When selecting the study sites, we referred to the trajectories of air masses reaching Wuhan in January 2017 (Fig. S1 in the Supplement) and found that the north and northwest directions were dominant. Regarding the northerly direction, the air masses originated from the SE-NCP and Luohe and moved south along routes that are close to heavy polluted regions (Fig. 1). As seen in Fig. 1, there are two obvious channels that allow for the movement of PM2.5 between the SE-NCP and CC, which are differentiated by the mountains dividing the two regions. Therefore, in order to investigate the regional transport of air pollutants and also answer the question regarding whether the pollutants in CC can be transported to the SE-NCP in winter, five sites including Wuhan (WH), three borderline cities (Xiangyang – XY, Suixian – SX, and Hong'an – HA; from west to east in Fig. 1) between the NCP and CC, and a city (Luohe, LH) in the SE-NCP were selected. The observation site at WH is located on a rooftop at the Hubei Environmental Monitoring Centre, which is an urban site with no industrial emission sources nearby. The LH and XY sites are located in suburban areas, whereas the HA and SX sites belong to rural areas. The observation instruments were placed near the local environmental monitoring stations. Measurements from six routinely monitored air pollutants including PM2.5, PM10, NO2, SO2, CO and O3 were available. Black carbon measurement instruments including Magee Scientific-AE31, AE33 and AE51 devices were deployed (Table 1). The observation periods began on 8 January after a regional snowfall event and ended on 25 January 2018 before another snowfall event began. The durations of the observations at the five sites are summarized in Table 1.

The AE31 instrument continuously collects ambient BC on a quartz tape and measures light transmitted through a “sampled spot” (I) and a “reference spot” (I0) on the tape. The light attenuation (ATN) is then defined as 1ATN=-100ln⁡II0. This assumes a linear relation between the BC mass loading and the delta of the ATN as a result of BC deposited on the tape. The BC mass concentration is calculated as follows: 2BC=d(ATN)MAC×AV, where MAC is the mass specific attenuation cross section (m2 g-1), A is the area of the sampled spot (1.67 cm2) and V is the volume of the sampled air passing through the tape. An disadvantage of the AE31 is the filter loading effect, which needs to be compensated for via a correction (Petit et al., 2015). The BC absorption coefficient (babs, Mm-1) is calculated as 3babs=BC×MACC×R(ATN), where C is the calibration factor (2.14 for quartz material tape), and R (ATN) is a correction factor for the shadowing effect; this correction factor is empirically determined using the following compensation parameter f (Weingartner et al., 2003): 4RATN=1f-1ln⁡ATN-ln⁡10ln⁡50-ln⁡10+1. To overcome the limitation of the loading effect, the AE33 (dual spot) was developed. This instrument also simultaneously measures the ATN at seven wavelengths. In contrast to the AE31, the AE33 measures BC on two parallel spots on the (Teflon-coated) fibre tape at a different flow rate: 5BC1=BC×(1-k⋅ATN1)6BC2=BC×(1-k⋅ATN2). The loading compensation k is calculated according to Eqs. (5) and (6), whereas the BC mass concentration is calculated as follows: 7BC=A[d(ATN)/100]F11-φMAC⋅C1-kATN1dt. For the AE33, the sample spot area (A) is 0.785 cm2, and the enhancement parameter (C) is 1.57 for the Teflon-coated fibre. The absorption coefficient (σabs, Mm-1) for the AE33 is estimated by multiplying the BC mass concentration by the MAC. More details regarding the BC concentration calculation, parameters (i.e., f and MAC for different wavelengths) and the differences between the AE31 and AE33 instruments can be found in a previous study (Rajesh and Ramachandran, 2018). The AE51 measures the absorbance (ATN) of the loaded spot (3 mm diameter) and the reference portion of a Teflon-coated borosilicate glass fiber using a stabilized 880 nm LED light source. The flow rate of the AE51 was set to 100 mL min-1; more information about AE51 can be found online (https://aethlabs.com/microaeth/ae51/tech-specs, last access: 20 June 2018).

