Long-term variations in surface NOx and SO2 mixing ratios from 2006 to 2016 at a background site in the Yangtze River Delta region, China

Abstract. China has been experiencing rapid changes in emissions of
air pollutants in recent decades. Increased emissions of primary
particulates and reactive gases caused severe haze in several polluted
regions including the Yangtze River Delta (YRD). Measures implemented in
recent years for improving air quality have reduced the emissions of
NOx, SO2, etc. The emission changes in these gases are reflected
by tropospheric columns from satellite observations and surface measurements
of surface concentrations from urban sites. However, little is known about
the long-term variations in regional background NOx and SO2. In
this study, we present NOx and SO2 measurements from the Lin'an
station (LAN; 30∘18' N, 119∘44′ E; 138.6 m a.s.l.), one of the Global Atmosphere Watch (GAW) stations in China. We
characterize the seasonal and diurnal variations and study the long-term
trends of NOx and SO2 mixing ratios observed at LAN from 2006 to
2016. We also interpret the observed variations and trends in terms of
changes in meteorological conditions as well as emission of these gases. The
overall average mixing ratios of NOx (NO2) and SO2 during
2006–2016 were 13.6 ± 1.2 ppb (12.5 ± 4.6) and 7.0 ± 4.2 ppb, respectively. The averaged seasonal variations showed maximum values of
NOx and SO2 in December (23.5 ± 4.4 ppb) and January (11.9 ± 6.2 ppb), respectively, and minimum values of 7.1 ± 0.8
and 2.8 ± 2.3 ppb (both in July), respectively. The average diurnal
variation characteristics of NOx and SO2 differed considerably
from each other, though the daily average mixing ratios of both gases were
significantly correlated (R2=0.29, P<0.001). The annual
average mixing ratio of NOx increased during 2006–2011 and then
decreased significantly at 0.78 ppb/yr (−5.16 %/yr, P<0.01). The
annual 95th and 5th percentiles of hourly NOx mixing ratios
showed upward trends until 2012 and 2014, respectively, before a clear
decline. The annual average mixing ratio of SO2 decreased significantly
at 0.99 ppb/yr (−8.27 %/yr, P<0.01) from 2006–2016. The annual
95th and 5th percentiles of hourly SO2 mixing ratios all
exhibited significant (P<0.001) downward trends at 3.18 and
0.19 ppb/yr, respectively. Changes in the total NOx and SO2
emissions as well as the industrial emissions in the YRD region were
significantly correlated with the changes in annual NOx and SO2 mixing ratios. The significant decreases in NOx from 2011 to 2016 and
SO2 from 2006 to 2016 highlight the effectiveness of relevant control
measures on the reduction in NOx and SO2 emissions in the YRD
region. A decrease in annual SO2 / NOx ratio was found, suggesting
a better efficacy in the emission reduction in SO2 than NOx. We
found gradual changes in average diurnal patterns of NOx and SO2,
which could be attributed to increasing contributions of vehicle emissions
to NOx and weakening impacts of large sources on the SO2
concentration. This study reaffirms China's success in controlling both
NOx and SO2 in the YRD but indicates at the same time a necessity
to strengthen the NOx emission control.



Introduction
China's economy has experienced decades of rapid development, resulting in considerable pollutant emissions from coal combustion and motor vehicles, which affect ambient air quality and human health (Kan et al., 2009(Kan et al., , 2012Liang et al., 2019). NO x and SO 2 are two major gaseous pollutants that are essential precursors to secondary aerosol formation and acidification . Therefore, the changes in NO x and SO 2 emissions have been receiving increasing attention in China (Zhao et al., 2013. To improve air quality, the Chinese government has promulgated a series of policies and regulations on SO 2 and NO x control, especially since 2006 and 2011, respectively (Zheng et al., 2015).
Long-term observations of NO x and SO 2 are critical not only for the integrated assessment of air quality and atmosphere-biosphere interactions (Swartz et al., 2020a), but also for the analysis of their reduction effects on PM 2.5 , nitrate, sulfate and near-surface O 3 , providing a basis for further improvement of atmospheric protection policies . At a regional scale, long-term, reliable NO x and SO 2 observations can also provide data to enable the scientific community to predict the future state of the atmosphere and assess environmental policies, serving to reduce environmental risks and enhance climate, weather and air quality prediction capabilities (World Meteorological Organization, 2017). Numerous studies have evaluated the effectiveness of NO x and SO 2 control in China from a long-term perspective by using emission inventories, satellite retrieval data and ground-monitoring data. For example, Sun et al. (2018) used a unified source emission inventory approach to quantify the historical emission trends of SO 2 and NO x in China from 1949 to 2015; the results indicated that these pollutants reached an inflection point in 2006 and 2011, respectively. Source emission inventories by Kurokawa and Ohara (2020) revealed similar patterns. During the period from January 2005 to December 2015, the column concentration of NO 2 from Ozone Monitoring Instrument (OMI) satellite retrievals indicated an increasing trend in most of China until a gradual or slight decrease in 2011 or 2012 (Cui et al., 2016).  used ground-based NO 2 observations to assess the effectiveness of pollution control policies in a south-western city cluster and revealed fluctuations in NO 2 mixing ratios from 2008 to 2013, followed by an irregular declining trend after 2013. All these studies reported that NO x and SO 2 mixing ratios have been effectively controlled in China despite the increasing economic development over the past decades.
