Due to excessive anthropogenic emissions, heavy aerosol pollution episodes
(HPEs) often occur during winter in the Beijing–Tianjin–Hebei (BTH) area of
the North China Plain. Extensive observational studies have been carried out
to understand the causes of HPEs; however, few measurements of vertical
aerosol fluxes exist, despite them being the key to understanding vertical
aerosol mixing, specifically during weak turbulence stages in HPEs. In the
winter of 2016 and the spring of 2017 aerosol vertical mass fluxes were
measured by combining large aperture scintillometer (LAS) observations,
surface PM
Recently, due to the country's rapid development of industrialization and urbanization, China has experienced heavy aerosol pollution episodes, particularly in the Beijing, Tianjin, and Hebei (BTH) region, which is one of the most polluted areas in China (Zhang et al., 2012). The pollution episodes often last for a long duration in the BTH region and cover a wide area, particularly in winter; they also severely reduce near-ground visibility (Lei and Wuebbles, 2013) and can have detrimental effects on public health (He et al., 2018; Cao et al., 2012). This heavily polluted environment has received extensive attention in recent years, and many observational studies have been carried out (Zhong et al., 2018b; Sun et al., 2014; Wang et al., 2015; Guo et al., 2011; X. Y. Zhang et al., 2009; Huang et al., 2014). Modelling studies have also been performed to examine the regional transport of pollutants (Wang et al., 2014) and to study the important role of large-eddy convective turbulent mixing in the vertical transfer of pollutants from a field campaign in Beijing (Li et al., 2018). However, few studies on the turbulence contribution of the aerosol turbulent flux in the surface layer have been conducted.
Ground pollutant emissions are known as the main source of aerosols in the atmosphere. However, in previous studies, no measurements of ground emissions during heavy pollution events were collected. Surface emission data are currently required for model verification and pollution predictions, and these data are primarily obtained through emission inventories (Wu et al., 2012; Bond et al., 2004). The establishment of emission inventories is primarily based on emission activity and emission factor (EF) data (Akagi et al., 2011; Lu et al., 2011; Roden et al., 2006; Zhang and Tao, 2009). Emissions data are mainly obtained from statistical yearbooks (Q. Zhang et al., 2009). Some studies have used fixed EFs while others have implemented dynamic EFs (Bond et al., 2004; Q. Zhang et al., 2009). Many factors are considered in dynamic EFs, such as the size of a city, the degree of economic development, the type of fuel, the kind of technology, product energy consumption, the control technology, and so on, as well as estimates based on actual measured meteorological parameters and aerosol parameters (Chen et al., 2015; Karvosenoja et al., 2008; Shen et al., 2013). A numerical model has also been used to estimate average fleet emission factors in typical urban conditions (Ketzel et al., 2003; Krecl et al., 2018). The uncertainties in the emissions of primary aerosols for inventories are high due to the highly uncertain contributions from the residential sector (Li et al., 2017), and the error in aerosol fluxes based on the use of emission inventories is huge (Liu et al., 2017; Zheng et al., 2017). Emission inventories constructed using the EF method provide only the total emission amount of atmospheric pollutants within a region. However, the emission data should be gridded to a suitable scale for air quality modelling and pollution predictions. Thus, near-surface aerosol emission data with a higher temporal and spatial resolution are urgently needed.
Many methods have been used to obtain aerosol flux data. For the upward transport of aerosols near the surface layer, the aerodynamic approach was adopted in the early years. The aerosol concentration gradient at different heights was measured and then calculated based on the similarity theory of the near-surface layer or calculated by the boundary layer box model, which can be based on meteorological data (Ceburnis et al., 2016; Hourdin et al., 2015; Zhang and Li, 2014). The emission rates of bioaerosols were also estimated from spore counts and molecular tracers (Elbert et al., 2007). The abundance of microbes and meteorological data were measured, and an estimate may be derived from the sea–air exchange of microorganisms (Mayol et al., 2014).
With the use of instruments for measuring the number of aerosol particles
(for example, a condensational particle counter, abbreviated as CPC by TSI),
the eddy covariance (EC) method has been applied, and measurement of the
aerosol particle number flux has become possible (Buzorius et
al., 1998). The vertical turbulent flux of the aerosol particle number
density
Although measurements of urban aerosol particle number density fluxes have been collected, the current eddy covariance method only provides fluxes for the aerosol particle number density at a certain point. We know that the underlying surface of a city is very complex, and thus the aerosol particle flux is not homogeneous in the horizontal. For a complex underlying surface such as a city, these point measurements are not representative of wider area. Therefore, it is of great importance to design an aerosol flux measurement system with larger spatial representation.
