Measurement report: Fast photochemical production of peroxyacetyl nitrate (PAN) over the rural North China Plain during cold-season haze events

Photochemical pollution over the North China Plain (NCP) are attracting considerable concern. Peroxyacetyl nitrate (PAN) is usually viewed as the second most important photochemical pollutant featuring high mixing ratios during warm seasons. Our observations at a background site in the NCP identified high PAN concentrations even during coldseason haze events. The abrupt increasing rates of PAN by 244% and 178% over the morning hours (8:00–12:00) on 10/20 20 and 10/25, 2020 were 10.6 and 7.7 times those on clean days. The pollution days were characterized by higher temperature and humidity, accompanied by anomalous southerlies. Enhanced local photochemistry has been identified as the dominant factor that controls PAN increase in the morning at the rural site, as the time when prevailing wind turned to southerlies was too late to facilitate direct transport of PAN from the polluted urban region. By removing the effect of direct transport of PAN, we provide a quantitative assessment of net PAN chemical production rate of 0.45 ppb h-1 on the polluted morning, 25 also demonstrating the strong local photochemistry. Using observations and calculated photolysis rates, we find that oxidation of acetaldehyde by hydroxyl radical (OH) is the primary pathway of peroxyacetyl radical formation at the rural site. Acetaldehyde concentrations and production rates of HOx (HOx = OH + HO2) radical on pollution days were 2.8 and 2 times that on clean days, respectively, leading to the abrupt increase of PAN in the morning. Formaldehyde (HCHO) photolysis dominates the daytime HOx production thus contributing to fast photochemistry of PAN. Our observational results fully 30 explain the cause of rapid increase of PAN during cold days at a rural site of the NCP, as well as provide the evidence of important role of HCHO photolysis in secondary pollutants at lower nitrogen oxide emissions. This highlights the imperative to implement strict volatile organic compounds controls out of summer seasons over the NCP. 1 https://doi.org/10.5194/acp-2021-359 Preprint. Discussion started: 23 June 2021 c © Author(s) 2021. CC BY 4.0 License.


Introduction 35
Since the late 1960s, peroxyacetyl nitrate (PAN) has been identified as a key photochemical pollutant in the atmosphere, having adverse effects in human health and vegetation (Heuss and Glasson, 1968;Taylor, 1969). It is a secondary pollutant formed through reactions between peroxyacetyl radical (CH3C(O)O2, PA) and nitrogen dioxide (NO2) (Xue et al., 2014). The PAN can be thermally decomposed back to PA and NO2 which is the major removal pathway in the lower troposphere: villages. The mean PM2.5 concentration in 2020 at the SDZ site is 26 μg m -3 , much lower than present air quality standard in  ). An integrated observation experiment was performed at the SDZ site from 10/13 to 10/27, 2020, 100 including measurements of PAN, VOCs, O3, HONO, NOx, CO and photolysis rates. Besides, we also conducted PAN observations at an urban site in Beijing, located on the campus of the Minzu University of China (39. 95°N, 116.32°E). The urban site is between the second and third ring roads in downtown Beijing, mainly affected by traffic and residential sources.

