Our observations were conducted at the Atmospheric Observation Supersite (24.61° N, 118.06° E) of the Institute of Urban Environment (IUE), Chinese Academy of Sciences in Xiamen, from 18 November 2023 to 6 January 2024. Xiamen is a coastal city in southeastern China. The observation site is located approximately 5 km from the shore and is surrounded by residential, commercial areas and major traffic arteries. Instruments were deployed on the roof of the IUE building at approximately 70 m above ground level. Virtually no high buildings obstructed the path between the site and the coast, making it a typical coastal urban site.

A suite of atmospheric parameters was observed simultaneously. Ambient HONO was measured online by a water-based long-path absorption photometer (Zhichen Beijing, China) (Xuan et al., 2024). Sulfuric acid (H₂SO₄) and HCl were measured using a long time-of-flight chemical ionization mass spectrometer, equipped with a nitrate reagent ion source (Aerodyne Research Inc., USA) (Yang et al., 2023). The chemical composition of PM_2.5 was analyzed using both a MARGA ADI 2080 (Metrohm Applikon, Switzerland) and an ACSM (Aerodyne Research Inc., USA) (Liu et al., 2022; Chen et al., 2022). Particle number size distribution (PNSD) in the range of 2–300 nm was measured by two scanning mobility particle sizers (TSI inc., USA) equipped with a long-DMA (2.5–62 nm) and a nano-DMA (7–300 nm), respectively (Li et al., 2025). Additionally, a range of commercial instruments were employed to monitor trace gases (NO, NO₂, SO₂, O₃, and VOCs), meteorological parameters (ambient temperature, RH, wind-speed, wind-direction and ultraviolet radiation), and photolysis rates (JO¹D, JNO₂, JHONO, JNO₃, JHCHO and JH₂O₂) (Liu et al., 2022). Detailed descriptions of instrument operation, species identification, and calibration are provided in Sect. S1 in the Supplement. Hourly BLH data were obtained from the ERA5 reanalysis dataset (Hersbach et al., 2020).

Sea-land breeze days were identified based on observed wind-speed and wind-direction according to the criteria established in our previous work (Chen et al., 2023). Briefly, the coastline in Xiamen is characterized by a northeast-southwest trend, the sea breeze is blowing from the east-south, and the land breeze is blowing from the west-north-northeast. According to observed wind direction, a natural day could be divided into four periods. Specifically, 01:00–08:00 LT (land breeze period), 13:00–20:00 LT (sea breeze period), and 09:00–12:00 LT and 21:00–24:00 LT (mixed sea-land breeze period). A day was classified as an SLB day if the following three conditions were met: (1) the 24 h mean wind speed was <10 m s⁻¹; (2) during the land breeze period, the land breeze persisted for ≥ 4 h and the sea breeze lasted ≤ 2 h; and (3) during the sea breeze period, sea breeze persisted for ≥ 4 h and the land breeze lasted ≤ 2 h. As shown in Fig. S1 (light blue areas), a total of 23 SLB days were identified during the observation period. The remaining days were classified as non-SLB days. Figure S2 illustrates that the wind patterns on SLB days differed from those on non-SLB days, characterized by a higher prevalence of southeasterly sea breezes.

