Temporal characteristics of atmospheric ammonia and nitrogen dioxide over China based on emission data, satellite observations and atmospheric transport modeling since 1980

Abstract. China is experiencing intense air pollution caused in large part by anthropogenic emissions of reactive nitrogen (Nr). Atmospheric ammonia (NH3) and nitrogen dioxide (NO2) are the most important precursors for Nr compounds (including N2O5, HNO3, HONO and particulate NO3− and NH4+) in the atmosphere. Understanding the changes in NH3 and NO2 has important implications for the regulation of anthropogenic Nr emissions and is a requirement for assessing the consequence of environmental impacts. We conducted the temporal trend analysis of atmospheric NH3 and NO2 on a national scale since 1980 based on emission data (during 1980–2010), satellite observation (for NH3 since 2008 and for NO2 since 2005) and atmospheric chemistry transport modeling (during 2008–2015). Based on the emission data, during 1980–2010, significant continuous increasing trends in both NH3 and NOx were observed in REAS (Regional Emission inventory in Asia, for NH3 0.17 and for NOx 0.16 kg N ha−1 yr−2) and EDGAR (Emissions Database for Global Atmospheric Research, for NH3 0.24 and for NOx 0.17 kg N ha−1 yr−2) over China. Based on the satellite data and atmospheric chemistry transport model (CTM) MOZART-4 (Model for Ozone and Related chemical Tracers, version 4), the NO2 columns over China increased significantly from 2005 to 2011 and then decreased significantly from 2011 to 2015; the satellite-retrieved NH3 columns from 2008 to 2014 increased at a rate of 2.37 % yr−1. The decrease in NO2 columns since 2011 may result from more stringent strategies taken to control NOx emissions during the 12th Five Year Plan, while no control policy has focused on NH3 emissions. Our findings provided an overall insight into the temporal trends of both NO2 and NH3 since 1980 based on emission data, satellite observations and atmospheric transport modeling. These findings can provide a scientific background for policy makers that are attempting to control atmospheric pollution in China. Moreover, the multiple datasets used in this study have implications for estimating long-term Nr deposition datasets to assess its impact on soil, forest, water and greenhouse balance.