The concentration-weighted trajectory (CWT) analysis is always used to assess the regional transport of air pollutants (Kong et al., 2018; Zheng et al., 2018). This method is based on backward trajectory analysis. Prior to CWT analyses, the calculation of backward trajectories was firstly carried out for each sampling site. The input wind data sets for HYSPLIT were downloaded from the Nation Oceanic Atmospheric Administration (NOAA) (ftp://arlftp.arlhq.noaa.gov/pub/archives/gdas1/, last access: 25 April 2018). For backward trajectory analysis, the air masses reaching each observation site during the sampling period were calculated 24 times with a 1 h resolution each day (starting from 00:00 to 23:00) at 200 m a.g.l. (Fig. S2). These trajectories were than clustered according to their geographic origins (Fig. 1). For CWT analysis, a user-friendly Igor-based tool “ZeFir” was used (Petit et al., 2017a). The domain covered by the trajectories was divided into thousands of cells with a 0.2∘ × 0.2∘ resolution. More details regarding the CWT analysis can be found in the Supplement (Text S1).

An hourly meteorological data set including sea level pressure, temperature, relative humidity, wind speed, wind direction and visibility was acquired form the China Meteorological Data Service Centre (CMDC) (http://data.cma.cn, last access: 26 January 2018). The 3-hourly boundary layer height (BLH) was acquired from the NOAA's READY Archived Meteorology online calculating program (http://ready.arl.noaa.gov/READYamet.php, last access: 8 April 2018). Figure S3 shows the hourly averaged meteorological parameters at the five sites. Meteorological conditions at the five sites followed similar trends, although significant differences (p<0.01) between these parameters were found (Table S3). For instance, the average pressure, temperature and relative humidity at WH were significant higher (p<0.01) than those at LH. Regarding the BLH, the mean values of the five sites showed no significant differences.

Information regarding six air pollutants (PM10, PM2.5, SO2, NO2, CO and O3) was available, and the data were downloaded from the China Environmental Monitoring Centre (http://www.cnemc.cn, last access: 10 April 2018). Figure S4 shows the hourly variations in these species during the observation period. The major air pollutant throughout the observation campaign was PM2.5. According to the ambient air quality standards (GB3095-2012), the air quality can be classified as clean, lightly polluted or heavily polluted when PM2.5 mass concentrations are less than 75, between 75 and 250, and greater than 250 µg m-3, respectively. Similar air quality classifications have also been reported elsewhere (Zheng et al., 2015; Zhang et al., 2018). Detailed information regarding the daily air quality of each site is shown in Fig. S5.

Time series and box plots of the eBC concentrations (measured at 880 nm) at the five sites are shown in Fig. 2. The highest eBC concentration was observed at LH (8.48±4.83 µg m-3), followed by XY (7.35±3.45 µg m-3), HA (5.54±2.59 µg m-3), SX (4.47±2.90 µg m-3) and WH (3.91±1.86 µg m-3). As shown in Table S1, BC was generally higher in North China, whereas lower BC levels were found in remote and coastal areas (Fig. 3a). Q. Wang et al. (2014) analyzed ambient BC at an urban site in Xi'an during winter and found that the average mass concentration was 8.8±3.7 µg m-3, which was higher than concentrations observed in this study. Compared with other regions, BC levels in this study were higher than those in a remote area of Lulang in the southeastern part of the Tibetan Plateau (0.31±0.55 µg m-3; Q. Wang et al., 2018) as well as coastal areas, such as Hong Kong (1.4±1.1 µg m-3; J. Wang et al., 2017), and a rural site in Shenzhen (2.6±1.0 µg m-3; Huang et al., 2012). Based on the BC emission inventory, northern and central China display higher BC emission intensity (Qin and Xie, 2012; Yang et al., 2017). Emission values in Hubei and Henan provinces were about 0.6–1.0 g C m-2 yr-1, which makes them higher than values in other regions (Yang et al., 2017). Simulation results also suggested that the near-surface concentrations of BC (6–8 µg m-3) in Hubei and Henan were higher than those in southern China (4–6 µg m-3) during winter (Yang et al., 2017). Compared with the data from other countries (Table S1), BC levels in this study were higher than those in Finland (Hyvärinen et al., 2011), France (Petit et al., 2017b), Canada (Ontario – Healy et al., 2017) and South Africa (Chiloane et al., 2017).

