2.1 WRF-Chem model configuration

Version 3.8.1 of the WRF-Chem model (Grell et al., 2005) was adopted in this study to simulate the emission, chemical transformation, and deposition of aerosols, as well as their interactions with radiation. We demonstrated in previous studies (Gao et al., 2016a, b, 2020b, c) that the spatiotemporal variations of air pollutants over China could be reproduced effectively by WRF-Chem. WRF-Chem enables multiple options for gas phase chemistry and aerosol modules (Grell et al., 2005). We employed the carbon bond mechanism version Z (CBMZ) gas phase chemistry (Zaveri and Peters, 1999) coupled with the model for simulating aerosol interactions and chemistry (MOSAIC) (Zaveri et al., 2008) aerosol module in this study. MOSAIC treats size-resolved aerosol species, and we used 8 bins version in this study, corresponding to the particle diameter ranges of 0.039–0.078, 0.078–0.156, 0.156–0.312, 0.312–0.625, 0.625–1.25, 1.25–2.5, 2.5–5.0, 5.0–10.0 µm, respectively. Secondary organic aerosol (SOA) formation in MOSAIC was simulated with a volatility basis set (VBS) (Shrivastava et al., 2011). We configured two nested domains with horizontal resolutions of 81 and 27 km, and 31 vertical layers up to a pressure level of 50 hPa. The configured domains cover most areas of East Asia and focus on the North China region (same as Fig. 1 in Gao et al., 2017). Other chosen options for key physical parameterizations follow Gao et al. (2016b). Meteorological initial and boundary conditions were provided by the NCEP $\mathrm{1}{}^{\circ }\times \mathrm{1}{}^{\circ }$ degree final reanalysis data set (FNL), and chemical initial and boundary conditions were obtained from the MOZART global chemistry simulations (Emmons et al., 2010). We used the YSU scheme in WRF-Chem to simulate the evolution of boundary layer, and the YSU scheme defines the top of the boundary layer using the Richardson number (Hong et al., 2010). To allow the effects of aerosol on meteorological conditions in the model, we did not apply observational nudging or reanalysis nudging.

Figure 1Scatter plots of modeled and observed near-surface meteorological variables (a: T, b: RH, c: wind speed); modeled AAOD with the core-shell model and an external mixing assumption, and the comparisons against observations (d); modeled and observed AOD (e).

Download

Anthropogenic emissions of particles and gases in China in the model were taken from the multiresolution emission inventory for China (MEIC) for year 2014 developed by Tsinghua University (Zheng et al., 2018). Anthropogenic emissions for areas outside China were obtained from the MIX Asian emission inventory developed for MICS-Asia and HTAP, which combines five emission inventories for Asia (Li et al., 2017). Both MEIC and MIX data sets provide monthly emissions of air pollutants at $\mathrm{0.25}{}^{\circ }\times \mathrm{0.25}{}^{\circ }$ grids, which were interpolated to WRF-Chem modeling domains in this study. We adopted the MEGAN model version 2.04 to estimate biogenic emissions of gases and particles online (Guenther et al., 2006). The global fire emissions database version 4 (GFEDv4) (Giglio et al., 2013) was used for open fire emissions.

We simulated the period from 16 October to 13 November and discarded the first seven days as spin-up to avoid the influence of initial conditions. To explore the influences of coordinated emission control measures on BC absorption, we conducted multiple sets of simulations, as described in Table 1. For the NOCTL experiments, simulations were conducted with no perturbations in emissions. For the CTL experiments, emissions of SO₂, NO_x, PM₁₀, PM_2.5, VOCs, and other species in Beijing were reduced by 39.2 %, 49.6 %, 66.6 %, 61.6 %, 33.6 %, and 50 %, respectively, over the November 3-11 period. Emissions in Inner Mongolia, Shanxi, Hebei, Tianjin, and Shandong were reduced by 35 %. These perturbation factors were taken from the BMEPB reports (Gao et al., 2017). The locations of these provinces are marked in Fig. 1 in Gao et al. (2017).