Instruments and measurements
The detailed information of the instruments used in this study has been summarized in Table S1. PAN concentration is measured with an online gas chromatograph equipped with an electron capture detector (GC-ECD). The capillary column is mounted in a compact temperature-controlled oven whose temperature is fixed at 12°C. The ECD temperature is maintained 110 at 50°C ± 0.2°C. The time resolution is 5 minutes, and the detection limit is 20 ppt. More information about the configuration of this instrument has been presented in our previous work (Qiu et al., 2020a). A regular calibration check was conducted every month to guarantee the quality of the PAN results since August 2015, usually on a day with low PAN concentration.
The proton transfer reaction-time of flight-mass spectrometer (PTR-ToF-MS) is used to measure concentrations of 115 HCHO, CH3CHO, acetone, propene (C3H6) and isoprene. To obtain a higher signal/noise ratio of the VOC species, all mass spectra are stored at a relatively low time resolution of one hour. Detailed descriptions of the PTR-ToF-MS configuration and calibration methods are listed in Sheng et al. (2018). HONO measurements are carried out using a long-path absorption photometer (LOPAP-03, QUMA). The HONO gas is collected in the atmosphere and then absorbed by solutions. The sampling rates of gas flow and liquid flow are set to 1.3 L min −1 and 0.30-0.34 mL min −1 , respectively. We also calibrated 120 the LOPAP instrument every week to guarantee data quality during the observation period.
Online measurements of O3, NOx, PM2.5, and CO are also conducted using a UV photometric O3 analyzer (model 49i, Thermo Electron Corporation, USA), a NOx analyzer (model 42i, Thermo Electron Corporation, USA), a TEOM-1405 analyzer, and a cavity ring-down spectrometer (G2401, Picarro, Inc., USA), respectively. Photolysis rates (J), including J(O 1 D), J(HCHO), and J(HONO) are simultaneously measured by the PFS-100 Photolysis Rate Analyzer (Focused 125 Photonics Inc., China). The analyzer receives solar radiation with a quartz probe and transfers the radiation to the spectrum via optical quartz fibers. The spectrum data is evaluated and compared with reference data via a mathematical approach to obtain the photolytic rate.

Other data
As the SDZ site is also a national meteorology observatory, meteorological variables, including temperature (T), 130 relative humidity (RH), sea level pressure (SLP), wind direction and speed, were continuously measured during the observation period. In addition, we use hourly European Centre for Medium-Range Weather Forecasts Reanalysis v5 (ERA5) data (0.25°×0.25°) to assess the impact of atmospheric circulation on pollutant levels at the SDZ site. The ERA5 data are accessed from https://cds.climate.copernicus.eu/. In addition, concentrations of Ox (O3+ NO2) obtained from the Beijing Municipal Ecological and Environmental Monitoring Center (http://www.bjmemc.com.cn/) are also utilized. Here, we 135 average the mixing ratios over 8 stations in the urban region and at the MiYunshuiku (MY) site to represent the Ox levels in urban region and rural region near SDZ. Compared with recent studies in China (Table S2), the observed PAN concentration during autumn at the SDZ site is generally lower than PAN levels over the urban NCP (Zhang et al., 2017;Liu et al., 2018) and in southwestern China (Sun et al., 2020) but comparable to those in the suburban NCP region (Qiu et al., 2019a;Zhang et al., 2019). In addition, the PAN 145 level at the SDZ site is remarkably higher than those obtained from the southern coastal region (Zhu et al., 2018;Zeng et al., 2019a;Hu et al., 2020) and Tibet (Xu et al., 2018), implying a severe photochemical pollution level over the NCP on a regional scale. We find two pollution events (10/20 and 10/25-10/26) occurring at the SDZ site with hourly PAN concentrations in 155 excess of 3 ppb (Figure 2a). Meanwhile, daytime O3 concentrations during the two pollution episodes were 16−39 % higher than those on clean days ( Table 1). Similar increase was also found in PM2.5, which was strongly correlated with PAN with a correlation coefficient (R) of 0.9 during the observation period. Daily mean NOx concentrations on pollution days were 16-42% higher than clean days as shown in Table 1, implying the potential role of regional transport under unfavorable meteorological conditions. Abrupt increases of PAN and related pollutants could be confirmed on the mornings of 10/20 and 160 10/25. enhanced local photochemistry at the SDZ site on pollution days during autumn despite reduction in observed photolysis rate due to aerosol and cloud radiative effects.   Previous studies also reported rapid increases of PAN accompanied with PM2.5 enhancement during cold seasons over the NCP region (Liu et al., 2018;Zhang et al., 2019). For example, Zhang et al., (2019) showed that PAN concentration was doubled at noon during a haze episode at a suburban site in Beijing in comparison with that in the morning, and the 175 synchronous increases of PM2.5 and O3 concentrations were also found. Liu et al. (2018) also reported rapid growth of PAN during a wintertime pollution event in urban Jinan along with high PM2.5 concentration; however, they showed rather lower O3 concentration during winter haze days because of high NOx concentrations and intense NO titration effects in the urban region. Therefore, synchronous increases of PAN and PM2.5 can occur over the whole NCP region during cold days, while co-occurrence of O3 enhancement just exits in suburban and background regions where NOx emissions are rather low. 180

