Winter ClNO2 formation in the region of fresh anthropogenic emissions: seasonal variability and insights into daytime peaks in northern China

Nitryl chloride (ClNO2) is an important chlorine reservoir in the atmosphere that affects the oxidation of volatile organic compounds (VOCs) and the production of ROx radicals and ozone (O3). This study presents measurements of ClNO2 and related compounds at urban, polluted rural, and polluted lower tropospheric (mountaintop) sites in the winter of 2017–2018 over the North China Plain (NCP). The nocturnal concentrations of ClNO2 were lower at the urban and polluted rural sites but higher at the polluted lower tropospheric site. The winter concentrations of ClNO2 were generally lower than the summer concentrations that were previously observed at these sites, which was due to the lower nitrate radical (NO3) production rate (P (NO3)) and the smaller N2O5 uptake coefficients (γ (N2O5)) in winter, despite the higher ratios of dinitrogen pentoxide (N2O5) to NO3 in winter. Significant daytime peaks of ClNO2 were observed at all the sites during the winter campaigns, with ClNO2 mixing ratios of up to 1.3 ppbv. Vertical transport of ClNO2 from the residual layers and prolonged photochemical lifetime of ClNO2 in winter may explain the elevated daytime concentrations. The daytime-averaged chlorine radical (Cl) production rates (P (Cl)) from the daytime ClNO2 were 0.17, 0.11, and 0.12 ppbv h−1 at the polluted rural, urban, and polluted lower tropospheric sites, respectively, which were approximately 3–4 times higher than the campaign-averaged conditions. Box model calculations showed that the Cl atoms liberated during the daytime peaks of ClNO2 increased the ROx levels by up to 27 %–37 % and increased the daily O3 productions by up to 13 %–18 %. Our results provide new insights into the ClNO2 processes in the lower troposphere impacted by fresh and intense anthropogenic emissions and reveal that ClNO2 can be an important daytime source of Cl radicals under certain conditions in winter.

NO 3 (g) + NO 2 (g) ↔ N 2 O 5 (g) (R10) NO 3 (g) + NO(g) ↔ 2NO 2 (g) (R11) NO 3 (g) + VOCs(g) → products (R12) N 2 O 5 (g) + Cl − (aq) → ClNO 2 (g) + NO − 3 (aq) ClNO 2 (g) + hv → Cl(g) + NO 2 (g) Field observations of ClNO 2 were first reported in the marine boundary layer off the coast of the Houston-Galveston area in the United States (Osthoff et al., 2008). Subsequent studies demonstrated the worldwide ubiquity of ClNO 2 and confirmed its significant role in photochemistry (Thornton et al., 2010;Mielke et al., 2011;Phillips et al., 2012;Edwards et al., 2013;Bannan et al., 2015;Wild et al., 2016;Bannan et al., 2019;Eger et al., 2019). The role of ClNO 2 in the radical budget could be more important than that of OH in winter, because OH production is reduced in winter owing to lower concentrations of O 3 and H 2 O vapor in this season. A limited number of winter observations of ClNO 2 have been conducted on various platforms, including on aircraft over northern Europe  and the eastern United States (Haskins et al., 2018(Haskins et al., , 2019; on a tall tower in Boulder, United States (Riedel et al., 2013); on a mountaintop in Hong Kong ; and at ground sites in Alberta, Canada (Mielke et al., 2016), and Heshan, China (Yun et al., 2018). These studies found high ClNO 2 mixing ratios of up to 7.7 ppbv (Yun et al., 2018) in winter and a contribution of ClNO 2 to Cl liberation of up to 83 %  in urban Manchester, and that ClNO 2 was a more dominant radical source than OH in both the early morning and the whole day in the polluted marine boundary layer downwind of the northeast United States (Haskins et al., 2019). ClNO 2 usually exhibits higher concentrations in aged and polluted air masses than in clean air and in regions subject to significant fresh NO emissions Osthoff et al., 2018).
The chemical production of ClNO 2 in winter has some unique features compared with that in warmer seasons. Long winter nights provide more time for ClNO 2 production and accumulation. Lower temperatures in winter shift the N 2 O 5 -NO 3 equilibrium to the N 2 O 5 side (Brown et al., 2003) and increase the γ (N 2 O 5 ) on aerosols ). In addition, NO x has longer lifetimes in winter compared with summer due to less abundant OH radicals in winter and its slower reaction rate with OH (Kenagy et al., 2018). However, P (NO 3 ) might be lower in winter due to reduced O 3 concentrations. The availability of aerosol Cl − also varies in winter and summer. More Cl − is emitted due to coal burning in winter (McCulloch et al., 1999;Fu et al., 2018). In places like East Asia, the winter monsoon brings air masses from the interior of the continent, thereby suppressing the transport of sea salt to inland areas. Because of the contrasts in the availability of aerosol chloride and the variability in meteorology and NO x emissions that affect the N 2 O 5 chemistry, it is not clear whether ClNO 2 formation is more prevalent in winter.