For the aerosol absorption properties measured at seven wavelengths using an Aethalometer, the characteristics (i.e., temporal variation) are generally consistent with one another; however the properties are mostly discussed at a wavelength of 520 nm (Zhuang et al., 2015, 2017; Y. Wang et al., 2017). Then, we only discussed the absorption properties at λ=520 nm. Figure 4a shows the frequency distribution of absorption coefficients (σabs) at three sites. σabs measured at HA, LH, and WH exhibited a single peak pattern. The average values of σabs measured at HA, LH and WH were 86.0, 132 and 60.6 Mm-1, respectively. Similar to the spatial distribution of the BC level, higher σabs values were found in northern and central China, whereas lower values were observed in coastal areas and the Tibetan Plateau (Fig. 3b and Table S2).

Figure 4b also shows the average absorption spectra measured at seven wavelengths for different sites. The power law fit was used to calculate the AAE (Zhu et al., 2017). The highest average AAE value was found at LH (1.37), followed by HA (1.32) and WH (1.29). These results indicated that the AAE was different at urban, suburban and rural sites. Generally, the AAE values from coal combustion (2.11–3.18; Sun et al., 2017) and biomass burning (1.85–2.0; Petit et al., 2017b) were higher than those from traffic sources (0.8–1.1; Sandradewi et al., 2008; Olson et al., 2015). Therefore, the AAE at different sites suggested a different energy consumption structure, and that more coal or biomass was burned in North China (i.e., LH in this study).

Figure 5 displays the eBC concentrations under different air quality conditions. It clearly shows that the eBC concentrations increased with the deterioration of air quality. At LH, the average eBC concentrations were 3.39±2.06, 8.31±4.55 and 13.0±4.59 µg m-3 when the air quality was clean, lightly polluted and heavily polluted, respectively. The average values of eBC increased by 163 %, 139 %, 96.2 %, 51.8 % and 26.4 % at SX, XY, LH, HA and WH, respectively, from clean to polluted periods. eBC enhancement along with air quality deterioration has also been reported elsewhere (Q. Wang et al., 2014; H. Wang et al., 2014; Liu et al., 2016; Y. Liu et al., 2018). At LH and HA, the enhancement of the eBC level from clean to polluted periods was due to both elevated BC emissions from biomass burning (BCbb) and fossil fuel combustion (BCff) (Fig. 5b, c). BCff accounted for a higher contribution to eBC and the percentages of BCbb to eBC decreased during the haze episodes (Fig. 4d). At WH, both the concentration and percentage contribution of BCbb decreased from clean to polluted conditions, which suggested that more BCff was emitted during haze episodes. This finding was different to those of a previous study conducted in Beijing, which reported that the absolute concentration and percentage contribution of biomass burning and coal combustion to eBC were higher than those from traffic sources and increased from clean conditions to polluted episodes (Y. Liu et al., 2018). These difference suggest that the control of fossil fuel combustion (vehicle emissions), not coal or biomass burning, should be the priority during haze episodes at WH, whereas biomass burning and coal combustion control should be made a priority in North China to prevent air pollution.

Additionally, the aerosol optical properties (σabs and AAE) also exhibited different levels under different air quality conditions. Similar to the eBC levels, the σabs increased by 11.7 %–254 % as the air quality switched from clean to polluted conditions (Fig. 5e). Our observation (Fig. S6) and previous research both found that there are more secondary aerosols (i.e., sulfate and nitrate) during pollution episodes (Huang et al., 2014). The increase in secondary aerosols would result in more of these secondary aerosol species being adsorbed onto (coating) BC particles, which would enhance the absorption properties of BC via the lens effects of these coating materials (Jacobson, 2000; Moffet and Prather, 2009). On the contrary, the AAE showed higher values during clean days compared with polluted periods (Fig. 5f). This decrease in the AAE from clean to polluted days has also been reported elsewhere (Y.-L. Zhang et al., 2015), and it can be partly attributed to source variation. The AAE for biomass burning is about 2.0, whereas the AAE for fossil fuel combustion is about 1.0 (Sandradewi et al., 2008). Higher AAE values during clean periods suggested that more BC may be from biomass burning, whereas lower AAE values indicated the dominance of fossil fuel combustion during the polluted periods (Fig. 5c). The AAE is also sensitive to other factors such as the particle size. Previous studies have suggested that the particle diameter and number concentration increases from clean to polluted conditions due to several factors such as coagulation, hygroscopic growth, emissions and meteorological conditions, i.e., planetary boundary layer and wind speed (Guo et al., 2014; Zhang et al., 2017). These studies have also suggested that the particle diameter is generally larger during polluted periods. Furthermore, lab combustion experiments and numerical simulations have proven that BC particles with larger geometric median diameters have lower AAE values (Singh et al., 2016; C. Liu et al., 2018); hence, a lower AAE value was observed during pollution episodes in this study.

