On the Relationship Between Aerosol and Boundary Layer Height in Summer in China Under Different Thermodynamic Conditions

Abstract

The observed relationships between boundary layer height (BLH) and PM_2.5 on a national scale remain unclear due to the dearth of observations. Here we investigated this relationship from a unique perspective of thermodynamic stability in the planetary boundary layer (PBL), using summertime (June–August) soundings from China for the period from 2014 to 2017. For all three times of soundings (0800, 1400, and 2000 Beijing time [BJT]), positive (negative) PM_2.5 concentrations anomalies were found to correlate with negative (positive) BLHs anomalies relative to daily means. The negative correlation was strongest at 1400 BJT, followed by 2000 BJT and 0800 BJT. Overall, the PM_2.5 was found to nonuniformly anticorrelate with BLH across China at 0800 and 2000 BJT. The strongest anticorrelation occurred in the North China Plain at 1400 BJT, in sharp contrast to the much weaker correlation in other regions characterized by much less polluted regions. The averaged PM_2.5 in neutral boundary layers was higher than that in convective boundary layers (CBLs). The CBL, where the anticorrelation was the strongest, was conducive to dissipating more aerosol in the heavily polluted area in China than neutral boundary layer. The higher CBL formed under low cloud cover, low surface humidity, and strong wind speed was favorable for the dispersion of aerosol, in contrast to the stable boundary layers that happen under the highest cloud cover. Also, positive correlation was seen between stable boundary layer and PM_2.5. The findings call for attention that the thermodynamical condition of PBL should be considered when examining the aerosol‐PBL interactions.

Introduction

As the lowest part of the troposphere, the planetary boundary layer (PBL) is strongly influenced by the turbulence and complex geophysical processes due to its close contact with the Earth's surface (Garratt, ; Stull, ). The coupling of the PBL and the surface is largely modulated by turbulent exchange processes, including the intensive vertical mixing of momentum, water, heat, and air pollution between surface and troposphere (Li et al., ; Luan et al., ; Mao et al., ; Zhu et al., ). The role of aerosol in modulating the evolution of PBL has been increasingly appreciated due to the considerable radiative effect (Guo, Deng, et al., ; Li et al., ; Patil et al., ; Wang et al., ; Xu et al., ; Zheng et al., ), especially the dome effect caused by the absorbing aerosol (e.g., black carbon) that tends to stabilize the PBL (Ding et al., ; Jiang et al., ).

The dispersion of atmospheric pollutants depends mostly on atmospheric turbulence and horizontal wind, which accelerates the vertical and horizontal diffusion of aerosol (Ding et al., ; Pal et al., ; Zheng et al., ). On the other side, except for the contributions of local pollutant emissions, atmospheric transport, and atmospheric chemistry (Guo et al., , ; Qin et al., ; Yao et al., ; Yim et al., ), the role of meteorological conditions cannot be ignored, such as strong temperature inversion or large‐scale high‐pressure systems (Miao, Guo, Liu, Liu, Li, et al., ). The increases in absorbing aerosol tend to stabilize the PBL and suppress the PBL, which consequently exacerbates the haze pollution (Ding et al., ; Wang et al., ). This positive feedback mechanism of aerosols on PBL evolution is a key factor in the accumulation of air pollution (Petäjä et al., ; Zhong et al., ). Therefore, the PBL and its evolution cannot be neglected when examining the meteorological factors associated with aerosol pollution (Garratt, ; Seibert et al., ).

