Formation drivers and photochemical effects of ClNO₂ in a coastal city of Southeast China

Nitryl chloride (ClNO₂) is an important precursor of chlorine (Cl) radical, significantly affecting ozone (O₃) formation and photochemical oxidation. However, the key drivers of ClNO₂ production are not fully understood. In this study, the field observations of ClNO₂ and related parameters were conducted in a coastal city of Southeast China during the autumn of 2022, combining with machine learning and model simulations to elucidate its key influencing factors and atmospheric impacts. Elevated concentrations of ClNO₂ (>500 ppt) were notably observed during nighttime in late autumn, accompanied by increased levels of dinitrogen pentoxide (N₂O₅) and nitrate (${\mathrm{NO}}_{\mathrm{3}}^{-}$). Nighttime concentrations of ClNO₂ peaked at 3.4 ppb, while its daytime levels remained significant, reaching up to 100 ppt and sustaining at approximately 40 ppt at noon. Machine learning and field observations identified nighttime N₂O₅ heterogeneous uptake as the predominant pathway for ClNO₂ production, whereas ${\mathrm{NO}}_{\mathrm{3}}^{-}$ photolysis may contribute to its daytime generation. Additionally, ambient temperature (T) and relative humidity (RH) emerged as primary meteorological factors affecting ClNO₂ formation, mainly through their effects on thermal equilibrium and N₂O₅ hydrolysis processes, respectively. Ultraviolet (UV) radiation was found to play a dual role in ClNO₂ concentrations around noon. Box model simulations showed that, under high-ClNO₂ conditions, the rates of alkane oxidation by Cl radical in the early morning exceeded those by OH radical. Consequently, volatile organic compound (VOC) oxidation by Cl radical contributed ∼19 % to RO_x production rates, thereby significantly impacting O₃ formation and atmospheric oxidation capacity. This research enriched the understanding of ClNO₂ generation and loss pathways, providing valuable insights for the regulation of photochemical pollution in coastal regions.

1 Introduction

Chlorine (Cl) radical, as an important atmospheric oxidant, can react with volatile organic compounds (VOCs) to affect RO_x (including OH, HO₂, and RO₂) radicals and ozone (O₃) formation (Yi et al., 2023), thereby perturbing atmospheric chemical components and air quality (Peng et al., 2021; Li et al., 2020). The reaction rates between Cl radical and some alkanes are several orders of magnitude faster than those involving OH radical (Atkinson et al., 2006). Furthermore, the related studies indicated that the production rates of Cl radical in the early morning could significantly exceed the production rates of OH radical formed via O₃ photolysis (Phillips et al., 2012; Tham et al., 2016), thereby enhancing the atmospheric oxidation capacity (AOC).

Nitryl chloride (ClNO₂) is one of the major Cl radical precursors in the tropospheric atmosphere (Thornton et al., 2010; Xue et al., 2015; Liu et al., 2017). It is mainly generated by the heterogeneous uptake of dinitrogen pentoxide (N₂O₅) on chloride-containing aerosols (Finlayson-Pitts et al., 1989; Thornton et al., 2010), among which N₂O₅ is produced through the equilibrium reaction with nitrogen dioxide (NO₂) and nitrate (NO₃) radical. Since Osthoff et al. (2008) first detected over 1 ppb of ClNO₂ in the urban outflows of America (Osthoff et al., 2008), significant production of ClNO₂ has been widely observed in the polluted coastal and inland areas with abundant anthropogenic emissions and chloride sources, with concentrations ranging from tens of parts per thousand (ppt) to several parts per billion (ppb) (Riedel et al., 2012; Mielke et al., 2013, 2011; Phillips et al., 2012; Bannan et al., 2015; Wang et al., 2016, 2022; Xia et al., 2020, 2021; Yun et al., 2018; Li et al., 2023). For the diurnal profile of ClNO₂, its concentrations generally peaked and accumulated at midnight, then rapidly decreased to low levels due to strong photolysis after sunrise (Ma et al., 2023; Mielke et al., 2011; Xia et al., 2020). However, elevated daytime concentrations of ClNO₂ have been observed in field studies, mainly attributed to reduced photolysis rates under heavy cloud or fog cover and to contributions from horizontal and vertical transport (Tham et al., 2016; Xia et al., 2021; Jeong et al., 2019; Mielke et al., 2013; Bannan et al., 2015). Notably, the recent laboratory research demonstrated that nitrate (${\mathrm{NO}}_{\mathrm{3}}^{-}$) photolysis can generate ClNO₂ alongside Cl₂ (Dalton et al., 2023), yet this mechanism has not been confirmed under real atmospheric conditions.

