Campaign A was conducted from 9 to 19 December 2019 on the R/V Dongfanghong-3 with a displacement tonnage of 5000. The research vessel was still within its testing period and used state-of-the-art combustion technology with low-sulfur diesel. Campaign B started from 27 December 2019 to 17 January 2020 and was organized by another research team. During 20–22 December, the vessel was anchored at the port while the sampling continued. The 44 h were referred to as the transition period between campaigns A and B. A standard-sized air-conditioned container was set up on the front deck to house a suite of instruments including the AIM-IC, a fast-mobility particle sizer (FMPS, Tsi), a cloud condensation nuclei counter (CCN-100, Droplet MT), and a single-particle aerosol mass spectrometer (SPAMS 05, Hexin). No human activities occurred on the front deck during cruising, excluding anchoring at the port. Even during the anchoring period, human activity on the front deck was rare. The use of the container on the front deck effectively minimized the self-vessel contamination by NH3gas and gaseous amines. The front deck was approximately 10 m a.s.l., and the container height was 2.8 m.

To ensure that the onboard AIM-IC was operated properly, it was housed in a mobile air-conditioned mini-container, which was further placed in a standard container with a 1 m stainless steel sampling probe connected to the ambient air. The inlet of the sampling probe extended from the top corner of the standard container facing the sea. The AIM-IC consists of two major parts: an ambient air sampling system and an ion chromatography analysis system. For the sampling system, the AIM-IC was equipped with a PM2.5 cyclone and operated at a rate of 3 Lmin-1. The sampled gases and particles in the water solution were stored in two syringes prior to their injection for analysis. The ion chromatography analysis system measured the semi-continuous concentrations of chemically reactive gases. These included NH3gas, gaseous amines, and acidic gases such as SO2 and HNO3, along with their particulate counterparts, at a temporal resolution of 1 h. This facilitated the identification of possible interference from onboard dew evaporation, which typically occurs with sunrise (Teng et al., 2017).

An automatic weather system providing real-time meteorological data is available on the R/V Dongfanghong-3. The heading wind was corrected to determine the true wind speed and direction. The surface seawater temperature was not measured during this cruise campaign and typically had a delay of a few hours compared to the ambient air temperature (Deng et al., 2014).

On 1–9 August, 12 September to 1 October, and 16 November to 1 December 2019, the AIM-IC was set up at a coastal site in Qingdao (36.34∘ N, 120.67∘ E) to conduct routine measurements (Fig. 1). Coastal measurement data were obtained from 2 weeks to 4 months before the winter cruise campaign. The sampling site was located in a new high-technology zone near the Yellow Sea, with the shortest distance from the sea being approximately 1 km in the south. The AIM-IC was housed in a research lab on the fifth floor of a building, approximately 16 m a.g.l. The sampling probe extended out of the window and was directly connected to the ambient air. Typically, higher biogenic emissions of reduced nitrogen compounds over the continents are expected in the summer than in the winter owing to the temperature effect (Yu et al., 2016; Teng et al., 2017).