Introduction
Reactive nitrogen (N r ) emissions have increased significantly in China due to anthropogenic activities such as increased combustion of fossil fuels, over-fertilization and high stocking rates of farm animals (Canfield et al., 2010;Galloway et al., 2008;Liu et al., 2013).Elevated N r in the environment has led to a series of effects on climate change and Published by Copernicus Publications on behalf of the European Geosciences Union.
L. Liu et al.: Temporal characteristics of atmospheric ammonia and nitrogen dioxide ecosystems, e.g., biodiversity loss, stratospheric ozone depletion, air pollution, freshwater eutrophication, the potential alteration of global temperature, drinking water contamination, dead zones in coastal ecosystems and grassland seed bank depletion (Basto et al., 2015;Lan et al., 2015;Shi et al., 2015).Atmospheric reactive N emissions are dominated by nitrogen oxides (NO x = NO + NO 2 ) and ammonia (NH 3 ; Li et al., 2016;Galloway et al., 2004).Atmospheric NO 2 and NH 3 are the most important precursors for N r compounds including N 2 O 5 , HNO 3 , HONO and particulate NO − 3 and NH + 4 in the atmosphere (Xu et al., 2015;Pan et al., 2012).Therefore, an understanding of both the spatial and temporal patterns of NO 2 and NH 3 is essential for evaluating N-enriched environmental effects, and can provide the scientific background for N pollution mitigation.
To investigate the spatial and temporal variations in atmospheric NO 2 and NH 3 , ground measurements are acknowledged to be an effective way of monitoring the accurate concentrations of NO 2 and NH 3 (Xu et al., 2015;Pan et al., 2012;Meng et al., 2010).Ground measurements of NO 2 concentrations in China, including about 500 stations in 74 cities, have been monitored and reported to the public since January 2013 (Xie et al., 2015).By the end of 2013, this network was extended with hourly NO 2 concentrations from more than 850 stations in 161 cities.However, there are fewer NH 3 measurements across China than NO 2 measurements.The China Agricultural University has organized a nationwide nitrogen deposition monitoring network (NNDMN) since 2010, consisting of 43 monitoring sites covering urban, rural (cropland) and background (coastal, forest and grassland) areas across China (Xu et al., 2015;Liu et al., 2011).Xu et al. (2015) reported the ground NH 3 concentrations throughout China for the first time, providing great potential to understand the ground NH 3 concentrations on a national scale.Other networks include the Chinese Ecosystem Research Network (CERN) which was established in 1988, including 40 field stations (Fu et al., 2010).However, to our knowledge, there are no detailed reports about ground NH 3 concentrations from CERN on a national scale.In addition, four Chinese cities (Xiamen, Xi'an, Chongqing and Zhuhai) have joined the Acid Deposition Monitoring Network in East Asia (EANET) since 1999.However, only one site (Hongwen, Xiamen) in EANET measured the ground NH 3 concentrations, and the data from it are not continuous.Finally, ground NH 3 concentrations at 10 sites in northern China from 2007 to 2010 have been reported by Pan et al. (2012).All of the aboveground measurements provide the potential to understand NH 3 and NO 2 concentrations on a regional scale.However, there is limited information on the spatial and temporal variations in NH 3 and NO 2 in the atmosphere across China.This is due to the limited observation sites and monitoring period, as well as given the uneven distribution of the monitoring sites.Importantly, atmospheric NH 3 and NO 2 monitoring based on ground-based local sites may have limited spatial representativeness of the regional scale as both NH 3 and NO 2 are highly variable in time and space (Clarisse et al., 2009;Wichink Kruit et al., 2012;Boersma et al., 2007).
In order to complement ground-based measurements, satellite observation of NH 3 and NO 2 is a welcome addition for analyzing the recent trends of NH 3 and NO 2 in the atmosphere.Satellite remote sensing offers an opportunity to monitor atmospheric NH 3 and NO 2 with high temporal and spatial resolutions (Warner et al., 2017;Li et al., 2017).NO 2 was measured by multiple space-based instruments, including the Global Ozone Monitoring Experiment (GOME), SCanning Imaging Absorption SpectroMeter for Atmospheric CHartographY (SCIAMACHY), Ozone Monitoring Instrument (OMI) and Global Ozone Monitoring Experiment-2 (GOME-2).The OMI NO 2 provides the best horizontal resolution (13 km × 24 km) among instruments in its class and near-global daily coverage (Boersma et al., 2007).OMI observations have been widely applied in environmental-related studies and for the support of emission control policy (Russell et al., 2012;Zhao and Wang, 2009;Castellanos et al., 2015;Lamsal et al., 2015;F. Liu et al., 2016;Foy et al., 2016).First measurements of NH 3 from space were reported over Beijing and San Diego areas with the Tropospheric Emission Spectrometer (TES; Beer et al., 2008) and in fire plumes in Greece with the Infrared Atmospheric Sounding Interferometer (IASI; Coheur et al., 2009).The first global map of NH 3 was created from IASI measurements by correlating the observed brightness temperature differences to NH 3 columns using the averaged datasets in 2008 (Clarisse et al., 2009).Shortly after that, many studies focused on developing techniques to gain more reliable NH 3 columns (Whitburn et al., 2016a;Van Damme et al., 2014a), validating the retrieved NH 3 columns using the ground measurements (Van Damme et al., 2015;Dammers et al., 2016) and comparing the data with the results of the atmospheric chemistry transport models (Van Damme et al., 2014b;Whitburn et al., 2016a), and the estimated NH 3 columns obtained from Fourier transform infrared spectroscopy (FTIR; Dammers et al., 2016).The retrieval algorithm of obtaining IASI NH 3 columns was based on the method described in Whitburn et al. (2016a).Two main steps were performed to derive the NH 3 columns from the satellite measurements.First, derive the spectral hyperspectral range index (HRI) based on each IASI observations (Walker et al., 2011;Van Damme et al., 2014a).Second, convert HRI to NH 3 columns based on a constructed neural network with input parameters including vertical NH 3 profile, satellite viewing angle, surface temperature and so on (Whitburn et al., 2016a).The progress made in satellite techniques provides a possibility for understanding both the spatial and temporal variations in NH 3 and NO 2 in the atmosphere.
In addition to satellite observations, the emission data are also very important for investigating the temporal trends of NH 3 and NO 2 such as the IIASA inventory (Cofala et al., 2007), EDGAR (Emission Database for Global Atmospheric Research, RAINS-Asia (Regional Air Pollution Information and Simulation) and Asia REAS (FRCGC, 2007).REAS is considered the first inventory to integrate historical, current and future emissions data for Asia based on a consistent methodology (Ohara et al., 2007)

NH 3 and NO 2 emissions
We examined the REAS (Regional Emission Inventory in Asia) emission inventory dataset for Asia with 0.5 • × 0.5 • resolution for the period 1980-2010 and analyzed the temporal trends of NO x and NH 3 over China.REAS v1.1 is believed to be the first inventory of integrating past, present and future dataset in Asia based on a consistent methodology.The REAS datasets have been validated by several emissions, and denote agreement with the recent growth status in Chinese emissions (Ohara et al., 2007).We also collected NO x and NH 3 emission data from EDGAR (Emissions Database for Global Atmospheric Research) v4.3.1, which was developed by the Netherlands Environmental Assessment Agency and European Commission Joint Research Centre (Jgj et al., 2002).The EDGAR emissions are calculated on the basis of a point emissions inventory conducted by the International Energy Agency.EDGAR also covers a long time period , with the highest spatial resolution globally (0.1 • × 0.1 • ; http://edgar.jrc.ec.europa.eu/overview.php?v=431).