Other meteorological variables in the PBL were also found to significantly affect the aerosol concentrations as well. The southerly winds in the PBL generally were accompanied by moderate to severe haze pollution in Beijing (Guo, He, et al., ; Gu et al., ; Liu et al., ). This region generally suffered severe air quality when both calm surface winds and stagnate air prevailed, which were favorable for the formation of fog and haze (Yang et al., ). Besides, the temperature inversion induced by light‐absorbing aerosol was able to impair the dispersion and transport of aerosol particles, especially in the North China Plain (NCP; Xu et al., ) and over central China (Liu et al., ). This inversion was strengthened by cool advection near surface and warm advection above the top of PBL, which in turn resulted in further deterioration of aerosol pollution (Miao, Guo, Liu, Liu, Li, et al., ). Besides, the enhancement of aerosol concentrations were tightly associated with increases in cloud cover (CLD; Miao, Guo, Liu, Liu, Li, et al., ). This is likely due to the fact that clouds tended to reduce the solar radiation reaching ground surface and thus suppress vertical turbulence and mixing (Guo, Miao, et al., ; Zhang et al., ). Among others, the increases in aerosol concentrations were accompanied by high relative humidity (RH) conditions, where enhanced hygroscopic growth and multiphase reactions facilitated the increases in particle size and mass (Miao, Guo, Liu, Liu, Zhang, et al., ; Tie et al., ; Zhong et al., ). High‐pressure systems were revealed to significantly prevent the diffusion of atmospheric pollutants, causing severe haze pollution (Chen et al., ; Leng et al., ; Wu et al., ; Yang et al., ; Zhang et al., ). In addition, the dynamic and thermodynamic features of the PBL can be substantially modified by aerosols, which tends to stabilize or invigorate the PBL depending on the aerosol types or the altitudes where aerosol particles aloft reside (Barbaro et al., ; Talukdar et al., ; Wilcox et al., ).

Elucidating the relationship between the PBL and air pollutions is not only significant for prevention and control of aerosol emissions in China but also indispensable for accurate air quality prediction and numerical weather prediction. This has been widely studied using the BLH estimates from lidar observations, radiosonde measurements, satellite remote sensing, and model simulation (Li et al., ; Miao et al., ; Pasch et al., ; Su et al., ; Tsaknakis et al., ; Wang et al., ). Previous studies involved in the interaction between aerosol and PBL in China were either limited to a specific region or limited time period (Du et al., ; Leng et al., ; Quan et al., ). This is partly due to the lack of national‐level high‐resolution observations available for the BLH estimates in China (Zhang et al., ). To the best of our knowledge, there are not any previous studies that have attempted to examine the relationship between aerosol and BLH by considering the PBL regimes. Typically speaking, the PBL can be divided into convective boundary layer (CBL), neutral boundary layer (NBL), and stable boundary layer (SBL). The NBL generally refers to the neutrally stratified residual layer that starts from the ground surface (Liu & Liang, ). The spatial and temporal variability of these three PBL regimes over China were revealed by Zhang et al. (). Nevertheless, the knowledge regarding how these three dominant PBL types interact with aerosol pollution remains poorly understood.

The China radiosonde network operated by the China Meteorological Administration (CMA) provides us a unique opportunity to study the features of PBL across China (Guo, Miao, et al., ; Zhang et al., ). Given the ignorance of PBL regimes in previous studies, one of the principle goals is to examine the interaction between aerosol and BLH under CBL, NBL, and SBL conditions. Additionally, the potential association among aerosol pollution, PBL, and other meteorological conditions will be explored.

Data and Methods

In the present study, we used a variety of meteorological observations, including the L‐band radiosondes, PM_2.5 concentration and ground‐based meteorological data in summer for the period of 2014 to 2017. The high‐resolution (1.2 s) sounding measurements were obtained from China radiosonde network operated by CMA, which comprised 120 operational radiosonde stations (Guo, Miao, et al., ). These data began to report the profiles of temperature, pressure, RH, wind speed (WS), and wind direction at 1.2‐s resolution since 2011. The World Meteorological Organization mandates that the sounding balloon was launched twice a day at 0000 UTC (0800 Beijing time [BJT]) and 1200 UTC (2000 BJT). For promoting the prediction of high‐impact weather, additional soundings were made at 0600 UTC (1400 BJT) and 1800 UTC (0200 BJT) during intensive observing periods in the summertime (June–August) at most radiosonde sites. The quality control and assessment of sounding measurements were detailed in Guo, Miao, et al. (). Technically, the L‐band soundings are good enough to be used for the estimates of BLHs.

