Special Topic : Air Pollution and Control Aerosol and boundary-layer interactions and impact on air quality

Aerosol and boundary-layer interactions and impact on air quality Zhanqing Li1,2,∗, Jianping Guo3,∗, Aijun Ding4, Hong Liao5, Jianjun Liu2, Yele Sun6,7, Tijian Wang4, Huiwen Xue8, Hongsheng Zhang8 and Bin Zhu9 1State Key Laboratory of Earth Surface Processes and Resource Ecology, GCESS, Beijing Normal University, Beijing 1000875, China; 2Department of Atmospheric and Oceanic Sciences, University of Maryland, MD 21029, USA; 3State Key Laboratory of Severe Weather, Chinese Academy of Meteorological Sciences, Beijing 100081, China; 4School of Atmospheric Sciences, Nanjing University, Nanjing 210023, China; 5School of Environmental Science and Engineering, Nanjing University of Information Science and Technology, Nanjing 210044, China; 6State Key Laboratory of Atmospheric Boundary Layer Physics and Atmospheric Chemistry, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing 100029, China; 7Center for Excellence in Regional Atmospheric Environment, Institute of Urban Environment, Chinese Academy of Sciences, Xiamen 361021, China; 8Department of Atmospheric and Oceanic Sciences, Peking University, Beijing 100871, China and 9School of Atmospheric Physics, Nanjing University of Information Science and Technology, Nanjing 210044, China


INTRODUCTION
Aerosols are multi-facet agents that affect air quality, weather and climate through many mechanisms, as reviewed extensively in a series of IPCC reports.Aerosols are colloids of fine solid particles or liquid droplets suspended in the atmosphere.Through the effects of aerosol-radiation interactions (ARI), aerosol-cloud interactions (ACI) or both, aerosols can significantly affect Earth's climate by perturbing the Earth's radiation budget and water cycle processes [1][2][3][4][5][6][7][8].As some observational analyses have indicated, cloud and precipitation properties are remarkably affected by elevated aerosols, which suppress light rainfall but intensify heavy rainfall and lightning in the coastal regions of Southeast China [9].Aerosol pollution in the planetary boundary layer (PBL) adversely affects human health [10,11].
Chemical reactions can occur on the surface of non-gaseous particles or within the body of liquid droplets, and these processes are key components of the biogeochemistry of our planet [12].
The ARI effect is concerned with the scattering and absorption of solar radiation by aerosol particles.Aerosols can substantially reduce the amount of solar radiation reaching the ground [13,14], and thus reduce sensible heat fluxes that drive the diurnal evolution of temperature and the PBL [15].This in turn leads to weaker turbulence in the PBL, and a reduction of entrainment of dry air into the PBL from the free troposphere, which leads to more moisture in the PBL.The combined effects of lowering nearsurface temperature and increased moisture can increase relative humidity (RH).The increased RH tends to favor the hygroscopic growth of aerosols and enhances the scattering of solar radiation.The REVIEW Li et al. 811 increased RH could also enhance the formation of secondary aerosols.The aqueous-phase reactions of NO 2 and SO 2 in fog/clouds or aerosol water are important for sulfate formation under hazy conditions in China [16][17][18].It has been found that the discrepancy in aerosol types leads to huge differences in estimates of aerosol direct forcing at the top and bottom of the atmosphere over the Indian Ocean [14] and in China [19].Also, it is well recognized that various aerosol types exhibit quite different ARI effects.Among all types of aerosols, absorbing aerosols, consisting primarily of organic carbon, black carbon (BC) and brown carbon [21], have the strongest interaction with the PBL [22].In a polluted environment, BC can be transformed into fully compact particles and becomes a much stronger absorbing agent [23,24].As such, absorbing aerosols can alter the PBL more effectively than other types of aerosol, which may be another major factor on top of the formation of new particles in triggering severe haze events in China [25].
The PBL is inherently connected to air pollution because of the bulk of aerosols residing in the PBL, and the strong interactions or feedbacks between aerosols and the PBL [26].These interactions can considerably exacerbate air pollution, even if emission rates remain the same.Observational and modeling studies suggest that aerosol-PBL feedbacks influence air quality significantly.Surface dimming (by all types of aerosols) and upper-PBL warming (by absorbing aerosols) help stabilize the PBL and weaken turbulence mixing, leading to a decrease in the boundary-layer height (BLH), which correctively favors the accumulation of air pollutants in a shallower PBL [15,24,27].
The absorption of solar radiation by aerosols can induce a temperature inversion (TI) at the top of the PBL that is often associated with severe pollution episodes [28].When the temperature difference between the top and bottom of a TI layer is greater than 20 • F, severe pollution ensues [29].Some of the most severe pollution events in history have been associated with elevated TIs.In general, high aerosol concentrations tend to occur in the atmosphere with a TI [30][31][32].Gas pollutants such as SO 2 [33] and NO 2 [32,34] have also been found to be closely linked to TIs.
In this article, we comprehensively review the studies with regard to aerosol, pollution, PBL and their interactions, starting with the section entitled 'Overview of air pollution and aerosols in China'.PBL processes and observations are given in the section entitled 'Fundamentals and observations of the PBL'.Aerosol-PBL-convection interaction schemes are described in the section entitled 'Pro-cesses governing aerosol and PBL interactions'.The section entitled 'The trend and fluctuation of air pollution: the roles of circulation, PBL, climate change and weather' elaborates on the general roles of PBL, climate changes and weather regimes on surface air pollution.The section entitled 'Concluding remarks' concludes the paper.

