oceanic NOx emissions (6.4 ± 0.8 Tg N yr−1 based on the known oil combustion) [14-20] and the land NOx emissions (53.3 ± 4.6 Tg N yr−1) [43,53-58], the ocean-to-land atmospheric transport of NOx accounts for only 1.2% of land NOx emissions and thus is often assumed negligible [35]. Accordingly, the δ15Np-NO3- values observed at land sites can be directly used to differentiate dominant sources of NOx emissions (Equation 5 in the online Supplementary Data).

However, the δ15Np-NO3- signal observed at ocean sites cannot represent the NO3− purely derived from ocean NOx emissions. Because the land has much higher NOx emissions and a smaller area, and thus a higher concentration than the ocean [57,59,60], the net transportation of atmospheric NOx occurs from the land to the ocean. The modeled NOy (the sum of NOx, inorganic and organic nitrates in the atmosphere) transportation (11.0 Tg N yr−1) [61] is about 1.7 times the oceanic and accounts for 21% of land NOx emissions. Accordingly, the δ15N signals of p-NO3− derived from the land-to-ocean NOy transportation should be excluded (Equation 2 in the online Supplementary Data) to obtain the δ15N values of p-NO3− derived only from the ocean NOx emissions (Supplementary Fig. S1) to differentiate the relative contributions between oil combustion and microbial NOx emissions (Equation 6 in the online Supplementary Data). Besides, the land-derived NOx and p-NO3− are the dominant form of the land-to-ocean NOy transportation and between them, the p-NO3− is the main type to be transported because the lifetime of NOx is much shorter [35,61,62]. So far, no substantial isotope effect was assumed for the physical processes of atmospheric transportation [63,64]. Thus, we thought that the ocean p-NO3− produced by the land-derived NOx did not differ isotopically from the land p-NO3− and used isotope mass-balance calculations to obtain the δ15N values of p-NO3− derived only from the ocean NOx emissions (Equation 2 in the online Supplementary Data).

The calculated results revealed that the δ15N of p-NO3− purely derived from ocean NOx emissions averaged –12.5 ± 8.2‰ (Fig. 2), which was much lower than the δ15Np-NO3- observed for the land sites (4.7 ± 3.6‰; Fig. 2). The increase in 15 N/14N of p-NO3− over the land should be mainly influenced by 15N-enriched NOx sourced to coal combustion, which was distinctly elevated in δ15N values (mean = 14.2 ± 5.1‰, Supplementary Table S2). However, the lower 15 N/14N of p-NO3− derived from ocean NOx emissions revealed a microbial NOx source with distinctly lower δ15N values than other sources (mean = −37.0 ± 13.5‰, Supplementary Table S2). Our findings demonstrated the contrasting δ15N pattern between p-NO3− derived from the land and ocean NOx emissions. Moreover, the newly constrained δ15N of p-NO3− sourced to ocean NOx emissions provided a more accurate and straightforward opportunity to constrain source contributions and emission amounts via isotope modeling.

Relative contributions of dominant NOx sources to p-NO3−

δ15Np-NO3- values are determined by the δ15N of sources and their relative contributions to total NOx emission and isotope effects of the NOx transformation to p-NO3− (15Δi-NOx→p-NO3- values) [65]. Accordingly, we compiled δ15N values of dominant sources of NOx emissions (Supplementary Table S2), constrained 15Δi-NOx→p-NO3- values (Supplementary Fig. S2) and thereby constructed isotope mass-balance models to further evaluate the contribution of dominant NOx sources to p-NO3− in the land and ocean, respectively (detailed in ‘Materials and methods’ section).

For source δ15N end-members, we considered coal combustion, oil combustion, biomass burning and the microbial N cycle as dominant NOx sources of p-NO3− over the land [65], while oil combustion and the microbial N cycle are dominant NOx sources to p-NO3− over the ocean [2,20]. The δ15N of such sources differ significantly from each other (p < 0.05, Supplementary Table S2), which is a prerequisite to differentiating their relative contributions isotopically. We assumed the same δ15N value of each NOx source for both land and ocean sites due to no δ15N observations on NOx from oil combustion and microbial N cycle in the ocean (detailed in ‘Materials and methods’ section). We did not consider lightning a dominant NOx source because the NOx produced by lightning in the land and ocean atmosphere is negligible. First, the global NOx production from lighting is 5.2 ± 1.0 Tg N yr−1 (Supplementary Text S1), which accounted for ∼9.7% and ∼7.2% of global NOx emissions by modeling methods (51.9–58.0 Tg N yr−1) and by isotopic methods in this study (Fig. 3). Moreover, the meridional distribution of global lightning in the atmosphere shows three main lightning centers of the Americas, Africa and the maritime continent in Southeast Asia. The minima represent the oceanic regions where little lightning is observed [66]. This baseline assumption of the dominant NOx sources is supported by emission inventory and deposition modeling [10,41,42,45-47].

Figure 3.