The Yangtze River Delta (YRD) region is located in the central-eastern region of China, which has the largest economic output in China and has the sixth-largest urban agglomeration in the world. The region covers an area of 359 100 km 2 and has a population of 224 million, accounting for 16.08 % of the country's population (Fang et al., 2020).
Because of increases in population, urbanization and industrialization in recent decades, the air pollution in the YRD has exhibited complex and regional characteristics Wang et al., 2019), and the YRD has become one of the most polluted regions in the world (Xie, 2017b), with NO x and SO 2 being the main factors that influence air quality in the region (Yang and Luo, 2019). Xu et al. (2008) compared observational data in [2005][2006] with those 10 years earlier and concluded that as early as the mid-1990s, SO 2 and NO x mixing ratios had already become considerably high at the background station in the YRD; since then, anthropogenic emissions have caused a substantial increase in the NO x concentration, making NO x another major pollutant in addition to SO 2 . The implementation of pollution control policies and continual innovation in SO 2 pollution control technology have mitigated SO 2 pollution in the YRD, resulting in a consistent decrease in SO 2 mixing ratios (Qi et al., 2012); however, NO x mixing ratios remain high (Shi et al., 2018).
In this paper, we present 11 years of (2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016) surface NO x and SO 2 observation data from the Lin'an regional atmospheric background station. We analysed the long-term variations in NO x and SO 2 and their influencing factors in the YRD background area to (1) assess the effectiveness of pollution control in the area and (2) provide a scientific basis and reference for future pollution control strategies.

Site information
The Lin'an regional atmospheric background monitoring station (30 • 18 N, 119 • 44 E; 138.6 m a.s.l.; referred to as LAN) is located in Lin'an District, Hangzhou, Zhejiang Province ( Fig. 1) and is one of the regional atmospheric background stations operated by the China Meteorological Administration; it is also a World Meteorological Organization (WMO) Global Atmospheric Watch (GAW) member station. LAN is located on an isolated hilltop, surrounded by hilly and mountainous terrain, with no large villages within a 3 km radius. It is within the region of subtropical monsoon climate, with the most dominant wind direction from the north-east and the secondary from the south-west. The seasonal variations in meteorological elements, namely atmospheric pressure (P ), temperature (T ), wind speed (WS), relative humidity (RH), and rose maps of wind speed (WS) and wind direction frequency (WF), are presented in Fig. 2.

Observations and quality control methods
At the LAN station, observations of O 3 , NO x , SO 2 and CO are performed by an integrated observation and quality control system combining O 3 , NO x , SO 2 and CO analysers as well as calibration equipment and ancillary materials, such as standard gases and zero air supply . NO x and SO 2 were measured using a Model 42C-TL tracelevel chemiluminescent analyser and a Model 43C-TL tracelevel pulsed fluorescence analyser (Thermo Fishier Scientific, MA, USA), respectively. In the Model 42C-TL tracelevel chemiluminescent analyser, NO 2 is converted to NO by a molybdenum NO 2 -to-NO converter heated to about 325 • . The converter efficiency was checked annually using gasphase titration (GPT). If the converter efficiency is less than 96 %, replace the converter. Data are recorded as 5 min averages. The meteorological parameters (WS, wind direction, T and RH) for a given period were obtained from the routine meteorological observations at the station. The main objective of operational observations of reactive gases at regional background stations is to obtain accurate trends in the measured reactive gases, for which reliable and comparable data are essential. Therefore, strict quality control measures were implemented during the observation process . The quality control measures mainly included the following: (1) daily zero and span checks (automatic); (2) monthly multi-point calibrations (≥5 points, including zero); (3) comparisons of reference SO 2 / N 2 and NO / N 2 gas mixtures to the standards of the National Institute of Standards and Technology before and after their usage (periodically) to ensure data traceability; (4) instrument self-diagnosis, manual testing, checking and maintenance (US EPA, 2017); and (5) data correction according to the quality control results, especially the results of zero and span checks and multipoint calibrations.
From 1 January 2006 to 31 December 2016, a total of 93 759 and 90 453 valid hourly average data points were obtained for NO x and SO 2 , respectively. Missing data totalled 2673 h and 5979 h for NO x and SO 2 , respectively. The missing NO x data were mainly for the period from 2 to 13 February 2007 and from 24 July to 8 October 2012. The missing SO 2 data were mainly for the period from 23 September to 21 December 2013, from 8 to 26 May 2014 and from 17 October 2014 to 24 January 2015.

Data processing methods
1. The daily means of NO x and SO 2 were calculated using the hourly average data, and only daily mean data calculated from at least 18-hourly data were used as valid daily means. The monthly means of NO x and SO 2 were calculated from the valid daily average data and considered valid if they were based on at least 21 valid daily averages (or at least 17 valid daily averages in February). Annual means were calculated on the basis of the complete monthly mean data each year. If a month's mean data were unavailable, we used an interpolating value from the corresponding monthly means in different years during the observation. In China, spring is from March to May, summer is from June to August, autumn is from September to November, and winter is from December to February.