We estimated two kinds of BLHs. The first BLH did not considersounding measurements were detailed in the static stability of atmosphere. Following the methods in Guo, Miao, et al. (), the bulk Richardson number method (Vogelezang & Holtslag, ) was utilized to estimate the BLH. This method can be applied for both CBL and SBL regimes. The second BLH was calculated separately to CBL, SBL, and NBL, according to the methods developed by Zhang et al. (). Specifically, the vertical resolution of 1 hPa in the original L‐band radiosonde measurement was resampled to 5 hPa. Then, the PBL regimes were identified by calculating the near‐surface potential temperature difference (PTD) between the fifth and second levels with the threshold value is 0.1 K, and the bulk Richardson number based on the original lowest 100‐m radiosonde measurements. The PBL was identified as NBL if the PTD lied between −0.1 and 0.1 K or as SBL otherwise when the PTD was higher than 0.1 K and the bulk Richardson number was positive and then the left cases were considered as CBL.

As of November 2013, hourly PM_2.5 concentration data began to be released to the public, most of which were observed by the national‐level stations operated by the Ministry of Environmental Protection of China. The National Meteorological Information Center of CMA provides all of the quality‐controlled meteorological data. To better match hourly PM_2.5 with sounding‐derived BLH, a circle with 60‐km radius centered at a radiosonde station was defined, within which all the PM_2.5 measurements made at the Ministry of Environmental Protection stations were averaged over the hours with radiosonde observations. We normalized the PM_2.5 concentration by the monthly mean in each site to avoid the effect of long‐term trends and spatial variability using the approach proposed by Wang et al. (). Additionally, the sites with the sample size less than 30 here were excluded for further analyses at 1400 BJT.

It is noteworthy that only those radiosonde stations with PM_2.5 observational sites within that circle were further analyzed in this study, as spatially illustrated supporting information Figure S1. Besides, the collocated PM_2.5‐BLH data pairs were removed for further when the hourly rainfall amount was greater than 0.1 mm. As such, there were 94 radiosonde sites along with 586 PM_2.5 stations available at 0800 BJT and 2000 BJT, as compared with 91 sounding sites with 559 PM_2.5 stations available at 1400 BJT (Table ). And Figure  showed the spatial distribution of radiosonde stations with simultaneously observed PM_2.5 and BLHs at 0800 BJT, 1400 BJT, and 2000 BJT. Additionally, other surface meteorological variables associated with aerosol pollution were collected at the same radiosonde stations, including total CLD, 2‐m temperature, 2‐m RH, and 10‐m WS, for the sake of better understanding of the factors modulating the PBL development.

Statistics of the Ground‐Based PM_2.5, Radiosonde Measurements, and Surface Weather Observations at 0800 BJT, 1400 BJT, and 2000 BJT in Summer (June–August) From 2014 to 2017

Data	# of available station	Time resolution	Source
0800/2000 BJT	1400 BJT
PM2.5	586	559	Hour	MEP & CMA
Radiosonde	94	91	Second	CMA
Surface weather observation	522	506	Hour	CMA

Note. CMA = China Meteorological Administration; MEMinistry of Environmental Protection of China.

Fig. 1. Spatial distribution of the radiosonde stations with simultaneously observed PM2.5 and BLHs at 0800 BJT (a), 1400 BJT (b), and 2000 BJT (c) in summer during the period from 2014 to 2017. The radiosonde stations with more than 30 observations at 1400 BJT are highlighted in black in panel (b); otherwise, the radiosonde with less than 30 were excluded for further analysis (in gray). BJT = Beijing time.