The influence of BC absorption under different assumptions, including external and core-shell mixing and with or without emission reductions (Δ_BC-Ext-NOCTL, Δ_BC-Ext-CTL, Δ_BC-CS-NOCTL, Δ_BC-CS-CTL, and Δ_BC-CYSN-CS), can be derived with Eqs. (1)–(5) below. The description of each simulation is documented in Table 1.

The influence of emission reductions during APEC on changes in the light absorption of BC and associated changes in meteorological and pollution conditions under external and core-shell mixing assumptions (Δ_emission-Ext and Δ_emission-CS) can be inferred with Eqs. (6)–(7) below. We use Eq. (8) to derive the impact of changed BC aging processes by comparing the differences between core-shell simulation and external mixing simulation. The influences of reduced coating due to emission control measures during APEC are calculated with Eq. (9). We use Eq. (10) to derive the influence of changes in the light absorption enhancement (E_ab) of BC.

Table 1Descriptions of model simulations.

Download Print Version | Download XLSX

2.2 Calculation of aerosol optical properties in WRF-Chem

WRF-Chem uses Mie theory to calculate layer aerosol optical depth (AOD), single scattering albedo (SSA), and asymmetry factor (g). First, the size parameter and spectral refractive index are used to calculate the Mie extinction efficiency Q_e. Then, the extinction coefficient σ_e(λ) is provided by the integral of Q_e with consideration of the geometric size of the particle (πr²) and the particle number size distribution n(r) (Eq. 11). σ_e(λ) is a equation of wavelength λ. Similarly, absorption coefficient σ_a(λ) and scattering coefficient σ_s(λ) can be obtained with Mie absorption efficiency Q_a and Mie scattering efficiency Q_s. The value of SSA can be calculated with Eq. (12) using σ_a(λ) and σ_s(λ).

The calculated optical properties vary with the assumption of mixing state of aerosols. For external mixing, each particle is assumed to be a single chemical species. There are several models proposed for internal mixing, and the commonly used ones include the volume averaging model and core-shell model. In the volume averaging model, all species are assumed to be well mixed, while the core-shell model assumes that BC is coated by a well-mixed shell of another species (Jacobson, 2001). The volume-weighted refractive index m is obtained with the equation below:

In Eq. (13), V_i denotes the volume of species i and m_i represents the refractive index of species i. For core-shell internal mixing, similar averaging processes are applied to the core and shell separately. The scattering efficiency, absorption efficiency, and asymmetry parameter are then obtained using the core-shell Mie theory documented in Ackerman and Toon (1981). Core-shell Mie calculations request the core radius, shell radius, refractive index of the core, and refractive index to the shell as inputs (Ackerman and Toon, 1981; Toon and Ackerman, 1981). The official version of WRF-Chem does not calculate the optical properties of aerosols with an external mixing assumption. To assess the influence of mixing with coating particles on the light absorption of BC, the estimated light absorption of pure BC is required. We modified the optical calculation module in WRF-Chem so that it does not mix BC with other chemical species in the calculation of optical properties. In the calculation of optical properties with an internal mixing assumption, the volume-weighted refractive index are inputs of the Mie code. In the calculation of the optical properties of BC with an external mixing assumption, we only allowed BC to go through the Mie code. The mass and particle number of BC, denoted in M_i and N_i in each bin i (1 through 8) are computed first, and the volume (V_i) is obtained then by dividing by the density of BC. During the calculation of physical diameter (Eq. 14), other chemical species are not considered. Accordingly, BC does not mix with other chemical species, and we designated it as optical properties of BC with an external mixing assumption here. It should be noted that this calculation is different from treatments of external mixing in other models where all particles are separated from each other. In our calculation, BC within each bin are internally mixed, although it does not mix with other chemical species. As the purpose of this study is to explore how coating particles on BC would affect the absorption of BC, this treatment would not be an issue.

