Aerosol (PM_2.5 and PM₁₀) online monitor (TEOM 1405DF, Thermo Fisher Scientific, Waltham, MA, USA) and a set of trace gas instruments (Thermo Fisher Scientific, USA), including 42i NO₂ analyzer, 43i SO₂ analyzer, 48i CO analyzer, and 49i O₃ analyzer, were installed in the air-conditioned container to measure mass concentrations of aerosol and gaseous pollutants along the Yangtze River. Quality control and assurance including regular cleanness and weekly and monthly calibration of the instruments were followed according to the technical guidance established by U.S. Environmental Protection Agency (U.S. Environmental Protection Agency, 1998).

Pre-combusted (at 500°C for ∼5 h) quartz filters (90 and 47 mm in diameter, Whatman International Ltd., Maidstone, UK) were applied for the sampling of PM_2.5 and particles with different sizes. The filter samples were weighed before and after sampling under the constant temperature (23°C ± 2°C) and relative humidity (RH) (40% ± 3%) for 24-h conditioning. All the filter samples were collected using the medium-volume sampler by TH-150C (Wuhan Tianhong Ltd., Wuhan, China) (flow rate 100 L/min). During the YRC, a total of 15 daily PM_2.5 samples were obtained. Size-resolved particles were collected by a 28 L/min eight-stage Anderson Cascade impactor (Thermo Electron North America, Palm Beach, FL, USA; model E-0162). The 50% cutoff points of the impactor were 9.0, 5.8, 4.7, 3.3, 2.1, 1.1, 0.65, and 0.43 μm (aerodynamic diameter). A total of eight sets of size-resolved filter samples were obtained. To obtain sufficient particles at each stage for the chemical analysis, PM_2.5 and size-resolved particles samplings both lasted continuously for 24 h starting at 8:00 a.m. All the filter samples were weighed by using a sensitive microbalance (Mettler Toledo, Greifensee, Switzerland; model AX205) with a resolution of 0.1 mg. Before chemical analysis, all quartz filter samples were stored in a freezer at −20°C to prevent the volatilization of particles.

Four anions (SO₄ ²⁻, NO₃ ⁻, Cl⁻, and F⁻) and five cations (Na⁺, NH₄ ⁺, K⁺, Mg²⁺, and Ca²⁺) in particles were measured in the extracted solution of the filter samples with ion chromatography (Dx-120, Dionex Ltd., Sunnyvale, CA, United States). CS12A column (Dionex Ltd.) with 20 mM of methanesulfonic acid MSA eluent and AS11-HC column (Dionex Ltd.) with 8 mM KOH eluent were used to measure cations and anions, respectively (Chen et al., 2019). The relative SD of each ion was less than 2% from three reproducibility tests. Five blank samples were analyzed with the same processes to remove possibly contaminations. Organic carbon (OC) and elemental carbon (EC) were measured with a thermal optical carbon analyzer (Desert Research Institute, DRI, Reno, NV, USA; model 2001) following the IMPROVE protocol and the thermal-optical reflectance (TOR).

In addition to OC and EC mass concentrations, the light attenuation (ATN) was also measured by the DIR analyzer to estimate the mass absorption efficiency (MAE) of EC, and MAE is calculated as M A E ( m 2 / g ) = A T N E C S × C × R ( A T N ) (1) where EC _s is the elemental carbon loading on the filter (µg/cm²). C and R (ATN) are the two empirical factors for correcting the artifacts due to multiple scattering and shadowing effects, respectively. In this study, an empirical factor of 2.14 (C) was adopted for correcting the artifact due to multiple scattering according to previous studies (Ram and Sarin, 2009; Cheng et al., 2011). In order to rectify shadowing effects, Weingartner et al. (2003) estimated R (ATN) based on the aethalometer data: R ( A T N ) = ( 1 f − 1 ) × ln ( A T N ) − ln ( 0.1 ) ln ( 0.5 ) − ln ( 0.1 ) + 1 (2) Ram and Sarin calculated the shadowing effects (660 nm) by setting f as 1.103 during wintertime and 1.114 for the rest of the seasons.

The aerosol extinction coefficient (including scattering and absorption coefficient) for particles at a given size (b _ext ; including b _sca and b _abs ) is calculated with the Mie theory (Bohren and Huffman, 1998; Cheng and He, 2015): b e x t = ∫ π [ D p × g 2 ] 2 × Q e x t [ m ,   D p ,   λ ] × N ( D p ) × g d D p (3) where m is the aerosol refractive index; g is the hygroscopic growth factor; Q _ext is the extinction efficiency for a single spherical particle with diameter D _p and can be calculated with the Mie theory by inputting D _p , m, and the incident wavelength (λ); N(D _p ) is the number concentration of particle with diameter D _p . Three typical models are typically proposed to represent the particle mixing state including external, internal, and core–shell mixture (Jacobson, 2001; Seinfeld and Pandis, 2006; Tan et al., 2016). The methods of calculating the parameters including m and N (D _p ) are different for the three mixed states, and the details can be referred to Ding et al. (2015).