The AIM-IC includes an ICS-1100 ion chromatograph, wherein an analytical column (Ion Pac CS17A – 2×250 mm) was used to measure cations, including Na+, NH4+, protonated dimethylamine (DMAH+), and protonated trimethylamine (TMAH+), and an AS11-HC (2×50 mm) was used to measure anions, including SO42-, NO3-, Cl-, and organic ions. Methanesulfonic acid solution (5 mM) was used as the eluent for cation analysis, while potassium hydroxide solution (varying from 3 to 40 mM) was used as the gradient eluent for anion analysis. Each analysis took 26–28 min to obtain a complete ion spectrum. The volume of the injection loop installed on the low-pressure valve was 250 µL, which substantially reduced the limits of detection for all ions. The limits of detection for NH4+, DMAH+, and TMAH+ were 0.0004, 0.004, and 0.002 µgm-3, respectively, in ambient air. The limits of detection for NO3- and SO42- were 0.05 and 0.015 µgm-3, respectively, in ambient air. The ICS-1100 was calibrated onboard prior to obtaining regular measurements, and the second calibration was conducted when the vessel was anchored at the port. The AIM-IC analysis was not affected by ambient water vapor, as the device directly measured the ions. Detailed information regarding the AIM-IC analysis is provided in Teng et al. (2017) and Xie et al. (2018). Notably, strong K+ interference occurred unexpectedly and occasionally and then disappeared during different campaigns. When the interference occurred, DMAH+ and TMAH+ were undetectable because of the increased baseline at the corresponding residence time in the ion chromatograph (Fig. S1 in the Supplement); consequently, some PM2.5 DMAH+ and TMAH+ concentration data are unavailable in Fig. 1. However, the concentrations of gaseous amines were still detected correctly, with a low baseline at the residence. K+ interference remains to be investigated. Additionally, a few surface seawater samples were collected from different sea zones. The NH4+ and aminium ion concentrations in the samples were not measured, as the analytical methods were still hindered by high sea-salt ion contents.

Before analyzing the basic gases and their counterparts in the marine atmosphere, we initially presented their continental concentrations at the coastal site facing the Yellow Sea as important evidence to facilitate the analysis of the contributors to these species in the marine atmosphere. Figure 1a and b show that the TMAgas and TMAH+ concentrations in PM2.5 were mostly below the detection limit, varying at approximately 0.001±0.001µgm-3 (average ± standard deviation), regardless of the presence of offshore or onshore winds during short-term measurements in the three seasons of 2019. The DMAgas and DMAH+ concentrations varied at 0.018±0.021 and 0.017±0.013µgm-3, respectively, which were approximately 1 order of magnitude larger than those of TMAgas and TMAH+. TMAgas and TMAH+ concentrations in the upwind continental and coastal atmospheres were substantially lower than those reported in the literature, by up to a few tens of ngm-3 (Ge et al., 2011). However, Gibb et al. (1999) reported a low average TMAgas (0.5 ngm-3) and particulate TMAH+ (0.5 ngm-3) in the marine atmosphere over the Arabian Sea on 16 November to 19 December 1994. Xie et al. (2018) reported that TMAH+ concentrations were comparable to those of DMAH+ in atmospheric particles collected at two other coastal sites located approximately 20 km from the study area, as listed in Table S1 in the Supplement. The cause of this change is beyond the scope of this study but may be due to the large decrease in manure application, based on our recent survey in the Qingdao area.

The DMAgas and DMAH+ concentrations in PM2.5 with offshore winds from the north were substantially higher than those with onshore winds from the south or southeast (top of Fig. 1a), suggesting that their continental emissions and related secondary sources were stronger. Moreover, the concentrations of DMAgas and DMAH+ were moderately correlated with those of NH3gas and NH4+, namely, [DMAgas]=5.6×10-3×[NH3gas] (R2=0.79, P<0.01) and [DMAH+]PM2.5=5.9×10-3×[NH4+]PM2.5 (R2=0.84, P<0.01). Generally, the DMAgas and DMAH+ concentrations were approximately 1/200 of those of the corresponding NH3gas and NH4+.