The North China Plain (NCP) -home to Beijing and several other megacities -is one of the most industrialized and populous regions of China and frequently suffers from severe haze pollution in winter Fu et al., 2020). ClNO 2 concentrations have been measured over the NCP Zhou et al., 2018), but only one study was conducted in winter . The present study presents recent field observations of ClNO 2 concentra-tions from three campaigns conducted in winter and early spring at three sites in the NCP. The results were compared with those obtained in the previous summer campaigns at the same locations. We examined the factors controlling ClNO 2 formation, i.e., P (NO 3 ), which is the nocturnal reactivity of NO 3 , and N 2 O 5 , γ (N 2 O 5 ), and ϕ(ClNO 2 ). We then focused on the unexpected daytime peaks of ClNO 2 concentrations that were observed at the sites and evaluated their impact on the daytime atmospheric oxidative capacity using a chemical box model.

Observation sites
Field campaigns were performed in Wangdu, Beijing, and Mt. Tai in sequence during the winter-early spring of 2017-2018 (Table 1). The locations of the three sites are shown in Supplement Fig. S1. The sites were selected for investigation of ClNO 2 in urban, rural, and mountain areas of the NCP. The winter indoor-heating period lasts from early November to 15 March of the following year (Ran et al., 2016), and thus the observations were made mostly during the heating period during which coal is intensively used. Detailed descriptions of the measurement sites are available in previous studies Xia et al., 2019), and a brief introduction is given here.
Our observations at the Wangdu site were part of the Campaign of Oxidation Potential Research for air Pollution in winter (COPPER). The Wangdu site is located in Dongbaituo Village, Hebei Province. Local villagers use coal stoves for cooking and heating during winter. National road G4 and provincial road S335 are 1 and 3 km to the west of the sampling site, respectively. Many heavy-duty trucks passed through G4 and S335 every night during the study period, emitting a large amount of NO x and particulate matter. Therefore, the site experienced heavy pollution from coal burning and road traffic (Peng et al., 2020).
The Beijing site is located at the Chinese Research Academy of Environmental Science (CRAES), which is 15 km northeast of the city center. The sampling site is surrounded by intra-city roads, commercial buildings, and residential buildings with few industrial facilities. When the prevailing wind originates from the north (i.e., remote mountainous regions), the site is upwind from the Beijing downtown area and thus is less polluted. However, when the wind originates from the south, the site receives pollutants from Beijing's urban areas in the NCP (Xia et al., 2019).
Mt. Tai is located approximately 40 km south of Jinan City (population: 8.9 million) and 15 km north of Tai'an City (population: 5.6 million) (Wen et al., 2018). Measurements were taken at Mt. Tai meteorological station (1534 m a.s.l.). The site is isolated from the anthropogenic emissions of tourist areas and is not affected by local emissions. The ob-servation period, i.e., March to April, was in early spring in the NCP. However, considering the low temperature (4.6 ± 6.3 • ) observed on top of Mt. Tai, this study considered the observation period to be winter to early spring.

Measurements of N 2 O 5 and ClNO 2 concentrations
N 2 O 5 and ClNO 2 were simultaneously measured by a chemical ionization mass spectrometer with a quadrupole mass analyzer (Q-CIMS; THS Instruments). The principles and calibrations of the Q-CIMS measurements are available in previous studies Tham et al., 2016;Xia et al., 2019). Briefly, the primary ions used in the Q-CIMS were iodide (I − ) and its water clusters, which were generated using CH 3 I with an inline ionizer ( 210 Po). The iodide adducts, namely IN 2 O − 5 and IClNO − 2 , were then detected by the mass spectrometer. An example of the mass spectrum is shown in Fig. S2. The integration time of the signals recorded by the Q-CIMS is shown in Table S1. The isotopic ratios of I 35 ClNO − 2 and I 37 ClNO − 2 in the ambient data were used to confirm the identity of ClNO 2 (Fig. S3). Gas-phase mixtures of NO 2 and O 3 produced N 2 O 5 for N 2 O 5 calibration. The synthetic N 2 O 5 was converted to ClNO 2 by passage through a humidified NaCl slurry for ClNO 2 calibration. The dependence of the N 2 O 5 sensitivities on ambient RH was tested once in each campaign and used to calibrate the N 2 O 5 data (Fig. S4a). ClNO 2 sensitivities were found to be not affected by RH (Fig. S4b). Single-concentration calibrations of N 2 O 5 and ClNO 2 were performed every 1-2 d, which showed stable sensitivities of N 2 O 5 and ClNO 2 (Text S1 and Fig. S5). And the linearity of the N 2 O 5 and ClNO 2 signals to concentrations was checked via multiconcentration calibrations in the Mt. Tai campaign (Fig. S6). Backgrounds of N 2 O 5 and ClNO 2 were determined by passing the ambient air through glass wool once a day at different time. The background signals of N 2 O 5 (3.3-7.7 pptv) and ClNO 2 (1.0-7.5 pptv) were stable and independent of the time of the day (Fig. S7). The detection limits of N 2 O 5 and ClNO 2 were 6.9-7.3 and 3.8-5.3 pptv, which is defined here as 3 times the standard deviation of the background signals in 5 min (Table S2 in the Supplement). A virtual-impactor design (Peng et al., 2020) was adopted, and the sampling tube was replaced daily to minimize inlet artifacts.

