Numerical modeling on the mechanisms of chlorine chemistry in snowpack and their impact on secondary atmospheric pollution

Snow with high albedo enhances atmospheric photochemical reactions, influencing key oxidative processes. Nitryl chloride (ClNO₂), as a strong oxidizing species, is generated by the heterogeneous reaction between dinitrogen pentoxide (N₂O₅) and chloride adsorbed on aerosol and the ground surfaces. After sunrise, the photolysis of ClNO₂ rapidly releases highly reactive chlorine radicals (Cl•), which contributes to the formation of secondary pollutants. However, the pollution mechanisms in high-latitude, snow-covered regions associated with increased chlorine emissions remain unclear. In this study, we employed the WRF-CAMx model (Weather Research and Forecasting Model-Comprehensive Air Quality Model with extensions) with a modified chemical mechanism (CB6r2h_lts, Carbon Bond 6 revision 2 with heterogeneous chemistry for low-temperature and snow-covered conditions) that incorporated heterogeneous N₂O₅ reactions and ClNO₂ photolysis on ground surfaces to assess their impact on regional atmosphere under snow-covered conditions in Northeast China. Our findings reveal that under snow-covered conditions, the YU20 aerosol scheme (from study by Yu et al., 2020) outperforms the BT09 scheme (from study by Bertram and Thornton, 2009) in simulating N₂O₅ and ClNO₂ concentrations within the CAMx model. Incorporating anthropogenic chlorine emissions and ground surface chemistry significantly improved model performance for ClNO₂, reducing the mean bias (MB) from −105.78 to 2.66 pptv and increasing the index of agreement (IOA) from 0.39 to 0.86. These processes resulted in a maximum hourly increase of 3.65 µg m⁻³ in PM_2.5 (relative contribution: 15.34 %) and 3.41 ppbv in MDA8 O₃ (5.68 %). Notably, ground surface chemical processes were identified as the dominant source of nocturnal ClNO₂, contributing approximately 28 % to nighttime accumulation across Northeast China. These findings not only highlight the pivotal role of chlorine chemistry in atmospheric processes under snow-covered conditions, but also provide crucial support for the refinement of the mechanisms governing the flux exchange of chemical substances between the atmosphere and the cryosphere.

1 Introduction

Reactive nitrogen and chlorine atoms are prominent active gases in the atmosphere, influencing the budget of the atmospheric oxidizing capacity (AOC) (Yang et al., 2022; Sun et al.,2026). N₂O₅ serves as a key species in nocturnal chemical reactions within the troposphere. It undergoes heterogeneous reactions with chlorine-containing aerosols and the surfaces of various media (vegetation, soil, snow, and buildings, etc.) to form nitryl chloride (ClNO₂) (Wang et al., 2017, 2020; Jeong et al., 2023). Photolysis of ClNO₂ in the atmosphere produces highly reactive chlorine atoms (Cl) which react with alkanes at rates approximately two orders of magnitude faster than hydroxyl radicals (OH) (Jeong et al., 2023). Therefore, accurately quantifying the impact of N₂O₅ absorption coefficient (γ(N₂O₅)) and ClNO₂ yield (ϕ(ClNO₂)) at different interfaces is crucial for clarifying the contribution of ClNO₂ to pollutants formation.

The γ(N₂O₅) represents the net probability of N₂O₅ undergoing irreversible uptake by an aerosol surface upon collision. Accurately quantifying γ(N₂O₅) values under varying environmental conditions remains challenging. This parameter demonstrably fluctuates in response to changes in nitrate concentration, liquid moisture content, chloride abundance, and organic matter content (Mentel et al.,1999; Thornton and Abbatt, 2005; Bertram and Thornton, 2009; Mielke et al., 2013; Gaston et al., 2014; McDuffie et al., 2018b; Tham et al., 2018; Yu et al., 2020).

Prominent among existing parameterizations is that of Bertram and Thornton (2009), who proposed a method for calculating the uptake coefficient of N₂O₅ on particulate matter surfaces (Reactions R1–R4) and the production yield of ClNO₂ (ϕ(ClNO₂)) from environmental chamber experiments. This framework has been implemented in numerous regional air quality models (Dai et al., 2020; Li et al., 2016; Yu et al., 2020). However, field-based determinations consistently yield lower values than laboratory-derived estimates. Yu et al. (2020) synthesized observational data from multiple regions across China to refine γ(N₂O₅) estimates, demonstrating that laboratory-based parameterizations systematically overestimate this coefficient when compared against ambient measurements. Subsequent field investigations in North China confirmed that the suppression of γ(N₂O₅) is mainly driven by reduced aerosol liquid water content (ALWC), high particulate nitrate (PNO₃), and particle morphological characteristics (Wang et al., 2020). This finding offers a mechanistic explanation for the discrepancies between model simulations and ambient observations.

$$\begin{array}{}\text{(R1)}& {\displaystyle }{\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{5}}\left(\mathrm{g}\right)\stackrel{k_{\mathrm{1}}}{\to }{\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{5}}\left(\mathrm{aq}\right)\text{(R2)}& {\displaystyle }{\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{5}}\left(\mathrm{aq}\right)+{\mathrm{H}}_{\mathrm{2}}\mathrm{O}\left(\mathrm{l}\right)\stackrel{k_{\mathrm{2}}}{\to }{\mathrm{H}}_{\mathrm{2}}{{\mathrm{ONO}}_{\mathrm{2}}}^{+}\left(\mathrm{aq}\right)+{\mathrm{NO}}_{\mathrm{3}}^{-}\left(\mathrm{aq}\right)\text{(R3)}& {\displaystyle }{\mathrm{H}}_{\mathrm{2}}{{\mathrm{ONO}}_{\mathrm{2}}}^{+}\left(\mathrm{aq}\right)+{\mathrm{H}}_{\mathrm{2}}\mathrm{O}\left(\mathrm{l}\right)\stackrel{k_{\mathrm{3}}}{\to }{\mathrm{HNO}}_{\mathrm{3}}\left(\mathrm{aq}\right)+{\mathrm{H}}_{\mathrm{3}}{\mathrm{O}}^{+}\left(\mathrm{aq}\right)\text{(R4)}& {\displaystyle }{\mathrm{H}}_{\mathrm{2}}{{\mathrm{ONO}}_{\mathrm{2}}}^{+}\left(\mathrm{aq}\right)+{\mathrm{Cl}}^{-}\left(\mathrm{aq}\right)\stackrel{k_{\mathrm{4}}}{\to }{\mathrm{ClNO}}_{\mathrm{2}}\left(\mathrm{aq}\right)+{\mathrm{H}}_{\mathrm{2}}\mathrm{O}\left(\mathrm{l}\right)\end{array}$$