Ground surface measurements
Aerosols have been measured extensively across China, chiefly after 2000, through national operational networks and field experiments, which have been comprehensively reviewed by Liao et al. [6] and Li et al. [7].Prior to 2000, a few direct aerosol optical depth (AOD) measurements were made, which were inferred from clear-sky radiation data [35,36,], revealing a rapid deterioration of air quality from the 1960s to the 1990s.This was confirmed by concomitant satellite-derived AOD observations [37].More accurate AOD measurements can be provided by ground-based sunphotometers.After 2000, several ground-based aerosol observation networks were established across China, including the Chinese Sun Hazemeter Network [38], the China Atmosphere Watch Network (CAWNET) [39], the China Aerosol Remote Sensing Network [40,41], and the Campaign on Atmospheric Aerosol Research network of China (CARE-China) by the Chinese Academy of Sciences [42].Intensive field experiments measuring rich aerosol properties have been increasingly conducted in China, such as the East Asian Study of Tropospheric Aerosols: an International Regional Experiment (EAST-AIRE) [43] and the East Asian Study of Tropospheric Aerosols and Impact on Regional Climate [44].
These data have been widely used to derive the nationwide distribution of AOD [38] and aerosol single scattering albedo (SSA) [45].Aerosol loading is exceptionally heavy in the eastern half of China.Aerosol absorption is particularly strong in centralwest China and Northeast China due to the high consumption of coal (Shanxi is the capital of coal mining in China).There is much weaker absorption in southeastern China due presumably to its industrial emissions containing high proportions of sulfate and nitrate whose absorption is weak.Combining ground-and satellite-based measurements, Li et al. [20] for the first time derived the estimate of aerosol radiative forcing at the top, bottom, and within the atmosphere.The daily and annual mean atmospheric absorption and surface cooling due to aerosols REVIEW (primarily in the PBL) are −20 W m −2 and 18 W m −2 , respectively.
Air pollution typically characterized by high concentration of aerosol particles in aerodynamic diameter less than 2.5 μm (PM 2.5 ) concentrations is one of the major environmental concerns in China.Although air quality in megacities has continued to improve since 2013, the annual mean PM 2.5 concentration in most cities in northern China is still much higher than the National Ambient Air Quality Standard (i.e.35 μg m −3 as an annual average), e.g.70.3 μg m −3 in Beijing in 2016.These results clearly indicate that air pollution in megacities in China is still severe, and it is even worse during the winter season because of the increased emissions by coal combustion for residential heating along with the frequent stagnant meteorological conditions.As a result, extensive ground measurements have been conducted during the past decade to characterize the chemical composition, sources and formation mechanisms of aerosol particles, with most of them taking place in the four most polluted regions, including the Pearl River Delta, the Yangtze River Delta, the North China Plain and the Sichuan Basin.This includes several international field campaigns by involving tens of research teams around the world, such as the Campaigns of Air Quality Research in Beijing and Surrounding Regions (CAREBeijing), the Program of Regional Integrated Experiments of Air Quality over Pearl River Delta (PRIDE-PRD) [39], and the Haze Observation Project Especially for Jing-Jin-Ji Area (HOPE J 3 A) [46].
Numerous results and findings on aerosol composition, sources and processes have been presented.Overall, organic aerosols account for the major fraction of PM 2.5 followed by ammonium sulfate or ammonium nitrate, and black carbon.While the sources of primary emissions including traffic, coal combustion, biomass burning and cooking are identified and quantified, secondary aerosols are found to be more important in the formation of severe haze episodes [16,47,48].It has been well recognized that the high anthropogenic emissions and rapid secondary aerosol formation are the key factors leading to the frequent occurrence of severe haze episodes, characterized by efficient new particle formation and growth [49].The efficient aerosol nucleation, combined with aerosol growth, is closely associated with severe haze episodes in China.It is quite distinct from those typically observed in the urban regions of other countries and pristine environments worldwide [50], which are modulated by interactions between sulfuric acid and organics [51][52][53][54].Typically, there exist clear diurnal variations in the PM number, size and mass concentration [55], reflecting the interplay between primary emissions, new particle formation, photochemical growth, removal and the PBL variation.In contrast, haze events in China typically exhibit a periodic cycle of 4-7 days [56].
In addition, recent studies have also illustrated the important roles of stagnant meteorology, which are typically characterized by low BLH, weak wind speed, strong temperature inversion and high RH in the formation of these events [57,58].In fact, the favorable weather conditions induced by the increases in greenhouse gas emissions or the changes in the boreal cryosphere, particularly in the global environment, are mainly responsible for the more frequent haze episodes in winter on the North China Plain [59,60].The most recent results from air pollution studies in China are summarized in [61], and the extensive ground measurements of aerosol particles based on real-time techniques are presented in [62].

Aerosol vertical distributions and the PBL
The aerosol vertical distribution is key in determining aerosol radiative forcing.To date, our understanding of aerosol radiative forcing is still very poor due to the assumption of vertically constant profiles in radiative-transfer models, among other factors.Aerosol radiative forcing strongly depends on the vertical distribution of aerosols relative to clouds, especially for dust layers [63].
Using aircraft observations over the Beijing-Tianjin-Hebei region of North China, different types of aerosol vertical distributions in association with the PBL evolution have been revealed [64][65][66].When the PBL is well developed, the aerosol number concentration (N a ) is homogeneously distributed throughout the whole PBL, leading to a sharp drop above the PBL top [66].The vertical profile of aerosol number concentration (N a ) can then be approximated by a piecewise function [64].
Our understanding of the aerosol vertical structure has been improved tremendously since the advent of the Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) onboard the Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations satellite, which has become increasingly recognized as a valuable sensor to elucidate the altitude-resolved structure of aerosol particles such as smoke, dust and polluted dust [67].While ground-based lidar and aircraft-borne in situ measurements have provided major insights into regional aerosol structure, CALIOP-based observations have been extensively used to derive the 3D structure of aerosols on a global scale [68].By defining the most probable height (MPH) where aerosol particles most likely reside, Huang et al. [69] generated a global MPH distribution for dust and smoke aerosol types.

REVIEW
Li et al. 813 On regional scales, large discrepancies and uncertainties remain when it comes to the vertical distribution of aerosols at altitudes within the PBL up to the bottom of the free troposphere [70], depending on the regions and seasons of interest [71][72][73][74].Specifically, Adams et al. [75] presented the 3D structure of aerosols across the trans-Atlantic region.Based on 10 dust cases observed by the CALIOP, Huang et al. [76] found that summertime dust storms occurred more frequently than previously thought on the remote northwestern Tibet Plateau, and that the dust layer reached altitudes of up to 4-7 km.Based on CALIOP observations, along with surface meteorological data and the Hybrid Single Particle Lagrangian Integrated Trajectory Model (HYSPLIT), Guo et al. [77] found that dust storms originating in northwestern China were transported eastward to Beijing at a rate of 1200 km per day, mostly at altitudes of 3-5 km.As shown in Fig 1, two zonal transport pathways were revealed from synergetic observations made by the CALIOP and the Moderate Resolution Imaging Spectroradiometer (MODIS): one dust belt across northern China and a smoke belt across southern China [78].Recently, based on long-term satellite-and groundbased measurements, both seasonal and spatial variations of the profiles of aerosol extinction coefficients over China have been identified.In particular, the lapse rates of aerosol extinction in the polluted regions were much greater than those in the pristine regions, most likely due to more stable atmosphere caused by absorptive aerosols in the polluted regions [79].
The aerosol vertical distribution in the PBL, subject to the PBL dynamics and large-scale weather systems, exhibits strong temporal (seasonal and diurnal) variations and spatial differences, as revealed in the meteorological tower measurements of PM 2.5 and PM 10 in Beijing in 2003 [80].Further subsequent measurements were made on the same tower to characterize the vertical distributions of trace gases (e.g.ozone (O 3 ), nitrogen dioxide (NO 2 ), and sulfur dioxide (SO 2 )), PM 2.5 and filter-based aerosol composition [81][82][83][84].The results showed that the mixing ratio of O 3 often peaked at ∼120 m and remained high in the residual layer at nighttime [85], while SO 2 was found to have the highest mixing ratio at ∼50 m [82].More recently, Sun et al. [86,87] conducted simultaneous real-time observations of aerosol particle composition at two different heights on a tower, i.e. ground level and 260 m above ground level (AGL), using two similar aerosol mass spectrometers.The results illustrated very complex vertical distributions of aerosol species that interact closely with boundary-layer meteorology.In general, the vertical differences between ground level and 260 m AGL are reduced substantially in the daytime due to the elevated BLH associated with stronger vertical mixing.They also observed the interactions of different air masses at different heights that affect the vertical gradients.For example, the temperature inversion during the clearing stage of a severe pollution event results in a delay of the cleaning of air pollutants between ground level and 260 m AGL, while the stably stratified layer associated with a fog event is often characterized by much higher concentrations of aerosol species in the lower atmosphere and rapid decreases on the top of the layer.Further analysis shows that the vertical differences in meteorology (e.g.T and RH) and gas precursors can also affect secondary aerosol formation at different heights.For example, higher concentrations of nitrate at higher altitudes were clearly associated with REVIEW the lower T and higher RH that facilitated the gasto-particle formation mechanisms.
Although the vertical distributions of air pollution and its interaction with the boundary layer in the low atmosphere have been extensively characterized based on the tower measurements, our understanding of these characteristics and interactions at high altitude is far from complete.The vertical distribution of aerosol number concentration in the PBL over Beijing has been found to be strongly related to different weather systems, based on in situ aircraft measurements [66].Distinct vertical profiles of aerosol scattering coefficient and precursors were found over cities in eastern, northwestern, and northeastern China [88].Similar vertical distribution characteristics of aerosol optical properties were reported based on micropulse lidar data from the Yangtze River Delta of China [89].Multi-layer BC in the PBL, including local accumulation near the surface and regional transport from upwind megacities in the upper PBL, was revealed using a tethered balloon platform [90].In 2014, the vertical profiles of O 3 and size-resolved aerosol number concentrations were measured at a rural site on the North China Plain using an unmanned aerial vehicle [91].The results showed higher O 3 levels in the residual layer than the mixed layer, while this is reversed for aerosol number concentrations.Further analysis illustrated that the vertical profiles of air pollutants are influenced by not only PBL meteorology, but also anthropogenic emissions on local and regional scales.

