Changes in PM2.5 sensitivity to NO x and NH3 emissions due to a large decrease in SO2 emissions from 2013 to 2018

ABSTRACT The authors evaluated and compared the behavior of PM2.5 with respect to NO x and NH3 emission changes in high (the year 2013) and low (the year 2018) SO2 emission cases. Two groups of simulations were conducted based on anthropogenic emissions from China in 2013 and 2018, respectively. In each group of simulations, a respective 25% reduction in NO x and NH3 emissions were assumed. A sensitivity factor (β) was defined as the relative change in PM2.5 concentration due to 1% change in NO x or NH3 emissions. In the high SO2 emissions case, PM2.5 was more sensitive to NH3 (0.31) emissions change than NO x (0.21). Due to the significant decrease in SO2 emissions from the high to low SO2 emissions case, the sensitivity of PM2.5 to NO x increased to 0.33, while its sensitivity to NH3 decreased to 0.22. The result implies that now and in the future, PM2.5 is/will be less sensitive to NH3 emissions change, while NO x emissions control is more effective in reducing the surface PM2.5 concentration. Seasonally, in the low SO2 emissions case, the sensitivities of PM2.5 to NO x and NH3 in winter were higher than those in summer, indicating that to deal with severe winter haze more attention should be paid to the emissions control of inorganic PM2.5 precursors, especially NO x . Graphical abstract


Introduction
Secondary inorganic aerosols (sulfate, nitrate, and ammonium; simplified to SNA) are important constituents of PM 2.5 , occupying up to a half of the total PM 2.5 mass (Pan et al. 2016;Wang et al. 2014). Sulfur dioxide (SO 2 ), nitrogen oxides (NO x ), and ammonia (NH 3 ) are precursors of SNA. In the atmosphere, SO 2 and NO x are oxidized to sulfuric acid (H 2 SO 4 ) and nitric acid (HNO 3 ), and then they are neutralized by NH 3 . NH 3 prefers to form ammonium sulfate ((NH 4 ) 2 SO 4 ) first for its stability. Ammonium nitrate (NH 4 NO 3 ) is semi-volatile, and the low temperature and high humidity are favorable for NH 4 NO 3 staying in the condensation state. Emission changes of SO 2 , NO x , and NH 3 will lead to changes in SNA concentration and then PM 2.5 concentration.
The sensitivity of PM 2.5 to SO 2 , NO x , and NH 3 emission changes is estimated to evaluate emission control strategies and to design effective emission control policies (Wang et al. 2013;Zhang et al. 2019Zhang et al. , 2015b. Wang et al. (2013) reported that from 2000 to 2006, a large increase in SO 2 (60%) and NO x (80%) emissions resulted in a 60% lifting of the SNA concentration over China. Since then, from 2006 to 2015, changes in SO 2 (−16%) and NO x (+16%) emissions have led to an overall decrease in the surface SNA concentration, while an increase in NH 3 emissions has offset the benefit of SO 2 and NO x emissions control (Wang et al. 2013). Fu et al. (2017) and Wang et al. (2013) pointed out the necessity of NO x and NH 3 emissions control for the high sensitivity of PM 2.5 with respect to NO x and NH 3 emissions change. However, these studies were conducted at a relatively high SO 2 emissions level. From 2013 through 2017, observations from satellites have indicated a significant decrease in the SO 2 column over China ( Figure  S1), and SO 2 emissions from China were estimated to have decreased by 59% (Zheng et al. 2018). Considering the thermodynamic equilibrium among H 2 SO 4 , HNO 3 , and NH 3 , a sharp reduction in SO 2 emissions will alter the sensitivity of the PM 2.5 response to NO x and NH 3 emission changes (Wang et al. 2013).
The purpose of this paper is to assess the influence of the large SO 2 emissions reduction from 2013 to 2018 on the sensitivity of PM 2.5 to NO x and NH 3 emissions control and make recommendations for future emission control policies with respect to PM 2.5 precursors. We introduce the model and simulation scenarios in section 2. In section 3, we discuss the sensitivity of PM 2.5 to NO x and NH 3 emission changes at different SO 2 emission levels. Concluding remarks are given in section 4.