Meteorological conditions
On synoptic scale, PAN mixing ratios are largely influenced by meteorological conditions. Figure 3 shows the atmospheric circulation for clean days and two pollution days (10/20 and 10/25), focusing on variations of SLP and wind in the boundary layer. During the observation period, the SDZ site was affected by a high-pressure system in the west associated with northwesterly. Similar weather pattern was also identified on clean days, facilitating pollutant diffusion. 185 During the two pollution days, southwesterly prevailed over the NCP. On 10/20, the SDZ site was in the south of a strong low-pressure system, leading to southwesterly resulted from pressure gradient force. The southwesterly on 10/25 was caused and 10/25, indicating hot and wet weather conditions. Although higher temperature can promote thermal decomposition of PAN, it also accelerates photochemistry thus increases PAN mixing ratios. Higher RH has been proved to inhibit heterogeneous reactions of PAN on soot, leading to increase of PAN concentration in the atmosphere (Zhao et al., 2017).
Negative SLP anomalies and positive V anomalies on 10/20 and 10/25 could contribute to pollution accumulation and 200 transport to the north, coinciding with the ERA5 results shown in Figure 3. J(O 1 D) was reduced by 20−34% on the two pollution days due to aerosol and cloud radiative effects, which was unfavorable for photochemistry. As noted above, we can conclude that the meteorological conditions during pollution events are virtually conducive to formation and accumulation of PAN at the SDZ site, though reductions are identified in photolysis rates.  The Ox concentration at the MY site during 9:00−12:00 a.m. of 10/20 was slightly lower than that at the SDZ site (Figure   5c), implying the similar photochemical pollution level at the two sites. Consequently, despite the observed high PAN concentration at the urban site (Figure 5a), the abrupt increase of PAN on the morning of 10/20 was not likely caused by 225 direct PAN regional transport. On another pollution day of 10/25, the prevailing wind turned to southerlies after 12:00 a.m.. Direct PAN transport from urban region could not explain the fast PAN growth in the morning. In addition, we also exclude possible impact of PBL evolution in the morning on increases of PAN, because our previous observation in an urban city of the NCP region reported that nighttime PAN concentration in boundary layer was just 9.5% higher than surface-layer PAN concentration (Qiu et al., 2019a). The slightly higher PAN concentration in upper layer could not be the cause of rapid 230 increases of PAN on the mornings of pollution days.

235
To quantitatively assess the impacts of regional transport and local photochemistry on PAN levels at the SDZ site, we choose CO as a tracer. CO is chemically inert and greatly affected by anthropogenic sources, thus it can well represent the physical transport of pollutants from the urban region (Gao et al. 2005;Worden et al., 2013;. Detailed calculation method has been described in Method S1. It should be noted that the calculation of physical transport impact 240 aims at PAN that has been formed outside of the SDZ site. SDZ site. The positive PAN change rate (Chem + Phys) was only found in the morning (8:00−12:00), which was attributed to the high net chemical production rate (Chem) with a value of 0.14 ppb h -1 . In the afternoon (12:00−16:00) and evening (16:00−20:00), regional transport (Phys) by southwesterlies contributed to increases in PAN with rates of 0.05 ppb h -1 and 0.02 ppb h -1 , respectively. However, the negative net chemical production rates (−0.07 ppb h -1 and −0.06 ppb h -1 ) originating 245 from strong thermal loss rates (L[PAN], −0.57 ppb h -1 and −0.34 ppb h -1 ) completely overcome the PAN increases from transport. This evidence manifests local photochemical formation of PAN in the morning, coinciding with PAN's diurnal variation. The impacts of photochemistry and regional transport on PAN were both largely enhanced during pollution days (Figure 6 and Figure S1). The net chemical formation rate (Chem) was 0.45 ppb h -1 on the morning of pollution days, which was 6.3 times that on clean days. This again demonstrates that strong local photochemical reactions contributed most 250 to PAN enhancement on the mornings of the two pollution days instead of direct transport.

