Vertical Structure of Air Pollutant Transport Flux as Determined by Ground-Based Remote Sensing Observations in Fen-Wei Plain, China

Abstract

Air pollutant transport plays an important role in local air quality, but field observations of transport fluxes, especially their vertical distributions, are very limited. We characterized the vertical structures of transport fluxes in central Luoyang, Fen-Wei Plain, China, in winter based on observations of vertical air pollutant and wind profiles using multi-axis differential optical absorption spectroscopy (MAX-DOAS) and Doppler wind lidar, respectively. The northwest and the northeast are the two privileged wind directions. The wind direction and total transport scenarios were dominantly the northwest during clear days, turning to the northeast during the polluted days. Increased transport flux intensities of aerosol were found at altitudes below 400 m on heavily polluted days from the northeast to the southwest over the city. Considering pollution dependence on wind directions and speeds, surface-dominated northeast transport may contribute to local haze events. Northwest winds transporting clean air masses were dominant during clean periods and flux profiles characterized by high altitudes between 200 and 600 m in Luoyang. During the COVID-19 lockdown period in late January and February, clear reductions in transport flux were found for NO₂ from the northeast and for HCHO from the northwest, while the corresponding main transport altitude remained unchanged. Our findings provide better understandings of regional transport characteristics, especially at different altitudes.

1. Introduction

Severe haze events caused by particulate matter have occurred regularly in China during recent decades, bringing health hazards and reducing atmospheric visibility [1,2,3,4]. These events have drawn worldwide attention, with many studies reporting frequent severe pollution in China’s most developed regions [5,6]. Beside the influence of local emissions and further atmospheric chemical reactions [7], regional transport also plays a key role in pollution episodes [8,9]. Cross-boundary transport may transport precursors or pollutants from their source regions to another area [10], leading to some severe haze days in Beijing [11] and ozone pollution episodes in North China [12]. A better understanding of transport characteristics is of great importance for the joint control of regional pollution. Air pollution transport takes place near the ground, as well as in the upper boundary layer [13], making vertical profiles of pollutant and wind distributions (rather than in situ measurements alone) essential to further investigations of transport flux characteristics [10]. However, previous research mainly focused on the near-ground pollutants variations; vertical structures of transport characteristics and pollutants distribution still need further investigations.

Some approaches have been applied to the estimation of transport and its effect on local air pollution. Backward trajectory models based on the reanalysis meteorological field data provide the most probable transport pathway of the air mass before its arrival at the target location. They can help us to analyze the transport effect, such as the effect from emission source in pathway [14]. Since the models do not take pollutant concentrations into account, transport flux cannot be quantified this way [15]. Some researchers have estimated transport flux based on simulated meteorology fields and air pollutant concentrations from chemical models [16], but this cannot represent real situations due to a lack of actual measurements. Atmospheric conditions simulated by chemical models are limited by parametric schemes and existing atmospheric processes, such as aerosol–cloud–radiative processes [17], so they may not provide precise variations in pollutant and meteorological parameters. Ground-based instruments have been used to estimate vertical flux profiles in various ways, such as determining water vapor flux by combining a water vapor lidar and a wind lidar near Broomfield, Colorado [18], vertical aerosol flux by combining an aerosol Raman lidar and a Doppler wind lidar in Leipzig, Germany [19], and particulate matter transport flux within the mixing layer in Beijing by combining a ceilometer and a wind radar [20]. However, these studies mainly focused on particulate matter transport and did not consider other gaseous pollutants transport characteristics.

Vertical distributions of pollutants and the wind field are essential for the study of transport flux profiles. Multi-axis differential optical absorption spectroscopy (MAX-DOAS) is a ground-based remote sensing technique that can synchronously retrieve vertical profiles of aerosol and trace gases (such as NO₂, HCHO, SO₂, HONO and CHOCHO) in the troposphere (typically up to ~4 km). Doppler wind lidar is another important remote sensing technique to obtain the vertical wind profiles in target altitude range. MAX-DOAS and lidar have wide applications in scientific studies (e.g., the evolution of pollutants and vertical structure of boundary layer) [21,22,23,24,25], but research on transport flux by combining both two technologies has not been reported yet.