Model description
We used the GEOS-Chem global chemical transport model (http://www.geos-chem.org), v11-1. The global version with a horizontal resolution of 2°× 2.5°was employed, driven by the MERRA-2 meteorology field. All simulations in this study used meteorology fields from December 2012 to December 2013, and the first month was for spin-up. The aerosol species considered in the GEOS-Chem model include SNA aerosol, carbonaceous aerosol, dust aerosol, sea salt, and marine POA (primary organic aerosols). SNA aerosol simulation coupled to gasphase chemistry was originally described by Park et al. (2004). SNA thermodynamics are calculated using the module of ISORROPIA II, developed by Fountoukis and Nenes (2007). The default anthropogenic emissions in GEOS-Chem are from EDGAR v4.2 (EC-JRC/PBL, http:// edgar.jrc.ec.europa.eu), and NH 3 emissions are from the MASAGE inventory developed by Paulot et al. (2014). We overwrote the anthropogenic emissions in East Asia with the MIX inventory (Li et al. 2017).
GEOS-Chem has been used to simulate PM 2.5 and its constituents in China, and it has been proven to have good skill in simulating the spatiotemporal distribution and concentration level of aerosols. Zhang et al. (2015a) reported a spatial correlation coefficient (R) of 0.65 between modeled and simulated PM 2.5 in China, and Zhang et al. (2018) reported an R of 0.85 for the day-to-day variation of PM 2.5 . Both found an underestimation of~40% in the PM 2.5 concentration simulated by the model, but the bias was within a reasonable performance level for chemical models to simulate PM 2.5 (Eder and Yu 2006;Stewart et al. 2019). GEOS-Chem also performs well in simulating SNA aerosols. Wang et al. (2013) and Zhang et al. (2015b) reported a good ability of GEOS-Chem to simulate SO 4 2− at weekly, monthly, and seasonal scales. However, the model was found to underpredict the SO 4 2− concentration during severe haze days for a large percentage of them (Wang et al. 2014). It was improved through adding a heterogeneous oxidation procedure of SO 2 to H 2 SO 4 in the model, which was included in the simulations carried out in this work. NO 3 − and NH 4 + overestimation is a typical problem of chemical models (Stewart et al. 2019;Wang et al. 2013). Heald et al. (2012) recommended to reduce the HNO 3 concentration when it is read into the SNA aerosol chemical module in the model, which was also included in this study.

Simulation scenarios
Two groups of simulations were conducted based on the emissions in the years 2013 and 2018, which are referred to as the high and low SO 2 emission cases, respectively. We assumed a separate 25% decrease in NO x and NH 3 emissions from China and calculated the sensitivity of PM 2.5 to them. As mentioned, the default MIX inventory in GEOS-Chem is for the year 2010, and we scaled it up to obtain the emissions for 2013 and 2018. According to Zheng et al. (2018), from 2010 to 2013, the emissions of the main PM 2.5 precursors, i.e. SO 2 , NO x , non-methane volatile organic carbon (NMVOC), NH 3 , elemental carbon (EC), and organic carbon (OC), changed by −9%, 5%, 9%, 4%, 1%, and −3%, respectively. After the Air Pollution Prevention and Control Action Plan was delivered in 2013, emissions of the above species dropped largely from 2013 through 2017, especially for SO 2 . The relative emissions change of the six precursors were −59% for SO 2 , −21% for NO x , 2% for NMVOC, −3% for NH 3 , −28% for EC, and −32% for OC. We

