A black carbon peak in the free troposphere of Beijing induced by cyclone lifting and transport from Central China

Observations suggest that the vertical distributions of air pollutants, such as black carbon (BC), present as various types depending on the emission sources and meteorological diffusion conditions. However, the formation process and source appointment of some special BC profiles are not fully understood. In this paper, by using the Weather Research and Forecasting model coupled with chemistry (WRF-Chem) with a BC-tagging technique, we investigate the formation mechanism and regional sources of a BC peak in the free troposphere observed by aircraft flight in Beijing (BJ) on May 5th, 2


Introduction
Black carbon aerosol (BC) has been a research hotspot in recent years owing to its significant environmental and climate effects (IPCC, 2013). Unlike scattering aerosols (sulfate, etc.), BC has a strong ability to absorb solar radiation, which accelerates global warming (Jacobson, 2001;Weingartner et al., 2003;Bond et al., 2013). Additionally, BC can heat the 35 upper air of the planetary boundary layer (PBL), inhibit its development, and promote regional air pollution (Ding et al., 2016). Air quality is affected not only by local emissions but also by long-distance transport (Zhang et al., 2017a). For example, BC in South China is mainly transported from Southeast Asia and North China (Fang et al., 2020), and most of the BC in the Arctic is transported from low-latitude regions (Keegan et al., 2014). Therefore, it is necessary to study the temporal and spatial distributions of BC and quantitatively analyse its regional sources to provide a reference for BC 40 emission reduction.
The annual carbon emissions in China account for approximately one-quarter of the global carbon emissions (Bond et al., 2004). As the capital of China, Beijing (BJ) has experienced frequent air pollution in recent years (Van Pinxteren et al., 2009). In BJ and its surrounding areas, straw is often burned in early summer; hence, BC pollution has become an environmental problem that cannot be ignored (Bergin et al., 2001;He et al., 2001). However, most of the previous 45 measurements of BC were performed at ground level, and the quantitative source tracking of BC was not thorough enough.
For example, based on radiocarbon ( 14 C) measurements, Zhang et al. (2015) found that fossil emissions dominated BC with a mean contribution of 75 ± 8% in BJ, but regional source research was lacking. To identify the regional sources, Liu et al. (2018) used a potential source contribution function (PSCF) model to determine that SX and HB were the main sources of BC in BJ but failed to quantify their separate contributions. Wen et al. (2020) used the Weather Research and Forecasting 50 model coupled to chemistry (WRF-Chem) to simulate the regional contribution and found that regional transport contributes more than 40% to BC in BJ, but there was no detailed delineation of external source regions.
In addition to research on horizontal BC, many vertical observations have also been conducted in recent years, showing that vertical distributions of BC are strongly affected by meteorological conditions and regional transport. The aircraft observation experiment carried out in BJ showed that regional transport of BC can enhance air pollution, and transport 55 occurs not only near the surface but also in the middle levels of the PBL (0.5-1 km) (Zhao et al., 2015). Moreover, when the BC profile has a homogeneous or negative gradient distribution, the diurnal evolution of the PBL is the leading factor. When there is a positive BC gradient from the surface to the top of the PBL, it is mainly caused by surrounding emissions from high stacks and regional transport (Lu et al., 2019;Shi et al., 2021). However, these studies were mostly qualitative inferences, and it is difficult to quantitate the formation process of vertical distributions and BC sources in detail. In addition, 60 the effect of BC on PBL varies greatly owing to its altitude. BC will inhibit the development of the PBL above the morning residual layer and conversely promote its development (Ma et al., 2020). Therefore, it is worth emphasizing that using only BC surface concentrations to calculate radiative forcing and heating rate and ignoring the vertical distribution of BC will induce great uncertainties (Shi et al., 2021). are not sufficiently separated; additionally, research on the reasons and source tracking of the vertical distribution of BC is even scarcer. Based on the BC profile observed in BJ, this study uses the air quality model WRF-Chem with a BC-tagging technique to track the sources of BC, with the hope to explain the observed BC peak in the free troposphere and evaluate BC sources and direct BC regional emission reduction measures.

