Impact of modified turbulent diffusion of PM 2.5 aerosol in WRF-Chem simulations in Eastern China — Source link

Pollutant vertical mixing in the nocturnal boundary layer enhanced by density currents and low-level jets: two representative case studies Turbulent Diffusion and Turbulent Thermal Diffusion of Aerosols in Stratified Atmospheric Flows in stratified atmospheric flows Pollutant transport in a Convective Boundary Layer with LES Turbulent dispersion from an elevated line source: measurements of wind-concentration moments and budgets Abstract Correct description of the boundary layer mixing process of particle is an important prerequisite to understanding the mechanism of heavy pollution episodes. Turbulent mixing process of particles is usually denoted by the turbulent diffusion relationship of heat, meaning that the turbulent transport of particles and heat are similar. This similarity has, however, never been verified. Here we investigate the 15 dissimilarity between particles and heat, indicating that the unified treatment of all scalars in the model is questionable. Using mixing-length theory, the turbulent diffusion relationship of particle is established, embedded in the model and verified on a long-term scale. Simulated results of PM 2.5 concentration were improved by 8.3% (2013), 17% (2014), 11% (2015) and 11.7% (2017) in Eastern China, respectively. However, under the influence of complex topography, the turbulent diffusion process is insensitive to the 20 simulation of the pollutant concentration. In addition to the PM 2.5 concentration, the simulation of the CO concentration has also been improved, which shows that the turbulent diffusion process is extremely critical to the change in the concentration of pollutants. of transport between dust, heat and momentum. The only studies assumed that particles were considered passive scalars with the same 45 source/sink as heat, and that they used similarity to correct particle flux from the heat flux (Damay et al., 2009; Deventer, Held et al., 2015). A key question is whether the turbulent transport between temperature and particles is similar. This similarity has, however, never been verified, due to the lack of observational turbulence data of particles.


Introduction
Along with the intensive urbanization and tremendous economic development, numerous incidents of 25 aerosol pollution have frequently occurred in China (An et al., 2019;Q. Zhang et al., 2019;. Aerosol pollution, characterized by PM2.5, occurs primarily in the planetary boundary layer (PBL). The horizontal transportation and vertical distribution of pollutants are obviously affected by the PBL mixing process, associated with intricate turbulence eddies Ren et al., 2018;Wang et al., 2018;Du et al., 2020). Turbulent transport, as a vital process, controls the exchange of momentum, 30 heat, water vapor and pollutants through turbulence eddies within the PBL (Stull, 1988).
Turbulent transports of temperature, water vapor and CO2 has long been considered similar (Kays et al., 2005). However, this statement is usually invalid and is regarded as applicable only under neutral stratifications. Previous researchers have demonstrated that temperature-humidity dissimilarity, and such a disparity between the effectiveness of heat and water vapor transport, is due to different mechanisms 35 of scalar transport (Katul et al., 2008;van de Boer et al., 2014;Guo et al., 2020).For example, the effect of advection (Assouline et al., 2008), entrainment at the top of PBL (Cava et al., 2008;Gao et al., 2018) and heterogeneity in sources and sinks (Detto et al., 2008;Wang et al., 2014;Guo et al., 2016). Moriwaki and Kanda (2006) also indicated that the differences of turbulent transport between heat and CO2 were due to both by the active role of temperature and the heterogeneity of the source distribution. Li and Bou-40 Zeid (2011) revealed that the transport dissimilarity between the momentum and the scalar likely resulted from the topology of turbulent structures. As a result, there are differences between turbulent transport of vectors and scalars, or between scalars. However, less attention has been paid to turbulent transport of particles. Dupont et al. (2019) have proven that the turbulent dissimilarity of transport between dust, heat and momentum. The only studies assumed that particles were considered passive scalars with the same 45 source/sink as heat, and that they used similarity to correct particle flux from the heat flux (Damay et al., 2009;Deventer, Held et al., 2015). A key question is whether the turbulent transport between temperature and particles is similar. This similarity has, however, never been verified, due to the lack of observational turbulence data of particles. https://doi.org/10.5194/acp-2021-435 Preprint. Discussion started: 7 July 2021 c Author(s) 2021. CC BY 4.0 License.
The turbulent diffusion processes of all scalars (including active and passive scalars) are dealt with in a 50 unified manner in the current model. To date, only a few studies have shown that pointed out the meteorological fields and pollutants can be changed by adjusting the minimum value of turbulent diffusion coefficient (TDC) (Savijarvi et al., 2002;Wang et al., 2018;Du et al., 2020;Liu et al., 2021), increasing turbulent kinetic energy (TKE) (Foreman and Emeis, 2012) and modifying experiment expressions (Sušelj and Sood, 2010;Huang and Peng, 2017). Recently, Jia et al. (2021) obtained the TDC 55 of particles by using high-resolution vertical flux data of particles based on the mixing length theory.
Additionally, this relationship has been embedded into the WRF-Chem model to calculate the PBL mixing process of pollutants separately. This work has initially improved the overestimation of pollutant concentration at night in winter 2016 in Eastern China. However, a series of heavy pollution incidents have occurred and attracted much attention since 2013 (Yang et al., 2018;. Therefore,