In Reactions (R1)–(R4), (g), (aq), and (l) denote the gas phase, aqueous phase, and liquid phase, respectively. However, compared to aerosols, research on the heterogeneous hydrolysis process of N₂O₅ on snowpack remains limited (George et al., 1994; Hanson and Ravishankara, 1991; Lopez-Hilfiker et al., 2012; McNamara et al., 2021). Field observations have confirmed that ice and snow surfaces in the nighttime boundary layer of high-latitude cold regions catalyze the heterogeneous hydrolysis of N₂O₅, leading to ClNO₂ production (Apodaca et al., 2008; Huff et al., 2011; Wang et al., 2020). An upward net flux of ClNO₂ was observed in near-surface snowpack regions, suggesting that saline snowpack may be a source of ClNO₂ (McNamara et al., 2021).

From the perspective of snowpack modeling, accurately quantifying the N₂O₅ adsorption coefficient and the ϕ(ClNO₂) on the snowpack, and coupling them into a one-dimensional numerical model, can provide valuable insights into the contribution of surface snow to this process (McNamara et al., 2020; Wang et al., 2020; Kulju et al., 2022; Jeong et al., 2023). In urban areas, winter snowpack can contribute up to 60 % of near-surface ClNO₂, according to one-dimensional numerical models (Jeong et al., 2023). However, one-dimensional models have limitations in representing the complexity of air-snow exchanges. In contrast, our use of a three-dimensional air quality model, combined with dynamic calculations of surface chemical processes, allows for a more accurate representation of the crucial bidirectional flux exchange between the atmosphere and snow cover.

Northeast China experiences a long period of snow cover in winter, and the heating season lasts up to half a year. Large-scale coal burning has significantly increased chlorine emissions in the atmosphere (Liu et al., 2018; Li et al., 2024). These chlorides adsorb onto particulate matter surfaces, some of which are deposited onto snow, where they undergo heterogeneous reactions with N₂O₅ to produce ClNO₂ (McNamara et al., 2019; Wang et al., 2019; Jeong et al., 2023). These conditions establish the region as an ideal setting for investigating chlorine chemistry fluxes and for performing numerical simulations of air-snow interactions. In this study, we modified the WRF-CAMx model to include a parameterization for N₂O₅ hydrolysis on aerosols and a coupled (3D) surface chemistry module to analyze the impacts on regional particulate matter and ozone formations under snow-cover condition in Northeast China.

2 Methodology

2.1 Field Observations

Field observations of N₂O₅, ClNO₂ and many other chemical species were conducted at the monitoring station of Northeast Institute of Geography and Agroecology (NIGA), Chinese Academy of Sciences from 23 February to 3 March 2024 in Northeast China. The DLS site is in the experimental farmland in the northern suburbs of Changchun City (Fig. 1), with coordinates (44.0° N, 125.4° E). During the observation period, the campaign was characterized by recurrent snowfall events that maintained a continuous snow cover at the surface. N₂O₅ and ClNO₂ were measured using an iodide-adduct chemical ionization mass spectrometer (CIMS).

Figure 1 Sampling site (DLS) information: (a) Satellite-derived snow cover image (data source: National Snow and Ice Data Center, https://nsidc.org/data, last access: 17 September 2026). (b) Traffic conditions around the sampling site at the Northeast Institute of Geography and Agroecology, Chinese Academy of Sciences (NIGA), Jilin Province, China (Imagery © 2026 Airbus, Map data © 2026 Google, https://earth.google.com/, last access: 17 September 2026). (c) The observation station surrounded by snow-covered farmland.

2.2 WRF-CAMx Model and Configuration

The CAMx v7.1 model, incorporating the newly developed low-temperature halogen chemical mechanism (CB6r2h_lts), was employed to simulate the spatio-temporal distributions of atmospheric pollutants (Ramboll, 2021; Emery et al., 2024). The modeling domain was configured to cover the whole Northeast region of China, consisting of 76 × 78 grid cells with a horizontal resolution of 27 km × 27 km and 29 vertical layers, with the model top located at 17 083 m (50 hPa). The model simulation period with a 13 d spin-up was set from 10 February to 2 March 2024. Boundary and initial conditions were derived from the Whole Atmosphere Community Climate Model (WACCM, https://www2.acom.ucar.edu/gcm/waccm, last access: 17 September 2026). Two anthropogenic emission inventories were employed: the Multi-resolution Emission Inventory for China (MEIC; http://meicmodel.org.cn/, last access: 17 September 2026), which was processed to represent the Northeastern China region, and the Emissions Database for Global Atmospheric Research (EDGAR; https://edgar.jrc.ec.europa.eu/, last access: 17 September 2026), used for areas outside China. The anthropogenic chlorine emissions inventory for China (ACEIC) compiled by Sun Yat-sen university was adopted (Li et al., 2024). Chlorine emissions from anthropogenic sources are set to zero in the emission inventory (Table S1 in the Supplement). Meteorological input data for the CAMx model were obtained fromWRFV3.7.1, which utilizes a data nudging method (guv, gt, gq = 0.0001 s⁻¹). More detailed configuration of WRF-CAMx is listed in Table S2.