Model configuration
The model used in this study, WRF-Chem 3.9.1.1, is coupled with a BC-tagging technique , which is similar to the particle source analysis technology in the Comprehensive Air Quality Model with Extensions (CAMx) model.  Table 1, respectively (the source regions with less contribution will be merged in the subsequent analysis, defined as Others). The BC of each source region is regarded as an independent variable, and it is marked from the beginning of the BC emission, ∆Emisi. The concentration of the total emissions in region i is defined as follows: ∆Emis i (x, y, z) = ∆Emis i (x, y, z) 0 If BC is in region i, then ∆Emisi (x, y, z) = ∆Emis (x, y, z); otherwise, ∆Emisi (x, y, z) = 0. Subsequently, the newly defined 80 variables will participate in the physical and chemical calculation process in the model accompanied by the origin variables in the model. Thus, we can obtain the BC concentrations of each source region at any grid of the model and at any time.
Compared with sensitive experiments, this method can more accurately quantify the sources of BC with fewer errors.
Previous studies have used a similar technique to study the source of air pollutants, such as PM2.5 and O3 (Gao et al., 2016;Zhang et al., 2017b;Gao et al., 2020). from the surface up to a pressure limit at 50 hPa, where 10 layers were located below 1 km. Notably, the carbon bond mechanism Z (CBM-Z, Zaveri and Peters, 1999) was applied as the gas-phase chemical mechanism in this study.

105
Correspondingly, the Model for Simulating Aerosol Interactions and Chemistry with eight bins (MOSAIC-8bins; Zaveri et al., 2008) was chosen as the aerosol chemistry mechanism. Other parameterization settings are listed in Table 2.

Model validation
Although the WRF-Chem model is widely used in air quality research, there are significant differences in the simulation 110 results with different parameterization schemes. In this study, time series of both simulated and observed PM2.5, BC concentration, temperature (T) and wind (Ua and Va) are shown in Fig. 2 to evaluate the performance of the model. Figure 2 illustrates that the model reproduces the numerical magnitude and variation characteristics of PM2.5, BC and meteorological factors well. Additionally, we calculate several common metrics (correlation coefficient (R), index of agreement (IOA), mean bias (MB), root mean square error (RMSE), mean normalized bias (MNB), mean fractional bias (MFB) and total error 115 (TE)) to validate the model performance on meteorological factors and air pollutants. The benchmarks shown in brackets in Table 3 follow the recommended values suggested by Emery and Tai (2001) and EPA (2007). The validation of the modelling profiles of BC and meteorological factors is also presented in section 3.2.2.
As shown in Table 3, meteorological factors (T, Ua and Va) showed high values of the mean IOAs (≥ 0.75), indicating that the simulation agreed well with the observations. In addition, the MB and RMSE of meteorological elements were 120 comparable with those in another modelling study  and were both within the threshold range. However, the MNB and MFB of Ua were beyond the scope of its benchmark, suggesting an overestimation of Ua. For air pollutants, good agreement was found between the simulations and observations since the IOAs of PM2.5 and BC were 0.93 and 0.70, respectively. The MFB of PM2.5 was within the scope of its benchmark. Notably, all the metrics of air pollutants also indicated that the model performance for air pollutants was acceptable. Generally, the model reproduces the 125 pollutants and main meteorological elements well, which provides a good basis for subsequent analysis.

Tracking observed BC from the surface to the free troposphere
We conducted aircraft observations in BJ from 10:00 to 11:00 on May 5th, 2018 (Beijing Time, BJT). The maximum flight altitude of the aircraft was 6000 m (Fig. 3a), and high BC values were observed both near the ground and in the free troposphere (~4000 m) (Fig. 3b). To reveal the cause of the distinct BC profile, the WRF-Chem model with a BC-tagging 130 technique was used for quantitative explanation.