60
we conducted a series of simulations for the heavy pollution periods in winter from 2013 to 2017 in this study. The difference between this study and previous work is that previous work focused on the analysis of observations, while this study mainly explores the uncertainty of the influence of the model on the turbulent diffusion of particles.

Data
In this study, the aerosol pollution level is denoted by the hourly surface PM2.5 concentration that is  China. In addition to PM2.5 observations, the hourly concentrations of CO were acquired from the National Air Quality real-time publication platform (http: //106.37.208.233:20035, last access: 20 May 2021). Aside from this, the hourly meteorological observation data, including temperature, pressure, relative humidity, wind and visibility from the national automatic weather stations (AWS) provided by the National Meteorological Information Center of China Meteorological Administration (NMICMA)

75
(illustrated by gray crosses in Fig. 1b). The time period of the data selected is from 1 January 2013 to 31  Fig. 1b). Identical eddy-covariance systems were operated, including three-dimensional sonic anemometer-thermometer (IRGASON, Campbell Scientific, USA) 80 and CO2/H2O open-path gas analyzer (LI7500, LI-COR, USA). These instruments measured three components of wind speed, potential temperature, water vapor and CO2 concentrations with a frequency of 10 Hz. The turbulence data finally was split into 30-min segments. In addition, a continuous particle measuring instrument E-sampler () and a high-frequency sampling visibility sensor CS120A () were used to obtain PM2.5 mass concentration every minute and visibility of 1 Hz. The calculation of 30-min 85 vertical flux of PM2.5 is based on the nonlinear relationship between PM2.5 concentration and visibility (Ren et al., 2020). Detailed background and calculation principle of this method were presented in Ren et al. (2020), so we only describe key steps. Firstly, we separate PM2.5 concentration (C) and visibility datasets (V) into mean and turbulent deviations (i.e., ' c c c =+ and ' V V V =+ ). Secondly, we get the fitted coefficients by using exponential correlation between the PM2.5 concentration and visibility  (Miao et al., 2018), which proved a fine resolution (1 Hz) profiles of temperature, relative humidity and wind speed two times (0800 and 2000 BJT) a day during winter. To eliminate the error caused by the difference of calculation methods of PBLH, Richardson number method is used to calculate the PBLH in both observation and simulation. The Richardson number is defined as follows: where z is the height above ground, g is the gravity, θv is the virtual potential temperature, and u and v are the component of wind. The subscript "s" denotes the surface level. The height at which the Richardson number equals 0.25 is defined as the PBLH.  (Subin et al., 2012;Gu et al., 2015), ACM2 planetary boundary layer scheme (Pleim, 120 2007), Grell-3D cumulus scheme (Grell and Devenyi, 2002). And the chemical mechanism is the RADM2-MADE/SORGM scheme (Ackermann et al., 1998;Schell et al., 2001). The initial and boundary conditions of meteorological fields were set up using the National Centers for Environmental Prediction The TDC is parameterized by the mixing length (l) and the function of Richardson number (f(Ri)), that