Throughout Campaign A, the TMAgas concentrations varied at approximately 0.031±0.009 µgm-3 (Fig. 2a–c), with three peaks occurring at 4–5 d intervals (gray shadowing in Fig. 2c). Peaks 1 and 2 were generally associated with offshore winds, while peak 3 was mostly associated with onshore winds (Fig. 2b). The peaks lasted from tens to dozens of hours and were not induced by the onboard dew evaporation at sunrise. For example, the highest value (0.060 µgm-3) occurred at 23:00 LT (local time; UTC+08:00) on 16 December. The observed TMAgas concentrations were 1 order of magnitude higher than those measured in the coastal atmosphere during the summer, fall, and winter. This suggested that the TMAgas observed during Campaign A was largely derived from marine sources rather than from long-range continental transport. The same conclusion can be drawn by analyzing the three peaks of TMAgas and its temporal variations during the port-anchoring period. For example, during peak 1 (Fig. 2a), the concentrations of TMAgas increased by approximately 100 % from 20:00 LT on 9 December to 11:00 LT on 10 December, with an approximately 30 % decrease in the non-sea-salt SO42- (nss-SO42-) concentration (from 22 to 16 µgm-3; Fig. 2b). Moreover, the peaks in the TMAgas concentrations corresponded to troughs in the nss-SO42- concentrations during peak 3, as shown in Fig. 2c and d. The self-vessel emissions of nss-SO42- in PM2.5 were negligible because of the use of low-sulfur diesel, which is discussed later. The increased nss-SO42- concentrations in PM2.5 may be a good indicator of continental transport and vice versa.

The concentrations of DMAgas varied at approximately 0.006±0.006µgm-3 (Fig. 2d) and were significantly higher than those of TMAgas (P<0.01). Unlike TMAgas, continental transport was likely an important contributor to the DMAgas and NH3gas observed in the marine atmosphere, particularly during peak 1, when higher nss-SO42- concentrations were observed in PM2.5 (Fig. 2c–e). The DMAgas and NH3gas concentrations were negatively correlated with those of TMAgas during peak 1, namely, R2=0.35 (P<0.01) between TMAgas and DMAgas and R2=0.17 (P<0.01) between TMAgas and NH3gas. This suggested that most of the DMAgas and NH3gas were likely derived from continental transport rather than marine sources. During peak 2, increased TMAgas, DMAgas, and NH3gas concentrations were concurrently observed with increasing nss-SO42- concentrations, suggesting that both marine emissions and continental transport simultaneously contributed to the observed DMAgas and NH3gas. During the port-anchoring period from 20 to 22 December, the DMAgas and NH3gas concentrations varied slightly and were moderate and low, respectively. However, the TMAgas concentrations continuously increased by over 100 % as the ambient temperature increased (Fig. 2c and f). Additionally, the nss-SO42- concentrations of PM2.5 varied greatly and followed a bell-shaped pattern during the port-anchoring period.

Additionally, the NH3gas concentrations varied at approximately 0.53±0.53 µgm-3 from 9 to 22 December. The variation narrowed to approximately 0.24±0.07 µgm-3 during the port-anchoring period from 20 to 22 December. When the data during Campaign A were used for the analysis, the NH3gas concentrations were significantly correlated with those of DMAgas, namely, [DMAgas]=9.2×10-3×[NH3gas] (R2=0.71, P<0.01). However, there was no correlation between NH3gas and TMAgas concentrations.

Figure 3a–f show the spatiotemporal variations in the TMAH+, DMAH+, and NH4+ concentrations of PM2.5 throughout Campaign A from 9 to 22 December, during which the TMAH+ concentrations varied greatly at approximately 0.28±0.18 µgm-3. However, they narrowed at approximately 0.21±0.04 µgm-3 during the port-anchoring period. The TMAH+ concentrations generally increased from 0.13±0.05 µgm-3 on 9 December to 0.46±0.05 µgm-3 on 16 December (Fig. 3a) and subsequently decreased to approximately 0.2 µgm-3 thereafter, excluding some strong peaks from 0.62 to 1.24 µgm-3 at 03:00–05:59 LT and from 1.02 to 1.81 µgm-3 at 14:00–16:59 LT on 18 December (gray shadowing representing peak 4 in Fig. 3a–d). The peaks reproduced the episodes observed in the marine atmosphere over the Yellow Sea in May 2012 (Hu et al., 2015) and were repeatedly observed during Campaign B (Gao et al., 2021). However, they were not observed in several other marine cruise campaigns conducted across the marginal seas of China and the northwestern Pacific Ocean (Hu et al., 2018; Xie et al., 2018).

