Variations in N_cn and N_ccn over marginal seas in China related to marine traffic emissions, new particle formation and aerosol aging

In this study, a cruise campaign was conducted over marginal seas in China to measure the concentrations of condensation nuclei (N_cn), cloud condensation nuclei (N_ccn) and other pollutants from day of year (DOY) 110 to DOY 135 of 2018. The ship self-emission signals were exhaustively excluded, and the mean values of N_ccn during the cruise campaign were found to slightly increase from $\mathrm{3.2}\pm \mathrm{1.1}\times {\mathrm{10}}^{\mathrm{3}}$ cm⁻³ (mean ± standard deviation) at supersaturation (SS) of 0.2 % to $\mathrm{3.9}\pm \mathrm{1.4}\times {\mathrm{10}}^{\mathrm{3}}$ cm⁻³ at SS of 1.0 %, and the mean value for N_cn was $\mathrm{8.1}\pm \mathrm{4.4}\times {\mathrm{10}}^{\mathrm{3}}$ cm⁻³. Data analysis showed that marine traffic emissions apparently largely contributed to the increase in N_cn in the daytime, especially in the marine atmospheres over heavily traveled sea zones; however, the fresh sources made no clear contribution to the increase in N_ccn. This finding was supported by the quantitative relations between N_cn and N_ccn at SS = 0.2 %–1.0 % against the mixing ratios of SO₂ in the ship self-emission plumes – i.e., a 1 ppb increase in SO₂ corresponded to a 1.4×10⁴ cm⁻³ increase in N_cn but only a 30–170 cm⁻³ increase in N_ccn, possibly because of abundant organics in the aerosols. Smooth growth can be observed in the marine-traffic-derived particles, reflecting aerosol aging. The estimated hygroscopicity parameter (κ) values were generally as high as 0.46–0.55 under the dominant onshore winds, suggesting that inorganic ammonium aerosols likely acted as the major contributor to N_ccn largely through aerosol aging processes of decomposing organics. Moreover, the influences of the new transported particles from the continent on the N_cn and N_ccn in the marine atmosphere were investigated.

1 Introduction

Oceans occupy approximately 2∕3 of the Earth's surface, and water evaporation from oceans is a major source of moisture in the atmosphere. Aerosol–cloud interactions in marine atmospheres, ranging from tropical to polar regions, have attracted great attention in the past few decades due to their impact on climate change (Huebert et al., 2003; Yu and Luo, 2009; Quinn and Bates, 2011; Wang et al., 2014; Brooks and Thornton, 2018; Rosenfeld et al., 2019). However, large uncertainties still exist in various marine atmospheres, e.g., the sources of aerosols, the concentrations of bulk cloud condensation nuclei (CCN) and aerosol CCN activation under various supersaturations. (Clarke et al., 2006; Decesari et al., 2011; Quinn and Bates, 2011; Saliba et al., 2019; Rosenfeld et al., 2019). These uncertainties are mainly determined by limited observations in marine atmospheres, although a few additional observations of the number concentrations of aerosols (N_cn) and CCN (N_ccn) were recently reported in different marine atmospheres, e.g., over the Mediterranean (Bougiatioti et al., 2009), Sea of Japan (Yamashita et al., 2011), Bay of Bengal (Ramana and Devi, 2016), coast of California (Ruehl et al., 2009) and northwest Pacific Ocean (Wang et al., 2019).

In addition to sea-spray aerosols and secondarily formed aerosols from sea-derived gaseous precursors (O'Dowd et al., 1997; Clarke et al., 2006; Quinn and Bates, 2011; Blot et al., 2013; Fossum et al., 2018), marine traffic emits large amounts of aerosols and reactive gases (Chen et al., 2017). These pollutants may directly or indirectly contribute to CCN to some extent (Langley et al., 2010). In addition, the long-range transport of continental aerosols has been widely reported to act as an important source of CCN in marine atmospheres (Charlson et al., 1987; Huebert et al., 2003; Fu et al., 2017; Royalty et al., 2017; Sato and Suzuki, 2019; Wang et al., 2019). The continent-derived aerosol particles observed in marine atmospheres usually mix with different sources, such as biomass burning, dust and anthropogenic emissions (Feng et al., 2017; Lin et al., 2015; Guo et al., 2014, 2016). An appreciable fraction of organics reportedly exists in marine aerosols and continental aerosols upwind of oceans (O'Dowd et al., 2004; Feng et al., 2012, 2016; Quinn et al., 2015; Song et al., 2018; Ding et al., 2019). However, ammonium sulfate aerosols have been frequently reported to dominantly contribute to CCN-related aerosols in many marine atmospheres and lead to hygroscopicity parameters (κ) larger than 0.5 (Mochida et al., 2010; Cai et al., 2017; Fu et al., 2017; Royalty et al., 2017; Phillips et al., 2018). A question is naturally raised, i.e., where do particulate organics go in the marine aerosols enriched in ammonium sulfate? Anthropogenic emissions in China such as SO₂ and NO_x have generally increased since the 1980s and recently started to decrease, i.e., SO₂ started to decrease in 2006 (Li et al., 2017), whereas NO_x started to decrease in 2011 (Li et al., 2017; Liu et al., 2016). Together with the influence of the Asian monsoon, the marginal seas near China are inevitably affected by the outflow of continental aerosols (Guo et al., 2016; Feng et al., 2017). Observations of N_cn and N_ccn in marine atmospheres over marginal seas near China help to address the data scarcity, understand the sources and dynamic changes in these parameters, and study their potential climate impacts.

In this study, cruise campaigns were conducted to measure the N_ccn, N_cn, particle number size distributions, gaseous pollutants and aerosol composition of water-soluble ionic species over the marginal seas from 20 April 2018 (day of year – DOY 110) to 15 May 2018 (DOY 135), traveling from the East China Sea to the South China Sea and returning to the Yellow Sea. Spatiotemporal variations in the N_cn, N_ccn and CCN activities of the aerosol particles were studied. The Kappa values of the aerosol particles from DOY 110 to DOY 118 over the marine environments were calculated and analyzed. Finally, we tried to establish relationship of N_cn and N_ccn with the mixing ratios of SO₂ in ship self-plumes and ambient marine air. The regression equations are valuable for the estimation of N_cn and N_ccn from SO₂ when the direct observations of N_cn and N_ccn are not available.