Satellite observations
IASI is a passive remote-sensing instrument operating in nadir mode that measures the infrared radiation emitted by the Earth's surface and the atmosphere (Clarisse et al., 2009).It covers the entire globe twice a day, crossing the Equator at a mean solar local time of 09:30 and 21:30 and has an elliptical footprint of 12 by 12 km up to 20 by 39 km depending on the satellite-viewing angle.In this study we use daytime satellite observations as these are more sensitive to NH 3 and are associated with a large positive thermal contrast and a significant amount of NH 3 (Van Damme et al., 2014a;Whitburn et al., 2016a).The availability of measurements is mainly driven by the cloud coverage as only observations with cloud coverage lower than 25 % are processed to be a good compromise between the number of data kept for the analysis and the bias due to the effect of clouds.As the quantity of daily data is not always sufficient to obtain meaningful distributions (due to cloud cover or the availability of the temperature profiles from the EUMETSAT operational processing chain; Van Damme et al., 2014a), it is more appropriate to consider monthly or yearly averages for this trend analysis.We consider IASI observations with a relative error below 100 % or an absolute error below 5 × 10 15 molec cm −2 for analysis over China.For the error, the filtering depends on the use of the data.For this, low columns typical for background conditions with a large relative error but a small absolute error are also taken into account.For other applications, such as comparing with ground measurements, we would recommend to use a threshold of 75 % or even 100 % relative error.We gained the data upon request from the Atmospheric Spectroscopy Group at Université Libre De Bruxelles (http://www.ulb.ac.be/cpm/atmosphere.html).These data can be gridded on 0.1 • latitude × 0.1 • longitude (Dammers et al., 2016), 0.25 • latitude × 0.25 • longitude (Whitburn et al., 2016a) and 0.5 • latitude × 0.5 • longitude (Whitburn et al., 2016b) or even coarser resolutions depending on the usage of the data.For IASI NH 3 , we firstly divided China into a 0.5 • latitude × 0.5 • longitude grid.For each grid cell, we calculated the monthly arithmetic mean by averaging the daily values with observations points within the grid cell.Similarly, we calculated the annual arithmetic mean by averaging the daily values with observations points within the grid cell over the whole year.
The NO 2 columns are obtained from the OMI instrument on NASA's EOS Aura satellite globally every day.We used the generated products by the project "Derivation of Ozone Monitoring Instrument tropospheric NO 2 in near-real time" (DOMINO) to analyze the temporal trends of NO 2 columns over China.In DOMINO products, only the observations with a cloud radiance fraction below 0.5 were processed for analysis.The retrieval algorithm is described in detail in the previous work (Boersma et al., 2007) and recent updates can be found in the DOMINO Product Specification Document (http://www.temis.nl/docs/OMI_NO2_HE5_1.0.2.pdf).We used tropospheric NO 2 retrievals from the DOMINO algorithm v2.0.The retrieval quality of NO 2 products is strongly dependent on different aspects of air mass factors, such as radiative transfer calculations, terrain heights and surface albedo.The OMI v2.0 data were mainly improved by more realistic atmospheric profile parameters, and include more surface albedo and surface pressure reference points than before (Boersma et al., 2011(Boersma et al., , 2016)).The DOMINO NO 2 datasets are available from http://www.temis.nl/airpollution/no2.html.We should state in particular that we used directly the DOMINO v2.0 products of monthly means from 2005 to 2015 over China for the trend analysis.The DOMINO NO 2 columns were gridded at a resolution of 0.125 • latitude × 0.125 • longitude grid globally, which has been widely used for scientific applications (Ma et al., 2013;Ialongo et al., 2016;Castellanos et al., 2015).
To illustrate measurement availability, we presented here some measurement statistics.A total number of cloud-free daytime observations as characterized by the operational IASI processor by year were retrieved in China during 2008-2015 for NH 3 (Fig. 1b).We retrieved more observation numbers after 2010 than those during 2008-2009.In 2010, the update of the improved air temperature profiles, cloud properties products and cloud detection, which are important for calculating the thermal contrast, increased the quality of retrieval (Van Damme et al., 2014a, b).In September 2014, there was another update of the air temperature profiles, cloud properties products and cloud detection for calculating the thermal contrast.The version of IASI NH 3 columns used in the present work was based on the method described in Whitburn et al. (2016a).We did not use the IASI NH 3 after September 30 in 2014 for the trend analysis because an update of the input meteorological data on 30 September 2014 has caused a substantial increase in the retrieved atmospheric NH 3 columns.For the updates of the IASI-NH 3 data, the reader can refer to Van Damme et al. (2014a, b), and Whitburn et al. (2016a).The monthly observation numbers are also presented in Fig. 1a, showing that spring (March, April and May), summer (June, July and August), autumn (September, October and November) and winter (December, January and February) months represent 29, 26, 23 and 21 %, respectively.Compared with large variations in observation numbers for NH 3 , the observation numbers for NO 2 varied less by year; the winter season had the least, while other seasons varied little.