Other measurements
The trace gases, particle number size distribution (PNSD), and ionic composition of aerosols and other species were simultaneously measured (Table S3). Online non-methane hydrocarbons were measured by gas chromatography-flame ionization detection-mass spectrometry (GC-FID/MS; Chromatotec Group) at the Beijing site  and Wangdu site . At Mt. Tai, we used canisters to collect air samples, which were analyzed using GC-FID/MS. In Wangdu and Mt. Tai, oxygenated volatile or- ganic compound (OVOC) samples were collected on DNPHcoated sorbent cartridges followed by post-campaign analysis using high-performance liquid chromatography. The ionic compositions of PM 2.5 (e.g., NH + 4 , NO − 3 , SO 2− 4 , and Cl − ) were quantified by the Monitor for AeRosols and GAses in ambient air (MARGA, Metrohm) at the Beijing and Mt. Tai sites (Wen et al., 2018). An aerosol chemical speciation monitor (ACSM, Aerodyne Research Inc.) was utilized at the Wangdu site to monitor the non-refractory components of these ions in PM 2.5 . The concentrations of the NO − 3 , SO 2− 4 , and NH + 4 measured simultaneously by the MARGA and ACSM were in good agreement, whereas the concentration of Cl − measured by the ACSM was slightly lower than that measured by the MARGA, which was possibly due to the significant proportion of refractory chloride, e.g., NaCl, present in the aerosols (Xia et al., 2020). We assumed that the particles sampled by a wide-range particle spectrometer (WPS) were spherical in shape and calculated the aerosol surface area density (S a ) and volume density (V a ). A parameterization was adopted to consider the hygroscopic growth factor (GF) of aerosol sizes, as follows: GF = a × b + 1 1−RH 1/3 (Lewis, 2008), where the parameters a and b were derived as 0.582 and 8.460, respectively, in a previous field study over the NCP (Achtert et al., 2009).

Calculation of N 2 O 5 loss and ClNO 2 production
Some analytical metrics were calculated from the observation data. P (NO 3 ) was calculated using Eq. (1), where k 1 represents the rate constant of Reaction (R9) (Atkinson and Lloyd, 1984).
(1) k(NO 3 ) during the night was calculated using the measured mixing ratios of NO and non-methane hydrocarbons that were measured by GC (Sect. 2.3). As most OVOCs react with NO 3 at much slower rates compared to those with hydrocarbons, especially alkenes (Atkinson and Arey, 2003), the OVOCs were not included in the calculation of k(NO 3 ). Nonetheless, the k(NO 3 ) might be slightly underestimated here.
where k i is the rate constant for a specific VOC + NO 3 reaction and is adopted from Atkinson and Arey (2003), and k NO+NO 3 represents the rate constant for Reaction (R11) (De-More et al., 1997). The ambient concentrations of NO 3 were estimated by assuming that NO 3 and N 2 O 5 were in dynamic equilibrium (DeMore et al., 1997).
