Nitrogen isotope differences between atmospheric nitrate and corresponding nitrogen oxides: A new constraint using oxygen isotopes

Science of the Total Environment 701 (2020) 134515

Contents lists available at ScienceDirect

Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv

Short communication

Nitrogen isotope differences between atmospheric nitrate and corresponding nitrogen oxides: A new constraint using oxygen isotopes Wei Song a, Xue-Yan Liu a,⇑, Yan-Li Wang b,c, Yin-Dong Tong d, Zhi-Peng Bai b, Cong-Qiang Liu a a

Institute of Surface-Earth System Science, Tianjin University, Tianjin 300072, China State Key Laboratory of Environmental Criteria and Risk Assessment, Chinese Research Academy of Environmental Sciences, Beijing 100012, China c Chinese Academy for Environmental Planning, Beijing 100012, China d College of Environmental Science and Engineering, Tianjin University, Tianjin 300072, China b

h i g h l i g h t s

g r a p h i c a l a b s t r a c t


N isotope differences (e) between

NO3 and NOx were evaluated by 17O signals. 17
The O-based e values were lower than those evaluated by 18O signals.
Ameliorated e values improved isotope analysis of atmospheric NO3 sources.

a r t i c l e

i n f o

Article history: Received 28 July 2019 Received in revised form 3 September 2019 Accepted 16 September 2019

Editor: Xinbin Feng Keywords: Nitrate aerosol Source apportionment Oxygen isotopes Nitrogen isotopes Isotopic fractionation

⇑ Corresponding author. E-mail address: [email protected] (X.-Y. Liu). https://doi.org/10.1016/j.scitotenv.2019.134515 0048-9697/Ó 2019 Elsevier B.V. All rights reserved.

a b s t r a c t Tracking of reactive nitrogen (N) sources is important for the effective mitigation of N emissions. By combining the N and oxygen (O) isotopes of atmospheric NO 3 , stable isotope mixing models were recently applied to evaluate the relative contributions of major NOx sources. However, it has long been unresolved how to accurately constrain the d15N differences between NO 3 and corresponding NOx (e(NO2?NO3) val17 ues). Here, we first incorporated the HC oxidation (NO2 ? NO 3 ) pathway by using D O values to evaluate  the e(NO2?NO3) values, performed on NO3 in PM2.5 collected during the day and at night from January 4–13, 2015 at an urban site in Beijing. We found that the D17O-based e values (e17O-based(NO2?NO3)) (15.6 ± 7.4‰) differed distinctly from d18O-based e values (e18O-based(NO2?NO3)) (33.0 ± 9.5‰) so did not properly incorporate the isotopic effects of the HC oxidation (NO2 ? NO 3 ) pathway. Based on the e(NO2?NO3) values, d15N values of NOx from coal combustion (CC), vehicle exhausts (VE), biomass burning (BB), and the microbial N cycle (MC), as well as NO 3 in PM2.5, we further quantified the source contributions by using Stable Isotope Analysis in R (the SIAR model). We found that the respective fractional contributions of CC-NOx and MC-NOx were underestimated by 64% and were overestimated by 216% by using e18O-based(NO2?NO3) values. We concluded that the new e17O-based(NO2?NO3) values reduced uncertainties in contribution analysis and the evaluation method for atmospheric NO 3 sources. Ó 2019 Elsevier B.V. All rights reserved.

2

W. Song et al. / Science of the Total Environment 701 (2020) 134515

1. Introduction Nitrate (NO 3 ) is a key component of atmospheric particulates and of reactive N in both dry and wet deposition (Liu et al., 2013; Huang et al., 2014). The enhanced deposition of anthropogenic NO 3 in terrestrial and aquatic ecosystems (Liu et al., 2013) has caused many negative effects, such as acidification, declining biodiversity and water quality (Clark & Tilman 2008; Guo et al., 2010). Atmospheric NO 3 is mainly derived from nitrogen oxides (NOx) (nitrogen dioxide (NO2) and nitric oxide (NO)) from fossil fuel combustions, biomass burning, microbial N cycles in agricultural soils, and human wastes (Richter et al., 2005; Huang et al., 2014). To complement effective NOx emission mitigation, it is important to differentiate the contributions of the above major NOx sources to atmospheric NO 3 deposition. Natural N isotopes (expressed as d15N values) have been recognized as a useful tool to trace atmospheric NO 3 sources (Kendall et al., 2007; Morin et al., 2008; Liu et al., 2018). Since the 1950s, there have been many measurements of the d15N signatures of major NOx emission sources (d15Nsource-NOx) (Table S1) and several isotopic studies facilitated identification of major NO 3 sources based on d15N similarity between the deposited NO 3 and the major NOx sources (Yeatman et al., 2001; Elliott et al., 2009; Proemse et al., 2012; Kamezaki et al., 2019). Recently, isotope mixing models (e.g., Stable Isotope Analysis in R (SIAR) and IsoSource) have been explored to quantitatively evaluate the relative contributions of major NOx sources to atmospheric NO 3 (Liu et al., 2017; Zong et al., 2017; Chang et al., 2018; He et al., 2018). However, the d15N values of atmospheric NO 3 (expressed as 15 d NNO3 values) integrate both the d15N values of atmospheric NOx sources and nitrogen isotopic effects (e(NO2?NO3)) during the conversion of NOx to NO 3 (Freyer 1978; Freyer 1991; Freyer et al., 1993; Walters and Michalski 2015, 2016; Liu et al., 2017; Zong et al., 2017). In other words, corresponding d15N values of 15 the NOx that has been converted to NO 3 (d NNOx) should be estimated by considering the e values with respect to the observed d15NNO3 values (Eq. (1)).