Atmospheric transport chemistry model
Atmospheric transport chemistry models are also of central importance in modeling the tropospheric NO 2 and NH 3 .We applied the widely used atmospheric global atmospheric transport chemistry model MOZART-4 to simulate the tropospheric NO 2 and NH 3 columns during 2008-2015 in accordance with the time period of IASI NH 3 measurements.The MOZART-4 model is driven by the meteorological data from the NASA Goddard Earth Observing System Model, version 5 (GEOS-5), at a resolution of 1.9 • latitude × 2.5 • longitude spatially.The emission data applied for driving the simulations are based on the updated EDGAR emission inventories.Twelve bulk aerosol compounds, 39 photolysis, 85 gas species and 157 gas-phase reactions were integrated in MOZART-4.The chemical mechanism on N compounds including the NO 2 , NH 3 and aerosols are detailedly integrated to MOZART-4, which is considered to be suitable for tropospheric chemical compositions (Emmons et al., 2010;Pfister et al., 2008;Sahu et al., 2013).The output data used in the current work temporally vary 6 h every day, which were available upon request from Louisa Emmons at the National Center for Atmospheric Research (NCAR).The monthly means of NO 2 and NH 3 columns were averaged by the daily data, and then used for the trend analysis over China.For more details about MOZART-4, the reader should refer to previous studies (Emmons et al., 2010;Brasseur et al., 1998;Beig and Singh, 2007).

NH 3 and NO 2 emissions during 1980-2010
We conducted the temporal analysis of NH 3 and NO x emissions since 1980 based on REAS and EDGAR.Significant continuous increasing trends in both NH 3 and NO x were observed from REAS (for NH 3 0.17 and for NO x 0.16 kg N ha −1 yr −2 ) and EDGAR (for NH 3 0.24 and for NO x 0.17 kg N ha −1 yr −2 ) over China (Fig. 2).We found a relatively consistent increase in NO x emission from EDGAR and REAS over China, i.e., 0.17 kg N ha −1 yr −2 vs. 0.16 kg N ha −1 yr −2 , but inconsistency in the magnitude of NH 3 emissions from EDGAR and REAS over China, i.e., 0.24 kg N ha −1 yr −2 vs. 0.17 kg N ha −1 yr −2 .The increase rate in NH 3 emissions over China from EDGAR was much higher than that from REAS, indicating that the magnitude of increase trend in NH 3 over China remains a debate, although their thread values (the slope in Fig. 2) of 0.24 kg N ha −1 yr −2 (EDGAR) vs. 0.17 kg N ha −1 yr −2 (REAS) both reflected a continuous increasing trend (in this regard they are consistent).It implies that the NH 3 emissions are indeed increasing during 1980-2010.We also conducted a simple correlation analysis of the NH 3 (Fig. 2a) and NO x (Fig. 2b) from REAS and EDGAR, showing agreement in the magnitude (slope = 1.06) and temporal trend (R 2 = 0.96) for NO x , but with some inconsistency in the increase rate (slope = 1.33) for NH 3 .
The discrepancy in the magnitude of NH 3 increase rate from REAS and EDGAR (0.24 kg N ha −1 yr −2 vs. 0.17 kg N ha −1 yr −2 ) in China since 1980 may be caused by the different emission factors considered for estimating NH 3 emissions.The EDGAR v4.3.1 NH 3 emissions were calculated based on a variety of sectors, including agriculture, shipping, waste solid and wastewater, energy for buildings, process emissions during production and application, power industry, oil refineries, transformation industry, combustion for manufacturing, road transportation, railways, pipelines and off-road transport, while the REAS v1.1 NH 3 emissions focused mainly on agriculture sources (i.e., manure management of livestock and fertilizer application; Crippa et al., 2016;Ohara et al., 2007).Moreover, the fundamental methodology on estimating the REAS v1.1 NH 3 emissions did not consider the seasonal agricultural variations compared with that of EDGAR v4.3.1 NH 3 emissions (Kurokawa et al., 2013), and the removal efficiency (as a key element to estimate NH 3 emissions) was also reported to be much higher in REAS v1.1 than in EDGAR v4.3.1 (Kurokawa et al., 2013).
A previous study (Liu et al., 2013) summarized published data on the national anthropogenic NH 3 and NO x emissions with multiple periods in China (Wang et al., 2009;Wang et al., 1997;Streets et al., 2003;Klimont et al., 2001;Sun and Wang, 1997;Olivier et al., 1998;FRCGC, 2007) and also analyzed the temporal pattern of NH 3 emissions.Their results showed that the NH 3 emissions had increased at an annual average rate of 0.32 Tg N yr −2 (about 0.33 kg N ha −1 yr −2 ).The increase rate of NH 3 emissions (0.33 kg N ha −1 yr −2 ) by Liu et al. (2013) was double that in REAS (0.17 kg N ha −1 yr −2 ), implying that the NH 3 increase rate in China is still an open question and should be further studied.3.2 Satellite NH 3 and NO 2 over China in the recent decade