2 Experimental design

2.1 Instruments and data sources

A cruise campaign was conducted across marginal seas in China from DOY 110 to DOY 135 of 2018 (Fig. 1a, b). A suite of instruments including a fast mobility particle sizer (FMPS; TSI Model 3091), CCN counter (CCNC; DMT Model 100), condensation particle counter (CPC; TSI Model 3775), gas analyzers, ambient ion monitor-ion chromatography (AIM-IC), etc., were onboard the commercial cargo ship Anqiang 87 for measurements. The FMPS was used to measure the particle number size distributions with mobility diameters from 5.6 to 560 nm in 32 channels at 1 s temporal resolution with an inlet flow of 10 L min⁻¹. The CPC was used to report the N_cn ranging from 4 nm (50 % efficiency) to 3000 nm (N_cn) in 2 s time resolution with an inlet flow of 1.5 L min⁻¹. The N_cn was then used to calibrate the particle number size distributions simultaneously measured by the FMPS on the basis of the procedure proposed by Zimmerman et al. (2015). Due to the severe oceanic conditions and humid weather conditions, the FMPS and CPC were out of service after DOY 118 and DOY 122, respectively. Prior to the campaign, the CCNC was calibrated with ammonium sulfate particles based on the standard procedure detailed by Rose et al. (2008). The calibration curve is shown in Fig. S1 in the Supplement. The total flow rate of CCNC was 0.45 L min⁻¹, with a ratio of sample to sheath at 1∕10, and five supersaturations (SS) conditions were selected, including 0.2 %, 0.4 %, 0.6 %, 0.8 % and 1.0 %. More detailed information about the measurement of N_ccn can be found in Wang et al. (2019).

Figure 1 The ship track during the campaign of 2018, where the blue arrows represent the sailing direction, with the (a) southward track and (b) northward track.

During the experiment, ambient particles were first sampled through a conductive tube (TSI, US) and a diffusion dryer filled with silica gel (TSI, US) and then split for analysis by means of different instruments with a splitter. All instruments were placed in an air-conditioned container on the deck of the ship, with an inlet height of approximately 6 m above sea level. Regarding the gas analyzers, the ambient O₃ (Model 49i, Thermo Environmental Instruments Inc., USA C-series), SO₂ (Model 43i, Thermo Environmental Instruments Inc., USA C-series) and NO_x values (Model 42i, Thermo Environmental Instruments Inc., USA C-series) were measured in mixing ratios with a temporal resolution of 1 min. The CCNC and gas analyzers were operated properly throughout the entire campaign. The same was true for the AIM-IC, which was used to measure the water-soluble ionic species in the ambient particles sized smaller than 2.5 µm.

In this study, the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model from the NOAA Air Resources Laboratory was used to track the particle sources. The input of HYSPLIT, such as wind speed and wind direction, was obtained from the National Center for Environmental Prediction (NCEP) Global Data Assimilation System (GDAS) with a spatial resolution of 0.5^∘.

The hygroscopicity parameter (κ) was calculated according to the method proposed by Petters and Kreidenweis (2007).

$$\mathit{\kappa }={\displaystyle \frac{\mathrm{4}{A}^{\mathrm{3}}}{\mathrm{27}{D}_{\mathrm{d}}^{\mathrm{3}}{\ln }^{\mathrm{2}}{S}_{\mathrm{C}}}},A={\displaystyle \frac{\mathrm{4}{\mathit{\sigma }}_{\mathrm{s}/\mathrm{a}}{M}_{\mathrm{w}}}{RT{\mathit{\rho }}_{\mathrm{w}}}},$$

where D_d is the dry diameter, S_C is the supersaturation, M_w is the molecular weight of water, σ_s∕a represents the surface tension over the interface of the solution and air with the value of 0.072 J m⁻² applied in this study, R is the universal gas constant, T is the ambient temperature and ρ_w is the water density. D_d was not measured directly and was assumed to be equal to the critical diameter for CCN activation (D_crit).D_crit was defined as the particle diameter which is considered as the lower limit of the integral on particle number, together with the upper limit setting as the largest particle diameter, yielding the total integrated particle number concentration equal to the CCN concentration (Hung et al., 2014; Cheung et al., 2020). The FMPS had a low size resolution, particularly at sizes greater than 90 nm, which did not allow accurate calculation of the Kappa values at SS = 0.2 %. At SS = 0.6 % and 0.8 %, the Kappa value was not calculated considering the complication in the explanation of the value, possibly reflecting the combined effects of particle size, mixing state and chemical composition.

2.2 Separating ambient signals of N_cn and N_ccn from ship self-emissions

The data measured during the cruise campaign were frequently subject to interference from self-emission signals from the ship. The N_cn and N_ccn over the marginal seas were first distinguished based on the source of the ambient environment or the ship self-emissions. The data measured at 18:00–24:00 LT on DOY 115 are used to illustrate the separation in Fig. 2, and the size distribution of the particle number concentration during DOYs 110–118 is shown in Figs. S2–S10 in the Supplement. At 18:00–21:11 LT (local time), a low N_cn of $\mathrm{5.8}\pm \mathrm{0.4}\times {\mathrm{10}}^{\mathrm{3}}$ cm⁻³ was observed. The accumulation mode dominated the particle number concentration with a median mobility mode diameter of 105±4 nm (Fig. 2a). Afterwards, the N_cn rapidly increased by over 1 order of magnitude (Fig. 2b). The dominant particle number concentration mode changed from accumulation mode to Aitken mode, with the median mobility diameter of the Aitken mode stabilized at 47±4 nm for approximately 90 % of the time. The rapid increase in N_cn and the change in the mode size indicated the signal of the emissions of the ship itself. The ship self-emission interference after 21:11 was supported by additional evidence, e.g., a large decrease in the activation ratio (AR), defined as the quotient of N_ccn and N_cn, from >0.5 to <0.2 at SS = 0.4 % (Fig. 2c) due to a large increase in N_cn but a much smaller magnitude enhancement of N_ccn (Fig. 2b) and a rapid increase in NO_x from <10 ppb to 192±99 ppb, NO∕NO₂ from <0.1 to 0.7±0.3, and SO₂ from <2 ppb to 6.2±2.4 ppb. Large changes were expected because the ship smoke stack was only approximately 10 m away from these detectors. Thus, based upon the features described above, certain criteria were designed in this study to identify ship self-emission signals to separate them from ambient signals, i.e., a net increase in N_cn beyond 5×10⁴ cm⁻³ in 5 min, a median mobility mode diameter of approximately 50 nm, NO₂ > 30 ppb and NO∕NO₂ > 0.5.