At present, the observation studies of ClNO₂ focused on investigating its influencing factors, such as the N₂O₅ uptake coefficient and the production yield of ClNO₂ (Thornton et al., 2003; Tham et al., 2018). The field and laboratory studies have indicated that ClNO₂ production was mainly affected by ambient temperature (T), relative humidity (RH), and particle components (e g., chloride (Cl⁻), ${\mathrm{NO}}_{\mathrm{3}}^{-}$, and liquid water content) (Bertram and Thornton, 2009; Wang et al., 2023, 2020). In addition to influencing factors, the photochemical effects of ClNO₂ photolysis have been extensively evaluated (Xue et al., 2015; Xia et al., 2021; Tham et al., 2016). Cl radical released by ClNO₂ photolysis will oxidize VOCs to promote the formation of RO₂ radical and O₃, greatly compensating for the underestimation of RO₂ radical and O₃ generation in model simulations (Peng et al., 2021; Ma et al., 2023). The field measurements of ClNO₂ have been conducted in different atmospheric environments, while the key drivers of ClNO₂ chemistry were still not well recognized. Moreover, it is pertinent to explore whether there are additional and unrecognized sources of ClNO₂ beyond its heterogeneous generation from N₂O₅.

In this study, the comprehensive measurements of ClNO₂ and related parameters were conducted in a coastal city of Southeast China during the autumn of 2022. Field observations, combined with a machine learning model, were used to reveal the key driving factors of ClNO₂ formation. Furthermore, we further investigated the potential mechanisms driving daytime ClNO₂ generation. Additionally, we also assessed the photochemical impacts of ClNO₂ based on a box model. Overall, this study underscored the important role of ${\mathrm{NO}}_{\mathrm{3}}^{-}$ in the ClNO₂ chemistry.

2 Materials and methods

2.1 Field measurements

The intensive field measurements of ClNO₂, related precursors, and meteorological parameters from 9 October–5 December 2022 were performed at an urban site (Institute of Urban Environment, Chinese Academy of Sciences) in a coastal city (Xiamen) of Southeast China (Fig. S1 in the Supplement). Here, ClNO₂, N₂O₅, gaseous pollutants (volatile organic compounds (VOCs), NO_x, SO₂, CO, and O₃), aerosol mass concentrations, ionic components, size distribution, and meteorological factors were simultaneously detected. Meanwhile, an iodide-adduct time-of-flight chemical ionization mass spectrometer (I⁻-ToF-CIMS) was used to measure ClNO₂ and N₂O₅. The principles and settings of the I⁻-ToF-CIMS were similar to previous studies (Ma et al., 2023; Yan et al., 2023). Detailed descriptions of this observation site and these instruments have been provided in previous work (Chen et al., 2024; Hu et al., 2022; see Text S1 and Table S1 in the Supplement). For the calibrations of ClNO₂ and N₂O₅, ClNO₂ was produced by passing Cl₂ (6 ppm in N₂) through a moist mixture of sodium nitrite (NaNO₂) and sodium chloride (NaCl) (Thaler et al., 2011; Wang et al., 2022), and N₂O₅ was synthesized by the reactions of O₃ and excessive NO₂ (Tham et al., 2016; Wang et al., 2016). The dependences of ClNO₂ and N₂O₅ sensitivities on relative humidity are presented in Fig. S2 in the Supplement. The uncertainties of the ClNO₂ and N₂O₅ measurements were estimated to be ∼15 % and 12 %, respectively. The details of ClNO₂ and N₂O₅ calibrations and uncertainty analysis are displayed in Text S2 in the Supplement.

