Interannual variation of reactive nitrogen emissions and their impacts on PM2.5 air pollution in China during 2005–2015

Emissions of reactive nitrogen as ammonia (NH3) and nitrogen oxides (NO x ), together with sulfur dioxide (SO2), contribute to formation of secondary PM2.5 in the atmosphere. Satellite observations of atmospheric NH3, NO2, and SO2 levels since the 2000s provide valuable information to constrain the spatial and temporal variability of their emissions. Here we present a bottom-up Chinese NH3 emission inventory combined with top-down estimates of Chinese NO x and SO2 emissions using ozone monitoring instrument satellite observations, aiming to quantify the interannual variations of reactive nitrogen emissions in China and their contributions to PM2.5 air pollution over 2005–2015. We find small interannual changes in the total Chinese anthropogenic NH3 emissions during 2005–2016 (12.0–13.3 Tg with over 85% from agricultural sources), but large interannual change in top-down Chinese NO x and SO2 emissions. Chinese NO x emissions peaked around 2011 and declined by 22% during 2011–2015, and Chinese SO2 emissions declined by 55% in 2015 relative to that in 2007. Using the GEOS-Chem chemical transport model simulations, we find that rising atmospheric NH3 levels in eastern China since 2011 as observed by infrared atmospheric sounding interferometer and atmospheric infrared sounder satellites are mainly driven by rapid reductions in SO2 emissions. The 2011–2015 Chinese NO x emission reductions have decreased regional annual mean PM2.5 by 2.3–3.8 μg m−3. Interannual PM2.5 changes due to NH3 emission changes are relatively small, but further control of agricultural NH3 emissions can be effective for PM2.5 pollution mitigation in eastern China.


Introduction
Nitrogen (N) is an essential element for life, but most N in the Earth cannot be used directly by ecosystems. The productivity of the ecosystem depends on the abundance of reactive N (Nr or fixed N) (Vitousek et al 2002). Excessive reactive N will, however, induce negative environmental effects, such as causing soil acidification, eutrophication, decreasing the diversity of ecosystems (Galloway 2001, Galloway et al 2003, and increasing nitric oxide (NO) and nitrous oxide (N 2 O) emissions contributing to air pollution and global warming (Pilegaard et al 2006, Eickenscheidt et al 2011. China is one of the regions with the most intensive reactive nitrogen emissions in the globe due to its rapid industrialization, urbanization, as well as high demand for food production (Liu et al 2013). Recent satellite observations have recorded significant changes in atmospheric ammonia (NH 3 ) and nitrogen dioxide (NO 2 ) levels over China since the 2000s (Qu et al 2017, van der A et al 2017, Warner et al 2017, Liu et al 2018a. Here we will use these satellite observations to constrain Nr emissions in China over [2005][2006][2007][2008][2009][2010][2011][2012][2013][2014][2015] and to assess their impacts on the PM 2.5 air quality. Emissions of Nr to the atmosphere are mainly in the forms of NH 3 and nitrogen oxides (NO x = NO + NO 2 ). NH 3 is the most abundant alkaline gas in the atmosphere. Over 85% of NH 3 is emitted from agricultural activities (Huang et al 2012, Paulot et al 2014, Zhang et al 2018, including nitrogen fertilizer application to farmland (Sha et al 2021) and livestock husbandry systems (Bai et al 2016). Chemical industry, residential, human wastes, and traffic are other important anthropogenic sources (Kean andHarley 2000, Sun et al 2017). NH 3 can also be released from rewettening processes of natural soils (Hickman et al 2018). NO x is mainly emitted as a byproduct of combustion at high temperature, such as from industry and transportation sectors. Soil and lightening are natural sources of NO x (Boersma et al 2005, Ciais et al 2014. Both NH 3 and NO x are precursors of secondary aerosols. NH 3 in the atmosphere reacts with sulfuric acid (H 2 SO 4 ; produced from the oxidation of SO 2 ) and nitric acid (HNO 3 ; produced from the oxidation of NO x ) to form ammonium sulfate and ammonium nitrate aerosols.