As the TMAH+ concentrations were approximately 2 orders of magnitude higher than those observed at the coastal site during the three seasons of 2019, the observed TMAH+ was largely derived from marine sources. The TMAH+ concentrations followed a spatiotemporal pattern that clearly differed from those of DMAH+ and NH4+, while the latter two ions exhibited a similar spatiotemporal pattern during most periods of Campaign A (Fig. 3a–c). A significant negative correlation (P<0.01) was observed between the concentrations of TMAH+ and NH4+ in PM2.5 (not shown). The spatiotemporal pattern of the TMAH+ concentration also significantly differed from those of nss-SO42- (Fig. 2d) and SO2 (Fig. 3b), which are regarded as tracers of long-range transported continental pollutants and fresh-vessel plumes. For example, extremely strong TMAH+ peaks occurred concurrently with low nss-SO42-, NH4+, and SO2 concentrations accompanied by high Na+ concentrations under high wind speeds, which are common indicators of sea-spray aerosols (Feng et al., 2017). Moreover, the TMAH+ concentrations were approximately 1 order of magnitude larger than those of TMAgas, and no significant correlation was observed between them (P>0.05). This suggests that the observed TMAH+ may not be derived from the neutralization reactions of TMAgas with acids in the marine atmosphere and may have been derived from primary sea-spray organic aerosols (Hu et al., 2015, 2018). Primary sea-spray organic aerosols mainly contain primary and degraded biogenic organics (Ault et al., 2013; Prather et al., 2013; Quinn et al., 2015; Dall'Osto et al., 2019).

The DMAH+ concentrations varied at approximately 0.065±0.068µgm-3 from 9 to 22 December; however, they varied at approximately 0.10±0.04 µgm-3 during the port-anchoring period. The 25th percentile value of DMAH+ during Campaign A was 0.021 µgm-3, suggesting a low background concentration in the marine area. The DMAH+ concentrations were significantly correlated with those of NH4+ (R2=0.71, P<0.01; data not shown). When the data obtained at 03:00–05:59 LT and 14:00–16:59 LT on 18 December (strong peaks of TMAH+ with a simultaneous increase in DMAH+) were removed for correlation, the R2 value improved to 0.78. Unlike TMAH+, the observed DMAH+ may have been partially derived from acid–base neutralization reactions in the ambient air in addition to the primary sea-spray organic aerosols. For example, a large increase in DMAH+ concentrations occurred concurrently with strong peaks in the TMAH+ concentrations (gray shadowed peak 4 in Fig. 3a and b).

The NH4+ concentrations of PM2.5 varied greatly at approximately 4.7±7.2 µgm-3 during Campaign A (Fig. 3c). However, the 25th percentile values were as low as 0.21 µgm-3, suggesting low marine background values. The 50th percentile value was also only 1.2 µgm-3, which was considerably smaller than the average owing to the presence of strong peaks in NH4+ concentrations. The increased NH4+ concentrations associated with NO3- and nss-SO42- during Campaign A were likely due to the long-range transport from the upwind continents.

As mentioned above, the observed TMAgas likely originated from marine sources. We plotted the concentrations of TMAgas against the ambient air temperature (T) in Fig. 4a, which generally increased with increasing T. We further separated the average hourly wind speed (WS) into three categories: WS≤5.0, 5.0<WS≤9.0, and WS>9.0ms-1. At WS>9.0ms-1, the data obtained from 15:00 LT on 16 December to 01:00 LT on 19 December, including peaks 3 and 4, were separately considered as half-full symbols in Fig. 4a. The concentrations of TMAgas (half-full symbols) generally exceeded those of the other gases at the same T, with which they exhibited a moderately good exponent correlation ([TMAgas]=0.03×e0.04T with R2=0.72). From 15:00 LT on 16 December to 01:00 LT on 19 December, stronger emission potentials of TMAgas to the marine atmosphere were expected in the corresponding marine zone. However, the measured concentrations of TMAH+ and seawater pH in surface seawater are required to confirm this.