2.2 Machine learning model

Here, the extreme gradient boosting (XGBoost) model coupling with the Shapely additive explanations (SHAP) model (the XGBoost–SHAP model) was used to identify the key influencing factors of ClNO₂ formation. Meanwhile, the XGBoost model was applied to establish the predictive model of ClNO₂ based on the observed data of gaseous precursors and meteorological factors; the SHAP model was employed to evaluate the importance of each feature affecting the simulated concentrations of ClNO₂. The SHAP model is an interpretability tool designed to analyse the contributions of individual features to model predictions. It employs an additive explanatory framework that regards all features as contributors, drawing inspiration from cooperative game theory. For each predicted instance, SHAP assigns a Shapley value, representing the cumulative contribution of each feature. Positive SHAP values indicate that a feature increases the model's predicted outcome, signifying a positive contribution. Conversely, negative SHAP values suggest that the feature reduces the predicted value, reflecting a negative contribution. The absolute value of the SHAP score reflects the magnitude of the contribution, regardless of direction, offering insight into the overall importance of the feature. The true value, on the other hand, reveals the direction of the contribution (positive or negative), facilitating a clearer understanding of the relationship between the feature and the prediction. Besides, the partial dependence plot (PDP) analysis offers a visual representation of the marginal effect that the factors have on the model's predicted outcome. It is based on the principle of stabilizing the values of non-target features, and it systematically altered the target feature's values according to the model's algorithmic framework to derive the predicted values.

ClNO₂ concentrations served as the dependent variable, with trace gases (SO₂, CO, NO₂, NO, O₃, and N₂O₅), PM_2.5 and its inorganic compositions (${\mathrm{NO}}_{\mathrm{3}}^{-}$, ${\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}$, ${\mathrm{NH}}_{\mathrm{4}}^{+}$, and Cl⁻), and meteorological parameters (T, RH, UV, wind speed (WS), wind direction (WD), and boundary layer height (BLH)) acting as independent variables. The simulated ClNO₂ concentrations by the XGBoost model were highly similar to the observed values (R²=0.91), indicating the good performance of the XGBoost model (Fig. S3 in the Supplement). Detailed introductions and settings of the XGBoost–SHAP model are provided in Text S3 in the Supplement.

Figure 1 The time series of ClNO₂, related precursors, and meteorological parameters during the autumn observation period.

Figure 2 Diurnal variations in ClNO₂ and other related parameters for the highest concentrations of ClNO₂ (case) on 28 November (a) and the observation-average conditions (from 9 October to 5 December) (b).

2.3 The box model

The observation-based model (OBM) was utilized to assess the impacts of ClNO₂ on photochemically atmospheric oxidation. As delineated in earlier studies (Xue et al., 2015; Tham et al., 2016; Xia et al., 2021; Peng et al., 2021, 2022), the Master Chemical Mechanism (MCM; version 3.3.1) was adopted, and established chlorine chemistry mechanisms have been integrated. The Tropospheric Ultraviolet and Visible (TUV) radiation model was used to calculate ClNO₂ photolysis rates (JClNO₂) under clear-sky conditions. The simulated JClNO₂ values were then scaled based on field-measured JNO₂ values. A thorough exposition of the box model configuration can be found in our previous publications (Liu et al., 2022b, a) and in Text S4 in the Supplement. Observation data, including ClNO₂, VOCs, HCHO, HONO, CO, O₃, NO, NO₂, and SO₂, along with meteorological factors as constraint, were input into the box model at an hourly resolution (Table S2 in the Supplement). Two scenarios were examined: one representing observation-average conditions from 9 October–5 December and the other reflecting a high-ClNO₂ case observed on 28 November.

This study focused on elucidating the influence of ClNO₂ on the formation of RO_x radical and O₃. The O₃ production rate minus the O₃ loss rate was used to calculate the net O₃ production rate (Eqs. S1–S3 in the Supplement). The AOC is calculated by the sum of the rates of CH₄, CO, and VOCs oxidized by atmospheric oxidants (O₃, OH, Cl, and NO₃ radical) (Eq. S4 in the Supplement) (Xue et al., 2015; Yi et al., 2023). Both scenarios were evaluated with and without including ClNO₂ inputs to assess its impacts on these processes.