Previous studies have shown that the SNA (sulfate-nitrate-ammonium) aerosols account for 20%-57% of PM 2.5 (particulate matter with an aerodynamic diameter less than 2.5 µm) in Chinese cities (Wang et al 2011, Huang et al 2014, Liu et al 2018b. Changes in Nr and SO 2 emissions can thus strongly affect the SNA fraction of PM 2.5 . Surface measurements have recorded large decreases in sulfate aerosol concentrations and weak decreases or even increases in nitrate aerosol levels over North China in recent years (Li et al 2019a, Zhai et al 2021, as driven by the recent clean air actions targeting emissions of SO 2 , NO x , and primary aerosols (Zheng et al 2018, Zhang et al 2019. NH 3 emissions have not been effectively regulated so far in China, and they have drawn increasing attentions for understanding their mitigation potentials and impacts on PM 2.5 air pollution (Liu et al 2019, Guo et al 2020. For example, Liu et al (2019) reported that 50% NH 3 emission reductions combined with 15% reductions of SO 2 and NO x emissions could remove 11%-17% of the total PM 2.5 in China. Here we aim to understand how interannual variations of Chinese NH 3 emissions might have contributed to changes in PM 2.5 concentrations over China in the recent past.
In this work, we extend our bottom-up estimates of Chinese NH 3 emissions (Zhang et al 2018) to the years 2005-2016, and we evaluate the GEOS-Chem model simulated atmospheric NH 3 concentrations against satellite NH 3 observations. Satellite observations of NO 2 and SO 2 columns are also applied to constrain the interannual variations of Chinese NO x and SO 2 emissions over this time period (2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015). We further quantify the changes of PM 2.5 concentrations in China contributed by the interannual changes in anthropogenic NH 3 , NO x , and SO 2 emissions, respectively, over 2005-2015 using the GEOS-Chem model simulations.

The GEOS-chem model
Here we use GEOS-Chem v12.1.1 (http://acmg. seas.harvard.edu/geos/), a three-dimensional global chemical transport model to simulate the chemical and physical processes of Nr in the atmosphere from 2005 to 2015 in China. The model is driven by the MERRA-2 assimilated meteorological data provided by the Global Modeling and Assimilation Office at the National Aeronautics and Space Administration (NASA). Meteorology fields such as temperature, relative and specific humidity, vertical pressure velocity, and surface pressure have a temporal resolution of 3 h, and sea level pressure, tropopause pressure, and other surface variables are at 1 h resolution. The model has 47 vertical layers from surface to 0.01 hPa, and the lowest layer is centered at 58 m above sea level.
GEOS-Chem simulates a detailed tropospheric O 3 -NO x -hydrocarbon-aerosol-halogen chemistry as described by Park et al (2004) and Mao et al (2010). The chemistry system fully couples the H 2 SO 4 -HNO 3 -NH 3 inorganic aerosol thermodynamics system from the ISORROPIA-II thermodynamical model (Fountoukis and Nenes 2007). NH 3 preferably reacts with sulfuric acid to form ammonium bisulfate and ammonium sulfate, then excessive NH 3 would combine with nitric acid to form ammonium nitrate (Binkowski and Roselle 2003). NO x is also a precursor tropospheric ozone, affecting atmospheric oxidizing capacity and the formation of secondary organic aerosols (SOA). In this study we do not analyze SOA, and focus on the SNA components that are directly related to Nr emissions. Both Nr gases and aerosols deposit to the surface via wet deposition (convective scavenging and largescale precipitation) following the parameterization of Liu et al (2001) and dry deposition using a standard resistance-in-series model (Wesely 1989, Zhang et al 2001. Emissions in GEOS-Chem v12.1.1 are processed through HEMCO (Harvard-NASA Emission Component) (Keller et al 2014). We used the Community Emissions Data System (Hoesly et al 2018) for global anthropogenic emissions, overwritten by regional emissions inventories including 2011 NEI (National Emissions Inventory) from EPA (United States Environmental Protection Agency) for US (NEI-2011), EMEP (European Monitoring and Evaluation Programme; www.emep.int/index.html) emissions over Europe, and Canada's Air Pollutant Emissions Inventory. Emissions over Asia are overwritten by the MIX inventory  that includes the MEIC inventory (Multi-resolution Emission Inventory for China; http://meicmodel.org/) over China except for NH 3 , NO x , and SO 2 emissions in China as will be described below. Natural NO x sources from soil and lightning are also included (Lu et al 2021).
