The ship-based atmospheric observation campaign along the marginal seas of China was conducted from 19 April to 16 May 2018. The latitude and longitude ranges of the entire campaign covered 21.12–35.89∘ N and 110.67–122.16∘ E. The detailed voyage records of the observation ship are shown in Table 1. An integrated and fully automated MAX-DOAS instrument was installed on board the stern deck of the ship (Fig. S1a in the Supplement). To ensure that the instrument was always kept in a horizontal position, a photoelectric gyro was used. The angle between the observation and heading direction of the ship was always maintained at 135∘ during the whole campaign. The telescope unit of the instrument pointed towards the sea during cruise no. 3 and no. 6. The telescope unit pointed towards the land during cruise no. 1, no. 4, and no. 5. During cruise no. 2, the observation telescope always pointed to Chongming Island. The measurement ship only sailed during daytime from 19 April to 2 May and continuously sailed during the entire daytime and nighttime from 3 to 16 May 2018. The ship docked in the port of Daishan on 9–10 May, and no observations were conducted during these 2 d.

The aim of this campaign was to learn the vertical differences in the NO2 heterogeneous reaction to produce HONO in marginal seas of China and compare the influence mechanism of that in inland cities. To fully understand the differences of the impacts of RH, temperature, and aerosol on the HONO secondary formation in land and sea conditions, the Chinese Academy of Meteorological Sciences (CAMS) and Southern University of Science and Technology (SUST) MAX-DOAS stations were selected as inland and coastal areas for analysis, respectively. CAMS is located in the urban area of Beijing (39.94∘ N, 116.32∘ E), and SUST is located in Shenzhen (22.60∘ N, 114.00∘ E) (Fig. S2). This study will provide scientific guidance for understanding regional oxidation capacity and controlling the secondary air pollution.

Aerosol and trace gas (i.e., NO2 and HONO) vertical profiles are retrieved from MAX-DOAS measurements using the algorithm reported by Liu et al. (2021). The inversion algorithm is developed based on the optical estimation method (OEM) (Rodgers, 2000), which employs the radiative transfer model VLIDORT as the forward model. The detailed retrieval procedure is displayed in Sect. S1 and Fig. S3.

In this study, an exponentially decreasing a priori profile with a scale height of 1.0 km was used as the initial profile for both the aerosol and trace gas retrievals (Fig. S4). The surface concentrations of aerosol, NO2, and HONO were set to 0.2 km-1, 3.0 ppb, and 1.0 ppb, respectively. We assume a fixed set of aerosol optical properties with an asymmetry parameter of 0.69, a single scattering albedo of 0.90, and ground albedo of 0.05. Furthermore, the uncertainty of the aerosol and trace gas a priori profile was set to 100 %, and the correlation length was set to 0.5 km. The averaging kernels indicated that the sensitivity of the profile retrieval tended to decrease with increasing altitude and the retrieval especially sensitive to the layers within 0–1.5 km (Fig. S5). The sum of the diagonal elements in the averaging kernel matrix is the degree of freedom (DOF), which denotes the number of independent pieces of information contained in the measurements.

For profile retrieval, the error sources can be divided into four different types: smoothing error, measurement noise error, forward model error, and model parameter error (Frieß et al., 2006). However, in terms of this classification, some errors are difficult to calculate or estimate. For example, forward model error, which is caused by an imperfect representation of the physics of the system, is hard to quantify due to the difficulty of acquiring an improved forward model. Given calculation convenience and contributing ratios of different errors in total error budget, we mainly took into account error sources based on the following classification, which were smoothing and noise errors, algorithm error, cross-section error, and uncertainty related to the aerosol retrieval (only for trace gas). Here, we estimated the contribution of different error sources to the trace gas vertical column densities (VCDs) and aerosol optical depth (AOD), as well as near-surface (0–200 m) trace gas concentrations and aerosol extinction coefficients (AECs). The detailed demonstrations and estimation methods are displayed below, and the final results are summarized in Table 3. a.