FUNDAMENTALS AND OBSERVATIONS OF THE PBL PBL processes
A comprehensive review of the theory, nature and modeling of PBL was given in Baklanov et al. [92].The physical processes in the PBL are concerned with turbulent diffusion, wind speed, the atmospheric thermodynamic state (i.e.temperature, humidity), and adiabatic heating due to aerosols and other absorbing agents.The wind and temperature profiles along with the BLH are the main factors affecting turbulent diffusion [81].The PBL structure is also dictated by large-scale weather regimes [93].
Based on the Monin-Obukhov similarity theory of continuous turbulent motion, the relationship between flux (e.g.momentum, sensible and latent heat fluxes) and atmospheric profile (e.g.wind, humidity, temperature, gradient) in the atmospheric surface layer is key for air pollution diffusion, especially under stable stratification conditions.The structure of a stable boundary layer remains a puzzle in the PBL study.A series of experiments to tackle this issue have been carried out, such as the Stable Atmospheric Boundary-Layer Experiment in Spain [94], the Cooperative Atmospheric-Surface Exchange Study in Kansas [95], and the Surface Heat Budget of the Arctic Ocean Experiment in the Antarctic [96].Turbulent intermittency frequently occurs in a stable boundary layer, which tends to scale up the transport of heat, water vapor, and momentum in the vertical, as opposed to the reduced vertical transport in the unstable boundary layer characterized by continuous turbulent motion [97].The characteristics of such scalar quantities as PM 2.5 and NO x are affected by the turbulent velocity field and present a complex, chaotic structure on spatiotemporal scales, which makes it difficult to study the turbulent transport of pollutants [98].The Hilbert-Huang transform method [99] is a valid method to solve atmospheric turbulence problems in a stable boundary layer and has been widely applied [100,101].Currently, most topics associated with generalized PBL parameterization schemes in models are related to a stable boundary layer.For example, a systematic overestimation of simulated near-surface wind velocities under stable atmospheric conditions has been reported in addition to an overestimated turbulent diffusion capacity of air pollutants in a stable layer, resulting in the underestimation of pollution [102,103].It is therefore imperative to explore and understand the structure of the stable boundary layer and its interaction with and influence on the air pollutant transport process.

Detection and variations of BLH
Knowledge of the PBL is crucial for understanding the interactions between air pollution and meteorology.A key question remaining unclear is how the diurnal evolution of the PBL interacts with the vertical distributions of aerosols.To address this question, we need to have a good knowledge of the BLH and its evolution.The determination of BLH is a nontrivial task because it is not directly measured by routine meteorological instruments.The most common PBL observations are obtained by radiosondes, which can provide vertical temperature, moisture and wind profiles from the surface up to the ∼50-hPa level [104][105][106].Various algorithms have been developed to retrieve the BLH; these have been completely summarized in Seidel et al. [107].
Guo et al. [108] obtained the first BLH climatology in China using long-term fine-resolution (1-s resolution) atmospheric soundings from the radiosonde observation network operated by the China Meteorological Administration.In addition to the spatial and seasonal variability of the BLH, .A large discrepancy between sounding-and reanalysis-derived BLH was found, particularly in North China at 1400 Beijing time.This may undermine our ability to observe and simulate surface PM 2.5 because the latter relies on reanalysis-derived BLH data [40,[109][110][111].
Although radiosondes are widely deployed around the world, operational radiosonde launches are done only twice a day (0000 UTC and 1200 UTC), prohibiting observation of the diurnal variation of the BLH.To overcome this limitation, various types of measurements have been used to derive the BLH, including more frequent launches of radiosondes that usually happen during a field experiment, ground-based remote sensing using high-resolution infrared sounder, multi-channel microwave radiometer, lidar, sodar, etc.
Among the various observation techniques, the lidar measurements have been most widely used to retrieve the PBL.The wavelet covariance transform method is one of the widely used algorithms; this is detailed by Davis et al. [112] and Brooks [113].The algorithm allows for comparisons between the backscatter sounding and a step function, such that the largest gradient in backscatter with altitude is marked as the top of the PBL.The wavelet co-variance transform is suitable for the automatic detection of the PBL because it requires only the lidar backscatter information.However, the wavelet covariance transform method often detects cloud backscatter and elevated aerosol plumes instead.To avoid this problem, Steyn et al. [114] fit the entire backscatter profile to an idealized curve.The algorithm uses an iterative curve-fitting routine to minimize the root-mean-square difference between the backscatter profile and the idealized curve.Simulated annealing allows the iterative process to bypass local solutions and returns a more robust BLH [115].It also introduces a small random element that must be filtered out to avoid occasional unrealistic jumps in BLH values.
Sawyer and Li [116] proposed a versatile method that can be applied to any type of atmospheric profile data.Moreover, it takes advantage of the merits of the aforementioned techniques while overcoming their limitations.Figure 3 presents a comparison of the BLH detected by lidar, radiosonde and temperature profiles from a high-resolution infrared sounder.The BLH values derived from these different approaches agree generally well, although some consistent differences exist.
The mast or meteorological tower is an important observational platform used to continuously measure profiles of turbulent flux (latent and sensible heat), and profiles of the atmospheric mixing ratio within the lower part of the PBL at a reasonably high resolution [84,104,117,118].The main shortcoming of this platform is its limited range, which is typically below 300 m.As a simple remote sensing instrument, the sodar is suitable for routine operations [104,119].The sodar can capture typical PBL features by measuring structure parameters such as the acoustic refractive index, irrespective of stable or convective atmospheric conditions.With Doppler capability, it can determine vertical velocity variance profiles.For instance, Yang et al. [120] characterized the daytime evolution of the PBL based on observed vertical velocity variance profiles using one Doppler sodar on the Tibetan Plateau during the dry season.The sodar, however, is limited to an altitude range of 500-1000 m and is highly sensitive to environmental noise [104].
Boundary-layer wind profilers have been widely deployed in field campaigns to investigate the PBL structure [121][122][123] as well.Based on wind profiler measurements, the features of low-level jets in Shanghai and Tianjin have been elucidated by Du et al. [124] and Wei et al. [125], respectively.The wind profiler observed backscatter signals in clear air are proportional to the structure function parameter of the refractive index, which can be used to estimate the BLH under convective conditions [126,127].Orbiting around the Earth, the space-borne lidar has the unique merit of providing global BLHs.Using CALIOP-attenuated backscatter observations, the BLH was investigated over China by Liu et al. [128] and Zhang et al. [129], in the southeastern Pacific by Ho et al. [130], and over Europe by Leventidou et al. [131].However, under stable PBL conditions, large uncertainties in estimating BLHs from space-borne lidar like CALIOP ensue due to the weak vertical gradients in the aerosol loading [104].

