Measurement report: Atmospheric nitrate radical chemistry in the South China Sea influenced by the urban outflow of the Pearl River Delta

The nitrate radical (NO₃) is a critical nocturnal atmospheric oxidant in the troposphere, which widely affects the fate of air pollutants and regulates air quality. Many previous works have reported the chemistry of NO₃ in inland regions of China, while fewer studies target marine regions. Here, we present a field measurement of the NO₃ reservoir, dinitrogen pentoxide (N₂O₅), and related species at a typical marine site (Da Wan Shan Island) located in the South China Sea in the winter of 2021. Two patterns of air masses were captured during the campaign, including the dominant airmass from inland China (IAM) with a percentage of ∼ 84 %, and the airmass from eastern coastal areas (CAM) with ∼ 16 %. During the IAM period, the NO₃ production rate reached 1.6 ± 0.9 ppbv h⁻¹ due to the transportation of the polluted urban plume with high NO_x and O₃. The average nocturnal N₂O₅ and the calculated NO₃ mixing ratios were 119.5 ± 128.6 and 9.9 ± 12.5 pptv, respectively, and the steady-state lifetime of NO₃ was 0.5 ± 0.7 min on average, indicating intensive nighttime chemistry and rapid NO₃ loss at this site. By examining the reaction of NO₃ with volatile organic compounds (VOCs) and N₂O₅ heterogeneous hydrolysis, we revealed that these two reaction pathways were not responsible for the NO₃ loss (< 20 %) since the NO₃ reactivity (k(NO₃)) towards VOCs was small ($\mathrm{5.2}\times {\mathrm{10}}^{-\mathrm{3}}$ s⁻¹) and the aerosol loading was low. Instead, NO was proposed to significantly contribute to nocturnal NO₃ loss at this site, despite the nocturnal NO concentration always below the parts per billion by volume level and near the instrument detection limit. This might be from the local soil emission or something else. We infer that the nocturnal chemical NO₃ reactions would be largely enhanced once without NO emission in the open ocean after the air mass passes through this site, thus highlighting the strong influences of the urban outflow to the downwind marine areas in terms of nighttime chemistry. During the CAM period, nocturnal ozone was higher, while NO_x was much lower. The NO₃ production was still very fast, with a rate of 1.2 ppbv h⁻¹. With the absence of N₂O₅ measurement in this period, the NO₃ reactivity towards VOCs and N₂O₅ uptake were calculated to assess NO₃ loss processes. We showed that the average k(NO₃) from VOCs (56.5 %, 2.6 ± 0.9 × 10⁻³ s⁻¹) was higher than that from N₂O₅ uptake (43.5 %, 2.0 ± 1.5 × 10⁻³ s⁻¹) during the CAM period, indicating a longer NO₃ $/$ N₂O₅ lifetime than that during IAM period. This study improves the understanding of the nocturnal NO₃ budget and environmental impacts with the interaction of anthropogenic and natural activities in marine regions.

1 Introduction

Reactive nitrogen compounds, especially the nitrate radical (NO₃) and dinitrogen pentoxide (N₂O₅) play an essential role in nocturnal atmospheric chemistry (Wayne et al., 1991; Brown and Stutz, 2012). NO₃ is mainly generated via the oxidation of NO₂ by O₃ (Reaction R1), and then NO₃ further reacts with NO₂ to produce N₂O₅ (Reaction R2) with a thermal equilibrium. The temperature-dependent equilibrium constant, K_eq, regulates the equilibrium, favoring NO₃ and NO₂ at higher temperatures (Osthoff et al., 2007; Chen et al., 2022). During daytime, the NO₃ mixing ratio is generally low as its lifetime is very short (< 5 s) due to the fast photolysis (Reactions R3 and R4) and rapid reaction with NO (Reaction R5) (a rate constant of 2.6 × 10⁻¹¹ cm³ molec.⁻¹ s⁻¹ at 298 K, Atkinson et al., 2004). At night, NO₃ accumulates and can reach tens to hundreds of parts per trillion by volume (pptv), making it the major nocturnal oxidizing agent (Wang et al., 2015).

$$\begin{array}{}\text{(R1)}& {\displaystyle }{\mathrm{NO}}_{\mathrm{2}}+{\mathrm{O}}_{\mathrm{3}}\to {\mathrm{NO}}_{\mathrm{3}}+{\mathrm{O}}_{\mathrm{2}}\text{(R2)}& {\displaystyle }{\mathrm{NO}}_{\mathrm{2}}+{\mathrm{NO}}_{\mathrm{3}}\rightleftharpoons {\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{5}}\text{(R3)}& {\displaystyle }{\mathrm{NO}}_{\mathrm{3}}+hv\to {\mathrm{NO}}_{\mathrm{2}}+\mathrm{O}\text{(R4)}& {\displaystyle }{\mathrm{NO}}_{\mathrm{3}}+hv\to \mathrm{NO}+{\mathrm{O}}_{\mathrm{2}}\text{(R5)}& {\displaystyle }\mathrm{NO}+{\mathrm{NO}}_{\mathrm{3}}\to \mathrm{2}{\mathrm{NO}}_{\mathrm{2}}\end{array}$$

During nighttime, NO₃ is the most important oxidant for alkenes (Mogensen et al., 2015; Edwards et al., 2017), particularly in rural, remote, or forested environments, where it predominantly reacts with unsaturated biogenic volatile organic compounds (VOCs), especially isoprene and monoterpene (Ng et al., 2017; Liebmann et al., 2018b, a), to form alkyl nitrates (RONO₂) that ultimately lead to secondary organic aerosols (SOAs) (Brown and Stutz, 2012). The observations and model simulations showed that the measured particulate organic nitrates were largely attributed to the nocturnal NO₃ oxidation across Europe (Kiendler-Scharr et al., 2016). The NO₃ oxidation was also reported to play an important role in aerosol formation in the southeastern United States, with high isoprene and monoterpene emissions (Xu et al., 2015). These studies highlighted the critical role of the reaction of NO₃ with VOCs in NO₃ budget and organic aerosol pollution. In addition, NO₃ also reacts with dimethyl sulfide (DMS) over the ocean, affecting the marine sulfur cycle and thus cloud formation and global climate (Aldener et al., 2006; Brown and Stutz, 2012; Barnes et al., 2006; Rosati et al., 2022). In high aerosol loading regimes, the N₂O₅ heterogeneous uptake becomes a significant indirect NO₃ loss pathway. The hydrolysis reaction produces nitrate (NO${}_{\mathrm{3}}^{-}$) and nitryl chloride (ClNO₂) on chloride-containing aerosol surfaces (Osthoff et al., 2008; Thornton et al., 2010), in which ClNO₂ activates the Cl radical and further enhances the photochemistry and ozone pollution in the following day (Riedel et al., 2012, 2014; Behnke et al., 1993).