2. Monthly satellite-based NO 2 OMI data were provided by Lin's research group at Peking University; the data were retrieved using an optimized inversion algorithm (Lin et al., 2014(Lin et al., , 2015. A grid range of 27. 125-35.875 • N and 115.125-122.875 • E was selected to cover the entire YRD region.

Concentration-weighted trajectory method
We used the concentration-weighted trajectory (CWT) method to identify potential source areas (PSAs) of NO x and SO 2 because this method can effectively distinguish the relative strength of potential sources (Xin et al., 2016). In the CWT method, the study area is divided into i × j small grids with equal size, and each grid (i, j ) is assigned a weighted concentration according to the following equation: where k denotes the indicator of a trajectory, m denotes the total number of trajectories, C k denotes the concentration observed when trajectory k arrives, and τ ij k is the residence time of trajectory k in the ij th grid cell. To reduce errors in the more distant grids, an empirical weighting factor W ij is introduced (Deng et al., 2020), with the following equation: 3n ave < n i, j < 1.5n ave 0.42 (1.5n ave < n i, j < n ave ) 0.05 n i, j < n ave . (3) Here, where D denotes the number of days included, t denotes the number of trajectories per day, n denotes the trajectory endpoints of each trajectory, and i × j denotes the total number of grids. We used a hybrid single-particle Lagrangian integrated trajectory model (Hysplit4.9) from the National Oceanic and Atmospheric Administration, USA, to calculate the 24 h backward trajectories at 10 m above ground level over LAN during 2006-2016; the NCEP-NCAR reanalysis meteorological data set (https://ready.arl.noaa.gov/archives.php, last access: 14 October 2020) was used to calculate the trajectories and atmospheric mixed-layer heights. The computed backward trajectories were subsequently processed using the TrajSat plug-in for CWT in MeteoInfo software (Wang, 2014), covering the region located within 20-40 • N and 110-130 • E and with a grid size resolution of 0.5 • × 0.5 • .

Observational levels and comparison with other sites
The hourly average SO 2 mixing ratios ranged from 0.1 to 128.6 ppb, which were all below the GB3095-2012 secondary standard limit for SO 2 (190 ppb). The hourly average NO x mixing ratios at LAN ranged from 0.4 to 165.6 ppb, with NO 2 mixing ratios ranging from 0.2 to 106.8 ppb. Only 3 h data exceeded the secondary standard limit value for NO 2 (106 ppb) as stated in the national ambient air quality stan-dard (GB3095-2012). It should be mentioned that the measurement of NO 2 was via conversion to NO by a molybdenum NO 2 -to-NO converter heated to about 325 • C, which was known to suffer from the interference of other NO y compounds such as PAN and HNO 3 (Steinbacher et al., 2007;Jung et al., 2017). This implies that the measured NO 2 mixing ratios were higher than actual values. However, it is impossible to quantify the overestimated parts due to the lack of other information. The interference might be enhanced with the increasing ratios of PAN to NO x (PAN / NO x ). Qiu et al. (2020) reported an increasing PAN / NO x from 2011 to 2018 at a background site in the North China Plain, but it is not clear if there was a similar increase in PAN / NO x in the YRD. During the transport of air masses to the background site, HNO 3 should be reduced by deposition or partitioning in the particulate phase and intercepted by filters before NO x was measured. Since NO z (NO y -NO x ) was produced by NO x oxidation, the overestimation of NO x by partial conversion of NO z , in turn, might be a positive offset in the difference between the measured mixing ratios and the emission of NO x when discussing their long-term trends. Table 1 presents annual statistics of the NO 2 , NO x and SO 2 mixing ratios observed at LAN between 2006 and 2016. The overall average mixing ratios with ± 1 standard deviation of for NO x (NO 2 ) and SO 2 from 2006 to 2016 were 13.6 ± 1.2 (12.5 ± 4.6 ppb) and 7.0 ± 4.2 ppb, respectively, with the highest NO x (NO 2 ) value being observed in 2012 and the highest SO 2 in 2006. NO 2 was the dominant form of NO x , accounting for 82.2 % of NO x (according to the slope value from the reduced major axis regression on hourly average NO 2 and NO x data). The average NO 2 mixing ratio was 12.5 ± 4.6 ppb, which was below the primary annual limit of 21.2 ppb in GB3095-2012. Some information on NO 2 (NO) can be seen in the Supplement (Table S1). The average SO 2 mixing ratio from 2006 to 2016 is close to the primary annual limit of 7.6 ppb in GB3095-2012. However, the annual average SO 2 mixing ratios (10.6-14.6 ppb) from 2006 to 2008 were much higher than the limit of the primary standard though lower than the limit of the secondary standard (22.8 ppb). Table 2 compares the levels of NO x and SO 2 