Aerosol-induced adiabatic heating in the PBL
To gain an insight into the radiative effect of aerosols on the PBL, it is essential to compute the radiative heating rate with known aerosol vertical extinction profiles and SSA profiles.The SSA profile is most difficult to get, although there are some aircraft in situ measurements of SSA.Several methods have been used to calculate the columnar SSA.One method is to use a combination of satellite-measured spectral reflectance and surface-measured transmittance, which can be collected over large regions such as across China [45].Since the bulk of aerosols are well mixed in the PBL, one may assume to the first order of approximation that SSA derived for the entire atmospheric column is the same as that for the PBL.
SSA values can then be used to compute the radiative heating rate by virtue of lidar-observed aerosol extinction or even backscattering profiles with certain assumptions regarding the backward scattering ratio [89].Based on a single-channel elastic-scattering lidar, the backscattered radiation can be calculated according to the following equation [132,133]: where r represents the range, and β(r) and σ (r) denote the backscattering and extinction coefficients caused by both aerosol and molecular factors, respectively.Other variables are detailed by Campbell et al. [132] and Welton et al. [71].The normalized relative backscatter, P NRB (r), or NRB, can be formulated as (2) where C can be solved using a technique constrained by the co-located AOD at a range where molecular scattering is determined above the surface-detected aerosol layer under cloud-free conditions [71].One can then solve for layer-averaged extinctionbackscatter ratios and aerosol extinction coefficient profiles [71,134].The radiative forcing of aerosols at the top, bottom, and within the atmosphere has been derived from this information [89].The aerosol particles largely reside below 2 km AGL, most of which

Aerosol-PBL interactions due to ARI
The strong contrast between warming and cooling in the atmosphere and the surface can drastically impact atmospheric stability and the PBL, which can affect weather and dynamics.By means of model (NCAR/CAM3) tests, it has been demonstrated that increases in aerosols of moderate absorption can reduce wind, weaken atmospheric circulation and even the monsoon system [7,135].Accompanying the rapid degradation of air quality is the steady decrease in surface wind speed [136] that has been attributed at least partially to the effect of increasing aerosols, as shown in recent studies using long-term meteorological data [137,138].Aerosols and the PBL interact inherently in a variety of ways [7,22,26,139]: (1) Aerosols cool the surface and lower atmosphere by the reduction of shortwave, sensible heat, and latent heat fluxes.
(2) Absorbing aerosols above the PBL warm up the air of the free troposphere, and thus stabilize the boundary-layer inversion cap, inducing and prolonging the temperature inversion in the upper PBL, and suppressing diffusional dispersion of pollutants near the surface.(3) Absorbing aerosols within the PBL may not strengthen the atmospheric stratification since there is strong vertical mixing in the PBL due to aerosol-induced atmospheric heating.Therefore, absorbing aerosol within the PBL does not naturally feed back to lower BLH.In this case, the lower BLH may be caused by the decrease of sensible heat at the surface.(4) Aerosols weaken surface winds and atmospheric circulation within the PBL and increase atmospheric circulation above the PBL.
Long-term visibility data, a proxy for aerosol loading, have been used to infer the impact of aerosols on meteorological variables in the PBL [137,138,140].Long-term (>50 years) trends in many meteorological variables were analyzed in central-western China, where aerosols are abundant and strongly absorbing [45].By virtue of the special topography of the region (a large mountain range REVIEW and a broad plain), changes due to background dynamics (mountain stations), the aerosol or pollution effect (contrasting trends between mountain top and bottom), and urbanization (the big city of Xi'an and a rural station located on the plain) were successfully differentiated.Below are the major observational findings and conclusions that offer some clues about the potential influences of aerosols on the PBL and meteorological variables inside the PBL layer [137,138]: (1) The difference in the daytime maximum temperature between plain stations (inside the PBL) and mountain-top stations (outside of the PBL) has decreased, implying cooling of the surface during the daytime, due presumably to aerosols.
(2) Wind speed has decreased over the plain stations, but increased at a neighboring mountaintop station, with the largest difference occurring around noon (most sensitive to aerosols), implying a trend of stabilization inside the PBL and destabilization outside the PBL.This has an important implication for the dispersion of pollutants from low lands, in particular, basins.(3) The number of thunderstorms in the plains has decreased substantially, but little change is observed at the neighboring mountain-top station, implying weakening convection over the plains.(4) The number of rainy days has decreased, implying the likely suppression of PBL clouds, which also has a bearing on the removal of pollutants by rain scavenging.
The finding of decreasing surface winds and increasing winds aloft is consistent with the hypothesis originally proposed by Jacobson and Kaufman [139] based on their model simulation with absorbing aerosols.The finding of decreasing thunderstorms is at odds with the aerosol invigoration effect, suggesting a dominant role of the aerosol thermodynamic effect due to strong aerosol absorption.The hypothesis was confirmed using Tropical Rainfall Measuring Mission (TRMM) thunder data in southeast China, where sulfate aerosols are more prevalent [140].The trend is opposite to that found in central China, where aerosols are more strongly absorbing.
The effect of aerosols on atmospheric thermodynamics is more clearly revealed by the relationships between aerosol loadings and near-surface temperature inversions derived from continuous measurements from an atmospheric emitted radiance interferometer (AERI) deployed at the US Southern Great Plains site for over a decade.Atmospheric temperature profiles derived from the AERI have been used to study the climatology of temperature inversions [28].It is expected that the thermal inversions in the lowest troposphere, especially near the ground, are substantially affected by aerosols, particularly absorbing ones.This was confirmed by the observed enhanced frequency of inversions with increasing aerosol loading but declining frequency of inversions with increasing aerosol SSA, as shown in Fig 5.This indicates that temperature inversions tend to occur at high and absorbing aerosol conditions.Among the absorbing aerosols, BC is one of the most important categories, accounting for ∼10%-50% of the total tropospheric aerosol particles.The atmospheric effects of BC largely refer to the interference with radiative transfer, visibility impairment, PBL stabilization, and alteration of cloud formation, which are strongly sensitive to the aging processes and mixing states with other aerosol constituents [141][142][143].
Comprehensive measurements of atmospheric chemical composition, the PBL, meteorological parameters and surface flux at ground-based stations provide opportunities to gain further insights into aerosol-PBL interactions [144].Based on such measurements at the Station for Observing Regional Processes of the Earth System (SORPES), a 'golden' case that occurred on 10 June 2012 was studied by Ding et al. [15].Compelling evidence of aerosol-PBL-weather interactions was revealed.A mixed layer of heavy biomass burning and urban pollution plumes substantially reduced the amounts of solar radiation, latent and sensible heat fluxes, and air temperature on the ground and in the lower PBL, which suppressed convection and the formation of a would-be major rain event in Nanjing.Increased hygroscopic effects related to increased RH associated with decreased air temperature [15] and a faster secondary aerosol formation [16] could enhance aerosol-meteorology interactions.By analyzing long-term continuous measurements of aerosols, radiation and fluxes at the SORPES station in Nanjing and guided by theory, quantitative relationships were established between PBL turbulence flux, aerosol concentration and RH [27].Suppression of vertical turbulence mixing confined aerosols to a shallower PBL, causing a positive feedback loop between aerosols and the PBL that further lowers the BLH [145].
Intensive measurements made throughout the day help understand the aerosol-PBL relationship.For example, using vertical profiles of BC measured by a micro-aethalometer, Ran et al. [146] found weak turbulence and an inversion layer under high concentrations of aerosols in a shallow PBL.Aerosols were diluted by the fast development of the PBL and a uniform distribution of aerosols was typically observed during the daytime in an unstable PBL.In the evening, aerosols quickly build up near The method and dataset used are described in [28].
the surface and decline exponentially with height, followed by the collapse of the mixing layer and then the formation of a stable nocturnal boundary layer [146,147].