d15 NNOx = d15 NNO3 -eðNO2!NO3Þ

eðNO2!NO3Þ ¼ eP1ðNO2!NO3Þ  f P1 þ eP2ðNO2!NO3Þ  f P2 þ eP3ðNO2!NO3Þ  f P3

ð2Þ

where fP1 + fP2 + fP3 = 1. Clearly, it is very important to accurately estimate the process-based f and e(NO2?NO3) values. So far, estimates have been derived by Walters and Michalski (2015, 2016) based on formulae (detailed in Section 2.3). A few studies have calculated the eP1(NO2?NO3), fP1, eP3(NO2?NO3), fP3 values using the d15N and d18O of atmospheric NO 3 , but with no consideration of the eP2(NO2?NO3) because the fP2 values could not be calculated from the d18O values (Zong et al., 2017; Chang et al., 2018). In previous studies the e18O-based(NO2?NO3) values were calibrated by multiplying by a constant in previous studies (Zong et al., 2017; Chang et al., 2018). According to Zong et al (2017), the e18O-based value after multiplying 0.52 is a method to evaluate the isotope effect, although proper physical and/or chemical meanings of this value remain unclear. Because the e18O-based values after multiplying 0.52 has also been adopted in few studies (Chang et al., 2018; Song et al., 2019), it was also considered in our study and compared with our new method based on D17O values. It remains highly uncertain whether the actual e(NO2?NO3) values are different when the HC oxidation pathway was considered, i.e., the eP2(NO2?NO3) and fP2 values were directly calculated. It would also be highly useful to verify whether more accurately constrained e(NO2?NO3) values would influence the final estimated results of source apportionment. To address these issues, we calculated the fP2 values using the D17O values of atmospheric NO 3 and incorporated both the eP2 (NO2?NO3) and fP2 values into the calculations of e(NO2?NO3) values (Eq. (2)). We evaluated the differences between e17O-based(NO2? NO3) and e18O-based(NO2?NO3) values, as well as corresponding estimated results of major NO 3 sources in PM2.5. These investigations were performed on NO 3 in PM2.5 collected during the day and at night from January 4–13, 2015 at an urban site in Beijing. The main objective of this work is to constrain the e(NO2?NO3) values more accurately and to reduce estimated uncertainties in contributions of major NOx sources to NO 3 in atmospheric particulates.

ð1Þ

However, how to constrain the e(NO2?NO3) value has not been well resolved; it has even been ignored in some studies (Liu et al., 2017). The e(NO2?NO3) value is determined by proportional contributions (fP1, fP2, and fP3, respectively) and isotopic effects (eP1(NO2?NO3), eP2(NO2?NO3), and eP3(NO2?NO3), respectively) of the three main pathways of NOx oxidation to NO 3 (Eq. (2)), including the OH pathway (NO2 + OH ? HNO3(g); P1 hereafter) and O3 pathways (including the HC pathway (NO3 + HC ? HNO3(g); P2 hereafter) and the N2O5 (N2O5 + H2O(surface) ? 2NO 3 ; P3 hereafter) pathway) (Alexander et al., 2009; Alexander & Mickley 2015).

P1 :NO2 + OH ! HNO3ðgÞ

ðR1Þ

P2 :NO2 + O3 ! NO3 + O2

ðR2Þ

NO3 + HC ! HNO3ðgÞ

ðR3Þ

P3 :NO3 + NO2 $ N2 O5

ðR4Þ

N2 O5 + H2 OðsurfaceÞ ! 2HNO3

ðR5Þ

The P2 pathway indicate the HC/DMS pathway according to Alexander et al (2009). However, the DMS pathway occurs mainly in the marine atmosphere because of little DMS in the urban atmosphere. The HC is much more important than DMS in the urban atmosphere for nitrate formation, thus the P2 pathway was expressed as NO3 + HC ? HNO3(g) in our study.

2. Materials and methods 2.1. Study site, sampling, and chemical analyses The sampling site is located in the courtyard of the CRAES site (Chinese Research Academy of Environmental Sciences) (40°040 N, 116°420 E) in the Chaoyang District of Beijing (detailed information of the sampling site has been described in Song et al. 2019). PM2.5 samples were collected every 12 h (n = 20) using a particulate sampler (Leckel, MVS6, Germany) equipped with a sizesegregating impactor with a flow rate of 38.3 L min1, which the size cut-point of this impactor is less than 2.5 lm. In parallel with the PM2.5 sampling, concentrations of NO2, CO and O3 were also measured.  2+ 2+ 2 + Major anions (NO 3 , SO4 , and Cl ) and cations (NH4, Ca , Mg , K+, and Na+) in extracts from PM2.5 samples were measured by ion chromatography (ICS5000, Dionex, USA). The NO 2 has been measured together with the NO 3 by the Ion Chromatography. The NO 2 accounted for 0–12% (mean = 1.2 ± 1.6%) in the total pool of   NO 2 plus NO3 in PM2.5 (Fig. S1). Thus, the NO2 may not substantially influence the overall d15N pattern and source analyses based  on the d15N values of NO 2 and NO3 mixture. Besides, the chemical removal of NO may trigger substantial NO 2 3 contamination from chemicals to samples, we did not conduct this step before the isotope analysis. Nitrogen and oxygen isotopes of NO 3 were analyzed using the bacterial denitrifier method (Kaiser et al., 2007) in the stable isotope laboratory at the University of Washington (UW).

3

W. Song et al. / Science of the Total Environment 701 (2020) 134515

Briefly, the NO 3 in the solutions was converted into N2O gas by the bacteria Pseudomonas aureofaciens, and the N2O was carried on-line by helium gas into a heated gold tube where it was thermally decomposed to N2 and O2. The N2 and O2 were then separated by gas chromatography, and the isotopic ratios of N2 (15N/14N) gas and O2 gas (18O/16O and 17O/16O) were measured on an isotope ratio mass spectrometer (Finnigan Delta Plus Advantage, Thermo Fischer Scientific, Waltham, MA). The standard deviations of replicate analyses of an individual sample were ±0.2‰, ±0.5‰ and ±0.1‰ for d15N, d18O and D17O (D17O = d17O  0.52  d18O), respec15 18 tively. Nitrogen and O isotopes of NO 3 (d N and d O values) were expressed as: d = (Rsample/Rstandard  1)  1000, where R denotes the 15 N/14N, 17O/16O or 18O/16O values in the sample (Rsample) and in the standards (Rstandard, atmospheric N2 for d15N; Vienna Standard Mean Ocean Water (VSMOW) for d18O). The statistical analysis was performed by SPSS version 16 for Windows (SPSS Inc., Chicago, IL, USA). Data for concentration, isotope, and contribution data were expressed as mean ± standard deviation (SD). A p-value < 0.05 was considered to indicate significant differences in mean data values, unless otherwise specified. 2.2. Calculating the fractional contributions of P1, P2 and P3 to NO 3 The fractional contributions of P1, P2, and P3 (fP1, fP2, fP3; %) to NO 3 were calculated using Eq. (3).

h

i

h

D17 O  NO3 ¼ D17 O  NO3  f P1 þ D17 O  NO3 P1 h i þ D17 O  NO3  f P3

i

P2

 f P2

P3

ð3Þ

where fP1 + fP2 + fP3 = 1, and D17O-NO 3 is the observed value in our study.   17 17 The [D17O-NO 3 ]P1, [D O-NO3 ]P2, and [D O-NO3 ]P3 endmember values were estimated by Eqs. (4)–(6) (Alexander et al., 2009).