In this study, we characterized the vertical structure of four pollutants (aerosol, NO₂, HCHO and SO₂) and estimated the corresponding transport flux by combining MAX-DOAS and Doppler wind lidar measurements in Luoyang, Fen-Wei Plain (FWP), China. Increasing PM_2.5 concentrations, especially during autumn and winter, have been reported for FWP in recent studies [26,27]. The increasing trend is especially obvious in the Luoyang basin, as documented by both in situ and satellite measurements, drawing significant attention from the public [28]. Luoyang is located in the southeast edge of FWP, close to the megacity Zhengzhou in the North China Plain to the northeast, locating it in a transportation corridor pathway between two megalopolises, which provide a unique opportunity to study the transport flux and its corresponding effect on local air quality. In the following sections, different characteristics of vertical distribution and flux were obtained under different degrees of pollution. We tried to evaluate the transport characteristics and the transport pathways around Luoyang, China. We further investigated pathway transport scenarios during an extremely-low-emission period. This work estimated the variations in the transport flux profiles of four pollutants (including aerosol and precursors) in a single study. Our results also provide a better understanding of air pollution in FWP and can be used to design joint pollution prevention and control strategies at a regional scale.

2. Measurements and Methodology

2.1. Overview of the Measurement Station

Luoyang, which has >7 million habitants, is surrounded by the Mang Mountains to the north and the Wan’an Mountains and some hills to the south and east, forming an east–west basin whose orientation encourages pollutant transport from the east or west (Figure 1). MAX-DOAS and Doppler wind lidar were co-located on the roof of the 14-story HuanBao building (112.46° E, 34.67° N, ~60 m a.g.l.) and measurements were taken from 16 January to 8 March 2020. They were set next to each other within a range of less than 10 m. MAX-DOAS measurements provided the daytime (typically 08:00–18:00 local standard time) vertical profiles of NO₂, SO₂, HCHO and aerosol extinction, while the Doppler lidar derived the all-day 3D wind profiles in the boundary layer.

2.2. MAX-DOAS Measurements

A commercial MAX-DOAS instrument, Airyx SkySpec-2D (Airyx, Heidelberg, Germany, https://www.airyx.de/ (accessed on 10 September 2021), was used to continuously collect scattered sunlight with an azimuth angle of 350° with respect to the north. The MAX-DOAS contains outdoor and indoor parts. The outdoor part consists of a telescope unit to collect the scattered sunlight and two motors to control the elevation and azimuth angles of the telescope, respectively. The telescope field of view is less than 0.3°, while the angle accuracy of these two motors is less than 0.1°. The indoor part consists of two spectrometers (AvaSpec-ULS2048L-USB2) with built-in CCD chip (charge coupled device, Sony ILX511, with 2048 individual pixels) and a controlling computer. The UV spectrometer covers 300–409 nm with a resolution of 0.45 nm at 334 nm, while the VIS one covers 400–495 nm with a resolution of 0.41 nm at 436 nm. The sunlight collected by the telescope was coupled into the spectrometer through a prism reflector and quartz fibers.

The elevation angles were set to 1°, 2°, 3°, 4°, 5°, 6°, 8°, 10°, 15°, 30° and 90° to form a measurement sequence lasting ~11 min. The exposure time and number of scans for each measurement were adjusted automatically depending on the intensity of received scattered sunlight to achieve a similar signal-to-noise ratio. MAX-DOAS was controlled by computer and operated in unattended automatic mode collecting scattered sunlight during daytime and dark current and offset spectra at night. The pollutants’ vertical profiles were retrieved by spectral analysis and a profile inversion algorithm based on the Lambert–Beer law and the optimal estimation method.

2.2.1. DOAS Analysis of O₄, NO₂, HCHO and SO₂ Differential Slant Column Densities (dSCDs)

The scattered sunlight at different elevation angles measured by MAX-DOAS was firstly analyzed using the DOAS method based on the Lambert–Beer law [29]. To analyze a collected spectrum, a Fraunhofer reference spectrum (FRS) should be included in the analysis procedure, then the differential slant column densities (dSCDs) can be retrieved, which are the slant column densities difference between measured spectrum and FRS. In this study, the zenith sky spectrum of the same measurement sequence, that is, the last spectrum of each sequence, was chosen as FRS in the DOAS analysis. The QDOAS software package developed by the Royal Belgian Institute for Space Aeronomy (BIRA-IASB, Brussels, Belgium, http://uv-vis.aeronomie.be/software/QDOAS/ (accessed on 10 September 2021) was used to analyze the collected spectrum. The wavelength range, trace-gas absorption cross sections and many other parameters of the DOAS fit for O₄, NO₂, HCHO and SO₂ are listed in

Table 1. Summary of the DOAS analysis settings used for the O₄, NO₂, HCHO, SO₂ dSCDs retrieval.