Figures 6 and S7 show the diurnal variations of eBC and absorption coefficients under different air quality conditions. The diurnal cycles of black carbon and absorption showed similar variation patterns. The BC mass concentrations are discussed here. At HA, LH and SX, after sunrise, an increase and a peak value were observed at about 09:00 local time (LT). This variation was more obvious during polluted periods due to the higher eBC levels. The morning peak may have been related to the combined effects of increased biomass burning and fossil fuel combustion emissions (Fig. S7). Additionally, the low mixing layer height in the morning also favored the accumulation of eBC. After sunrise, with the elevation of the BLH, the eBC levels decreased and a minimum occurred at about 15:00 LT. In the evening, eBC showed an increasing trend and peaked at about 21:00 LT. Similar diurnal patterns of eBC have also been reported in other areas (Verma et al., 2010; Ji et al., 2017; Y. Liu et al., 2018). In contrast, the diurnal variations of eBC at WH and XY exhibited different patterns during clean and polluted periods. The diurnal pattern of eBC at WH was likely not controlled by the development of the mixing layer height, which generally leads to the maximum and minimum values of air pollutants occurring at sunrise and in the afternoon, respectively. Hence, the unexpected lower value in the morning (about 09:00 LT) and a higher value in the afternoon (15:00 LT) at WH require further research.

BC/PM2.5 and BC/CO ratios are widely used to identify the sources of BC (Zhang et al., 2009; Wang et al., 2011; Verma et al., 2010; Chow et al., 2011). Generally, ratios of BC/PM2.5 from mobile sources (0.059–0.74) and area sources (0.032–0.33) are higher than those from industrial sources (0.0046–0.03). For instance, mobile sources present the highest ratios of BC/PM2.5 (0.33–0.77) and cement kilns show lower ratios (0.03) (Chow et al., 2011). The BC/CO ratios (µg m-3 / ppbv) also vary for different sources, i.e., traffic (0.0052), industry (0.0072), power plants (0.0177) and residential (0.0371) (Zhang et al., 2009). In this study, the BC, PM2.5 and CO were well correlated with one another (Fig. S8). The correlation coefficients (r2) between BC and PM2.5 were 0.67, 0.30, 0.44, 0.37 and 0.48 at LH, HA, WH, SX and XY, respectively. Significant correlations (p<0.05) between BC and CO were found with r2 values ranging from 0.27 (XY) to 0.71 (LH). These good correlations indicated that BC, PM2.5 and CO may be from similar sources (except at HA where r values as low as 0.06 were observed).

Overall, the BC observed in this study was likely not from industrial emissions (Fig. 7a), as the BC/PM2.5 ratios (µg m-3 / µg m-3) (0.045–0.083) were higher than those typically seen from industry (0.0046–0.03) (Chow et al., 2011). Instead, BC/PM2.5 ratios at the five sites were all within the range of oil combustion (0.03–0.136). Additionally, the BC/PM2.5 ratios at LH and SX were in line with the ratio values seen from residential wood combustion. With respect to BC/CO ratios, BC was more likely from biomass burning (crop residue: 0.0056–0.016) at HA and LH, whereas it was mainly from gasoline combustion at SX, WH, and XY (Fig. 7b). Quantified calculation results using the equations in Sect. 2.3.1 also suggested that the fractions of BC from biomass burning at HA (27.6±9.40 %) and LH (29.5±9.14 %) were significantly higher (p<0.01) than those at WH (25.4±11.8 %). Compared with other urban areas, the ratios of BC/CO (µg m-3 / ppbv) at SX (0.004) and WH (0.0044) were lower than those in Beijing (0.0058) (Han et al., 2009), Guangzhou (0.0054) (Verma et al., 2010), Gwanjun (0.006) (Park et al., 2005) and Tokyo (0.0057) (Kondo et al., 2006) as well as Mt. Huang (0.0065) (Pan et al., 2011), whereas ratios at HA (0.0091) and LH (0.0076) were higher than the values observed in these studies.