2.3 Localization of N₂O₅ adsorption reaction coefficients and ClNO₂ productionon aerosol surfaces

The γ(N₂O₅) is influenced by the particulate nitrate (PNO₃), organic matter, Cl⁻, and aerosol liquid water content (ALWC) (Reactions R5–R7). Currently, the heterogeneous reaction parameters of N₂O₅ on aerosols in most air quality models (such as CAMx, CMAQ, etc.) are based on laboratory experiments (Bertram and Thornton, 2009). However, based on field observations, Yu et al. (2020) directly measured γN₂O₅ in ambient aerosols from two rural areas in northern China and Nanjing using an in-situ aerosol flow tube system. The parameters of aerosol surfaces in the above two studies (BT09 and YU20) are shown in Table S3.

$$\begin{array}{}\text{(R5)}& {\displaystyle }\begin{array}{rl}\mathit{\gamma }\left({\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{5}}\right)& ={\displaystyle \frac{\mathrm{4}}{\mathrm{c}}}{\displaystyle \frac{V_{\mathrm{a}}}{S_{\mathrm{a}}}}{K}_{\mathrm{H}}{k^{\prime }}_{\mathrm{2}\mathrm{f}}\\ & \left(\mathrm{1}-{\displaystyle \frac{\mathrm{1}}{\left(\frac{k_{\mathrm{3}}\left[{\mathrm{H}}_{\mathrm{2}}\mathrm{O}\right(l\left)\right]}{k_{\mathrm{2}}b\left[{{\mathrm{NO}}_{\mathrm{3}}}^{-}\right]}\right)+\mathrm{1}+\left(\frac{k_{\mathrm{4}}\left[{\mathrm{H}}_{\mathrm{2}}\mathrm{O}\right(l\left)\right]}{k_{\mathrm{2}}b\left[{\mathrm{Cl}}^{-}\right]}\right)}}\right)\end{array}\text{(R6)}& {\displaystyle }{k}_{\mathrm{2}\mathrm{f}}^{\prime }=\mathit{\beta }-\mathit{\beta }{e}^{(-\mathit{\delta }[{\mathrm{H}}_{\mathrm{2}}\mathrm{O}\left]\right)}\text{(R7)}& {\displaystyle }\mathit{\varphi }\left({\mathrm{ClNO}}_{\mathrm{2}}\right)=(\mathrm{1}+{\displaystyle \frac{k_{\mathrm{3}}\left[{\mathrm{H}}_{\mathrm{2}}\mathrm{O}\right(l\left)\right]}{k_{\mathrm{4}}\left[{\mathrm{Cl}}^{-}\right]}}{)}^{-\mathrm{1}}\end{array}$$

Where $V_{\mathrm{a}}/S_{\mathrm{a}}$ represents the volume-to-surface-area ratio of particles (units: m), serves as a key morphological parameter alongside K_H, the dimensionless Henry's law constant (K_H= [N₂O₅]aq $/$ [N₂O₅]g). To account for H₂O limitation in nitrate-free particles, the rate coefficient for R_2f is treated as a function of liquid water content and redefined as k^′2f in Reaction (R6). The yield of ClNO₂, ϕ(ClNO₂), is given by Reaction (R7). The fitted constants k₃/k_2b, k₄/k_2b, k₃/k₄ are showed (Table S2).

2.4 Coupling Heterogeneous Chemical Processes of N₂O₅ on the ground Surface

The surface chemistry module is activated to simulate gas flux exchange between the atmosphere and the ground, accounting for the chemical reactions of N₂O₅ on surfaces such as snow, vegetation, buildings and soil (Karamchandani et al., 2015). However, the $S/V_g$ parameter in the surface chemistry module is calculated as a one-dimensional plane, leading to an underestimation of the results of the model. To address this, we revised the surface chemistry module to dynamically calculate the $S/V_g$ parameter based on 3D surfaces within each grid cell. To support this revision, we gathered data on surface area of buildings in Northeast China through field investigation and satellite retrieval. The detailed description of revising the surface chemistry module is shown in Fig. S1.

N₂O₅ deposited on snowpack surfaces undergoes simultaneous reactions with hydrochloric acid and water (Table 1), similar to the chemical reactions on aerosol surfaces. The calculation processes for γ(N₂O₅) and ϕ(ClNO₂) on snow grains, which depend on the temperature of the snow, are presented in Text S1 in the Supplement. Under low-temperature conditions, ϕ(ClNO₂) is relatively high, reaching values over 0.80. At a temperature close to approximately 273 K (0 °C), both the γ(N₂O₅) and ϕ(ClNO₂) showed a decreasing trend toward zero (Fig. S2).

Table 1 Revised heterogeneous N₂O₅ reactions in CAMx model.

T represents the surface temperature; K is the reaction rate constant (s⁻¹); J indicates the photolysis rate constant (s⁻¹). ∅ (ClNO₂₎ denotes the yield, without units.

2.5 Scenarios setting

This study investigates the impact of various aerosol schemes, ground chemistry models, and anthropogenic chlorine emissions through six scenarios: B1, Y0, Y1, Y2, Y3, and Y4 (Table 2).

Table 2 Scenario Design Schemes.

Sensitivity analyses were conducted among different simulation scenarios, with the specific configurations detailed as follows: Y3–Y0: The combined effects of anthropogenic chlorine and heterogeneous chemistry of N₂O₅; Y2–Y0: The impact of anthropogenic chlorine emissions; Y2–Y1: The contribution of N₂O₅ heterogeneous chemistry on aerosol surfaces; Y3–Y2: The sensitivity analysis of ground surfaces chemical processes; Y4–Y3: The contribution of ClNO₂ surface photolysis.

3 Result and discussion

3.1 Overview of Wintertime Observations

The time series of N₂O₅ and ClNO₂ concentrations is depicted in the Fig. 2 during the observation period. N₂O₅ concentrations ranged from 0.76 to 424.15 pptv, with an average of 58.73 ± 83.37 pptv. ClNO₂ exhibited an average concentration of 102.64 ± 146.43 pptv with a peak value of 902.45 pptv on 27 February. The peaks of N₂O₅ and ClNO₂ exhibit synchronicity at certain time points during T1, T2, T3, T5, and T6. This phenomenon is attributed to the role of N₂O₅ as the primary precursor of ClNO₂. In winter, the northeastern region of China exhibits abundant chloride-containing aerosols, providing a non-limiting reservoir of reactive chloride. Consequently, the formation of ClNO₂ is governed predominantly by the availability of N₂O₅ rather than by chloride supply. This is evidenced by the observation that during T4, when N₂O₅ mixing ratios remained below ∼120 pptv, no discernible ClNO₂ peaks were detected, despite the persistent presence of chloride species. The positive correlation between N₂O₅ and ClNO₂ further supports that the latter's production is precursor-limited under the prevailing winter conditions. Figure 2b presents the average daily variation patterns of these two species, with a general accumulation at night and rapid depletion during the day, and the similar temporal pattern over snow-covered areas has been widely reported in previous studies (Xia et al., 2020; Kulju et al., 2022; Jeong et al., 2023; Li et al., 2025). The average daily values of N₂O₅ and ClNO₂ were 49.21 ± 37.51 and 56.51 ± 30.70 pptv, respectively. ClNO₂ concentrations exhibited a pronounced diurnal cycle, accumulating at night and reaching a maximum of 113.45 pptv around 07:00 (UTC+8). Following sunrise, rapid photolysis of ClNO₂ led to a sharp decline in its concentrations during daylight hours, accompanied by the release of chlorine atoms (Thornton et al., 2010; Jeong et al., 2023).