Parameter	Cross Section	Species
		O4	NO2	HCHO	SO2
Fitting interval (nm)		338–370	338–370	322.5–358	307.5–330
NO2	Vandaele et al. [32] 220 K, 294 K, I0-correction (SCD of 1017 molecules/cm2)	✓	✓	✓(only 294 K)	✓(only 294 K)
SO2	Vandaele et al. [33], 298 K				✓
HCHO	Meller and Moortgat [34], 297 K	✓	✓	✓	✓
O3	Serdyuchenko et al. [35], 223 K, 243 K, I0-correction (SCD of 1020 molecules/cm2)	✓	✓	✓	✓
O4	Thalman and Volkamer [36], 293 K	✓	✓	✓
BrO	Fleischmann et al. [37], 223 K	✓	✓	✓
Ring	Ring spectra calculated with QDOAS according to Chance and Spurr [38]	✓	✓	✓	✓
Polynomial degree		Order 5	Order 5	Order 5	Order 5
Intensity offset		Constant	Constant	Order 1	Order 1
Wavelength calibration	Based on a high resolution solar reference spectrum (SAO2010 solar spectra) [39]

Solar I0-correction [40].

(O₄: collision complex of ground state oxygen molecules [30], its absorption features could be detected by DOAS method and further used in aerosol extinction retrieval [31]). A typical example of the DOAS analysis of a spectrum measured at 11:56 20 January 2020 local standard time (LST) at the elevation angle equal to 2 degrees is shown in Figure 2. The dSCDs results with root mean square (RMS) larger than $1\times {10}^{-3}$ (O₄, NO₂ and HCHO) and $2\times {10}^{-3}$ (SO₂) were not included in further analysis.

2.2.2. Vertical Profile Retrieval

Aerosol extinction and trace gases vertical profiles were retrieved from dSCDs using the profile retrieval algorithm [41,42,43], which is based on the optimal estimation method (OEM) [44]. The vertical concentration profile C(z) is retrieved from the dSCDs obtained from the DOAS analysis in one sequence by minimizing the cost function ${\chi }^2$ in Equation (1); then, the optimal a posteriori state vector x is the profile C(z).

$${\chi }^2={\left(y-F\left(x,b\right)\right)}^T{S}_{\epsilon }^{-1}\left(y-F\left(x,b\right)\right)+{\left(x-x_a\right)}^T{S}_a^{-1}\left(x-x_a\right)$$

(1)

where y is the measurement vector, which is the dSCDs in one measurement sequence, F(x, b) is a simulated vector simulated by the forward model libRadtran [45], which describes the function between measurement and atmospheric state, x is the posteriori state vector, b refers to other atmosphere parameters, i.e., vertical temperature, pressure profile and aerosol properties, which are necessary for forward model simulations, and $S_{\epsilon }$ and $S_a$ are covariance matrices of y and a priori vector x_a, which describe the uncertainties of the measurement and a priori, respectively. In this study, we set an exponential decreasing a priori with scaling height of 0.5 km for both aerosol extinction and trace gases profile retrieval and the atmosphere from surface to 4 km was divided into 25 layers, with a vertical resolution of 100 m from surface to 1 km and 200 m from 1 km to 4 km. Then, the aerosol optical depth (AOD) and tropospheric vertical column density (VCD) corresponding to the total amount of aerosol extinction and trace gases were calculated by integrating the concentrations of all layers. All profiles were validated before further analysis; the comparison results and details are shown in Appendix A.1. Validations of MAX-DOAS results.

2.3. Doppler Wind Lidar Measurements

The Doppler wind lidar (WindMast PBL) used in this study was jointly developed by the Ocean University of China and Qingdao Leice Transient Technology Co., Ltd., Qingdao, China. It operates at a wavelength of 1.5 um and measures the Doppler frequency shift of laser light that has been backscattered by particles in the atmosphere. The WindMast can retrieve near-surface temperature, pressure and 3D vertical wind profiles from 42 m to 3 km (a.g.l.) with a spatial resolution of 14 m and a temporal resolution of 1 min (further technical parameters are given in

Table 2. Parameters for the Doppler wind lidar.