Different NO₃ loss pathways produce different air pollutants, thus characterization of NO₃ budget is essential to clarifying the NO₃ chemistry in air pollution under various environments. Observations of N₂O₅ and NO₃ in different regions and evaluation of their loss processes have been reported in numerous studies (Crowley et al., 2011; Geyer et al., 2001; Brown et al., 2011; Dewald et al., 2022; Niu et al., 2022; Brown et al., 2016; Wang et al., 2020a; Tham et al., 2016; Aldener et al., 2006; Lin et al., 2022). In general, the NO₃ loss process shows significant regional differences. In urban areas featuring intensive anthropogenic NO_x emissions and moderate (or high) aerosol loading, N₂O₅ uptake is comparable or even dominates the NO₃ loss (Wang et al., 2013). While in rural and forested areas with abundant biogenic VOC (BVOC) emissions, the NO₃ loss processes were usually dominated by BVOCs (Dewald et al., 2022; Geyer et al., 2001; Brown et al., 2011). As for the coastal areas, which were jointly affected by the polluted air mass from inland and the relatively clean air mass from the ocean, the dominant NO₃ loss process varies greatly depending on the air mass origin (Aldener et al., 2006; Niu et al., 2022; Brown et al., 2016; Crowley et al., 2011). For instance, Crowley et al. (2011) found on the Atlantic coast of southern Spain (forested area) that when the air mass mainly originated from the Atlantic, NO₃ was mainly consumed by BVOCs (mainly monoterpenes) emitted from nearby forests, while when the air mass came from the continent, NO₃ loss was mainly due to reactions with anthropogenic VOCs (AVOCs).

China has been recently proven to be a hot spot of nocturnal chemistry with a high NO₃ production rate (H. Wang et al., 2023). Many studies have reported the mechanisms, budget, or impacts of NO₃–N₂O₅ chemistry in different regions, where most of them were conducted in urban regions (Wang et al., 2013; Yan et al., 2021; H. Wang et al., 2020; X. F. Wang et al., 2017; Z. Wang et al., 2017). For example, H. Wang et al. (2017b) showed a significant contribution of N₂O₅ uptake to nitrate pollution in summer and winter, and they also highlighted the fast organic nitrate production rate observed in Beijing rural region in summer (Wang et al., 2018b). Only some studies have focused on nighttime oxidation in coastal cities like Shanghai, Shenzhen, and Hong Kong (Zhu et al., 2022; Niu et al., 2022; Yan et al., 2019), which have shown different patterns of NO₃ chemistry compared with urban regions. Even fewer field studies have been conducted on islands that are far away from coastal cities where the interactions of the oceanic atmosphere and urban plumes can significantly affect the NO₃ budget and impacts. Given the diversity of air masses in inland and coastal areas, studies are needed to gain a comprehensive understanding of NO₃ losses in different atmospheric environments, particularly in coastal and marine areas.

Therefore, we conducted an intensive field observation on Da Wan Shan Island (DWS) in the winter of 2021, which is a typical island site in the north of the South China Sea, and downwind of the city clusters in the Pearl River Delta (PRD), China, during the winter monsoon periods. The island features a subtropical oceanic monsoon climate, and the north and northeast synoptic winds from inland PRD and eastern China coast are generally predominant in winter (Liu et al., 2019; Wang et al., 2018a). This allows us to further investigate the interactions between anthropogenic emissions and marine emissions from the perspective of nighttime chemistry. In this study, the measurements of N₂O₅ and the related species during the DWS winter campaign are reported. We have identified two types of air masses from both mainland China and coastal areas. Finally, the NO₃ budget and loss processes in different air masses are characterized.

2 Methods

2.1 Site description

The field campaign was conducted at Da Wan Shan Island (21^∘55^′57^′′ N, 113^∘43^′15^′′ E) from 9 November to 16 December 2021. Figure 1 shows the location of the study site, which is approximately 60 km southwest of Hong Kong, 40 km southeast of Zhuhai, and about 100 and 80 km away from the megacities Guangzhou and Shenzhen, respectively. This island is dominated by mountainous terrain with an area of 8.1 km² and has a small population of about 3000. Anthropogenic emissions are sparse and no industrial pollution sources were identified, though numerous ships engaging in local fishing activities were observed, potentially affecting the local atmosphere. During the measurement, local airflow was consistently from either the northwest or southeast (Fig. 1a) due to the winter monsoon, with wind speeds most frequently ranging from 1.8 to 7.9 m s⁻¹ (10th–90th percentiles) and an average of 4.5 ± 2.6 m s⁻¹. This wind direction is indicative of the mixing of air masses from both continental and coastal areas. The Hybrid Single-Particle Lagrangian Integrated Trajectory model (HYSPLIT) was adopted to investigate the historical trajectory. The HYSPLIT model was run for 48 h backward in time at 20:00, 24:00, and 04:00 (all times given are CNST, Chinese National Standard Time, UTC + 8), and at a height of 70 m above sea level. It confirmed that the airmass during nighttime mostly came from inland China (fresh urban emissions, IAM, 84 %) and the coastal areas (aged urban emissions, CAM, 16 %). IAM featured the outflow from inland China, such as Guangzhou and Changsha, while CAM featured the outflow of coastal cities like Hong Kong and Shenzhen. No air masses free of pollution from the South China Sea were observed during the measurement period. All measurement instruments were placed in the DWS Atmospheric-Marine Research Station, located on the rooftop with inlets approximately 10 m above ground level and about 72 m above sea level, with sunrise around 06:40 and sunset at 17:40.

Figure 1 A map of the field measurement site of Wanshan Island (red star) and the surrounding environment (© Microsoft). Two coastal sites, Hok Tsui (Yan et al., 2019) and Yangmeikeng (Niu et al., 2022), and an urban site, Heshan (Wang et al., 2022; Yun et al., 2018b), are denoted with the blue circles and diamond, respectively. The blue triangle denotes Mt. Tai Mo Shan (957 m), a mountainous site that recorded the nighttime chemistry in the nocturnal residual layer (Brown et al., 2016). The inset  (a) shows the wind rose for the sampling site during the campaign. Panel (b) shows the clustering results of the 48 h backward trajectory calculations at nighttime using the HYSPLIT model throughout the campaign.