Figure 2 Field observations of N₂O₅ and ClNO₂ at the DLS station from 23 February to 3 March 2024. (a) Time series. The gray shaded periods (T1, T2, T3, T5, and T6) represent episodes with concurrent elevated N₂O₅ and CINO₂ peaks, while the cyan shaded period (T4) indicates an episode where no such concurrent peak was observed. The data gap between 14:00 (UTC+8) on 25 February and 10:00 (UTC+8) on 26 February is due to instrument maintenance, during which no measurements were available. (b) Diurnal variation.

3.2 Model performance of the updated heterogeneous N₂O₅ chemistry on aerosol/ground surfaces

3.2.1 Comparative analysis of aerosol schemes on heterogeneous N₂O₅ reaction

This study compared the performance of the BT09 and YU20 aerosol schemes within the CAMx model during snow-covered periods (Fig. S3). The YU20 scheme demonstrated significant improvements in the statistical indicators for N₂O₅ and ClNO₂ compared to the BT09 scheme (Table S4). For N₂O₅, the mean bias (MB) decreased from 114.62 to 69.66, the normalized mean bias (NMB) from 1.92 to 1.16, and the root mean square error (RMSE) from 201.39 to 139.03, while the IOA increased from 0.49 to 0.61. Similarly, for ClNO₂, the MB decreased from −29.28 to −16.08, the NMB from −0.28 to −0.16, the RMSE from 100.66 to 99.80, with the index of agreement (IOA) from 0.81 to 0.84. The YU20 scheme is a set of parameter values derived from local observation data in China, making it more representative of the evolution of atmospheric chemical species in the region. Given the above enhancements, the YU20 and BT09 schemes were selected for snow-covered and snow-free regions, respectively, in the subsequent scenario simulations.

The parameters γ(N₂O₅) and ϕ(ClNO₂) are critical parameters for understanding N₂O₅ heterogeneous chemistry on aerosol surfaces. The BT09 scheme is believed to overestimate the γ(N₂O₅) (Wagner et al., 2013) and ϕ(ClNO₂) (McDuffie et al., 2018a). Yang et al. (2022) further demonstrated that incorporating anthropogenic chlorine and biomass burning emissions into the YU20 parameterization effectively enhanced γ(N₂O₅) values, emphasizing the advantage of localized schemes for improving model accuracy (Yang et al., 2022). However, Xie et al. (2025) found that the BT09 scheme outperformed YU20 at the Wangdu station in Hebei province of China under snow-free conditions using the WRF-CAMx model. Similarly, Dai et al. (2020) evaluated the emission of PCl from the South China Sea and also found that the BT09 aerosol scheme better captured the peak value of ClNO₂ compared to the YU20 scheme.

In contrast, our study found that YU20 performed better than BT09 during snow-covered conditions. This improvement can be attributed to the significant impact of atmospheric humidity changes under snow cover conditions, along with an increase in anthropogenic chlorine emissions (e.g., Cl₂, PCI, HCI, and HOCI) during winter in high-latitude regions. These factors together drove changes in species concentrations (chloride ions, and nitrate ions) on aerosol surfaces during the observation period, directly affecting the γ(N₂O₅) and ϕ(ClNO₂) values in the N₂O₅ aerosol schemes (BT09 or YU20).

3.2.2 Comparison of model simulations with observation

A comparison of atmospheric N₂O₅ and ClNO₂ concentrations before and after model modification is presented in Fig. 3. In the Y0 scenario, N₂O₅ concentrations averaged 141.81 ± 148.62 pptv, with a peak value of 540.37 pptv. In the Y3 scenario with modified modeling, N₂O₅ concentrations ranged from 0.0 to 513.03 pptv, with an average value of 33.58 ± 143.92 pptv. While the MB of N₂O₅ decreased from 64.10 to 56.53 pptv, the NMB from 0.98 % to 0.87 %, and the RMSE from 123.17 to 118.39 pptv, the IOA remained at 0.66 (Table 3). This suggests no significant improvement in model accuracy, likely due to the relatively low re-emission rate of N₂O₅ deposited on the ground surface, which has a minimal impact on atmospheric N₂O₅ concentrations.

Table 3 Statistical indicators for the simulation results of N₂O₅ and ClNO₂ between Y0 and Y3 scenarios.

Figure 3 Comparison of simulated and observed values for YU20_A (Y0) and YU20_A_G_Chl (Y3) the DLS station in Northeast China: (a) N₂O₅, (b) ClNO₂.

In contrast, the revised model significantly improved the simulation of ClNO₂, particularly in capturing peak concentrations. In Y0 scenario, the average ClNO₂ concentrations is 7.73 ± 6.96 pptv with a peak value of 35.59 pptv. In contrast, the Y3 scenario showed an expanded range of 0.07 to 503.30 pptv and an increased average concentration of 110.30 ± 117.45 pptv. Additionally, the MB of ClNO₂ improved from −105.78 to 2.66 pptv, the NMB increased from −0.93 % to 0.02 %, and the RMSE decreased from 170.26 to 88.28 pptv. The IOA also improved significantly, rising from 0.39 to 0.86 (Table 3). These results demonstrate that incorporating chlorine emissions and surface N₂O₅ chemical processes into the CAMx model substantially enhanced the simulation accuracy for ClNO₂, while the impact on N₂O₅ remained limited. In addition, we further examine the effects of the revised model on secondary pollutants (PM_2.5 and O₃) at this site, which are also pronounced. For PM_2.5 the MB improves from −10.56 to −1.39 µg m⁻³, and the IOA increases from 0.68 to 0.91; for O₃, the MB decreases from 19.42 to 13.91 µg m⁻³, while the IOA improves from 0.72 to 0.77 (Table S5). These findings underscore that the updated parameterization not only refines ClNO₂ simulation but also yields significant improvements in simulating secondary aerosol and ozone formation.