Temporal trends
We referred to the method of a previous study (Russell et al., 2012) to conduct the temporal trend analysis by calculating the average values during cold months (October-March) and warm months (April-September), respectively.We herein concentrated more on the temporal analysis of satellite observations during warm months because of the relatively lower uncertainty in comparison with that during cold months.Figure 3 [2006][2007][2008][2009][2010], the Chinese government undertook a series of strategies to increase energy efficiency and to reduce NO x emissions, but NO x emissions were not successfully restrained, which created a big challenge for improving air quality over the country (Xia et al., 2016).During the 12th FYP period (2011-2015), more stringent strategies were implemented to control NO x emissions, including the application of selective catalytic/noncatalytic reduction (SCR/SNCR) systems in the power sector; staged implementation of tighter vehicle emission standards; and a series of standards with aggressive emission limits for power, cement, and the iron and steel industries.These strategies are believed to have helped achieve national targets of NO x emission abatement (Xia et al., 2016).However, the satellite-retrieved NH 3 columns increased with a slope of 0.118 × 10 15 molec cm −2 yr −1 (2.37 % yr −1 ) in warm months from 2008 to 2014 (Fig. 3).The percent increase rate for NH 3 by year (2.37 % yr −1 ) from 2008 to 2014 is lower than that for NO 2 (4.07 % yr −1 ) from 2005 to 2011, although the absolute NH 3 increase rate of 0.118 × 10 15 molec cm −2 yr −1 from 2008 to 2014 was higher than absolute NO 2 increase rate of 0.063 × 10 15 molec cm −2 yr −1 from 2005 to 2011.An increase in NH 3 columns from IASI may be due to decreased NH 3 removal, leading to a larger fraction maintained in the gaseous state for a long time rather than changing to the condensed phase.Specifically, NH 3 is considered an important alkaline gas that is abundant in the atmosphere and is able to neutralize acidic components such as HNO 3 and H 2 SO 4 through the oxidation of NO x and SO 2 , respectively (Li et al., 2014;Liu et al., 2011;X. Liu et al., 2017;Xu et al., 2015).The decreased NH 3 removal to some degree can be attributed to continuous decreased acidic gases, including NO 2 and SO 2 , over China under a strong control policy in 12th FYP, which can largely decrease the fraction of the chemical conversion to (NH 4 ) 2 SO 4 and NH 4 NO 3 in the atmosphere.A increasing trend in NH 3 columns may be associated with continuous N fertilizer use for guaranteeing an increase in crop production (Erisman et al., 2008).Although there was no strong NH 3 emission control regulation, N fertilizer efficiency should be further improved over China.In 2015, the Ministry of Agriculture formally announced its "Zero Increase Action Plan" for national fertilizer use by 2020, which requires that the annual increase in total fertilizer use be less than 1 % from 2015 to 2019, with no further increase from 2020 (Liu et al., 2015).
If the Zero Increase Action Plan for N fertilizer can be effective, future NH 3 emissions should be consistent with the current NH 3 emissions.In addition, due to strong emission control of NO x , the NO x emissions were believed to decrease significantly from 2011 to 2015.We can reasonably make two major conclusions.First, the atmospheric NO 2 , as a key indicator of oxidized N compounds (NO 2 , HNO 3 and NO − 3 ), decreased since 2011, and will continue to decrease under the current policy.Second, the atmospheric NH 3 , as a key indicator of reduced N (NH 3 and particulate NH + 4 ), will slightly increase or stay at the current level in the future with the Zero Increase Action Plan.Thus, due to a decreasing trend of oxidized N (NO x -N), ammonia N (NH x -N) should still dominate N r deposition (oxidized N plus reduced N) in China, and is expected to play a more significant role in N r deposition.Therefore, monitoring the reduced N on a regional scale is encouraged to assist in enacting effective measures to protect the environments and public health, with respect to air, soil and water quality.

Spatial pattern
High NH 3 columns were found in Beijing, Hebei, Henan, Shandong, Hubei and Jiangsu provinces and in eastern Sichuan Province (Fig. 4a), which were consistent with their high NH 3 emissions due to intensive fertilizer application and livestock (Huang et al., 2012).Guangdong, Guangxi, Hunan and Jiangxi provinces also showed high NH 3 columns, due to high volatilization from paddy fields in these regions, with rice being the dominant crop and contributing the most emissions.High NH 3 columns in southern China are in agreement with the high percentage of paddy farmland area (Fig. S1a in the Supplement), and the high NH 3 columns in northern China are in agreement with the high percentage of dry farmland area (Fig. S1b).In addition, the NH 3 emissions from vehicles in urban areas could also contribute to the observed high NH 3 columns.For example, in Beijing, the contribution of vehicles equipped with catalytic converters, particularly since the introduction of three-way catalysts, to non-agricultural NH 3 emissions has recently been considered and might be the most important factor influencing NH 3 concentrations in urban cities (Meng et al., 2011;Xu et al., 2017).In addition, Xinjiang Province also emits remarkable NH 3 emissions related to sheep manure management (Huang et al., 2012;Kang et al., 2016;Zhou et al., 2015;Liu et al., 2017a).The lower NH 3 columns are located mostly in the Tibetan Plateau area, where there is a minimal amount of arable land and low use of synthetic nitrogenous fertilizers.
NO 2 columns (Fig. 4b) show significantly higher values over vast areas covering North China, East China, and the Sichuan Basin.The NO 2 columns also show high values over the Pearl River Delta, the southern part of Northeast China, and some areas in Northwest China.High NO 2 columns are mostly distributed in populated areas (Fig. S2), where there is a mix of various anthropogenic NO x sources, such as vehicles and industrial complexes (Wang et al., 2012;Xu et al., 2015;Meng et al., 2010).It should be noted that an enhanced  emission intensity from transportation is confirmed since 2005, even with staged implementation of tightened emission standards for on-road vehicles (Wang et al., 2012).For example, NO x emissions from transportation grew to 30 % for the whole country in 2014, and the values reached 44, 55, and 33 % for Beijing, Shanghai, and Guangdong, respectively (Xia et al., 2016).Therefore, transportation is believed to play an increasingly important in regional NO 2 pollution, especially when emissions from stationary sources are gradually controlled through increased penetration of selective catalytic/non-catalytic reduction (SCR/SNCR) systems.