Results and Discussion

Spatial and Temporal Variation of BLH and PM_2.5

Figure  showed the geographic distribution of the difference of hourly PM_2.5 and BLH relative to their respective daily means at 0800, 1400, and 2000 BJT across China. Except for over East China where it was susceptible to the influences of sea breezes, the BLHs were generally lower than the daily means in the mornings (0800 BJT). By comparison, positive PM_2.5 anomalies shrouded most parts of China, particularly in Northeast China and NCP, as shown in Figures a and d. The opposite spatial patterns were observed in the afternoons (1400 BJT). That is to say, the negative anomalies of PM_2.5 concentrations corresponded very well to the positive anomalies BLHs (Figures b and e). In contrast, there were little differences between BLHs in the evenings (2000 BJT) and their daily means of BLH in eastern and southeastern China, while negative PM_2.5 still emerged (Figures c and f). Therefore, these shallow (deep) PBLs at least were partially responsible for the high (low) PM_2.5 concentrations, largely depending on the observational timing.

Fig. 2. The difference of (a–c) PM2.5 concentration and (d–f) BLH at 0800 BJT (left column), 1400 BJT (middle column), and 2000 BJT (right column) relative to daily means. Note that rainy days were excluded. BLH = boundary layer height; BJT = Beijing time.

Figures a–c illustrated the temporal evolution of monthly mean BLH and PM_2.5 at 0800 BJT, 1400 BJT, and 2000 BJT in summer from 2014 to 2017 in China. It was apparent that the summertime aerosol pollution exhibited large interannual cycle with the maximum PM_2.5 occurring in 2014, followed by almost linear decreasing trend in the following years to be analyzed, no matter it was at 0800, 1400, or 2000 BJT (Figures a–c). The lowest PBL led to rapid growth of PM_2.5, which reached about 50 μg/m³ at afternoon in July of 2014. The PM_2.5 dropped quickly and was accompanied by the rise of BLH in June of 2015. Similarly, the sharp increases in BLH resulted in the lowest PM_2.5 of 25 μg/m³ at 1400 BJT during the summertime of 2016. Also, it was noticeable that the diurnal variation of PM_2.5 in summer was obvious. The PM_2.5 concentration decreased from 10–80 μg/m³ at in the mornings (0800 BJT) to 10–40 μg/m³ in the afternoons (1400 BJT) then to rise slightly to 10–60 μg/m³ in the early evenings (2000 BJT; Figures d–f).

Fig. 3. The temporal evolution of monthly PM2.5 concentrations (blue line) and BLH (black line) averaged over all stations in China at 0800 BJT (a), 1400 BJT (b), and 2000 BJT (c) in summer (June–August) from 2014 to 2017. The shaded region represents the standard deviation. The 2‐D histogram plots showed the relationships between PM2.5 concentration and BLH in China at 0800 BJT (d), 1400 BJT (e), and 2000 BJT (f), respectively. The color shading represented the frequency of occurrence samples, overlaid with the regression lines (black lines) between PM2.5 and BLH. The corresponding correlation coefficients (R) and the number of samples (N) were given at the top of each panel as well. R with asterisks indicated statistically significant trends at a 99% confidence level. Note the occurrence frequency smaller than 0.05 was not displayed. BLH = boundary layer height; BJT = Beijing time.

The 2‐D frequency of occurrence diagrams (Figures d–f) showed more details with regard to the correlation between all available collocated samples of BLHs and PM_2.5 at three times in a day. The daily peak BLHs occurred in the afternoons, which mainly concentrated on the heights below approximately 3.5 km, with high frequency ranging from 0.7–2.3 km. The BLH was generally higher at 2000 BJT than those at 0800 BJT, mostly below about 1.2 and 0.8 km, respectively (Figure ), which was in good agreement with the findings in Guo, Miao, et al. (). The reason could be discrepancy in sensible heat fluxes at 0800 and 2000 BJT. In the afternoon, the solar radiation effectively heated the surface, the resultant PBL over land was capable of getting full development due to the much vigorous turbulence. After sunset, PBL ceased to grow and began to decay for the disappearance of buoyancy forcing acting on the turbulence. More importantly, there were incomplete dissipated turbulences still in the atmosphere to form the residual layer. When night falls, the bottom portion of the residual layer was transformed into stable boundary layer by the surface radiation inversion (Stull, ). By comparison, the PBL began to rise when the temperature inversion was destroyed by sunrise in the next day. Therefore, the BLHs at 0800 BJT in mainland of China were lower than those of 2000 BJT with the daily peak at 1400 BJT (Figures d–f). Combined with the daily features of PM_2.5 concentration, the result suggested that the daily cycle of BLH played an important role in modulating intraday air pollution.