3.2.3 The Spatial impacts of chlorine emissions and chlorine chemistry

In northern regions of China, residential heating typically extends from early November through March of the following year. During this period, coal combustion releases substantial quantities of particulate chloride (PCl) and HCl into the atmosphere (Liu et al., 2018). To assess the impact of model modifications, we further examined the spatial concentration differences and ratios of PCl, HCl, N₂O₅, and ClNO₂ between Y0 and Y3 (Fig. 4). The spatial concentration differences (e.g., Y3–Y0) are used to quantify the magnitude of changes in simulated pollutant concentrations between different scenarios, while contribution ratios (e.g., (Y3–Y0) $/$ Y3) are employed to assess the relative contributions of model modifications. For PCl and HCl, peak spatial concentrations in the Y0 scenario reached 0.43 µg m⁻³ and 76.96 pptv, respectively. In the Y3 scenario, these ranges increased to 0–0.97 µg m⁻³ for PCl and 0–77.79 pptv for HCl. The spatial differences in PCl concentrations between the Y0 and Y3 scenarios ranged from 0 to 0.59 µg m⁻³, with the most significant differences simulated in Heilongjiang and Liaoning provinces, where the contribution rate was approximately 72 %. Meanwhile, the spatial differences in HCl concentrations between the Y0 and Y3 scenarios ranged from 0 to 8.21 pptv with the highest contribution rate was 99.20 % in Heilongjiang province. Regarding N₂O₅ and ClNO₂, the N₂O₅ concentration peaked at 310.65 pptv in the Y0 scenario, while the maximum spatial concentration increased to 343.29 pptv in the Y3 scenario. The spatial differences in N₂O₅ between the Y0 and Y3 scenarios ranged from −24.04 to 0.0 pptv (Fig. 4k) with a maximum contribution ratio of 44.87 %, indicating net consumption of N₂O₅. ClNO₂ concentrations peaked at 116.82 pptv in the Y0 scenario and increased to 179.69 pptv in the Y3 scenario. Spatial differences of ClNO₂ between the Y0 and Y3 scenarios ranged from 0 to 173.49 pptv (Fig. 4o) with the contribution rate of ClNO₂ concentrations exceeding 90 % in most simulated regions. These findings highlight the significant impact of incorporating anthropogenic chlorine emissions and revised chlorine chemical reaction mechanisms on regional ClNO₂ concentrations.

Figure 4 The spatial distribution of the impact of chlorine emissions and chlorine chemical reactions on pollutants (PCl, HCl, N₂O₅, and ClNO₂).

3.3 The relative contribution of chlorine emissions and ground surface chlorine chemistry

The comprehensive impact of anthropogenic chlorine emissions and heterogeneous chemical reactions of surface N₂O₅ on ClNO₂ concentration had been quantitatively evaluated by comparing Y0 and Y3 scenarios. Here, we further analyzed the individual effects of chlorine emissions and surface chemical reactions separately.

3.3.1 The impact of chlorine emission on ClNO₂

To isolate the contribution of anthropogenic chlorine emissions, the Y0 and Y2 scenarios were established. In the Y2 scenario, the spatial concentrations of PCl and HCl ranged from 0.0 to 0.95 µg m⁻³ and 0.0 to 77.60 pptv, respectively (Fig. 5b and f). The spatial differences in PCl between the Y0 and Y2 scenarios ranged from 0.0 to 0.58 µg m⁻³ (Fig. 5c), with an average contribution ratio of 61.02 %. Similarly, the spatial differences in HCl concentrations between the Y0 and Y2 scenarios ranged from 0 to 30.56 pptv, with the highest contribution rate of 99.11 % simulated in Heilongjiang Province, China. Notably, the peak ClNO₂ concentration in the Y2 scenario reached 166.82 pptv, with a maximum difference of 151.67 pptv. In most regions of northeastern China, ClNO₂ contributions exceeded 50 %.

Figure 5 The spatial distribution of the impact of chlorine emissions on pollutants (PCl, HCl, N₂O₅, and ClNO₂). Y0 and Y2 denote the YU20_A and YU20_A_chl scenarios, respectively.

For snow-free conditions, Liu et al. (2018) first utilized the ACEIC chlorine emission inventory in the Community Multi-scale Air Quality (CMAQ) modeling system and found that anthropogenic chlorine emissions contributed approximately 100 pptv to ClNO₂ concentrations, with a contribution range of 30 %–50 %. Similarly, Hong et al. (2020) using the CMAQ model with ACEIC inventory, further reported that the maximum contribution of chlorine emissions to ClNO₂ ranged from 50 to 100 pptv in Liaoning Province, corresponding to a contribution rate of 20 %–50 %. In contrast, the simulated contribution concentration range in Jilin and Heilongjiang Provinces was only 10–50 pptv, with a corresponding contribution rate of 10 %–20 %. However, under snow-covered conditions in this study, simulations using the CAMx model and an updated ACEIC inventory yielded ClNO₂ concentrations ranging from 0 to 151.67 pptv, with a contribution range of 50 %–90 %. This suggests that snow cover with higher albedo plays a significant role in the photochemistry of the ClNO₂ generation. Additionally, differences in air quality models and simulation periods may also contribute to the variations observed among these studies.

3.3.2 The impact of ground surface chlorine chemistry on ClNO₂

The existing air quality models typically underestimate ClNO₂ concentrations compared to observed values, suggesting the presence of unknown sources (Wang et al., 2022). To address this gap, we further incorporated two new ground surface chemical reactions including ClNO₂ production and consumption reactions (Table 1) into a 3D air quality model to investigate the contribution of these unknown ClNO₂ sources.