Aerosol and convection interactions
PBL-aerosol interactions also affect convection, due to aerosol-induced changes in the atmospheric profile of heating, and surface latent and sensible heat fluxes that significantly affect the evolution of the boundary layer.Gu et al. [148] found that for East Asia, the modeled radiative impact of boundarylayer aerosols suppressed tropical convection.
Convective potential available energy is controlled by boundary-layer temperature and humidity [149].The source of energy for both the convective boundary layer and deep convective cells ultimately originates from surface heating and moisture [150,151].
The PBL also affects deep convection [3,152,153], especially if it hosts absorbing aerosols that can provide enhanced potential energy above the PBL [22].Wang et al. [154] shows that absorbing aerosols residing in the PBL can destabilize the lower atmosphere at the periphery of tropical cyclones, enhance convection in the rainband region, but cut off the energy inflow to the eyewall.More recently, the conditional enhancement of instability by absorbing aerosols was simulated by Fan et al. [155] and Lin et al. [156].In those simulations, due to the radiative heating caused by soot particles in the PBL, both relative humidity and convection strength associated with shallow cumuli are reduced during daytime.However, the altered daytime temperature and moisture conditions act to reserve the energy, thereby facilitating the development of nighttime deep convection.

Aerosol-PBL-chemistry interactions
Variations in the macro-and micro-physics of the PBL may alter the photochemical and thermal chemical reaction conditions in the PBL and thus affect the near-surface air quality.Any variations in photolysis rate, temperature, humidity and concentration of species and structures of the PBL result in the adjustment of photochemical and thermal chemical reaction rates, the chemical equilibrium constant and even changes in chemical products.Moreover, small modifications of solar radiation, atmospheric stability and the structure of the PBL could induce significant changes in the chemical environment, for instance in the ozone photochemistry [157] or on new particle-formation processes [158,159].Figure 6 shows that the circulations and boundarylayer structures over the downstream Kunshan are closely associated with the upstream urban surface (Shanghai), which further affects the O 3 concentration by redistributing O 3 and its precursors.To be specific, the horizontal transport of O 3 and its precursors, from upstream Shanghai to downstream Kunshan, are suppressed in the lower PBL but strengthened in the upper PBL due to the strong circulation caused by the urban heat island effect.A significant attenuation (>50%) in ultraviolet (UV) radiation due to atmospheric aerosols in polluted and urban areas has been reported [160].Such an attenuation of UV radiation by aerosols can exert a significant influence on photolysis and species chemical cycles, especially photochemical reaction processes.The brown carbon emitted from biomass burning diminishes UV-B radiation so strongly that it can reduce the net production rate of ozone by up to 18% and the mass concentrations of HO 2 , radicals OH and RO 2 by up to 15%, 17% and 14%, respectively [161].From observations and simulations using a radiative-transfer/air-quality model, UV-scattering particles in the PBL tend to accelerate photochemical reactions, as opposed to the inhibition effect caused by UV-absorbing aerosols [162].The hygroscopic growth of aerosol particles normally occurs in polluted air with relatively high humidity, which can largely affect their SSA.In comparison with the dry state condition, the calculated J NO2 at RH = 98% at 1 km AGL increased by 30.4% due to the UV radiation enhancement induced by the larger-sized humidified scattering aerosol particles [163].The influence of aerosol hygroscopic growth on the J NO2 profile inhibits photolysis at the surface and accelerates it in the upper PBL by a similar mechanism described by Dickerson et al. [162].This amplification of J NO2 in the upper PBL likely brings about high aerosol concentration and more ozone production in the polluted upper PBL and free troposphere [163].
The direct and indirect effect of aerosols can alter the photochemical reactions and ozone concentration [164][165][166].For instance, absorbing aerosols were found to be able to reduce the photolysis rate and weaken ozone generation, while scattering aerosols showed opposite change [167,168].Li et al. [169] found that PBL O 3 was reduced by 5% in highly polluted regions in summertime in Central Eastern China.Deng et al. [160] found that UV radiation/ozone is negatively correlated with PM 10 , and at least half of the UV radiation was attenuated by the atmospheric aerosols.Li et al., [170] found a reduction of about 2%-17% in surface ozone during the daytime in Mexico City due to the changes in REVIEW Li et al. 821 photolysis rates caused by aerosols.The simulation by Cai et al. [171] shows that an elevated concentration of particulate matter can increase the AOD by 20%-40% and reduce photolysis rates of NO 2 and O 3 by 20%-30%, resulting in a reduction of 30%-40% in the net photochemical production rate of O 3 .Aerosols also exert an important influence on ozone concentrations by heterogeneous reactions.Ravishankara [172] determined the role, rates and media of heterogeneous reactions in the troposphere.Jacob [173] proposed that heterogeneous reactions have an impact on O 3 concentrations by affecting generation and consumption of NO x , HO x , O 3 and halogen radicals.Heterogeneous reactions on sea salt and soot surfaces have significant effects on the trace gas [174][175][176][177]. Kleffmann et al. [178] and Kaiser et al. [179] noted that BC has relatively larger surface area, favoring heterogeneous chemistry.
Liao and Seinfeld [180] found that the surface O 3 concentrations in eastern China can be reduced by 25%-30% because of the heterogeneous reactions on wet surfaces of sea salt, nitrate, sulfate, ammonium, mineral dust and organic carbon aerosols.Overall, the heterogeneous reactions reduce annual mean O 3 in eastern China by 10%-18% [182].These reactions take up ozone precursors such as NO x and N 2 O 5 , leading to reduced O 3 concentrations, which explains in part why O 3 concentrations have kept increasing while PM 2.5 concentrations have decreased, which has been happening in China in recent years as a direct result of emission control measures for aerosols.Note that the impact of heterogeneous processes is strongly dependent on aerosol concentrations and the surface uptake coefficients [181].Heterogeneous reactions also likely play a central role in the formation of major aerosol ingredients, including sulfate, nitrate and organics [183][184][185].
Urbanization also affects local and downstream air quality in two major ways.First of all, land surface properties and meteorological fields modified by urban canopies alter the chemical reaction, dry and wet depositions, and the spatial distribution of primary and secondary atmospheric pollutants.Second, urbanization processes tend to enhance local human activities, thereby scaling up anthropogenic emissions.Both pathways will exert non-negligible influences on air quality on local and regional scales.Zhang et al. [186] showed that without an upstream city, the urban heat island (UHI) effect over Baltimore would be 1.25 • C weaker and the PBL would be 200 m shallower, which could redistribute the air pollutants throughout Baltimore.Anthropogenic heat emissions can significantly change the UHI and urban-breeze circulations in cities in the Yangtze River Delta region, which in turn changes the spa-tial and vertical distributions of the simulated air pollutants [187].Urban aerosols may also contribute to the UHI based on satellite observations and urban climate simulations [188].The urban-rural difference in haze pollution levels is one of the key factors determining nighttime UHI across China.
Rapid vertical mixtures in the convective boundary layer (CBL), along with a temperaturedependent partitioning of atmospheric nitrate between the gas and aerosol phases, results in complex interactions between dynamics and aerosol formation [189].Near the top of the CBL (cooler), gaseous nitric acid and ammonia condense on ammonium nitrate and the gas-aerosol equilibrium shifts towards the aerosol phase.Close to the surface (warmer), ammonium nitrate evaporates into gaseous nitric acid and ammonia, shifting the equilibrium towards the gas phase.Using a large-eddy simulation (LES) model coupled with radiation, chemistry and the surface exchange of aerosols, Barbaro et al. [190] highlighted that the close connection of the gas-aerosol conversion of nitrate to CBL (thermo-)dynamics produces highly nonlinear concentration and vertical profiles of turbulent flux.
Although a great deal of phenomena related to aerosol-PBL-chemistry interactions have been reported, most of the explanations for the underlying mechanisms are qualitative and a deeper insight is still warranted.

