PM2.5 and PM1.0 were sampled from 25 November to 5 December in 2015. Detailed information is also summarized in Table 1. The average mass concentrations of PM1.0 and PM2.5 during the YRC were 96.69±22.18 and 119.29±33.67 µg m-3, respectively. The average ratio of PM1.0 / PM2.5 was 0.8±0.085, implying that PM2.5 was mainly dominated by fine particles with a size of <1.0 µm. Detailed meteorological information, including T, RH, pressure, wind direction and wind speed, and trace gases in different episodes, is also summarized in Table 2. The peak concentrations of PM2.5 were observed in EP-4 and EP-7 (Table 1). However, there were obvious differences between EP-4 and EP-7 in the meteorological parameters and trace gas levels, indicating that these two pollution events were completely different. As mentioned in Sect. 3.1, sampled air masses in EP-4 mainly originated from local emissions, whereas EP-7 was influenced by long-range transport of air pollution.

As shown in Table 2, the average concentrations of CO, SO2, and NOx varied dramatically in the different pollution episodes. Average concentrations of CO and SO2 (993.96±387.34 and 9.32±4.33 ppbv, respectively) were slightly lower than those in the cities in winter, including Wuhan (1024.00 and 13.30 ppbv) (Wang et al., 2017), Nanjing (1096.00 and 13.09 ppbv) (Sun et al., 2017), Chengdu (1440.00 and 12.60 ppbv) (Liao et al., 2017), and Shanghai (1067.20 and 18.90 ppbv) (Huang et al., 2012a). CO levels continued to rise since the start of the YRC and finally peaked in EP-6 and EP-7. Meanwhile, the SO2 and NOx levels were much lower in these two episodes, which were identified as the BB event. As previously reported, biomass burning could produce large amounts of CO, while NOx and SO2 were not the major gaseous pollutants released from BB (Huang et al., 2012a; X. Ding et al., 2013). The mean CO concentration in EP-7 reached 1224.88 ppbv, which was close to the level recorded at Shanghai during the harvest season of wheat (June) (Huang et al., 2012a). The SO2 concentrations in EP-3 and EP-8 greatly increased and were close to the SO2 level in the haze event in Shanghai (Huang et al., 2012a). This was partly caused by local fresh emissions (the high T / B in EP-3 and EP-8). Except for EP-6 and EP-7 (BB event), the NOx concentration almost exceeded 50 ppbv along this cruise. The average mass concentration of NOx in this cruise is 63.74±41.08 ppbv, which was much higher than the mean level in Shanghai (42.40 ppbv, 2012) (Han et al., 2015) and Guangzhou (39.14 ppbv, 2012) (Zou et al., 2015) that represented the typical urban NOx level. The NOx concentration peaked in EP-3, which was identified to mainly come from local emissions. The high NOx level during the YRC was identified to come from strong regional emissions. It could be derived that multiple sources of air pollution were distributed on both banks of the Yangtze River.

The MLYR region is one of the most polluted areas in China, and the spatial distribution of various pollutants was apparently different from coastal to inland regions. As shown in Fig. 2a, the high average values of AOD retrieved from MODIS MOD04 were observed in eastern Jiangsu and Shanghai, etc., where human and industrial activities were concentrated, suggesting that anthropogenic emissions were dominated. However, there was much missing data for AOD in central China due to heavy clouds. As presented in Fig. S4 of the MODIS true-color imagery on 28 November, thick clouds covered central China. In addition, the average of AOD was about 0.45, which was slightly lower than that in Shanghai in winter (0.55) (He et al., 2012) and background (0.65) in the North China Plain (Xu et al., 2011). The AOD value in northern China was higher than that in southern China. As plotted in Fig. 2b, the CO surface mixing ratio calculated by MOPITT revealed that Shandong, Henan, and Anhui were exposed to elevated CO column concentrations. CO is an important tracer for incomplete combustion sources, such as BB and fossil fuel combustions (Girach et al., 2014). BB should be a major source for CO in grain-producing areas (Huang et al., 2012a; A. J. Ding et al., 2013). As mentioned in Sect. 3.2.1, the peak CO level was also observed in Anhui and the west of Jiangsu, which was characterized by agriculture emissions (EP-6 and EP-7). However, the sources of CO in northern China should be further studied in the future. The high levels of SO2 were mainly observed in the east in Anhui and stretched to the Shanghai area (Fig. 2c), suggesting high-sulfur fossil fuels were still widely utilized over the MLYR region. Conversely, SO2 levels in Nanjing urban areas were measured at background pollution levels. In general, NO2 was regarded as a tracer for the local emission source due to a short lifetime in the atmosphere (Geng et al., 2009; Xu et al., 2011). NOx emissions mainly originated from vehicles and power plants (Fu et al., 2013; Geng et al., 2009). One can see that NO2 emissions were characterized by strong local sources in north China and the YRD region (Fig. 2d), which is in good agreement with previous reports (Lin, 2012; B. Zhao et al., 2013).