Due to the absence of agricultural fields near the sampling site and its proximity to major traffic arteries, the primary emission of HONO was attributed to vehicles. Only nighttime data were used in our study, the uncertainties associated with photochemical process, such as NO3- photolysis were excluded. Therefore, the nighttime corrected HONO (HONO_corr) is calculated as follows: 1HONOcorr,t=HONOt-Evehicle,t-PNO+OH,t where HONO_t is the observed HONO concentration, Evehicle,t is the HONO emission from vehicles, and PNO+OH,t is the HONO production from homogeneous reactions. Evehicle,t is calculated as the NOx concentration multiplied by the emission factor 0.005 (Zhang et al., 2025b). PNO+OH,t is calculated as follows: 2PNO+OH,t=kNO+OH×[NO]×[OH]×3600s where kNO+OH (cm³ molec.⁻¹ s⁻¹) is calculated in the Master Chemical Mechanism 3.3.1 (MCM v3.3.1 https://mcm.york.ac.uk/MCM/, last access: 4 September 2026), 3600 s represents the 1 h integration interval, and the concentration of OH radicals is estimated based on the steady-state balance of H₂SO₄ (Nie et al., 2022). Because H₂SO₄ is primarily produced by the reaction of SO₂ and OH and is mainly lost via condensation sink (CS) onto particle surfaces, the OH concentration can be estimated as below: 3[OH]=[H2SO4]×CSkOH+SO2×[SO2] where kOH+SO2 is 5 × 10⁻¹³ cm³ s⁻¹ (Seinfeld et al., 1998), and CS (s⁻¹) is calculated as follows (Kulmala et al., 2012): 4CS=2πDv∫0dp,maxdpβn(dp)ddp where Dv is the diffusion coefficient of H₂SO₄, dp is the particle diameter derived from the measured PNSD, β is the transitional correction factor, and n(dp) is the particle number concentration of diameter dp.

In line with previous studies (Zhang et al., 2025b, 2020, 2024), the heterogeneous conversion of NO₂ to form HONO_corr at night is considered a pseudo-first-order reaction. The reaction rate constant (kNO2, s⁻¹) is calculated based on the conversion of NO₂ during a given period (from t1 to t2) as follows: 5kNO2=HONOcorr,t2-HONOcorr,t1NO2‾(t2-t1) where NO2‾ is the mean concentration of NO₂ between t1 and t2.

To evaluate the sensitivity of the results to the assumed emission factor, the calculation of HONO_corr and kNO2 (Fig. S3) were calculated using emission factors of 0.005, 0.008 and 0.012 (Xu et al., 2015), while the observational inputs, analysis periods and calculation procedures held constant.

Random Forest (RF) algorithm, a machine learning technique extensively applied in atmospheric (Yang et al., 2024; Lin et al., 2026b; Zhang et al., 2025b), was employed to elucidate the primary drivers of kNO₂. The model was built using the “scikit-learn” library (https://github.com/scikit-learn/scikit-learn/blob/fe2edb3cd/sklearn/ensemble/, last access: 10 June 2026) in a python environment. Drawing upon previously identified factors influencing kNO₂ (e.g., pH, RH, NH₃) (Liu and Abbatt, 2021; Gen et al., 2024; Zhang et al., 2025a, 2023) alongside the specific characteristics of the coastal urban environment, we incorporated meteorological parameters (RH and T), atmospheric reactive species (proxy of effective particle-phase chloride (Cl_total), NH₃, and NO₂), and heterogeneous-related parameters (pH, SA, and ground surfaces areas) as input variables for the RF model. Cl_total was defined as the sum of particulate Cl⁻ and gaseous HCl. It was used as a depletion-corrected proxy for the particulate Cl⁻ burden before acid displacement. During NO₂ uptake, the generated H⁺ can protonate particulate Cl⁻ and release HCl to the gas phase. Consequently, measured particulate Cl⁻ represents the residual fraction after depletion, whereas HCl retains information about part of the chloride transferred from the particle phase. Their sum therefore represents, to some extent, the chloride burden before depletion. To evaluate the suitability of Cl_total as a proxy for the depletion-corrected Cl⁻ burden, two additional RF models were constructed by replacing Cl_total with either Cl⁻ or gaseous HCl. Variance inflation factors (VIFs) were calculated for the predictor set of each RF model. A VIF below 10 was used to indicate the absence of severe multicollinearity. Only nighttime data were used to train the model to eliminate the effects of photochemical processes on kNO₂ quantification, thereby enabling a more precise identification of key drivers. Detailed RF hyperparameter configurations and model evaluation are provided in Sect. S2. In this study, the dominant factors were identified based on RF factor importance, which quantifies the percentage contribution of each variable to kNO₂ (Breiman, 2001; Gregorutti et al., 2017). Furthermore, one-dimensional partial dependence plots (PDPs) were employed to delineate the main effects of individual variables, while bivariate PDPs were utilized to visualize and access the interaction effects between pairs of factors on kNO₂ (Goldstein et al., 2015).