2.3 Observations

Both observations of meteorological variables and air pollutants were used to evaluate the performance of the model over the APEC study period in Gao et al. (2017) and in this study. The meteorological measurements were retrieved from the National Centers for Environmental Information website (https://gis.ncdc.noaa.gov/maps/ncei#app=cdo, last access: 22 July 2021), which includes near-surface temperature, relative humidity (RH), wind speed, and wind direction. The hourly surface concentrations of PM_2.5 and daily PM_2.5 chemical compositions were measured at the Institute of Atmospheric Physics (IAP), Chinese Academy of Sciences (CAS) site (Z. Liu et al., 2017; Pan et al., 2012; Yang et al., 2020). We also obtained absorption aerosol optical depth (AAOD) from the AERONET network (Dubovik and King, 2000; Holben et al., 1998) to evaluate model performance. We derived AAOD at 440 nm based on an angstrom exponent relation (Schuster et al., 2006) to make it consistent with observations. Data from more than 500 sites across the world are provided online at the AERONET website (http://aeronet.gsfc.nasa.gov, last access: 22 July 2021).

3.1 Model evaluation

Model evaluation was conducted with surface observations of meteorological variables, PM_2.5, PM_2.5 chemical components, and AAOD. Data at two meteorological sites in urban Beijing were averaged, and were compared against the model values for the domain grid cell containing the monitoring site. Figure 1 indicates that the daily mean temperature and relative humidity (RH) are captured well by the model. Observed strong wind conditions are slightly underestimated, which is a common issue due to inaccurate land use inputs or other problems in the model (Gao et al., 2018a). Our previous investigation (Gao et al., 2017) suggested that temperature and RH were lower, and northerly winds became more frequent from before APEC compared with during APEC periods, contributing to pleasant air quality. Figure 2a displays the simulated and observed hourly PM_2.5 concentrations in urban Beijing. Before APEC, observed high PM_2.5 concentration is captured well by our model. During APEC, the NOCTL case overestimates PM_2.5 concentrations, while the CTL case exhibits better agreement with observations. Implementing emission reductions in the model lowers the mean bias of the model from 30.8 to −4.0 µg m⁻³. The performance of WRF-Chem in simulating wintertime PM_2.5 chemical compositions was explored extensively in our previous investigations (Gao et al., 2016b, 2018a). Similarly, measured high concentrations of inorganic aerosols (sulfate, nitrate, and ammonium) are underestimated, which could be partly due to missing sulfate formation pathways (Cheng et al., 2016). We used the updated version with heterogeneous sulfate formation (Gao et al., 2016a) to reduce the underestimation of sulfate in this study. Simulated BC concentration shows a high degree of consistency with observations, while OC is slightly underestimated due to large uncertainties in the current status of SOA modeling (Fig. 2f). In general, the temporal variations and magnitudes of air pollutants are represented effectively in our model. Figure 1d compares simulated AAOD with an external mixing assumption and a core-shell model against AERONET inferred AAOD during the APEC study period. AAOD simulated with an external mixing assumption exhibits much lower values than with observation. With the core-shell model, this underestimation is largely reduced. However, AAOD is still underestimated by the model, which might be caused by missing sources of absorbing particles in the model. Currently, the absorption of organics is not treated in the WRF-Chem model, which is likely to underestimate the light-absorbing capability of carbonaceous aerosols in the atmosphere (Andreae and Gelencsér, 2006). Uncertainties in the aerosol size distribution in emissions may also contribute to this mismatch between the model and observations (Matsui, 2016).

Figure 2Modeled and observed time series of PM_2.5 concentrations in urban Beijing (a); Scatter plots of modeled and observed near-surface daily mean concentrations of sulfate, nitrate, ammonium, BC (black carbon), and OC (organics) in Beijing (b–f).