As illustrated in Fig. 3, the mass concentrations of SNA with an average of 38.35±15.17 µg m-3 increased dramatically from coastal to inland cities and exhibited the highest level (no. 6, 79.06 µg m-3) in the Wuhan region (EP-4), accounting for nearly 50 % of the local PM2.5 mass loading. As mentioned above, SO2 and NOx also presented the highest concentrations in this region. Furthermore, Wuhan and the surrounding regions were controlled by a low-pressure system with low wind speed and high-RH conditions (Fig. S3), which have been verified to cause haze episodes (X. J. Zhao et al., 2013; Quan et al., 2011; Wang et al., 2010). In addition, the ratio of NO3-/SO42- in the Wuhan area was close to the values of cities in northern China (relatively low) (Fig. 4), suggesting that the stationary sources (such as coal-fired power stations or stove emissions) dominated in this region. Heavy clouds and high humidity in central China suggested aqueous-phase transformation processes were probably the main reaction path of SO42- from SO2 (Wang et al., 2016; X. J. Zhao et al., 2013). In addition, the mass fractions of SNA in PM2.5 also peaked in rural regions (EP-2 and EP-6), which was in accord with the low ratio of T / B in these regions, suggesting that aerosol particles in rural regions were well aged.

Meanwhile, trace elements for coal combustion (component 1) also had the highest concentrations in EP-4 and EP-5 (Fig. 5b) when the ship anchored in Wuhan and traveled through the Jiujiang area. As illustrated in Fig. 6a–d, As, Cd, Pb, and Se showed similar source distribution. The higher PSCF values in Hubei, Hunan, and Jiangxi provinces coincided well with the uneven regional distribution of residential coal consumption (Fig. S7) in central China, suggesting coal-related PM pollution was quite serious in this region during this cruise. The peak mass fraction of Cl- and SO42- in PM2.5 in Wuhan also confirmed this result Thus, we conclude that coal combustion contributed significantly to serious pollution with a high SNA loading in Wuhan and the surrounding regions during sampling.

In contrast to SNA distribution, the concentration of Ca2+ along this cruise increased from the mainland to the coast of the East China Sea (Fig. 3). The peak concentration of crustal elements (component 2) and Ca2+ mass fraction of PM2.5 occurred in EP-7 when a cold front and associated northeast winds arrived, accompanied by floating dust (Figs. 3 and 5b). The dust episode was verified by the MODIS true-color image on 2 and 3 December (Fig. S4) and was further confirmed by a drastic decrease in RH with the prevailing northwest wind (Table 2 and Fig. S3). As shown in Fig. 6e–h, the YRD region and the Loess Plateau with the highest PSCF values were identified as important source regions and/or pathways for crustal elements of Al, K, Mg, and Ca. Meanwhile, central China also showed distribution of K and Mg, for which the coal combustion in this region could be primarily responsible. Furthermore, Ca showed high EFs (EFs >100), suggesting that the crustal element may not derive from a natural source but from anthropogenic resuspension of dust from road and/or construction activities along the Yangtze River. To further evaluate the impact of anthropogenic Ca, the equation below was applied: Caanthropogenic=Catotal-Altotal×(Ca/Al)crust. (Ca/Al)crust is the ratio of Ca to Al in the crust, and its value is 0.5. According to this method, the average Caanthropogenic concentration was 2.15 µg m-3, and the peak level reached 3.42 µg m-3 on 3 December. If all Caanthropogenic values in the samples of other cities and cruises (Table 4) were calculated according to the same method, the level in this cruise was much higher than those in other samples, suggesting that anthropogenic dust dominated and was distributed in the YRD region during the period.