mixing ratios at LAN with those corresponding SO 2 / NO x ratios at other background stations in seven geographic regions of China: North, East, South, Northeast, Northwest, Southwest and Central China. The NO x mixing ratio at LAN was slightly higher than that at Shangdianzi (12.7 ± 11.8 ppb) in North China; equal to that at Dinghushan (13.6 ppb) in southern China; and much higher than those at Wuyishan (2.70 ppb) in East China, Fukang (8.3 ppb) in Northwest China, Changbai Mountain (4.7 ppb) in Northeast China, Jinsha (5.6 ± 5.5 ppb) in Central China and south-western Mount Gongga (0.90 ppb). These results indicate that LAN recorded the highest level of NO x among the regional atmospheric background stations in China, which could be attributed to the developed economy of the YRD region. The SO 2 mixing ratio at LAN was close to that at Shangdianzi (7.6 ± 10.2 ppb) in North China; higher than that at Dinghu Mountain (6.5 ppb) in South China; and much higher than those at Wuyishan (1.48 ppb) in East China, Changbai Mountain (2.1 ppb) in Northeast China, Fukang (2.2 ppb) in Northwest China, Mount Gongga in Southwest China (0.19 ppb) and Jinsha (2.8 ± 5.5 ppb) in Central China. The regional difference in NO x and SO 2 was closely related to the diverse levels of economic development in China's regions because it was broadly characterized by a higher level in the eastern than in central and western regions. The SO 2 / NO x ratio at LAN was at a high level in China, which reflects the different energy structures to some extent. Figure 3 illustrates the average seasonal variations in NO x and SO 2 mixing ratios at LAN. The maximum monthly average mixing ratios of NO x and SO 2 were observed in December and January, at 23.5 ± 4.4 and 11.9 ± 6.2 ppb, respectively. The minimum values both occurred in July, at 7.1 ± 0.8 and 2.8 ± 2.3 ppb, respectively. The average monthly variations in NO x exhibited significant correlations with the monthly NO 2 satellite data (R 2 = 0.82, P <0.001). Seasonal variation patterns of NO x and SO 2 look alike, showing a concave shape with its minimum in summer. The highest mixing ratios occurred in winter (NO x : 19.5 ppb; SO 2 : 10.1 ppb), followed by spring (NO x : 13.4 ppb; SO 2 : 7.8 ppb), autumn (NO x : 13.6 ppb; SO 2 : 6.7 ppb) and summer (NO x : 8.1 ppb; SO 2 : 3.3 ppb). The monthly average mixing ratios of both NO x and SO 2 showed a dip in February -a phenomenon also observed in NO x and SO 2 Xue et al., 2020) and NO − 3 and SO 2− 4 in PM 2.5 in Shanghai (Duan et al., 2020). The source emission inventory data indicated that NO x and SO 2 emissions from industry, transportation and coal-fired power plants were all lower in February than in January and March throughout China (Li et al., 2017), which may be related to decreased emissions due to lower economic activity during the Chinese Spring Festival. In addition, the higher RH in February (Fig. 2) might have led to higher NO x and SO 2 removal rates. Figure 4 shows the annual and seasonal average diurnal variations in NO x and SO 2 at LAN from 2006 to 2016, along with the annual average diurnal variations in NO x and SO 2 at some other sites in the YRD. The overall diurnal profile of NO x displayed a double-peak and double-valley pattern (Fig. 4a). The valley values occurred at 05:00-06:00 and 13:00 (all times in the text are UTC+8), with mixing ratios of 12.3 and 10.0 ppb, respectively, and the peak values occurred at 09:00 and 19:00, with mixing ratios of 13.1 and 14.4 ppb, respectively. Surrounding areas -such as Chongming, Pudong (Xue et al., 2020) and Xujiahui (Gao et al., 2017) in Shanghai; Hangzhou  in Zhejiang Province; and Nanjing  in Jiangsu Province -also exhibited a double-peak and double-valley type of average diurnal variation in NO x (Fig. 4a), indicating a regional NO x pollution characteristic. However, at most atmospheric background stations, the average diurnal variations in NO x exhibited a single-peak and single-valley pattern, such as those at Xinglong in North China (Yang et al., 2012), Tianhu in the Pearl River Delta (Shen et al., 2019), Dae Hung District in Seoul (South Korea) (Pandey et al., 2008) and Mount Cimone in Italy (Cristofanelli et al., 2016),  suggesting a more complex anthropogenic influence in the YRD region. In summer, the seasonal average diurnal variation in NO x showed a morning peak at 08:00, 1 to 2 h earlier than in other seasons (Fig. 4c).

Diurnal variations
SO 2 at LAN showed relatively small average diurnal variation (Fig. 4b), with higher mixing ratios from midnight to noontime and lower ones during later afternoon and evening. The average diurnal amplitude of SO 2 at LAN was much smaller than those found in Nanjing and Jiaxing. The seasonal average diurnal profiles of SO 2 at LAN were similar to the annual average one except for that in winter, which had a peak around noon (Fig. 4d).