The use of eddy covariance principles to measure sensible heat fluxes has been widely performed (Lee, 2004). Current sensible heat fluxes can also be obtained using a large aperture scintillometer (LAS) based on the light propagation theory and atmospheric surface layer similarity theory (Zeweldi et al., 2010). This configuration makes it possible to achieve aerosol mass flux measurements using the same principles. Recently, we measured the imaginary part of the atmospheric equivalent refractive index structure parameter based on the light propagation theory (Yuan et al., 2015). The results showed that the imaginary part of the atmospheric equivalent refractive index structure parameter is related to turbulent transport and the spatial distribution characteristics of aerosols. Experiments also showed that there is a strong correlation between the imaginary part of the atmospheric equivalent refractive index and the mass concentration of aerosol particles (Yuan et al., 2016). Thus, similar to the temperature structure parameter reflecting the sensible heat flux, the structural parameter of the imaginary part of the atmospheric equivalent refractive index can reveal the mass flux of aerosol particles. This paper attempts to measure the aerosol mass flux in the BTH area, especially during heavy aerosol pollution episodes.
Generally, based on the PM
To gain a deeper understanding of the interaction between atmospheric heavy
pollution and weather in the BTH region, joint observations have been
carried out in the BTH region since the winter of 2016 (Zhong et al.,
2018b, c; Wang et al., 2018; Shen et al., 2018). Based on
meteorological causes of the increase or decrease in PM
The second section of this paper introduces the theory of aerosol vertical turbulent flux measurements, the third section introduces the experiment, the fourth section gives the experimental results, and, finally, the conclusion and discussion are presented in the fifth section.
The principles for calculating the vertical flux of aerosol particles and the approach for calculating the friction velocity and characteristic temperature using the temperature and wind profiles is presented in the following subsections.
According to the micrometeorological principle (Stull, 1988), similar
to the estimation method of the sensible heat flux, the aerosol flux
In Eqs. (3) and (4),
The following describes the method to deduce the aerosol mass concentration
structure parameter
Although the aerosol particles are dispersed in the air, the macroscopic
behaviour of the gas-particle two-phase mixture is the same as if it is
perfectly continuous in structure and physical quantities, such as the mass
and refractive index associated with the matter contained within a given
small volume, which will be regarded as being spread continuously over that
volume. The aerosol particles and gases in the atmosphere can be considered
as an equivalent medium, and an atmospheric equivalent refractive index
(AERI)
For visible light, there is a robust linear relationship between the
variation of the real part of the AERI and the variation of the atmospheric
temperature, namely,
According to Eqs. (5) and (6), we have the following:
The measurement of relevant parameters is performed based on the light propagation theory. When light is transmitted in an equivalent medium, the AERI fluctuation will cause light fluctuations in light intensity. When the attenuation caused by scattering and absorption along the propagation path is very weak, light intensity fluctuation depends on the fluctuation of the real part of the AERI along the propagation path. When the attenuation caused by scattering and absorption along the propagation path is relatively strong, the light intensity fluctuation is also related to the fluctuation of the imaginary part of the AERI along the propagation path. With the spectral analysis method, the LAS light intensity fluctuations can be separated into the contributions of the real and imaginary parts of the AERI. The contribution of the real part of the AERI corresponds to the high frequencies, whereas the contribution of the imaginary part of the AERI corresponds to the low frequencies, suggesting that the variances resulting from the real and imaginary parts are independent. Therefore, the light intensity variances induced by the real and imaginary parts can be detected separately at high frequencies and low frequencies from the LAS measurements (Yuan et al., 2015). Thus, the real and imaginary structure parameters of the AERI can be calculated by our developed LAS.
So far, we have completed the estimation of the aerosol mass turbulent flux.
According to the previous derivation and analysis, there are two calculation
schemes for determining the aerosol mass flux:
According to Eqs. (10)–(12), the vertical turbulent flux of aerosol particles is related to the strength of turbulent fluctuations of temperature and aerosol mass concentration fluctuations.
In this study, data from a weather tower in the north of Beijing were used. The weather
tower is 6.1 km far from the CAMS site. The meteorological observation data
from the weather tower show that applicability of the Monin–Oubhov
similarity theory under stable conditions causes a significant error for
Based on the discussion above, the LAS technique is capable of determining the magnitude of the flux but not the sign. In general, the aerosols are very heterogeneous in space and the measured fluxes typically show large variation in magnitude, including the sign. Over the polluted areas, which behave as the source, the emissions presumably overwhelmingly exceed the deposition sinks (Ripamonti et al., 2013). Therefore, a rough quantification of the deposition sink would allow the conclusion that the sink term is negligible and the flux quantified by LAS can be assumed to represent the upward fluxes.