Resembling the Ca2+ distribution pattern, the maximum concentration and mass fraction of Na+ and K+ in PM2.5 were also measured during EP-7. Significant correlation between Ca2+ and K+ suggested that dust could be the major source of K+ in PM2.5 sampled during the YRC (Fig. 3). In general, it is well known that dust particles with high alkalinity could first neutralize SO42- and NO3- in aerosol particles and then atmospheric ammonia was absorbed. The concentrations and mass fractions of SNA in PM2.5 slightly increased at the end of the cruise (EP-7 and EP-8) (Fig. 3) since carbonate in aerosol could enhance the uptake of acidic gases on particles (Huang et al., 2010). Meanwhile, the increasing mass ratio of NO3-/SO42- in EP-7 and EP-8 was attributed to two main reasons (Fig. 3). The mobile sources (such as vehicle emissions) increased and released a huge amount of NOx when the vessel was close to megacities (Huang et al., 2013). Furthermore, NO2 could transform into NO3- via heterogeneous processes on the dust aerosol surface (Nie et al., 2012).

Heavy metals have toxic effects on plants, animals, and human beings. However, there is no uniform standard concentration as a control indicator (Sharma and Agrawal, 2005). The trace elements (component 3 and component 4), with high EFs ranging from 24 to 1213, were considered to mainly come from heavy oil and industry, respectively. The high concentrations of V and Ni were observed when the ship berthed in the Waigaoqiao port region of Shanghai (EP-8) (Fig. 5b), where some field observations also identified that heavy oil combustion exerts a significant impact on the local air quality (M. Zhao et al., 2013; Fu et al., 2014; Ding et al., 2017; Liu et al., 2017). It is also reported that the transition metals of Ni and V were greatly enriched in smaller particles with a diameter of <1.0 µm (Jang et al., 2007). Fine-particle Ni (Fig. 6i) had almost the same spatial distribution as Cr (Fig. 6j), and Shanghai, Jiangsu, and the east of Anhui were identified as the major potential source regions and/or pathways, owing to ship emissions, nonferrous metal mining, and smelting industries. The Mongolian plateau was also a source region, indicating that natural dust may be another possible source for Cr and Ni. However, the high PSCF values of fine-particle V were only derived from the YRD region and Mongolian plateau (Fig. 6j). V was considered to originate from heavy oil combustion, while Ni and Cr probably have other sources (Table S1) (M. Zhao et al., 2013).

The temporal variations in component 4 nearly peaked in Wuhan and Shanghai (EP-4, EP-7, and EP-8) (Fig. 5b) where the China Baowu steel industry and numerous shipyards are located (Ivošević et al., 2016). Fine-particle Fe, Co, Mn, and Zn displayed similar regional distribution (Fig. 6l–o), and the high PSCF levels were observed in the YRD region, indicating that steel industries and/or shipyards were densely distributed in the east of Anhui, Jiangsu, and Shanghai. In addition, the high PSCF value for Zn (Fig. 6l) was also exhibited in Hubei, Henan, and Shanxi provinces, probably due to the influence of coal combustion and nonferrous metal smelting activities in these regions (T. Li et al., 2015). Overall, it should be noted that anthropogenic sources in megacities (WNS) were dominant origins for trace elements in fine particles collected during this cruise.

Numerous studies have also confirmed that levoglucosan mostly originates from BB (Liang et al., 2017; X. Ding et al., 2013; Wan et al., 2017; Wang et al., 2014). The distribution of levoglucosan is irregularly parabolic from inland to coastal areas in Fig. 3. The maximum value of levoglucosan (0.18 µg m-3) was observed in the rural of Anhui Province (EP-6), while its level in the YRD region (EP-8) was very low. The elevated levels of CO and low concentrations of SO2 and NOx also confirmed BB in EP-6 and EP-7 (Table 2). However, fire points could not be apparently observed in the satellite-detected fire maps (https://firms.modaps.eosdis.nasa.gov/map, last access: 1 June 2017) due to heavy cloud cover on 27 November and 1 December. During the whole observation period, there was only one sample (no. 12, Fig. S8) collected during a BB event. It was verified by MODIS fire points due to a cold front blowing heavy clouds away (Fig. S4). The slightly higher levoglucosan concentration was observed in the night and was attributed to the lower boundary layer at night and BB for heating and cooking in the rural regions.