Smoothing errors arise from the limited vertical resolution of profile retrieval. Measurement noise errors denote the noise in the spectra (i.e., the fitting error of DOAS fits). They can be quantified by averaging the error of retrieved profiles, as the error of the retrieved state vector equals the sum of these two independent errors. We calculated the sum of smoothing and noise errors on near-surface concentrations and column densities, which were 14 % and 5 % for aerosols, 16 % and 17 % for NO2, and 20 % and 22 % for HONO, respectively in the sea scene. The corresponding values were 13 % and 5 % for aerosols, 14 % and 16 % for NO2, and 18 % and 20 % for HONO, respectively, at SUST and 13 % and 5 % for aerosols, 15 % and 17 % for NO2, and 19 % and 21 % for HONO at CAMS.

Meteorological data (including temperature, pressure, relative humidity, visibility, solar radiation intensity, wind speed, and wind direction) with a temporal resolution of 1 min were measured by the weather station installed on the ship. NO was measured using an NO analyzer (Thermo Scientific model 42i) with a 1 min resolution. The speed of the ship was calculated referring to the GPS data.

The temperature and relative humidity of two ground-based stations (i.e., CAMS and SUST) were collected from the Weather Underground website, the temporal resolution of which is around 3 h.

The backward trajectory was calculated using HYSPLIT (Hybrid Single-Particle Lagrangian Integrated Trajectory) developed by the National Oceanic and Atmospheric Administration Air Resource Laboratory (NOAA-ARL). The meteorological data with a 1∘ × 1∘ spatial resolution and 24 layers were collected from the Global Data Assimilation System (GDAS).

The radiative transfer model SCIATRAN was used to convert SCDs of NO2 and HONO to their tropospheric VCDs. The vertical profiles of aerosol, NO2, and HONO retrieved from MAX-DOAS; the temperature and pressure vertical profiles simulated using a dynamical chemical model (WRF-Chem); and the geo-position data collected by GPS were introduced as inputs in SCIATRAN for the NO2 and HONO air mass factor (AMF) calculation. Missing data are due to power and instrument system failure, interference of ship plume, unfavorable weather conditions (i.e., heavy rain), and night sailing. During the cruise of Chongming to Zhanjiang, NO2 VCDs varied from 1.05 × 1014 to 4.02 × 1016 molec. cm-2 with an averaged value of 3.90 × 1015 molec. cm-2. From Zhanjiang to Qingdao, NO2 VCDs varied from 1.08 × 1014 to 2.60 × 1016 molec. cm-2 with an averaged value of 4.27 × 1015 molec. cm-2. From Chongming to Zhanjiang, HONO VCDs varied from 1.00 × 1014 to 2.58 × 1015 molec. cm-2 with a mean value of 2.39 × 1014 molec. cm-2. From Zhanjiang to Qingdao, HONO VCDs varied from 1.01 × 1014 to 2.61 × 1015 molec. cm-2 with a mean value of 2.74 × 1014 molec. cm-2.

Figure 2 shows the spatial distribution of NO2 and HONO VCDs over the marginal seas of China. Five enhanced tropospheric NO2 VCD hot spots were observed during the whole campaign, i.e., the coastal areas of the Yangtze River Delta, Taiwan Strait, Guangzhou–Hong Kong–Macau Greater Bay Area, port of Zhanjiang, and port of Qingdao. In the coastal areas of Yangtze River Delta, the hot spots were mainly distributed in the Yangtze River estuary, Hangzhou Bay, port of Ningbo, port of Taizhou, and port of Wenzhou. These areas are mostly important shipping channels or shipping ports and are great NO2 emission sources. The averaged NO2 VCDs in the above five areas reached 1.07 × 1016, 1.30 × 1016, 7.27 × 1015, 5.34 × 1015, and 3.12 × 1015 molec. cm-2, respectively (Fig. S6a). HONO exhibited similar spatial distribution characteristics as NO2, and the averaged HONO VCDs in the above five hot-spot areas reached 1.01 × 1015, 7.91 × 1014, 6.02 × 1014, 5.36 × 1014, and 5.17 × 1014 molec. cm-2, respectively (Fig. S6b). It indicates that NO2 is an important precursor of HONO. Earlier studies reported that HONO can be generated from NO2 through heterogeneous reactions on the surface of aerosol and the sea (Yang et al., 2021). However, there are obvious differences in the concentration distribution of HONO and NO2 in the southeast coastal area of Jiangsu (from Qidong to Dongtai). In this area, NO2 showed a higher concentration (1.66 × 1016 molec. cm-2, which is 4 times higher than the mean NO2 VCD), while HONO showed a lower concentration (2.06 × 1014 molec. cm-2, which is ∼ 80 % of the mean HONO VCD). It may be the fresh ship emission plume on the route enhancing the NO2 concentration and HONO had not been fully formed from the NO2 heterogeneous reaction in time, since the observations from the ship-based MAX-DOAS are instantaneous.