Technical Index	Parameters
Wavelength	1.5 um
Temporal resolution	1 min
Spatial resolution	14 m (vertical)
Wind speed Measurement range	0–75 m/s
Detection height	30~3000 m
Wind speed accuracy	Less than 0.1 m/s
Wind direction accuracy	Less than 3°

). The 3D wind profile for each measurement was expressed by layers composed of a horizontal wind vector expressed by a wind speed and direction with respect to the north and a vertical wind speed expressed by a wind speed value with positive representing upward and negative representing downward. Invalid wind results were filtered according to the corresponding signal-to-noise ratio.

2.4. Flux Calculation

The calculation of regional transport flux requires pollutant mass concentration and wind profile data [46]. We calculated the transport flux from the combination of the pollutant concentration retrieved from MAX-DOAS and the corresponding wind speed observed by the Doppler wind lidar. Since the latter had a higher temporal and spatial resolution, the vertical wind speeds were temporal and the spatial vector averaged according to the resolution of MAX-DOAS results following Equations (2)–(4):

$$\{\begin{array}{l}u(z)=ws(z)\times \sin \left(wd\left(z\right)\times \frac{\pi }{{180}^{\circ }}\right)\\ v(z)=ws(z)\times \cos \left(wd\left(z\right)\times \frac{\pi }{{180}^{\circ }}\right)\end{array}$$

(2)

$$\{\begin{array}{l}\overline{u_j}=\frac{1}{z_2-z_1}{\displaystyle {\int }_{z_1}^{z_2}u(z)dz}\\ \overline{v_j}=\frac{1}{z_2-z_1}{\displaystyle {\int }_{z_1}^{z_2}v(z)dz}\\ \overline{w_j}=\frac{1}{z_2-z_1}{\displaystyle {\int }_{z_1}^{z_2}w(z)dz}\end{array}$$

(3)

$$\{\begin{array}{l}W{S}_j=\sqrt{{\overline{u_j}}^2+{\overline{v_j}}^2}\\ W{D}_j=\arctan \left(\frac{\overline{u_j}}{\overline{v_j}}\right)\\ WS{Z}_j=\overline{w_j}\end{array}$$

(4)

where $ws(z)$ and $wd(z)$ are the horizontal wind speed (m/s) and wind direction (degree, with respect to the north) at height $z$(m), respectively, while $w(z)$ is the vertical wind speed (m/s) at height $z$, $u(z)$ and $v(z)$ are the E–W and N–S portions of horizontal wind (m/s), respectively, $\overline{u_j}$ and $\overline{v_j}$ are averaged portions of wind speed (m/s) for the MAX-DOAS layer $j$, $\overline{w_j}$ is the averaged vertical wind (m/s), $W{S}_j$ and $W{D}_j$ are spatial averaged wind speed (m/s) and wind direction (degree, with respect to the north) for layer $j$, respectively, and $WS{Z}_j$ is the spatial-averaged vertical wind (m/s).

The transport flux was calculated by multiplying the vectors for pollutant concentration, wind speed and cross-sectional area [10]. Since we used wind and pollutant profiles at a single point (in the city center) rather than the information for 3D cross-sections, we defined the flux intensity through the concurrent measurement of pollutants and wind profiles. The flux intensity corresponding to layer j was thus calculated by Equation (5) [47,48]:

$$flux{\left(T\right)}_j=\overline{W{S}_j\times c_j}$$

(5)

where $W{S}_j$ and $c_j$ are the time series of the mean wind speed (horizontal flux using horizontal wind vector and vertical transport flux using vertical wind vector) and the mean retrieved pollutant concentration for layer j (km⁻¹ for aerosol and molecules/cm³ for other pollutants), respectively, and the bar means a temporal average within the time period. The flux unit is also the product of two units, wind and concentration. So, the flux intensity $flux{(T)}_j$ is defined by the ks⁻¹ for aerosol and $molecules/(c{m}^2\cdot s)$ for NO₂, HCHO and SO₂ (abbreviated as $mole/(c{m}^2\cdot s)$ in figures).