Numerical simulation
where ss is the wind shear (i.e.,

140
(1962) is widely used in the model (Louis, 1979;Liu and Carroll, 1996;Lin et al., 2008;Pleim, 2016 where fh and fm denote the function of heat and momentum, respectively, and these functions have been (ii) For the unstable conditions (Ri<0), Equation (2) is rewritten as: Considering that the pollution is usually accompanied by the stable boundary layer, we mainly modify 3 Temperature-particles transport dissimilarity 160 The PM2.5 concentration frequently reached hazardous levels above 100 μg m -3 during six heavy pollution episodes (marked by HPE1-HPE6 in Fig. S1). The turbulent characteristics of PM2.5 concentration have been demonstrated in Ren et al. (2020), and the turbulence characteristics of heat and particles were markedly different (Jia et al., 2021;Ren et al., 2021). Based on the previous studies, the turbulent correlation coefficient is used to evaluate the transport efficiencies for heat, water vapor, momentum and 165 particles (Stull, 1988;Li and Bou-Zeid, 2011;Dupont and Patton, 2012). The expression as follows: Rwp denotes the correlation coefficient between the fluctuations of w' and p', while p stands for the temperature T, specific humidity q, longitudinal velocity component u and particles c. This value is between -1 to 1 (negative correlation to positive correlation), and zero indicates that the two parameters are uncorrelated. The σw and σp are the standard deviations of vertical velocity and parameter p (i.e., T, 170 q, u, c) over a 30-min interval, respectively. If the MOST is applicable, it indicates the turbulent mechanisms of heat, water vapor and particles are the same, i.e., Rwt=Rwq=Rwc . Previous studies have investigated different mechanisms of scalar transport between temperature and humidity (Moriwaki and Kanda 2006;Katul et al., 2008;van de Boer et al., 2014;Guo et al., 2016Guo et al., , 2020. Lacking profile data for the PM2.5 concentration (Yuan et al., 2019;Ren et al., 2020), there is little about the 175 transport efficiency of fine particles (i.e., PM2.5). The correlation coefficient of heat flux and particle flux can be defined as: are the standard deviations of w't' and w'c', respectively. The correlation coefficients of the heat (Rwt), fine particles (Rwc), and heat flux and particle flux (Rwt,wc) are presented in Fig. 2. Clearly, there is an obvious difference between Rwt and Rwc. Whether transport efficiency is Rwt or Rwc, transport 180 efficiency can exhibit the greatest variability at night during the HPEs, probably suggesting an increasing complexity of turbulent structures at night ( Fig. 2b and 2c). High correlation exists between heat and fine particles fluxes at night (especially at the wee hours) in the HPEs (Fig. 2d), which indicates that these fluxes are performed by the same motions within the PBL. Previous research has noted that the atmospheric vertical mixing is mainly controlled by the large-scale eddies' percentage at night during the 185 HPEs (Li et al., 2020). However, it should be mentioned that the correlation coefficient between heat and fine particles fluxes (Rwt,wc) changes dramatically at night (Fig. 2d). This means that these two fluxes transported with different eddies in a short time, or transported at different time periods by the same eddy when the correlations diminish. Consequently, there is a difference between the transport of heat and fine https://doi.org/10.5194/acp-2021-435 Preprint. Discussion started: 7 July 2021 c Author(s) 2021. CC BY 4.0 License. particles fluxes. Whether scalar is temperature or particle, it is debatable that the mixing process of all 190 scalars are dealt with a unified manner within the PBL. As a result, we urgently need to develop a TDC of particles, which is used only to calculate the mixing process of pollutants within the PBL.