Tracking BC sources at the surface
The near-surface pollution in BJ lasted from May 5th to 6th (Fig. 2). In the synoptic chart of Fig. 4a, there was a weak cyclone system in SX and NWCHN at 17:00 BJT on May 4th, leading to convergence and upward movement there. At 08:00 BJT on May 5th (Fig. 4b), the regions controlled by the weak cyclone system had expanded, including SX, NWCHN, HB 135 and HN, while BJ was in front of the weak low pressure. There was convergence near the surface, and air pollutants in the surrounding areas were likely to accumulate, leading to an increase in BC concentration in BJ ( Fig. 4d and e). Moreover, there was an obvious 'transport channel' from SX to BJ in the free troposphere (Fig. 4h), and BC was transported from SX to BJ by westerly winds. Subsequently, the low-pressure system moved south, and the BJ was controlled by the uniform pressure field at 08:00 BJT on May 6th (Fig. 4c), which was not beneficial for the diffusion of BC in the horizontal and 140 vertical directions (Fig. 4f). At an altitude of approximately 4000 m, BC was transported southwards due to clean north winds (Fig. 4i).
Moreover, the sources of BC were quantitatively traced (Fig. 5) to explain the cause of pollution in BJ. The mean BC concentration was 2.29 g·m -3 in this case (Fig. 5a), and the diurnal evolution of BC showed a characteristic of 'lower in the daytime and higher in the nighttime', which is related to the daily change in the PBL (Huang et al., 2018). Furthermore, BC 145 near the surface mainly originated from BJ, HB, HN, TJ and SD, whose mean contribution rates were 52.0%, 23.9%, 6.1%, 5.7% and 4.4%, respectively (Fig. 5b). The BC contribution rate of the BTH region to BJ exceeded 80%, further confirming that when controlled by weak low pressure, air pollutants in surrounding areas are likely to accumulate in BJ, and local pollutants have difficulty diffusing outward (Chen et al., 2008).

150
Near-surface BC mainly originated from BJ and its surrounding areas, while the source of BC in the free troposphere (~4000 m) showed different characteristics. The model grasped the key features of vertical BC, T, relative humidity (RH) and upper wind wells (Fig. 6a-d). The BC concentration presents a decreasing trend from the ground to approximately 700 m but increases from approximately 3000 m, forming a peak at an altitude of approximately 4000 m (Fig. 6a). As shown in Fig. 6e, the total contribution rate of BJ and HB below 700 m was as high as 96.7%, which is related to the easterly wind near the 155 surface (Fig. 6d). For the BC peak in the free troposphere (~4000 m), almost all BC was transported from external regions, including SX, NWCHN, SWCHN, HB, HN, and even SCHN, and the contribution rates were 24.7%, 23.5%, 10.5%, 9.4%, 8.9 and 7.1%, respectively. In addition, Fig. 6d illustrates that this altitude was mainly controlled by westerly winds (~225°), and the wind speeds increased significantly above 3000 m, which is beneficial to the long-distance transport of BC.

Air mass trajectory analysis
To determine the source of BC in depth and to validate the source tracing results above, we utilized the Hybrid Single

Particle Lagrangian Integrated Trajectory (HYSPLIT) model, developed by the National Oceanic and Atmospheric
Administration (NOAA), to analyse the air mass backward trajectory (24 h) at 08:00 BJT on May 5th (Fig. 7). As presented in Fig. 7, the air mass near the surface (red line) originated from BJ and its surrounding areas, which is consistent with the 165 source tracking results (Fig. 6). The upper air mass (blue line) originated in SX and reached BJ via HB, corresponding to the results of the cross section (Fig. 8). In addition, the blue line indicates that from 08:00 BJT on May 4th to 08:00 BJT on May 5th, there was an air mass lifting from the surface in SX along a southwestern path and reaching the upper levels of BJ (~4000 m). During this time, the China central (including SX, HB and HN) was controlled by a cyclone system ( Fig. 4a and   b), leading to convergent and elevating motion there, which is conducive to the uplift of near-surface air masses and 170 consistent with the backward trajectory model results.