To calculate the aerosol vertical turbulent flux, according to Eq. (10), the
values of the friction velocity
Based on Eqs. (13)–(18), the friction velocity
Observations were collected at two locations (two rectangles in Fig. 1a)
from December 2016 to March 2017: a rural site in Gucheng (GC site), Hebei
Province and an urban site at the Chinese Academy of Meteorological Sciences
(CAMS site) in Beijing. The distance between the two locations is
approximately 100 km. According to the theoretical methods defined in the
preceding section, to estimate the aerosol turbulent flux, the ratio of the
aerosol mass to the imaginary part of the AERI, the ratio of the temperature
to the real part of the AERI, the real and imaginary parts of the
atmospheric equivalent refractive index structure parameter (AERISP,
Photographs of the measurement site.
Two sets of LASs developed by our research group were installed at the top of the building of the Chinese Academy of Meteorological Sciences (point A in Fig. 1b) and at the top of a two-storey building in the farm of the Central Meteorological Bureau of Gucheng Town, Baoding City (point D in Fig. 1c). The light intensity sampling frequency of the receiving end was 500 Hz, and a file was recorded every 20 min. Then, the real and imaginary parts of the AERISP were calculated.
In the CAMS site, the transmitter end of the LAS was placed on the roof of a
building on the east side of the Chinese Academy of Meteorological Sciences,
and the receiver end was placed at the top of the Chinese Academy of
Meteorological Sciences. The propagation path was along an east–west
direction. The distance between the two ends was 550 m as shown in Fig. 1d. The light beam passed over urban buildings, residential areas, and urban
roads. The beam height was 43 m. The average height of the building
below the beam was 24 m; thus, the zero-displacement was 18 m (
In the GC site (point D in Fig. 1c, namely, the LAS position) of Gucheng,
Baoding, Hebei, the transmitter of the LAS was placed on the roof of a
two-storey building with a height of 8 m, and the receiving end was located
in a room in a three-storey building on the west side of National Highway 107
at the same height as the transmitting end. The distance between the
transmitting end and the receiving end was 1670 m. The terrain between
the transmitting end and the receiving end was flat, with farmland, a
national road, and sporadic trees below the beam, as seen in Fig. 1e. Near
the light beam, there was a 30 m high meteorological observation tower,
in which the temperature, relative humidity (RH), and wind speed were
measured at five levels (1, 3, 8, 18, and 28 m). The friction speed
and characteristic temperature were calculated according to the temperature
wind speed profile. Visibility observations were made in Xushui District
near the LAS position (see point E in Fig. 1c). The PM
There are two types of variables, namely mean variables and fluctuation
variables. Mean variables include temperature, wind speed, wind direction,
PM
Peaks in the light intensity fluctuation data appear because the received signal quickly increases when the light signal is blocked, such as due to birds along the transmission path. The data processing program automatically determines this situation. When this happens, the current 20 min period is rejected. For the real and imaginary parts of the AERISP and the aerosol flux data, (a) 3 times the standard deviation (SD) of the anomaly and (b) 3 times the SD of the difference between adjacent moments (AMD) were determined. A trend of 2 h averages, namely, 6-point moving averages, was first obtained. Then, the difference between the measured value and the trend at each moment was calculated, and the mean and SD of the difference were also calculated. The data with differences from the trend exceeding 3 times the SD were considered as spikes. The method for judging 3 times the SD of the AMD was first to calculate the AMD and then to calculate the mean and SD of the AMDs. Any data whose AMD deviated from the mean of the AMD by more than 3 times the SD of the AMD were considered an error. Less than 5 % of the data were found to contain spikes or errors.
The data determined to be errors were supplemented with the average of the nearby observations. Of course, if data were missing over a long period, the missing gap could not be filled. For this situation, further gap-filling was not considered.
Other errors exist in the measurements using a LAS due to specific reasons (Moene et al., 2009); for example, the impact of the deviation of the shape of spectrum from von Karman's scheme and the intermittent variations in the characteristics of that spectrum on the LAS signal were not considered in this study.
First, the visibility and PM
To obtain the ratio of the atmospheric aerosol mass concentration to the
imaginary part of the AERI
The maximum PM
Relationship plot of aerosol mass concentration
Furthermore, Fig. 2a and b show that although there is a significant
scattering between PM
The following provides the results of the aerosol turbulent flux under typical weather conditions in Beijing and Baoding for the period from 10 to 17 March 2017.