3 Results and discussion

3.1 Overview of observations

Figure 1 displays the time series of ClNO₂, N₂O₅, and related parameters, including O₃, NO_x, PM_2.5, Cl⁻, ${\mathrm{NO}}_{\mathrm{3}}^{-}$, and meteorological parameters, during the autumn observation period. Our observation shows a decline in T and UV values from October to November, with average RH values increasing from ∼60 % in October to ∼70 % in November (excluding rainy days). During the entire measurement period, ClNO₂ concentrations exhibited significant variability, with elevated levels (>500 ppt) frequently observed in late autumn, particularly after 10 November. The elevation of ClNO₂ concentrations coincided with increased levels of N₂O₅ and ${\mathrm{NO}}_{\mathrm{3}}^{-}$ during late autumn. The concentrations of ClNO₂ at our study site reached several parts per billion (ppb), compared with previous field measurements conducted at urban, suburban, rural, background, and mountain sites (Table S3 in the Supplement), indicating its widespread presence in diverse atmospheric environments. The highest concentrations of ClNO₂ were detected during the night of 27 November, with a maximum hourly average of 3.4 ppb. Peak concentrations of N₂O₅ and ${\mathrm{NO}}_{\mathrm{3}}^{-}$ were also observed on that night (Fig. 1). On the evening of 27 November, N₂O₅ concentrations rapidly decreased after 07:00 p.m. (local time, LT), while ClNO₂ and ${\mathrm{NO}}_{\mathrm{3}}^{-}$ concentrations significantly increased, reflecting fast N₂O₅ heterogeneous hydrolysis and effective formation of ClNO₂. Notably, on the following day (28 November) (Fig. 2a), ClNO₂ concentrations sustained above 100 ppt around noon, partially related to weakened UV values ($\sim \mathrm{14}\mathrm{W}{\mathrm{m}}^{-\mathrm{2}}$) under heavy fog and cloud cover, with the RH values of ∼70 % at that time. Similar research in California has shown ClNO₂ concentrations exceeding 100 ppt 4 h after sunrise due to reduced photolysis (Mielke et al., 2013).

The average diurnal changes in ClNO₂ and related parameters during the entire measurement campaign are depicted in Fig. 2b. As expected, ClNO₂ exhibited a distinct diurnal variation, peaking and accumulating after sunset and decreasing in the early morning. However, ClNO₂ concentrations remained ∼40 ppt around noon, which is different to some other studies, in which ClNO₂ concentrations decreased to near the detection limit around midday (Wang et al., 2022; Niu et al., 2022). A similar observation in North China declared ClNO₂ concentrations above 60 ppt in the afternoon (Liu et al., 2017). Previous studies have indicated that abundant ClNO₂ may be transported from the upper atmosphere or air mass, contributing to the elevated ClNO₂ concentrations in the early morning (Tham et al., 2016; Xia et al., 2021; Jeong et al., 2019). However, the explanations for the concentrations of ClNO₂ around noon remained elusive.

To evaluate the contribution of the heterogeneous N₂O₅ uptake to daytime ClNO₂ levels, we calculated ClNO₂ production using a box model, considering (1) the contribution of heterogeneous N₂O₅ uptake to ClNO₂ production and (2) ClNO₂ loss via photolysis, aerosol uptake, and reaction with OH (Text S5 and S6 in the Supplement). We used a γ(N₂O₅) value of 0.06, a ϕ(ClNO₂) value of 1.0, and a γ(ClNO₂) value of 0.006 in our calculations, which represent upper-end estimates based on previous field studies (Mcduffie et al., 2018a, b; Tham et al., 2016). However, as shown in Fig. 3, the simulated daytime ClNO₂ concentrations were lower than the observed values. Therefore, we believe that the observed daytime ClNO₂ levels, particularly around noon, cannot be adequately explained by heterogeneous N₂O₅ uptake alone, suggesting the presence of additional sources contributing to the formation of daytime ClNO₂.

Figure 3 Comparisons between daytime ClNO₂ levels in observations and in simulations with a ϕ(ClNO₂) of 1.0 and a γ(N₂O₅) of 0.06 (ϕγ=0.06).

3.2 Key drivers of ClNO₂ formation

The XGBoost–SHAP model was employed to investigate the major drivers of ClNO₂ production during the whole observation period. The average absolute SHAP value of each feature was ranked to determine the key drivers of ClNO₂ formation, with larger SHAP values suggesting greater contributions (Fig. 4a). Additionally, features with positive SHAP values (depicted as red points) indicate that higher values of those features positively affect ClNO₂ concentrations and vice versa (Fig. 4b). Overall, N₂O₅, ${\mathrm{NO}}_{\mathrm{3}}^{-}$, T, RH, and UV were the most important features affecting ClNO₂ concentrations. Notably, these factors exhibited varied behaviours between daytime and nighttime periods.

Figure 4 Relative importance of each feature to ClNO₂ using XGBoost–SHAP during the autumn observation period. The mean absolute SHAP value (a) and a summary plot of SHAP values of each feature (b).