The aerosol extinction coefficient is an optical property that describes the extent to which light attenuation in the atmosphere is caused by aerosol. It correlates with particle mass depending on the aerosol type. The contributions of PM_2.5 chemical components to aerosol extinction were relatively constant for all pollution levels in FWP [49]. A linear relationship between them could be obtained in some previous studies [48,49] and our two-month observations (Figure A1a), so the transport flux intensity of aerosol (calculated based on aerosol extinction) could quantify the strength of regional transport.

2.5. Date from China National Environmental Monitoring Center (CNEMC)

The in situ monitoring data, including PM_2.5, PM₁₀, NO₂, SO₂ and O₃, were released from the website of the China National Environmental Monitoring Center (http://www.cnemc.cn/, accessed on 10 September 2021). The CNEMC network is composed of more than 1300 monitoring sites throughout China at present and automated monitoring systems were installed at each monitoring station, mainly composed of a series of sample analysis instruments from Thermo Fisher Scientific Inc., Waltham, MA, USA. Each station measures the near-surface concentration of six pollutants in near-real time automatically and the Air Quality Index (AQI) was calculated according to each pollutant’s concentration to describe the air quality status. The hourly data at each monitoring station were published from 2013 and they have been reported in detail in many previous studies [50,51]. The near-surface data of PM_2.5, PM₁₀, NO₂, SO₂, O₃ and AQI of Luoyang were used to analyze the temporal variation in the pollutants and validate the MAX-DOAS results. In this study, we filtered the released CNEMC data using the z-score method for quality assurance purposes before using the data in further analysis. More details for the z-score method are provided in Appendix A.2. Description of Z-score method.

3. Results

The transport flux observed at a boundary station could capture the cross-boundary transport effect on target location [10]. Single-point flux observations in the city center instantaneously monitor transport variations from all directions at a regional scale [20]. In this study, we obtained the transport flux profiles and further characterized the transport structures and transport pathway.

3.1. Temporal Variations of Pollutants in Boundary Layer

The daytime vertical distribution of pollutants from near surface to 4 km was divided into 25 layers, while AOD or VCD represented the total amount of aerosol or trace gases within the whole troposphere, respectively. To better understand temporal variations in pollutants at different altitudes, we chose the concentrations of the 0–100 m, 500–600 m and 900–1000 m layers (retrieved from MAX-DOAS) to represent the lower, middle and upper boundary layers, respectively, and determined the daily time series of aerosol extinction, NO₂, HCHO and SO₂ in the chosen layers from 16 January to 8 March (Figure 3).

The pollutant concentrations in each layer could be clearly distinguished throughout the study period, with the highest concentrations at the surface and the lowest in the upper boundary layer. The study period began with a severe haze episode, during which aerosol extinction and NO₂ reached 4.3 km⁻¹ and 35 ppbv, respectively. After this, all pollutants except for SO₂ gradually decreased, until the second haze episode started on 20 February. The two haze episodes observed during the study period were characterized by pronounced continuous aerosol extinction higher than 2 km⁻¹, as well as HCHO and NO₂ concentrations significantly higher than the mean value across the entire period. The haze episode from 20 to 29 January 2020 was defined as episode 1 (EP1) and that from 19 to 29 February 2020 as episode 2 (EP2). EP1 began with a rapid increase in NO₂ and aerosol extinction on 21 January, reaching a maximum of 4.3 km⁻¹ on 26 January. Compared to the preceding clean period, the surface, 500 m and 900 m layer aerosols, as well as AOD, increased by 188%, 143%, 129% and 162%, respectively. EP2 began on 19 February and peaked on 26 February, with a smaller increase in aerosol extinction than EP1, with increases in the surface, 500 m and 900 m layer aerosols, as well as AOD, of 170%, 120%, 122% and 142%, respectively. The time series of aerosol, NO₂ and HCHO exhibited a ‘U’ shape, but SO₂ varied little throughout the period, with a mean concentration of 3.5 ppbv near the surface and a mean VCD of 7.89 × 10¹⁵ molec/cm². These differences can be attributed to different sources; SO₂ mainly originates from primary emissions, such as emissions from household fuel burning and industrial power generation during wintertime [52], while particulate matter and NO₂ mainly originate from human activities and vehicle emissions [53].