Figure 6 presents the simulated spatial distributions of monthly mean, nighttime monthly mean, and daytime monthly mean ClNO₂ concentrations, along with their corresponding contribution ratios from ground surface chemistry. The spatial differences in ClNO₂ between the Y2 and Y3 scenarios ranged from 0.0 to 38.96 pptv, with the contribution ratios of up to 48.20 % simulated in Heilongjiang, Jilin, and the northeastern regions of Liaoning Province (Fig. 6c and d). The higher albedo of snow can facilitate the surface photochemical reactions and lead to rapid ClNO₂ photolysis (Chen et al., 2019), which explains significant differences in the concentration of ClNO₂between daytime and nighttime. During the daytime, ClNO₂ concentrations in both the Y2 and Y3 scenarios ranged from 0 to 70.60 pptv. The maximum spatial difference in ClNO₂ between the Y2 and Y3 scenarios was 7.51 pptv, with monthly average contribution ratios reaching up to 23.02 % in the eastern regions of Heilongjiang and Jilin Provinces. In contrast, nighttime ClNO₂ concentrations exhibited an accumulation trend. Scenarios Y2 and Y3 simulated maximum nighttime ClNO₂ concentrations of 271.07 and 310.04 pptv, respectively. The nighttime spatial maximum difference (72.14 pptv) was approximately ten times greater than that simulated during the day, with a contribution ratio of around 92 % in Heilongjiang and Jilin Provinces.

Figure 6 The quantitative contribution of ground surface chemistry to hourly atmospheric ClNO₂ over northeastern China.

The flux exchange between ground surface chemical processes and the atmosphere is a critical component of atmospheric chemistry. Pollutants deposited on surfaces (soil, snow, vegetation, and buildings) can also undergo heterogeneous chemical reactions.

Snow, in particular, provides a larger reaction interface for heterogeneous chemical while low temperatures influence chemical equilibria and increase atmospheric moisture content, favoring gas deposition processes (McNamara et al., 2021). Consequently, higher ClNO₂ and N₂O₅ concentrations are observed during winter compared to summer (Xia et al., 2021), primarily due to elevated N₂O₅ levels enhancing ClNO₂ production, while lower winter temperatures favor the equilibrium shift of the N₂O₅–NO₃ reaction toward N₂O₅ formation. Additionally, longer winter nights in high latitudes facilitate N₂O₅ accumulation under dark conditions (Wagner et al., 2013).

The ClNO₂ yield and flux variations serve as key indicators for assessing the significance of surface heterogenous chemical processes. During the snow-covered periods, the φ(ClNO₂) values are higher compared to snow free periods, ranging from 0.065 to 1.00 during most winter months in the Northern Hemisphere (Table 4). Notably, over snow surfaces, Wang et al. (2020) and Jeong et al. (2023) reported the φ(ClNO₂) values even exceeding 0.80 during winter.

Table 4 Summary of literature on the yield of ClNO₂ in winter.

From the perspective of ClNO₂ flux variations, McNamara et al. (2021) conducted vertical gas profile observations and snow chamber experiments in Kalamazoo, Michigan, reporting a daily averaged ClNO₂ flux of 3 × 10⁷ molec. cm⁻² s⁻¹ over snow-covered surfaces, compared to a negative flux of −24 × 10⁷ molec. cm⁻² s⁻¹ over snow-free surfaces. Jeong et al. (2023) further integrated a one-dimensional atmospheric boundary layer model with snow modules to investigate the vertical distribution and impact of urban snow cover on ClNO₂ emissions in Kalamazoo, Michigan. Their findings showed that the ClNO₂ flux was positive on snow-covered nights, averaging 3.7 × 10⁸ molec. cm⁻² s⁻¹, but turned negative on snow-free nights, averaging −(2.8 × 10⁹) molec. cm⁻² s⁻¹. These findings indicate that under snow-covered conditions, the flux of ClNO₂ is positive, whereas under snow-free conditions, it turns negative. Furthermore, the positive flux of ClNO₂ during nighttime snow conditions is significantly higher than the daily average flux.

As mentioned above, snow-covered ground surfaces exhibit higher φ(ClNO₂) values and larger positive ClNO₂ fluxes (Jeong et al., 2023; Kulju et al., 2022; McNamara et al., 2021), particularly at night. In this study, we further integrated the surface chemistry module into the 4D Eulerian model of CAMx, establishing a link for flux exchange between the ground surfaces and the atmosphere. This integration provides a bridge to explore the spatiotemporal variations in ClNO₂ concentration and allows for a more nuanced quantification of the impacts of chlorine chemistry on atmospheric oxidizing capacity and regional air quality.

3.4 Contribution of Reactive Chlorine to AOC, PM_2.5, and O₃

3.4.1 Contribution on Atmospheric Oxidizing Capacity (AOC)

According to previous studies (Lu et al., 2014; Zhu et al., 2022), the incorporation of chlorine chemical reactions can enhance the AOC. The spatial concentrations of oxidizing species (OH, HO₂, and RO₂) in scenarios Y0, Y1, Y2, and Y3 are shown in (Fig. S5). The contributions of anthropogenic chlorine emissions and N₂O₅ heterogeneous chemistry under different conditions are illustrated in Fig. S6. Note that Y3–Y0, Y2–Y1, and Y3–Y2 represent contributions from N₂O₅ heterogeneous chemistry, aerosol surface heterogeneous chemistry, and ground surface heterogeneous chemistry, denoted as Chl_het_N₂O₅_a+g, Het_N₂O₅_a, and Het_N₂O₅_g, respectively.

In the Y3–Y0 scenario, the maximum OH concentration difference reached 3.69 × 10⁵ cm⁻³ with a maximum relative contribution of 7.59 %. This result is consistent with Wang et al. (2020), who reported a 6 % contribution of anthropogenic chlorine emissions to the AOC in Northeast China for 2014 (Wang et al., 2020). The spatial concentration difference in RO₂ ranged from −(4.70 × 10⁶) to 0.17 × 10⁶ cm⁻³ with a maximum relative contribution of 6.56 %. The HO₂ concentration was more regionalized where it primarily concentrated in coastal areas of the Beijing-Tianjin-Hebei region and the Bohai Sea, with the range of −(10.96 × 10⁵) to 7.95 × 10⁵ cm⁻³ and a relative contribution of approximately 3 % in Northeast China.

In the Y2–Y1 scenario, the spatial OH concentration difference ranged from −(0.15 × 10⁴) to 1.88 × 10⁴ cm⁻³ with a maximum relative contribution of 6.34 %. The maximum HO₂ concentration difference (9.04 × 10⁵ cm⁻³) exceeded that simulated in the Y3–Y0 scenario. Furthermore, the RO₂ difference range −(6.98 × 10⁸) to 0.29 × 10⁸ cm⁻³ was also larger than in the Y3–Y0 scenario.