The diurnal variation in pollutants emitted at ground level is closely related to the intensity of emissions, atmospheric transport, diurnal development in boundary layer height and atmospheric photochemical reactions (Resmi et al., 2020). The mixed-layer depth (MLD) was much lower at night than during the daytime, as shown in Fig. 4b. Low MLDs at night are not conducive to pollutant dispersion, whereas high MLDs during the daytime are conducive to pollutant dispersion. This day-night difference in the MLD is one of the factors causing lower levels of SO 2 and NO x during afternoon hours. Photochemistry during the daytime also contributes to rapid chemical transformation of SO 2 and NO x , which results in low NO x and SO 2 mixing ratios in the afternoon. Overall, the morning peak of NO x was lower than the evening peak; the morning peak of SO 2 was higher than the evening subpeak; and the morning peak of SO 2 was not as protruding as and occurred slightly later than that of NO x , reflecting the differences in their sources. The morning peak of NO x may be influenced by vehicle emissions during the morning rush hour, and the early peak of SO 2 may be more influenced by vertical changes during the developing mixedlayer depth (Qi et al., 2012). The evening peaks of NO x and SO 2 were relatively similar because both were closely related to the MLD decrease and for NO x likely also vehicle emissions during the evening rush hour.

Influence of meteorological factors
Changes in meteorological factors have considerable effects on the levels of air pollutants. In this section, we investigate the influences of meteorological factors on the variations in NO x and SO 2 mixing ratios through statistical plots showing relationships between pollutant concentrations and meteorological factors as well as correlation analysis. The variation characteristics of hourly average mixing ratios of NO x and SO 2 along with meteorological parameters are presented in The variation characteristics of NO x and SO 2 with WS ( Fig. 5a, b) were consistent during period I, showing decreases in NO x and SO 2 with increasing WS. Higher WS facilitated the dilution of NO x and SO 2 and vice versa. How-ever, the situation for SO 2 was different during period II and III, when the SO 2 level was stable with the change in WS. The correlation of T between the two pollutants varied considerably, with the SO 2 mixing ratios decreasing nearly monotonically with increasing T (Fig. 5d), whereas NO x increased with increasing T in the low temperature range and decreased with increasing T in the high temperature range (Fig. 5c). Figure 5c indicates a positive correlation between NO x and T in winter and negative correlations in other seasons, but the positive correlation in winter is weak and insignificant (Table 3). Pandey et al. (2008) reported that low T might facilitate the increase in NO x emissions from motor vehicle exhaust. The variations in NO x and SO 2 with RH ( Fig. 5e, f) exhibit a convex pattern, and the former patterns in three different periods are very consistent, but the latter ones are not at low RH. The correlation between SO 2 and RH was stronger than that of NO x and RH ( Table 3). The variation characteristics of NO x and SO 2 mixing ratios with the MLD exhibited diverse patterns (Fig. 5g, h). The mixing ratio of NO x decreased with increasing MLD. However, the SO 2 levels during period II and III remained nearly stable in the whole MLD range, and a slight decline in SO 2 with increasing MLD was only observed during period I. The difference in NO x and SO 2 mixing ratios with the MLD implies that the NO x sources mostly impacting the LAN site should be mainly in the near-surface layer, such as emissions from motor vehicles and small burners, whereas SO 2 may originate from the vertical exchange of elevated sources transported in the higher altitude layer (200-1300 m). Figure 6 displays the rose diagrams of NO x and SO 2 mixing ratios in different seasons. There are some seasonal differences in the dependence of NO x and SO 2 on wind direction. In summer, the high mixing ratios of NO x and SO 2 were mainly from the NW-NNE and SSW-NW sectors, respectively (Fig. 6b). In other seasons, relatively high NO x and SO 2 values were mainly from the N-ENE and S-WSW directions, respectively, under the influences of the dominant and subdominant wind directions (Fig. 2b, d). Overall, NO x and SO 2 observed at LAN originated mainly from the NW-ENE and SSW-NW sectors, respectively. However, this result provides only little information about the actual geographic distributions of major NO x and SO 2 sources influencing LAN. Therefore, we used the CWT method to identify the PSAs for NO x and SO 2 . Figure 7 presents the areas from which NO x and SO 2 observed at LAN originated. Although the PSAs covered the entire YRD, the PSAs for the highest NO x and SO 2 levels appeared mainly in the eastern coastal region, which is closely related to the booming local economy. More obvious provincial differences were observed in a higher PSA for NO x than that for SO 2 . Temporally, the high PSA (>10 ppb) of NO x and SO 2 was most extensive in winter, followed by spring and autumn, with the least extensive PSA in summer. The NO x PSAs over coastal areas were more extensive than those for SO 2 in each season. The YRD is one of the five major port clusters in China;  thus, this region's ship emissions might be a major cause of this difference (Fan et al., 2016;Wan et al., 2020). The CWT analysis indicated that SO 2 was mainly influenced by industrial emissions from inland areas, whereas NO x was mainly influenced by both inland and marine traffic.