3.2. Statistics of Wind and Pollutant Transport Characterization

Statistics of wind directions and wind speeds for the whole observation period (including day and night) are shown in Figure 4. The wind speed generally increased with altitude and the prevailing wind direction gradually turned northwest, meaning that the air quality in the higher boundary layer may be more easily affected by air pollutants from the northwest. In the near-surface layer, the wind blew most frequently from the northeast at less than 5 m/s, though west winds were also common (reaching 13.5 m/s). However, these varied in the middle and upper layers. In the former, prevailing winds from the northwest and the east had mean wind speeds of 6.6 and 4.9 m/s, respectively. Winds from the southeast and the northwest mainly contributed to the upper boundary layer.

Wind is a kind of physical atmospheric process that can transport atmospheric components [47], including pollutants or clean air mass. The northeast and northwest are two privileged wind directions, which indicate two potential transport pathways in the local area. We further plotted the polar plots of four pollutants during the observation period to investigate whether both two directions contribute pollutants to haze events in the local area (Figure 5). Bivariate polar plot is a common method used in source characterization [54]. It explores the relationship between the chemical constituents and meteorological parameters. Results suggest that transport plays an important role in local air quality. Higher concentration experienced with northeast winds indicate that wind from the northeast may contribute to local pollution events. Locally sourced pollution during low wind speed was also potentially indicated for all pollutants at the near-surface layer. As for middle and upper boundary layer, transport from the east or northeast play an even more import role than local emissions (except for SO₂). Transport effect from the east or northwest is obvious for SO₂ in 500 m latitude and local emissions or transport from other directions cannot be clearly distinguished in the upper boundary layer. Although two main privilege wind directions were found in the local area, air masses from the northeast direction may contribute more pollutants to local severe haze events.

3.3. Vertical Structure of Transport Flux from Privileged Wind Directions

Pollutant transport may take place near the ground, as well as in the upper boundary layer [13]. To investigate the vertical structures of transport flux from two major privileged wind directions, the transport fluxes from the northeast and northwest in different layers were gathered and averaged to obtain the transport flux profiles (Figure 6). The transport strength was generally smaller in the northwest direction and was mainly located in the middle boundary layer. The total transport fluxes in the boundary layer from the northeast were generally 60%, 72% and 56% larger than fluxes from the northwest for aerosol, NO₂ and HCHO, respectively, while fluxes for SO₂ were similar in both directions. The largest transport strength occurred at the near-surface layer for the northeast pathway, while the main transport altitude was between 200 and 600 m for the northwest pathways, which means the two privileged directions have different vertical transport structures. The structure differences between these two directions may be attributed to the meteorological conditions of the wind field. For each specific direction, more transport fractions in the middle boundary layer can be found in SO₂ profiles, which indicates that SO₂ distributed in the middle and upper boundary layers is more obvious than other pollutants (this point can also be proven in

Table A1. Statistics of vertical distribution of various pollutants.

	Aerosol Extinction	NO2	HCHO	SO2
surface	84.92%	79.72%	88.48%	53.96%
elevated	15.08%	20.28%	11.52%	46.04%

).

3.4. Structure of the Transport Flux under Different Pollution Degrees

Transport may play a more important role during severe haze events than on clear days [11]. To investigate the transport scenarios and characteristics under different pollution levels, we compared the mean concentration and flux profiles after grouping the total observation period into four levels, following a previous study on PM_2.5 concentrations [55]: PI (<35 ug/m³), PII (35–75 ug/m³), PIII (75–115 ug/m³) and PIV (>115 ug/m³). The wind direction was dominantly northwest during PI and PII, turning to the northeast during the PIII and PIV periods, indicating different privileged wind directions during clear and polluted days (Figure 7). The horizontal wind speeds during clear days (PI conditions) were higher than in other phases. The measurement site is located at an east–west basin (Figure 1). A narrow pathway in the northwest direction connects Luoyang with other cities in Fen-Wei Plain. The valley makes the wind blow at a relative high speed, while the North China Plain in the northeast direction may mostly cause the wind to blow at a lower speed.