Limitations of satellite observations
It is difficult to gain whole coverage over China based on the daily data for both IASI NH 3 and OMI NO 2 .For daily NO 2 , the spatial coverage gained by OMI was influenced by cloud radiance fractions, surface albedo, solar zenith angles, www.atmos-chem-phys.net/17/9365/2017/Atmos.Chem.Phys., 17, 9365-9378, 2017 row anomaly and so on (Russell et al., 2011;De Smedt et al., 2015).The "row anomaly" issue resulting from the OMI instrumental problem had an impact on approximately half of the rows undergoing unpredictable patterns in cross-track directions relying on latitudes and seasons and prevented obtaining convincing daily product with continuous coverage (Boersma et al., 2011(Boersma et al., , 2016)).For NH 3 , the satellite instruments were strongly dependent on the meteorological conditions such as cloud fractions or the availability of the temperature profiles (Van Damme et al., 2014a;Boersma et al., 2011), and we cannot retrieve the whole coverage based on daily data over China.It will be beneficial to analyze a very local region with enough numbers of observations, but not appropriate to analyze such large coverage over China.Facing this big challenge, we used the monthly data for the trend analysis over China.The uncertainty in DOMINO v2.0 NO 2 columns has been well documented in Boersma et al. (2011), and the relative error is reported lower than 20-30 % in East Asia by an improved altitude-dependent air mass factor look-up table, a more realistic atmospheric profile, an increased number of reference vertical layers and advanced surface albedos (Boersma et al., 2011).The reader is strongly suggested to refer to Boersma et al. (2011) for more details on the uncertainty analysis.
The potential uncertainty of IASI NH 3 columns resulted from IASI observation instruments and retrieval algorithms.In this paper, the NH 3 datasets were generated based on the recent-updated robust and flexible NH 3 retrieval algorithms, which were designed to overcome some shortcomings of the current algorithms (Whitburn et al., 2016a).The current algorithms were designed firstly to calculate the hyperspectral range index (HRI), a measure for the NH 3 signature strength in the spectrum, and then converted to IASI NH 3 columns by using the thermal contrast (TC) and lookup tables (LUTs) of the (HRI, TC) pair corresponding to NH 3 columns.The retrieval of HRIs is strongly dependent on the amount of NH 3 and the thermal state of the atmosphere (Whitburn et al., 2016a).The quality of the IASI NH 3 product has been validated by atmospheric chemistry transport models (CTMs), ground-based and airborne measurements, and NH 3 total columns obtained with ground-based Fourier transform infrared spectroscopy (FTIR).A first validation of the IASI NH 3 using the LOTOS-EUROS model was conducted over Europe, indicating the respective consistency of IASI measurements and model simulations (Van Damme et al., 2014b).A first evaluation of IASI NH 3 dataset using ground-based measurements was then made worldwide, presenting consistency with the available ground-based observations and denoting promising results for evaluation by using independent airborne data (Van Damme et al., 2015).A first validation of IASI NH 3 dataset using ground-based FTIR derived NH 3 total columns was evaluated, demonstrating a mean relative difference of −32.4 ± (56.3) %, a correlation r of 0.8 with a slope of 0.73 (Dammers et al., 2016).