Aerosol, PBL and cloud interactions
The interactions between cloud-nucleating aerosols and the PBL involve complex feedbacks between cloud microphysics, precipitation and PBL turbulence.Precipitation is a key component regulating the PBL evolution.When the aerosol concentration increases, more but smaller cloud droplets are formed and precipitation is suppressed [2], leading to significant changes in PBL evolution [191][192][193].
In a stratocumulus-topped PBL, the effects of heavy drizzle on PBL turbulence and structure were simulated with a LES model developed by Stevens et al. [191].Significant evaporative cooling in the sub-cloud layer ensued and weakened mixing from the cloud layer to the sub-cloud layer.Less turbulent kinetic energy (TKE) was generated in the PBL, leading to a weaker cloud-top entrainment.As a result, the growth rate of the PBL was slowed down and the PBL became thinner.PBL turbulence and cloud-top entrainment can become significantly stronger with increasing aerosols [193].The growth rate of a stratocumulus-topped PBL increases with aerosol loading.Because the air in the inversion layer REVIEW above the PBL is normally warm and dry, the enhanced cloud-top entrainment can cause warming and drying of the PBL and reduce the cloud liquid water path (LWP) [194].
However, in a stratocumulus case with very weak drizzle, the evaporative cooling and moistening right below the clouds can create instability for the subcloud layer.The weak drizzle then helps to produce TKE in the PBL.When aerosols are increased, the reduced drizzle results in weaker PBL turbulence [192].This is because there is less evaporative cooling and less moistening below clouds, and therefore weaker instability in the sub-cloud layer.The PBL then has a less effective supply of surface water vapor to clouds and therefore a much lower LWP.The reduced drizzle in the high cloud condensation nuclei case is believed to weaken the coupling between the surface and air in the PBL, and to reduce the vertical fluxes of heat and the transport of water vapor.This weakens any cloud development that may ensue.
In a cumulus-topped PBL, the aspect ratio, the turbulent mixing process, cloud organization and formation mechanisms of shallow cumulus clouds all differ from the stratocumulus cloud case.As cumulus clouds introduce liquid water into the inversion layer, the subsequent evaporation provides moisture and the dry inversion-layer air gradually becomes moist and takes on the characteristics of the cloud layer, therefore deepening the PBL [195].Also, the non-precipitating cumulus-topped PBL grows proportionally with time.It was hypothesized that the precipitation from shallow cumulus can arrest the growth of the PBL by the removal of liquid water from the cloud top.However, to what extent the aerosol-induced changes in precipitation can affect the PBL growth rate has not been well studied.
As aerosols increase in the cumulus-topped PBL, PBL turbulence is affected by two processes.Xue and Feingold [196] described one process in which increasing aerosols results in smaller cloud droplets, and therefore faster evaporation at the cloud top and edges.The faster evaporation tends to generate stronger evaporative cooling and stronger downdrafts at cloud edges, which produces more TKE and enhances the evaporation of the clouds.This results in a smaller cloud fraction and thinner clouds.In another process described by Grabowski et al. [197], increasing aerosols leads to precipitation suppression, less efficient condensate removal, and therefore weaker buoyancy due to the water loading, resulting in shallower clouds.Although both of the mechanisms suggest shallower clouds under higher aerosol-loading conditions, their underlying reasons are completely different.
The patchy precipitation in cumulus clouds can cause a temperature anomaly (cold pool) and a moisture anomaly (more moisture) in the sub-cloud layer [198].The evaporative cooling and associated downdrafts lead to divergence at the center of the precipitating cells at the surface and convergence at the edges of the precipitating cells.This facilitates the formation and development of new clouds, resulting in a mesoscale open cellular structure [199,200].In the polluted case, increased aerosols can suppress precipitation and the formation of open cellular structures.
In a cumulus-topped PBL, cumulus clouds are also very efficient at vertically transporting aerosols [201].When the source is at the surface, aerosols are transported upwards mainly through the updraft regions of cumulus clouds.When the source is in the inversion layer, which means that aerosols are transported to the studied region by the free atmosphere, aerosols can be transported downward mainly in the downdrafts of the shell regions of cumulus clouds.
In summary, most studies have shown that increasing aerosols can suppress precipitation, as expected from the traditional theory of cloud microphysics.However, it has feedbacks to PBL turbulence, surface processes and radiation through several pathways.Further studies are needed to reveal the mechanisms underlying these interactions in the PBL.

THE TREND AND FLUCTUATION OF AIR POLLUTION: THE ROLES OF CIRCULATION, PBL, CLIMATE CHANGE AND WEATHER
Concentrations of pollutants (aerosols and precursor gases) are driven by emission, transformation (e.g.gas to particle conversion), transport and deposition.These drivers may be classified as being chemical (e.g. the availability of oxidants), meteorological (e.g.wind speed, temperature, humidity, precipitation, soil moisture, solar radiation) and biological (e.g.vegetation cover and properties), etc.While all these factors are at work in dictating air pollution at any time and location, we argue that the following factors are most essential: (1) Emissions of precursor gases, primary and secondary aerosol particles; (2) Gas-to-particle, or new particle formation and growth; (3)   Since the first two topics will be addressed in separate review articles of the same special issue, we will just elaborate more on (3) to (5).