Tracking BC sources in the cross section
In addition to the backward trajectory, the BC sources in the vertical cross section along the path from SX to BJ (in Figs. 1 and 7) can more convincingly explain the BC uplift and transport phenomenon. Referring to the backward trajectory, the white line shown in Fig. 1b can be regarded as the transport path of the air mass. To illustrate the uplift and transport 175 mechanism of BC, cross sections of BC concentration, wind vector, and BC contour lines (including BJ, SX and NWCHN) along the aforementioned line are shown in Fig. 8. From 08:00 to 18:00 BJT on May 4th, the southwest wind prevailed along the SX-BJ line ( Fig. 8a and b), which contributed to the transport of BC. BC originated in the NWCHN transported to SX and mixed with BC emitted in the SX region. By 16:00 BJT on May 4th, the ascending movement in SX was vigorous, resulting in the elevation of near-surface BC (Fig. 8b), which is consistent with the initial uplifting position presented in Fig.   180 7. BC originating from NWCHN and SX was uplifted up to 3000 m and transported to HB and BJ to form high values (Fig.   8c). However, the wind direction near the surface changed from westerly to easterly, while westerly winds still prevailed in the upper air (Fig. 8d). As a result, near-surface BC in BJ diffused to the surrounding areas, such as HB, but in the upper air, BC originating in NWCHN and SX was still transported to BJ and lifted further to approximately 4000 m (Fig. 8d). Thus far, there were high BC values both near the surface and in the free troposphere above BJ (~4000 m), which is consistent with 185 the observation and source tracing results in Fig. 6.

Physical process analysis
In this section, we further quantified the dominant transport processes of BC in the source region (SX) and receptor region (BJ), including horizontal advection (HADV), vertical advection (VADV) and vertical mixing (VMIX). In Fig. 9, the results show that HADV and VADV played an important role in the convergent and upward movement induced by cyclones and the 190 transport of BC. From 14:00 to approximately 18:00 BJT on May 4 th , in the boundary layer (<2 km), HADV made positive contributions to BC concentration in SX, while the VADV had the opposite effect, suggesting that convergent and upward motion existed therein ( Fig. 9a and b). At the same time, in the free troposphere (~4 km), the contribution of VADV and HADV to BC concentration was opposite to that in the boundary layer, indicating that BC originating in SX was lifted from the surface to the free troposphere by VADV ( Fig. 9a and b), which is consistent with the cross-sectional analysis (

4 Conclusions
This paper utilizes the air quality model WRF-Chem with a BC-tagging technique to study the formation mechanism of a special BC profile observed by an aircraft flight in BJ. The major findings are summarized as follows: In this case, the mean BC concentration was 2.29 g·m -3 . BC at the surface mainly originated from BJ, HB, HN, TJ and SD, with contribution rates of 52.0%, 23.9%, 6.1%, 5.7% and 4.4%, respectively. The BC contribution rate of the BTH region to 205 BJ exceeded 80%, further confirming that when controlled by weak low pressure, air pollutants in surrounding areas are likely to accumulate in BJ, and local air pollutants have difficulty diffusing outward.
Local sources dominated BC in BJ from the surface to approximately 700 m (78.5%), while BC was almost entirely imported from external sources (99.8%) in the free troposphere (~4000 m). BC in the free troposphere mainly originated from SX, HB, HN, and NWCHN, and the contribution rates were 24. 7%, 9.4%, 8.9% and 23.5%, respectively. 210 Figure 10 illustrates the formation mechanism of the special BC profile. HADV and VADV processes played an important role in the convergent and upward movement and the transport of BC. Near-surface BC that originated from SX, HB, HN and NWCHN was uplifted by a cyclone system approximately 16 hours previously, transported to a height of approximately 3000 m above BJ, and then lifted by the VADV process to approximately 4000 m. At the surface, BJ and its surroundings were in the field of a weak pressure gradient, leading to the accumulation of BC.