2.2 Instrument setup

Various parameters were measured in this study, including N₂O₅, NO, NO₂, O₃, VOCs, particle number size distribution (PNSD), and meteorological parameters with different instruments. Detailed information about these instruments is listed in Table 1. The N₂O₅ measurements were performed using a cavity-enhanced absorption spectrometer (CEAS) which has been deployed over several field campaigns (H. Wang et al., 2017a, b; Wang et al., 2018b; Wang et al., 2020b). In brief, ambient N₂O₅ was thermally decomposed to NO₃ in a perfluoroalkoxy alkane (PFA) tube (length: 35 cm, i.d.: 4.35 mm) heated to 130 ^∘C, and NO₃ was detected within a 110 ^∘ PFA resonator cavity. NO was injected to destroy NO₃ from N₂O₅ thermal decomposition in a 5 min cycle, and the result was used as the reference spectrum to avoid the influence of ambient water vapor. A pair of high-reflectivity (HR) mirrors (Layertec GmbH, Mellingen, Germany) with a diameter of 25.0 mm (+0.00 / −0.10 mm) was used to enhance the effective optical pathlength. Mirror reflectivity (R(λ)) was calibrated with high-purity He and N₂ in the current experimental setup during the field measurements. R(λ) was calibrated to be 0.99997, and the effective pathlength of the optical resonator was 13.96 km. A Teflon polytetrafluoroethylene (PTFE) filter was used to remove ambient aerosol particles, and the inlet flow rate was 1.0 L min⁻¹. The loss of N₂O₅ in the sampling line and filter was considered in the data correction according to previous work (H. Wang et al., 2017a). Here the CEAS measurement encompasses the combined concentration of ambient [N₂O₅ + NO₃] and effectively represents N₂O₅ under high NO_x (or low temperature) conditions when the ratio of NO₃ to N₂O₅ is likely to be low. Accounting for the instrument's transmission efficiency and the thermal transformation between NO₃ and N₂O₅, the contribution of NO₃ is sufficiently negligible in comparison to N₂O₅. Nevertheless, we have taken it into account during the N₂O₅ data correction. The limit of detection (LOD) was 2.7 pptv (1σ), and the measurement uncertainty was ±19 %.

Table 1 The information of observation instruments used during the DWS campaign.

NO_x and O₃ were measured by commercial instruments (model T200U and model T400U, Teledyne API Inc., respectively) calibrated with zero air before the measurement. The nitrogen oxide analyzer uses the chemiluminescence detection method to measure the original NO and converted NO₂, and the LOD was 0.4 ppbv for each species. Aerosol surface area density (S_a, µm² cm⁻³) was calculated based on the particle numbers and geometric diameter, which were calculated from the results measured by a laboratory-assembled scanning-mobility particle sizer (SMPS) according to Mcmurry et al. (2000). This SMPS system consists of two differential mobility analyzers (DMAs, “nano-DMA” mode 3081A, and “regular-DMA” mode 3085A, TSI Inc.) in parallel and a condensed particle counter (mode 3787, TSI Inc.) as the detector. The combination of nano-DMA and conventional mode 3085A DMA enables the SMPS to have better detection performance for particles below 50 nm. In this measurement, SMPS measured the particle size distribution in 5–300 nm with a time resolution of 5 min, and S_a can be regarded as the lower limit value. A growth factor f(RH) = 1 + 8.8 × (RH$/$100)^9.7 (Liu et al., 2013) was used here to correct dry state S_a to wet state S_a.

VOCs were measured by proton transfer reaction time of flight mass spectrometry (PTR-TOF-MS, Ionicon Analytik GmbH, Innsbruck, Austria) with a time resolution of 10 s. At the end of this campaign, background measurements and instrument calibration were conducted with high-purity nitrogen and multi-component VOC gas standards, respectively. The instrument calibration results yielded strong linear relationships (R²=0.98) between the proton transfer reaction rate constants and the sensitivities of 10 calibrants: acetaldehyde, acetone, dimethyl sulfide, isoprene, methyl ethyl ketone, benzene, toluene, styrene, o-xylene, and trimethylbenzene. The sensitivities of the uncalibrated species were determined through the rate constants of the proton transfer reactions and their correlation coefficients with sensitivity. Meanwhile, the VOCs were also sampled by canister and analyzed by a gas chromatograph equipped with a mass spectrometer or flame ionization detector (GC-MS) for some ozone-polluted days. In the absence of nocturnal data from canister samples, the following analysis was based on the PTR-TOF-MS measurement, except the weight of α-pinene and β-pinene detected by GC-MS. Since monoterpene species cannot be distinguished by PTR-TOF-MS, the reaction rate constant of the sum monoterpene reaction with NO₃ was weighted by the campaign-averaged weight of α-pinene and β-pinene detected by GC-MS. Meteorological parameters (i.e., temperature (T), relative humidity (RH), wind speed, and wind direction) were routinely monitored with a time resolution of 5 min.

2.3 The calculation of NO₃ budget and lifetime

With the observation of N₂O₅, NO₃ can be calculated according to their temperature-dependent equilibrium relationship (Eq. 1) (Brown and Stutz, 2012). Lifetimes are commonly expressed as the ratio of their concentrations to the NO₃ production rate as determined by Eqs. (2) and (3), assuming the production and loss are in dynamic balance at night (Brown et al., 2003; Brown and Stutz, 2012). The production rate of nitrate radical, P(NO₃), is commonly expressed by Eq. (4), where $k_{{\mathrm{NO}}_{\mathrm{2}}+{\mathrm{O}}_{\mathrm{3}}}$ represents the temperature-dependent reaction rate constant of NO₂ and O₃ (Atkinson et al., 2004). In general, the nocturnal NO₃ losses typically include three main pathways (Eq. 5): (1) the reaction with NO, (2) the reactions with VOCs, and (3) N₂O₅ uptake.

$$\begin{array}{}\text{(1)}& {\displaystyle }\begin{array}{rl}& \left[{\mathrm{NO}}_{\mathrm{3}}\right]=\left[{\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{5}}\right]/K_{\mathrm{eq}}\left(\mathrm{T}\right)\left[{\mathrm{NO}}_{\mathrm{2}}\right]\\ & K_{\mathrm{eq}}=\mathrm{5.50}\times {\mathrm{10}}^{-\mathrm{27}}\times \exp (\mathrm{10724}/T)\end{array}\text{(2)}& {\displaystyle }{\mathit{\tau }}_{{\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{5}}}={\displaystyle \frac{\left[{\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{5}}\right]}{P\left({\mathrm{NO}}_{\mathrm{3}}\right)}}={\displaystyle \frac{\left[{\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{5}}\right]}{k_{{\mathrm{NO}}_{\mathrm{2}}+{\mathrm{O}}_{\mathrm{3}}}\left[{\mathrm{NO}}_{\mathrm{2}}\right]\left[{\mathrm{O}}_{\mathrm{3}}\right]}}\text{(3)}& {\displaystyle }{\mathit{\tau }}_{{\mathrm{NO}}_{\mathrm{3}}}={\displaystyle \frac{\left[{\mathrm{NO}}_{\mathrm{3}}\right]}{P\left({\mathrm{NO}}_{\mathrm{3}}\right)}}={\displaystyle \frac{\left[{\mathrm{NO}}_{\mathrm{3}}\right]}{k_{{\mathrm{NO}}_{\mathrm{2}}+{\mathrm{O}}_{\mathrm{3}}}\left[{\mathrm{NO}}_{\mathrm{2}}\right]\left[{\mathrm{O}}_{\mathrm{3}}\right]}}\text{(4)}& {\displaystyle }\mathrm{P}\left({\mathrm{NO}}_{\mathrm{3}}\right)=k_{{\mathrm{NO}}_{\mathrm{2}}+{\mathrm{O}}_{\mathrm{3}}}\left[{\mathrm{O}}_{\mathrm{3}}\right]\left[{\mathrm{NO}}_{\mathrm{2}}\right]\text{(5)}& {\displaystyle }\begin{array}{rl}\mathrm{L}\left({\mathrm{NO}}_{\mathrm{3}}\right)& =\sum k_i\left[{\mathrm{VOC}}_i\right]\left[{\mathrm{NO}}_{\mathrm{3}}\right]\\ & +k_{\mathrm{NO}+{\mathrm{NO}}_{\mathrm{3}}}\left[\mathrm{NO}\right]\left[{\mathrm{NO}}_{\mathrm{3}}\right]+k_{\mathrm{het}}\left[{\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{5}}\right]\end{array}\end{array}$$