3.5 Long-term variations in NO x and SO 2 mixing ratios  (Fig. 8a). The decreasing rate was less than that found in urban Shanghai (−2.1 ppb/yr; Gao et al., 2017). Selecting 2006 as the base year, we compared the annual percentage change in NO x at LAN (−0.49 %/yr) during 2006-2016 with those of other regions over the same period. The Ecological and Environmental Status Bulletin (Department of Ecology and Environment of Shanghai city, 2021a-j; Department of Ecology and Environment of Zhejiang Province, 2021a-k; Department of Ecology and Environment of Jiangsu Province, 2021a-k) reported a similar change of −0.45 %/yr in the YRD region (without data for Anhui Province), reflecting the suitable regional representativeness of LAN. The annual percentage decrease in NO x at LAN and in the YRD was much smaller than those in many regions -for example, the Pearl River  Davuliene et al., 2021) and in New York City in the United States (−3.46 %/yr; Squizzato et al., 2018) -but more favourable than those in some other regions, such as Wuhan in China (+2.08 %/yr;  and Amersfoort (+6.50 %/yr) and Louis Trichardt in South Africa (+1.85 %/yr; Swartz et al., 2020b). Compared with other background regions in China, the annual change in NO x at LAN was less favourable than that in North China (−3.34 %/yr) with a base year of 2005 (Bai et al., 2015) and more favourable than that in Northwest China (+12.98 %/yr) with a base year of 2010 . Figure 8 also presents the NO x emission data from the China Ecological Environment Bulletin in different years. The change in the annual average NO x mixing ratio was highly correlated with the total NO x emissions (R 2 = 0.92, P <0.001) and total industrial emissions (R 2 = 0.94, P <0.001) in the YRD region. The peak surface NO x mixing ratio was observed in 2011. Since China began to control and reduce NO x emissions as part of the 12th Five-Year Plan plants and heavy industry operations , resulting in a decrease in annual NO x emission since 2011. As seen in Fig. 8a, the total and the industrial NO x emission showed increasing trends with 5.84 %/yr (R 2 = 0.91, P = 0.011) and 6.3 %/yr (R 2 = 0.91, P = 0.006), respectively, from 2007-2011 and −7.63 %/yr (R 2 = 0.91, P = 0.003) and −8.30 %/yr (R 2 = 0.84, P = 0.011), respectively, from 2011-2016. The seasonal long-term trends of NO x always resembled the annual trend. While seasonal NO x mixing ratios in winter, autumn and spring increased before 2011 and then decreased, just like the annual NO x mixing ratio did, the seasonal NO x mixing ratio in summer exhibited a nearly monotonic decreases from 2006 to 2016 at 0.11 ppb/yr (R 2 = 0.20, P = 0.09) (Fig. 8c). Regarding the seasonal linear fitting trends, the highest increasing and declining trends were observed in winter (+1.29 ppb/yr, R 2 = 0.52, P = 0.06; −2.33 ppb/yr, R 2 = 0.94, P <0.01), followed by autumn (+1.24 ppb/yr, R 2 = 0.65, P = 0.02; −0.41 ppb/yr, R 2 = 0.12, P = 0.30) and spring (+0.31 ppb/yr, R 2 = 0.93, P <0.001; −1.16 ppb/yr, R 2 = 0.76, P = 0.09). We found a significant correlation (P <0.05) between surface NO 2 mixing ratio and OMI NO 2 vertical column density over the YRD (Fig. S3b). To better compare the changes in the two over the same period, we have fitted a linear fit to the data from 2006 to 2011 and from 2011 to 2016, respectively (Fig. S3a). The surface and the OMI NO 2 increased at 2.23 %/yr (R 2 = 0.264, P = 0.17) and 5.87 %/yr (R 2 = 0.855, P <0.01) (based on 2006), respectively, during the up period and decreased at −4.98 %/yr (R 2 = 0.823, P <0.01) and −4.22 %/yr (R 2 = 0.897, P <0.01), respectively, during the declining period.
Annual mean SO 2 mixing ratios revealed a significant decreasing trend (−0.99 ppb/yr, R 2 = 0.92, P <0.001) during 2006-2016 (Fig. 8b). The annual decreasing rate of SO 2 at LAN (−8.27 %/yr) was more rapid than those in the whole YRD (−6.65 %/yr), in the background area in North China (−0.78 %/yr; Bai et al., 2015) and in Northwest China (−5.4 %/yr; . Different from NO x , the annual average of SO 2 at LAN decreased more rapidly than in most of the aforementioned regions (Table 4), which demonstrates the effectiveness of the policies in controlling SO 2 emission during the observation period in the YRD.