In the Y3–Y2 scenario, the OH concentration difference ranged from −(0.12 × 10⁴) to 3.63 × 10⁴ cm⁻³, and the contribution regions were primarily concentrated in Heilongjiang Province and the four eastern regions of Inner Mongolia with a maximum relative contribution of 7.37 %. The maximum HO₂ contribution was 5.98 %, while the RO₂ contribution was the smallest at 1.99 %. The results illustrate that under snow cover conditions, chlorine chemistry generally promotes OH generation, inhibits RO₂ production, and has varying effects on HO₂ production, inhibiting it in Chifeng city while promoting it in other regions of Northeast China (Fig. S6).

The greater spatial distribution differences simulated in Y2–Y1 compared to Y3–Y0 can be attributed to the CB6r2h-lts chemical mechanism, which includes 22 gas-phase chlorine reactions, eight of which Reactions (R8–R15) directly impact the generation and removal of OH, HO₂, and RO₂. The analysis reveals that the Y2–Y1 scenario had the most significant impact, followed by Y3–Y0, while the contribution of Y3–Y2 was relatively minor.

$$\begin{array}{}\text{(R8)}& {\displaystyle }\mathrm{HOCl}+\mathrm{hv}=\mathrm{Cl}+\mathrm{OH}\text{(R9)}& {\displaystyle }\mathrm{OH}+\mathrm{HCl}=\mathrm{Cl}\text{(R10)}& {\displaystyle }\mathrm{OH}+\mathrm{FMCl}=\mathrm{Cl}+\mathrm{CO}\text{(R11)}& {\displaystyle }\mathrm{FMCl}=\mathrm{Cl}+\mathrm{CO}+{\mathrm{HO}}_{\mathrm{2}}\text{(R12)}& {\displaystyle }\mathrm{ClO}+{\mathrm{MEO}}_{\mathrm{2}}=\mathrm{Cl}+\mathrm{FORM}+{\mathrm{HO}}_{\mathrm{2}}\text{(R13)}& {\displaystyle }\begin{array}{rl}\mathrm{Cl}+\mathrm{ETHA}& =\mathrm{HCl}+\mathrm{0.991}{\mathrm{AlD}}_{\mathrm{2}}\\ & +\mathrm{0.991}{\mathrm{XO}}_{\mathrm{2}}\mathrm{H}+\mathrm{0.009}{\mathrm{XO}}_{\mathrm{2}}\mathrm{N}+{\mathrm{RO}}_{\mathrm{2}}\end{array}\text{(R14)}& {\displaystyle }\begin{array}{rl}\mathrm{Cl}+\mathrm{PRPA}& =\mathrm{HCl}+\mathrm{ACET}+\mathrm{0.97}{\mathrm{XO}}_{\mathrm{2}}\mathrm{H}\\ & +\mathrm{0.03}{\mathrm{XO}}_{\mathrm{2}}\mathrm{N}+{\mathrm{RO}}_{\mathrm{2}}\end{array}\text{(R15)}& {\displaystyle }\begin{array}{rl}\mathrm{Cl}+\mathrm{ISOP}& =\mathrm{FMCl}+\mathrm{ISPD}+\mathrm{0.96}{\mathrm{XO}}_{\mathrm{2}}\mathrm{H}\\ & +\mathrm{0.04}{\mathrm{XO}}_{\mathrm{2}}\mathrm{N}+{\mathrm{RO}}_{\mathrm{2}}\end{array}\end{array}$$

Where FMCL denotes formyl chloride (HC(O)Cl); MEO₂ represents the methyl peroxy radical; AlD₂ refers to acetaldehyde and higher aldehydes; XO₂H indicates NO-to-NO₂ converting peroxy radicals from HCs; XO₂N signifies NO-to-organic nitrate converting peroxy radicals; ISPD represents Isoprene product/Isoprene-derived peroxy radical; FORM denotes formaldehyde; ACET indicates acetone; and ISOP represents isoprene.

3.4.2 Contribution on PM_2.5

Chlorine emissions and the associated chlorine chemical reactions significantly influence AOC, the formation of particulate matter, and ozone levels. Figure S7 presents the spatial concentration distributions of these species with concentration ranges of 0–128.64 µg m⁻³ for PM_2.5, 0–16.94 µg m⁻³ for PNH₄, 0–13.06 µg m⁻³ for PSO₄, and 0–47.53 µg m⁻³ for PNO₃, respectively. Figure 7 exhibited the differences and contribution ratios of scenarios (Y3–Y0), (Y2–Y1), and (Y3–Y2) on PM_2.5 and its components (PNH₄, PSO₄, and PNO₃), respectively.

Figure 7 The quantitative contribution of different pathways of chlorine chemistry on PM_2.5, and its chemical components.

In the Y3–Y0 scenario, chlorine chemistry suppressed PM_2.5 formation in the eastern regions of Northeast China with a maximum suppression of 1.94 µg m⁻³. Conversely, chlorine chemistry promoted PM_2.5 formation in most regions of Heilongjiang, Jilin, and Liaoning Provinces with a significant increase of 3.65 µg m⁻³ and a maximum relative contribution of 15.34 %. The simulation results show that chlorine addition promoted the formation of PNH₄ (increased by 0.73 µg m⁻³) and PSO₄ (increased by 0.37 µg m⁻³) which were primarily concentrated in urban areas such as Harbin and Changchun. In contrast, PNO₃ was suppressed, with a minimum difference value of −1.78 µg m⁻³ and a corresponding relative contribution of 23.77 %.

In the Y2–Y1 scenario, PM_2.5 was inhibited in the eastern regions of Heilongjiang and Liaoning Provinces, while an increase was simulated in certain areas of Jilin Province, with concentration differences ranging from −2.64 to 2.16 µg m⁻³ and a maximum contribution of 3.52 %. Aerosol surface heterogeneous chemistry inhibited PNH₄ and PNO₃ formation across most of the central and northeastern regions, with maximum relative contributions of 3.14 % and 26.25 %, respectively. The spatial concentration difference in PSO${}_{\mathrm{4}}^{\mathrm{2}-}$ was not significant in the Northeast region with the maximum value of 0.09 µg m⁻³.

In the Y3–Y2 scenario, the formations of PM_2.5, PNO₃, PNH₄, and PSO${}_{\mathrm{4}}^{\mathrm{2}-}$ were enhanced, with maximum differences of 0.18, 0.16, 0.05, and 0.01 µg m⁻³, respectively. Compared to Y3–Y0 and Y2–Y1, the spatial concentration difference in Y3–Y2 was relatively minor for these species.