Large-scale circulation and local-scale aerosol-PBL interactions
Air pollution is intimately related with large-scale circulation and local-scale aerosol-PBL interactions [66,[202][203][204][205][206][207].High-resolution mesoscale meteorological models, e.g. the MM5 and the WRF, have been widely used to study their joint roles.For example, MM5 simulations suggested that the evolution of sea-land breezes and PBL dynamics were influenced by an approaching northwest Pacific typhoon [205].Based on WRF-FLEXPART simulations, it was found that the PBL air pollutants from the North China Plain can be transported by cyclones and its associated warm conveyor belts to the free troposphere over Northeast China [206].This study showed that Lagrangian modeling connected to high-resolution meteorological output from mesoscale models could well demonstrate detailed air pollution transport and dispersion mechanisms under specific synoptic weather conditions.Also, based on offline air-quality models [208,209], many previous studies have been conducted to un-derstand PBL aerosols and their impacts in typical regions of China under different synoptic conditions [210][211][212][213][214][215].
In comparison with 'offline' air-quality models that often underestimate extreme aerosol peaks during severe haze episodes [215], 'online' coupled models considering the aerosol-radiation-PBLweather feedback can improve the forecast capabilities of severe pollution events taking place in Beijing during wintertime [216,217].For the WRF-Chem simulation with mixed biomass burning with urban plumes [15], it showed that daytime mixed biomass burning plumes not only 'burn off' daytime precipitation but also enhance nighttime precipitation in downwind regions [218].It was also found that dust aerosols from Northwest China could have a strong aerosol-PBL feedback and influence the PBL structure along transport pathways, which could even influence the emission and deposition of dust in the source and downwind regions [219].
Using the online-coupled WRF-Chem numerical model, Ding et al. [24] first singled out the role of BC aerosol and PBL interactions in polluted events and attributed an extreme haze episode in East China to the positive feedback between absorbing aerosol and PBL.The feedback tends to lower the BLH, which was thus referred to as the 'dome effect', as illustrated in a simplified conceptual scheme (Fig 7) that is valid under the static state of the PBL.Under general conditions, mixing takes place due to any inhomogeneous heating of the atmosphere.The extra heating by aerosol absorption in the upper PBL may be mixed down into the interior of the PBL to lower the PBL.
The hypothesized 'dome effect' seems to prevail even on the decadal trend of AOD retrieved from satellites (MODIS and MISR) at different altitudes.As shown in Fig 8a, Dong et al. [145] found opposite trends for AOD: decreasing and increasing below and above ∼0.5 km in northern China, which is dominated by strong absorbing aerosols with low SSA (45).The increase in the lower PBL is a testimony to the positive feedback caused by a reduced temperature lapse rate or even inversion that keeps pollutants accumulating, whereas above the layer of maximum absorption conditions the PBL becomes more unstabilized and more favorable for diffusional transport of pollutants.As such, even though the overall basin-wide column total AOD may not have changed much, the aerosol loading near the surface in the center of the basin has increased drastically.To reinforce the argument, they also analyzed similar trends in southern China (cf.change, due presumably to the lack of the positive feedback process.
These findings suggest that absorbing aerosols not only alter atmospheric thermodynamics and stability, but can also effectively push down the BLH to severely exacerbate air pollution near the surface.Apparently, such a deterioration of air quality has little to do with the emission of pollutants, but more to do with their accumulation in a thinning PBL.This could be an important mechanism that ought to be accounted for in understanding and forecasting air pollution.
The mechanism may also help explain the systematic difference in the PBL height between the US South Great Plains and Hefei, Anhui, China, as shown in Fig. 9.The two sites have somewhat similar meteorology, but aerosol concentrations and types are distinctly different.AOD is much larger at Hefei than at the SGP, and the opposite is true for the PBL height.Besides, aerosols in China are generally more absorbing than the rural area of US due to much less consumption of coal.
The mechanism governing aerosol and PBL interactions for absorbing aerosols is illustrated in Fig. 10.As a result of reduction in surface fluxes and atmospheric heating, the lapse rate decreases and probability of inversion increases.These are unfavorable for the dispersion of pollutants that would further enhance atmospheric stability, etc.

Aerosols and climate change
Increased temperature can enhance the chemical production of sulfate [220] but reduce nitrate formation through shifting gas-particle equilibrium [221,222].All aerosol species are found to be very sensitive to changes in precipitation scavenging [223][224][225].As such, understanding of any long-term changes in air pollution must take into account the changes in climate [6,7], as well as the day-to-day change in air quality [103].In Asia, the monsoon is the most important dynamic regime dictating the Asian atmospheric environment whose changes are key to understanding any long-term trend of air pollution [7].On the other hand, climate changes in Asia, especially in China, are also strongly affected by anthropogenic activities.
Climate change influences aerosol concentrations in China on different timescales, including seasonal [226], interannual [135,[227][228][229][230] and decadal variations [135,227,231,232].Summer monsoon rainfall, cross-equatorial flows that carry clean air from the oceans and the relatively higher BLH lead to generally lower aerosol concentrations in summer than in winter [226].In North China (32-42 • N, 110-120 • E), interannual variations of 11%−17% relative to the mean pollution concentrations were simulated over 2004−2012 [228], indicating that the aerosol effect on the PBL could vary greatly year by year.As special cases of aerosol interannual variations, seasonal mean aerosol concentrations in eastern China have been reported to correlate negatively with East Asian monsoon strength for both summer and winter [7,135,227].For decadal variations of aerosols, Yang et al. [231] reported that, although changes in anthropogenic emissions dominated the increasing trend in wintertime PM 2.5 concentrations over eastern China, variations in meteorological parameters contributed 17(±14)% to the overall increasing trend in PM 2.5 concentrations in this region from 1985 to 2005.
Climate change also influences severe fog and haze events, especially in northern China in winter.Niu et al. [135] found that decadal and persistent decreases in visibility were associated with a longterm reduction in cold-air outbreaks from Siberia, and an increase in calm or low-wind days, which were further attributed to global warming by means of global climate model (GCM) simulations [135].Cai et al. [59] developed an effective haze weather index to represent such favorable weather conditions by using observed long-term PM 2.5 daily concentrations at Beijing and daily reanalysis of meteorological fields, and showed that favorable weather conditions for severe haze increased by 10% over 1982-2015 relative to 1948-1981, owing to global warming by greenhouse gases.Other studies also showed consistent conclusions.As a result of global REVIEW warming, reductions in Arctic sea-ice cover correlated with the increases in wintertime haze days in northern China [60,233], and reduced mineral dust aerosol emissions led to more stagnant conditions favorable for haze days over eastern China [234].In addition to climate change by anthropogenic forcing, natural climate variability was found to influence haze days.Decadal variability in the occurrence frequency of wintertime haze in central eastern China was reported to be closely associated with the Pacific Decadal Oscillation [235].

Aerosols and weather regimes
The vertical distribution of aerosol particles is significantly affected by meteorological conditions [66,236] such as convective transport and removal processes [237], and the PBL structure and processes [238][239][240][241][242]. Aerosol concentrations and effective radii typically differ within and above the PBL since the BLH and the intensity of the turbulence affect to some degree the 3D distribution of air pollutants [243,244].It has been observed that fine particles (< 2.5 μm in aerodynamic diameter) tend to be well mixed vertically during the daytime coarse particles tend to reside close to the ground and surface due to gradation settling [245].
Modeling studies have suggested that the vertical distribution of aerosols tends to be modulated by local atmospheric circulations [241,246,247] such as mountain-valley breeze circulation and sea-breeze circulation.In mountain-valley/plain regions, the presence of a mountain-valley breeze circulation can modify the PBL structure over valleys/plains through dynamic advection and the accompanying thermal alteration [241,248,249].Meanwhile, the upslope wind of a valley breeze circulation may bring aerosols to the top of the PBL and form an elevated pollution level there [246,250,251].In coastal regions, the updraft induced by the frontal movement of a sea breeze may also bring near-surface aerosols to the top of the PBL [241,252,253].