The change in the annual SO 2 mixing ratio was closely correlated with changes in thermal power plant SO 2 industrial emission (R 2 = 0.89, P <0.001), industrial SO 2 emission (R 2 = 0.76, P <0.001) and total SO 2 emission (R 2 = 0.78, P <0.001) in the YRD (Fig. 8b). In 2011, the SO 2 mixing ratio rebounded slightly, with an increase of 9 % compared with the value in 2010. This seemed to be consistent with the variation in industrial SO 2 emission. The weakening impact of the global financial crisis and the recovery of industry in the YRD region may explain this slight rebound in SO 2 emissions (Xie, 2017b). Seasonally, the SO 2 mixing ratio exhibited the strongest decreasing trend (−1.69 ppb/yr, R 2 = 0.90, P <0.001) in winter, followed by spring (−1.05 ppb/yr, R 2 = 0.97, P <0.001) and autumn (−0.99 ppb/yr, R 2 = 0.93, P <0.001), with the smallest de-   In the annual statistics, the 95th and 5th percentile of the pollutants' concentrations can be regarded as influenced by polluted and clean air masses, respectively. The annual trends of the 95th percentile of NO x and SO 2 (Fig. 9a) exhibited similar patterns to the corresponding trends in annual average mixing ratios (Fig. 8a, b), but the peak of the 95th percentile of NO x occurred in 2012 instead of in 2011. Hao and Song (2018) noted that the NO x emissions from vehicles peaked in Hangzhou and Ningbo in 2012, which may explain the peak of the 95th percentile occurring later than that in the annual data. Moreover, the 95th percentile of the SO 2 mixing ratio decreased at a remarkable rate (−8.9 ppb/yr) from 2007 to 2009, which is approximately 2.8 times as strong as the overall rate of decrease during the 11-year period (−3.2 ppb/yr). Substantial decreases were also found in the 95th percentiles of the CO mixing ratio  and the NO x mixing ratio from 2007 to 2009 at LAN. It is highly possible that this phenomenon was caused by reduced industrial productions and related emissions following the 2008 global financial crisis. As displayed in Fig. 9b, the level of NO x in cleaner air mass arriving at LAN exhibited an increasing trend, with a rate of +0.17 ppb/yr, from 2006 to 2014 (R 2 = 0.86, P <0.001) and then declined after 2014. This is inconsistent with the trend of the 95th percentile of the NO x mixing ratio, suggesting that the polluted and relative clean air masses arriving at LAN were impacted by different emission sources of NO x . Interestingly, the 5th percentile of the NO x level was significantly corre- Figure 9. Annual variations in the 95th percentile concentration (a) and the 5th percentile concentration (b) of NO x and SO 2 at LAN; data of NO 2 road emissions in the YRD are obtained from the REASv3.2 data sets in the Regional Emission Inventory in Asia (Kurokawa and Ohara, 2020). lated (R 2 = 0.74, P <0.001) with the road emissions of NO 2 in the YRD (Kurokawa and Ohara, 2020), suggesting that the lower end of NO x mixing ratios was mainly determined by long-range transported background air containing NO x from road emissions, while the high end was mainly associated with emissions from industrial production as well as power generation. The level of SO 2 in cleaner air masses exhibited a decreasing trend at a rate of −0.2 ppb/yr (R 2 = 0.61, P <0.01). Figure 10 displays the scatter plot of the daily average SO 2 and NO x mixing ratio during period I, II and III at LAN. Reduced major axis regressions were performed on three data subsets. The daily mean mixing ratios of NO x were significantly (R 2 = 0.29, P <0.001) and positively correlated with those of SO 2 . The ratios of SO 2 to NO x (SO 2 / NO x ) were 0.96, 0.53 and 0.33 (slopes in the regression lines) during period I, II and III, respectively. The decreasing SO 2 / NO x suggests that SO 2 emissions were more efficiently reduced than NO x emissions. Such a change in emis-sion ratio not only affected ambient SO 2 / NO x but also the ratios of sulfate/nitrate in PM 2.5 in Shanghai from 2009 to 2012  and SO 2− 4 / NO − 3 in rainwater in Hangzhou (Yang, 2018;Xu et al., 2019). These results indicate that NO x has been gaining a more important role in the processes of precipitation acidification and secondary inorganic aerosol formation in the YRD region. Therefore, NO x emission reduction should be further strengthened in subsequent air pollution control measures and legislation in the YRD region. Figure 11 reveals the average diurnal variations in NO x and SO 2 during the period I, II and III. During these three periods, the average diurnal curves in NO x exhibited a valley around 13:00, with minimum values of 7.5, 11.2 and 9.2 ppb, respectively. The morning and evening NO x peaks, which occurred at 09:00 and 19:00, respectively, became increasingly distinct over time (Fig. 11a, c, e). The morning and evening peak NO x values were 9.8 and 10.9 ppb during period I, 14.6 and 15.8 ppb during period II, and 12.3 and 13.6 ppb during period III. The gradual protruding of the morning and evening peaks should be mainly caused by increasing vehicle emissions during the morning and evening rush hours. According to the 2010 Annual Report on China's Motor Vehicle Pollution Prevention and Control, the state introduced a series of policies to promote automobile and motorbike ownership in response to the international financial crisis and to ensure economic growth; these policies effectively stimulated the automobile market (Mi and Qin, 2011;Hao and Song, 2018) and led to an increase in vehicle emissions and atmospheric oxidation in the YRD region . Thus, the NO x mixing ratios around the morning and evening peaks were much higher than those at night during period II (Fig. 11e), which differs much from the pattern during period I (Fig. 11a). The disappearance of the small peak around 01:00 at night during 2012-2016 may be related to the introduction of stricter air pollution control policies for factories that emit at night. Small peaks in NO x and SO 2 occurred between 01:00 and 02:00, which might be related to nighttime emissions from unscrupulous enterprises (Fan et al., 2016) or more production activities with lower electricity prices after midnight in response to the financial pressure of the 2008 economic crisis and the corresponding increase in electricity prices for industrial users (Sun et al., 2018). In spite of these two reasons, however, it is really hard to tell exactly why these small peaks dominate after midnight.