In this study we have conducted four sets of GEOS-Chem model simulations for 2005-2015 at the global 2 • latitudes by 2.5 • longitude resolution. The standard simulation (BASE) uses the emission setting as described above that accounts for the interannual variations of Chinese Nr and SO 2 anthropogenic emissions. Three sets of sensitivity simulations by fixing the Chinese anthropogenic NH 3 , Nr (NH 3 + NO x ), and additionally SO 2 (Nr + SO 2 ) emissions, respectively, to the year 2005 are conducted (i.e. FixNH 3 , FixNr, and FixALL), and their differences with the BASE simulation and with each other estimate the impacts from their emissions' interannual variations. For better evaluating the model results, we have also conducted a nested GEOS-Chem simulation using the BASE emission conditions for the years 2005-2015 at a higher 0.5 • latitude × 0.625 • longitude resolution over Asia and 2 • latitudes × 2.5 • longitude for the rest of the world. All the simulations are initiated after one year spin-up.

Observations
We have compiled an ensemble of surface and satellite observations to evaluate our estimates of Nr emissions and resulting model simulations. We use observed surface NH 3 concentration data at 53 sites for 2015 over China We also use satellite observations of NO 2 and SO 2 columns from the ozone monitoring instrument (OMI). OMI aboard the NASA's Aura satellite launched in 2004 measures backscattered solar radiation with a nadir-scanning resolution of 13 × 24 km 2 and at local passing time of 13:45 (Levelt et al 2006). We use the NASA standard products of daily Level-3 NO 2 tropospheric column (Krotkov et  We further evaluate the model simulated surface PM 2.5 concentrations with available PM 2.5 datasets. We use the nationwide PM 2.5 surface measurements for 2013-2015 obtained from the China National Environmental Monitoring Center (CNEMC; https:/ /air.cnemc.cn:18007/). Ground-level PM 2.5 products derived from satellite observations of aerosol optical depth over 2005-2015 are also applied (Hammer et al 2020; https://sites.wustl.edu/acag/datasets/surface-pm2-5/#V4.CH.03). The products have a spatial resolution of 0.1 • × 0.1 • and are regridded to 0.5 • × 0.625 • .
In addition, the land cover datasets retrieved from the moderate resolution imaging spectroradiometer (MODIS) aboard the Terra and Aqua satellites are applied to identify the changes of cropland distributions in China. We use the Terra-Aqua combined land cover product (MCD12Q1) (https:/ /modis-land.gsfc.nasa.gov/landcover.html), in which the types of land cover at 500 m resolution are identified to 17 classes, including 11 natural vegetation classes, three human-altered classes, and three nonvegetated classes. We extract the grids classified as cropland and calculate the cropland areas at the model 0.5 • × 0.625 • resolution. The MCD12Q1 products covers the period of 2005-2012, and the latest available year 2012 data is applied to the years afterwards in this study. The NH 3 emissions from fertilizer use are based on practical fertilizer application information, including cropland area, fertilizer application timing and rate for 21 types of crops, vegetables, and fruits. Emission factors are calculated as a function of fertilizer type, application mode, soil pH, and soil cation exchange capacity, and further modulated by local surface temperature and wind speed. We previously used a baseline cropland area for the year 2000 (Zhang et al 2018). Here we scale the baseline cropland area to match those observed by MODIS as described in the section above to account for their interannual changes in China. The fertilizer application amount and types have also changed substantially during this period. We use the year-specific statistics of provincial fertilizer application amounts from the China Rural Statistical Yearbook (National Bureau of Statistics (NBSC) 2006-2017), and associated changes in the fractions of synthetic fertilizer types are estimated by the International Fertilizer Association (www.fertilizer.org).