The multiphase chemical box model was established based on the Framework for 0-D Atmospheric Modeling (F0AM) (Wolfe et al., 2016). As shown in Tables S1 and S2, comprehensive HONO and NO3- budgets were incorporated into the F0AM model to simulate the variations of HONO and NO3-. The ISORROPIA II model (Seinfeld et al., 1998) was integrated into the model to determine aerosol pH, aerosol liquid water content and the partitioning ratio of HNO₃ and NO3-. Six simulation scenarios were conducted: simulations without heterogeneous NO₂ uptake for (1) SLB and (2) non-SLB days; simulations with NO₂ uptake for (3) SLB and (4) non-SLB days, using the mean kNO₂ derived from SLB days for the SLB simulation and that derived from non-SLB days for the non-SLB simulation, respectively; and two additional sensitivity simulations for SLB days using the (5) 25th and (6) 75th percentiles of the kNO₂ distribution derived from SLB days. The ability of the model to reproduce real atmospheric processes was evaluated by comparing the simulated HONO and NO3- with observations. Due to the underestimation of total particulate nitrate concentrations measured by instruments, the performance of NO3- simulations was evaluated by comparing the trends of the simulated NO3- production rate with the observed NO3- concentration. In contrast, as HONO predominantly exists in the gas phase, the simulated HONO was evaluated directly against the observed HONO concentrations.

As for the model simulation, each simulation was preceded by a 3 d spin-up to allow intermediate species concentrations to stabilize. The mean diurnal pattern of trace gas concentrations, meteorological parameters, and the results from the ISORROPIA II model during SLB days and non-SLB days were used as input data to constrain the box model. Notably, the concentrations of HONO and NO3- were not constrained by observed values. More details about the model incorporating mechanisms, modules, and simulation settings are provided in Sect. S3 and our previous study (Lin et al., 2026a).

Figure S1 shows the time series of the observed parameters. Based on the identification criteria detailed in the Methods section, 23 d were classified as SLB days, accounting for 48 % of entire observation period. The diurnal patterns of major observed parameters are summarized in Fig. S4, revealing that the primary differences between the two types of days predominantly emerged at night. Compared to non-SLB days, SLB days exhibited rapid nighttime accumulation of HONO and NO3-, accompanied by elevated chlorine levels, NOx concentrations, and aerosol surface area (SA) concentrations. Additionally, consistent with our previous study, SLB days were characterized by higher temperature and RH at night (Chen et al., 2022). Figure 1 and Table S3 further summarizes the nighttime observational results on SLB and non-SLB days. By definition, SLB days are influenced by persistent sea breezes, which likely enhances the impact of sea-salt aerosols. As shown in Fig. 1a, the Cl⁻ concentrations were comparatively higher on SLB days (0.154 µg m⁻³) than on non-SLB days (0.139 µg m⁻³). Specifically, this overall increase (Fig. S5) was primarily driven by elevated levels during the sea breeze (18:00–20:00 LT) and mixed sea-land breeze (21:00–00:00 LT) periods, while concentrations declined during the land breeze (01:00–06:00 LT) period. The elevation pattern of Cl⁻ indicates that the more abundant Cl⁻ transported by stronger sea breezes on SLB days can persist into the subsequent mixed sea-land breeze periods. Additionally, Fig. S6 shows a moderate positive correlation between Na⁺ and Cl⁻ on SLB days (r=0.67), compared with a weaker relationship on non-SLB days (r=0.28). Because sodium chloride (NaCl) is a major component of sea salt, this contrast is consistent with a greater sea-salt influence on Cl⁻ on SLB days. As for gaseous chlorine, Fig. 1b shows that the concentration of HCl was significantly higher on SLB days (96.2 ppt) than on non-SLB days (67.5 ppt). This further confirms that, under the influence of sea-salt aerosols on SLB days, chlorine levels are distinctly elevated in the nighttime coastal troposphere.