Interaction Among BLH, PM_2.5, and Other Meteorological Factors

On the monthly time scale, the BLHs at all times were negatively correlated with PM_2.5, and the correlation coefficient (R) was statistically significant at a 99% confidence level (Figures d–f). The correlation was strongest at 1400 BJT, followed by 2000 BJT when R was still much larger than that at 0800 BJT. The strong anticorrelation between BLH and PM_2.5 took place in the afternoons (1400 BJT), compared to that at 0800 and 2000 BJT, because the aerosols emitted into the PBL were dispersed and finally became completely mixed over this layer, due to the thriving turbulence in the afternoon. However, the weakening solar radiation at 0800 BJT and 2000 BJT generally came with weak turbulence with sporadic residual layers.

To learn more about the relationship for each site from the national scale, the spatial distribution of correlation coefficient between the BLH and normalized PM_2.5 concentration at three times from the 4 years was shown in Figure . Generally, the negative correlation between PBL and PM_2.5 concentration uniformly spread across China, albeit with quite different R values. Overall, the correlation was found to be relatively weaker in western China, compared with eastern China, irrespective of the observational times (0800, 1400, and 2000 BJT). Contrary to our expectations, there existed several radiosonde sites with positive correlation at 1400 BJT, which were mostly located in eastern China, the inland cities, including Ankang, Naqu, and Yushu and existed at several sporadic coastal sites. The maximum PM_2.5 concentration ranged from 55 to 75 μg/m³ in the region of the NCP, the Yangtze River Delta, and central China in the afternoons and then diminished toward surrounding clean area located in the northern frontier and southwesternmost and southernmost regions of China (Figure S2). Interestingly, the strongest anticorrelations between PBL‐PM_2.5 tended to exist in heavily polluted regions. Compared to the distribution at 0800 BJT, the radiosonde sites in Xinjiang with positive correlation disappeared at 2000 BJT.

Fig. 4. The spatial distribution of correlation coefficient (R) between normalized PM2.5 and boundary layer heights at 0800 BJT (a), 1400 BJT (b), and 2000 BJT (c). The shaded dots represented negative correlation, whereas the plus signs represented a positive correlation. Symbols outlined in black circle indicated that values were statistically significant (p < 0.05). BJT = Beijing time.

Since the correlation at 1400 BJT was the strongest, the PBL parameters analyzed later were derived from the radiosonde measurements at 1400 BJT in the rest of this paper, unless noted otherwise. In order to understand how the meteorological factors impacted the relationship between BLH and PM_2.5, we selected four city sites, including Beijing (BJ), Wuhan (WH), Shanghai (SH), and Lhasa (LS). Negative correlations existed between BLH and PM_2.5 in BJ and WH, as opposed to LS and SH. Figure  manifested the time series of potential temperature overlaid by BLH and the corresponding PM_2.5 concentration for BJ, WH, SH, and LS in 2015. The R and p value were given, which was roughly similar to the correlations revealed in Figure . The days with heavy pollution in Beijing were characterized by warmer air masses above 1.5‐km above ground level and shallower PBL, whereas the relatively clean days were often connected with cool air masses above and high BLH (Figure a). Although the peak value of PM_2.5 (~75 μg/m³) was nearly the half of Beijing, the increases and drops of BLH accompanied by inverse changes of PM_2.5 concentration could also be found in Wuhan, especially during the period of 4–14 July (Figure b). Figure c demonstrates no obvious relationship between BLH and PM_2.5 in SH where rather limited soundings were available. As mentioned above, the effect of the rainy period (Meiyu) cannot be neglected due to scavenging effect induced by precipitation in preceding hours of 1400 BJT, although the data with precipitation were deleted. Also, there was no striking connection between BLH and PM_2.5 in Lhasa in summer since the background PM_2.5 kept at low level at most of the time, which made the correlation between PM_2.5 and BLH less sensitive (Su et al., ). In particular, when PM_2.5 concentration exceeds 35 μg/m³, a sudden drop of BLH took place.