The loss rates of NO 3 due to NO and VOCs were then cal- The loss rate coefficient of N 2 O 5 on the aerosol surface (k(N 2 O 5 )) is expressed as follows.
where c(N 2 O 5 ) represents the average molecular velocity of N 2 O 5 . The rate constants (k 1 , k i , and k NO+NO 3 ) and equilibrium constant (K eq ) are calculated as temperature-dependent parameters. γ (N 2 O 5 ) and ϕ(ClNO 2 ) were estimated using steadystate analysis in applicable cases (Brown et al., 2006). This method assumes a steady state of N 2 O 5 , which means that the production rate of N 2 O 5 is equal to its loss rate. We adopted the criteria described by Xia et al. (2020) to select the cases, namely low concentrations of NO, an increasing trend of ClNO 2 concentrations, and stable air masses. Equation (5) was then established by plotting τ ( as the slope and k(NO 3 ) as the intercept in the linear regression (Brown et al., 2003). Here, the derived γ (N 2 O 5 ) was accepted when the regression had R 2 > 0.5 and k(NO 3 ) > 0.
ϕ(ClNO 2 ) was then calculated using the following equation: where d[ClNO 2 ] / dt and [N 2 O 5 ] represent the increasing rate of ClNO 2 production and the average concentration of N 2 O 5 , respectively, within the selected cases.

Box model
An observation-based chemical box model was utilized to simulate the concentrations of Cl and RO x radicals and the production and loss pathways of O 3 . The detailed model description is available in Peng et al. (2020). Based on Master Chemical Mechanism (MCM) v3.3.1 (Jenkin et al., 2015), Peng et al. (2020) modified the chemical mechanisms to include up-to-date gas-phase chlorine and bromine chemistry. The observed N 2 O 5 , ClNO 2 , NO x , HONO, O 3 , j NO 2 , and related species were constrained in the model for every 10 min of model time, after interpolating or averaging the data (Table S4). The mixing ratios of non-methane hydrocarbons (NMHCs) and OVOCs (Sect. 2.3) were constrained every hour. As OVOCs were not measured in Beijing in this study, we adopted the concentrations of OVOCs measured in previous studies in winter Beijing Qian et al., 2019). We also assumed the CH 4 mixing ratio to be 2000 ppbv, which was the mean value in summer at Wangdu , for our Wangdu and Beijing sites. We acknowledge that the adoption of the summer CH 4 concentrations for our winter studies may underestimate the CH 4 level, but the effect on the RO x budget is expected to be insignificant due to much smaller contribution of CH 4 to RO x than NMHCs and OVOCs in polluted environments. The photolysis frequencies of ClNO 2 , O 3 , and other species were simulated according to the solar zenith angle using the tropospheric ultraviolet and visible (TUV) radiation model and scaled by the observed j NO 2 values. Numerical experiments were conducted by constraining (Case 1) and not constraining ClNO 2 data (Case 2) at each site. The differences in the radical concentrations and O 3 budgets between Cases 1 and 2 represented the effect of ClNO 2 . For example, the increase in RO x (%) due to ClNO 2 was calculated by (RO x _w -RO x _wo) / RO x _wo, where RO x _w represents the concentration of RO x in Case 1 with ClNO 2 constrained in the model, and RO x _wo represents the concentration of RO x in Case 2 without ClNO 2 constrained.

Overall measurements, diurnal patterns, and comparison with other studies
The time series of N 2 O 5 and ClNO 2 levels in the three campaigns are displayed in Fig. 1. Overall, elevated levels of N 2 O 5 and ClNO 2 were observed with different patterns at each site. The ground sites (Wangdu and Beijing) were characterized by high NO x levels (83.2 ± 81.3 ppbv and 35.6 ± 27.3 ppbv, respectively) and low O 3 levels (8.5 ± 8.8 ppbv and 17.3 ± 11.4 ppbv, respectively), whereas the mountain site, Mt. Tai, was marked by relatively lower NO x levels (2.4 ± 2.0 ppbv) and higher O 3 levels (64.6 ± 14.7 ppbv) (Fig. S8). The campaign-averaged mixing ratios of ClNO 2 were similar at the ground sites (71 ± 132 pptv and 76 ± 103 pptv in Wangdu and Beijing, respectively) and were significantly lower than that at Mt. Tai (179 ± 247 pptv). The nocturnal ratio of ClNO 2 / N 2 O 5 at each site displayed large day-to-day variability, which was positively dependent on the ambient RH ( Fig. S9) and, to a lesser extent, positively correlated with S a (figure not shown). The campaign-averaged diurnal patterns of the mixing ratios of N 2 O 5 , ClNO 2 , and related species are depicted in Fig. 2. ClNO 2 levels typically exhibited a daily cycle, peaking at night and decreasing during the day. The diurnal pattern of ClNO 2 at the Wangdu site in winter was an exception, with elevated concentrations (10th-90th percentiles) around midday (12:00-14:00 local time, LT), which resulted from a noontime peak in ClNO 2 concentrations during a few days at Wangdu. The detailed observation results from each site are separately introduced as follows.