Following the same approach, the DMAgas and NH3gas concentrations were plotted against T, as shown in Fig. 4b and c, respectively. These values generally increased with increasing T. The NH3gas concentrations (half-full symbols) were strongly correlated with T ([NH3gas]=0.05×e0.3T with R2=0.96). As lower concentrations of nss-SO42-, NH4+, and SO2 were generally observed simultaneously, the continental transport of NH3gas was greatly reduced; therefore, the observed NH3gas was mainly derived from the seas. Therefore, the seas were the net source of NH3gas at the time of measurement. However, at the same T, the NH3gas concentrations (half-full symbols) were generally lower than those during other periods in this study. The concentrations of NH4+ in surface seawater may have been lower at the time of measurement. However, this may not be the case, as higher concentrations of TMAH+ were expected. Alternatively, the continental transport of NH3gas may have made an important contribution to the observed NH3gas during most other periods when the seas were the net NH3gas sink.

DMAgas exhibited an extremely good exponent correlation with T (half-full symbols) at the measurement time ([DMAgas]=0.001×e0.3T with R2=0.91). At the same T, the DMAgas concentrations (half-full symbols) were not always higher or lower than the others. We considered these two hypotheses: in hypothesis 1, the observed DMAgas concentrations exceeded those predicted by the regression equation using the ambient T as the input; the seas were the net sinks of DMAgas. In hypothesis 2, including all the others, measurements of DMAH+ in the surface seawater were required to confirm whether the seas were net sources or sinks of DMAgas.

To estimate the sea-derived DMAgas and NH3gas concentrations in the marine atmosphere, we plotted the DMAgas and NH3gas concentrations against TMAgas, as shown in Fig. 5a and b. The purple-red and dark-green markers represent the data obtained, with increasing concentrations of the three species from 10:00 LT on 14 December to 23:00 LT on 16 December (increasing period) and with decreasing concentrations from 23:00 LT on 16 December to 19:59 LT on 17 December (decreasing period) during peak 3, respectively; these were analyzed separately. A good correlation was obtained between DMAgas and TMAgas during the increasing period ([DMAgas]=0.64×[TMAgas]-0.01, R2=0.86, and P<0.01). The good correlation suggested that DMAgas was likely released with TMAgas from the seawater and facilitated the estimation of non-sea-derived DMAgas (DMAgas#) concentrations using the regression equation. We assumed that any data beyond the purple-red dashed line reflected the non-sea-derived DMAgas, which can be attributed to continental transport. Therefore, we assumed that the DMAgas# concentrations were equal to the observed values of DMAgas minus the predicted values obtained using [DMAgas]=0.64×[TMAgas]-0.01; the calculated DMAgas# values are shown in Fig. 5c. During peak 1, the calculated DMAgas# contributed over 40 % of the observed DMAgas for 12 h. Similar calculated results for DMAgas# were obtained during peak 2.

However, the equation for the decreasing period was as follows: [DMAgas]=1.4×[TMAgas]-0.05, R2=0.84, and P<0.01. The decreasing R2 value and the increasing slope suggest that the TMAH+ in the surface seawater may decompose into DMAH+ to different extents (Lidbury et al., 2015a, b; Xie et al., 2018). The two regression curves (purple-red and dark-green dashed lines in Fig. 5a and b) created a large triangular zone that likely reflected the different ratios of DMAgas/TMAgas in primary marine emissions on the cruise route. Based on the triangular zone in Fig. 5a, the aforementioned calculations should be considered the lower limit of DMAgas#.

The same approach was employed to analyze the NH3gas results, as shown in Fig. 5b and d. During peak 1, the calculated non-sea-derived NH3gas (NH3gas#) contributed over 40 % of the observed NH3gas for 17 h. During peak 2, the calculated NH3gas# contributed over 40 % of the observed NH3gas for 24 h.