The NO₃ reactivity towards VOCs, k(NO₃), is the first-order loss rate coefficient calculated from the products of the bimolecular rate coefficients k_i and the VOC concentrations, as shown in Eq. (6).

$$\begin{array}{}\text{(6)}& k\left({\mathrm{NO}}_{\mathrm{3}}\right)=\sum k_i\left[{\mathrm{VOC}}_i\right]\end{array}$$

k_het is the first-order loss rate coefficient of N₂O₅ uptake on the aerosol surface. It depends on the uptake coefficient γ(N₂O₅), the aerosol surface area density S_a (µm² cm⁻³), and the mean molecular speed c (Eq. 7). γ(N₂O₅) is influenced by chemical composition, physical properties of aerosol, and ambient conditions including related humidity and temperature (Yu et al., 2020; Wagner et al., 2013; Wang et al., 2018b; Bertram and Thornton, 2009; Tang et al., 2014; Kane et al., 2001). There are several kinds of methods proposed to quantify or estimate γ(N₂O₅) by using observed parameters. Given that some essential parameters were not directly measured during this campaign, only two approaches were employed to estimate the N₂O₅ uptake coefficient. The first method is the pseudo-steady-state method, which assumes that N₂O₅ and NO₃ have achieved a steady state (Brown et al., 2009). γ(N₂O₅) and $k_{{\mathrm{NO}}_{\mathrm{3}}}$ can be determined from the slope and intercept of linear regression of K_eq[NO₂] τ(N₂O₅)⁻¹ versus 0.25cS_aK_eq[NO₂], respectively, as shown in Eq. (8). The second is the parameterization method. As the aerosol compositions used to estimate the N₂O₅ uptake coefficients were not measured, only a simplified parameterization is available that is based on relative humidity (RH) and temperature (Eq. 9) (Hallquist et al., 2003; Kane et al., 2001; Evans and Jacob, 2005). Although simple, it had an overall reasonable performance in China (Wang et al., 2022; Tham et al., 2018; Wang et al., 2020a).

$$\begin{array}{}\text{(7)}& {\displaystyle }{k}_{\mathrm{het}}={\displaystyle \frac{\mathrm{1}}{\mathrm{4}}}c{S}_{\mathrm{a}}\mathit{\gamma }\left({\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{5}}\right)\text{(8)}& {\displaystyle }{K}_{\mathrm{eq}}\left[{\mathrm{NO}}_{\mathrm{2}}\right]\mathit{\tau }({\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{5}}{)}^{-\mathrm{1}}={\displaystyle \frac{\mathrm{1}}{\mathrm{4}}}c{S}_{\mathrm{a}}\mathit{\gamma }\left({\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{5}}\right)K_{\mathrm{eq}}[{\mathrm{NO}}_{\mathrm{2}}]+k_{{\mathrm{NO}}_{\mathrm{3}}}\text{(9)}& {\displaystyle }\mathit{\gamma }\left({\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{5}}\right)=\mathit{\alpha }\times {\mathrm{10}}^{\mathit{\beta }}\text{(10)}& {\displaystyle }\begin{array}{rl}\mathit{\alpha }& =\mathrm{2.79}\times {\mathrm{10}}^{-\mathrm{4}}+\mathrm{1.3}\times {\mathrm{10}}^{-\mathrm{4}}\times \mathrm{RH}-\mathrm{3.43}\times {\mathrm{10}}^{-\mathrm{6}}\\ & \times {\mathrm{RH}}^{\mathrm{2}}+\mathrm{7.52}\times {\mathrm{10}}^{-\mathrm{8}}\times {\mathrm{RH}}^{\mathrm{3}}\end{array}\text{(11)}& {\displaystyle }\mathit{\beta }=\mathrm{4}\times {\mathrm{10}}^{-\mathrm{2}}\times \left(T-\mathrm{294}\right)\left(T>\mathrm{282}\mathrm{K}\right)\text{(12)}& {\displaystyle }\mathit{\beta }=-\mathrm{0.48}\left(T<\mathrm{282}\mathrm{K}\right)\end{array}$$

3 Results and discussion

3.1 Measurement overview

The time series of N₂O₅, related trace gases, and selected meteorological parameters for the study period are depicted in Fig. 2. The air masses are categorized into IAM and CAM according to the backward trajectories at 20:00, 00:00, and 04:00 each day as illustrated in Fig. 1. Detailed information of two kinds of air masses is listed in Table 2. Data gaps for N₂O₅ were caused by technical problems, mirror reflectivity calibration, or instrumental maintenance, which usually took place in the daytime. In this campaign, meteorological conditions featured a typical subtropical winter climate with average temperature and RH values of 20.1 ± 1.9 ^∘C and 52.0 % ± 13.6 %, respectively.

Figure 2 Time series of 1 h average N₂O₅, NO₃, P(NO₃), NO, NO₂, S_a, temperature, and relative humidity. NO₃ was calculated by measured N₂O₅ according to the thermal equilibrium. The light gray shadow indicates the nighttime period. The ribbon at the top separates the air masses into two categories, yellow for IAM and blue for CAM. The x axis shows the time of day and date given in month-day format.

Ozone exhibited the characteristics of afternoon photochemical peaks, especially when the airmass came from inland. The average and maximum concentrations of ozone were 48.2 ± 18.2 and 120.1 ppbv, respectively. Once the maximum hourly average O₃ exceeded the Chinese national air quality standard (200 µg m⁻³, equivalent to 93 ppbv), we marked this day as an O₃ pollution day. During the campaign, 6 out of 37 days were O₃ polluted days, and all occurred during IAM periods. Meanwhile, the mixing ratio of NO, NO₂, and S_a usually increased during these days, indicating that this site was strongly affected by regional transport from the PRD city clusters. Previous observations by Wang et al. (2018a) also found high O₃ levels in autumn on the same island due to the weak NO titration and high O₃ production rate.