Fig. 5. Time series of PM2.5 concentrations (red curves) and BLHs (black curves) at 1400 BJT in BJ, WH, SH, and LS from June to August 2015. Also shown were the time‐height cross sections (color shaded) of BLH/PM2.5 PT in (a) BJ, (b) SH, (c) XC, and (d) LS. The Pearson correlation coefficient (R) between BLH and PM2.5 concentration and the number of samples were also displayed. BLH = boundary layer height; PT = potential temperature; BJ = Beijing; WH = Wuhan; SH = Shanghai; LS = Lhasa.

To elucidate the variation of BLH in summer in a thorough way, the associations of BLH with six relevant meteorological factors, including the 10‐m WS, 2‐m temperature, RH, surface pressure, lower tropospheric stability (LTS), and CLD, were examined using correlation analyses (Figure ). The sites with a small number of samples were removed in the same manner as Figure . The LTS was an indicator to describe the thermodynamic state of the lower troposphere, which was defined as the difference in potential temperature between 700 hPa and the surface (Guo, Miao, et al., ). The daily BLH was positively correlated with the surface wind and temperature (Figures a and b) and was anticorrelated with RH, LTS, and CLD (Figures c, e, and f). There was no significant correlation between BLH and surface pressure in the summer. The stabilized stratification in summer would limit the exchanges of momentum and heat between the surface and free troposphere, leading to low PBL. The reduction of PBL gave rise to high‐surface RH. Furthermore, CLD suppressed turbulence and further impeded the evolution of PBL. However, there was no obvious relationship between normalized PM_2.5 concentration and meteorological factors from a national perspective, although strong correlations existed at several sites (Figure ). To further clarify the joint influences of BLH and meteorological factors on pollution, the distribution of the normalized PM_2.5 with BLH and meteorological factors were examined in Figure . The BLHs were classified into 10 discrete cells by 10 decile bins as a vertical axis and 10 decile bins as a horizontal axis, in which approximately equal sample sizes were ensured among the cells. The mean values of normalized PM_2.5 were calculated in each cell. In Figure b, the normalized PM_2.5 concentration was found to increase with decreases in BLH and WS, which was totally opposite to the variation pattern of BLH (Figure e). As the BLH was fixed, the normalized PM_2.5 climbed up and then declined with increasing LTS. However, we were not able to find apparent regular distribution among BLH, normalized PM_2.5, and other meteorological factors. This indicated that the effect of meteorological variables on the correlation between BLH and PM_2.5 is more complicated than expected and deserved further analysis based on model simulation.

Fig. 6. Spatial distribution of Pearson correlation coefficient (R) between BLH and (a) 10‐m WS, (b) 2‐m T, (c) RH, (d) Ps, (e) LTS, and (f) CLD. Symbols outlined in black indicated values that were statically significant (p < 0.05). Note that the stations in which Ps is less than 700 hPa were discarded in panel (e). BLH = boundary layer height; WS = wind speed; T = temperature; RH = relative humidity; Ps = surface pressure; LTS = lower tropospheric stability; CLD = cloud cover.

Fig. 7. Same as Figure  but for the relationship between normalized PM2.5 and six meteorological variables. WS = wind speed; RH = relative humidity; LTS = lower tropospheric stability; CLD = cloud cover; Ps = surface pressure.

Fig. 8. Joint dependence of the normalized PM2.5 on BLH versus (a) CLD, (b) 10‐m WS, (c) 2‐m T, (d) RH, (e) LTS, and (f) Ps. The bold number in each cell indicated the number of samples in the cell. The color bar denoted the normalized PM2.5 averaged in each cell. BLH = boundary layer height; WS = wind speed; RH = relative humidity; LTS = lower tropospheric stability; CLD = cloud cover; Ps = surface pressure.