Figure 1c indicates that the concentration of NO₂ on SLB days (mean: 28.3 ppb) was significantly higher than that on non-SLB days (23.2 ppb). Within our study area, ship emissions, cross-sea bridge traffic, and road traffic contribute substantial NOx, and the enhanced local circulation on SLB days can promote NO₂ accumulation (Chen et al., 2023). The mean concentrations of HONO and NO3- on SLB day were 0.836 ppb and 3.61 µg m⁻³, respectively, which were significantly higher than the corresponding values of 0.497 ppb and 3.13 µg m⁻³ on non-SLB days (Fig. 1d, e). To better understand the elevated HONO levels, HONO_corr concentrations, defined as the observed HONO concentration subtracts contributions from primary emissions and gas-phase production, were calculated to represent the nocturnal HONO produced from heterogeneous conversion. The ratio of HONO_corr / NO₂ was notably higher on SLB days (0.030) than on non-SLB days (0.021), indicating more active heterogeneous conversion of NO₂to HONO on SLB days. Multiple studies have confirmed that NO₂ uptake is a primary source of nocturnal heterogeneous HONO formation (Zhang et al., 2023, 2022). Here, the higher SA concentrations on SLB days were also conducive to heterogeneous reactions (Fig. 1g). Furthermore, Fig. 2a–b indicate a moderate positive correlation between HONO_corr and NO3-, with the correlation coefficient (r) increasing from 0.56 on non-SLB days to 0.76 on SLB days. The HONO_corr concentrations under the two conditions overlapped mainly below 1.5 ppb. Within this overlapping range, the correlation on SLB days remained significant (r=0.63, p<0.001, Fig. S7a) but was close to that on non-SLB days (r=0.56), indicating that the wider concentration range partly contributed to the higher full-range correlation on SLB days, while the positive relationship between HONO_corr and NO3- persisted within the overlapping range. As shown in Reaction (R1), NO₂ uptake via disproportionation simultaneously produces HONO and HNO₃, the latter of which is further converted to NO3- via gas-particle partitioning. Thus, nocturnal NO₂ uptake is also likely an important contributor to the formations of NO3- on SLB days. Figure 2c, d show that the full-range correlation between HONO_corr and gaseous HCl was stronger on SLB days than on non-SLB days (r=0.58 versus 0.16). Within the overlapping HONO_corr range below 1.5 ppb, the correlation on SLB days decreased to (r=0.47) but remained significant (p<0.001, Fig. S7b) and appreciably higher than that on non-SLB days. This result indicates that the stronger HCl-HONO_corr relationship during SLB periods cannot be explained solely by the wider HONO_corr range. The H⁺ generated by NO₂ uptake can combine with Cl⁻ (Reactions R2–R3). Simultaneously, the aerosol pH in the coastal region of southeastern China was relatively low (approximately 1.92 ± 0.34 during nocturnal SLB periods), chloride depletion could easily occur, leading to the observed positive correlation between HONO_corr and HCl. The enhanced chloride depletion can also explain our observations of a significant increase in gaseous HCl but not in Cl⁻concentrations on SLB days. Similarly, the rapid decline in Cl⁻ concentration during land breeze periods (Fig. S4) could also be explained by chloride depletion. R12NO2+H2O(ad)+aerosol/ground surface→HNO3+HONOR2NaCl(s)→Na+(aq)+Cl-(aq)R3H+(aq)+Cl-(aq)→HCl(aq)⇋HCl(g)