In our study, N₂O₅ was identified as the most important influencing factor, consistent with its role in ClNO₂ formation through heterogeneous uptake processes (Thornton et al., 2010; Finlayson-Pitts et al., 1989). After sunset, ClNO₂ concentrations markedly increased due to active nighttime N₂O₅ chemistry, while this heterogeneous uptake process was hindered after sunrise as N₂O₅ concentrations decreased significantly (Fig. 1) (Niu et al., 2022; Wang et al., 2020; Tan et al., 2022). Indeed, the concentrations of ClNO₂ were evidently increased when N₂O₅ concentrations exceeded ∼13 ppt, predominantly during the nighttime (Fig. 5a). Conversely, in northern Europe, the ClNO₂ concentrations were mainly controlled by O₃ and NO₂, rather than by the heterogeneous uptake of N₂O₅ (Sommariva et al., 2018). In Heshan of South China, chloride and PM_2.5 were the major factors affecting ClNO₂ formation (Wang et al., 2022). Differently, the relative importance of ${\mathrm{NO}}_{\mathrm{3}}^{-}$ derived from the XGBoost–SHAP result indicated that elevated ClNO₂ concentrations were associated with high concentrations of ${\mathrm{NO}}_{\mathrm{3}}^{-}$ besides N₂O₅. According to Fig. 5b, high ${\mathrm{NO}}_{\mathrm{3}}^{-}$ concentrations ($>\mathrm{3.7}\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}$) are accompanied by the elevation of ClNO₂, especially its concentrations reaching 6.2 µg m⁻³. Previous studies suggested that increased concentrations of ${\mathrm{NO}}_{\mathrm{3}}^{-}$ decreased γ(N₂O₅), which would limit the production of ClNO₂ (Wahner et al., 1998; Mentel et al., 1999; Bertram and Thornton, 2009). As depicted in Fig. S4 in the Supplement, the dependence of γ(N₂O₅) on ${\mathrm{NO}}_{\mathrm{3}}^{-}$ concentrations follows the nitrate suppression effect. Therefore, the importance of nighttime ${\mathrm{NO}}_{\mathrm{3}}^{-}$ for ClNO₂ levels is that they are co-products from the processes of N₂O₅ heterogeneous uptake. As shown in Fig. 1, compared to low-${\mathrm{NO}}_{\mathrm{3}}^{-}$ conditions, ClNO₂ production was enhanced in high-${\mathrm{NO}}_{\mathrm{3}}^{-}$ conditions. Especially in late autumn, increased aerosol abundances and N₂O₅ levels enhanced N₂O₅ uptake, further promoting ClNO₂ and ${\mathrm{NO}}_{\mathrm{3}}^{-}$ production. Considering the limited contribution of N₂O₅ hydrolysis to daytime ${\mathrm{NO}}_{\mathrm{3}}^{-}$ levels (Yan et al., 2023; Zang et al., 2022; Chen et al., 2020), the impact of high ${\mathrm{NO}}_{\mathrm{3}}^{-}$ concentrations on daytime ClNO₂ concentrations warrants further analysis.

Figure 5 Isolation plots of PDP for N₂O₅ (a), ${\mathrm{NO}}_{\mathrm{3}}^{-}$ (b), T (c), RH (d), and UV (e). The average variations in simulated ClNO₂ with changes in factors are indicated by the splines with yellow and black curves, and blue curves present all situations during the whole observation period.