A conditional bivariate probability function (CBPF) plot was used to identify and quantify the impact of likely source regions of air pollutants as defined by wind direction and speed (Carslaw and Ropkins, 2012). Figure S9 shows the eBC levels under different wind speeds and directions at the five sites. As shown in Fig. 1, SX and HA are located northwest of WH, and high eBC levels were found northwest of SX, HA and WH when northerly winds dominated. In contrast, when southerly winds dominated, BC was blown from south to north, and high levels were found in the south at WH and HA. However, at LH and XY, higher levels of BC were only found from the south. In addition to eBC levels, BCbb and BCff concentrations under different wind speeds and directions are also discussed at HA, LH and WH (Fig. 8). At HA, the CBPF plot of BCff was in line with eBC, and high levels were observed from both the northwest and the south, whereas high levels of BCbb (>1.8 µg m-3) were only found in the southeast. Similar results were also found at WH. This high level of BCbb was due to more biomass burning southeast of HA and WH (Fig. S10). At LH, the CBPF plots of BCbb and BCff were the same as those for eBC, as discussed above.

In order to describe the downwind (from upwind to downwind) BC transportation, we used Eq. (13) in Sect. 2.3.3 to calculate the surface transport (ST) of eBC (Fig. 9). The calculated average ST values of BC were -0.69±10.2, -0.06±12.0, -0.17±5.33, 0.29±6.14 and 0.99±17.8 µg s-1 m-2 for HA, LH, SX, WH and XY, respectively . The negative values at HA, LH and SX suggested that the transportation intensity of BC from the south (southeast) to the north (northwest) was higher, whereas the positive values observed at WH and XY indicated that more BC was transported from north to south. The large standard deviation in surface transport (ST), as mentioned above, reflected strong fluctuations in transport, which were due to wind speed, wind directions and BC levels (Z. Wang et al., 2018).

Employing the CWT method, the potential geographic origins of eBC for the five sites were explored (Fig. S11). Overall, the CWT results of eBC at the five sites suggested that high eBC levels were found both north and south of LH and WH, whereas the high levels (i.e., >4 µg m-3) of eBC were only found northeast of HA, SX and XY (Fig. S11). Additionally, the potential geographic source regions of BCbb and BCff at HA, LH and WH are also discussed as shown in Fig. 10. At HA, the CWT results showed that high levels of eBC (i.e., >3 µg m-3) were from north/northeast. However, the hot spots of BCbb and BCff were different, with higher levels of BCbb from both the south and north and higher levels of BCff from the north. Furthermore, higher levels of BCbb and BCff were found south of LH. In contrast to the CWT results at HA, the hot spots of BCbb were only found southeast of WH, and high levels of BCff were found north and south of WH. The CWT results at WH were in line with the CBPF plots in Sect. 3.4. The unity of the CWT and CBPF results at WH suggested that there were intensive biomass burning activities south of WH during the observation period, which was verified by the MODIS fire points distribution map (Fig. S10).