To analyse the aerosol turbulent flux characteristics, we present the time
series of the conventional meteorological parameters. The time series of
temperature, RH, wind speed, wind direction, PM
Temporal variations in
The LAS at the CAMS site was located in the roughness layer, so the local similarity theory should in principle be applied to flux calculation. Because there was no measurement of wind speed and temperature profiles near the LAS measurement location, the friction velocity and characteristic temperature could not be calculated. We (Yuan et al, 2015) conducted a test experiment for aerosol vertical flux in Hefei, China, using free convection assumptions and local similarity theories to calculate aerosol fluxes. Comparison of the calculation results of the two methods shows that very unstable conditions account for about 62 % of the time, and the relative difference is about 5 %. Under weak unstable and stable conditions, the relative error is about 15 %.
From the aerosol flux time series given in Fig. 3h, the aerosol flux is
large at noon and small in the morning and at night, which is mainly because
of the strong convection at noon. However, large aerosol fluxes also
occurred on the nights of 11 and 12 March , which were related to high
wind speeds. The mean aerosol flux measured at this observation point during
this period was 0.0039 mg m
Similarly, Fig. 4a–d provide the time series of temperature, RH, wind
speed, and wind direction at 3 and 18 m for the GC site, and
Fig. 4e–h show the PM
Temporal variations in
Figure 4e shows the PM
Figure 4h shows the aerosol mass vertical flux changes over time. The
aerosol flux has a significant diurnal variation characteristic associated
with turbulent transport near the surface. The mean aerosol flux measured at
the GC site during this period was 0.0016 mg m
In the winter of 2016, there were several HPEs. A heavy pollution event began on 1 December 2016 and ended on 10 January 2017. Relevant observational experiments were performed in the Beijing and Baoding areas, including observations of meteorological parameters and aerosol parameters, to understand the causes of the heavy pollution.
According to the definition of HPEs and classification, there were seven TS stages in the 2016 winter heavy pollution event, and the AS stage appeared immediately after four TS stages. These included 00:00 LT on 1 December to 03:20 LT on 4 December, 18:40 LT on 15 December to 00:00 LT on 22 December, 00:00 LT on 29 December to 2 January, and 00:00 LT and 08:40 LT on 2 January to 00:00 LT on 5 January.
During this period, we used a LAS to conduct an observational study of the vertical aerosol flux in the GC site, which was from 00:00 LT on 1 December, 2016, to 00:00 LT on 22 December 2016. No corresponding observations were made at the Beijing site during this period. Here, we first discuss the observation results of the GC site, Baoding City, as shown in Fig. 5. Figure 5a shows the time series of the aerosol vertical turbulent flux. Figure 5b–g indicate the time series for the real and imaginary parts of the AERISP, the temperature and RH at 18 m, and the wind speed and direction. Purple curves indicate the TS stages, red curves show the AS stages, and grey curves show the RS stages.
Temporal variations in
According to Fig. 5a, in the TS stages and the RS stages, the aerosol flux
exhibited diurnal variations, while the AS stage did not show a diurnal
variation. There were some peaks in the TS stage. The average aerosol flux
of the TS stages was 0.00065 mg m
According to Fig. 5b–c, the imaginary structure parameters and the real structure parameters of the refractive index in the TS and RS stages exhibited diurnal variations, while the AS stage did not exhibit a diurnal variation. Figure 5d shows that except for the second AS event (22:00 LT on 19 December to 00:00 LT ON 22 December 2016), the temperature showed a diurnal variation. During the AS stage, the RH (see Fig. 5e) was close to 100 %, while the RH during the TS and RS stages was lower. Moreover, Fig. 5f shows that during this time, the wind speed was relatively weak, although the wind speed was slightly stronger on 5 December. As shown in Fig. 5g, during the TS and AS stages, southerly winds prevailed, while during the RS period, northerly winds prevailed. The high wind speed and convection in the TS and RS stages contributed to the upward transport of aerosol particles, whereas the low wind speed and stable stratification in the AS stage were not conducive to the upward transport of aerosol particles.