The surface concentrations of NO2 and HONO were extracted from their corresponding vertical profiles. As shown in Fig. 3, the total averaged near-surface NO2 concentrations under sea-oriented and land-oriented measurements were 8.46 and 11.31 ppb, respectively. The total averaged near-surface HONO concentrations were 0.23 and 0.27 ppb under sea-oriented and land-oriented measurements. The total averaged near-surface HONO / NO2 ratios in sea-oriented and land-oriented measurements were 0.027 and 0.024, respectively. Earlier studies reported that vehicle and ship emissions were the main primary HONO sources on land and sea, respectively, and NO2 heterogeneous reactions on the surfaces of the ground, sea, vegetation, and aerosol were the important secondary HONO sources (Liu et al., 2021). Additionally, Yang et al. (2021) found that the surface HONO concentrations in the sea cases were lower than that those in the land cases, especially in the morning and evening. Figure 4 shows the time series of AOD, the surface concentrations of NO2 and HONO, and the surface HONO / NO2 during the whole campaign. We found that the time series of AOD and NO2 were similar. The high AOD and NO2 usually appeared in busy shipping channels and ports, and the obvious high-value areas were the coast of the Yangtze River Delta, the Taiwan Strait, port of Xiamen, port of Zhanjiang, and port of Qingdao (with mean AOD and NO2 of 1.28 and 18.90 ppb, respectively). HONO always appeared under high AOD and NO2 conditions; however, high AOD and NO2 were not necessarily accompanied by a high HONO concentration. This was because the heterogeneous formation of HONO requires suitable meteorological conditions (i.e., RH and temperature) in addition to its precursor (NO2) and the reaction surface (aerosol) (Liu et al., 2019). The high HONO / NO2 values were found on 2, 13, and 14 May with an average value of 0.45. Furthermore, we found the high values of HONO / NO2 always appeared from 11:00 to 14:00 LT (UTC+8)​​​​​​​ during the whole campaign.

Sun et al. (2020) reported that HONO concentrations could increase by up to 40 %–100 % over the shipping routes and international ports, and Huang et al. (2017) reported vehicle exhaust could contribute to ∼ 12 %–49 % of the atmospheric HONO budget. Since the direct emissions of the measurement ship were removed before data analysis, the primary source of HONO during the whole campaign was mainly from the direct emissions of cargo ships. By subtracting the average marine background of NOx and HONO from the ship plume emission values, the impact of background values is reduced and the emission ratio of ΔHONO / ΔNOx can be obtained, and this emission ratio can be used for quantifying the primary HONO (Sun et al., 2020; Xu et al., 2015). In this study, we used an averaged 0.46 ± 0.31 % emission ratio of ΔHONO / ΔNOx referring to Sun et al. (2020) to understand the primary source of HONO on the sea surface during the campaign. The NO was measured using an in situ instrument, and sea surface NO2 was extracted from the retrieved NO2 vertical profiles (NOx = NO + NO2). Additionally, the calculation method of emission ratios of ΔHONO / ΔNOx in CAMS and SUST was taken from Xu et al. (2015), Liu et al. (2019), and Xing et al. (2021) (Sect. S2). The averaged emission ratios in CAMS and SUST were 0.82 ± 0.34 % and 0.79 ± 0.31 %, respectively. The direct emissions were deduced in the following study of the secondary formation of HONO. The ratios of HONO / NO2 in CAMS, SUST, and the ship-based campaign can be found in Fig. S7. Furthermore, the main secondary formation pathway of HONO is considered the heterogeneous reaction of NO2 on the surface. The linear regression between HONO and NO2 in the land and static sea scenarios is shown in Fig. 5. We found the fitting slopes in static sea scenes were ∼ 8–10 times larger than those in land scenes, especially for sea-oriented measurements under static weather conditions (slope ≈ 0.06). The correlation coefficients (R) in inland and static sea scenarios were all > 0.62, except in SUST (R = 0.58), which indicates the formation rate of secondary HONO from NO2 heterogeneous reaction in static sea scenarios may be faster than that in land scenarios. The corresponding temperature and RH conditions of each spot are displayed in Fig. S8, which roughly reveals the impact of RH and temperature on the process of NO2 forming HONO through heterogeneous reactions.