Relationships Between Aerosol‐PBL Under Various PBL Thermodynamic Conditions

We will first analyze the relationships between different type of BLH and PM_2.5 concentration in the afternoon in this section and followed by the analyses on how several key meteorological variables affected these PBLs.

As expected, the height of CBL was, on average, much higher than that of NBL (Figure S3). The occurrence of SBL, despite not too much, was most likely caused by cloudy or overcast conditions (Zhang et al., ). Paradoxically, there were several circumstances that PM_2.5 concentration under SBL was relatively lower, probably due to the rare sampling in several clean sites (Figure S4).

Figures a–c demonstrated the distribution of PM_2.5 differences for CBL, NBL, and SBL relative to hourly averaged PM_2.5 in the afternoons with valid colocated BLH retrievals for all PBL regimes in summer during 2014 to 2017. As a well‐mixed layer, CBL had a better ability to dissipate pollution in most regions of China, especially in eastern China afflicted by heavy atmospheric pollutions (Ding et al., ; Guo et al., ). For the NBL, the air quality was generally inferior to the average in similar regions. In the majority of sites in China, PM_2.5 concentration for the SBL was much above the average.

Fig. 9. Spatial distributions of the differences of PM2.5 for CBL (a), NBL (b), and SBL (c) regimes relative to those for all PBL regimes at 1400 Beijing time in summer from 2014 to 2017. Note that the sites highlighted with black cycles indicated the difference was statistically significant (p < 0.05), and those sites without enough samples were not shown. CBL = convective boundary layer; NBL = neutral boundary layer; SBL = stable boundary layer; PBL = planetary boundary layer.

The data binning method was an effective way to remove the expected day‐to‐day atmospheric variability from sampling uncertainty (Seidel et al., ). Figures a–c showed the violin plots of binned CBL, NBL, and SBL in the afternoons. The R and regression equation were calculated according to the average values (BLH and PM_2.5) of each bin. For CBL and NBL regimes, the BLH was found to decline with the increasing aerosol loadings, while the BLH for the SBL increases with the increasing aerosol loadings. As expected, the most significant negative correlation was observed for the CBL height and PM_2.5, followed by NBL. In contrast, the SBL height was found to be positively correlated with PM_2.5, likely owing to the fact that SBL often came with much frequent turbulent intermittency in the PBL (Mahrt, ). This tended to cause extremely complex vertical distribution of aerosol particles, and the complicated feedback of PM_2.5 on the development of PBL more or less contributed to the positive correlation between BLH and ground‐level PM_2.5 (Wei et al., ).

Fig. 10. Violin plots showing the relationship between normalized PM2.5 and BLHs under CBL (a), NBL (b), and SBL (c) regimes at 1400 Beijing time in China. Each bin had the same number of samples. Box plots show the interquartile range (the distance between the bottom and the top of the box), median (the band inside the box), and 95% confidence interval (whiskers above and below the box) of the data. The maximum (the end of the whisker above), minimum (the end of the whisker below), and mean (blue dot) values were shown in each bin as well. Blue “+” represents outliers. The yellow‐shaded areas represent the kernel estimations, which indicated the frequency distribution of paired BLH and PM2.5. The linear regression relationships (blue lines) were shown between PM2.5 and BLH. Also given were the Pearson correlation coefficient (R) and p values. BLH = boundary layer height; NBL = neutral boundary layer; CBL = convective boundary layer; SBL = stable boundary layer.

Figure  displayed the box‐and‐whisker plots of CLD, LTS, RH, and WS for all PBL regimes. To figure out how the variation in cloud influences the structure of PBL, the LTS, RH, WS, BLH, and PM_2.5 concentration were compared under different CLD conditions. Here the overcast condition was thought to be the sky with more than 80% of total CLD at 1400 BJT, as opposed to the clear‐sky condition when the CLD was less than 20%. As shown in Figure , the NBL and SBL tended to form under high CLD conditions, among which the SBL had the most cloud amount of about 70%. In contrast with the NBL, the higher CBL height was formed under low CLD (Figure a). On average, the CBLs were characterized by relatively more unstable stratification and lower surface RH, all of which tended to facilitate the development of PBL and thus lead to the dispersion and turbulent mixing of aerosols (Figure ). However, it was noteworthy that wind was much stronger under overcast condition, due to the strong cloud‐induced static stability. By contrast, high cloud inhibited the development of SBL, which in turn led to heavy pollution.