Emissions of significant land and ocean NOx sources (in black and blue) based on natural isotope methods (detailed in ‘Materials and methods’ section). Data of the NOy deposition and transportation (in red) were cited from Refs [35,61,62,70,89,90].

Regarding isotope effects, we estimated 15Δi-NOx→p-NO3- values under two independent scenarios (detailed in ‘Materials and methods’ section) and found no significant differences between them (11.3 ± 2.1‰ and 13.1 ± 3.8‰, respectively) (Supplementary Fig. S2). Accordingly, we used the mean 15Δi-NOx→p-NO3- estimate (12.2 ± 2.2‰) in our subsequent isotope mass-balance calculations (Supplementary Fig. S2). The mean 15Δi-NOx→p-NO3- value in this study (12.2 ± 2.2‰) did not differ from the ϵNO→p-NO3- value estimated by Li et al. [67] (∼15‰) and was also comparable with the global mean 15Δi-NOx→p-NO3- value (16.7 ± 2.3‰) [65]. The calculation of the global mean 15Δi-NOx→p-NO3- value by Song et al. [65] was based on the theoretical framework of computation established by Walters and Michalski [68,69], which combined natural 15N and 17O isotopes with environmental parameters relating to the NOx oxidization to p-NO3−. Relative contributions of dominant NOx sources were calculated using the Stable Isotope Analysis model in R programming language (detailed in ‘Materials and methods’ section). Results showed that the NOx from coal combustion, oil combustion, biomass burning and microbial N cycle accounted for 23 ± 7%, 27 ± 11%, 26 ± 10% and 24 ± 4% on the land, respectively (Supplementary Fig. S3a). In contrast, the NOx from oil combustion and microbial N cycle accounted for 42 ± 3% and 58 ± 3% in the ocean, respectively (Supplementary Fig. S3a). Generally, high fractions of microbial NOx emissions revealed the vital contribution of this pathway to both land and ocean NOx emissions into the global atmosphere.

Total and microbial NOx emissions on the land

Based on statistical data on quantities and NOx emission factors of coal and oil combustions in the land system, previous studies have estimated global fossil-fuel NOx emissions with a relatively high degree of certainty [7,43,50,70,71]. Global fossil-fuel NOx emissions averaged 28.4 ± 1.8 Tg N yr−1, showing a relatively low variation over past decades (25.6–30.0 Tg N yr−1) [7,43,50,70,71]. By using the fraction and amount of fossil-fuel NOx emissions in the land (50 ± 14% and 28.4 ± 1.8 Tg N yr−1, respectively, Supplementary Fig. S3a), we estimated that total land NOx emissions were 56.8 ± 18.6 Tg N yr−1 (Fig. 3 and Supplementary Fig. S3b). Our estimate falls in the range of the total land NOx emissions (50.0–61.4 Tg N yr−1; averaging 55.6 ± 2.9 Tg N yr−1) estimated by optimized modeling methods by considering more microbial sources of NOx emissions [54,57,58]. However, our estimate is higher than the total land NOx emissions (39.7–51.0 Tg N yr−1; averaging 43.8 ± 5.0 Tg N yr−1) estimated using the global NO2 satellite column concentrations [43,55,56]. Due to no consideration of the influences of atmospheric NO2 transformations, the estimates based on the satellite data were thought to underestimate global NOx emissions [72-74].

Based on the fraction and amount of total land NOx emissions (24 ± 4% and 56.8 ± 18.6 Tg N yr−1, respectively, Fig. 3 and Supplementary Fig. S3), microbial NOx emissions on the land were calculated as 13.6 ± 4.7 Tg N yr−1 (Fig. 3 and Supplementary Fig. S3b). So far, observations on microbial NOx emissions on the land showed a relatively lower flux of 7.9 ± 1.5 Tg N yr−1 (5.0–11.0 Tg N yr−1; data compiled from Refs [43,55,75-84]) than our estimate, because these observations have been conducted mainly on fertilized soils and merely on unfertilized soils and other land substrates. Besides, few modeling studies showed distinctly higher fluxes of land microbial NOx emissions ≤20.4 Tg N yr−1 [80] and 23.6 Tg N yr−1 [85] than the observation results and our estimate, due to overestimated N inputs in cropland and natural ecosystems and largely overlooked the influence of NOx sink uncertainties on the satellite-derived NOx fluxes. However, by considering more substrates of microbial N cycles on the land to optimize the modeling methods, some studies showed the land microbial NOx emissions as 11.5–13.6 Tg N yr−1 (12.4 ± 0.7 Tg N yr−1) [53,71,86,87], which is very comparable with our estimate. The isotopic method in our study offers a comprehensive and accurate constraining on microbial NOx emissions.