Figure 2 (a) Contour plot of particle number size distribution with the median mobility mode diameter shown in black hollow circles; (b) time series of per minute N_cn and half-hourly N_ccn at SS = 0.4 %; (c) half-hourly AR values at SS = 0.4 %, (d) SO₂ and NO_x at nighttime on DOY 115.

3 Results and discussion

3.1 Spatiotemporal variations in ambient N_cn during the cruise period

Figure 3 shows the time series of minutely averaged distributions of N_cn, N_ccn and AR at SSs of 0.4 % and 1.0 % from DOY 110 to DOY 135, 2018, after the ship self-emission signals were exhaustively removed.

Figure 3 Time series of (a) per minute N_cn from DOY 110 to 122; (b) per minute N_ccn at SS of 0.4 % and 1.0 % during DOYs 110–135 and hourly ${\mathrm{NH}}_{\mathrm{4}}^{+}$ during DOYs 113–135; (c) per minute AR at SS of 0.4 % and 1.0 % during DOYs 110–122 and daily Kappa values at SS of 0.4 % and 1.0 % from DOYs 110 to 118 due to data availability. Please note that for (c), most Kappa values were based on a daily scale, except on DOY 110, during which two Kappa values were calculated from 00:00–06:00 and 08:00–21:00.

When the spatiotemporal variations in N_cn were examined during the first half of the cruise period (Fig. 3a), it was found that N_cn spanned a broad range of 0.2–4.5×10⁴ cm⁻³ with an average value of $\mathrm{8.1}\pm \mathrm{4.4}\times {\mathrm{10}}^{\mathrm{3}}$ cm⁻³. Specifically, N_cn was only $\mathrm{6.5}\pm \mathrm{0.8}\times {\mathrm{10}}^{\mathrm{3}}$ cm⁻³ at 00:00–06:00 LT on DOY 110 when the ship anchored at the Yangtze River estuary near Shanghai (Fig. 1). The low N_cn values were comparable to the mean value of N_cn (5.4×10³ cm⁻³) in the marine-air cases during January–December 2010 in Shanghai reported by Leng et al. (2013). The N_cn greatly increased to $\mathrm{1.9}\pm \mathrm{0.7}\times {\mathrm{10}}^{\mathrm{4}}$ cm⁻³ at 08:00–21:00 LT on DOY 110 when the ship cruised across the Yangtze River estuary. The measured particles in the number concentration were dominantly distributed in the Aitken mode on that day, while the median Aitken mode diameter shifted from 49±5 nm at 00:00–06:00 to 40±5 nm at 08:00–21:00 (Fig. S2). The Yangtze River estuary contains several world-class ports and is heavily traveled by marine traffic in the daytime (Chen et al., 2017). Since the onshore wind dominated on that day (not shown), the increase in N_cn and the decrease in the median Aitken mode diameter at 08:00–21:00 LT possibly reflected the increased contribution from marine traffic emissions. The marine traffic visibly decreased when the ship left the Yangtze River estuary toward the south. The N_cn value then significantly decreased, i.e., to $\mathrm{9.5}\pm \mathrm{4.4}\times {\mathrm{10}}^{\mathrm{3}}$ cm⁻³ in the marine atmosphere over the sea zone in Zhejiang Province (from 07:00 LT on DOY 111 to 17:00 LT on DOY 114), with P<0.01. The N_cn further decreased to $\mathrm{5.8}\pm \mathrm{1.7}\times {\mathrm{10}}^{\mathrm{3}}$ cm⁻³ in the marine atmosphere over the sea zone in Fujian Province (from 18:00 LT on DOY 114 to 14:00 LT on DOY 117). All these values were, however, 1–2 orders of magnitude greater than the background values in remote clear marine atmospheres, e.g., <300 particle cm⁻³ without the influence of industrial activities in the western Pacific and the tropical Pacific (Ueda et al., 2016) and those reported by Quinn and Bates (2011) and Saliba et al. (2019), indicating overwhelming contributions from non-sea-spray aerosols including marine traffic emissions, long-range continental transport, newly formed particles in marine atmospheres, etc. As reported, the atmospheric particles over marginal seas in China can be further transported to the remote northwest Pacific Ocean (NWPO) in spring under westerly winds – e.g., the N_cn observed over the NWPO in March–April 2014 was as high as $\mathrm{2.8}\pm \mathrm{1.0}\times {\mathrm{10}}^{\mathrm{3}}$ cm⁻³ and approximately half of that over marginal seas in China observed in March 2014 (Wang et al., 2019).

The mean value of N_cn ($\mathrm{8.1}\pm \mathrm{4.4}\times {\mathrm{10}}^{\mathrm{3}}$) observed in this study was close to that of $\mathrm{7.6}\pm \mathrm{4.0}\times {\mathrm{10}}^{\mathrm{3}}$ cm⁻³ (the number concentrations of particles larger than 10 nm) observed over the eastern part of the Yellow Sea in spring 2017 according by Park et al. (2018). The authors attributed the high number concentrations of particles within nucleation and Aitken modes to the long-range transport of air pollutants over eastern China under the influence of westerly winds. Consistently, larger values of N_cn were frequently observed in the continental atmospheres upwind of the Yellow Sea, e.g., mean values of $\mathrm{1.8}\pm \mathrm{1.4}\times {\mathrm{10}}^{\mathrm{4}}$ cm⁻³ in May 2013 in Qingdao, a coastal city in proximity to the Yellow Sea (Li et al., 2015), 3.18×10⁴ cm⁻³ in February–August 2014 in Beijing (Dal Maso et al., 2016) and 1.0×10⁴ cm⁻³ in continental-air cases during January–December 2010 in Shanghai (Leng et al., 2013).