Download

3.2 Reductions in the concentrations of BC and coating pollutants and changes in BC aging degree

Previously, the reductions of air pollutants were estimated by comparing concentrations of air pollutants during the APEC period with those during other periods. Given the differences in meteorological conditions, such a comparison is not able to indicate the influence of emission control measures. As displayed in Fig. 2a, the concentrations during 24–25 October can be two times of those during 26–27 October, although no emission reduction measures were implemented. Previously, we concluded that the meteorological conditions during the APEC week were generally favorable for good air quality compared to it during the week before the APEC week (Gao et al., 2017). Thus, we perturbed emissions in this study to examine how it would affect concentrations of air pollutants, including both BC and its coating pollutants. As displayed in Fig. 3a–b, mean concentrations of SO₂ and NO₂ in urban Beijing declined by 38.7 % and 36.3 %, respectively, in response to short-term emission control measures. Based on observations, Zhang et al. (2018) reported that SO₂ concentrations decreased by 35 % (67 %), and NO₂ concentrations decreased by 34 % (45 %) compared with that before (after) APEC. These declines in aerosol precursors would have modified secondary aerosol formation during the APEC week. Our model indicates that sulfate and nitrate declined by 40.0 % and 28.2 %, respectively. Given the slight underestimation of sulfate and nitrate, these values might have been moderately underestimated. Mass concentrations of BC declined by 34.6 %, while the abundance of OC in the atmosphere exhibited a larger reduction (44.2 %).

We used the ratio of the sum of pollutants (primary as well as secondary) to black carbon concentrations (rBC) to track the relative abundance of BC and non-BC particles, which is essential in the C-S calculation:

As shown in Fig. 3c, the impacts of emission reductions during APEC on rBC behave differently at different sizes. For ultrafine particles, emission reductions generally lower the aging degree of BC. This is consistent with the observational evidence that smaller BC cores show larger reductions in aging degree as a result of emission control measures during APEC (Zhang et al., 2018). As most secondary aerosols are in smaller sizes, the effect of emission reduction on BC aging is more significant for smaller particles. Zhang et al. (2018) reported only the changes in sizes below 0.2 µm, our modeling results suggest, however, that the aging degree of BC might be enhanced under emission reductions for relatively larger particles (Fig. 3c). The impact of emission reductions on rBC behaves differently near the surface and at higher layers (Fig. 3d). The aging degree is lowered in the CTL case near the surface, mainly due to reductions in coating materials. However, at layers higher than 200 m, the aging degree of BC increases with emission reductions. In-situ near-surface measurements also indicate that rBC was reduced during APEC, and the reduction was most likely caused by lower photochemical production (Zhang et al., 2018).

3.3 Changes in AAOD and the light-absorption enhancement (E_ab) of BC during APEC

rBC values describe the aging degree of BC, while the exploration of how emission reductions affect light absorption of BC requires a sophisticated calculation of optical properties of BC. Mie theory is commonly used to calculate the light absorption enhancement of BC (E_ab) from the lensing effect with a core-shell model. Zhang et al. (2018) estimated E_ab by dividing the light-absorption cross section of the whole BC-containing particle by that of the BC core at a certain wavelength. Here we follow the method in Curci et al. (2019), and calculate E_ab as the ratio of BC AAOD estimated assuming core-shell internal mixing to that calculated with an external mixing assumption (enhancement due to mixing with coating non-BC particles):

Inferred vertical profiles of layer E_ab values in the CTL and NOCTL scenarios are displayed in Fig. 3e. At the layers below 5 km, mean E_ab values are 1.96 and 1.95 for CTL and NOCTL scenarios, respectively. Below 500 m in the troposphere, emission reductions during APEC lower E_ab from 2.11 to 2.06. A previous study by Jacobson (2001) suggests a global average BC absorption enhancement factor of 2, which is consistent with the current study. However, a wide range of enhancement factors have been reported, from negligible (Cappa et al., 2012) to as high as 2.4 (Peng et al., 2016). D. Liu et al. (2017) pointed out that the enhancement factors depend on the particles' mass ratio of nonblack carbon to black carbon. Our model results also indicate that the reductions in light absorption enhancement of BC are concentrated at lower layers, while enhancement could happen at higher layers (Fig. 3e). This is consistent with the vertical profile of rBC where it decreases due to emission reductions near the surface, while increasing at higher layers (Fig. 3d).