Impacts of precursors on rapid increase of PAN
A key question in explaining the abrupt increase of PAN at the SDZ site is how its precursors change. PAN is directly formed through the reaction between PA radical and NO2. As NO2 is much more abundant than PAN in the atmosphere, PA formation through VOC oxidation and photolysis may have greater impacts on PAN. As noted above, the dominant three  , where the photolysis rates J are estimated using the tropospheric ultraviolet and visible radiation (TUV) model described in Madronich and Flocke (1999). The acetone concentration is collected from PTR-ToF -MS measurements, and the MGLY concentration is obtained from Qiu et al., (2020c) in which the modified modeling MGLY concentration near the SDZ site in autumn of 2018 was about 0.012 ppb. As seen in Figure 7, CH3CHO oxidation by OH radical plays a dominant role in PA production at the 270 SDZ site rather than photolysis of acetone and MGLY. This result represents a combination of a relatively lower photolysis rate of acetone (~10 -7 s -1 ) and previously reported low MGLY concentration (~0.012 ppb) during autumn at the rural site (Qiu et al., 2020c). Our results are consistent with previous studies, in which they also confirm the dominant role of CH3CHO+OH in PA formation in eastern China (Zeng et al., 2019a;Zhang et al., 2020b) and even on global scale (Fischer, et al., 2014). 275

280
CH3CHO could be directly emitted to the atmosphere and also formed through oxidations of alkenes, such as C3H6. Case 2 (10/22−10/26). Here, we use the data at 8:00 a.m. to represent the VOC concentration level before intense photochemistry during the daytime. On 10/16, the NCP region was influenced by a high-pressure system with northerlies (Figure 4e), which was conducive to pollutant diffusion. From 10/17 to 10/20, persistent stagnant conditions with 285 southerlies during late morning to evening contributed to pollutant accumulation and transport from the urban region, though northerlies prevailed from night to morning. Thus, the C3H6 and CH3CHO concentrations during 10/16 to 10/20 both exhibited increasing trends, demonstrating the cumulative effect. Similar in appearance to that from 10/22 to 10/25 (Figure   8d-e), during which the C3H6 and CH3CHO concentrations increased by 180 % and 196 %. These indicate that cumulative effect with persistent stagnant weather conditions elevates the concentration level of PAN precursors, causing high VOC 290 levels before the abrupt increases occurred on the mornings of 10/20 and 10/25. Furthermore, the high PAN/NO2 ratios on 10/20 and 10/25 when abrupt increases of PAN appeared (Figure 8c, f, i) also enabled the identification of strong photochemistry with a relatively high precursor level. As shown in Figure 6, the net chemical formation rate on the morning of pollution days was 6.3 times that on clean days. However, the increase ratios for CH3CHO and C3H6 were just about 2.8 times higher than clean days (Figure 8g-h). That is to say, increase of VOC concentration could not fully explain the abrupt 295 increase of PAN, although it indeed promoted the photochemical formation of PAN during the two pollution events.