Atmospheric chemistry transport model NO 2 and NH 3 columns since 2008
Satellite NO 2 and NH 3 columns were observed at overpass time as an instantaneous point in a day (at 09:30 for IASI NH 3 and at 13:45 for OMI NO 2 local time).These instantaneous satellite observations may not be representative of the temporal trend analysis over China.We further retrieved the monthly variations in NO 2 and NH 3 columns since 2008 from MOZART varying 6 h every day (00:00, 06:00, 12:00, 18:00).We compared the temporal trend analysis of NO 2 from MOZART at 12:00 with that gained from satellite at the overpass time (OMI 13:45 local time) as well as for NH 3 .
For NH 3 , we found the percent increase rate at 12:00 during warm months between 2008 and 2015 was 1.30 % yr −1 from MOZART (Fig. 5), which was lower than that (2.37 % yr −1 ) from IASI during 2008-2014.The percent increase rate by daily average (at 00:00, 06:00, 12:00 and 18:00) during warm months between 2008 and 2015 was 1.36 % yr −1 from MOZART (Fig. 5).In MOZART-4, the alkaline gaseous NH 3 and the acidic gaseous NO 2 (the precursor for HNO 3 ) and SO 2 are very important precursors for bulk NH 4 NO 3 and (NH 4 ) 2 SO 4 particles, which form the primary system of gas-particle partitioning (NH 3 -NH + 4 -NO x -NO − 3 -SO 2 -SO 2− 4 ).The chemical shifts between particulate NH 4 NO 3 and gaseous NH 3 and NO x are correlated with the abundance of NH 3 and NO x and meteorological factors.The decreased abundance of NO x between 2011 and 2015 may also contribute to an increase in the NH 3 abundance in the gas stage resulting from decreased conversion to particulate NH 4 NO 3 .3.4 Implications for estimating long-term N r deposition datasets and recommendations for future work We found both the NO x and NH 3 over China increased continuously from 1980 to 2010 based on emissions data from REAS and EDGAR.In recent years, based on satellite observations, we found an increase of 2.37 % yr −1 in NH 3 columns during 2008-2014.We also found high-level NO 2 columns over China from 2005-2011 (4.07 % yr −1 ) but a decrease from 2011 to 2015 (−3.62 % yr −1 ).Despite the decline, the NO 2 columns during 2011-2015 were still at a high level, with an average of 1.87 × 10 15 molec cm −2 yr −1 , compared with that (1.65 × 10 15 molec cm −2 yr −1 ) during 2005-2010.Notably, these emissions certainly lead to the deposition of atmospheric N r in the form of dry and wet processes into aquatic ecosystems and terrestrial, with implications of effects on ecosystem and human health, biological diversity and greenhouse gas balances (Lu et al., 2016).Hence, it is very crucial to estimate N r deposition with high spatiotemporal resolutions in order to drive ecological models such as the DeNitrification-DeComposition (DNDC) model and Integrated BIosphere Simulator (IBIS) to assess its impact on soil, forest, water and greenhouse balance.Here, we call for a long-term dataset of N r depositions both regionally and globally to investigate how the N emissions affect the environment.A challenge still exits in estimating both the dry (NO 2 , HNO 3 particulate NO − 3 , NH 3 and particulate NH + 4 ) and wet (NH + 4 and NO − 3 in precipitation) depositions for a long-term dataset such as since 1980 or earlier, possibly due to the complex scheme of N transformations and transportation or limited available data from emissions, satellites and a limited number of ground measurements.
Satellite observations provide a new perspective for estimating N r depositions regionally and have been used to improve the estimation performance.For example, to improve the modeling performance in dry gaseous NO 2 depositions from GEOS-Chem (Goddard Earth Observing System chemical transport model), Nowlan et al. (2014) applied the OMI NO 2 columns to calibrate the simulated ground NO 2 concentrations, and then estimated the deposition between 2005 and 2007.Our previous work focusing on the dry particulate NO − 3 deposition over China was also based on the OMI NO 2 columns, MOZART simulations and monitored-based sources (Liu et al., 2017b).Geddes et al. (2017) used the satellite NO 2 columns from GOME, GOME-2 and SCIA-MACHY instruments to calibrate the NO x emissions in GEOS-Chem to estimate the NO x depositions since 1996.The simulations combining the satellite measurements and CTMs to derive N r depositions (Geddes and Martin, 2017;Nowlan et al., 2014) in recent years will provide relatively accurate datasets (certainly need to be validated and modified by ground measurements).
Despite progress in satellite techniques in recent decades (for NO 2 since 1997 by GOME and for NH 3 since 2008 by IASI), we can find very few studies concerning N r depositions before 1997 based on satellite observations.Thus, with the help of emissions data such as REAS and EDGAR, we can derive long-term N r depositions, especially before 1997.Long-term emissions data such as REAS and EDGAR will provide a valuable dataset to expand the modeling of N r depositions in recent years.In order to derive the N r depositions from the emission data, CTMs are frequently used through modeling the wet (simplified as the product of scavenging efficiency and precipitation amount) and dry processes (simplified as the inferential method by multiplying the deposition velocity and gaseous or particulate concentrations).However, we still lack a comprehensive dataset of gridded longterm N r depositions including both the dry (NO 2 , HNO 3 particulate NO − 3 , NH 3 and particulate NH + 4 ) and wet (NH + 4 and NO − 3 in precipitation) processes over China, which will be addressed in future work.
Another gap is that, all the abovementioned studies focused on the NO x depositions and did not derive the NH y (NH 3 and NH + 4 ) depositions over China.Our recent work (Liu et al., 2017a) using IASI NH 3 columns combining the vertical profiles from MOZART benefits our understanding of the ground NH 3 concentrations over China, and the satellite-derived ground NH 3 concentrations were generally in accord with the national measurements from NNDMN.To date, there are still no reports of using the satellite NH 3 columns to derive the temporal and regional NH y depositions over China, which dominated the total N r depositions (NO x plus NH y ; L. Liu et al., 2016;Liu et al., 2013).The gaps in modeling of NH y depositions by applying the satellite observations combining the CTM simulations require more efforts and further research.