The levoglucosan concentration and ratio of OC to levoglucosan (OC / levoglucosan) were also widely applied to estimate the contribution of BB to OC in PM2.5. An empirical model was utilized as proposed by Wan et al. (2017): OCBB=[lev][OC]ambient/[lev][OC]BB. The differences of the (lev / OC)BB ratio among different biomass fuels and combustion conditions were taken into account. Thus, the average (lev / OC)BB ratio of 8.14 % was selected to calculate the contribution of BB to OC (Wan et al., 2017). Figure S9 presents the variation in levoglucosan / OC ratio along the Yangtze River. The ratio of levoglucosan / OC during this cruise ranged from 0.03 % to 0.91 % with an average of 0.35±0.24 %, which was comparable to that of Lin'an in the YRD region (Liang et al., 2017). However, the ratio of levoglucosan / OC during the YRC was near an order of magnitude lower than its value in New Delhi (3.1±0.8 %) (Li et al., 2014) and Lumbini (3.34±2.53 %) on the northern edge of the Indo-Gangetic Plain (Wan et al., 2017), where BB plays an important role in air quality. Figure S9 also shows the time series of contribution of BB OC to OC. The average contribution of BB OC / OC was 4.26±2.89 %, while the mean mass fraction of OC to PM2.5 was slightly higher than 20 %. The peak contribution of OC derived from BB to total OC of PM2.5 nearly accounted for 11 % in EP-6, which approached that of the Pearl River region sites (13 %) (Ho et al., 2014). Here, it is emphasized that our method based on empirical formula and value is just a rough estimation. Hence, the radiocarbon measurement (14C) of carbonaceous aerosol and air quality model simulation will need to confirm this result in the future.

Over the past few decades, China's rapid economic development has led to increasingly busy shipping transportation in the Yangtze River. However, there is lack of data related to ship emissions along the Yangtze River, especially in inland areas. The ratio of V to Ni was used to judge whether ship emissions could influence air quality (Isakson et al., 2001). The average ratio of V / Ni during the cruise is 1.27, which was in good agreement with previous studies (Pandolfi et al., 2011; Zhang et al., 2014). Emission factors of heavy metals from different types of fuel oil were also analyzed in our group (Table S1). Only heavy oil contained V, while the V levels emitted from other diesel and petrol sources were under the detector limits. In this study, only V was regarded as a tracer for heavy oil combustion. However, it was still difficult to distinguish V from refinery and ship emissions. Hence, the high-resolution back trajectory and high-resolution of the ship position from the AIS data were applied to investigate ship plumes during this cruise. As plotted in Fig. 7, the numbers of the ship from AIS were closely related to the V concentrations. From the inland region to the East China Sea at Shanghai, the concentration of V of PM2.5 generally increased and reached the highest level of 0.06 µg m-3 on 4 December when the vessel berthed in the anchorage of the Yangtze River estuary. Meanwhile, air masses on this evening originated from the port and anchorage (Fig. S10). Hence, V of PM2.5 sampled in the port of Shanghai could be attributable to ship emissions, especially from oceangoing vessels.

The contribution of primary ship emissions to PM2.5 could be calculated by the equation developed by Agrawal et al. (2009): PMa=〈a〉×〈r〉×Va/〈FV,HFO〉, where PMa represents the primary PM2.5 concentration estimated (µg m-3); 〈a〉 is a coefficient of the fraction of V from ship emissions in fine particles in China (0.85); 〈r〉 is the average ratio of PM2.5 to normalized V emitted (ppm); Va represents the V amount of the samples (µg m-3) during the YRC and is the V content of heavy oil on average from the vessels (ppm). The value of 〈r〉 was set as 8205.8 ppm as Agrawal et al. (2009) reported. The value of 〈FV,HFO〉 was set as 65.3 ppm, which represents the average V content (M. Zhao et al., 2013). The average concentration of primary ship emissions was 1.01±1.41 µg m-3, ranging from 0.02 to 6.27 µg m-3, which is higher than that at Tuoji Dao (0.65 µg m-3) (Zhang et al., 2014). The peak level of primary ship emissions was observed in the Shanghai harbor.