Nitrogen oxides (NO_x= NO + NO₂) were at a moderate level, with an average value of 13.1 ± 8.2 ppbv, which is much lower than the values in PRD regions (usually > 20 ppbv; Wang et al., 2022; Yang et al., 2022; Yun et al., 2018b) and higher than those on the remote islands in South China Sea (< 5 ppbv, Chuang et al., 2013). The mixing ratio of NO at nighttime was low and showed small peaks during daytime. With the O₃ accumulating throughout the day, NO decreased to below the instrument detection limit in the first half of the night, and it began to increase as the O₃ concentration decreased in the second half of the night. Given that the lifetime of NO is only a few minutes in the presence of several tens of ppbv of O₃ (Dewald et al., 2022), we inferred that NO likely originates from a local source such as soil emission, boats, cooking, and so on.

Table 2 Summary of parameters on the two air mass types (mean ± standard deviation).

Note: ^a NO₃ is calculated by the thermal equilibrium between NO₂, NO₃, and N₂O₅. ^b “All day” means the 24 h average and “Nighttime” means the time between 18:00 and 06:00 local time. ^c Without N₂O₅ measurements in the daytime and limited N₂O₅ data during the CAM period, N₂O₅, NO_3, and their lifetimes were not valid here.

N₂O₅ was at a moderate level on most days, with a nocturnal average of 119.5 ± 128.6 pptv, with high concentrations (> 400 pptv in 1 h average) in the first 3 d of the campaign. During the nights of 9 to 12 November, the N₂O₅ concentrations were significantly higher than those on other nights, with a maximum of 657.3 pptv at midnight of 12 November. The NO₃ concentration (calculated based on the thermal equilibrium with N₂O₅) was also moderate, with an average mixing ratio of 9.9±12.5 pptv, which was higher than that reported on a nearby coastal site of Hong Kong Island (Yan et al., 2019). Table 3 compares the N₂O₅, NO₃, and P(NO₃) values found in other coastal (or island) and continental regions of Europe, the United States, and China. In our study, N₂O₅ and NO₃ were at a moderate level compared to other coastal regions when they were affected by emission plumes from continental regions, such as northwestern Europe (Morgan et al., 2015), the east coast of the USA (Brown et al., 2004), and Shenzhen, China (Niu et al., 2022), and were comparable with urban regions (H. Wang et al., 2017b; Wang et al., 2018b). The concentrations of NO₃ precursors (NO₂ and O₃) at this site were very similar to some rural areas, leading to a high NO₃ production rate with a daily average of 1.5±0.9 ppbv h⁻¹ and a maximum of 5.9 ppbv h⁻¹. The average value is much higher than that reported in Beijing in winter (0.4 ppbv h⁻¹, Wang et al., 2021), comparable to autumn (1.4 ± 1.7 ppbv h⁻¹, H. Wang et al., 2017b) and even higher than that in summer Taizhou (1.01 ± 0.47 ppbv h⁻¹, Wang et al., 2020a). The nocturnal average P(NO₃) during this campaign was 1.4±0.7 ppbv h⁻¹, which is higher than the average value in the warm season of China of 1.07 ± 0.38 ppbv h⁻¹ (H. Wang et al., 2023). The high reaction rate constant for NO₂ and O₃ due to the high temperature at this site is a potential explanation for the high P(NO₃) values observed in this study (i.e., at the same NO₂ and O₃ level if the temperature increased from 10 to 20 ^∘C, the reaction rate constant would increase from 2.27 × 10⁻¹⁷ to 3.05 × 10⁻¹⁷, which means P(NO₃) would be 1.34 times faster). The high P(NO₃) and the low concentrations of N₂O₅ and NO₃ indicate intensive atmospheric oxidation capacity and fast NO₃ and N₂O₅ removal over the Pearl River Estuary.

The mean diurnal profiles of N₂O₅, together with relevant species are shown in Fig. 3. Daytime N₂O₅ and NO₃ in the IAM period were shown as NaN due to the absence of observations. Because of limited N₂O₅ data for the CAM period, neither N₂O₅ nor NO₃ is shown in Fig. 3. NO exhibited similar diurnal variations in both periods and the mixing ratio was higher in the IAM period. The wind rose plot (Supplement Fig. S1) showed high concentrations of NO originating from the north characterized by the outflow from PRD regions. However, NO₂ differed in the two periods, showing high anti-correlation with O₃ only in the IAM period and little diurnal variation in the CAM period.

Figure 3 Mean diurnal profiles of N₂O₅, NO₃, P(NO₃), and relevant parameters in the two types of air masses. NO₃ was calculated from N₂O₅. Neither N₂O₅ nor NO₃ was shown during the CAM period because of limited N₂O₅ measurement. The x axis shows the time of day.

Ozone exhibited a typical diurnal pattern for all air masses, gradually increasing until its peak at 16:00 and then slowly decreasing throughout the night until its lowest mixing ratio was reached at about 06:00. Compared to the CAM period, the lower minimum hourly O₃ concentration and a small peak of NO₂ in the early morning indicated that NO titration effect was stronger in the IAM period, and the higher maximum of O₃ concentration in IAM indicated that photochemical formation of O₃ and/or transport was faster to completely offset the titration. In addition, the higher NO_x and VOC concentrations (as shown in Table S1 in the Supplement) in the IAM period facilitated O₃ formation. With the elevated precursor concentrations (NO₂ and O₃) in the IAM period, N₂O₅ and NO₃ accumulated rapidly after sunset, reaching their peak values (492.1 and 49.6 pptv for each). P(NO₃) was highly consistent with O₃ in diurnal variation and reached the peak at 16:00, with peak values of 2.7 ppbv h⁻¹ (IAM) and 1.9 ppbv h⁻¹ (CAM), as well as a nocturnal average value of 1.5 ± 0.8 ppbv h⁻¹ (IAM) and 1.2 ± 0.6 ppbv h⁻¹ (CAM), respectively. The P(NO₃) of CAM was consistent with the observation when the air mass over eastern Shenzhen was transported from the clean area or sea surface (1.2 ± 0.3 ppbv h⁻¹, Niu et al., 2022).

Table 3 Summary of field-observed N₂O₅, NO₂, and O₃ concentrations and NO₃ production rate.

Notes: UAM means air masses coming from continental areas, and BAM means air masses coming from background marine areas. Mean values are in the form of mean ± standard deviation or single data. The maximum was noted in the table.