The nocturnal production of ClNO 2 was insignificant in Wangdu despite the presence of abundant Cl − (3.3 ± 3.2 µg m −3 throughout the observation), which likely originated from the intensive residential coal combustion in the area (Peng et al., 2020). The Wangdu site experienced high mass concentrations of PM 2.5 (a maximum of approximately 450 µg m −3 ) and very large mixing ratios of NO (a maximum of approximately 350 ppbv). The wind rose analysis showed that the high concentrations of NO originated Figure 2. Diurnal average levels of N 2 O 5 , ClNO 2 , O 3 , and j NO 2 observed at the Wangdu, Beijing, and Mt. Tai sites throughout the campaign in winter (this study) and previous summer field studies ( Table 1). The shaded areas indicate the 10th and 90th percentiles. from the west of the sampling site where two major roads were located. Numerous heavy-duty trucks on these roads were responsible for high NO concentrations. The presence of abundant NO inhibited N 2 O 5 formation by consuming O 3 and NO 3 at the Wangdu site. When the ambient concentrations of NO substantially decreased, e.g., on 10 December, the N 2 O 5 mixing ratios increased to 1 ppbv. The mixing ratios of ClNO 2 were mostly low (< 200 pptv) during the night. The relationship between nighttime levels of ClNO 2 and grouped NO and NO x concentrations is shown in Fig. 3. ClNO 2 showed higher levels when the NO mixing ratios were below 10 ppbv and NO x mixing ratios ranged from 10-20 ppbv (Fig. 3a, d). However, significant daytime peaks in ClNO 2 mixing ratios were observed on 14 and 28 December, reaching approximately 0.8 and 1.3 ppbv, respectively. The daytime peaks in ClNO 2 concentrations at the three sites are discussed in detail in Sect. 3.3. For comparison, the ambient mixing ratios of NO in the summer campaign at Wangdu were much lower (mostly 0-10 ppbv), and O 3 mixing ratios were much higher (i.e., exceeded 90 ppbv on most days), which favored the production of N 2 O 5 and ClNO 2 .
The winter Beijing observations showed that there was significant production of N 2 O 5 but limited conversion of N 2 O 5 to ClNO 2 in dry conditions. The observation period in Beijing was divided into polluted days (24 h PM 2.5 > 75 µg m −3 ; China's Grade II air quality standard for PM 2.5 ) and clean days (24 h PM 2.5 < 35 µg m −3 ; Grade I standard). The polluted periods were characterized by simultaneous high levels of PM 2.5 and NO, e.g., on 19 January. The clean periods were marked by relatively high mixing ratios of O 3 and low levels of PM 2.5 and NO x , e.g., from 8 to 11 January. Both polluted and clean conditions were unfavorable for ClNO 2 formation owing to the high concentrations of NO on the polluted days and the low concentrations of NO 2 and aerosols on the clean days. Moreover, the RH observed in Beijing was typically below 40 %, which indicated relatively slow heterogeneous loss of N 2 O 5 and slow ClNO 2 formation. Consequently, N 2 O 5 mixing ratios frequently accumulated to elevated levels, exceeding 0.4 ppbv on 10 of the 26 observation nights, and the mixing ratio of ClNO 2 was mostly below 0.4 ppbv. Nighttime levels of ClNO 2 in winter Beijing were higher when NO mixing ratios ranged from 0-10 ppbv and NO x mixing ratios ranged 20-50 ppbv (Fig. 3b, d). The highest mixing ratios of ClNO 2 were observed (up to approximately 0.8 ppbv) when the site occasionally intercepted air masses with a higher RH (approximately 75 %), e.g., on the night of 22 January. This result is similar to the previous observation in Beijing (Xia Figure 3. The relationship between nighttime levels of ClNO 2 and grouped NO (a, b, c) and NO x (d, e, f) mixing ratios in the winter (green color) and summer (purple color) campaigns. The difference in the scale of ClNO 2 in (c) and (f) is caused by statistic factors, since only the 10th to 90th percentiles of ClNO 2 data are shown here. et al., 2019), in which the ratio of ClNO 2 to N 2 O 5 increased significantly from late spring with a low RH (10 %-30 %) to early summer with a higher RH (20 %-80 %). The overall mixing ratios of ClNO 2 in the present Beijing study in winter were also significantly lower than those reported in summer (maximum of 1.4 to 2.9 ppbv) in other studies (Le Zhou et al., 2018).