3.2 Spatiotemporal variations in ambient N_ccn during the cruise period

N_ccn data were generally available during the entire campaign (Fig. 3b). The mean values of N_ccn over marginal seas in China during DOY 110 to DOY 135, 2018, ranged from $\mathrm{3.2}\pm \mathrm{1.1}\times {\mathrm{10}}^{\mathrm{3}}$ to $\mathrm{3.9}\pm \mathrm{1.4}\times {\mathrm{10}}^{\mathrm{3}}$ cm⁻³ under SSs ranging from 0.2 % to 1.0 % (Table 1), which is 2 to 4 times larger than the N_ccn at the same SS over the NWPO in March–April 2014 (Wang et al., 2019) and much higher, i.e., 1–2 orders of magnitude, than the pristine marine background values (Quinn and Bates, 2011). As discussed in the previous section, the mean N_cn in this study ($\mathrm{8.1}\pm \mathrm{4.4}\times {\mathrm{10}}^{\mathrm{3}}$ cm⁻³) was comparable to that of N_cn ($\mathrm{7.6}\pm \mathrm{4.0}\times {\mathrm{10}}^{\mathrm{3}}$ cm⁻³) over the Yellow Sea in spring 2017 in Park et al. (2018); however, the comparison of the mean N_ccn reveals that the mean value ($\mathrm{3.6}\pm \mathrm{1.2}\times {\mathrm{10}}^{\mathrm{3}}$ cm⁻³) at SS of 0.6 % in this study was approximately 25 % smaller than that (4.8×10³ cm⁻³ at a similar SS of 0.65 %) in Park et al. (2018), which was likely a result of long-range transport, considering the observations made a relatively long distance (i.e., 500–600 km) from the land depicted in Fig. 1 of Park et al. (2018) and the subsequently higher extent of aerosol aging. The N_ccn under SS of 0.2 % in this study ($\mathrm{3.2}\pm \mathrm{1.1}\times {\mathrm{10}}^{\mathrm{3}}$) is comparable to that ($\mathrm{3.1}\pm \mathrm{1.9}\times {\mathrm{10}}^{\mathrm{3}}$) of Li et al. (2015) in the continental atmosphere of Qingdao in May 2013; however, the increment of N_ccn with increasing SS was much weaker in our study, resulting in an average of 36 % smaller for N_ccn under SSs of 0.4 % to 1.0 % compared to that of Li et al. (2015). The sensitivity differences in N_ccn to SS between the relatively clean (i.e., N_cn of $\mathrm{8.1}\pm \mathrm{4.4}\times {\mathrm{10}}^{\mathrm{3}}$ in this study) and polluted (with N_cn of $\mathrm{1.8}\pm \mathrm{1.4}\times {\mathrm{10}}^{\mathrm{4}}$ cm⁻³) environments in Li et al. (2015) were also reported by Nair et al. (2020), which found little sensitivity in N_ccn to changes in SS over the equatorial Indian Ocean (<6^∘ N) with relatively clean air and much larger enhancement of N_ccn with increasing SS in polluted marine atmospheres (>6^∘ N).

Table 1 N_cn and N_ccn, AR and SO₂ mixing ratios on DOYs 110–135, 2018, over marginal seas in China. Please note that N_cn and AR are from DOYs 110–122, 2018.

In addition, the N_ccn at SSs from 0.1 % to 1.0 % during the period with high ${\mathrm{NH}}_{\mathrm{4}}^{+}$ (17:00 LT on DOY 114 to 10:00 LT on DOY 120) is statistically significantly higher (P<0.01) than that during the low-${\mathrm{NH}}_{\mathrm{4}}^{+}$ period (11:00 LT on DOY 120 to 07:00 LT on DOY 136; Fig. 3b). More specifically, a large increase in ${\mathrm{NH}}_{\mathrm{4}}^{+}$ concentration, with a mean concentration of 6.3±2.5 µg m⁻³, can be observed during the period from 17:00 LT on DOY 114 to 10:00 LT on DOY 120 (Fig. 3b). The mean N_ccn during this period varied from $\mathrm{3.5}\pm \mathrm{1.0}\times {\mathrm{10}}^{\mathrm{3}}$ to $\mathrm{4.0}\pm \mathrm{1.1}\times {\mathrm{10}}^{\mathrm{3}}$ cm⁻³ at SSs ranging from 0.2 % to 1.0 %. In contrast, after DOY 120, the concentration of ${\mathrm{NH}}_{\mathrm{4}}^{+}$ (0.67±0.70 µg m⁻³) substantially decreased by almost 90 %, during which the mean N_ccn at each SS showed statistically significant decreases of 8 % to 15 %, implicative of the vital contribution of secondary ammonium salt aerosols to CCN.