To further investigate the height dependence of HONO / NO2 under land and sea conditions, two cases in the Pearl River Delta (PRD) were selected from the whole campaign. As shown in Fig. 9, “A” and “B” had a similar aerosol level (the extinction coefficients in the surface layer being 0.45–0.60 km-1) and vertical distribution structure and were all observed from 10:00 to 11:00 LT. The instrument viewing the sea accompanied with sea wind in “A” is named the sea scene, and the instrument viewing the land accompanied with land wind in “B” is named the land scene. The NO2 concentrations in the sea and land scenes have a similar vertical structure, and the NO2 concentrations in the land scene are larger than those in the sea scene except on the surface layer. The HONO have the same vertical distribution structure in the above two scenes, and the HONO concentration in the land scene is always larger than that in the sea scene. In Fig. 9e, we found that HONO / NO2 under 0–400 m in the land scene is higher than that in the sea scene; however, the HONO / NO2 values are obviously lower in the land scene than those in the sea scene above 400 m. Furthermore, the growth rate of HONO / NO2 with the increase in height in the sea scene is significantly faster than that in the land scene above 400 m. This indicates the generation rates of HONO sourced from NO2 heterogeneous reaction on the aerosol surface in the sea scene are larger than those in the land scene above 400 m. Under 400 m, the HONO generation rates in the land scene are larger than those in the sea scene.

Additionally, we selected inland cases (CAMS) to learn the difference in height dependence of HONO / NO2 compared with sea scenes with different aerosol loads. As shown in Fig. 10, the sea and inland scenes had similar aerosol levels (low aerosol level: < 0.2 km-1) and vertical structure. Furthermore, the NO2 and HONO in the sea and inland scenes have a similar vertical structure, although their concentrations in the sea scene are all larger than those in the inland scene. In Fig. 10d, we found that the HONO / NO2 in the sea scene was obviously larger than that in the inland scene above 400 m. The HONO / NO2 in the sea scene was about 4.5 times larger than that in the inland scene especially above 600 m. As shown in Fig. 11, the aerosols in the sea and inland scenes exhibited similar extinction levels (relatively high level: ∼ 0.8 km-1) and vertical structure. The NO2 concentration in the sea scene was higher than that in the inland scene but with a similar vertical structure. The HONO concentration in the sea scene was lower than that in the inland scene under 400 m, while the concentration in the sea scene was larger than that in the inland scene above 400 m. In Fig. 11d, we found the HONO / NO2 in the inland scene was larger than that in the sea scene under 600 m, while the HONO / NO2 in the sea scene was about 2 times larger than that in the inland scene above 600 m. All the above cases indicated that the HONO generation rate from NO2 heterogeneous reaction in the sea scene was larger than that in the inland scene in higher atmospheric layers above 400–600 m. The high-altitude (> 400–600 m) atmospheric parameters in the sea scene were more conductive to promote the HONO formation through the heterogeneous reaction of NO2. As shown in Fig. S9, the ratio of HONO / NO2 also generally increased with the increase in height above 0.2 km during the whole ship-based campaign. The greatest sensitivity under 1.5 km and the high DOF for aerosol, NO2, and HONO gave confidence in the retrieval results (Fig. S10).

The important factors and precursors to drive the formation of HONO through heterogeneous reactions had complex evolution and transport characteristics. To further clarify the role of these parameters in the heterogeneous process of NO2 to form HONO, three typical processes were selected to reveal the favorable conditions for HONO formation in the sea scene.