Elevated mixing ratios of ClNO 2 (i.e., above 0.5 ppbv) were frequently recorded at the Mt. Tai station in winter. Nighttime levels of ClNO 2 were slightly higher when NO levels were below 0.5 ppbv (Fig. 3c) and showed a positive correlation with NO x levels (Fig. 3f). High concentrations of PM 2.5 (34.5 ± 27.3 µg m −3 ) and high RH (63.6 ± 27.1 %) favored the ClNO 2 formation at Mt. Tai. The maximum level of ClNO 2 (approximately 1.7 ppbv) was observed just before midnight on 18 March, which was slightly lower than the highest concentration observed at Mt. Tai in the summer of 2014 (Z. . The elevated concentrations of ClNO 2 observed in the previous summer study at Mt. Tai were due to emissions from distinct coal-fired power plants, whereas this winter study found that coal burning had less effect on concentrations of ClNO 2 . The campaign-averaged levels of SO 2 and particulate SO 2− 4 were 1.6 ± 1.6 ppbv and 3.6 ± 2.9 µg m −3 , respectively, during the winter observations, which were significantly lower than those observed in the summer campaigns (2.9 ± 3.7 ppbv and 14.8 ± 9.0 µg m −3 , respectively). The decreases in SO 2 and sulfate were attributed to strengthened emission control for coal-fired power generation during 2014-2018 and also less transport of emissions from the ground to the Mt. Tai site (1534 m a.s.l.) in late winter and early spring compared with that in summer.
We compared the observed winter concentrations of ClNO 2 with those reported in previous studies in Asia, North America, and Europe (Fig. 4). The highest winter concentrations of ClNO 2 to date were observed in southern China, with a maximum level of 4.7 ppbv at a mountain top in Hong Kong in aged urban/industrial plumes from the Pearl River Delta (PRD)  and 8.3 ppbv during a severe pollution episode within the PRD (Yun et al., 2018). The high- concentration ClNO 2 events in southern China were due to concurrent high levels of PM 2.5 and O 3 (e.g., 400 µg m −3 and 160 ppbv found by Yun et al., 2018), which contrast the high concentrations of PM 2.5 and low concentrations of O 3 over northern China during the cold winter. The winter mixing ratios of ClNO 2 in the United States and Europe range from approximately 0.3 ppbv in urban California (Mielke et al., 2016) and urban Manchester , respectively, to 1.3 ppbv in the outflow of coastal urban areas (Riedel et al., 2013;Haskins et al., 2019). In general, the winter concentrations of ClNO 2 over northern China were comparable to or slightly higher than those observed in the United States and Europe.

NO 3 production and loss pathways
To gain insight into the processes controlling the variability in concentrations of ClNO 2 , nocturnal P (NO 3 ) and NO 3 loss pathways were compared using Eqs. (1-5) in Sect. 2.4. The average P (NO 3 ) was comparable at the three sites in winter, ranging from 0.15 to 0.25 ppbv h −1 , and these rates were significantly lower than the respective summer values (Fig. 5a). The lower P (NO 3 ) in winter was caused by both lower k 1 and lower [NO 2 ] × [O 3 ] in winter (see Eq. 1). The lower k 1 in winter is caused by lower temperature in winter, while the lower [NO 2 ] × [O 3 ] in winter is mainly caused by less photochemical production of O 3 and more NO that consumes the available O 3 in winter (Table S5) (Table S6 and Fig. S10) were used in Eq. (4). The nighttime NO 3 loss via NO titration and VOC oxidation was greater than the N 2 O 5 heterogeneous loss in all the winter and summer campaigns (Fig. 5b). These were the campaign average results. In contrast, the N 2 O 5 loss was greater than the NO 3 loss in selected cases in summer at Mt. Tai (Wang et al., 2017c). To determine the nocturnal loss of NO 3 , we further compared the N 2 O 5 / NO 3 ratio and γ (N 2 O 5 ) at the three sites. The thermal decomposition of N 2 O 5 was suppressed in winter and resulted in high ratios of N 2 O 5 / NO 3 ( Fig. 6a; up to approximately 1000), which favored N 2 O 5 loss over NO 3 loss. However, the γ (N 2 O 5 ) in winter was systematically lower than that in summer (Fig. 6b), which indicated slower N 2 O 5 loss in winter. A previous field study in winter Beijing also reported small values of γ (N 2 O 5 ), ranging from < 0.001 to 0.02 . This result differs from previous laboratory studies, which reported larger γ (N 2 O 5 ) on (NH 4 ) 2 SO 4 aerosols at lower temperatures (Hallquist et al., 2003;Griffiths and Anthony Cox, 2009). It is possible that other factors, such as RH and aerosol composition (aside from (NH 4 ) 2 SO 4 ), had a large influence on γ (N 2 O 5 ). The limited number (two to four) of γ (N 2 O 5 ) values obtained in each winter campaign (Table S6) may have also caused a bias in the estimation of the overall γ (N 2 O 5 ). The opposite effects -a higher N 2 O 5 / NO 3 ratio and lower γ (N 2 O 5 ) in winteroffset each other in Wangdu (Fig. 6b) but favored N 2 O 5 loss in Beijing and NO 3 loss at Mt. Tai compared with those in the respective summer campaigns. The higher concentrations of ClNO 2 at Mt. Tai during the winter campaigns may be attributable to higher ϕ(ClNO 2 ) values in Mt. Tai (Fig. 6c).