The simulated concentrations of ClNO₂, based on the XGBoost–SHAP model, were significantly elevated when ${\mathrm{NO}}_{\mathrm{3}}^{-}$ concentrations were higher than 3.7 µg m⁻³ (Fig. 5b). Consequently, the average daily concentrations of ${\mathrm{NO}}_{\mathrm{3}}^{-}$ were classified as high ($>\mathrm{3.7}\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}$) and low ($<\mathrm{3.7}\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}$) to further elucidate the impacts of ${\mathrm{NO}}_{\mathrm{3}}^{-}$ on the formation of ClNO₂. Figure 6 presents the diurnal variations in the relative importance of factors based on the SHAP values under high and low ${\mathrm{NO}}_{\mathrm{3}}^{-}$ concentrations. Unexpectedly, daytime ${\mathrm{NO}}_{\mathrm{3}}^{-}$ was the dominant influencing factor for daytime ClNO₂ (Fig. 6a). High concentrations of daytime ${\mathrm{NO}}_{\mathrm{3}}^{-}$ positively affected the daytime concentrations of ClNO₂, independent of N₂O₅ uptake processes. As depicted in Fig. 6a, daytime N₂O₅ did not promote the elevation of daytime ClNO₂. Negative SHAP values for N₂O₅ during the daytime indicate that the contribution of N₂O₅ chemistry to daytime ClNO₂ levels was limited. Therefore, it is very likely that high concentrations of daytime ${\mathrm{NO}}_{\mathrm{3}}^{-}$ participated in daytime ClNO₂ production. A recent study suggested that nitrate photolysis produced ClNO₂ in addition to Cl₂ (Dalton et al., 2023), but this has not been verified by field observations. Figure 7 shows that daytime ClNO₂ concentrations correlated well (R=0.62) with the product of a proxy of ${\mathrm{NO}}_{\mathrm{3}}^{-}$ photolysis $({\mathrm{NO}}_{\mathrm{3}}^{-}\times J{\mathrm{NO}}_{\mathrm{2}}\times S_{\mathrm{a}}$) on aerosol surfaces, implying that the photolysis of ${\mathrm{NO}}_{\mathrm{3}}^{-}$ likely resulted in the daytime formation of ClNO₂ at our study site. Furthermore, high concentrations of ${\mathrm{NO}}_{\mathrm{3}}^{-}$ and Cl⁻, along with large values of S_a (Figs. 7a–c and S5 in the Supplement) in the daytime accelerated ${\mathrm{NO}}_{\mathrm{3}}^{-}$ photolysis, promoting the formation of ClNO₂, while ClNO₂ concentrations exhibited a weak correlation with JNO₂. It should be emphasized that the weak correlation between JNO₂ and ClNO₂ concentrations does not deny the potential contribution of nitrate photolysis, which could be explained by the fact that ClNO₂ concentrations are affected by both its production and loss processes. Specifically, photolysis rates exert dual effects on daytime ClNO₂ concentrations: positive effects through photochemical production pathways and negative effects through direct ClNO₂ photolytic loss. Given the short daytime lifetime of ClNO₂, we calculated the missing ClNO₂ production rate (production rate minus loss rate) to assess the contributions from unknown sources (Text S6). The production rates of unknown sources showed a good correlation with JNO₂ (R=0.41) (Fig. S6 in the Supplement), indicating that photochemical processes may enhance ClNO₂ production. Notably, the strong correlation between the observed concentrations of ClNO₂ and the ${\mathrm{NO}}_{\mathrm{3}}^{-}$ photolysis proxy (${\mathrm{NO}}_{\mathrm{3}}^{-}\times J{\mathrm{NO}}_{\mathrm{2}}\times S_{\mathrm{a}}$) has revealed the possibility of the contribution of ${\mathrm{NO}}_{\mathrm{3}}^{-}$ photolysis to the unknown daytime ClNO₂ source.

Figure 6 The diurnal variations in the relative importance of factors to ClNO₂ based on the SHAP values under the high ($>\mathrm{3.7}\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}$) (a) and low ($<\mathrm{3.7}\mathrm{\mu }\mathrm{g}{\mathrm{m}}^{-\mathrm{3}}$) (b) ClNO₂ concentrations.

Figure 7 The relationship of daytime ClNO₂ concentrations (12:00–15:00 local time) and a proxy of nitrate (${\mathrm{NO}}_{\mathrm{3}}^{-}$) photolysis (${\mathrm{NO}}_{\mathrm{3}}^{-}\times J{\mathrm{NO}}_{\mathrm{2}}\times S_{\mathrm{a}}$). The colours of the dots represent ${\mathrm{NO}}_{\mathrm{3}}^{-}$ (a), S_a (b), Cl⁻ (c), and JNO₂ (d) amounts.

In terms of meteorological factors, UV, T, and RH were the major influencing factors. The photolysis was the most important sink of ClNO₂ in the daytime, leading to a rapid reduction in ClNO₂ concentrations, particularly in the early morning (Fig. 5e and Fig. 6). However, it is crucial to understand the dual role of photolysis intensity in determining daytime ClNO₂ levels. As mentioned before, photolysis can contribute to the generation of ClNO₂ by promoting ${\mathrm{NO}}_{\mathrm{3}}^{-}$ photolysis, while also causing the rapid decomposition of ClNO₂. As reported in California (Mielke et al., 2013), reduced photolysis rates even increased daytime ClNO₂ levels by decreasing ClNO₂ loss through photolysis. The impact of ambient temperature on ClNO₂ was probably reflected in its thermal equilibrium with N₂O₅. Elevated daytime ambient temperature suppressed the formation of N₂O₅, resulting in low N₂O₅ concentrations, which further limited the contribution of heterogeneous N₂O₅ uptake to daytime ClNO₂ generation (Figs. 5c and 6). During the whole observation period from October to November, the drop in ambient temperature facilitated ClNO₂ production by decreasing the thermal decomposition process. Increased RH values provided favourable conditions for the nighttime N₂O₅ hydrolysis reactions, thereby affecting ClNO₂ production (Figs. 5d and 6), while high RH (>80 %) also weakened the generation of ClNO₂. Notably, Cl⁻ was not the most important factor of ClNO₂ formation at our study site (Fig. 4), likely attributed to the abundant chlorine source in coastal regions (Peng et al., 2022).