To in-depth characterize the contribution of the ship emissions to secondary fine particles, a lower limit of the SO42- / V, NO3- / V, EC / V, and OC / V ratios (equal to the average minus 1 standard deviation) was applied to estimate the particulate from heavy oil combustion along the Yangtze River (Becagli et al., 2017). As presented in Fig. S11a–b, the mass ratio of SO42- / V and NO3- / V decreased rapidly with increasing V concentration. According to ship traffic numbers, weather conditions, and the EFs of different types of oils (Table S1), the samples with V >15 ng m-3 were mainly considered to come from ship emissions.

The limit ratio of SO42- / V, NO3- / V, and OC / V, and the estimation of ship emission contributions to SO42-, NO3-, OC, and PM2.5, is summarized in Table S2 in the Supplement. The minimum ratio of NO3- / V in this cruise was nearly two times larger than the limit ratio for SO42- / V, which was contrary to the previous results with higher SO42- from ship emissions observed in summer on the island of Lampedusa (35.5∘ N, 12.6∘ E) in the central Mediterranean. In general, SO42- and NO3- in aerosol were mainly formed through gas precursors of SO2 and NOx, respectively, both of which were completely different for lift time and chemical processes in the atmosphere. High UV radiation and humidity could accelerate the reaction rate of SO2 to SO42- (Zhou et al., 2016). However, NO3- was in gas-aerosol equilibrium with gaseous HNO3 (g). Low temperature and humidity would shift the gas-aerosol equilibrium to the particle phase (NO3-) (Matthias et al., 2010; Wang et al., 2016). One reason for this discrepancy was probably meteorological and photochemical conditions, which led to lower sulfur conversion rate and particulate NO3- domination in low temperatures and moisture in winter during this cruise (Table 2). Conversely, NO3- may have other sources in the Shanghai port, whereas Lampedusa was a remote site (Becagli et al., 2017). The average estimated concentration of minimum SO42- derived from ship emissions was 1.38 µg m-3 during the YRC, which was similar to the value (1.35 µg m-3) measured in the Mediterranean (Becagli et al., 2017, 2012).

EC and OC were also estimated using the same methods for SO42- and NO3-, and the lower limit for the OC / V and EC / V ratios is also presented in Fig. S11c–d. In addition, significant correlation between V and EC (R2=0.71, P<0.01) suggested that V and EC have the same sources (Agrawal et al., 2009). In this cruise, organic matter (OM) of PM2.5 was estimated from OC by multiplying a conversion factor of 1.4, due to typical fresh emissions and weak light in winter (Becagli et al., 2017). The estimated lower limit for the average of ship emissions was 7.65 µg m-3, contributing 6.41 % of PM2.5 during the YRC. The peak ship contribution could reach up to 36.04 % of total PM2.5 when the vessel berthed in the Waigaoqiao port of Shanghai, which was slightly above the value (20 %–30 %) estimated by Liu et al. (2017) during ship-plume-influenced periods. It should be noted that the ship emissions decreased from the Shanghai port to the inland area. One reason for this corresponded to the density of ships in the Yangtze River channel. However, fuel oils were completely different between the ships traveling on inland waterways and oceangoing vessels. In general, light diesel with low EFs of heavy metals (such as V and Ni) was widely used by the ships on the river, whereas heavy oil with a high content of V and Ni was widely burned onboard marine vessels (Table S1). Oceangoing ship emissions were probably the major air pollution sources in the Shanghai port. Hence, it is urgent to establish emission control areas (ECAs) in Shanghai ports. However, it is worth noting that our estimation based on empirical values was also limited by meteorological conditions and sample numbers. Hence, long-term observation and high-resolution model simulations of ship emissions should be strengthened as part of the control of air quality along the Yangtze River, especially in the Shanghai harbor cluster.