Another feature depicted in Fig. 3b is that the N_ccn during the low-${\mathrm{NH}}_{\mathrm{4}}^{+}$ period may even exceed the maximal value of N_ccn during the high-${\mathrm{NH}}_{\mathrm{4}}^{+}$ period. To elucidate the underlying mechanism, the N_ccn values under each SS were composited and compared for the days with ${\mathrm{NH}}_{\mathrm{4}}^{+}$ concentrations higher than the upper quartile and the days in the lower quartile, yielding some interesting findings. At SS = 0.2 %, the composited N_ccn under the high-${\mathrm{NH}}_{\mathrm{4}}^{+}$ period was higher than that during the low-${\mathrm{NH}}_{\mathrm{4}}^{+}$ period with a statistical significance level of 0.01. There was no significant difference between the N_ccn values of the two composite periods at SS values of 0.4 % and 0.6 %. However, the composited N_ccn (i.e., only selection of the upper quartile) during the high-${\mathrm{NH}}_{\mathrm{4}}^{+}$ period was significantly lower than the composited value during the low-${\mathrm{NH}}_{\mathrm{4}}^{+}$ period for P<0.01, e.g., $\mathrm{5.1}\pm \mathrm{0.5}\times {\mathrm{10}}^{\mathrm{3}}$ cm⁻³ versus $\mathrm{5.3}\pm \mathrm{0.7}\times {\mathrm{10}}^{\mathrm{3}}$ cm⁻³ at SS = 0.8 % and $\mathrm{5.2}\pm \mathrm{0.5}\times {\mathrm{10}}^{\mathrm{3}}$ cm⁻³ versus $\mathrm{5.7}\pm \mathrm{0.7}\times {\mathrm{10}}^{\mathrm{3}}$ cm⁻³ at SS = 1.0 %. During the low-${\mathrm{NH}}_{\mathrm{4}}^{+}$ period, the marine atmospheres over the observational zones may sometimes receive strong continental inputs and/or marine traffic emissions, leading to the larger N_ccn. The enhanced formation of ammonium salt aerosols during the high-${\mathrm{NH}}_{\mathrm{4}}^{+}$ period likely canceled out or even overwhelmed the effects of the continental inputs and/or marine traffic emissions on N_ccn at SS = 0.2 %.

In addition, fresh marine traffic emissions likely made a negligible contribution to N_ccn in the marine atmosphere because of the large amounts of aged aerosols from various sources therein. For example, the mean values of N_ccn were 3.2×10³ and 4.5×10³ cm⁻³ at SS = 0.4 % and 1.0 % at 08:30–11:30 on DOY 110, respectively. These values were almost the same as the 3.2×10³ cm⁻³ at SS = 0.4 % and 3.8×10³ cm⁻³ at SS = 1.0 % before 06:00 on that day. The mean values of N_cn, however, greatly increased from $\mathrm{6.5}\pm \mathrm{0.8}\times {\mathrm{10}}^{\mathrm{3}}$ cm⁻³ before 06:00 to $\mathrm{1.3}\pm \mathrm{0.3}\times {\mathrm{10}}^{\mathrm{4}}$ cm⁻³ at 08:30–11:30 when the ship cruised across the Yangtze River estuary (Fig. 3b).

3.3 Spatiotemporal variations in CCN activation and Kappa values

The AR values at SSs of 0.4 % and 1.0 % are examined in this section, as shown in Fig. 3c. At SS = 0.4 %, the AR values largely varied from 0.06 to 0.92 with a median value of 0.51. Specifically, the AR values narrowly varied around 0.51±0.04 at 00:00–06:00 LT on DOY 110. At 08:00–21:00 LT on that day, when the ship cruised across the Yangtze River estuary, the AR values substantially decreased to 0.26±0.06 concurrently with an approximate 200 % increase in the N_cn values, i.e., N_cn values of $\mathrm{6.5}\pm \mathrm{0.8}\times {\mathrm{10}}^{\mathrm{3}}$ cm⁻³ at 00:00–06:00 LT and $\mathrm{2.0}\pm \mathrm{0.7}\times {\mathrm{10}}^{\mathrm{4}}$ cm⁻³ at 08:00–21:00 LT on DOY 110 (Fig. 3a). The AR values then exhibited an oscillating increase from DOY 111 to DOY 113. A low AR value of 0.12±0.04 was suddenly observed at 10:00–18:00 LT on DOY 114 in the presence of strong new particle signals transported from the upwind continental atmosphere, as discussed later. The AR values, however, reached 0.34±0.04 at 06:00–08:00 LT and 0.39±0.08 at 19:00–24:00 LT on DOY 114, with the new particle signals largely decreased. Even excluding the AR values on DOY 114, a significant difference was still obtained between the AR values of 0.61±0.12 during the high-${\mathrm{NH}}_{\mathrm{4}}^{+}$ period and those of 0.55±0.17 during the low-${\mathrm{NH}}_{\mathrm{4}}^{+}$ period. The enhanced formation of ammonium salts seemingly increased the CCN activity to some extent. At SS = 1.0 %, the AR values showed large fluctuations with a median value of 0.57±0.17 (Fig. 3c), and the temporal trend was similar to that at SS = 0.4 %.

To minimize the impact from the particle sizes, the Kappa values were further investigated. As reported by Phillips et al. (2018), Kappa values at a high time resolution usually exhibit a broad distribution, reflecting the complexity due to various factors. To reveal the key factors in determining the Kappa values on a large spatiotemporal scale, the daily Kappa values of atmospheric aerosols were estimated on the basis of the daily mean N_ccn and the size distributions of the particle number concentration from DOYs 110–118 (Fig. 3c). Please note that for DOY 110, considering the large differences in the particle number concentration between 00:00–06:00 and 08:00–21:00 (Fig. S2), the Kappa values were calculated separately for these two periods. At SS = 0.4 % (green dashed line in Fig. 3c), the estimated Kappa values were as high as 0.66 at 00:00–06:00 LT, while they decreased to 0.37 at 08:00–21:00 LT on DOY 110. The Kappa value varied narrowly from 0.46 to 0.55 on DOYs 111–113, 115 and 117, implying that inorganic aerosols such as completely and incompletely neutralized ammonium salts may make large contributions to the N_ccn. These values were generally consistent with the reported observations in most marine atmospheres. For example, Cai et al. (2017) reported a Kappa value of approximately 0.5 for particles with sizes of 40–200 nm at a marine site in Okinawa and that sulfate was the dominant component of aerosol particles on 1–9 November 2015, and a similar Kappa value in spring 2008 was reported by Mochida et al. (2010) over this site. Royalty et al. (2017) reported Kappa values for 48, 96 and 144 nm dry particles of 0.57±0.12, 0.51±0.09 and 0.52±0.08 in the subtropical North Pacific Ocean and sulfate-like particles contributing at most 77 %–88 % to the total aerosol number concentration. Over the Atlantic Ocean, Kappa values of approximately 0.54±0.03 were observed for 284 nm particles (Phillips et al., 2018).