Positive matrix factorization (PMF) is a multivariate statistical tool that had been widely used in the study of source apportionment of aerosol (Kim and Hopke, 2004). In this technique, the potential sources and their respective contributions to the mass concentration of aerosols were estimated. In total, 45 aerosol samples in this study were analyzed, and the mass concentrations of water-soluble ions and a carbonaceous component of aerosols were used for source appointment analysis. The source contributions of primary particles were obtained using the PMF model, and those of secondary inorganic aerosol (SIA) were further determined based on estimates of the nitrogen oxides (NO_x) and sulfur dioxide (SO₂) emissions in a local inventory. We followed the methods described in the PMF manual and Chen et al. (2020) to calculate the chemical component uncertainties in the measurement dataset. Criteria including the optimum number of factors and the minimization of an objective function Q were determined based on the principles described in previous studies (Watson et al., 2015; Tian et al., 2016) and applied in the model to obtain the best PMF solution. More details on the PMF solution were described in our previous study (Chen et al., 2020).

Figure 4 presents the size distributions of particles and their main chemical species. As shown in Figure 4A, the size distributions of particles (PM), sulfate, and ammonium were similar, which were a bimodal pattern that peaked at 0.65–1.1 μm and 4.7–5.8 μm. Sulfate and ammonium were mainly concentrated in fine particles (0.43–2.1 μm), accounting for a larger proportion (65.0% and 67.0%) of the total mass (Figure 4A). A previous study indicated that the peak in the droplet mode (0.56–1.0 μm) was due to the formation of sulfate through in-cloud processes under high SO₂ concentration and high RH conditions (Zhuang et al., 1999; Liu et al., 2008). The fact that high SO₂ by ship emission and high RH on the river in this study would benefit the liquid phase reaction of SO₂ and thereby the high sulfate concentration existed in the fine particle mode. Compared with nitrate and chloride, ammonium was easily combined with sulfate, since ammonium sulfate was more stable (Liu et al., 2008). As shown in Supplementary Figure S4, the high correlation between NH₄ ⁺ and SO₄ ²⁻ was found in the fine mode (0.43–1.1 μm), with R ² equal to 0.82. The mole ratio of NH₄ ⁺ and SO₄ ²⁻ was 1.1, indicating that NH₄ ⁺ cannot completely neutralize SO₄ ²⁻ and ammonium existed mainly in the form of (NH₄)₂SO₄ along the Yangtze River. Different from previous studies in urban (Wang et al., 2016b; Honglei Wang et al., 2016), no obvious difference was found in the size distribution of NO₃ ⁻ concentration, and there was even no peak in accumulation mode in this study, which probably is less affected by vehicle NO₂ emissions. Moreover, high temperatures during the sampling period may favor the dissociation of nitrate in the small particle size.

As shown in Figure 4B, the peak OC concentration was found at 0.43–0.65 μm, and EC showed a unimodal distribution that peaked at 0.65–1.1 μm. As shown in Supplementary Figure S5, there was a stronger correlation between OC and EC in fine particles (0.43–2.1 μm) (R ² = 0.82) compared to that in coarse particles (2.1–10.0 μm) (R ² = 0.41). This result indicated that the source of OC and EC in fine particles was similar. The OC and EC in fine particles were generally derived from exhaust emissions by moving source, coal combustion, and biomass burning (Zhang et al., 2015). The weak correlation between OC and EC in coarse mode indicates that there were different sources for OC and EC. A previous study suggested that coarse-mode OC was mainly attributed to the hygroscopic growth of water-soluble OC (WSOC)-containing particles from the emission of particulate OC (POC) by industrial sources (Wang et al., 2014). Coarse-mode EC could be formed through the mix of particles and hygroscopic growth by the resuspension of soil dust (Lan et al., 2011; Wang et al., 2014).

Figure 5 presents the MAE values measured by the carbon analyzer and those estimated by Mie theory. The simulated MAE values based on the assumption of external mixing states were much smaller than those based on core–shell and internal mixing states. The simulated MAE of the internal mixing mode was 27.0% larger than the measured MAE, whereas the MAE of the external mixing mode was 78.0% less than the measured value. Compared with core–shell and internal mixing, the simulated MAE value of the core–shell mode was much closer to the measured MAE, and its slope was 0.97, indicating that the core–shell mode was the main form of aerosol chemical component particles in the air along the Yangtze River channel.

The size distribution of aerosol extinction was then estimated based on core–shell state, and the result is shown in Figure 6A. Small-size particles dominated the light extinction, as shown in Figure 6B, and particles in the size ranges of 0.4–0.7, 0.7–1.1, and 1.1–2.1 μm contributed the most (90.5%) to the total extinction coefficient. The light scattering effect was the main part of light extinction, and the contribution of the scattering coefficient of 0.4- to 1.1-μm particles was up to 85.0% in the total extinction coefficient. It should be noted that the absorption coefficient contributed more than 20% to the total extinction when the particle size was in the range of 1.1–10.0 μm. Wu et al. (2021) indicated that ships at high speed emitted more black carbon (BC). Mass concentration is often proportional to particle size. Thus, BC of coarse particle size from high-speed vessels probably is the main cause for the high contribution of absorption effect. More works on the influence of ship driving conditions on particle size distribution are needed for a better understanding of radiative forcing along the Yangtze River.