Daytime peaks in ClNO 2 concentrations
In the winter campaigns, high concentrations of ClNO 2 were sustained after sunrise. Distinct peaks in ClNO 2 concentrations were observed on 3-4 d in each campaign, as shown in Fig. 7 displaying one case at each site. Other daytime cases from the three sites are shown in Figs. S11-13. The valid-ity of the daytime peaks was checked by performing isotopic analysis of ClNO 2 , background detection, and on-site calibration. The signals of I 35 ClNO − 2 and I 37 ClNO − 2 were well correlated (R 2 > 0.99) during daytime peaks in ClNO 2 concentrations (Fig. S3a-c) and calibrations (Fig. S3d-f). The ratio of I 37 ClNO − 2 to I 35 ClNO − 2 (0.32-0.35) was consistent with the natural isotopic ratio of 37 Cl to 35 Cl. The background signals of ClNO 2 were checked when its daytime peaks in concentrations were observed, and no increase in the background was found. We also checked the signal of primary ions (IH 2 O − ) and found no abnormal changes when ClNO 2 concentrations showed daytime peaks. These results confirmed that the daytime peaks in ClNO 2 concentrations were real atmospheric phenomena.
The daytime ClNO 2 episodes usually occurred from 10:00 to 11:00 LT at each site. The highest daytime mixing ratio of ClNO 2 was 1.3 ppbv (5 min average) observed at 11:30 on 28 December 2017 in Wangdu. In comparison, the daytime ClNO 2 concentration observed in the previous summer study at Wangdu (Tham et al., 2016) reached a maximum in the early morning (08:00 LT) and declined to several parts per trillion by volume at 11:00. Attenuated solar radiation was observed during the days with daytime peaks in ClNO 2 concentrations. For example, the daily maximum rates of j NO 2 (1 min average) for the Wangdu case shown in Fig. 7a (2.5 × 10 −3 s −1 ) was significantly lower than the highest rate observed during this campaign (6.0 × 10 −3 s −1 ). The attenuated solar radiation reduced the photolysis of ClNO 2 , which allowed it to persist for a longer period during the day. The chemical data showed contrasting features during the daytime peaks in ClNO 2 concentrations at the three sites. At Wangdu, ClNO 2 concentrations showed a sharp peak while the concentrations of other pollutants decreased (Fig. 7a); in Beijing, the daytime peak in ClNO 2 concentrations appeared with little simultaneous change in the NO − 3 , NO x , and O 3 levels after sunrise (Fig. S12a). In two cases, daytime peaks of ClNO 2 concentrations at Mt. Tai (Figs. 7c and S13c) occurred together with significant increases in NO − 3 , NO x , and PM 2.5 levels, whereas O 3 concentrations decreased after sunrise and resumed previous levels.
The daytime peaks in ClNO 2 concentrations were likely caused by the transport of air masses to the respective sites. In situ production of ClNO 2 was limited during the days on which significant daytime ClNO 2 occurred, because the mixing ratios of N 2 O 5 were near the detection limit of the instrument (several parts per trillion by volume). The photochemical lifetime of ClNO 2 at 10:00 am LT was estimated to be 1-2 h, based on the inverse of j ClNO 2 , which allowed the transport of ClNO 2 produced elsewhere to the observation sites. As daytime peaks in ClNO 2 concentrations appeared at both the ground and mountain sites, the high-ClNO 2 region may exist in the residual layer above the nocturnal mixing layers. At sunrise, ClNO 2 -rich air masses may be transported downward to the ground sites (Wangdu and Beijing) and upward to the mountaintop site (Mt. Tai). The downward trans- port of ClNO 2 at Wangdu in summer has been illustrated by Tham et al. (2016), and the upward transport to the top of Mt. Tai has also been implicated by the increasing daytime concentrations of O 3 and other pollutants (e.g., Gao et al., 2005;Zhou et al., 2009;Jiang et al., 2020). Measurements in the residual layers are needed to further investigate the transport of ClNO 2 within the entire boundary layer.