Overall, the DMAgas# and NH3gas# concentrations varied at approximately 0.001±0.002 and 0.18±0.39 µgm-3, respectively. The calculated average DMAgas# and NH3gas# values accounted for 16 % and 34 % of the observed averages of each species, respectively. The estimations suggested an appreciable continental contribution to the observed DMAgas and NH3gas during Campaign A.

We plotted the concentrations of DMAH+ against those of TMAH+ in PM2.5 (Fig. 6a) using the data obtained from 15:00 LT on 16 December to 01:00 LT on 19 December ([DMAH+]PM2.5=0.13×[TMAH+]PM2.5, R2=0.91, P<0.01). During this period, largely increased concentrations of DMAH+ and TMAH+ were observed under high wind speeds of 9–13 ms-1. The good correlation suggested that the observed DMAH+ was likely released with TMAH+ as amine-containing sea-spray aerosols in the atmosphere and facilitated the calculation of sea-derived DMAH+ using TMAH+ as a tracer of sea-spray aerosols. Thus, the non-sea-derived DMAH+ concentrations in PM2.5, marked as DMAH+#, were assumed to be equal to the observed DMAH+ values minus the predicted values (sea-derived DMAH+) using the regression equation. The calculated DMAH+# values are shown in Fig. 6b. The DMAH+# concentrations varied at approximately 0.042±0.070 µgm-3 throughout Campaign A, during which the calculated average DMAH+# accounted for 65 % of the observed average. Additionally, the calculated DMAH+# values accounted for over 80 % of the observed values in 26 % of the Campaign A period. The estimations suggested that the observed DMAH+ originated predominantly from long-range continental transport and/or secondary formation in the marine atmosphere. The analysis was supported by the good correlation between the concentrations of DMAH+# and those of NH4+, namely, [DMAH+#]PM2.5=0.0089×[NH4+]PM2.5 (R2=0.82, P<0.01; Fig. 6c). The slope of 0.0089 was approximately 50 % larger than that obtained in the coastal atmosphere (0.0059), suggesting more DMAgas partitioning in PM2.5 in the marine atmosphere than in the coastal atmosphere (Pankow, 2015; Xie et al., 2018).

Moreover, the decomposition of TMAH+ to DMAH+ may have occurred in surface seawater and/or the marine atmosphere, to an extent, and the estimated DMAH+# should be considered the upper limit. Notably, the NH4+ and TMAH+ concentrations were negatively correlated during Campaign A, and no primary particulate NH4+ from sea-spray aerosols was identified.

When the sea-spray particulate DMAH+ was deducted, the increased concentrations of DMAH+# were generally associated with increased nss-SO42- and SO2 concentrations. Combining this with the moderate correlation between DMAH+# and NH4+, we inferred that DMAH+# likely originated from concurrent secondary formation with NH4+. However, we separated the air pollutant plumes into two groups. Group 1 represented an increase in nss-SO42- and NH4+ together with SO2, while group 2 represented an increase in SO2 without increases in nss-SO42- and NH4+. Group 1 likely reflected the transport of aged air pollutant plumes from the continents, while group 2 may reflect self-vessel SO2 plumes. As shown in Figs. 6b and 3b–c, the concentrations of DMAH+# and NH4+ in the self-vessel SO2 plumes did not increase in the intervals between peaks 1 and 2 and between peaks 2 and 3. Therefore, no fresh formation of DMAH+# and NH4+ in the self-vessel emissions was detected. However, the concentrations of TMAH+ decreased in some self-vessel SO2 plumes. The TMAH+ concentrations were approximately 1 order of magnitude higher than those of TMAgas in the marine atmosphere. Assuming that the decreased TMAH+ was released from PM2.5 to the gas phase, a simultaneous large spike in TMAgas should be observed. However, this was not the case, as shown in Fig. 1c. The decreased TMAH+ may persist in PM2.5 but could not be detected by AIM-IC.