The values of kNO₂ on SLB days and non-SLB days were derived as Fig. S8. Overall, the average kNO₂ values on SLB days (9.70 × 10⁻⁶ s⁻¹) were significantly higher than those on non-SLB days (6.72 × 10⁻⁶ s⁻¹), by a factor of approximately 1.44. Across the tested emission factors, the mean HONO_corr decreased from 0.58 to 0.51 ppb on SLB days, whereas kNO₂ varied by no more than 6.2 %, indicating limited sensitivity of the derived rate constant to the assumed emission factor (Fig. S3). As presented in Table S4, a prolonged increase in the HONO_corr / NO₂ ratio was also observed during 17 of the 23 SLB days (see an example in Fig. S9). In contrast, this phenomenon was virtually absent on non-SLB days, indicating a sustained nocturnal enhancement of kNO₂ during SLB periods. Compared with existing field observations, the kNO₂ on SLB days was broadly consistent with the levels reported under the sea-case scenario in coastal cities of China such as Qingdao (1.25 × 10⁻⁵ s⁻¹) and Hong Kong (8.81 × 10⁻⁶ s⁻¹) (Yang et al., 2021; Zha et al., 2014), while it was markedly higher than those values observed in major non-coastal cities including Beijing (7.78 × 10⁻⁷–5.75 × 10⁻⁶ s⁻¹) (Wang et al., 2017b; Zhang et al., 2020, 2022; Xuan et al., 2023; Jia et al., 2020), Nanjing (1.19 × 10⁻⁶ – 2.22 × 10⁻⁶ s⁻¹) (Zheng et al., 2020; Liu et al., 2019), and Guangzhou (4.44 × 10⁻⁶–6.67 × 10⁻⁶ s⁻¹) (Su et al., 2008; Li et al., 2012; Qin et al., 2009). The consistently higher kNO₂ in coastal areas implies that coastal atmospheric environments are likely more conducive to NO₂ uptake.