Fig. 11. Box plots showing the statistics of several variables for SBL, NBL, and CBL regimes: (a) CLD, (b) LTS, (c) RH, (d) 10‐m WS, (e) BLH, and (f) PM2.5 concentration under clear‐sky (yellow) and overcast (blue) conditions. CLD = cloud cover; RH = relative humidity; WS = wind speed; LTS = lower tropospheric stability; BLH = boundary layer height; SBL = stable boundary layer; NBL = neutral boundary layer; CBL = convective boundary layer.

Fig. 12. Dependence of (a) CLD, (b) 10‐m WS, (c) RH, and (d) LTS on normalized PM2.5 (in units of %) on the convective boundary layer height. Note that all the data deviated from triple standard deviation were excluded in these statistics. CLD = cloud cover; WS = wind speed; RH = relative humidity; LTS = lower tropospheric stability; BLH = boundary layer height.

Conclusions

Using 4‐year (2014–2017) record of high‐resolution summertime L‐band radiosonde measurements, the relationship between PM_2.5 concentration and BLH in China were thoroughly examined. In particular, this work was one of first attempts to examine the response of aerosol to various PBL regimes (CBL, NBL, and SBL). As expected, the BLH was found to exhibit significant diurnal variability, consistent the pattern of PM_2.5. The correlations were strongest at 1400 BJT, followed by 2000 BJT and 0800 BJT. Overall, for both CBL and NBL regimes, the BLHs were found to negatively correlate with PM_2.5, regardless of the PBL regimes. We noticed that the positive correlation was found in SBL, compared to the strong negative correlation in CBL. Due to the sufficient solar radiation reaching the land surface, the CBL was characterized by intense vertical turbulence and uniformly mixed pollutants. Therefore, the CBL height was highly associated with changes in aerosols. Compared to the vertical turbulent mixing in CBL, the turbulence was nearly of equal intensity in all directions in NBL, and pollutants emitted into the layer tended to disperse at equal rates in vertical and lateral directions (Blay‐Carreras et al., ; Stull, ). Therefore, the relationship between NBL height and PM_2.5 got much complicated. By contrast, larger SBL height characterized by much frequent stratification and turbulent intermittency in the PBL was often found to come with the enhancement of ground level PM_2.5.

The correlations between BLH and PM_2.5 showed strong geographic dependence with the strongest anticorrelations (0.5) observed in the NCP in the afternoons. By comparison, the relationships between CBL height and aerosol seemed not significant in clean areas, due to its less sensitivity to aerosol. The PBL was found to be suppressed with the increases in PM_2.5 concentrations, irrespective of PBL regimes.

Besides, the dominant meteorological features were revealed under various PBL regimes. Particularly, the BLH was positively associated with the surface wind and temperature but negatively associated with RH, LTS, and CLD. The SBL, occurring mostly under overcast conditions, generally came with relatively high aerosol loadings. This could be due to the fact that reduced solar radiation under SBL impeded the vertical mixing of aerosols, resulting in accumulation of aerosol particles in the PBL. It remains challenging to elucidate the mechanism underlying the complex relationships between BLH, meteorology, and aerosols from observations alone, which merits further future work with the aids of explicit model simulation.

Acknowledgments

This work was under the auspices of the Ministry of Science and Technology of China under grants 2017YFC0212405 and 2017YFC1501401, the National Natural Science Foundation of China under grants 91544217 and 41771399, and the Chinese Academy of Meteorological Sciences under grant 2017Z005. The radiosonde measurements were provided by the National Meteorological Information Center of China Meteorological Administration (http://data.cma.cn/en/), and the PM_2.5 data were downloaded from China National Environmental Monitoring Centre (http://www.cnemc.cn).