Results and discussion
NH 3 emissions from livestock account for six categories of animals, including beef cattle, dairy cows, goat, sheep, pig, and poultry raised in three raising systems: free-range, intensive, and grazing (Zhang et al 2018). The most common system in the rural area is free-range, which is the traditional way to raise livestock leading to large NH 3 emissions. The intensive raising system becomes the most popular way near megacities due to its high efficiency and easiness to manage the shelter environment. Grazing mostly occurs in Northwest China. We use the livestock manure mass-flow methodology and emission factors from Huang et al (2012), and further consider the meteorology conditions (2 m air temperature and 10 m wind speed) effects on emission factors (Zhang et al 2018). The numbers of animals raised in intensive and grazing systems of each province are divided based on the Chinese Animal Husbandry and Veterinary Yearbook (Editorial Committee of China animal husbandry and veterinary Yearbook 2006Yearbook -2017. In addition to the agricultural sources, we also include NH 3 emitted from vehicles, biomass burning, residential burning, industry, and waste disposal estimated by Kang et al (2016) from 2005 to 2012. To further account for changes in some of these sources over 2013-2016, we assume that NH 3 emissions from vehicle, industry, and residential burning follow the same interannual variations as SO 2 emissions for each source, and apply the corresponding SO 2 emission changes estimated from the MEIC inventory. Figure 1 shows the spatial distribution of anthropogenic NH 3 emissions over China for the year 2015, and interannual changes in Chinese NH 3 emissions from different sources over 2005-2016. High NH 3 emissions can be found in the eastern China, in particular, over the key regions such as the North China Plain (including Beijing-Tianjin-Hebei (BTH) and surrounding areas), Yangtze River Delta (YRD), Pearl River Delta (PRD), and Sichuan Basin (SCB) (see figure S1 available online at stacks.iop.org/ERL/16/ 125004/mmedia for their locations). These regions are highly populated and also have intense agricultural activities. The annual total Chinese anthropogenic NH 3 emissions are 12.5 Tg NH 3 in 2015, and range from 12.0-13.3 Tg during 2005-2016, with 37%-42% from fertilizer application and 46%-53% from livestock manure management. The national total NH 3 emissions show relative weak interannual variations, while spatially they generally increase in western China and decrease in eastern China relative to 2005 for both the fertilizer application and livestock sources (figure S2). The increases of NH 3 emissions in western China are largely driven by increasing fertilizer application amount and livestock number, different from those in eastern China and national totals as discussed below.
We find that the national total NH

Top-down estimates of Chinese NO x and SO 2 emissions during 2005-2015
We use OMI satellite observations of NO 2 tropospheric columns and SO 2 PBL columns to constrain the interannual variations of NO x and SO 2 emissions in China over the period of 2005-2015. We follow the finite-difference mass-balance inversion method from Geddes and Martin (2017). We have conducted two sets of 11 year (2005-2015) GEOS-Chem simulations at the 2 • × 2.5 • resolution: one with the Chinese SO 2 and NO x emissions fixed to those from the MEIC inventory in the year 2010, and the other applies +30% perturbations on the Chinese SO 2 and NO x emissions based on the first simulation setup. We then calculate the changes in simulated NO 2 tropospheric columns and SO 2 PBL columns, and estimate the interannual variations of their emissions for years 2005-2015 using the formula below where E topdown is the top-down SO 2 or NO x emissions; E prior is prior SO 2 or NO x emissions based on the 2010 MEIC estimates; △E is 30% perturbation of E prior ; Ω prior is simulated SO 2 or NO x columns with prior emissions; △Ω is simulated changes of SO 2 or NO x columns between prior and 30% perturbation emissions; Ω sat is SO 2 or NO x columns observed from OMI.  (Zheng et al 2018). It should be noted that the top-down inferred NO 2 and SO 2 interannual changes are subject to uncertainties in the OMI retrievals, such as the reduced spatial coverage due to OMI row anomaly since 2007. The OMI products used in this study also treat aerosols implicitly in the retrieval algorithm, which may affect the observed interannual variations over polluted regions with changing aerosol levels (Lin et al 2015, Lamsal et al 2021. Figures S4-S7 evaluate the BASE simulation with the ensemble of measurements as described above. The BASE model in general reproduces the measured near surface NH 3 concentrations and nitrogen (NH 4 + and NO 3 − ) wet deposition fluxes with correlation coefficients of 0.51-0.82 and normalized mean biases (NMBs) within 30%. The model results underestimate the measured NH 3 concentrations and NH 4 + wet deposition fluxes in summer by 41. 