3.3 Impact of ClNO₂ photolysis on RO_x budget

The photochemical effects of ClNO₂ were evaluated under the observation-average conditions and the high-ClNO₂ case based on the box model. The largest Cl production rates (P(Cl)) contributed by ClNO₂ photolysis were 0.05 ppb h⁻¹ for the observation-average conditions, which were lower than 0.19 ppb h⁻¹ for the high-ClNO₂ case. The difference led to variable levels of atmospheric oxidation capacity induced by Cl radical. Cl radical released via the photolysis of ClNO₂ initiated the oxidation of VOCs. Among VOC groups (including alkanes, alkenes, alkynes, aromatics, and oxygenated volatile organic compounds (OVOCs)), Cl radical primarily oxidized alkanes (∼65.0 %), followed by OVOCs (∼12.7 %) for both the observation-average conditions and the high-ClNO₂ case (Fig. 8a and b). The contributions of Cl radical and other atmospheric oxidants (including OH radical and O₃) to daytime VOC oxidation were also compared (Fig. 8c, d and Table 1). In our study, the oxidation of alkanes by Cl radical for the observation-average conditions was about 11.7 %, which increased by 44.8 % for the high-ClNO₂ case and was higher than that in London (Bannan et al., 2015), Weybourne (Bannan et al., 2017), Boston (Rutherford et al., 1995), and LA (Fraser et al., 1997) and lower than that in Hong Kong (Xue et al., 2015). It should be noted that the rates of Cl radical reacting with alkanes even exceeded those of OH radical in the early morning for the high-ClNO₂ case. The largest rates of alkanes oxidized by Cl radical were approximately twice as high as those of OH radical at 10:00 a.m. (LT) (Fig. 8e and f), highlighting that the photochemical effects of Cl radical released via ClNO₂ photolysis were particularly important for VOC oxidation during the morning hours at our study site.

Figure 8 The impacts of Cl radials released by ClNO₂ photolysis and other atmospheric oxidants (including OH, NO₃, and O₃) on VOC oxidation under the observation-average conditions and high-ClNO₂ case, respectively. The contributions of different VOC groups oxidized by Cl radical during the observation average (a). The contributions of different VOC groups oxidized by Cl radical during the case (b). The contributions of different atmospheric oxidants (including OH, Cl, NO₃, and O₃) to VOC groups during the observation average (c). The contributions of different atmospheric oxidants (including OH, Cl, NO₃, and O₃) to VOC groups during the case (d). Comparisons of alkane oxidation rates ($\mathrm{molecules}{\mathrm{cm}}^{-\mathrm{3}}{\mathrm{s}}^{-\mathrm{1}}$) by OH and Cl radical during the observation average (e). Comparisons of alkane oxidation rates by OH and Cl radical ($\mathrm{molecules}{\mathrm{cm}}^{-\mathrm{3}}{\mathrm{s}}^{-\mathrm{1}}$) during the case (f).

Figure 9 The contributions of different production pathways to RO_x production rates under the observation-average conditions (a) and high-ClNO₂ case (b).

Table 1 Relative importance of Cl, OH, and O₃ to the daytime oxidation of VOC groups (including alkanes, alkenes, alkynes, aromatics, and OVOCs) around the world (Xue et al., 2015; Bannan et al., 2015, 2017; Rutherford et al., 1995; Fraser et al., 1997).

Figure 10 The impacts of Cl radials released by ClNO₂ photolysis on net O₃ production rates and the AOC levels under the observation-average conditions (a, c) and high-ClNO₂ case (b, d).

Table 2 The impacts of ClNO₂ photolysis on RO_x (OH, HO₂, and RO₂) levels, P(RO_x), and P(O₃) around the world (Xia et al., 2021; Wang et al., 2022, 2016; Tham et al., 2016; Xue et al., 2015; Bannan et al., 2017; Jeong et al., 2019).