Figure 3Mass concentrations of gaseous and condensed pollutants in the CTL and NOCTL cases (a–b); the distribution of rBC with sizes in the CTL and NOCTL cases (c); the distribution of rBC with height in the CTL and NOCTL cases (d) and the distribution of E_ab values with height in the CTL and NOCTL cases (e).

Download

Figure 4a presents the daytime mean (defined as the mean BC AAOD over the 10:00–17:00 time period) BC AAOD in Beijing inferred from simulations with different mixing assumptions and emission perturbations. In the NOCTL scenarios, BC AAOD simulated with a core-shell model exhibits higher values than those with an external mixing assumption (0.0220 for external and 0.0427 for core-shell). Due to reductions in emissions, these values decline to 0.0145 and 0.0283, respectively. Due to emission reductions (differences between the CTL and NOCTL scenarios), mean daytime BC AAOD decreases by 0.0075 during the APEC week, as a result of declines in mass concentration of BC (52.0 %, Table 2). However, the lensing effect of BC induces a further decline of 0.0069 (48.0 %, Table 2). The influence of the lensing effect is mostly due to the reductions in coating materials (39.4 %, Eq. 9, Table 2). The BC absorption enhancement (E_ab) factor decreased by 0.003 due to reductions in emissions (Fig. 4b). We further quantified that the reduced light absorption capability (E_ab) resulting from emission reductions during APEC contributed 3.2 % to the total reductions in AAOD (Eq. 10, Table 2).

Figure 4Daytime (10:00–17:00 local time) mean BC AAOD in Beijing inferred from different simulations (a) and the BC AAOD enhancement in the NOCTL and CTL scenarios (b); BC absorption induced changes in EPT (orange line indicates the vertical profile of EPT) (c), PBLH (d), O₃ (e), and PM_2.5 (f).

Download

Table 2The division of the impact of BC absorption into the impact of BC mass itself and the BC mixing state.

Download Print Version | Download XLSX

3.4 Influences on boundary layer process and air pollution

The vertical distribution of BC absorption plays an important role in modulating the temperature gradient and changing boundary layer meteorology (Ding et al., 2016). We conducted a series of numerical experiments to understand the influences of reshaped BC absorption due to emission reductions during APEC on boundary layer process and the formation of air pollution. Figure 4c illustrates the vertical profiles of BC absorption induced changes in the equivalent potential temperature (EPT), which is commonly used to indicate the stability of air in the atmosphere (Obremski et al., 1989). When EPT decreases with height, the atmosphere is unstable, and vertical motion and convection is likely to occur. In all experiments, BC absorption induces a positive impact on EPT in the air above ground acting to enhance the stability of the atmosphere (Fig. 4c). The maximum enhancement occurs at layers close to 1–2 km (Fig. 4c). At ∼ 2.6 km, the maximum ratio of changes with the core-shell model to those with an external mixing reach above 2.5, indicating the important effects of mixing state of BC in the upper boundary layer (Fig. 4c).

In urban Beijing, BC absorption induced mean changes of daytime planetary boundary layer height (PBLH) during the APEC week are −11.6 and −24.0 m for an external mixing and core-shell model, respectively (Fig. 4d). Under relatively clean conditions (CTL scenarios), these values change to −8.8 and −15.6 m for external mixing and the core-shell model for NOCTL emissions (Fig. 4d). Due to emission reductions, the impact of BC absorption on PBL inhibition decreases by 8.2 m on average during the APEC week (reduced emissions enhance PBLH by 8.2 m). The influence of reduced mass concentration of BC itself accounts for 35 % of the total changes, while the lensing effect of BC explains the rest (65 %, Table 2). The decreased coating due to emission reductions greatly contributes to the lensing effect of BC (47.4 %, Table 2).