The estimated Kappa values sometimes reached 0.66–0.67 (i.e., on DOY 116), which may be related to unidentified factors. For example, O'Dowd et al. (2014) proposed that some organics derived from sea-spray aerosols may also increase the N_ccn to some extent by reducing the surface tension, leading to an increase in the Kappa values. A small fraction of sea-salt aerosols in submicron particles may also increase the Kappa values since its Kappa value was as high as 1.3 (O'Dowd et al., 1997, 2004). A Kappa value of 0.29 was obtained on DOY 118, which is close to the Kappa values widely observed for continental atmospheric aerosols (∼0.3) (Andreae and Rosenfeld, 2008; Pöschl et al., 2009; Rose et al., 2010). The estimated Kappa value largely decreased to 0.15 on DOY 114 when new particle formation (NPF) occurred; see Sect. 3.5 for detailed discussion. Moreover, at an SS of 1.0 %, the estimated Kappa value was always smaller than 0.2. The Kappa values of organics were commonly assumed to be 0.1 (Rose et al., 2011; Cai et al., 2017; Singla et al., 2017). In general, the fraction of organics in the nanometer particles increases with decreasing particle size from ∼100 to ∼50 nm (Rose et al., 2010, 2011; Crippa et al., 2014; Cai et al., 2017). A combination of the two factors likely led the overall Kappa values estimated at SS = 1.0 % to be much lower. However, direct measurements of the chemical composition of nanometer particles are needed to confirm these arguments.

3.4 Particle number size distributions and CCN activation associated with marine traffic emissions and aerosol aging

The particle number size distributions during DOYs 110–118, shown in Fig. 4, can be generally classified into two categories. Category 1 occurred on DOYs 110–114, when particle number concentrations were mainly distributed in Aitken mode, whereas the accumulation mode was generally undetectable. Category 2 occurred on DOYs 115–118, when the accumulation mode could be clearly identified and generally dominated over the Aitken mode. Hoppel (1986) proposed that cloud-modified aerosols are mainly distributed at 80–150 nm in the remote tropical Atlantic and Pacific oceans. Cloud-modified aerosols are quite common in remote marine atmospheres, likely leading to the dominant accumulation mode particles being observed on DOYs 115–118. Occasionally, the Aitken mode dominated over the accumulation mode, such as on DOY 118. To further investigate the sources of different modes of particles, the days DOY 112, DOY 116 and DOY 118 were selected.

Figure 4 Contour plot of particle number size distribution on DOYs 110–118 with ship self-emission signals removed.

On DOY 112, the Aitken mode particles accounted for approximately 60 % of the total particle number concentration (Fig. 5a), with median Aitken mode diameters of approximately 54±8 nm. Similar to the observations over the Yangtze River estuary, the mean value of N_cn increased by approximately 50 % concurrently with a decrease in the median Aitken mode diameters by ∼9 nm at 05:30–11:40 LT compared to those in the early morning before 05:30 LT (Fig. 5b). Concomitantly, the AR values decreased to 0.31±0.09 at SS of 0.4 %, with similar AR decreases at SS of 1.0 %, and the lowest AR and Kappa values occurred at 06:00–07:00 LT at SSs of both 0.4 % and 1.0 %. All these results indicated that the increase in Aitken mode particles at 05:30–11:40 LT was likely derived from enhanced marine traffic contributions carried by the onshore wind from the south (Fig. S11). During other times on DOY 112, the onshore wind may also carry marine-traffic-derived particles to the observational sea zones. However, the marine-traffic-derived particles likely aged to some extent, e.g., the median Aitken mode diameters exhibited an oscillating increase from approximately 50 nm at 19:00 to approximately 70 nm at 24:00 LT with a particle growth rate of ∼4 nm h⁻¹. The AR values, however, narrowly varied around 0.47±0.03 at SS = 0.4 % and 0.52±0.05 at SS = 1.0 % during the particle growth period. The Kappa values at SS = 0.4 % gradually decreased from 0.56 at 19:00 to 0.41 at 23:00 LT, reflecting more aged marine-traffic-derived particles growing into CCN sizes.

Figure 5 (a, d, g) Daily average and (b, e, h)contour plot of particle number size distributions and (c, f, i) time series of hourly averaged AR at SS of 0.4 % and 1.0 % and Kappa value on DOY 112, DOY 116 and DOY 118. The bars represent the standard deviation, with the mean indicated by the hollow circles.

On DOY 116, the accumulation mode particles dominantly contributed to N_cn rather than Aitken mode particles (Fig. 5d) under the marine air influence from the northeast (Fig. S13). The median accumulation mode diameters narrowly varied around 135±5 nm at 01:00–13:00 LT and 102±5 nm at 16:20–24:00 LT with a transition period in between (Fig. 5e). The AR and Kappa values, however, showed no statistically significant differences during the two periods at SSs of 0.4 % and 1.0 %, implying that the size change in the accumulation mode particles had a negligible influence on CCN activation. The hourly variations in the AR and Kappa values may be associated with other factors, e.g., chemical composition and mixing state (Gunthe et al., 2011; Rose et al., 2011).

On DOY 118, under the influence of mixtures from marine and coastal areas from the northeast (Fig. S14), the accumulation mode particles generally dominated the contribution to N_cn, while the reverse was true on some occasions (Fig. 5g, h). The median accumulation mode diameters exhibited an oscillating increase from approximately 100 to 130 nm at 00:00–08:00 LT, narrowly varied around 133±5 nm at 08:00–13:00 LT and then exhibited an oscillating decrease down to approximately 100 nm at 20:00 LT. The AR values and Kappa values at SS = 0.4 %, however, exhibited an inverted bell shape with the lowest values at 0.31 and 0.11 at 13:00. The decreases in the AR values and Kappa may be related to organic condensation on the accumulation mode particles since the median accumulation mode diameters were almost the largest at 13:00. The number concentration of Aitken mode particles was evidently enhanced at 14:00–15:00, but the influence on the AR values and Kappa values at SS = 0.4 % was undetectable (Fig. 5i).