h h h

D17 O  NO3 D17 O  NO3 D17 O  NO3

i P1

i P2

i

ð‰Þ ¼ 2=3A  D17 O  O3

ð4Þ

ð‰Þ ¼ 2=3A  D17 O  O3 þ 1=3D17 O  O3

ð5Þ

Canada (Savard et al., 2017), 5.3 ± 2.6‰ in Yurihonjo of Japan (Kawashima 2014), 3.0 ± 3.5‰ in Rishiri of Japan (Nelson et al., 2018), and 3.5 ± 4.5‰ in Sapporo of Japan than corresponding d15N-HNO3(g) values, showing a total difference (etotal (HNO3MNO3)) of 5.1 ± 4.1‰. Therefore, the etotal(HNO3MNO3) value represents the d15N difference between total NO 3 produced by all three pathways and HNO3(g) that contributed only by P1 and P2. The real d15N differences between NO 3 and HNO3(g) for P1 and P2 (ereal(HNO3MNO3)) can be estimated by Eq. (7).

erealðHNO3$NO3Þ ¼ etotalðHNO3$NO3Þ =ðf P1 þ f P2 Þ

ð7Þ

and the respective ereal(HNO3MNO3) values for P1 and P2 can be calculated by Eqs. (8) and (9).

eP1realðHNO3$NO3Þ ¼ erealðHNO3$NO3Þ  f P1 =ðf P1 þ f P2 Þ

ð8Þ

eP2realðHNO3$NO3Þ ¼ erealðHNO3$NO3Þ  f P2 =ðf P1 þ f P2 Þ

ð9Þ

Then, the overall e(NO2?NO3) values for P1 and P2 can be calculated by Eqs. (10) and (11).

eP1ðNO2!NO3Þ ¼ eP1ðNO2!HNO3Þ þ eP1realðHNO3$NO3Þ

ð10Þ

eP2ðNO2!NO3Þ ¼ eP2ðNO2!HNO3Þ þ eP2realðHNO3$NO3Þ

ð11Þ

Finally, the e(NO2?NO3) value for all three pathways of atmospheric NO2 conversion to NO 3 (p) can be calculated by Eq. (2). The eP1(NO2?NO3), eP2(NO2?NO3), and eP3(NO2?NO3) values are the isotopic fractionations of P1, P2, and P3, respectively. According to Walters and Michalski (2016), d15N values of NO2 and HNO3 were estimated as 25.6 ± 10.6‰ and 7.2 ± 10.7‰, respectively, by which the isotopic fractionation of P2 (eP2(NO2?NO3) value) can be estimated as 18.4 ± 14.3‰. So far, this allowed us to consider the uncertainty of the eP2(NO2?NO3) value into the modeling results. The eP1(NO2?HNO3) values were calculated using the following mass-balance equation (Walters and Michalski, 2016).

eP1ðNO2!HNO3Þ ¼ 1000 







aNO2=NO  1 ð1  f NO2 Þ =   15  ð1  f NO2 Þ þ aNO2=NO  f NO2 15

ð12Þ

15

ð‰Þ ¼ 1=3A  D17 O  O3 P3   þ 1=2 2=3A  D17 O  O3 þ 1=3D17 O  O3 ð6Þ

The A value is the proportional contribution of O3 among the NO oxidation values. We have explained how A, fP1, fP2, fP3, [D17O17 17   NO 3 ]P1, [D O-NO3 ]P2, and [D O-NO3 ]P3 were estimated in great details in the main text of the other paper (Wang et al., 2019). The method and steps are all the same in this study. So, the above explicit procedures of all calculations were provided in the Supporting Information (Text S1). 2.3. Calculation methods for D17O-based e(NO2?NO3) values Atmospheric NO2 is first converted to HNO3 through three dominant reactions (i.e., reactions of NO2 with OH, NO3 with HC, and N2O5 with H2O), and HNO3 is then converted to NO 3 (p) (Alexander et al., 2009). The product of the N2O5 pathway (P3) is NO 3 , but the products of both P1 and P2 are first gaseous nitric acid 15 15 (HNO3), then NO 3 . The d N signatures of gaseous nitric acid (d N15  HNO3(g)) did differ from those of particulates NO (d N-NO ) 3 3 that were measured during the same time at the same sites. Based on the available data, d15N-NO 3 values were higher by 9.8 ± 3.5‰ in Jülich of Cermany (Freyer 1991), 2.7 ± 1.2‰ in New York, Pennsylvania, and Ohio of USA (Elliott et al., 2009), 6.2 ± 5.6‰ in Alberta of

where the aNO2/NO value is the equilibrium isotope fractionation factor between NO2 and NO, which is temperature (T) dependent (see Eq. (14)), and fNO2 is the fraction of NO2 in total NOx. The fNO2 values were 0.34 ± 0.27 in winter in Beijing (Liu and Zhu, 2013; Hu et al., 2016), which integrated the day/night variations of photochemistry. The standard deviations of fNO2 were estimated by the Monte Carlo method and were finally propagated into the uncertainties of the e(NO2?HNO3) values. Similarly, the eP3 values can be calculated from the following equation (Walters & Michalski 2016).

eP3ðNO2!HNO3Þ ¼ 1000 



15



aN2O5=NO2  1 ;

ð13Þ

in which 15aN2O5/NO2 refers to the equilibrium isotope effects between N2O5 and NO2, which are T-dependent (see Eq. (14)). In this study, the temperature (T) is shown in Fig. S1. The 15aNO2/NO and 15aN2O5/NO2 in Eqs. (12) and (13) (expressed the 15aX/Y) were calculated by Eq. (14):

 1000

15



aX=Y  1 ¼ A=T4  1010 þ B=T3  108 þ C=T2  106 þ D=T  104

ð14Þ

where A = 3.847, B = 7.680, C = 6.003, and D = 0.118 for 15aNO2/ 15 aN2O5/ NO; and A = 1.004, B = 2.525, C = 2.718, and D = 0.135 for NO2 (Walters & Michalski 2015). The standard deviations of each variable in the above calculations were estimated by the Monte Carlo method and were finally propagated into the uncertainties of the e(NO2?NO3) values.