Impacts of radicals on rapid increase of PAN
In addition to VOC precursors, the observed strong chemical formation on the mornings of pollution days could also relate to enhanced OH concentration level and atmospheric oxidation capacity. Photolysis of HONO, HCHO, and O3 can provide the major source of HOx radicals (Schnell et al., 2009;Edwards et al., 2014;Tan et al., 2018;Li et al., 2021). Here, we present the observed time series of HOx production through photolysis of HONO, HCHO, and O3 using comprehensive 305 measurements over the rural NCP (Figure 9). The calculation method has been described in Method S2. Unlike HONO and O3, photolysis of HCHO directly produces HO2. Recycling of HO2+NO contributes to OH formation and HO2+HO2 consumes HO2 in the atmosphere. The remarkably higher production rate of NO+HO2 than HO2+HO2 (Table S3) during the observation period at the SDZ site reveals the fast production of OH through immediate reaction of NO+HO2. Thus, we use HOx here to represent the OH level. On average, a maximum HOx production rate (P[HOx]) of 6.5×10 6 molec cm -3 s -1 was 310 observed at noon (Figure 9b). The P[HOx] could be large during periods of pollutions days (Figure 9a). Figure 10 13 https://doi.org/10.5194/acp-2021-359 Preprint. Discussion started: 23 June 2021 c Author(s) 2021. CC BY 4.0 License. compares the P[HOx] during clean days and pollution days. The P[HOx] at noon of pollution days (11.0×10 6 molec cm -3 s -1 ) was approximately 2 times that on clean days, implying enhanced atmospheric oxidation capacity. In the conventional view, photolysis of HONO could provide the major source of OH radical during cold days over the 320 NCP region (Hendrick et al., 2014;Tan et al., 2018). Here we present the observational evidence of the dominant role of HCHO photolysis in daytime HOx productions during autumn over the rural NCP. Photolysis of HONO was vital in the early morning, but it became less important after 9:00 a.m. due to its fast decomposition with increasing sunlight (Figure 9b). On average, the P[HOx] through HCHO photolysis reached 4.6×10 6 molec cm -3 s -1 at noon, accounting for 71% among the three pathways (Figure 9b). During pollution days (10/20 and 10/25−10/26), the HCHO photolysis rate reached 8.3×10 6 molec 325 cm -3 s -1 at noon, which was 140% higher than that on clean days (Figure 10). Moreover, PAN was strongly correlated with HCHO (R 2 =0.87) (Figure 9c). It proves the similar source of PAN and HCHO, also demonstrates a potential impact of HCHO photolysis on PAN increase at this rural site with accelerated photochemistry. The sources of HOx radical during cold days have been a subject of recent interest in the field of atmospheric chemistry over the NCP region (Tan et al., 2018;Xue et al., 2020), and most of them reported the importance of HONO in OH radical and atmospheric oxidation capacity during 330 cold days over the NCP. By comparisons, our results at the SDZ site presents a much lower HONO concentration (average: 0.15 ppb) than Tan et al. (2018) and Xue et al. (2020). As the SDZ site is located in the north border of the NCP region, which is much cleaner than the suburban and rural site in Tan et al. (2018) and Xue et al. (2020), then less affected by shortlive species, such as NOx and HONO. For another, Tan et al. (2018) and Xue et al. (2020)   The source of HCHO at the SDZ site can be largely affected by aged air mass from the urban NCP region instead of biogenic source because the observed mean concentration of isoprene during pollution days was relatively low (0.2 ppb). As shown in Table S4, the observed mean HCHO concentration of 4.6±3.8 ppb at the SDZ site is lower than most of 345 observations in urban and suburban Beijing (3.2−29.2 ppb) (Pang et al., 2009;Duan et al., 2012;Zhang et al., 2012;Rao et al., 2016;Sheng et al., 2018;Yang et al., 2018;Qian et al., 2019;Zhou et al., 2019) and urban Guangzhou in summer (7.6 ppb) (Ling et al., 2017). But it is slightly higher than most observations in other southern China (2.1−5.6) ppb (Guo et al., 2016;Wang et al., 2017;Yang et al., 2019;Zeng et al., 2019b) and the background NCP region (3.5−3.7 ppb) conducted 3−6 years before our experiments (Yang et al., 2017;Wang et al., 2020). The HCHO photolysis rate of 8.3×10 6 molec cm -3 s -1 at 350 noon on pollution days is higher than pervious results in a suburban site of Beijing during winter (~6.7×10 6 molec cm -3 s -1 at noon) (Tan et al., 2018) and an industrial zone of southeastern China during winter (1.6×10 6 molec cm -3 s -1 averaged over 7:00−16:00) (Zheng et al., 2020). This indicates that high reactive VOC emissions over the NCP region can drive fast photochemistry in cold seasons on a regional scale by acting as radicals in addition to precursors.