The oxidation of VOCs by Cl radical further affects the generation of RO_x ($\mathrm{OH}+{\mathrm{HO}}_{\mathrm{2}}+{\mathrm{RO}}_{\mathrm{2}}$) radicals. The RO_x radical production rates for the high-ClNO₂ case were evidently lower than those under the observation-average conditions, primarily due to reduced photolysis rates on that day. However, the total RO_x radical production rates averagely increased by 23.8 % with ClNO₂ photolysis for the high-ClNO₂ case, higher than a 4.9 % increase for the observation-average conditions (Fig. S7 in the Supplement). For the observation-average conditions, O₃ (32.7 %), HONO (31.7 %), and OVOC (21.6 %) photolysis were the most significant contributors to RO_x radical production in the early morning (07:00–10:00 a.m. (LT)), with VOC oxidation by Cl radical contributing only 3.7 % (Fig. 9a). However, for the high-ClNO₂ case, VOC oxidation induced by Cl radical in the early morning accounted for 19.1 % of RO_x radical production, which was higher than O₃ (7.4 %) and HCHO (4.1 %) photolysis, close to OVOC (19.0 %) photolysis (Fig. 9b). The contributions of ClNO₂ photolysis to the RO_x radical production rates in our study were larger than previous results observed in autumn in Heshan (Wang et al., 2022) and North China (Xia et al., 2021), similar to those in summer in Wangdu (Tham et al., 2016). Thus, the concentrations of OH, HO₂, and RO₂ radicals in the box model with ClNO₂ inputs averagely increased by 17.9 %, 34.6 %, and 54.3 % for the high-ClNO₂ case, higher than the increases of 3.7 %, 7.1 %, and 10.3 % contributed by the observation-average conditions, respectively (Fig. S8 in the Supplement). The uplift in the concentrations of RO_x radicals also accelerated the generation of O₃. The increase in the net O₃ production rates (P(O₃)) for the observation-average conditions reached 0.13 ppb h⁻¹ (15.8 %) in the daytime on average (Fig. 10a), while larger elevations in the net P(O₃) were observed for the high-ClNO₂ case (Fig. 10b), with a maximum of 0.64 ppb h⁻¹ (120 %) at 10:00 a.m. (LT). As a result, increased RO_x radical and O₃ greatly enhanced the atmospheric oxidation capacity (Fig. 10c and d), especially for the high-ClNO₂ case (up to 65 %).

Table 2 summarizes the impacts of ClNO₂ photolysis on RO_x radical and O₃ production in our study and previous observations around the world (Xia et al., 2021; Wang et al., 2022, 2016; Tham et al., 2016; Xue et al., 2015; Bannan et al., 2017; Jeong et al., 2019), indicating that the photochemical impacts of ClNO₂ were variable in different atmospheric environments. At our study site, the effects of ClNO₂ photolysis on RO_x radical production were important, especially in the early morning. The enhanced RO_x radical production induced by ClNO₂ photolysis accelerated the chemical generation of O₃. Primary RO_x radical production rates (including O₃, HONO, HCHO, OVOCs, and ClNO₂) were regarded as one of the most important parameters to O₃ formation (Lu et al., 2023). Therefore, the considerable contribution of ClNO₂ photolysis to primary RO_x radical production in the early morning may bring new challenges for O₃ alleviation.

4 Conclusions

In conclusion, we present 2 months of field measurements in the coastal area of Southeast China during the autumn, coupled with machine learning and model simulations, providing new insights into ClNO₂ chemistry. Our observation shows that the increase in the concentrations of ClNO₂ was accompanied by elevated concentrations of N₂O₅ and ${\mathrm{NO}}_{\mathrm{3}}^{-}$, low values of T and UV, and high values of RH. The nighttime heterogeneous uptake of N₂O₅ was identified as the major source of ClNO₂, while ${\mathrm{NO}}_{\mathrm{3}}^{-}$ photolysis served as a potential daytime ClNO₂ source. Cl radical released by ClNO₂ photolysis after sunrise had important photochemical effects in the early morning. The photolysis of high ClNO₂ concentrations resulted in net O₃ production rates and atmospheric oxidation capacity levels increasing by 120 % and 65 %, respectively. Our results enhanced the understanding of ClNO₂ chemistry in coastal regions, calling for more observations and laboratory research to fully reveal its exact role in different atmospheric environments.