CONCLUDING REMARKS
The PBL involves a number of chemical, physical and dynamic processes that are closely related to the accumulation of aerosol pollution, dispersion and transport.There are complex interactions between aerosol and PBL that jointly dictate air pollution.PBL-aerosol interactions and their feedbacks on meteorology thus complicate the forecasting of air quality.
This paper presents a comprehensive review of aerosols, the PBL, their interactions and impact on air quality.While we have learned a great deal on all pertinent fronts, there is still a wide range of fundamental PBL characteristics or key boundarylayer processes that are poorly understood or overlooked.Most work related to PBL processes is based primarily on time series of measurements made at fixed sites chiefly through field experiments that have a poor spatial coverage.A dense network of both aerosol and meteorological measurements is urgently needed to gain an explicit insight into the evolution of the PBL and the roles of aerosols played in combination with atmospheric processes driven by both large-scale and local-scale processes.This would improve PBL parameterizations and turbulence closures used in the models.A major challenge lies in the observation of the vertical profiles of atmospheric variables and aerosol properties, none being trivial at present but essential to improve our understanding of aerosol-PBL interactions.This requires more state-of-the-art ground-based and airborne measurements in different regions.For example, the BLH is a key variable whose observation is still far from being adequate.Sounding balloons are generally launched twice per day, making it impossible to study the diurnal variation of the PBL during a day, whereas space-borne lidar has a low signal-noise ratio, and thus very large uncertainty, due to strong attenuation, besides being available at nadir view only.Likewise, there is a severe dearth of vertical profile measurements of aerosol absorption that is the key for aerosol-PBL interactions.
In addition to boundary-layer measurements, PBL modeling efforts such as explicit PBL schemes have to be improved in operational weather or climate models.Current aerosol-climate models are able to simulate aerosol-climate interactions on seasonal, interannual and decadal timescales, but few studies have examined the changes in PBL from climate simulations.Attention has mostly been paid to atmospheric temperature, circulation and precipitation in a changing climate.PBL schemes are generally included as sub-models in atmospheric chemistry models, and numerical weather or climate prediction models, and are typically limited to a certain scenario.The coarse horizontal and vertical resolutions of climate models prohibit the accurate representation of boundary-layer processes.Many PBL schemes were developed mainly for applications in low-resolution models.These schemes would be difficult to apply to increasingly high-resolution models that require more detailed and explicit representations of physical processes and physiographical features.Considering that climate models are moving forward to higher resolutions, accurate representations of land use, sub-grid thermal and dynamical structures, as well as the radiative effects of aerosols on the boundary layer, are the keys to REVIEW Li et al. 827 improving simulations of boundary layer, air quality and climate.Continual development of PBL models and parameterization schemes is thus highly recommended.Forecasting of air quality may be improved significantly by accounting for the aerosol-PBL interactions, which further require both aerosol loading and aerosol-absorption property, including its vertical distribution.
While the aerosol-PBL interaction plays a key role in air pollution, it is by no means the sole mechanism dictating air pollution, which is affected by a lot of different mechanisms and processes, but they are not the foci of this review.Other pertinent factors and their associations with aerosol pollution are pursued in other review papers, such as the effects of new particle formation [25], and large-scale circulation associated with monsoon systems [7].

Figure 1 .
Figure 1.Annual mean 3D occurrence (color shaded) for (a) all aerosol types, (b) dust aerosols and (c) smoke aerosols, which are derived from level-2 CALIOP aerosol layer products, in combination with MODIS/Aqua AOD data (whiteblack shaded) for the period October 2006-September 2014.Note that only those with frequency greater than 5% are shown here.(Adapted from [78].)

Figure 2 .
Figure 2. Spatial distribution of mean BLHs derived from summertime sounding measurements at (a) 0200, (b) 0800, (c) 1400 and (d) 2000 Beijing time in summers of the period 2011 to 2015 throughout the L-band China Radiosonde Network as operated by the China Meteorological Administration (colored dots), overlaid with those derived from simultaneous ERA-Interim reanalyses (color shading).(Adapted from [108].)

Figure 3 .
Figure 3. Evolution of BLHs calculated using micropulse lidar (MPL), atmospheric emitted radiance interferometer (AERI) and radiosonde data, overlaid on the MPL backscatter field during a nine-day period (2-10 May 2004) at the Southern Great Plains site in the US.(Adapted from [116].)

REVIEW
Li et al.817

Figure 4 .
Figure 4. Seasonal daily averaged vertical profiles of aerosol particle heating rate (black solid lines), with their corresponding standard deviations (gray horizontal lines) at Taihu in the central Yangtze River Delta region of eastern China in (a) spring, (b) summer, (c) autumn and (d) winter during the period from June 2008 to May 2009.(Adapted from [89].)

Figure 5 .
Figure5.Impact of aerosol on temperature inversion in the lowest troposphere.Upper panel: A time series of temperature profiles retrieved from the atmospheric emitted radiance interferometer (AERI) measurements on 30 October 2008 over the US Southern Great Plains (SGP).Lower left: the frequency of occurrence of temperature inversion near the surface as a function of the number concentration of condensation nuclei (CN) derived from nine years of observations at the SGP site, which is part of the US Atmospheric Radiation Measurements (ARM).Lower right: same as the lower left panel but as a function of aerosol single scattering albedo.The method and dataset used are described in[28].

Figure 6 .
Figure 6.Vertical cross-section of the difference in simulated O 3 concentration (ppbv) and in-plane vectors, where the vertical speed is multiplied by a factor of 10, between urban land use and no urban land use in Shanghai simulations.(Adapted from [157].)

Figure 7 .
Figure 7.A schematic figure showing the aerosol-boundary-layer feedback loop for scenarios without (left) and with (right) black carbon (BC) emissions in a megacity.The black lines give air temperature profiles (solid, dotted and dash-dotted lines for the scenarios with BC, without aerosols, and with aerosols except for BC, respectively).The yellow dashed lines with arrows denote the reflection of solar radiation by the ground surface, clouds and aerosols.The red arrows show absorption of solar radiation by absorbing aerosols.The blue dash-dotted line indicates the top of the PBL.White arrows show the vertical ventilation of urban plumes induced by circulations or large eddies induced by the urban heat island effect.(Adapted from [24].)

Figure 8 .
Figure 8.The vertical profiles of AOD trends over the period of 2002-2014 as a function of altitude above sea level in northern (low SSA) and southern (high SSA) China derived from MODIS AQUA.(Adapted from [145].)

Figure 9 .Figure 10 .
Figure 9. Boxplots showing the distribution of PBL depths with increasing AOD, at SGP (left) and Hefei (right).Adapted from the Ph.D dissertation of V. Sawyer (2015).

Appendix 1 :
List of acronyms and abbreviations used in this paper.