Discussion and implication 355
From the results above, we conclude that the abrupt increase of PAN at the SDZ site during cold days is a result of local enhanced photochemistry from increased VOC precursor concentrations and HOx levels under a warmer, wetter atmosphere and southern wind anomalies. On the mornings of pollution days, the mean concentration of CH3CHO was 2.8 times that on clean days. This increase was due to persistent southern wind in the previous days, bringing polluted air mass from urban region to the SDZ site. Besides, the P[HOx] on pollution days was about 2 times that during clean days, owing to enhanced 360 photolysis of HCHO and HONO though weaker radiation at that time. The increases in CH3CHO level and P[HOx] on pollution days could virtually explain the enhanced net chemical formation rate (6.3 times that on clean days) with nearly constant reaction rate coefficient. Accelerated photochemistry by enhanced NOx and VOCs facilitates the rapid increase of PAN at the background site during cold seasons.
Particularly, our study demonstrates the dominant role of HCHO photolysis in HOx production during autumn at the 365 rural NCP region. The enhanced HCHO photolysis on pollution days not only promotes chemical production of PAN, but also accelerates formation of other secondary pollutants. This is evidenced by the synchronously increased concentrations of PAN, PM2.5 and O3 on pollution days, highlighting the importance of HCHO photolysis from VOC oxidation in secondary pollutant formation under low-NOx conditions even during cold days. On the other hand, HCHO itself is also a photochemical product mainly formed through oxidation by NMHCs. Its high level during pollution days and strong 370 correlation with PAN imply the potential role of accelerated photochemistry by enhanced NOx and VOCs in increases of photochemical pollutants.
The importance of HCHO photolysis on secondary pollutants as we demonstrated from the SDZ site measurement could also be applicable to regions with low NOx emissions but high reactive VOC emissions during cold seasons. The Chinese government has conducted effective NOx emission controls since 2013 and summertime VOC emission controls in 375 recent years. Our results show that it is also imperative to implement the VOC controls out of the summer season to avoid the unexpectedly enhanced photochemistry with decreasing NOx emissions over the NCP region.

Conclusion
We performed an integrated observation experiment at a rural site in the northern border of the NCP region in autumn 2020. The observed results show abrupt increases of PAN over two pollution days with increasing ratios of 244% and 178% 380 over the morning hours (8:00−12:00). Meteorlogical reanalysis data together with surface meteorological observations reveal that southwesterlies prevailed in the NCP during pollution days, accompanied by a warmer and wetter atmosphere. However, the abrupt increases of PAN on pollution days impossibly resulted from direct PAN transport from the urban region, as the time when the prevailing wind turned to southerlies was too late. Using CO as a tracer to exclude impact from physical transport of PAN from the urban region, we find that the net chemical formation rates was 0.45 ppb h -1 on the morning of 385 16 https://doi.org/10.5194/acp-2021-359 Preprint. Discussion started: 23 June 2021 c Author(s) 2021. CC BY 4.0 License. pollution days, which was 6.3 times that on clean days. Therefore, the strong local photochemistry is the main cause of PAN enhancement on the morning of the two pollution days.
Further investigation reveals that CH3CHO oxidation by OH is the major pathways of PA formation at the SDZ site.
The C3H6 and CH3CHO concentrations during 10/16 to 10/20 exhibited increasing trends, demonstrating the cumulative effect and regional transport from urban region under a meteorological condition of persistent southerlies during late 390 morning to evening. Statistical results show that the mean CH3CHO concentration on the morning of pollutions days was 2.8 times that on clean days. Additionally, the P[HOx] on pollution days was about 2 times that during clean days, owing to enhanced photolysis of HCHO and HONO despite weaker radiation at that time. Consequently, the abrupt increase of PAN at the SDZ site during cold days is a result of local enhanced photochemistry from increases of VOC precursor concentrations and HOx levels under a warmer and wetter atmosphere and south wind anomalies. Our study explores the 395 cause of abrupt increase of PAN concentration at a background site of the NCP region during cold days, and emphasizes the important role of HCHO in secondary pollutant formation in cold seasons, which is vital for understanding winter photochemistry under low NOx emissions.

Data availability
The ERA5 data are accessed from https://cds.climate.copernicus.eu/. The observation data for Ox at the MY and urban site 400 are obtained from the Beijing Municipal Ecological and Environmental Monitoring Center (http://www.bjmemc.com.cn/).
The observation data used in this study can be accessed via https://doi.org/10.7910/DVN/EPAGNB.