We also discussed the source region differences of BC under different air quality conditions (Fig. 11). The higher levels (>1 µg m-3) of eBC, BCbb and BCff mainly originated south of three sites when the air was clean, whereas during pollution episodes, air parcels from the north contributed high concentrations of these species. For instance, at WH, high levels of eBC (>2.5 µg m-3) were found from the south, whereas the source regions with high levels of eBC (>3 µg m-3) switched to northeast as the air quality deteriorated. Figure 12 shows the semiquantitative results of the transportation contribution results during clean and pollution periods. At the boundary sites (HA, SX and XY), BC was mainly from the south (accounting for 46.0 %–58.2 %) when the air quality was good (clean), and it was mainly from northeast/northwest (51.2 %–76.5 %) when the air quality deteriorated. At the SE-NCP site (LH), BC was dominantly from the south (47.8 %) during pollution episodes. At the CC site (WH), BC was mainly from the northeast (49.3 %–71.1 %). These results suggested that the northwest and northeast directions were the main transport pathways for air pollutants reaching WH during pollution episodes. Furthermore, to control local emissions during haze episodes, the emission sources (i.e., industry and open biomass burning) in the upwind direction should also be controlled to prevent the further deterioration of air quality in downwind areas.

To explore BC variations (i.e., mass concentration, sources and AAE) during transportation, we chose two case studies. LH and HA were selected due to the fact that the same instruments (AE33) were deployed at the sites, and the sites are representative of both the SE-NCP and CC. BC transportation from HA to LH and from LH to HA were both considered. Figure 13a shows the hourly backward trajectories reaching HA on 12 January 2018 and the trajectory at 13:00 (GMT; black line) as it was observed passing through LH (28 h prior). Therefore, the eBC mass concentration (including BCff and BCbb), σabs and AAE at the upwind site, LH, at 08:00 on 11 January 2018 (GMT) and the downwind site, HA, at 13:00 on 13 January 2018 (GMT) were compared (Fig. 13b). In Case 1, eBC, BCff and BCbb significantly increased (p<0.01) during the air transport from LH to HA. A BC absorption enhancement from 25.6±0.81 Mm-1 (LH) to 61.8±12.5 Mm-1 (HA) was also observed, whereas the AAE significantly decreased from 1.49±0.02 to 1.42±0.02 (p<0.01). Similarly, in Case 2, the air masses reaching LH at 07:00 on 13 January 2018 (GMT) had also passed through HA (black line) 31 h prior (Fig. 13c). The eBC, BCff and σabs increased during transport from upwind (HA) to downwind (LH), whereas BCbb and AAE decreased from 2.37±0.23 and 1.43±0.02 to 2.14±0.14 and 1.32±0.01 µg m-3, respectively (Fig. 13d). The eBC mass concentrations were enhanced during transportation regardless of the transport direction (CC to NCP or NCP to CC). Atmospheric removal of BC generally occurs in a few days to weeks via wet and dry deposition or contact with surfaces (Bond et al., 2013). In the two abovementioned cases, there were no precipitation events and the transport time was short (i.e., 28 and 31 h, respectively), which suggested lower removal rates. Therefore, new emission inputs along the trajectory enhanced the eBC mass concentration during transport. However, slight differences were found with respect to BCbb transport: BCbb increased from north (LH: 1.28±0.06 µg m-3) to south (HA: 2.57±0.47 µg m-3), whereas BCbb decreased from HA (2.37±0.23 µg m-3) to LH (2.14±0.14 µg m-3). This difference suggested that there were more intensive biomass burning emissions in Henan Province than in Hubei Province, which was also verified by the BC emission inventory (Qin and Xie, 2012; Qiu et al., 2016).

A previous study found that BC coagulation with nonrefractory materials becomes more significant when the aging timescale is greater than 10 h (Riemer et al., 2004). Chamber studies and field observations have also found that BC absorption is enhanced under polluted urban ambient air (Peng et al., 2016; Zhang et al., 2018; Y. Wang et al., 2018), suggesting that aging plays a role in modifying BC optical properties. In these two cases, the traveling time (aging time) from LH to HA and from HA to LH was 28 and 31 h, respectively, which suggested that the BC particles would have been coagulated through complex atmospheric processes. Therefore, the σabs was found to be higher in the downward direction. On the contrary, AAE values were found to decreased during transport. The AAE is sensitive to the particle size. A lab combustion experiment showed that particles with smaller diameters from fresh biomass burning have higher AAE values than larger particles (Singh et al., 2016). Simulations have also confirmed that the AAE of BC particles decrease with increasing geometric median diameter (C. Liu et al., 2018). Therefore, the diameter of BC particles increases during transportation due to the aging processes supported by the increased absorption coefficients and decreased AAE as discussed above.