3.2 The lifetimes of N₂O₅ and NO₃

Steady-state lifetime is one of the most common and useful diagnostics for NO₃ and N₂O₅ analysis in the atmosphere (Brown et al., 2003; Wang et al., 2018b; Wang et al., 2020a; Brown et al., 2016). As shown in Fig. 4, ${\mathit{\tau }}_{{\mathrm{NO}}_{\mathrm{3}}}$ was low during the whole campaign, with an average of 0.5 ± 0.7 min; ${\mathit{\tau }}_{{\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{5}}}$ showed a similar pattern to ${\mathit{\tau }}_{{\mathrm{NO}}_{\mathrm{3}}}$ but had a much higher value, ranging from 0 to 34.1 min with an average of 6.1 ± 6.5 min. The N₂O₅ lifetime was higher in the first half of the campaign (11.5 min, 9 to 14 November) than in the second half (3.5 min, 22 to 28 November). The difference was mainly due to the N₂O₅ mixing ratio rather than P(NO₃), as P(NO₃) showed no significant difference during the whole observation (Fig. 2).

Figure 4 Time series of N₂O₅ and NO₃ lifetimes (x axis showing the time of day and date given in month-day format.) (a) and variations in nocturnal N₂O₅ lifetime as a function of aerosol surface area density, S_a (b), and relative humidity, RH (c). The blue diamond represents the average ${\mathit{\tau }}_{{\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{5}}}$ and pink dots represent scatter data points of 1 min. The ribbon at the top separates the air masses into two categories, yellow for IAM and blue for CAM.

${\mathit{\tau }}_{{\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{5}}}$ values were comparable to those measured on the coastline of Finokalia, Greece, for a median of 5 min (Vrekoussis et al., 2004, 2007) but much lower than those previously reported in the residual layer in Hong Kong for 1–5 h (Brown et al., 2016). In comparison, the lifetimes were much longer than in inland urban areas, for example, 0.93 ± 1.13 min in Taizhou (Li et al., 2020) and 1.6±1.5 min in Changzhou (Lin et al., 2022) for Yangtze River Delta (YRD) regions, and 1.1–10.7 min (Zhou et al., 2018) and 4.5±4.0 min (Wang et al., 2018b) in Beijing. Typically, high aerosol loading, more intensive VOC, and NO emissions in these areas have led to enhanced N₂O₅ uptake and reactions of NO₃ with VOC, while at this site, measurements indicated that the peak diameter in the particle number distribution was small during the whole campaign and indicated no significant difference between the two air masses with respect to the aerosol diameters (Fig. S2). S_a values ranged from 29 to 557 µm² cm⁻³. All of these indicated the atmosphere was relatively clean (H. Wang et al., 2017b), making N₂O₅ uptake slow. Figure 4b shows that N₂O₅ lifetime decreased rapidly from 8.3 to 1.7 min when S_a increased up to 300 µm² cm⁻³ and then remained at relatively low constant levels though S_a still increased. Such a trend of ${\mathit{\tau }}_{{\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{5}}}$–S_a dependence was consistent with previous observations and varied in exact values (Zhou et al., 2018; Wang et al., 2018b; Li et al., 2020). Figure 4c showed that ${\mathit{\tau }}_{{\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{5}}}$ decreased as RH increased (> 40 %), possibly due to the hygroscopic aerosol growth and the dependence of the N₂O₅ uptake coefficient on the RH (Brown and Stutz, 2012). Overall, the trend is consistent with previous works, while the large discrepancy of the dependence implied that N₂O₅ uptake was not the dominant NO₃ loss process.

3.3 The NO₃ reactivity and N₂O₅ uptake coefficients

The concurrent high P(NO₃) and low NO₃ lifetime imply high NO₃ reactivity as well as a large nocturnal NO₃ loss process at DWS. The NO₃ reactivity towards VOCs, calculated by Eq. (4), was categorized into anthropogenic VOC and biogenic VOC (Gu et al., 2021). Throughout the campaign, k(NO₃) varied considerably (Fig. 5a), showing relatively high and fluctuated values when the air masses featured IAM. The k(NO₃) ranged from 1.6 × 10⁻³ to 2.4 × 10⁻² s⁻¹, with a daily average of 4.6 ± 2.8 × 10⁻³ s⁻¹. Low values of k(NO₃) were observed from 9 to 12 December when the air masses originated on the coast or offshore from the east and southeast, featuring the outflow of coastal cities like Hong Kong and Shenzhen.

Figure 5 NO₃ reactivity via VOCs during the campaign. (a) k(NO₃) time series from 25 November to 15 December 2021 (x axis showing the time of day and date given in month-day format), (b) mean diurnal profiles (x axis showing the time of day), and (c) the relative contribution in different categories. The ribbon at the top separates the air masses into two air masses types, yellow for IAM and blue for CAM.

Figure 5b shows the mean diurnal profile of k(NO₃), where a trend of high values in the daytime and low values at nighttime are observed. Anthropogenic VOC, especially cresol, dominated the daily trend of k(NO₃), while biogenic VOC-k(NO₃) showed no significant diurnal variation. Except cresol, other highly reactive VOC showed little change throughout the day. Regarding the biogenic VOC-k(NO₃), the concentrations of monoterpene, isoprene, and DMS changed smoothly although their emissions increased with elevated temperature and sunlight during daytime (Fuentes et al., 2000). The detailed contributions of VOC categories to k(NO₃) were shown in Fig. 5c. The k(NO₃) was 5.6 ± 2.8 × 10⁻³ and 3.7 ± 2.5 × 10⁻³ s⁻¹ on average for daytime and nighttime, respectively. The daytime distribution of k(NO₃) differed from that at the mountaintop of Tai Mo Shan in Hong Kong (Brown et al., 2016). During the nighttime, anthropogenic VOC-k(NO₃) tripled the biogenic VOC-k(NO₃) and was dominated by cresol (26.4 %). The nighttime k(NO₃) corresponded to a NO₃ lifetime of 4.5 min, which was about 10 times the lifetime derived from steady-state analysis, indicating that the reaction of NO₃ with VOC was not significant enough. The faster NO₃ loss rate also indicated the less aged air mass that was influenced by surface-level emissions.