9% (2015) and 25.0% (2008-2012 average), respectively, likely reflecting that our estimates of Chinese NH 3 emissions in summer are still biased low. The BASE model results are also in good agreement with the measured surface SNA aerosol concentrations ( figure S6). Further evaluations of the model simulated surface PM 2.5 concentrations with CNEMC measurements and satellite products show consistent spatial distributions in eastern China (figure S7). Figure 3 shows the IASI satellite observed spatial distribution of atmospheric NH 3 columns and the corresponding model results averaged over 2008-2015. The comparison of IASI observations and model results shows a high spatial correlation coefficient of 0.82. The model captures the observed high NH 3 levels over North China, while there is an overall low bias of −35.6% relative to the IASI observations, as can be seen from the comparisons over the southern China and western China. In addition to the possible low bias in the NH 3 emission estimates, IASI NH 3 measurements may also biased high due to the relative error weighted method, which tends to give more weight to high values (Van Damme et al 2015). We also compare the simulated monthly mean NH 3 concentrations in January, April, July and October 2005-2015 with AIRS observations at 918 hPa (Warner et al 2017) in figure S8. AIRS NH 3 observations mainly cover the intense agricultural regions of China. We find that compared with the AIRS observations, the seasonal variations of simulated NH 3 concentrations in China are consistent with a high correlation coefficient (0.85) and a small low bias (−5.5%). Figure 3 also compares IASI observed and model simulated 2008-2015 interannual variations of NH 3 columns over the four most densely populated Chinese regions: BTH, YRD, PRD, and SCB. Despite the small interannual variations and decreases since 2012 in the Chinese NH 3 emissions (figure 1), both IASI satellite observations and model results show increases in atmospheric NH 3 concentrations since 2011 over the four regions. The BASE model results well capture the IASI observed NH 3 interannual variation over these regions, although there are considerable underestimates (−43% ∼ −47%) of NH 3 concentrations over YRD and SCB. Previous studies have reported the increases in atmospheric NH 3 over the North China (i.e. BTH and surrounding areas) observed by IASI and AIRS, and mainly attributed such NH 3 increases to SO 2 emission reductions (Warner et al 2017, Liu et al 2018a. Here we use our sensitivity simulations (section 2.1) to identify drivers of NH 3 interannual variations over broader regions of China. We find that over the four regions SO 2 emission reductions are the dominant factor driving the NH 3 concentration increases (figure S9). Decreases in NO x emissions since 2011 also show small contributions over BTH and YRD. As explained by Liu et al (2018a), reductions in SO 2 and NO x emissions would lower the formation of SNA and thus enhance the NH 3 gas-phase partitioning. Changes in the NH 3 concentration are generally weak except for the sudden drop from 2006 to 2007 over BTH, consistent with the changes in NH 3 emissions (figure 1).

Impacts of interannual variations of Nr emissions
We now quantify the changes in PM 2.5 air pollution attributable to the interannual changes in NH 3 , NO x , and SO 2 emissions over China. Figure 4 shows the simulated seasonal and annual mean surface SNA aerosol concentrations over China for the year 2015, and also their changes due to the changes in Chinese  The national mean population-weighted SNA concentrations are 38.3 µg m −3 annually, peaking in winter (50.2 µg m −3 ) and being the lowest in summer (28.9 µg m −3 ). Such strong seasonal variations are largely caused by the higher pollutants' emissions and more frequent stagnant weather conditions over eastern China in winter than summer, and more effective removal of air pollution by precipitation in summer than winter as well (Liu et al 2018b).
Changes in Chinese NH 3 , NO x , and SO 2 emissions in 2015 relative to 2005 have led to large changes in the SNA aerosol concentrations. As shown in figure 4, due to the large reductions in SO 2 emissions, the annual mean population weighted SNA aerosol concentrations have reduced by 8.3 µg m −3 . There is a higher reduction of 9.9 µg m −3 in summer than 6.1 µg m −3 in winter, reflecting higher efficiency of SO 2 photochemical oxidation to sulfuric acid in summer. Spatially These changes in the SNA aerosol concentration largely reflect the changes of their precursor emissions, but can also be influenced by the changes in atmospheric oxidants affecting the SNA aerosol formation. For example, NO x emission changes would affect tropospheric ozone concentrations and thus affect the oxidation of SO 2 to form sulfate aerosol. In turn, changes in aerosol levels would also affect ozone concentrations via aerosol chemistry and photolysis pathways (Li et al 2019b). Figure S10 shows the simulated changes in seasonal and annual mean surface ozone concentrations over China caused by changes in Chinese NH 3 , NO x , and SO 2 emissions from 2005 to 2015. Compared with figure 4, the aerosol concentration changes driven by NH 3 and SO 2 emission changes tend to be negatively correlated with the ozone concentration changes, likely reflecting the suppression of ozone formation via aerosol radiative effect or the ozone sink via aerosol chemistry. The increases of NO x emissions from 2005 to 2015 could also decrease seasonal mean ozone levels due to titration at night and in winter, except in summer when photochemistry is active. Future work is needed to quantify the interaction of interannual changes in SNA aerosol and ozone concentrations.