Impact of daytime ClNO 2 on atmospheric oxidation capacity
We used the box model (Sect. 2.5) to show the impact of ClNO 2 on photochemical oxidation at the three sites ( Fig. 7a-c). In campaign-averaged conditions, the impact of ClNO 2 was minor, owing to the low daytime concentrations of ClNO 2 . The daytime-averaged P (Cl) (06:00-18:00 LT) from ClNO 2 photolysis was in the range of 0.03-0.06 ppbv h −1 , with the peak values of 0.07-0.12 ppbv h −1 , and the photolysis of ClNO 2 enhanced the daytime RO x concentrations by 1.3 %-3.8 % and net O 3 production by 1.3 %-6.2 % at the three sites (figures not shown). Such impacts were lower than those during summer at Wangdu . However, the impact of ClNO 2 increased considerably in the cases of daytime-peak concentrations, as shown in Fig. 8. The daytime-averaged P (Cl) values from ClNO 2 photolysis were 0.15 ± 0.13 (maximum of 0.46), 0.11 ± 0.09 (maximum of 0.32), and 0.19 ± 0.20 (maximum of 0.74) ppbv h −1 at Wangdu, Beijing, and Mt. Tai, respectively (Fig. 8a-c). The winter P (Cl) peak in Wangdu (Fig. 8a, 0.46 ppbv h −1 ) was twice the summer average value (0.24 ppbv h −1 ) . The P (Cl) during the daytime peaks of ClNO 2 in this study is significantly higher than that in Riedel et al. (2012) (maximum ∼ 0.08 ppbv h −1 ) but slightly lower than that in Haskins et al. (2019) (maximum ∼ 1.3 ppbv h −1 ). P (Cl) from other sources (e.g., the HCl + OH reaction) was minor (8.8 %-14.5 %) during these cases. The relative importance of ClNO 2 in primary radical production varied among these sites. ClNO 2 had a minor contribution in Beijing but became increasingly important in Wangdu and Mt. Tai (Fig. 8b, c). HONO photolysis was the most important source of OH at the two ground sites, whereas O 3 was also important at Mt. Tai.
The impact of Cl in the NCP is likely larger than the result shown above. Our model calculations considered photolysis of ClNO 2 (and HCl + OH) as the source of Cl, but not other photolabile Cl-containing gases. However, in the Wangdu field campaign, we frequently observed elevated daytime concentrations of bromine chloride (BrCl) and molecular chlorine (Cl 2 ), which dominated the Cl production (Peng et al., 2020). In addition, our ClNO 2 measurements were conducted at polluted ground-level sites and at a high mountain site (1534 m a.s.l.), which are not in the nocturnal residual layer where strong ClNO 2 production is expected to occur . It is thus highly desirable to measure ClNO 2 in the residual layer in future studies to comprehensively assess the role of ClNO 2 in the lower part of the atmosphere.  Observations of ClNO 2 and related species were conducted at urban, polluted rural, and polluted lower tropospheric sites in the winter of 2017-2018 in the NCP, which suffers from severe winter haze pollution. The winter measurements showed lower concentrations of ClNO 2 compared with those in previous summer observations. The campaign-averaged NO 3 loss via reaction with NO at night dominated over the N 2 O 5 loss at all the sites due to high NO concentrations, and in situ ClNO 2 formation was generally insignificant. However, high levels of daytime ClNO 2 (exceeding 1 ppbv) were observed at the three sites. We suggest that ClNO 2 was efficiently produced in the nocturnal residual layer and was transported to ground-level and high-elevation sites. The daytime concentrations of ClNO 2 had great effects on the production of Cl, RO x , and O 3 . Vertical measurements of the concentrations of ClNO 2 and related compounds are needed to better understand the distribution and impact of these species in the lower troposphere. Compared to the previous studies in the clean troposphere or in more aged air masses, our results provide new insights into ClNO 2 formation in the region affected by fresh and intense anthropogenic emissions.
Data availability. The datasets described in this study are available by contacting the corresponding author (cetwang@polyu.edu.hk).
Author contributions. TW designed this study. JC, YM, LX, JG, and HL provided field measurement sites. MX, XP, and WW conducted the CIMS measurements. CY, ZW, YJT, HC, CZ, PL, and XW provided supporting data. XP and WW performed the box model simulation. MX analyzed and virtualized the research data. MX and TW wrote the manuscript with discussions and comments from all co-authors.
Competing interests. The authors declare that they have no conflict of interest.
Disclaimer. Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.