In the Y3–Y2 scenario, the impact on particulate matter is relatively small, primarily due to two factors: (1) the concentration of re-emitted species was low, and the species involved were limited; and (2) highly oxidizing species such as HCl and ClNO₂ in rapidly decomposed into other chlorides within the snow, preventing their release into the atmosphere as oxidants and leading to a net loss of chlorine. Additionally, an analysis of ClNO₂ photolysis in snow cover revealed a maximum consumption value of 2.88 pptv, with a monthly average maximum concentration of 38.96 pptv on the ground, accounting for approximately 7.40 % (Fig. S8).

3.4.3 Contribution on MDA8 O₃

The anthropogenic chlorine emissions and N₂O₅ heterogeneous chemistry can significantly enhance ozone levels. Previous studies have quantified this effect: Wang et al. (2020) reported that anthropogenic chlorine emissions increased the maximum daily 8 h average (MDA8) O₃ concentration by 1.70 ppbv while Yi et al. (2021) found a 6.70 ppbv increase in winter MDA8 O₃ concentration in the Yangtze River Delta (YRD) region due to updates in chlorine chemistry processes. Furthermore, Yang et al. (2022) demonstrated that incorporating the anthropogenic chlorine emissions with biomass burning emissions and updating the parameterization of heterogeneous N₂O₅ and Cl related chemistries, including adjustments of γ(N₂O₅) and ϕ(ClNO₂), led to a 4.5 ppbv increase in the MDA8O₃ concentration over China (Yang et al., 2022). In this study, the maximum difference in monthly mean MDA8 O₃reached 3.41 ppbv (Fig. 8a), which is smaller than the value simulated (4.74 ppbv) in the Y2–Y1 scenario. This difference is primarily attributed to the adding of chlorine emissions, which suppresses atmospheric oxidation. For example, the maximum difference concentrations of RO₂ and HO₂ in Y2–Y1 are 0.29 × 10⁸ and 9.04 × 10⁵ cm⁻³, respectively, while in Y3–Y0, they are 0.17 × 10⁶ and 7.95 × 10⁵ cm⁻³, respectively (Fig. S6). The maximum contribution concentration of surface heterogenous processes Y3–Y2 to ozone was 1.30 ppbv (Fig. 8c).

Figure 8 The quantitative contribution of different pathways of chlorine chemistry on O₃.

To fully understand the dynamic exchange processes between the atmosphere and the ground surfaces, it is essential to quantify the mass of chemical species deposited on various surfaces, such as snow, vegetation, buildings, and soil, which provide sufficiently large reaction interfaces. Incorporating these deposition processes into the air quality models will offer a more comprehensive understanding of ground surface-atmosphere interactions and their impact on atmospheric chemistry. However, the current ground surface chemistry module remains relatively simplified, primarily focusing on the re-emission of species into the atmosphere from predefined reactions. To more accurately represent the full impact of surface chemistry on atmospheric processes, further development and refinement of the ground surface module are necessary.

4 Conclusions

The mechanisms of heterogeneous N₂O₅ reactions and ClNO₂ photolysis in snow remain incompletely understood. In this study, we employed numerical simulations to compare two different aerosol parameterization schemes involving chlorine chemistry (BT09 and YU20) and incorporated newly heterogeneous N₂O₅ reactions on ground surfaces in WRF-CAMx model to assess their impacts on regional – scale atmosphere pollution under snow-covered conditions in Northeast China.

The observation results reveal distinct diurnal cycles of N₂O₅ and ClNO₂, characterized by nighttime accumulation during winter, followed by rapid photolysis after sunrise. The formation of ClNO₂ is primarily limited by the availability of N₂O₅ rather than chloride supply, as evidenced by the absence of ClNO₂ peaks when N₂O₅ remained below ∼ 120 pptv. Ground surface chemical processes accounted for about 28 % of the simulated nighttime ClNO₂ accumulation in Northeastern China, which emphasizes the critical role of snow-covered surfaces in ClNO₂ production.

Furthermore, comparisons between different aerosol schemes demonstrated that the YU20 aerosol scheme outperforms the BT09 scheme in simulating N₂O₅ and ClNO₂ concentrations within the CAMx model. Incorporating anthropogenic chlorine emissions and surface N₂O₅ heterogeneous chemistry significantly improved the model's performance on the simulated ClNO₂ concentrations, reducing the RMSE from 170.26 pptv to 88.28 pptv and increasing the IOA from 0.39 to 0.86. These processes increased hourly PM_2.5 concentrations by up to 3.65 µg m⁻³ (15.34 %) and MDA8 O₃ by 3.41 ppbv (5.68 %), promoting particulate matter (PNH₄ and PSO₄) formation while suppressing PNO₃. Regarding regional impacts on the AOC Atmospheric Oxidizing Capacity, PM_2.5, and MDA8 O₃, the contribution of N₂O₅ heterogeneous chemistry on aerosol surfaces exceeded the combined effect of anthropogenic chlorine emissions and surface heterogeneous chemistry.

This study provides key insights into the role of the heterogenous chlorine chemistry in atmospheric processes over cold residential regions, and highlights unknown sources of ClNO₂ on the ground surfaces during snow cover periods. From a 3D air quality modeling perspective, quantifying the chemical deposition of atmospheric pollutants on diverse surfaces and integrating heterogenous chlorine related reaction processes into ground surface modules are crucial for accurately representing flux exchange between the ground and atmosphere. In the surface chemistry module, although the variation of rate constants for heterogeneous reactions at two snow layers is considered, further optimization is possible. By collecting snow samples at different depths in the field or conducting sensitive experiments in laboratory a long-term quantitative relationship among snow depth, total ion concentration (C_T), and liquid brine layer fraction (f_brine) can be established. This will allow for a more precise determination of the parameter values of γ(N₂O₅) and ϕ(ClNO₂) corresponding to different snow depths. Moreover, the snowmelt process requires special attention, as dynamic variations in liquid water content within snow/ice pores significantly affect heterogeneous reaction rate constants. Future studies should incorporate the numerical representation of these variations in models and evaluate the health impacts of regional atmospheric pollution driven by chemical exchanges between the atmosphere and the cryosphere.