Total and microbial NOx emissions in the ocean

Based on statistical data of quantities and NOx emission factors of oil combustions in the ocean system, ocean fossil-fuel NOx emissions have been estimated as 6.4 ± 0.8 Tg N yr−1 on average (5.0–7.8 Tg N yr−1; compiled from [14-20]). Using the fraction of the ocean fossil-fuel NOx emissions in our study (42 ± 3%, Supplementary Fig. S3a), we estimated the total ocean NOx emissions as 15.2 ± 2.3 Tg N yr−1 (Fig. 3 and Supplementary Fig. S3b). The ocean NOy deposition averaged 21.3 ± 1.8 Tg N yr−1 (18.0–23.0 Tg N yr−1; compiled from Refs [35,61,62,88-90]), which includes the land-to-ocean NOy transportation of 11.0 Tg N yr−1 [61]. Accordingly, the oceanic NOy deposition derived from oceanic NOx emissions was 10.3 ± 1.8 Tg N yr−1, which is lower than our study's total ocean NOx emissions. The generally higher NOx emissions than NOy deposition in the ocean might be attributed to other fates such as biological NOx uptake and atmosphere retention. Further, we calculated ocean microbial NOx emissions as 8.8 ± 1.5 Tg N yr−1 on average (Fig. 3 and Supplementary Fig. S3b). Our results updated the total and microbial NOx emissions in the marine environment.

Total and microbial NOx emissions in the globe

By integrating the land and ocean values together (detailed in ‘Materials and methods’ section), we calculated global total NOx emissions as 72.0 ± 18.1 Tg N yr−1 (Fig. 3 and Supplementary Fig. S3b). Before this work, the modeled total land NOx emissions (39.7–61.4 Tg N yr−1; compiled from Refs [43,53-58]) have been assumed as the global NOx emissions because the ocean NOx emissions have been unconstrained. Our results showed that oceanic NOx emissions accounted for ∼21% of the global NOx emissions. The global NOx emissions have been underestimated by 15–45% because oceanic NOx emissions have been unconsidered.

Moreover, we found that microbial NOx emissions accounted for 31 ± 12% of the total NOx emissions globally and reached up to 22.5 ± 4.7 Tg N yr−1 (Fig. 3 and Supplementary Fig. S3b). By comparison, microbial NOx emissions in the land (13.6 ± 4.7 Tg N yr−1), ocean (8.8 ± 1.5 Tg N yr−1) and globe (22.5 ± 4.7 Tg N yr−1) are ∼0.5, 1.4 and 0.6 times fossil-fuel NOx emissions in the land, ocean and globe, respectively (Fig. 3 and Supplementary Fig. S3b). Our results highlight a vital role of the microbial N cycle in global NOx emissions. In addition to the direct impacts of fossil-fuel combustion on global NOx emissions, other human activities such as inefficient fertilizer use in cropping systems, wastes and sewage discharge and treatments, N deposition and water N enrichment all can accelerate microbial NOx emissions in the land, inland water bodies, estuaries and ocean [13,91].

Our results offer an updated and isotopically grounded estimate of land- and ocean-to-atmosphere NOx emissions. Notably, our results revealed that previous reports have largely underestimated land-based microbial NOx emissions, constrained long-missing uncertainties over ocean microbial NOx emissions and therefore elevated the recognition of the substantial contribution of the microbial N cycle to global NOx emissions. Moreover, our findings highlight the unique significance of natural records of atmospheric N isotopes for understanding global N biogeochemical cycles. Currently, reducing NOx emissions to alleviate N pollution while sustaining economic development is a major challenge in the twenty-first century. Owing partly to unclear contributions of microbial processes to NOx emissions, many countries have been engaging in developing technologies and measures for reducing fossil-fuel NOx emissions to reduce airborne and water N pollution, with a focus on adjusting energy systems and increasing the chemical conversion of NOx to reduce emissions during fossil-fuel combustion. Our findings point to the need to consider the substantial contribution of the microbial N cycle to atmospheric NOx loadings while reducing fossil-fuel NOx emissions. Accordingly, the potential costs and impacts of reducing fossil-fuel NOx emissions need to be re-assessed when making more effective emission mitigation strategies—including the indirect effects of anthropogenic N on terrestrial and marine microbial processes. Moreover, the isotopically constrained microbial NOx emissions and updated total NOx emissions we provide are helpful for benchmarking atmospheric and earth system models that project the feedback between the biosphere, climate and global N cycle.

In summary, based on large-scale isotope observations of p-NO3− in the atmosphere, we established a simple but effective approach for estimating NOx sources in the atmosphere. Before, isotope mass-balance models have been constructed to successfully partition continental hydrologic fluxes and quantify the contributions of local evaporation and ocean-to-land water transportation to the land moisture [92,93]. Accordingly, the framework established in our study enriches the application of isotopic mass-balance approaches in quantifying processes and fluxes of global biogeochemical cycles. However, our method can only consider dominant sources of NOx emissions. Additional work on detailed measurements of δ15N values for all NOx emission sources could further refine our estimates. Isotope observations of p-NO3− in the atmosphere across more sampling areas will be critical to reducing uncertainties in our estimation and offering spatial tools to pinpoint source regions of great concern.

MATERIALS AND METHODS

Detailed materials and methods are given in the online supplementary materials.

Click here for additional data file.