3.5 The long-range transport of newly grown particles on DOY 114

No hour-long sharp increases were observed in the number concentration of the nucleation mode particles (<20 nm) during the period from DOY 110 to DOY 118, except on DOY 114 (Fig. 4). According to the conventional definition of NPF events (Kulmala et al., 2004; Dal Maso et al., 2005), the occurrence frequency of NPF events was low in this study. Unlike continental atmospheres where a high occurrence frequency of NPF events has been observed globally in spring (Kulmala et al., 2004; Kerminen et al., 2018), a low occurrence frequency reportedly occurred over the seas during the “Meiyu (plum-rain) season” in spring because of frequent rainy, foggy or cloudy weather conditions (Zhu et al., 2019). The lack of NPF events in the marine atmospheres implied that the contributions to N_cn and N_ccn were mainly from primary emitted aerosols and their aged products.

During the period of 10:00–18:00 LT on DOY 114, the large increase in the number concentrations of Aitken mode particles (Fig. 6a) likely reflected the long-range transport of newly grown particles from upwind continental atmospheres (Fig. S12). The size distributions of the particle number concentration showed a dominant Aitken mode at 10:00–18:00 LT, when the spatiotemporal variations in N_cn and median Aitken mode diameters exhibited bell-shaped patterns (Fig. 6b). The median Aitken mode diameters increased from 26 nm at 10:00 LT to 33 nm at 12:00–13:00 LT and then decreased to 20 nm prior to the signal disappearance, likely reflecting the growth and shrinkage of the Aitken mode particles (Yao et al., 2010; Zhu et al., 2019). The median Aitken mode diameters were evidently smaller than the values, i.e., 40–50 nm for the Aitken mode particles, observed over the Yangtze River estuary on DOY 112 (Fig. 5a). Moreover, the number concentrations of the 20–40 nm particles increased by 5.8 times at 12:00–13:00 LT compared to the mean value at 06:00–09:00 LT, while the total number concentrations of particles greater than 90 nm increased by only 67 %. These results implied that the large increases in the number concentrations of Aitken mode particles with a dynamic change in the mode diameter observed at 10:00–18:00 LT were not likely caused by primary emitted and aged particles from marine traffic emissions or other combustion sources. The observations of the gaseous and particulate species during the same period implied that the air masses were well aged and less polluted. For instance, the measured hourly average mixing ratios of SO₂ were no larger than 1.2 ppb (Fig. 6c), and the hourly average concentrations of ${\mathrm{NH}}_{\mathrm{4}}^{+}$ in PM_2.5 were smaller than 2 µg m⁻³ (Fig. 3b). In addition, the concentrations of K⁺ were below 0.3 µg m⁻³, suggesting negligible contributions from biomass burning (Fig. 6e).

Figure 6 (a) Contour plot of particle number size distributions for the DOY 114 2018; (b) the size distributions of the particle number concentration during 10:00–18:00 LT DOY 114, 2018; (c) time series of hourly averaged SO₂ and Kappa values at SS of 0.4 % and 1.0 %; (d) N_ccn at SS of 0.4 % and 1.0 %; and (e) AR values at SS of 0.4 % and 1.0 % and K⁺ for DOY 114, 2018.

Before 09:00 LT, a much weaker spike of nucleation mode particles was intermittently observed (Fig. 6a). The weak and intermittent NPF seemed to occur in the marine atmospheres before 09:00 LT when no apparent growth of new particles was observed. Possibly due to transport from the continent (Fig. S12) and an increase in the condensational sink at approximately 10:00 a.m. (Fig. 6a), the weak NPF signal gradually dropped to a negligible level 0.5 h later, concomitant with a large increase in the number concentrations of Aitken mode particles at 10:00–18:00 LT.

N_ccn at SS = 0.4 % increased from 1.2×10³ cm⁻³ at 06:00–09:00 LT to the peak value of 2.3×10³ cm⁻³ at 12:00 LT, with an increase of 92 %, and N_ccn at SS = 1.0 % increased from 1.6×10³ to 4.0×10³ cm⁻³, with an increase of 150 % (Fig. 6d). The net increase in N_ccn at SS = 0.4 % likely reflected the contribution from pre-existing particles since new particles with diameters less than 50 nm were unlikely to be activated as CCN at such low SS (Li et al., 2015; Wu et al., 2016; Ma et al., 2016). The larger net increase in N_ccn at SS = 1.0 % may reflect the contributions from pre-existing particles and newly grown particles. The high SS can activate particles as CCN with diameters down to 40 nm (Dusek et al., 2006; Li et al., 2015). The invasion of newly grown particles also led to the large decreases in the AR values from 0.3 to 0.1 at SS = 0.4 % and from 0.4 to 0.2 at SS = 1.0 % (Fig. 6e). After 18:00 LT, the AR values retuned to 0.3–0.4 at SS = 0.4 % and 0.4–0.6 at SS = 1.0 %. The calculated Kappa values were examined (Fig. 6c) and were found to decrease from 0.4 to 0.1–0.2 at SS = 0.4 %. This value returned to 0.3 at 18:00–19:00 LT (FMPS temporarily malfunctioned after 19:20 LT). The Kappa values were below 0.2 at SS = 1.0 % on that day. The decreases in the AR values and Kappa values at the two SS were likely caused by organic vapor condensed on pre-existing particles and new particles (Wu et al., 2016; Zhu et al., 2019).