Conclusions
Atmospheric ammonia (NH 3 ) and nitrogen dioxide (NO 2 ) play an important role in determining air quality, environmental degradation and climate change.The emission data, satellite observations and atmospheric transport modeling have great potential for understanding the temporal varia-tions in atmospheric NH 3 and NO 2 on a regional scale, with high spatial and temporal resolutions.This study analyzed the characteristics of atmospheric NH 3 and NO 2 over China since 1980 based on the multiple datasets.The major findings were as follows: 1. Based on emission data, a significant continuous increasing trend in both NH 3 and NO x were observed from REAS (for NH 3 0.17 and for NO x 0.16 kg N ha −1 yr −2 ) and EDGAR (for NH 3 0.24 and for NO x 0.17 kg N ha −1 yr −2 ) over China during 1980-2010.
2. Based on the satellite observations, we found highlevel NH 3 columns with the percent increase rate of 2.37 % yr −1 from 2008 to 2014.For NO 2 , we found continuous high-level NO 2 columns over China from 2005-2011 but a decrease from 2011 to 2015 (still at a high level).The decrease of NO 2 columns may result from more stringent strategies taken to control NO x emissions during the 12th Five Year Plan, including successful application of SCR/SNCR systems in the power sector, tighter emission standards on vehicles and a series of standards with aggressive emission limits.An increasing trend in NH 3 columns may be due to continuous N fertilizer use for guaranteeing a continuous increase in crop production.An increase in NH 3 columns may be due to decreased NH 3 removal leading to a larger fraction being maintained in the gaseous state for a long time rather than changing to the condensed phase, which may be related to continuously decreasing acidic gases, including NO 2 and SO 2 , over China under a strong control policy in 12th FYP.
4. The multiple datasets used in the current work have implications for estimating long-term N r deposition datasets.The simulations combining the satellite measurements and CTMs to derive N r depositions will provide relatively accurate datasets, and the REAS and EDGAR emissions have potential to expand the modeling N r depositions to long-term datasets.In particular, modeling of NH y depositions by applying satellite observations combining the CTM simulations requires more efforts and further research.

Figure 1 .
Figure 1.The satellite-derived observation numbers for NO 2 and NH 3 .Panel (a) shows the percentages of observations in each month in 2010 for NO 2 and in 2015 for NH 3 and (b) represents the total observation numbers for NO 2 and NH 3 over China.Notably, the NO 2 observation numbers were gained from DOMINO products with a cloud radiance fraction below 0.5, while the IASI observations with a relative error below 100 % or an absolute error below 5 × 10 15 molec cm −2 were processed for analysis over China.

Figure 2 .
Figure 2. The NO 2 and NH 3 emissions over China.Panel (a) shows the NO 2 and NH 3 emissions over China from 1980 to 2010 from REAS, (b) represents the NO 2 and NH 3 emissions over China from 1980 to 2010 from EDGAR, (c) demonstrates the relationship of NO 2 emissions over China from REAS and EDGAR and (d) shows the relationship of NH 3 emissions over China from REAS and EDGAR.

Figure 3 .
Figure 3.Time series of average OMI NO 2 and IASI NH 3 columns over China during warm months (April-September) and cold months (October-March).The time period of NO 2 columns was from 2005 to 2015, while the time span of NH 3 columns was from 2008 to 2014 over China.The associated mean error for each period is presented here as error bars.The percent increase or decrease rate (%) was the long-term mean, calculated as 100 ×( Y 2 −Y 1

Figure 5 .
Figure 5.Time series of MOZART NO 2 and NH 3 columns over China during average warm months (April-September) and cold months (October-March) from 2008 to 2015.The mean columns were calculated by averaging the columns at 00:00, 06:00, 12:00 and 18:00.The associated mean error for each period is presented here as error bars.
, and EDGAR is the global emission data with 0.1 by 0.1 grid, which has the highest spatial resolution among different datasets mentioned above.Thus, REAS and EDGAR are used to analyze the historical trends of NH 3 and NO 2 during 1980-2010 in this study.
shows the temporal trend of NO 2 columns during warm and cold months between 2005 and 2015 as well as monthly average values.From satellite observations, the NO 2 columns over China increased with a slope of 0.063 × 10 15 molec cm −2 yr −1 (4.07 % yr −1 ) in warm L.Liu et al.:Temporal characteristics of atmospheric ammonia and nitrogen dioxide months from 2005 to 2011 and then decreased with a slope of −0.072 molec cm −2 in warm months (−3.62 % yr −1 ) from 2011 to 2015 (Fig.3).The decreasing trends were consistent with NO x emissions since 2011 over China (decreasing from 24.04 × 10 6 t in 2011 to 20.78 × 10 6 t in 2014, China Statistical Yearbook, http://www.stats.gov.cn/).During the 11th Chinese Five Year Plan (FYP) period (