The four pollutant profiles all exhibited an exponentially decreasing shape with altitude (Figure 8) and were mainly distributed near the ground. The maximum increase in the boundary layer was found near the surface for all pollutants. The near-surface aerosol extinction, NO₂, HCHO and SO₂ increased by 700%, 300%, 250% and 150%, respectively, when mean concentrations for PIV were compared with PI, indicating explosive pollutant growth during heavily polluted days. Different distributions of pollutions may be attributed to their different sources, chemical reactions and transport effect. Particulate matter and NO₂ mainly originate from human activities and vehicle emissions [51], causing an exponentially decreasing trend with altitude. SO₂ mainly originates from primary emissions, such as emissions from household fuel burning and industrial power generation during wintertime [52]. Long-range transport from other areas may also contribute to SO₂ at a higher altitude [56]. Therefore, more fractions in the upper boundary layer can be found for SO₂. As for HCHO, primary emissions (e.g., biomass and industrial emissions) and secondary generation (photochemical oxidation) are two important sources in the atmosphere [57,58], so it can be produced in higher altitude because of vertical transport of its precursors.

The mean transport flux profiles were calculated by the vector average and are shown in Figure 8b. These profiles represent the total transport scenarios across the local area during the whole period (the transport effect may get cancelled if two transport directions are opposite to each other during one divided period). The horizontal flux directions were often different or even opposite between clear days (PI and PII) and polluted days (PIII and PIV). Transport from the west to the east dominated clear days for all pollutants, but could turn northeast during haze days. Great transport flux from the northeast was found for aerosol below 400 m during the PIV period, which may indicate more particulate matters were transported from the northeast to the southwest over the city on heavily polluted days. The mean NO₂ transport flux showed a nearly exponential decreasing trend with altitude during the PII, PIII and PIV periods from the northeast direction. As for HCHO, the mean transport intensities were similar during clear (PI) and severely polluted days (PIV) with almost opposite directions. The vertical transport fluxes calculated from pollutant concentration and vertical wind speed (Figure 8c) showed a decreasing trend with altitude at all pollution levels. Higher vertical fluxes were found in the lower-most layer during polluted days, meaning that the upward transport in the lower boundary layer would increase during polluted days. Turbulence is another important atmospheric process which mixes the pollutants well in the boundary layer [59]. The vertical transport flux may be larger than the turbulent flux at the surface layer.

4. Discussion

4.1. Transport Flux Variations during Severe-Haze Days

We further investigated the transport flux temporal variations across a haze episode (EP1, 20 to 29 January 2020). The time series of aerosol extinction profiles and pollutants transport fluxes profiles are shown in Figure 9 and Figure 10; surface measurement data, including data on NO₂, SO₂ and PM_2.5 released by CNEMC and temperature and relative humidity measured by lidar are shown in Figure 11. Wind blowing from two privileged wind directions occurred at the start, middle and end of this episode, respectively, which caused increases in transport flux intensities. On 20 January, NO₂ (Figure 10a, 20 January) was transported from the northeast to the southwest at an altitude <500 m. The near-surface PM_2.5 concentrations also increased after the night on 20 January. This transport process may dominate local precursors or pollutant accumulation before the pollution event. Aerosol extinction began to rise with the increase in NO₂ within the boundary layer on 21 January. The transport process of pollutants, including aerosol (Figure 9b), NO₂ (Figure 10a), HCHO (Figure 10b) and SO₂ (Figure 10c), occurred from the northeast to the southwest (Figure 9c) on 24 January. During this period, the near-surface PM_2.5 concentrations significantly increased and reached their peak of the whole episode at noon on 24 January. The flux throughout the boundary layer was dominated by a maximum wind speed of >9 m/s from the northeast to the southwest. PM_2.5 reached its maximum concentration (>300 ug/m³) around 12:00 on 24 January, which lasted for roughly two days, while the air quality reached a severely polluted level. Increased flux intensities could be found from 28 to 29 January from the northwest, including aerosol, NO₂ and SO₂, characterized by high-altitude transport. The horizontal flux was dominated by winds from the northwest to the southeast with a mean speed >10 m/s. The main transport altitude was located at 100–500 m a.g.l.