3.6 Relationship of N_cn and N_ccn with SO₂ in ship self-plumes and ambient air

When ship self-emission signals were detected, the observational values included a combination of contributions from ship self-emissions and ambient concentrations. Although the ambient N_cn was negligible in comparison with the N_cn derived from the ship self-emissions, this was not the case for N_ccn and SO₂. Based on the per minute data, the signal was considered to be vessel self-emissions when both N_cn was greater than 50 000 cm⁻³ and SO₂ was greater than 5 ppb. The composited data were then used to derive the hourly average N_cn, N_ccn and SO₂, which were then subtracted by the ambient hourly mean values during the preceding hour with relatively clean conditions (i.e., concentrations of N_cn lower than 10 000 cm⁻³ and SO₂ lower than 2.5 ppb). Please note that uncertainties exist in terms of the criteria and separation between the ship self-signals and ambient signals; however, minimal impact is expected in the relationship examined below.

Figure 7a shows the relationship of N_cn and N_ccn with the mixing ratio of SO₂ in the ship self-plumes, prefixed by Δ for N_cn, N_ccn and SO₂ to implicate the removal of ambient signals. A good correlation of 0.66 for R² (P<0.01) is obtained, and the slope indicates that the increase in N_cn by 1.4×10⁴ cm⁻³ for each 1 ppb increase in SO₂ resulted from ship emissions (Fig. 7a). High emissions of N_cn were generally reported in engine exhausts where high-sulfur-content diesel was used (Yao et al., 2005, 2007). The N_ccn at SS of 0.2 % to 1.0 % (Fig. 7b) increased from 30 to 170 cm⁻³ per 1 ppb increase in SO₂, showing a statistically significant correlation at the 99th confidence level. The contribution ratios of SO₂ to N_ccn were 0.002 (SS of 0.2 %), 0.004 (SS of 0.4 %) and 0.012 (SS of 1.0 %) to that of N_cn, which is in general consistent with a previous study by Ramana and Devi (2016), in which a range of 0.0012–0.57 was observed for CCN at 0.4 % in Bay of Bengal during 13–16 August 2012.

Figure 7 Relationship of hourly averaged N_cn and N_ccn with SO₂ at SS of 0.2 %, 0.4 % and 1.0 %. For (a) and (b), N_cn, N_ccn and SO₂ reflect the impact of the ship self-emission after the removal of the ambient concentration. For (c) and (d), each bar indicates the standard deviation with the mean value marked as hollow circles (or triangles, squares), and the interval of SO₂ is 0.2 ppb for each bar.

The relationship of hourly averaged N_cn and N_ccn with SO₂ in ambient air were examined and are shown in Fig. 7c, d. The data were segmented into pieces based on SO₂ with an interval of 0.2 ppb. A good correlation was obtained between the averaged N_cn and SO₂ with an R² of 0.80 (P<0.01), and a 1 ppb increase in SO₂ likely increased N_cn by 1.6×10³ cm⁻³ (Fig. 7c). The increase in N_cn with SO₂ may reflect the contribution from primary emissions. The intercept was, however, as large as 3.9×10³ cm⁻³, likely representing the contribution from well-aged aerosols.

The hourly averaged N_ccn at different SSs generally increased with increasing ambient SO₂ (Fig. 7d). A good correlation was obtained between the averaged N_ccn and SO₂, with R² = 0.78–0.91 (P<0.01). A 1 ppb increase in SO₂ likely increased N_ccn by 0.6×10³ to 0.8×10³ cm⁻³ at SSs from 0.2 % to 1.0 %. The increase in N_ccn with SO₂ may also reflect the contribution from primary emissions. The intercepts of 2.2×10³–2.7×10³ cm⁻³ at different SSs were likely contributed by well-aged aerosols. This relationship may be used as an estimation of the N_ccn in marine atmospheres over marginal seas in China when no measurements of CCN are available, whereas the ambient SO₂ can be estimated from web-based satellite data.

4 Conclusions

The spatiotemporal variations in ambient N_cn and N_ccn were studied during a cruise campaign on DOYs 110–135 over marginal seas in China. The mean values of N_cn (8.1×10³ cm⁻³) and N_ccn (3.2–3.9×10³ cm⁻³) at SSs of 0.2 %–1.0 % were approximately 1 order of magnitude larger than those in remote clear marine atmospheres, indicating overwhelming contributions from non-sea-spray aerosols such as marine traffic emissions, long-range continental transport and others.

The observed ship self-emission signals showed that fresh marine traffic emissions can be important sources of N_cn but minor sources of N_ccn in the marine atmosphere. The signals showed that a 1 ppb increase in SO₂ corresponded to a 1.4×10⁴ cm⁻³ increase in N_cn and a 30–170 cm⁻³ increase in N_ccn at SS = 0.2 %–1.0 %. Data analysis showed that marine traffic emissions largely increased N_cn over heavily traveled sea zones in the daytime.

In ambient marine air, the growth of marine-traffic-derived particles led to a decrease in the estimated bulk Kappa values at 0.4 %, possibly because some of these particles enriched in organics grew into CCN size. However, strong formation of ammonium salts led to aerosol aging and significantly increased N_ccn at SS of 0.2 %–1.0 % in comparison with those observed during the period poor in ammonium salt aerosols in PM_2.5 with P<0.01. The estimated bulk Kappa values from the daily average values varied from 0.46 to 0.55 at SS = 0.4 % in most marine atmospheres, indicating that inorganic ammonium aerosols may dominantly contribute to the N_ccn at SS of 0.4 %. The particle number size distributions showed that the high bulk Kappa values could be related to cloud-modified aerosols, which likely led to a large extent of degradation of organics and subsequent loss from the particle phase.

NPF events rarely occurred in the humid ambient marine air. The dominant onshore winds occurred during most of the measurement periods and likely carried primary aerosols and their aged products rather than secondarily formed aerosols to the observational zone. During an occasion when offshore winds blew from the northwest (Fig. S12), new particle signals transported from the continent can be clearly observed. However, the NPF in the marine atmosphere was too weak to be important. The new transported particles from the continent yielded the maximal increases in N_ccn of 92 % at SS of 0.4 % and 150 % at SS of 1.0 %. However, consistent with those reported in the literature, the estimated Kappa values largely decreased from 0.4 to 0.1–0.2 at SS of 0.4 % during most of the continent-transporting NPF event because the Kappa value of the organic condensation vapor was as low as 0.1.