Analyzing the interannual variations of emission changes (figure S11), we can see strong influences on the regional SNA aerosol concentration from changes in Chinese NO x emissions. We estimate that over the study period the Chinese NO x emissions peak in 2011. Increases in NO x emissions from 2005 to 2011 increased the annual mean PM 2.5 concentrations by 6.5 µg m −3 in BTH, 4.6 µg m −3 in YRD, and 2.8 µg m −3 in SCB, offsetting or even exceeding the PM 2.5 decreases due to SO 2 emission reductions. Estimated changes in Nr emissions and their effects on PM 2.5 over PRD are small over this period. Reductions in Chinese NO x emissions after 2011 have accelerated the PM 2.5 decreases.  Chinese NO x emission reduction reduced regional mean PM 2.5 by 3.8 µg m −3 in BTH, 3.2 µg m −3 in YRD, and 2.3 µg m −3 in SCB. Together changes in Chinese Nr and SO 2 emissions over 2011-2015 have decreased the regional mean PM 2.5 by 12.5 µg m −3 in BTH, 8.8 µg m −3 in YRD, and 9.7 µg m −3 in SCB, with changes in Nr emissions contributing 20%-36% of the decreases. According to the CNEMC measurements (figure S7), PM 2.5 concentrations have decreased rapidly in eastern China since 2013 due to stringent air pollution control actions, e.g. decreased by ∼29 µg m −3 over 2013-2015 in BTH. Our results suggest that ∼40% of the decreases can be attributed to changes in Nr and SO 2 emissions over China.

Conclusions
We have analyzed the interannual variations of Nr (NH 3 + NO x ) emissions in China over 2005-2015 and quantified their contributions to PM 2.5 air pollution. Satellite observations of NH 3 , NO 2 , and SO 2 atmospheric concentrations during this time period are used to constrain Chinese Nr and SO 2 emissions. We have applied a bottom-up approach to estimate Chinese NH 3 emissions, and a top-down approach for NO x and SO 2 emission estimates. The bottom-up estimates of NH 3 emissions resolve the interannual variations of agricultural activities such as changes in fertilizer application amounts and types and livestock numbers. We find small interannual changes in the total Chinese anthropogenic NH 3 emissions, ranging 12.0-13.3 Tg during 2005-2016, with 37%-42% from fertilizer application and 46%-53% from livestock manure management. Although fertilizer and livestock amounts have increased over the time period, the shifts of fertilizer types (from ABC to urea and nitrogen compounds) and livestock raising systems (from free-range to intensive systems) keep the total Chinese NH 3 emissions relatively stable or even slightly decreasing after 2012.
In contrast to NH 3 , Chinese NO x and SO 2 emissions as inferred from the satellite observations show large interannual changes. We estimate that Chinese Chinese NH 3 , NO x , and SO 2 emissions over 2005-2015, the GEOS-Chem model simulation can well capture the increases of atmospheric NH 3 levels after 2011 in eastern China as observed by the IASI satellite observations (although with some low biases in the model results) and attribute the atmospheric NH 3 increases mainly to the rapid SO 2 emission reductions.
We find that interannual variations of Nr and SO 2 emission changes have strongly influenced the regional SNA components of PM 2.5 in eastern China over 2005-2015. The Chinese NH 3 emission changes in 2015 relative to 2005 (−6%) lead to 0.9 µg m −3 reductions in the national population weighted mean SNA concentration, compared with an increase of 1.6 µg m −3 due to 7.5% NO x emission increases, and a decrease of 8.3 µg m −3 due to 48.3% SO 2 emission reductions. The 2011-2015 Chinese NO x emission reductions decreased regional mean PM 2.5 by 2.3-3.8 µg m −3 , also becoming an important driver of recent PM 2.5 air quality improvements. The Chinese air pollution control actions after 2011 have mainly focused on power plant, industry, and transportation sectors that have significantly lowered NO x and SO 2 emissions (Zheng et al 2018). Our analyses indicate that strengthening agricultural NH 3 emission reduction can achieve similar effectiveness for further improving PM 2.5 air quality in eastern China.

Data availability statement
All data that support the findings of this study are available from the corresponding author on reasonable request.