4

W. Song et al. / Science of the Total Environment 701 (2020) 134515

2.4. Estimating proportional contributions of major NOx sources to NO 3 in PM2.5 Corresponding d15N values of the NOx that has been converted 15 to NO 3 (d NNOx) should be estimated by considering the e values with respect to the observed d15NNO3 values (Eq. (15)).

d15 NNOx ¼ d15 NNO3  eðNO2!NO3Þ ¼ f CC  d15 N  NOx CC þ f VC  d15 N  NOxVC þ f BB  d15 N  NOxBB þ f MC  d15 N  NOxMC ;

ð15Þ

where fCC + fVC + fBB + fMC = 1, fCC, fVC, fBB, and fMC were the fractional contributions of NOx from coal combustion, vehicle exhausts, biomass burning and microbial N cycle to NO 3 of PM2.5, respectively. d15N-NOx CC, d15N-NOx VC, d15N-NOx BB, and d15N-NOx MC were the nitrogen isotope values of NOx from coal combustion, vehicle exhausts, biomass burning and microbial N cycle, respectively. d15N values of major NOx sources overlapped, though differences did exist between the available values (Table S1). So far, d15N values of different NOx sources in China were not available. We used the Stable Isotope Analysis in R (i.e., the SIAR model) to run the d15NNO 3 data in different time periods, which can evaluate the standard deviation for corresponding calculations. Particularly, the SIAR model allows us to input the mean ± SD value of isotope fractionation, mean ± SD values of source d15N signatures, and also the scattered PM2.5 d15N-NO 3 values in a given time period (Moore and Semmens, 2008; Davis et al., 2015). The above detail calculations were provided in the Supporting Information (Text S3).

3. Results and discussion 3.1. Isotopes of NO 3 in PM2.5 During the study period, concentrations of PM2.5 and chemical  + + 2+ 2+ + 2 components (NO 3 , NH4, SO4 , Cl , Na , K , Mg , and Ca ) did not differ significantly between day and night samples (Figs. 1 & S3). However, the d15N-NO 3 values of day samples (17.8 ± 2.7‰) were significantly higher than those of night samples 18  (14.7 ± 3.5‰) (Fig. 1). The d18O values of PM2.5 NO 3 (d O-NO3 values) were between those reported for O3 (90.0–120.0‰) and OH (20.0–0.0‰) (Dubey et al., 1997; Johnston & Thiemens 1997). d18ONO 3 values in this study (88.3 ± 6.9‰) were similar to that measured in winter on BH Island of the Bohai Sea (88.1 ± 10.1‰) (Zong et al., 2017), higher than those on Yongxing Island (83.2‰) (Xiao et al., 2015) and Dongshan Island (78.8‰) of southern China sea (Yang et al., 2014), Sanjiang Plain of northeastern China (63.6‰) (Chang et al., 2018), and Mt. Lulin (48 ± 22‰) in Taiwan 17  (Guha et al., 2017). The D17O values of PM2.5 NO 3 (D O-NO3 values) in this study (27.8 ± 2.1‰) resembled those measured in the winter of 2014 (30.6 ± 1.8‰, 27.5‰ to 33.9‰; He et al., 2018) in Beijing and on Sado Island of Japan (26.3‰, 18.6‰ to 32.4‰) (Tsunogai et al., 2016), were higher than that in Mt. Lulin (17.0 ± 7.0‰, 2.7‰ to 31.4‰) in Taiwan (Guha et al., 2017). Theoretically, both d18O and D17O values would be lower for NO 3 produced during the day (relatively more produced by the OH oxidation pathway) than those produced at night (relatively lesser produced by the OH oxidation pathway) (Alexander et al., 2009; 17  Nelson et al., 2018). However, neither d18O-NO 3 nor D O-NO3 val18 ues differed between day (88.3 ± 5.0‰ for d O and 28.0 ± 1.4‰ for D17O) and night samples (88.2 ± 8.5‰ for d18O and 28.0 ± 2.6‰ for D17O) (Fig. 1). Accordingly, our results suggested that both NO 3 produced in the day time and in the night time were well mixed and accumulated over the day and night time scales, which cannot be differentiated by using the PM2.5 samples collected every 12 h.

3.2. Fractional contributions of P1, P2 and P3 to NO 3 In our study, the values of fP1, fP2 and fP3 are 31 ± 11%, 34 ± 11% and 35 ± 20% for the day-time samples, and 28 ± 12%, 37 ± 12% and 35 ± 18% for the night-time samples (Fig. 2 & Table S2). The fP2 value was much higher than the recent GEOS-Chem modeling result (Alexander et al., 2019). We calculated fP1, fP2 and fP3 values by using D 17O-NO 3 values of the samples and the three major pathways. Different from our method, the GEOS-Chem model calculated the fractional contributions of each formation pathways first, and then calculated the D 17O-NO 3 value of each formation pathway. Using the fractional contributions and D 17O-NO 3 values of each formation pathways, they estimated the D 17O-NO 3 values in the real atmosphere. According to the study of Alexander et al. (2019), D 17ONO 3 values averaged about 30‰ in Beijing (Fig. 4 of Alexander et al., 2019), which are much higher than that of measured values (26.3‰) in Beijing in our previous study (Wang et al., 2019) and also higher than the mean value of this study (27.8‰ in January of winter). Most possibly, the GEOS-Chem model might overestimate the D 17 O-NO 3 values, at least for the case in Beijing. The fractional contributions of different pathways from the GEOS-Chem model may be applicable for regional or global scales, but may not be accurate or applicable when comparing with measured values at a specific site. The NO3 radical readily reacts with the volatile organic compounds (VOCs) (i.e., P2), as that with the NO2 (i.e., P3) (Geyer et al., 2001). Geyer et al (2001) found at Pabstthum near Berlin, German that about 30–50% of the NO3 radical was lost due to reaction with hydrocarbons, while the major part of the remaining loss probably can be attributed to the indirect loss via the reaction of N2O5 on aerosol surface. Womack et al (2019) found that the VOCs have a very effective control on the initial reduction of NH4NO3 aerosols in Salt Lake City. In our study, the contributions of P2 to atmospheric NO 3 formation were estimated as 34 ± 11% in daytime samples and 37 ± 12% in night-time samples during the sampling period (Jan-2015), which is very likely to occur in the winter time of Beijing because of coexisting high VOCs. For example, during the Jan-2012, the mean concentration of VOCs reached about 60 ppbv in Beijing, with the highest concentration of up to 200 ppbv (Wang et al., 2013), which were distinctly higher than those (ca. < 30 ppbv, with the maximum of 50 ppbv) measured in tropical forests (Kesselmeier et al., 2000, 2002; Yánez-Serrano et al., 2015). In Beijing, vehicular emissions and coal combustion have been identified as the main sources of high atmospheric VOCs concentrations, which can even occur in some time of the summer (Song et al., 2007; Wang et al., 2013). 3.3. Differences between d18O- and D17O-based e values The values of ereal(HNO3MNO3) are 7.9 ± 6.5‰ for the day-time samples and 7.9 ± 6.8‰ for the night-time samples, respectively. eP1-real(HNO3MNO3) values are 3.8 ± 3.6‰ for the day-time samples and 3.4 ± 3.7‰ for the night-time samples, while eP2-real (HNO3MNO3) values are 4.1 ± 3.7‰ for the day-time samples and 4.5 ± 4.3‰ for the night-time samples, respectively. Values of eP1 15 (NO2?NO3), eP2(NO2?NO3), eP3(NO2?NO3), and d N-NOx in daytime and night-time were presented in Table S2. The e values can be calculated by using the d18O-NO 3 values that consider only P1 and P3 (hereafter, e18O-based(NO2?NO3)), and then calibrated by multiplying by a constant of 0.52 (Zong et al., 2017; Chang et al., 2018) to obtain the highest probability distributions of source contributions (hereafter, e18O-based (NO2?NO3)  0.52). For comparison, both e18O-based and e18O-based(NO2? NO3)  0.52 values were also calculated for all of the samples in this study (detailed in Text S2 of the Supporting Information). As shown in Section 2.3 (main text), the calculation of e(NO2?NO3) values based on the D17O values can consider P1, P2, and P3 (hereafter,