Improvement of PM2.5 concentration
Based on the TDC relationship of particles in the previous study (Jia et al., 2021), this study applies this relationship to a long-term scale simulation for verification. Figure (Fig. 5). And the standard deviation (normalized) of the mean value is decreased by 0.2 (2013), 0.28 (2014), 0.14 (2015) and 0.16 (2017) (Fig. 5). As a whole, the new scheme can significantly improve the common phenomenon of overestimated pollutant concentration in the SBL in Eastern China (Fig. 5).
In addition to the changes in the pollutant concentration near the surface, we should also pay attention to the changes in the pollutant concentration in the vertical direction. Theoretically, increasing turbulent diffusion will reduce the pollutant concentrations near the surface-layer, and the pollutants will be more fully mixing in the vertical direction, which results in lower concentrations of pollutants in the near surface-layer and higher concentrations of pollutants in the upper layer. Actually, the pollutant 220 concentration is reduced in the surface-layer and it is increased in the upper layer at night (Fig. 6), which is consistent with the theory.

Meteorological parameters
Depending on the transport dissimilarity of heat and particles, the TDC of particles was added separately 225 in the model to calculate the turbulent mixing process of particles. For correctional approaches, it is important that a new scheme does not lead to worse performance than that with the original scheme. To verify the new scheme without affecting the simulation results of the meteorological parameters, the simulation results of the near-surface meteorological elements (i.e., 2-m temperature, 2-m relative humidity and 10-m wind speed) have been compared and analyzed. It can be seen from Figure S2-S4 230 that the correlation coefficients of meteorological parameters by two schemes are greater than 0.99, noting that the new scheme does not alter the performance of meteorological fields, which is an advantage of the new scheme. Compared with previous studies, modifying the turbulent diffusion coefficient of heat not only affects the simulation of temperature (Savijarvi and Kauhanen, 2002), but also influences the results of pollutants (Liu et al., 2021). Improving the parameterization scheme is a long and tough 235 process, making it difficult to improve the simulation results of all parameters at once. When the simulation results of one parameter are improved, we should first seek to ensure that the simulation results of other parameters are not deteriorated. Then, we are going to look at improving other parameters.
Although the aerosol-radiation two-way feedback process has been considered in the atmosphericchemistry two-way coupled model, the mean fractional change in PM2.5 concentration varying just a few Some turbulent characteristics (e.g., turbulence barrier effect) can be taken into consideration during the HPEs, reflecting a more realistic pollutant concentration evolution process. We think the next step is to solve this major problem.

PBL height
Although PBL height (PBLH) is widely used to determine the effective air volume and atmospheric environmental capacity for pollutant diffusion (Miao et al., 2018), the influence of PBLH on the pollution is uncertain. (1) There are various methods to determine the PBLH, either through observation or simulation (Jia and Zhang., 2020;Zhang et al., 2020). Various methods diagnose different PBLH, which 250 reinforces uncertainty about the PBLH as a criterion. (2) There does not necessarily reflect a negative correlation between pollutant concentration and PBLH. The relationship between the PBLH and PM2.5 pollution has been revealed on the basis of the four-year radiosonde measurements, and the results show that the correlation between PBLH and PM2.5 concentration is different in various regions (Miao et al., 2018). Moreover, when the PBLH is higher, the corresponding pollutant concentration is not necessarily 255 lower (Miao et al., 2021). When there is a transport stage during the HPEs with a high wind speed, the mechanical turbulence is strong, and the PBLH and pollutant concentration increase simultaneously. Therefore, the relationship between PBLH and PM2.5 pollution is intricate. The impact of PBLH is ultimately represented through the TDC in the model, because the PBLH is used to calculate TDC. If the pollutant concentration is clearly controlled by the PBLH, when the pollutant concentration is 260 overestimated and the PBLH is to be underestimated. However, the PBLH is reproduced well by the model, and the model does not underestimate the PBLH (Fig. 7). The new scheme does not disturb the simulation results of meteorological fields, and therefore does not affect the simulation results of PBLH