A machine learning technique was used to further identify key drivers of the enhanced kNO₂ on SLB days. The RF models using Cl_total reproduced the calculated kNO₂ well during both SLB and non-SLB periods, with R2 values of 0.91 and 0.87, respectively (Fig. S10c, d). Replacing Cl_total with either Cl⁻ or HCl resulted in the respective chlorine variable ranking third after RH and SA (Figs. S10–S11). Particulate Cl⁻ represents the residual particle-phase fraction, whereas HCl contains information on the fraction transferred to the gas phase through chloride depletion. Their respective importance therefore suggests that the Cl_total signal was not determined by either component alone, but reflected their complementary information on the Cl⁻ burden before depletion. All VIFs were below 10, indicating no severe multicollinearity between the chlorine metrics and the other predictors (Table S5). Because all three metrics retain part of the chlorine-related information, the RF models showed similar R2 values (0.89–0.91); nevertheless, the Cl_total-based model had the highest R2 and provided a more physically representative description of the association between depletion-corrected particulate chloride and kNO₂. Notably, the role of Cl_total and RH diminished substantially on non-SLB days (Fig. 3b), underscoring the profound modulation of these two influencing factors by the SLB circulation. Figure 3c–e further elucidate the specific impacts of these primary drivers on nocturnal kNO₂ under SLB conditions. Specifically, Cl_total exhibits a positive correlation with kNO₂ (Fig. 3c). This phenomenon is likely attributable to the preferential partitioning of chloride ions at the gas-aqueous interface of aerosols (Zhang et al., 2025a; Shen et al., 2025; Wang et al., 2025a). The surface-bound Cl⁻ can attract gaseous NO₂ molecules (Reaction R4) to form interfacial intermediates [Cl-NO₂]⁻, thereby facilitating the overall NO₂ uptake process. Subsequent interaction analysis was consistent with the mechanisms proposed in laboratory studies (Reactions R4–R5) and suggest that similar processes may operate in the coastal boundary layer. Figure 3f shows that at a fixed NO₂ concentration, elevated chloride levels amplify the positive effect of NO₂ molecules on kNO₂. This indicates that chloride ions actively promote the heterogeneous uptake of NO₂, consequently accelerating the kNO₂. Similar halogen-promoted effect has also been revealed in NO₂-initiated aqueous-phase oxidation of dissolved SO₂ at the air-aqueous interface and of nitrate photolysis at the aqueous-phase surface (Wang et al., 2025a; Shen et al., 2025). Figure 3d shows a significant positive relationship exists between RH and kNO₂, suggesting that moist sea breezes effectively supply the water molecules essential for the heterogeneous NO₂ uptake. This finding aligns with previously reported field observations highlighting RH-driven NO₂ uptake (Xuan et al., 2024). Analogous to Cl_total, the interaction between RH and NO₂ reveals that at a constant NO₂ level, the positive impact of NO₂ on kNO₂ is enhanced under elevated RH. It implies that RH could also facilitate NO₂ uptake by promoting the interaction between NO₂ molecules and heterogeneous interface. Furthermore, Fig. 3h shows a threshold-dependent joint enhancement: kNO₂ increased when RH exceeded approximately 80 % and Cl_total exceeded approximately 0.5 µg m⁻³, whereas the response was limited when either variable remained below these levels. Elevated RH can increase aerosol liquid water and maintain an aqueous particle phase, while Cl⁻ in wetted chloride-containing particles can be preferentially accessible at the air-aqueous interface and enhance NO₂ adsorption and uptake (Zhang et al., 2025a; Wang et al., 2025a). Therefore, the bivariate PDPs response is consistent with high RH providing a favorable aqueous interfacial environment for chloride-associated NO₂ uptake. This interaction may partly explain the higher importance of RH and Cl_total on SLB days. As a primary factor on both SLB and non-SLB days, SA positively correlates with kNO₂ (Fig. 3e). This indicates that an increase in aerosol surface area concentrations provides effective reaction interfaces for NO₂ uptake process, consistent with most atmospheric heterogeneous processes (Wang et al., 2017a; Ma et al., 2026). Moreover, Fig. S12 shows that on SLB days, concentrations of the major NO₂ uptake products (HONO and NO3-) increased with those key factors including Cl_total, RH, and SA, providing additional support for the above driving factor analysis. In general, the reasons underlying the enhanced kNO₂ on SLB days help to explain the multi-fold higher kNO₂ observed in coastal areas compared to non-coastal urban areas, highlighting ambient chlorine levels and RH as non-negligible contributors. R4Cl-(aq)+NO2→Cl-NO2-surfaceR5Cl-NO2-surface+NO2+H2O(ad)→Cl-(aq)+HONO+NO3-+H+

The above correlations analysis suggests that NO₂ uptake could play an important role in the production of HONO and NO3- on SLB days. Since the HONO and NO3- formation were influenced by multiple meteorological factors and many atmospheric reaction pathways, a multiphase chemical box model was employed to evaluate the role of enhanced NO₂ uptake. As described in Methods, six simulation scenarios were set up. Incorporating NO₂ uptake mechanism substantially improved the simulated diurnal HONO profile during SLB periods: r increased from 0.57 to 0.94, while mean absolute error (MAE) and root mean square error (RMSE) decreased from 0.58 to 0.11 ppb and from 0.64 to 0.13 ppb, respectively (Fig. 4a). For NO3-, incorporating NO₂ uptake increased the mean simulated net NO3- formation rate by 2.14 µg m⁻³ h⁻¹ and strengthened the temporal consistency between the simulated net formation rate and observed concentration, increasing r from 0.04 (p>0.05) to 0.64 (p<0.001, Fig. 4b). This improvement indicates that the added mechanism better represented the observed temporal variability in NO3- during SLB periods. Additionally, the NO₂ uptake could robustly explain the enhancements in both atmospheric reactive nitrogen species on SLB days across various kNO₂ quantiles (Fig. S13). In contrast, including NO₂ uptake under non-SLB conditions produced only limited improvements in the simulated HONO concentrations and the temporal covariation between NO3- production rates and observed NO3- concentrations (Fig. S14). This suggests that NO₂ uptake had less explanatory importance under non-SLB conditions than during SLB periods, consistent with the stronger observed HONO_corr-NO3- relationship on SLB days. Moreover, unlike on SLB days (Fig. S9), HONO_corr / NO₂ did not exhibit a sustained nocturnal increase on non-SLB days, which resulted in numerous negative estimates of kNO₂ estimated using Eq. (5). Excluding these negative values likely led to an overestimated mean kNO₂ and, consequently, to the overprediction of HONO and NO3- under non-SLB conditions.