W. Song et al. / Science of the Total Environment 701 (2020) 134515

5

Fig. 1. Concentrations and stable isotopes of NO 3 in PM2.5 collected during the day and at night from January 4–13, 2015 at an urban site in Beijing. Dots around each box show scatter values. The box encompasses the 25th–75th percentiles, whiskers and line in each box are the SD and mean values, respectively. Different letters above the boxes indicate significant differences at the p < 0.05 level.

Fig. 2. Fractional contributions of the OH pathway (P1), the HC pathway (P2), and the N2O5 pathway (P3) to NO 3 in PM2.5 collected during the day and at night from January 4–13, 2015 at an urban site in Beijing. Mean ± SD values are shown.

Fig. 3. d15N differences between NOx and NO 3 (e(NO2?NO3) values) estimated for PM2.5 collected during the day and at night from January 4–13, 2015 at an urban site in Beijing. Mean ± SD values are shown.

e17O-based(NO2?NO3)), which has not been performed before this study. We found that for both day and night samples, the e17O-

respectively). The e17O-based(NO2?NO3) values were still lower than corresponding e18O-based(NO2?NO3)  0.52 values (16.8 ± 6.3‰ for day- and 17.4 ± 7.1‰ for night samples) (Fig. 3). These results clearly demonstrated that the e values would be overestimated if P2 is not included.

based(NO2?NO3) values (15.9 ± 10.4‰ and 15.2 ± 10.8‰, respectively) (Fig. 2 & Table S2) were significantly lower than corresponding e18O-based(NO2?NO3) values (32.4 ± 12.0‰ and 33.5 ± 13.8‰,

6

W. Song et al. / Science of the Total Environment 701 (2020) 134515

3.4. Differences between source analyses by e18O-based and e17O-based values The main sources of NOx in the Beijing area have been verified as coal combustion, vehicle exhausts, biomass burning, and the microbial N cycle (Zhao et al., 2012). Scattered d15N values of major NOx sources measured at different sites overlapped to some extents (Elliott et al., 2019), but significant differences did exist between mean values of major sources (Table S1). So far, d15N values of most NOx sources in China were not available, extant d15N data of few NOx sources in China showed very similar values or ranges with corresponding values reported at other sites worldwide (Wang et al. 2016). d15N values of different biomass materials in Beijing are not available, we were unable to estimate accurate d15N-NOx values from biomass burnings in Beijing. According to Fibiger and Hastings (2016), d15N-NOx may differ among open burning, residential heating, cooking, etc, we will conduct related experiments in near future to accuracy of source contribution analyses, but this will not influence the conclusion of the present work focusing on the methodology of constraining isotope fractiona15 tions. By using the d15N-NO 3 values, d N end-member values of major NOx sources, and nitrogen isotopic effects (e(NO2?NO3)), the proportional contributions of major NOx sources to the PM2.5 NO 3 were analyzed using Stable Isotope Analysis in R (i.e., the SIAR model, the calculation method is detailed in Text S3 of the Supporting Information). Differences in source contributions were also compared between evaluations using the e18O-based(NO2?NO3), e18O-based (NO2?NO3)  0.52, and e17O-based(NO2?NO3) values (Fig. 4).

When based on e17O-based(NO2?NO3) values, NOx from coal combustion, vehicle exhausts, biomass burning and microbial N cycle contributed 34 ± 9%, 25 ± 9%, 29 ± 10%, and 12 ± 5%, respectively, to the NO 3 of day PM2.5 samples, and 30 ± 8%, 25 ± 10%, 28 ± 10%, and 17 ± 6%, respectively, to the NO 3 of night PM2.5 samples (Fig. 4). According to analytical results, higher contributions of NOx from coal combustion (Fig. 4) and higher isotope effects (Fig. 3) caused higher d15N-NO 3 values of daytime samples than nighttime samples (Fig. 1). The source proportional contributions of NOx from biomass burning in our study were higher than that of NOx emissions based on regional emission inventory data (Zhao et al., 2012). The Back-trajectory analysis suggested that the air mass during the sampling period were mainly from the north and northwest directions of the sampling site, with very few industries and vehicles and spanning areas dominated by grassland, forests and farmland (Fig. S8). In the winter of Beijing, the low temperature and weak microbial activities lead to less contributions of NOx from microbial N cycle along the air mass to the NO 3 of PM2.5. Besides, biomass (agriculture residues, grass, wood and so on) burning is also used for residential heating and cooking in the winter of rural areas in northern China (Streets et al., 2003). So contributions of NOx from biomass burning to PM2.5 nitrate can be substantial in the winter of study area. Compared with the estimated results of source contributions based on e17O-based(NO2?NO3) values (Fig. 4), we found that fractional contributions of NOx from coal combustion were underestimated by 64% and 6% and those of NOx from the microbial N cycle were overestimated by 216% and 12%, respectively, in evaluations based on e18O-based(NO2?NO3) and e18O-based(NO2?NO3)  0.52 values

Fig. 4. Fractional contributions of major NOx sources (a: coal combustion, b: vehicle exhausts, c: biomass burning, d: microbial N cycle) estimated by using different e values (e17O-based(NO2?NO3), e18O-based(NO2?NO3), and e18O-based(NO2?NO3)  0.52) for NO 3 in PM2.5 collected during the day and at night from January 4–13, 2015 at an urban site in Beijing. Mean ± SD values are shown.