Influence of other processes
Overestimating of pollutant concentrations has been improved in Eastern China, but Fig. 8), and the pollutant concentration is overestimated at some sites, which are away from the mountain (i.e., Tianjin and southeast of Hebei). Furthermore, we found that the TDC of particles in the new scheme is significantly smaller than the TDC of heat in the original scheme in the mountain area (red rectangle in Fig. S6). The terrain will disturb the turbulence fields, making the stable stratification 280 weakly stable/unstable. Theoretically, the reduced TDC will increase the pollutant concentration near the mountain, and improve the underestimation of pollutant concentration of the original scheme. However, the change of TDC does not improve the underestimation of pollutant concentration in the mountain area, which shows that the impact of other processes is more obvious in the mountain area. For instance, the advection process is strongly related to the wind and pollutant concentration gradients from upwind areas 285 to downwind areas . Figure S7 shows that the wind speed is much more overestimated in the mountain areas (two purple rectangles in Fig. S7i-l), but only the pollutant concentration is always underestimated in the BTH region (Fig. 3i-l). Clear gradients of wind speed and pollutant concentration exist in the BTH region (Fig. 3a-d; Fig. S7a-d), so these sites (i.e., closer to the mountain) may be significantly affected by the advection process. Hence, the influences of other processes or topography 290 in the mountain area deserve further consideration in the future.
Whether the simulation of chemical components has been improved, it cannot be well verified because of the lack of observational data. Although the observational components of PM2.5 are not available to evaluate the simulation results of new scheme, CO, as a representative of primary pollutants, can be https://doi.org/10.5194/acp-2021-435 Preprint. Discussion started: 7 July 2021 c Author(s) 2021. CC BY 4.0 License. compared to the observations. Results from new scheme with TDC of particles are more consistent with 295 the observations than the original scheme (Fig. S8), which supports the improvement of PM2.5 concentration ( Fig. 5 and S8).

Conclusions and prospects
Mesoscale model faces numerous challenges during the heavy pollution events. One of these challenges is the correct description of the turbulent mixing of pollutants. Although the model can reproduce the concentration is not specifically controlled by the PBLH, but by TDC. However, TDC has a negligible impact on the simulation of pollutant concentration at some sites with complex topography. PM2.5 315 components cannot be used to evaluate the results of the simulation of the new scheme, due to the lack of observational data. CO, however, as a representative of primary pollutants, can be compared to observations. Results from new scheme are more consistent with the observations than the original scheme, which supports the improvement of PM2.5 concentration. The new scheme could provide promising guidance during the heavy pollution events. The turbulent transport mechanism and scalar 320 parameterization is a complex topic (Smedman et al., 2007, Lemon et al., 2019Couvreux et al., 2020; https://doi.org/10.5194/acp-2021-435 Preprint. Discussion started: 7 July 2021 c Author(s) 2021. CC BY 4.0 License. Edwards et al., 2020), and beyond that, other processes also need in-depth understanding and exploration (Seinfeld et al., 2016;Shao et al., 2019;Emerson et al., 2020). Therefore, more research during the heavy pollution events, especially on the experimental side (e.g., extensive measurement campaigns), might shed more light on the turbulent mixing process of pollutants and their mechanisms.

Data availability
The surface PM2.5 concentration, meteorological data, turbulent datasets and turbulent flux data of PM2.5 are available by request (xiaoye@cma.gov.cn).

Author contributions
Development of the ideas and concepts behind this work was performed by all the authors. Model 330 execution, data analysis and paper preparation were performed by WJ and XZ with feedback and advice.

Competing interests
The authors declare that they have no conflict of interest.