The corresponding changes in daytime mean near-surface concentrations of O₃ and PM_2.5 in Beijing are displayed in Fig. 4e and f, respectively. The inhibited development of PBL due to BC absorption results in a higher abundance of PM_2.5 within the PBL (Ding et al., 2016; Gao et al., 2016b). Previously, we quantified that the cobenefits of reduced aerosol feedback could explain ∼ 11 % of the total decreases in PM_2.5 in Beijing during APEC. Here we focus on the light absorption of BC, and find that the lensing effect of BC decreases PM_2.5 concentration by 0.8 µg m⁻³ on average (Fig. 4f). On average, declines in BC mass concentration itself account for 64.3 % of the total impact of reduced light absorption of BC on PM_2.5, while 35.7 % is attributed to the lensing effect of BC. However, inhibited PBL development does not necessarily lead to enhanced levels of near-surface O₃, as the formation of O₃ is also affected by changes in aerosols and photolysis reactions above the ground (Chen et al., 2021; Lu et al., 2019). As displayed in Fig. 4e, near-surface O₃ concentrations in urban Beijing decrease in response to BC absorption.

The spatial distribution of Δ_emission-Ext and Δ_emission-CS in Fig. 5 reveals that external mixing and core-shell models estimate similar patterns of changes in AAOD, PBLH, near-surface O₃, and near-surface PM_2.5. However, the responses of these variables are larger in the core-shell model due to the lensing effects of coating materials. Reduced emissions of BC and its coating materials during APEC led to declined AAOD, less stabilized PBLH, decreased near-surface PM_2.5 concentrations, and enhanced near-surface O₃ concentrations in the North China Plain (Fig. 5). It was noted that PM_2.5 concentrations was enhanced in northwest China, particularly when we used the C-S model. This is related to absorption-modulated natural emissions of windblown dust.

Figure 5Spatial distribution of daytime (10:00–17:00 local time) mean BC AAOD (first row), and mean BC absorption induced changes after emission reductions (CTL minus NOCTL) in PBLH (second row), PM_2.5 (third row), and O₃ (fourth row); first, second, and third columns represent Δ_emission-Ext, Δ_emission-CS, and Δ_aging.

Figure 6Cross sections of daytime (10:00–17:00 local time) mean changes in BC absorption coefficient (first row, CTL minus NOCTL), E_ab (second row), BC absorption induced changes after emission reductions (CTL minus NOCTL) in PM_2.5 (third row); first, second, and third columns represent Δ_emission-Ext, Δ_emission-CS, and Δ_aging.

Figure 6a, b illustrates the cross sections in the northeast direction of changes in BC absorption coefficient due to emission reductions, as Beijing and polluted cities in South Hebei are covered. Pronounced declines are concentrated below 2 km, and the core-shell model estimates stronger reductions due to lensing effects (Fig. 6c). Emission control measures also reshaped the light absorption enhancement factor of BC, as indicated in Fig. 6f. Within the lower boundary layer, E_ab values were reduced with emission reductions during APEC. Light absorption of BC stabilizes boundary layer to accumulate PM_2.5, yet this effect is inhibited at lower emission levels during APEC. These relationships are reflected in Fig. 6g and h with negative changes in PM_2.5 near the ground.

The responses of O₃ to reduced light absorption of BC during APEC are in the opposite direction (Gao et al., 2018c), compared to those for PM_2.5. Strong absorption of BC tends to enhance photolysis above the aerosol layer, but to reduce photolysis near the ground. Figure 7d, g illustrate the changes in O₃ 1D and NO₂ photolysis rates with emission reductions inferred from an external mixing assumption. With emission control implemented, photolysis rates near the ground are enhanced due to lower light absorption of BC, while the photolysis rates above the aerosol layer are reduced. Similar patterns but with larger values are found using the core-shell model (Fig. 7e, h). The responses of O₃ are generally in line with the responses of O₃ 1D and NO₂ photolysis rates (Fig. 7a, b).

Figure 7Cross sections of daytime (10:00–17:00 local time) mean changes in BC absorption induced changes after emission reductions (CTL minus NOCTL) in O₃ (first row), O₃ 1D photolysis rate (second row), and NO₂ photolysis rate (third row); first, second and third columns represent Δ_emission-Ext, Δ_emission-CS, and Δ_aging.