We showed that NO₃ reactivity and its composition in this study exhibited significant differences compared to other urban or forested regions (Z. Wang et al., 2017; Ayres et al., 2015; Brown et al., 2016; Lin et al., 2022). Although anthropogenic VOCs played a dominant role, accounting for 66.1 %, the major contributors were not low-carbon alkenes but phenol (13.9 %, 0.64 ± 0.28 × 10⁻³ s⁻¹) and cresol (30.8 %, 1.4 ± 1.0 × 10⁻³ s⁻¹), which have received little attention in previous studies. Despite their relatively low concentrations, averaging 7 ± 3 and 4 ± 3 pptv respectively, their substantial contribution to k(NO₃) is notable due to their fast rate constants (3.8 × 10⁻¹² and 1.4 × 10⁻¹¹ cm³ molec.⁻¹ s⁻² at 298 K, respectively) for reaction with NO₃. Considering that the measured phenol and cresol concentrations were low and near the instrumental detection limit, we note this may bring some uncertainties in quantifying the contribution to the total NO₃ reactivity and NO₃ loss rate. These substances are mainly secondary species from aromatic compounds and higher concentrations have also been observed, such as in the Strasbourg area, France (14 pptv, Delhomme et al., 2010), and in Great Dun Fell, UK (16 pptv, Lüttke et al., 1997). Hence, these phenolic compounds are potentially important but their contributions to NO₃ reactivity in urban areas are often overlooked, and their reactions with NO₃ may also contribute to the formation of nitrophenol. These reactions warrant further attention in future research. Regarding biogenic VOCs, besides the contributors commonly observed in forest regions such as monoterpenes and isoprene, the marine emissions indicator, dimethyl sulfide (DMS), contributed 10.2 % to NO₃ reactivity (daily average). Previous studies have suggested that DMS may serve as a major direct sink for NO₃ in clean marine regions (Allan et al., 1999; Aldener et al., 2006; Brown et al., 2007). However, this study reveals that anthropogenic VOC emissions significantly enhanced the NO₃reactivity in marine areas, highlighting the crucial influence of anthropogenic activities on marine atmospheric chemistry.

As shown in Fig. 6a, k(NO₃) differed significantly between the inland and coastal air masses, with 5.2 ± 3.1 × 10⁻³ and 3.7 ± 1.9 × 10⁻³ s⁻¹ on average in IAM and CAM periods, respectively. Of which anthropogenic VOC-k(NO₃) in IAM (3.5±2.3 × 10⁻³ s⁻¹) was higher than in CAM (2.3±1.4 × 10⁻³ s⁻¹) and dominant in both air masses, while biogenic VOC-k(NO₃) was comparable (1.7 ± 0.9 × 10⁻³ and 1.4 ± 0.6 × 10⁻³ s⁻¹ for IAM and CAM, respectively). The difference indicated that this region was affected by long-range transport emissions to a certain extent. The pie charts in Fig. 6b show different VOC categories that contributed to k(NO₃) in two periods with AVOC dominant at all times. The change in the relative contribution of various VOCs to k(NO₃) varied simultaneously throughout the day, showing an increase in butene, phenol, and DMS and a decrease in cresol and monoterpene from daytime to nighttime.

Figure 6 (a) Distributions of k(NO₃) from AVOC and BVOC for both IAM and CAM periods. The error bar indicates the standard deviation. (b) The relative contribution of VOC categories to the k(NO₃).

N₂O₅ heterogeneous uptake on aerosol is one of the vital loss processes of NO₃, and the uptake coefficient varied greatly under different environmental conditions. For instance, γ(N₂O₅) can reach up to 0.072 in polluted urban regions (H. Wang et al., 2017b; Wang et al., 2018b; Lu et al., 2022; Li et al., 2020) while usually below 0.03 in coastal areas (Brown et al., 2016; Morgan et al., 2015; Niu et al., 2022). N₂O₅ uptake coefficient can be obtained from the pseudo-steady-state method by assuming that N₂O₅ and NO₃ have achieved a steady state (Brown et al., 2009), in which the fitted slope represents γ(N₂O₅) and the intercept represents the direct loss rate coefficient, k(NO₃) (as shown in Eq. 8). However, this approach failed to generate valid results in our study since a negative slope or intercept was observed (Fig. S4). These results indicated that a large NO₃ removal process existed at this site, making it unable to approach a stable state. The γ(N₂O₅) was also calculated from 9 to 16 November by using the simplified parameterization, as shown in Eq. (9). The parameterized average γ(N₂O₅) showed a large variation ranging from 0.0014 to 0.0299, with an average of 0.0095±0.0059. This value is within the range of < 0.0016 to 0.03 derived from the ambient observation around other coastal areas (Niu et al., 2022; Yun et al., 2018a; Brown et al., 2006, 2016; Morgan et al., 2015) and smaller than the polluted North China Plain (X. F. Wang et al., 2017; H. Wang et al., 2017b; Z. Wang et al., 2017; Tham et al., 2018).

3.4 The NO₃ loss budget

To assess the contribution of various loss processes to the total NO₃ removal, we calculated their loss rate and the loss ratio, LR(NO₃). The LR(NO₃) is defined as the ratio of the loss rate by process X (VOC or N₂O₅ uptake) to the total NO₃ loss rate; here the total NO₃ loss rate is represented by P(NO₃) since we cannot quantify the total NO₃ loss rate due to the NO concentration below the limit of instrument detection. Due to the data absence of measured VOCs or N₂O₅ during certain periods, the loss proportion of VOCs and N₂O₅ uptake in NO₃ loss are only presented from 26 November to 5 December 2021, during which all air masses originated from continental China. As shown in Fig. 7, a closer examination revealed that the nights can be divided into two periods: period I, 25 to 28 November when the loss ratio of VOC and N₂O₅ uptake remained below 3 %, and period II, 30 November to 4 December when the loss ratio was higher. Both periods had large nocturnal NO₃ production rates with an average of 2.1 ± 1.1 ppbv h⁻¹ in period I and 1.4 ± 0.6 ppbv h⁻¹ in period II, respectively.

Figure 7 Time series of (a) the loss rate of NO₃ reactions with AVOC, BVOC, and N₂O₅ uptake and (b) fractional contribution to the NO₃ loss during the nighttime by taking P(NO₃) as the total NO₃ loss in the IAM period (x axes showing the time of day and date given in month-day format). The mean diurnal profiles of NO₃ loss ratio in two periods (c) 25–28 November and (d) 30 November–4 December 2021 (x axis showing the time of day). Pie charts in the center showed the relative contribution of VOCs (blue) and N₂O₅ uptake (red) in NO₃ loss.

N₂O₅ uptake rate was larger in period I (0.01 ± 0.01 ppbv h⁻¹) than in period II (0.006 ± 0.004 ppbv h⁻¹), which can be explained by the increased RH, S_a, and N₂O₅ concentration, as shown in Fig. 2. The loss ratio of these processes is shown in Fig. 7b: the total NO₃ loss through reactions with VOCs and N₂O₅ uptake accounted for less than 20 %, with an average of 1.2 % (period I) and 5.3 % (period II), respectively. This result shows that the nighttime NO₃ chemistry may be almost negligible in the NO_x removal compared with the day OH + NO₂ pathway according to previous works reporting on urban regions (H. Wang et al., 2017b; Wang et al., 2020a). The diurnal variation of the NO₃ loss fraction of both periods is shown in Fig. 7c and d, revealing that NO₃ loss via N₂O₅ uptake and VOCs was slightly higher in the early evening and relatively stable in the late evening. The pie charts in the center were the relative contribution between VOCs and N₂O₅ uptake, showing that VOCs were overwhelming compared with N₂O₅ uptake during the two periods, with an average of 68.4 % and 91.7 % during the first and second periods, respectively.