Figure 4c, d illustrate the simulated HONO and NO3- budgets for the scenarios incorporating NO₂ uptake on SLB days. Despite potential overestimations of non-SLB kNO₂, the mean NO₂ uptake rate remained significantly higher on SLB days than on non-SLB days, with an average rate of 1.03 ppb h⁻¹ (Fig. S15). During nighttime, NO₂ uptake accounted for 83.8 % of HONO formation (Fig. 4e). The predominance of NO₂ uptake in nocturnal HONO formation has also been observed during PM pollution episodes in urban areas (Xuan et al., 2024; Jia et al., 2020). However, unlike the chlorine-driven reason identified here, those prior cases were primarily attributed to favorable heterogeneous reaction conditions, such as larger SA concentrations, lower BLH, and higher RH. In addition to NO₂ uptake, primary emissions from vehicle exhaust, coal combustion, and biomass burning are also recognized as major contributors to nighttime HONO in urban areas (Wang et al., 2025b; Zhang et al., 2022). In this study, primary emissions contributed only 15.5 % to nocturnal HONO formation, substantially less than the contribution from NO₂ uptake. As for nighttime NO3- formation, many studies had indicated that N₂O₅ uptake was the dominant pathway in urban areas (Chen et al., 2020; Wang et al., 2018). However, the situation changed under the impact of the enhanced NO₂ uptake. Figure 4f shows that NO₂ uptake contributed 46.4 % to nocturnal NO3- formation, which was almost comparable to that of N₂O₅ uptake (51.6 %). These values were obtained using the BT09 parameterization of γN₂O₅ (Bertram and Thornton, 2009b). Additional simulations using the EJ05 and GRI09 schemes yielded contributions of 47.0 %–47.9 % for NO₂ uptake and 50.0 %–50.9 % for N₂O₅ uptake (Fig. S16 and Table S6), indicating that their comparable importance was insensitive to the choice of γN₂O₅ parameterization (Evans and Jacob, 2005; Hallquist et al., 2003; Griffiths et al., 2009). While chlorine can also promote the uptake of N₂O₅ (Bertram and Thornton, 2009a), the relative enhancement in the contribution of NO₂ uptake to nocturnal NO3- formation was more pronounced. This might be related to the suppression of N₂O₅ uptake caused by increased NO3- generating directly from NO₂ uptake. The increase in NO3- component of aerosol can shift the equilibrium between N₂O₅ and NO3- toward N₂O₅, thereby reducing the contribution of N₂O₅ to NO3- formation (Bertram and Thornton, 2009a; Wagner et al., 2013). Additionally, the enhanced NO₂ uptake process also competed for NO₂, an important precursor of N₂O₅ formation, especially in the NO₂-limited regime for N₂O₅ formation (Lin et al., 2025), ultimately reducing the overall enhancement effect on N₂O₅ generation. Moreover, this comparable contribution of NO₂ and N₂O₅ uptake NO3- formation was also found during fog events (Xu et al., 2024). Although the observed phenomenon was similar, the dominant drivers differed between fog conditions and coastal SLB periods. The former was primarily attributed to favorable uptake conditions such as abundant aerosol water content and high surface area concentration, whereas the latter was more due to the enhancement effect of chlorine levels and RH on the NO₂ uptake process. Overall, these results reveal the high contribution of NO₂ uptake to HONO and NO3- formation during SLB circulation periods, demonstrating that enhanced NO₂ uptake can significantly influence the atmospheric reactive nitrogen cycle in coastal urban areas.