The average diurnal variation curve of SO 2 at LAN during period I (Fig. 11b) is of the single-valley type, with an average valley mixing ratio of 6.5 ppb. After 2010, the peak shape changed from the single-valley type to the double-peak and double-valley type (Fig. 11d, f). The valleys of SO 2 during period II occurred at 06:00 and 15:00, with average mixing ratios of 5.2 and 4.7 ppb, and the peaks occurred at 10:00 and 19:00, with average mixing ratios of 5.9 and 5.3 ppb, respectively. The NO x and SO 2 evening peaks occurred at the same time (19:00), but the SO 2 morning peak time was 1 h later than the NO x morning peak (09:00), indicating that the NO x and SO 2 morning peaks were influenced by different sources, whereas the evening peaks were from similar sources. The formation of the SO 2 morning peak may be mainly related to the vertical exchange during the development of the atmospheric boundary layer and the air in the upper layer with a higher SO 2 mixing ratio than that at the surface draining down. The formation of the evening peaks of NO x and SO 2 may be mainly related to the increase in motor vehicle and residential sources emissions, which are stronger in the rush and cooking hours, and those of SO 2 may be probably more due to the reduction in power plant emissions. Compared with those during period II, the SO 2 mixing ratios at the morning and evening peaks in period III were approximately 3 ppb lower, suggesting that the large emitters that release SO 2 all the time were emitting less and less.

Conclusions
In this study, we characterized the seasonal and diurnal variations and analysed the long-term trends in NO x and SO 2 mixing ratios in the YRD background area during the period of 2006-2016. We also tried to understand the variations and trends in terms of the changes in emissions and meteorological conditions. The hourly average mixing ratios of NO x (NO 2 ) and SO 2 at the LAN background station varied in the ranges of 0.4-165.6 (0.2-106.8) and 0.1-128.6 ppb, respectively. The levels of NO x and SO 2 were highest in winter, followed by spring and autumn, and lowest in summer. Although a significant correlation was observed between the daily average mixing ratios of NO x and SO 2 (R 2 = 0.29, P <0.001), their average diurnal variation characteristics differed from each other, with morning peaks in SO 2 occurring later than in NO x .
The annual average mixing ratio of NO x (NO 2 ) fluctuated upwards between 2006 and 2011 (+0.31 ppb/yr, P = 0.16) (+0.27 ppb/yr, P = 0.17), with a mean value of 13.8 ppb, and then began to decrease significantly from 2011 to 2016 (−0.78 ppb/yr, P <0.01) (−0.70 ppb/yr, P <0.01), with a mean value of 13.7 ppb (12.5 ppb). The annual average mixing ratio of NO x was significantly correlated with the in-dustrial (R 2 = 0.81, P <0.001, 2006-2016) and total (R 2 = 0.88, P <0.001, 2006-2016) NO x emissions in the YRD; so were NO x mixing ratios in LAN with OMI NO 2 column density over YRD (Fig. S3b, R 2 = 0.61, P <0.01). The annual 95th percentile of NO x mixing ratios followed a similar trend as the annual average, whereas the 5th percentile levels fluctuated upwards at +0.17 ppb/yr from 2006 to 2014, reflecting the increasing regional background level of NO x in the YRD during those years, which was related to the continued increase in vehicle numbers in the YRD. The annual average mixing ratio of SO 2 exhibited a rapid and significant decreasing trend (−0.99 ppb/yr, P <0.001) and was closely correlated to total SO 2 emission (R 2 = 0.78, P <0.001), total SO 2 industrial emission (R 2 = 0.76, P <0.001) and total thermal power plant SO 2 industrial emission (R 2 = 0.89, P <0.001) in the YRD. The reduced emissions resulted from the strong and effective introduction of national control policies. The yearly decrease in SO 2 / NO x ratios suggests a more effective reduction in SO 2 than in NO x . Thus, NO x emission control needs to be further strengthened in the future.
We found gradual changes in diurnal patterns of both gases. After 2010, both NO x and SO 2 showed diurnal patterns with two peaks and two valleys. The morning peak of NO x occurred at approximately 09:00, earlier than that of SO 2 (10:00), and the evening peak occurred at the same time as SO 2 (19:00). The morning and evening peaks of both gases protruded gradually. This phenomenon can hardly be attributed to changes in meteorological conditions (such as the MLD). We believe that changes in major sources of NO x and SO 2 should be the cause, with increasing NO x emission from vehicles resulting in higher NO x peaks during rush hours and reduced SO 2 emissions from power plants and other large point sources making the SO 2 peaks relatively protruding.
Author contributions. QY wrote the paper; WL and XX developed the idea, formulated the research goals and edited the paper. QM and JY carried out the measurement of NO x and SO 2 and analysed the meteorological data.
Competing interests. The contact author has declared that neither they nor their co-authors have any competing interests.

Disclaimer.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.