W. Song et al. / Science of the Total Environment 701 (2020) 134515

(Fig. 4). Values of isotopic fractionations are very important parameters to accurately calculate fractional contributions of NOx sources in the SIAR model. The sensitivity of source contributions to isotopic fractionations has been revealed in previous isotope studies (Chang et al. 2018), that is also why we focused on developing new method to constrain isotope fractionations in this work. Clearly, differences between d18O- and D17O-based e(NO2?NO3) values would cause substantial differences in apportionment of major NO 3 sources in PM2.5. 4. Remarks This work focused on constraining N isotope differences between atmospheric NO 3 and corresponding NOx, which is critical in conducting source contribution analysis based on isotope mixing models. We used D17O records to constrain processspecific contributions to NO 3 production and for the first time incorporate the explicit fractional contributions of the HC oxidation (NO2 ? NO 3 ) pathway into the calculations of e(NO2?NO3) values, which could not be achieved previously by using d18O values. New D17O-based e(NO2?NO3) values were distinctly lower than d18O-based e(NO2?NO3) values, which caused substantial differences in corresponding results of relative contributions of major NOx sources to the NO 3 in PM2.5. The amelioration of e(NO2?NO3) values in this work contributed to the methodology of differentiating major sources of atmospheric NO 3 based on dual N and O isotopes. Although this study represents a new exploration on constraining the e(NO2?NO3) values, substantial uncertainties exist in the whole estimation. First, regarding to relative contributions (for sources or pathways), the 10,000 possible solutions output from the SIAR model contain substantial uncertainties, although their mean and SD values were used and considered into subsequent calculations. For example, high fP2 values (34 ± 11%) estimated for atmospheric NO 3 production in Beijing cannot be quantitatively verified, though high HC compounds potentially contributed to the reduction of NO3∙ in Beijing city. Second, end-member values 15 of D17O in NO 3 produced by different pathways and d N in different NOx emission sources were measured directly in this work but cited from literatures. Third, some parameters (including etotal (HNO3MNO3) in Eq. (7), fNO2 in Eq. (12), and A, B, C, and D in Eq. (14)) used for calculating e values of each pathway were not measured but cited from literatures. We concluded that future studies should reduce and resolve the above uncertainties following this work, which is essential for improving isotope tracing methods of atmospheric NO 3 sources. Declaration of Competing Interest The authors declare no competing financial interests. Acknowledgements This study was supported by the State Key Project of Research and Development Plan (2017YFC0210101), the National Natural Science Foundation of China (Nos. 41730855, 41603007, 41522301), the Outstanding Youth Funds of Tianjin (No. 17JCJQJC45400), the Coordinated Research Project of IAEA (F32008), the 11st Recruitment Program of Global Experts (the Thousand Talents Plan) for Young Professionals granted by the central budget of China, the State Environmental Protection Commonweal Trade Scientific Research, Ministry of Environmental Protection of China (No. 201309010), and the foundation for Innovation team training in Higher Education of Tianjin (TD 12-5037). We thank Andrew J. Schauer, Zhongyi Zhang, Nengjian Zheng for