Flux variations indicate that clear days were dominated by wind blowing from the northwest; significant northeast transport at the surface occurred at the beginning or in the middle of severe-haze days and was accompanied by an increase in near-surface PM_2.5 concentrations. Pollutants transported from the surrounding area before haze events may promote local pollutant accumulations [11,49]. Combining the analysis with polar plots of bivariate statistics (Figure 5), we could infer that transport at the surface from the northeast contribute to local haze events; the wind from the northwest may transport fresh air at the middle boundary layer. These results also make up for an inability to observe high-altitude transport processes [10]. Different contributions of wind from the northeast and the northwest were discussed through the analysis of flux variations during a severe haze event. Further improvements in local air quality should be based on the relocation of severely polluting enterprises according to the prevailing wind direction and introducing strong emission reduction measures according to air quality forecasts.

4.2. Transport Flux during the Low-Emission COVID-19 Period

In early 2020, the rapid spread of the novel coronavirus (COVID-19) led the Chinese government to ask citizens to reduce activities nationwide, encouraging them to reduce travel and stay at home [60,61,62,63]. These lockdown restrictions provided a unique opportunity to investigate pollutant variation and pathway transport scenarios under the resulting significant reduction in anthropogenic emission levels. COVID-19 lockdown restrictions began on the Chinese New Year (25 January 2020); we defined the observation period from 26 January to 17 February 2020 as the lockdown period and from 16 to 24 January and from 18 February to 8 March as non-lockdown periods.

To investigate the impact of emission reductions on transport fluxes from the northeast and the northwest during the lockdown period, the mean vertical transport flux profiles for the two directions were obtained (Figure 12). The NO₂ transport flux from the northeast was clearly reduced during the lockdown period, which can be attributed to the lockdown of the highly industrialized and urbanized megacity Zhengzhou and the associated reduction in traffic emissions. At the same time, the aerosol transport flux from the northeast increased. Huang et al. comprehensively investigated the mechanisms of enhanced aerosol episodes under the large decreases in primary pollution during the lockdown period and suggested that the large decrease in NOx emissions increased the atmospheric oxidizing capacity and, in turn, promoted the formation of secondary particulate matter (ozone could be used to assess the oxidizing capacity of the atmosphere; an increasing trend in surface O₃ concentration after 26 January can also be found in Figure 11) [60]. This is also consistent with our observations. As for the northwest direction, a reduced transport profile was found for HCHO. The area northwest of Luoyang is dominated by rural areas with few manufactories, the closure of which reduced the primary HCHO emissions and led to a decrease in the HCHO transport flux. The lockdown restrictions greatly reduced primary pollutant emissions for a regional and even national scale [61,62], with the changes in the transport profile being also associated with the background and corresponding land use situations along different directions, while the main transport altitudes remained unchanged.

5. Summary and Conclusions

We obtained transport flux profiles of four kinds of pollutants (aerosol, NO₂, HCHO and SO₂) during winter in Luoyang, Fen-Wei Plain, China, using ground-based remote sensing to gather pollutant concentrations and wind profiles rather than single-point in situ measurements at the near-surface layer. The maximum concentrations of all four pollutants were found in the near-surface layer. Prevailing winds came from the northwest and the northeast during the study period due to the specific topography of the Luoyang surrounding area. The wind direction and total transport scenarios were dominantly northwest during clear days, turning to the northeast during the polluted days. Particulate matters were transported from the northeast to the southwest over the city on heavily polluted days. As for the transport flux profiles from the northeast and the northwest directions, the largest transport strength was found at the near surface for the northeast direction and the main transport altitude was between 200 and 600 m for the northwest pathway. Considering pollution dependence on wind directions and speeds, surface-dominated northeast transport potentially contributes to local haze events and northwest winds transport clean air mass during clean periods. During the COVID-19 lockdown period, primary emissions were greatly reduced and transport fluxes declined for NO₂ from the northeast and HCHO from the northwest. A smaller transport flux increment was found for aerosols from the northeast, while the main transport altitudes for each direction remained unchanged. Our findings provide a more thorough understanding of regional transport at different altitudes under different degrees of pollution. Further improvements in the local air quality should be based on relocating severely polluting enterprises according to the prevailing wind direction and introducing strong emission reduction measures according to air quality forecasts.

Flux observations in city centers allow the monitoring of pollutant transport from all directions, clarifying the transport characteristics related to different pollution degrees and pathways. Considering the dominant wind directions and unknown pathway cross-sectional areas, future studies could seek to achieve higher-precision and higher-resolution flux observations for whole days along city boundaries and transport pathways to better quantify the total amount of transport along pathways.