Sensitivity of PM2.5 to NO x and NH 3 emissions at different SO 2 emission levels and in different seasons
Here, we discuss the sensitivity of PM 2.5 with respect to NO x and NH 3 emissions control in different SO 2 emission cases and different seasons. We defined an efficiency factor (β) following the work of Zhang et al. (2019), Zhang et al. (2015b) to quantify and compare the PM 2.5 sensitivities to NO x emissions change. β is defined as follows: where ΔX X represents the relative changes in the PM 2.5 concentration and ΔE E denotes the relative change in NO x or NH 3 emissions. In this study, ΔE E was 0.25 for all the simulations. Figure 2 shows the annual mean PM 2.5 distribution in the high and low SO 2 emission cases and the difference between them. The PM 2.5 concentration was much higher in the east than in the northeast or west, for both the high and low SO 2 emission cases. Therefore, we focused on the east of China (22°-42°N, 102°-122°E) where the PM 2.5 concentrations were highest. The annual mean PM 2.5 concentration over East China decreased by 10.3 μg m −3 (19.6%) due to the emissions reduction from the high to low SO 2 emissions case, 55% of which was driven by a reduction in the sulfate concentration. Seasonally, the reduction in the PM 2.5 concentration was lower in winter (January, February, and December) than in summer (June, July, and August). Relatively, the PM 2.5 reduction rate was 17.8% in winter, which was lower than the annual mean level; and it was 28.2% in summer, which was much higher than the annual mean rate of decrease. This indicates that the formation of aerosols is more easily affected by the emissions of precursors in summer.
As shown in Figure 3, in the high SO 2 emissions case, the annual mean PM 2.5 sensitivity with respect to NO x emissions reduction (β) was 0.21 over the region of interest. Seasonally, β in winter (0.22) and summer (0.23) was close to that of the annual mean. There was a sharp increase in β (53.3%) from the high to low SO 2 emissions case at the annual mean level. In other words, under low SO 2 emissions, the behavior of PM 2.5 becomes more sensitive to  NO x emissions change. This indicates that now and in the future, NO x emissions control is/will be more effective in reducing the PM 2.5 concentration. The increase in β from the high to low SO 2 emissions case varied from season to season. For the East China regional mean, β increased by more than 80% in winter, while in summer the rate of increase was approximately 20%. Different to that in the high SO 2 emissions case, the wintertime β was much higher than the summertime one in the low SO 2 emissions case. In winter, a reduction in SO 2 emissions would release more free NH 3 into the atmosphere, making NH 4 NO 3 more sensitive to NO x emissions change. This explains the large increase in β in winter. The results imply that now and in the future, NO x emissions control is/will be more effective in dealing with severe winter haze than in previous years.
The behavior of PM 2.5 with respect to NH 3 emissions control is displayed in Figure 4. In the high SO 2 emissions case, PM 2.5 was 47.6% more sensitive to NH 3 than to NO x emissions change, with an annual mean β of 0.31. Seasonally, β was higher in winter (0.43) than in summer (0.24), in contrast to that in the NO x emissions control condition. The sensitivity of PM 2.5 to NH 3 emissions change fell by 30.1% from the high to low SO 2 emissions case, and the decrease was more significant in winter (32.6%) than in summer (20.7%). Notably, in the low SO 2 emissions case, the sensitivity of PM 2.5 to NH 3 emissions change (0.22) was 32.7% lower than it was to NO x emissions change. The result indicates that (1) when SO 2 emissions decrease, more free NH 3 is released to the atmosphere, making NH 3 emissions control less effective at reducing the surface PM 2.5 concentration; and (2) now and in the future, it is/will be more effective to reduce the PM 2.5 concentration by controlling NO x rather than NH 3 emissions.

Conclusion
The reduction in emissions of the precursors of PM 2.5 from 2013 to 2018 resulted in a 19.6% decrease in the surface PM 2.5 concentration over East China (22°-42°N, 102°-122°E), most of which was driven by the significant decrease in the sulfate concentration. The substantial decrease in SO 2 emissions and the sulfate concentration were expected to have altered the sensitivity of PM 2.5 to NO x and NH 3 emission changes. Therefore, we conducted model sensitivity studies to assess the changes in PM 2.5 sensitivities to NO x and NH 3 emission changes owing to the large decrease in SO 2 emissions from 2013 to 2018.
In the high SO 2 emissions case (the year 2013), PM 2.5 was 47.6% more sensitive to NH 3 emissions change than NO x , which was mostly driven by winter. From the high to low SO 2 emissions case, the PM 2.5 sensitivity factor (β) for NH 3 decreased by 30.1%, while for NO x it increased by 53.3%, and the PM 2.5 sensitivity was 48.6% higher to NO x than NH 3 . Seasonally, in the low SO 2 emissions case, the PM 2.5 sensitivities to both NO x and NH 3 were higher in winter than in summer.
Based on the results of this work, we conclude that now and in the future, the behavior of PM 2.5 is/will become less sensitive to NH 3 emission changes, while NO x emissions control is more effective at reducing the surface PM 2.5 concentration, especially in winter.

Disclosure Statement
No potential conflict of interest was reported by the authors.