7

their helps and assistances on field and laboratory works. We appreciate Dr. Robert Mark Ellam in SUERC for helping English corrections. Appendix A. Supplementary data Supplementary data to this article can be found online at https://doi.org/10.1016/j.scitotenv.2019.134515. References Alexander, B., Mickley, L.J., 2015. Paleo-perspectives on potential future changes in the oxidative capacity of the atmosphere due to climate change and anthropogenic emissions. Curr. Pollut. Res. 1, 57–69. Alexander, B., Hastings, M.G., Allman, D.J., Dachs, J., Thornton, J.A., Kunasek, S.A., 2009. Quantifying atmospheric nitrate formation pathways based on a global model of the oxygen isotopic composition (D17O) of atmospheric nitrate. Atmos. Chem. Phys. 9, 5043–5056. Alexander, B., Sherwen, T., Holmes, C.D., Fisher, J.A., Chen, Q., Evans, M.J., Kasibhatla, P., 2019. Global inorganic nitrate production mechanisms: comparison of a global model with nitrate isotope observations. Atmos. Chem. Phys. Discuss. https://doi.org/10.5194/acp-2019-422. Chang, Y., Zhang, Y., Tian, C., Zhang, S., Ma, X., Cao, F., Liu, X., Zhang, W., Kuhn, T., Lehmann, M.F., 2018. Nitrogen isotope fractionation during gas-to-particle conversion of NOx to NO 3 in the atmosphere – implications for isotope-based NOx source apportionment. Atmos. Chem. Phys. 18, 11647–11661. Clark, C.M., Tilman, D., 2008. Loss of plant species after chronic low-level nitrogen deposition to prairie grasslands. Nature 451, 712–715. Davis, P., Syme, J., Heikoop, J., Fessenden-Rahn, J., Perkins, G., Newman, B., 2015. Quantifying uncertainty in stable isotope mixing models. J. Geophys. Res.: Biogeo. 120 (5), 903–923. Dubey, M.K., Mohrschladt, R., Donahue, N.M., Anderson, J.G., 1997. Isotope specific kinetics of hydroxyl radical (OH) with water (H2O): testing models of reactivity and atmospheric fractionation. J. Phys. Chem. 101, 1494–1500. Elliott, E.M., Kendall, C., Boyer, E.W., Burns, D.A., Lear, G.G., Golden, H.E., Harlin, K., Bytnerowicz, A., Butler, T.J., Glatz, R., 2009. Dual nitrate isotopes in dry deposition: utility for partitioning NOx source contributions to landscape nitrogen deposition. J. Geophys. Res. 114, G04020. Elliott, E.M., Yu, Z., Cole, A.S., Coughlin, J.G., 2019. Isotopic advances in understanding reactive nitrogen deposition and atmospheric processing. Sci. Total Environ. 662, 393–403. Fibiger, D.L., Hastings, M.G., 2016. First measurements of the nitrogen isotopic composition of nox from biomass burning. Environ. Sci. Technol. 50, 11569– 11574. Freyer, H.D., 1978. Seasonal trends of NH4+ and NO3 nitrogen isotope composition in rain collected at Jülich, Germany. Tellus B 30 (1), 83–92. Freyer, H.D., 1991. Seasonal variation of 15N/14N ratios in atmospheric nitrate species. Tellus B 43 (1), 30–44. Freyer, H.D., Kley, D., Volz-Thomas, A., Kobel, K., 1993. On the interaction of isotopic exchange processes with photochemical reactions in atmospheric oxides of nitrogen. J. Geophys. Res.: Atmos. 98 (D8), 14791–14796. Geyer, A., Alicke, Björn, Konrad, S., Schmitz, T., Stutz, J., Platt, U., 2001. Chemistry and oxidation capacity of the nitrate radical in the continental boundary layer near Berlin. J. Geophys. Res.: Atmos. 106, 8013–8025. Guha, T., Lin, C.T., Bhattacharya, S.K., Mahajan, A.S., Ou-Yang, C.F., Lan, Y.P., Hsu, S.C., Liang, M.C., 2017. Isotopic ratios of nitrate in aerosol samples from MT. Lulin, a high-altitude station in Central Taiwan. Atmos. Environ. 154, 53–69. Guo, J.H., Liu, X.J., Zhang, Y., Shen, J.L., Han, W.X., Zhang, W.F., Christie, P., Goulding, K.W.T., Vitousek, P.M., Zhang, F.S., 2010. Significant acidification in major Chinese croplands. Science 327, 1008–1010. He, P., Xie, Z., Chi, X., Yu, X., Fan, S., Kang, H., Liu, C., Zhan, H., 2018. Atmospheric D17O(NO 3 ) reveals nocturnal chemistry dominates nitrate production in Beijing haze. Atmos. Chem. Phys. 18, 14465–14476. Hu, W., Hu, M., Hu, W., Jimenez, J.L., Yuan, B., Chen, W., Wang, M., Wu, Y.S., Chen, C., Wang, Z.B., Peng, J.F., Zeng, L.M., Shao, M., 2016. Chemical composition, sources, and aging process of submicron aerosols in Beijing: contrast between summer and winter. J. Geophys. Res.: Atmos. 121 (4), 1955–1977. Huang, R.J., Zhang, Y., Bozzetti, C., Ho, K.F., Cao, J.J., Han, Y., Daellenbach, K.R., Slowik, J.G., Platt, S.M., Canonaco, F., Zotter, P., Wolf, R., Pieber, S.M., Bruns, E.A., Crippa, M., Ciarelli, G., Piazzalunga, A., Schwikowski, M., Abbaszade, G., SchnelleKreis, J., Zimmermann, R., An, Z., Szidat, S., Baltensperger, U., Haddad, I.E., Prevot, A.S.H., 2014. High secondary aerosol contribution to particulate pollution during haze events in China. Nature 514, 218–222. Johnston, J.C., Thiemens, M.H., 1997. The isotopic composition of tropospheric ozone in three environments. J. Geophys. Res.-Atmos. 102, 25395–25404. Kaiser, J., Hastings, M.G., Houlton, B.Z., Röckmann, T., Sigman, D.M., 2007. Triple oxygen isotope analysis of nitrate using the denitrifier method and thermal decomposition of N2O. Anal. Chem. 79, 599–607. Kamezaki, K., Hattori, S., Iwamoto, Y., Ishino, S., Furutani, H., Miki, Y., Uematsu, M., Miura, K., Yoshida, N., 2019. Tracing the sources and formation pathways of atmospheric particulate nitrate over the Pacific Ocean using stable isotopes. Atmos. Environ. 209, 152–166.

8

W. Song et al. / Science of the Total Environment 701 (2020) 134515

Kawashima, H., 2014. Formation mechanism and source apportionment of ammonium and nitrate ions in aerosol using nitrogen isotopes. Earozoru Kenkyu 29, 110–116. Kendall, C., Elliott, E.M., Wankel, S.D., 2007. Tracing anthropogenic inputs of nitrogen to ecosystems. In: Michener, R.M., Lajtha, K. (Eds.), Stable Isotopes in Ecology and Environmental Science. second ed. Blackwell, Oxford, pp. 375–449. Kesselmeier, J., Kuhn, U., Rottenberger, S., Biesenthal, T., Wolf, A., Schebeske, G., Andreae, M.O., Ciccioli, P., Brancaleoni, E., Frattoni, M., Oliva, S.T., Botelho, M.L., Silva, C.M.A., Tavares, T.M., 2002. Concentrations and species composition of atmospheric volatile organic compounds (VOCs) as observed during the wet and dry season in Rondônia (Amazonia). J. Geophys. Res. Atmos., 107 Kesselmeier, J., Kuhn, U., Wolf, A., Andreae, M.O., Ciccioli, P., Brancaleoni, E., Frattoni, M., Guenther, A., Greenberg, J., De Castro Vasconcellos, P., Oliva, T., Tavares, T., Artaxo, P., 2000. Atmospheric volatile organic compounds (VOC) at a remote tropical forest site in central Amazonia. Atmos. Environ. 34 (24), 4063–4072. Liu, J., Zhu, T., 2013. NOx in Chinese Megacities. Disposal of Dangerous Chemicals in Urban Areas and Mega Cities. Springer, Netherlands, pp. 249–263. Liu, X.J., Zhang, Y., Han, W.X., Tang, A.H., Shen, J.L., Cui, Z.L., Vitousek, P., Erisman, J. W., Goulding, K., Christie, P., Fangmeier, A., Zhang, F.S., 2013. Enhanced nitrogen deposition over China. Nature 494, 459–462. Liu, X.Y., Koba, K., Koyama, L.A., Hobbie, S.E., Weiss, M.S., Inagaki, Y., Shaver, G.R., Giblin, A.E., Hobara, S., Nadelhoffer, K.J., Sommerkorn, M., Rastetter, E.B., Kling, G.W., Laundre, J.A., Yano, Y., Makabe, A., Yano, M., Liu, C.Q., 2018. Nitrate is an important nitrogen source for Arctic tundra plants. PNAS 115, 3398–3403. Liu, X.Y., Xiao, H.W., Xiao, H.Y., Song, W., Sun, X.C., Zheng, X.D., Liu, C.Q., Koba, K., 2017. Stable isotope analyses of precipitation nitrogen sources in Guiyang, Southwestern China. Environ. Pollut. 230, 486–494. Moore, J.W., Semmens, B.X., 2008. Incorporating uncertainty and prior information into stable isotope mixing models. Ecol. Lett. 11, 470–480. Morin, S., Savarino, J., Frey, M.M., Yan, N., Bekki, S., Bottenheim, J.W., Martins, J.M.F., 2008. Tracing the origin and fate of NOx in the arctic atmosphere using stable isotopes in nitrate. Science 322, 730–732. Nelson, D.M., Tsunogai, U., Ding, D., Ohyama, T., Komatsu, D.D., Nakagawa, F., Noguchi, I., Yamaguchi, T., 2018. Triple oxygen isotopes indicate urbanization affects sources of nitrate in wet and dry atmospheric deposition. Atmos. Chem. Phys., 1–30 Proemse, B.C., Mayer, B., Chow, J.C., Watson, J.G., 2012. Isotopic characterization of nitrate, ammonium and sulfate in stack PM2.5 emissions in the Athabasca oil sands region, Alberta, Canada. Atmos. Environ. 60, 555–563. Richter, A., Burrows, J.P., Nüss, H., Granier, C., Niemeier, U., 2005. Increase in tropospheric nitrogen dioxide over China observed from space. Nature 437, 129–132. Savard, M.M., Cole, A., Smirnoff, A., Vet, R., 2017. d15N values of atmospheric N species simultaneously collected using sector-based samplers distant from sources – isotopic inheritance and fractionation. Atmos. Environ. 162, 11–22. Song, W., Wang, Y.L., Yang, W., Sun, X.C., Tong, Y.D., Wang, X.M., Liu, C.Q., Bai, Z.P., Liu, X.Y., 2019. Isotopic evaluation on relative contributions of major NOx sources to nitrate of PM2.5 in Beijing. Environ. Pollut. 248, 183–190. Song, Y., Shao, M., Liu, Y., Lu, S., Kuster, W., Goldan, P., Xie, S.D., 2007. Source apportionment of ambient volatile organic compounds in Beijing. Environ. Sci. Technol. 41 (12), 4348–4353.