During the heavy pollution period from 1 December 2016 to 10 January 2017,
we did not conduct surface aerosol flux observations at the CAMS site. From
25 to 31 January, the pollution in the Beijing area also reached the
level of heavy pollution. During this HPE, a measurement of surface aerosol
fluxes at the CAMS site was conducted. Figure 6 shows the results of the
meteorological and pollutant observations for six days from 00:00 LT on
25 to 00:00 LT on 31 January 2017. According to Fig. 6, northerly winds
prevailed after noon on January 26, when the concentration of PM
Temporal variations in
According to the previous characteristics for the TS and AS stages, a period
of southerly winds can be determined as the TS stage. Thus, 27 January can
be designated as the TS stage, 28 January can be determined as the AS stage,
and 29 January can be determined as the RS stage. During Beijing's heavy
pollution event in January 2017 (25 to 31 January 2017), the mean aerosol
vertical flux in the TS stage was 0.0024 mg m
Even during heavy pollution events, the RH in Beijing was lower than in the outer suburbs. According to Fig. 6e, the RH exceeded 60 % in the period from 03:00 LT to 06:00 LT on 26 January, where the maximum value was 63 %, and the RH was less than 60 % in the remaining periods. In urban areas, when the RH is low, heavy pollution incidents can occur. In Beijing, during the AS stage, the vertical flux of aerosol was less than during the TS and RS stages.
During the winter of 2016 and the spring of 2017, HPEs frequently occurred in the BTH area.
This study investigated the aerosol vertical mass flux and compared its
magnitude during different stages of HPEs, including RSs, TSs, and ASs, in
two representative urban and rural sites, including the CAMS site in Beijing
and the GC site in Hebei Province. Based on the light propagation theory and
surface-layer similarity theory, the aerosol vertical mass flux was obtained
by combining LAS observations, surface PM
Based on our measurement results, it can be seen that from the TS to the AS, the aerosol vertical turbulent flux decreased, but the aerosol particle concentration with surface layer increased. It is inferred that in addition to the contribution of regional transport from upwind areas during the TS, suppression of vertical turbulence mixing confining aerosols to a shallow boundary layer increased accumulation.
In this study, the aerosol emission flux was also estimated in these two
rural and urban sites. Generally, compared with the emissions in spring, we
found that in winter, the near-ground emissions were weaker in suburban
areas and were similar in urban areas. In suburban areas, although the
aerosol concentrations were relatively high (Shen et al., 2018), the
upward emitted aerosol flux was smaller than in urban areas. During the ASs
of the HPEs, the aerosol emission flux from near-ground emission sources was
weaker than for the RSs and TSs at both the CAMS and GC sites, which
indicates that surface pollutant emissions are not the major cause of
explosive PM
Compared to the results (Yuan et al., 2016) from Hefei, China, a small to medium-sized provincial capital city in East China, the measured aerosol mass fluxes in Beijing are almost at the same amount. A series of measures and actions have been made for emission reduction in Beijing, and the main emission is from vehicles. The difference in aerosol mass flux may be small.
Due to the lack of necessary experimental conditions, such as meteorological towers and EC systems, current experimental results cannot be compared with EC methods. According to the literature data, the two methods have been compared indirectly, and the estimated aerosol flux under different measurement conditions is consistent in magnitude (Yuan et al., 2016). However, a direct comparison of the two methods is in development.
Compared with the EC method, the aerosol flux has high spatial representativeness based on the principle of light propagation, and there is no need to install a tall tower. However, the estimation of aerosol fluxes using the LAS method still has theoretical and practical deficiencies. At present, the LAS method for the aerosol flux regards the aerosol particles as ordinary scalar molecules. At the same time, based on the assumption of the equivalent medium, the imaginary part of the AERI is presumed to be proportional to the aerosol mass concentration. This is often not the case. The actual turbulence spectrum shape may deviate from the von Karman spectrum, and turbulence intermittent and scintillation saturation can also occur (Moene et al., 2009). The applicability of the near-surface layer similarity theory to the aerosol particle motion under stable layer conditions also has many problems. The formation of new particles and changes in aerosol particle size distribution also affect the scintillation in light propagation. There are also practical problems such as untimely maintenance, rainfall and low visibility, and platform vibrations required for observation. All these problems will cause errors in final estimates, so more theoretical and experimental research is needed.
Requests for data that support the findings of this study can be sent to rmyuan@ustc.edu.cn.
RY and XZ designed experiments and wrote the paper. RY, HL, YG, BS, YW, JZ, and XT carried out experiments. RY analysed experimental results. YL and ZG designed experiments and discussed the results.
The authors declare that they have no conflict of interest.
We also thank two anonymous reviewers for their constructive and helpful comments.
This research has been supported by the National Key Research and Development Program (grant no. 2016YFC0203306), and the National Natural Science Foundation of China (grant nos. 41775014 and 51677175).
This paper was edited by Tuukka Petäjä and reviewed by two anonymous referees.