To better understand the nocturnal oxidation of VOCs, we compared the nighttime oxidation of VOCs by NO₃ with O₃. Since OH was not measured and OH is often regarded as a vital daytime oxidant (Finlayson-Pitts and James, 2000; Lu et al., 2010), we did not consider OH oxidation in the nighttime. Figure S4 showed the diurnal pattern of VOC loss rate by NO₃ and O₃: NO₃ predominantly achieves its peak oxidation rates (0.07 ppbv h⁻¹) during the initial half of the night, accounting for 63.1 % of the total VOC oxidation on nocturnal average. Meanwhile, O₃ also makes a contribution to VOC oxidation, mainly owing to its relatively high nighttime concentration levels (42.9 ± 18.4 ppbv).

Due to the difficulty in experimental quantifying γ(N₂O₅), the estimation of N₂O₅ uptake in NO₃ loss may include some uncertainty. Considering the uncertainty both in parameterized γ(N₂O₅) and the NO₃ reactivity calculation, three sensitivity tests were conducted to assess the uncertainty in period II because of the relatively high loss ratio in the above analysis (Fig. 8), and the three cases were used to represent the upper limit of their contribution to NO₃ loss. Case 1 represents the overestimated contribution of N₂O₅ uptake by taking γ(N₂O₅)= 0.08, which was the high value reported from a high N₂O₅ and ClNO₂ plume of Shenzhen (Niu et al., 2022) and approximately 7 times the parameterized value at this site. In this case, the fraction of NO₃ + VOCs and N₂O₅ uptake was significantly elevated to account for approximately 30 % of NO₃ loss. Case 2 shows the total NO₃ reactivity reached an average of 5.0 × 10⁻³ s⁻¹ by taking β-pinene as the total monoterpene because of the higher reaction rate constant. The weak change in the loss ratio indicates the reactions of NO₃ with VOC may not be sensitive to the weights of monoterpenes, since the contribution of monoterpenes to the NO₃ reactivity is not dominant. Case 3 is the synthesis of Case 1 and Case 2 by considering a higher N₂O₅ uptake coefficient and higher k(NO₃) to represent the upper limit of N₂O₅ uptake and NO₃ reaction with VOCs to NO₃ loss, whose result is slightly higher than the contribution of Case 1. Nevertheless, the quantified upper contribution was still less than half. Thus, we conclude that most of the NO₃ loss was not well accounted for even considering the uncertainties.

Figure 8 Three sensitivity tests for the contribution of VOCs and N₂O₅ uptake to the NO₃ loss during the nighttime of period II (30 November–4 December 2021). Case 1 takes γ(N₂O₅) = 0.08, which is the high value reported in a previous study. Case 2 takes β-pinene as the total monoterpene with a higher reaction rate constant, and Case 3 is the synthesis of the above two cases to represent the upper limit of the contribution.

The NO₃ reaction with NO was often considered to be one of the dominant loss processes during the daytime since at nighttime NO decreased to low levels, thus not considered in the above analysis. However, by taking NO into consideration, although at low concentration levels below the detection limit of the instrument (0.4 ppbv), the contribution of NO to the nighttime NO₃ loss exceeded 100 % frequently, as shown in Fig. S5. Due to the rapid reaction between NO and NO₃, several pptv concentrations of NO could effectively account for most NO₃ loss in a relatively clean coastal environment (Crowley et al., 2011). Nevertheless, limited by precise NO measurement, we considered the following assessments to understand the total NO₃ loss processes (Fig. 9). By assuming NO at a constant value of 40–400 pptv, more than 80 % of the total NO₃ loss can be well explained. Although some loss remained unidentified, these results underline that NO, often considered to be important during daytime, was the predominant NO₃ loss pathway during nighttime at this study site. This also suggests accurate measurement of low NO concentrations is crucial for identifying removal pathways of nocturnal NO₃ oxidants and has significant implications for nighttime atmospheric chemistry. We can infer that the nocturnal chemical NO₃ reactions would be largely enhanced once without NO emission in the open ocean after the air mass passes through this site, indicating the strong influences of the urban outflow to the downwind marine areas with respect to nighttime chemistry.

Figure 9 Examples for the assessment of NO₃ loss process by assuming NO as constant values to approximately explain about 80 % of the budget. The x axes show the time of day for 3 (a) and 4 (b) December.

In the absence of measured N₂O₅ during the CAM period, we compared k(NO₃) and the reactivity of N₂O₅ uptake (k_hetK_eqNO₂) to indirectly reflect NO₃ removal process. Overall, the NO₃ reactivity values from VOCs and N₂O₅ uptake during nighttime were relatively comparable, 56.5 % and 43.5 %, respectively. This indicates that VOCs still had a slightly larger contribution than N₂O₅ uptake during the CAM period, which is consistent with the findings in southern China (Brown et al., 2016) and on the east coast of the USA (Aldener et al., 2006).

4 Summary and conclusion

This study presents the first observation of nocturnal nitrogen oxide species, N₂O₅, at a typical marine site (Da Wan Shan Island, Zhuhai) in the north of the South China Sea during the winter of 2021. Although Da Wan Shan Island was almost free of local anthropogenic emissions, the air pollutants from the megacities of the Pearl River Delta were transported to this area by northerly or northeasterly winds during the measurement period. The maximum ratio of N₂O₅ was 657.27 pptv (1 min average) and the nocturnal average was 119.5 ± 128.6 pptv. The NO₃ production rate was comparable to that in urban areas such as north China and the Yangtze River Delta, with an average value of 1.5 ± 0.9 ppbv h⁻¹ and a maximum of up to 5.84 ppbv h⁻¹, indicating an active nighttime chemical process in that area.

Further analysis of N₂O₅ and NO₃ steady-state lifetimes indicates that NO₃ had a very short average life of 0.5 ± 0.6 min, which was to some extent comparable to that in urban areas in summer. The combination of the high NO₃ production rate and short lifetime suggests a rapid NO₃ loss at night. While N₂O₅ uptake is inefficient in relatively clean air masses. The nighttime k(NO₃) corresponded to a NO₃ lifetime of 4.5 min, indicating that VOCs also contribute little to NO₃ loss. Both VOC and N₂O₅ uptake can only explain less than 20 % of the total loss. The fast NO₃ loss rate also indicated the air mass that was influenced by local surface-level emissions. We infer that the local weak NO emission may significantly change the near-surface chemical pattern of NO₃ chemistry, which may result in a huge difference between the observed results on the island and those on the sea surface. We suggested that future field measurements should be made on sea surfaces away from islands, such as ship-based cruise observations, to obtain a comprehensive understanding of the nocturnal NO₃ chemistry in the background marine regions.