Streets, D.G., Yarber, K.F., Woo, J.H., Carmichael, G.R., 2003. Biomass burning in Asia: annual and seasonal estimates and atmospheric emissions. Global Biogeochem. Cy. 17 (4). https://doi.org/10.1029/2003GB002040. Tsunogai, U., Miyauchi, T., Ohyama, T., Komatsu, Daisuke, D., Nakagawa, F., Obata, Y., Sato, K., 2016. Accurate and precise quantification of atmospheric nitrate in streams, draining land of various uses by using triple oxygen isotopes as tracers. Biogeosciences 13 (11), 3441–3459. Walters, W.W., Michalski, G., 2016. Theoretical calculation of oxygen equilibrium isotope fractionation factors involving various NOy, molecules, OH, and H2O and its implications for isotope variations in atmospheric nitrate. Geochim. Cosmochim. Acta 191, 89–101. Walters, W.W., Michalski, G., 2015. Theoretical calculation of nitrogen isotope equilibrium exchange fractionation factors for various NOy molecules. Geochim. Cosmochim. Acta 164 (7), 284–297. Wang, M., Shao, M., Lu, S.H., Yang, Y.D., Chen, W.T., 2013. Evidence of coal combustion contribution to ambient VOCs during winter in Beijing. Chinese Chem. Lett. 24 (9), 829–832. Wang, X., Ti, P.C., Luo, Y.X., Sun, X.L., 2016. Determination of 15N natural abundance in nitrogen oxides from major anthropogenic emission sources. Acta Pedol. Sin. 53, 1552–1562 (in Chinese). Wang, Y.L., Song, W., Yang, W., Sun, X.C., Tong, Y.D., Wang, X.M., Liu, C.Q., Bai, Z.P., Liu, X.Y., 2019. Influences of atmospheric pollution on the contributions of major oxidation pathways to PM2.5 nitrate formation in Beijing. J. Geophys. Res.: Atmos. https://doi.org/10.1029/2019JD030284. Womack, C.C., McDuffie, E.E., Edwards, P.M., Bares, R., de Gouw, J.A., Docherty, K.S., Dubé, W.P., Fibiger, D.L., Franchin, A., Gilman, J.B., Goldberger, L., Lee, B.H., Lin, J. C., Long, R., Middlebrook, A.M., Millet, D.B., Moravek, A., Murphy, J.G., Quinn, P. K., Riedel, T.P., Roberts, J.M., Thornton, J.A., Valin, L.C., Veres, P.R., Whitehill, A.R., Wild, R.J., Warneke, C., Yuan, B., Baasandorj, M., Brown, S.S., 2019. An odd oxygen framework for wintertime ammonium nitrate aerosol pollution in urban areas: NOx and VOC control as mitigation strategies. Geophys. Res. Lett. https://doi.org/10.1029/2019GL082028. Xiao, H.W., Xie, L.H., Long, A.M., Ye, F., Pan, Y.P., Li, D.N., Long, Z.H., Chen, L., Xiao, H. Y., Liu, C.Q., 2015. Use of isotopic compositions of nitrate in tsp to identify sources and chemistry in south china sea. Atmos. Environ. 109, 70–78. Yánez-Serrano, A.M., NoLscher, A.C., Williams, J., Wolff, S., Alves, E., Martins, G.A., Bourtsoukidis, E., Brito, J., Jardine, K., Artaxo, P., Kesselmeier, J., 2015. Diel and seasonal changes of biogenic volatile organic compounds within and above an Amazonian rainforest. Atmos. Chem. Phys. 15 (6), 3359–3378. Yang, J.Y.T., Hsu, S.C., Dai, M.H., Hsiao, S.Y., Kao, S.J., 2014. Isotopic composition of water-soluble nitrate in bulk atmospheric deposition at Dongsha Island: sources and implications of external N supply to the northern South China Sea. Biogeosciences 11 (7), 1833–1846. Yeatman, S.G., Spokes, L.J., Dennis, P.F., Jickells, T.D., 2001. Comparisons of aerosol nitrogen isotopic composition at two polluted coastal sites. Atmos. Environ. 35 (7), 1307–1320. Zhao, B., Wang, P., Ma, J.Z., Zhu, S., Pozzer, A., Li, W., 2012. A high-resolution emission inventory of primary pollutants for the Huabei region, China. Atmos. Chem. Phys. 12, 481–501. Zong, Z., Wang, X., Tian, C.G., Chen, Y.J., Fang, Y.T., Zhang, F., Li, C., Sun, J.Z., Li, J., Zhang, G., 2017. First assessment of NOx sources at a regional background site in North China using isotopic analysis linked with modeling. Environ. Sci. Technol. 51 (11), 5923–5931.